eLife Assessment

Using a combination of connectomics, optogenetics, behavioral analysis, and modeling, this study provides important findings on the role of two populations of inhibitory neurons in the generation of leg grooming movements in Drosophila. The data as presented provide incomplete evidence that the identified neuronal populations contribute to the alternation of flexion and extension by inhibiting specific sets of motor neurons while disinhibiting their counterparts. While the manuscript provides comprehensive details about the 13A/B neuronal populations involved in grooming control, updates on statistics, and explicit mentioning of experimental/modeling caveats would strengthen the study. The work will interest neuroscientists, and particularly those working on motor control.

Reviewer #1 (Public review):

Summary:

Syed et al. investigate the circuit underpinnings for leg grooming in the fruit fly. They identify two populations of local interneurons in the right front leg neuromere of ventral nerve cord, i.e. 62 13A neurons and 64 13B neurons. Hierarchical clustering analysis identifies 10 morphological classes for both populations. Connectome analysis reveals their circuit interactions: these GABAergic interneurons provide synaptic inhibition either between the two subpopulations, i.e., 13B onto 13A, or among each other, i.e., 13As onto other 13As, and/or onto leg motoneurons, i.e., 13As and 13Bs onto leg motoneurons. Interestingly, 13A interneurons fall into two categories, with one providing inhibition onto a broad group of motoneurons, being called "generalists", while others project to a few motoneurons only, being called "specialists". Optogenetic activation and silencing of both subsets strongly affect leg grooming. As well aas ctivating or silencing subpopulations, i.e., 3 to 6 elements of the 13A and 13B groups, has marked effects on leg grooming, including frequency and joint positions, and even interrupting leg grooming. The authors present a computational model with the four circuit motifs found, i.e., feed-forward inhibition, disinhibition, reciprocal inhibition, and redundant inhibition. This model can reproduce relevant aspects of the grooming behavior.

Strengths:

The authors succeeded in providing evidence for neural circuits interacting by means of synaptic inhibition to play an important role in the generation of a fast rhythmic insect motor behavior, i.e., grooming. Two populations of local interneurons in the fruit fly VNC comprise four inhibitory circuit motifs of neural action and interaction: feed-forward inhibition, disinhibition, reciprocal inhibition, and redundant inhibition. Connectome analysis identifies the similarities and differences between individual members of the two interneuron populations. Modulating the activity of small subsets of these interneuron populations markedly affects the generation of the motor behavior, thereby exemplifying their important role in generating grooming.

Weaknesses:

Effects of modulating activity in the interneuron populations by means of optogenetics were conducted in the so-called closed-loop condition. This does not allow for differentiation between direct and secondary effects of the experimental modification in neural activity, as feedforward and feedback effects cannot be disentangled. To do so, open loop experiments, e.g., in deafferented conditions, would be important. Given that many members of the two populations of interneurons do not show one, but two or more circuit motifs, it remains to be disentangled which role the individual circuit motif plays in the generation of the motor behavior in intact animals.

Reviewer #2 (Public review):

Summary:

This manuscript by Syed et al. presents a detailed investigation of inhibitory interneurons, specifically from the 13A and 13B hemilineages, which contribute to the generation of rhythmic leg movements underlying grooming behavior in Drosophila. After performing a detailed connectomic analysis, which offers novel insights into the organization of premotor inhibitory circuits, the authors build on this anatomical framework by performing optogenetic perturbation experiments to functionally test predictions derived from the connectome. Finally, they integrate these findings into a computational model that links anatomical connectivity with behavior, offering a systems-level view of how inhibitory circuits may contribute to grooming pattern generation.

Strengths:

(1) Performing an extensive and detailed connectomic analysis, which offers novel insights into the organization of premotor inhibitory circuits.

(2) Making sense of the largely uncharacterized 13A/13B nerve cord circuitry by combining connectomics and optogenetics is very impressive and will lay the foundation for future experiments in this field.

(3) Testing the predictions from experiments using a simplified and elegant model.

Weaknesses:

(1) In Figure 4, while the authors report statistically significant shifts in both proximal inter-leg distance and movement frequency across conditions, the distributions largely overlap, and only in Panel K (13B silencing) is there a noticeable deviation from the expected 7-8 Hz grooming frequency. Could the authors clarify whether these changes truly reflect disruption of the grooming rhythm? More importantly, all this data would make the most sense if it were performed in undusted flies (with controls) as is done in the next figure.

(2) In Figure 4-Figure Supplement 1, the inclusion of walking assays in dusted flies is problematic, as these flies are already strongly biased toward grooming behavior and rarely walk. To assess how 13A neuron activation influences walking, such experiments should be conducted in undusted flies under baseline locomotor conditions.

(3) For broader lines targeting six or more 13A neurons, the authors provide specific predictions about expected behavioral effects-e.g., that activation should bias the limb toward flexion and silencing should bias toward extension based on connectivity to motor neurons. Yet, when using the more restricted line labeling only two 13A neurons (Figure 4 - Figure Supplement 2), no such prediction is made. The authors report disrupted grooming but do not specify whether the disruption is expected to bias the movement toward flexion or extension, nor do they discuss the muscle target. This is a missed opportunity to apply the same level of mechanistic reasoning that was used for broader manipulations.

(4) Regarding Figure 5: The 70ms on/off stimulation with a slow opsin seems problematic. CsChrimson off kinetics are slow and unlikely to cause actual activity changes in the desired neurons with the temporal precision the authors are suggesting they get. Regardless, it is amazing that the authors get the behavior! It would still be important for the authors to mention the optogenetics caveat, and potentially supplement the data with stimulation at different frequencies, or using faster opsins like ChrimsonR.

Overall, I think the strengths outweigh the weaknesses, and I consider this a timely and comprehensive addition to the field.

Reviewer #3 (Public review):

Summary:

The authors set out to determine how GABAergic inhibitory premotor circuits contribute to the rhythmic alternation of leg flexion and extension during Drosophila grooming. To do this, they first mapped the ~120 13A and 13B hemilineage inhibitory neurons in the prothoracic segment of the VNC and clustered them by morphology and synaptic partners. They then tested the contribution of these cells to flexion and extension using optogenetic activation and inhibition and kinematic analyses of limb joints. Finally, they produced a computational model representing an abstract version of the circuit to determine how the connectivity identified in EM might relate to functional output. The study, in its current form, makes an important but overclaimed contribution to the literature due to a mismatch between the claims in the paper and the data presented.

Strengths:

The authors have identified an interesting question and use a strong set of complementary tools to address it:

(1) They analysed serial‐section TEM data to obtain reconstructions of every 13A and 13B neuron in the prothoracic segment. They manually proofread over 60 13A neurons and 64 13B neurons, then used automated synapse detection to build detailed connectivity maps and cluster neurons into functional motifs.

(2) They used optogenetic tools with a range of genetic driver lines in freely behaving flies to test the contribution of subsets of 13A and 13B neurons.

(3) They used a connectome-constrained computational model to determine how the mapped connectivity relates to the rhythmic output of the behavior.

Weaknesses:

The manuscript aims to reveal an instructive, rhythm-generating role for premotor inhibition in coordinating the multi-joint leg synergies underlying grooming. It makes a valuable contribution, but currently, the main claims in the paper are not well-supported by the presented evidence.

Major points

(1) Starting with the title of this manuscript, "Inhibitory circuits generate rhythms for leg movements during Drosophila grooming", the authors raise the expectation that they will show that the 13A and 13B hemilineages produce rhythmic output that underlies grooming. This manuscript does not show that. For instance, to test how they drive the rhythmic leg movements that underlie grooming requires the authors to test whether these neurons produce the rhythmic output underlying behavior in the absence of rhythmic input. Because the optogenetic pulses used for stimulation were rhythmic, the authors cannot make this point, and the modelling uses a "black box" excitatory network, the output of which might be rhythmic (this is not shown). Therefore, the evidence (behavioral entrainment; perturbation effects; computational model) is all indirect, meaning that the paper's claim that "inhibitory circuits generate rhythms" rests on inferred sufficiency. A direct recording (e.g., calcium imaging or patch-clamp) from 13A/13B during grooming - outside the scope of the study - would be needed to show intrinsic rhythmogenesis. The conclusions drawn from the data should therefore be tempered. Moreover, the "black box" needs to be opened. What output does it produce? How exactly is it connected to the 13A-13B circuit? The context in which the 13A and 13B hemilineages sit also needs to be explained. What do we know about the other inputs to the motorneurons studied? What excitatory circuits are there? Furthermore, the introduction ignores many decades of work in other species on the role of inhibitory cell types in motor systems. There is some mention of this in the discussion, but even previous work in Drosophila larvae is not mentioned, nor crustacean STG, nor any other cell types previously studied. This manuscript makes a valuable contribution, but it is not the first to study inhibition in motor systems, and this should be made clear to the reader.

(2) The experimental evidence is not always presented convincingly, at times lacking data, quantification, explanation, appropriate rationales, or sufficient interpretation.

(3) The statistics used are unlike any I remember having seen, essentially one big t-test followed by correction for multiple comparisons. I wonder whether this approach is optimal for these nested, high‐dimensional behavioral data. For instance, the authors do not report any formal test of normality. This might be an issue given the often skewed distributions of kinematic variables that are reported. Moreover, each fly contributes many video segments, and each segment results in multiple measurements. By treating every segment as an independent observation, the non‐independence of measurements within the same animal is ignored. I think a linear mixed‐effects model (LMM) or generalized linear mixed model (GLMM) might be more appropriate.

(4) The manuscript mentions that legs are used for walking as well as grooming. While this is welcome, the authors then do not discuss the implications of this in sufficient detail. For instance, how should we interpret that pulsed stimulation of a subset of 13A neurons produces grooming and walking behaviours? How does neural control of grooming interact with that of walking?

(5) The manuscript needs to be proofread and edited as there are inconsistencies in labelling in figures, phrasing errors, missing citations of figures in the text, or citations that are not in the correct order, and referencing errors (examples: 81 and 83 are identical; 94 is missing in text).

eLife Assessment

This study presents a valuable finding on the perturbed pyruvate metabolism in models of repetitive traumatic brain injury. The evidence supporting the main claims of the authors is solid, but much of the accompanying analysis and interpretation relies on incomplete evidence. The work will be of interest to those working on the imaging of traumatic brain injury.

Joint Public Review:

Summary:

The authors present a metabolic imaging study of pyruvate metabolism in a mouse model of repetitive traumatic brain injury in the chronic recovery stage. They measure pyruvate metabolism with hyperpolarised 13C magnetic resonance spectroscopic imaging. This is acquired alongside semi-quantitative MR imaging metrics, a behavioural measure, and postmortem measures of relevant enzyme activity and expression of metabolic transporter proteins. They find that the MRSI-measured cortical lactate/pyruvate ratio (and signal from pyruvate and lactate independently) can differentiate the rTBI group from the sham group. They additionally find that postmortem, cortical pyruvate dehydrogenase activity is a statistically significant discriminator. All other metrics (MRI and enzyme/transporter measures) are not significantly different between groups. Finally, using a machine learning approach, the authors investigate the predictive power of combinations of all measures.

Strengths:

The primary strength of this work is the likely robustness of the primary finding - that hyperpolarised 13C lactate/pyruvate metabolite ratios are perturbed in this chronic rTBI model compared to the sham control.

Weaknesses:

Focal alterations in blood-brain-barrier permeability may affect the primary lactate/pyruvate measures. Whilst 13C urea measures perfusion, urea remains purely extracellular; whilst in the metabolism of the healthy brain, pyruvate must be transported through two levels of monocarboxylate transporters (MCTs) - in the endothelium surrounding the capillary bed and then into the parenchyma. By mechanically disrupting the brain, tight junctions in the BBB may be disrupted, therefore increasing the flux of pyruvate across the BBB and increasing pyruvate availability. In this case, lac/pyr would be a poor measure of metabolism as "delivery" has changed. While the authors assess perfusion using HP urea, it is unclear whether or how this metric would change in the presence of BBB disruption in relatively large and well-vascularised voxels.

The finding that "HP [1-13C]pyruvate levels were 1.05 fold higher" indicates that delivery of pyruvate might be increased. It is unclear if normalisation to the combined amplitude of lactate and pyruvate is fair in the case that the volume fraction in the voxel might have increased. Ideally, the authors would estimate polarisation separately as a normalisation.

No estimate of uncertainty is provided for the primary metabolic measures. Note that the lactate-pyruvate ratio is not normally distributed (see doi: 10.1002/mrm.26615), and this should be accounted for when carrying out statistical tests.

All metabolic maps are shown masked to the brain and interpolated to the structural MRI resolution (around 20 times). Nor is there any characterisation of the spectroscopic imaging's voxel volume, including the effect of the point spread function. It is, therefore, hard to have confidence in any spatial effects or potential partial volume effects from the tissue surrounding the brain.

The t2-weighted and SWI MRI measures used in this work are not quantitative. Normalisation in each case is carried out without regard to any spatially variable transmit and receive coil sensitivities (B1{plus minus}), which vary per subject. This adds intersubject variance, which could mask any effect between groups. No quality metrics (SNR or uncertainty estimates) are given for the MRI metrics.

Spectroscopic imaging was conducted 16 s after injection. Given the high heart rate of a mouse, measures of perfusion (using urea) could , therefore, be considered in a steady state, lowering sensitivity to any changes in perfusion or metabolite delivery. Furthermore, it is unclear how any changes in BBB permeability would manifest with the relatively low spatial resolution of MRSI. Would signal always be dominated by vascular compartments?

There is no apparent attempt to understand if an immune response occurs at this chronic time point. Macrophages are glycolytic and could affect the pyruvate measurement. Furthermore, is there any evidence for cellular changes in this model, namely density, cell type fraction, or microstructure? Are there any expected changes in glucose uptake?

There is no information or references provided for the accuracy or precision of the postmortem assays or their correlation with in vivo processes. What is the effect of cell density changes after injury on the assay kits?

The proposed interpretation of T1 as a measure of oxidative stress would seem to ignore the many confounding interpretations of T1.

Aims and impact:

In summary, the authors broadly achieve one aim, which is to find that HP 13C measured lac/pyruvate is a biomarker for the chronic effects of rTBI in a mouse model. As the authors themselves highlight in the discussion, the interpretation of this finding is tricky alongside their post-mortem assay results. The MR imaging in this work seems inconclusive, given the potential for inter-subject variance in the normalisation method.

The work, therefore, continues to highlight that HP 13C MRSI is a highly promising avenue of investigation to identify, characterise, and understand traumatic brain injury. It suggests that HP 13C MRSI is more promising in this sense than some standard MRI contrasts. The work currently fails to convincingly interpret the HP 13C MR results in conjunction with the other metrics.

Reviewer #1 (Public review):

Summary:

The manuscript by Ray et al. provides a theoretical framework to study tissue mechanics and the solid-to-fluid transition phenomenon observed in many tissues. The authors advanced previous models by directly incorporating cell-cell adhesion in force calculation with flexible cell geometries. They performed an in-depth analysis of the model and found that reducing cell-cell adhesion in near-confluent tissues can result in spontaneous cell rearrangements and transition to tissue fluidity. This is in contrast with previous predictions of Vertex models, which require higher adhesion for solid-to-fluid transition.

Strengths:

The authors provided a more general formulation of a 2D active foam model by directly incorporating cell-cell adhesion and performed a careful analysis of cell dynamics and cell shape in their simulations. They measured various quantities such as the mean-squared displacement of the cell center and shape index, which was introduced in previous studies to analyze jamming transition in tissues. By careful analysis of their simulations, they found a universal length scale in their simulations, explaining the observed heterogeneity. They provided a qualitative connection to previous experimental observations, where a reduction in cell adhesion caused tissue fluidity.

Weaknesses:

The phenomenon of tissue fluidity is an important and open question in biology. While theoretical models provide guidance to study such complex phenomena, the details in these models should go hand-in-hand with quantitative comparison with experiments. The study by Ray et al. indeed provided a more detailed description of deformable and adhesive cell collectives, but without a quantitative comparison with experiment, it is not clear if one needs all these details, or maybe more is needed. For example, do we need a more detailed mechanical model of the vertices, how the friction with substrate should be incorporated in such models, and is there a feedback between cell dynamics and its internal cytoskeleton organization?

While the manuscript by Ray et al. is an interesting theoretical study, without a quantitative comparison with experiments, it is not clear if it truly advances our understanding of tissue mechanics.

Reviewer #2 (Public review):

Summary:

Ray and coworkers introduce a discrete model of cellular layers aimed at investigating the role of inter-cellular adhesion in collective cell migration. The model combines aspects of particle-based models, in which cells are treated as simple point-particles with pair-interactions, and "morphological models", where interactions primarily depend on the cellular shape. In this case, cells are modeled as rings of beads connected by springs, thus allowing for exploration of the role of cell morphology while treating intercellular interactions as particle-like. Upon exploring the parameter space of this model, the authors recover physical behaviors reminiscent of reconstituted cell layers, including the onset of collective cell migration, when the forces leading to cell propulsion overweight inter-cellular adhesion, and various signatures of glassy dynamics.

Strengths:

The model presented in the article is simple, easy to implement, and scalable. The analysis appears solid and delivers a number of clear physical properties that could be tested in more depth in experiments and future numerical studies (e.g., distribution of displacements, etc.). The authors make an appreciable effort to make contact with other models and share their ideas for further investigations.

Weaknesses:

I found two main weaknesses in the original version of this manuscript, which I strongly encourage the authors to address.

(1) The manuscript explicitly aims at resolving an apparent contradiction of tessellation-based models, such as the Vertex and the Voronoi model. Both models used the so-called shape index p0 - i.e. the ratio between the preferential perimeter and the preferential area of the cells - to drive a solid/liquid phase transition in the presence of Brownian and/or rotational noise. Specifically, for sufficiently large p0 values, these in silico cell layers undergo a transition to a state of collective migration, where a rigid junction network becomes unstable to T1 events. Because p0 is often interpreted as "adhesion strength", this leads to the paradoxical conclusion that cell intercalation is favored by intercellular adhesion. The paradox, however, only lies in this interpretation, which assigns to the shape index p0 a biophysical role that is too specific. To illustrate this concept, let us consider the energy of an individual cell of area A and perimeter P: i.e. e = (a-1)^2+c*(p-p0)^2, where a=A/A_0, with A_0 the preferred area, p=P/sqrt(A_0) and p_0 = P_0/sqrt(A_0), with P_0 the preferred perimeter. Expanding the square in the second term gives e ~ p^2 - 2p_0 p. Thus, increasing p_0, favors longer cell junctions, from which it appears reasonable to interpret p0 as a dimensionless measure of intercellular adhesion. Such an increase in the length of the junctions is, however, only a byproduct of the effect of p0 on the overall shape of the cell, which becomes progressively less rounded as p0 is increased (e.g., for a circle, p0≈3.55, for an equilateral triangle, p0≈4.56). The roundness of an individual cell, on the other hand, cannot single-handedly be ascribed to intercellular adhesion, despite intercellular adhesion being undoubtedly one of the biophysical properties affecting this geometrical feature. Moreover, the shape index p0 ​enters the energy functional at the single-cell level, implying that even in isolation, without intercellular adhesion, an increase in p0 leads to a less rounded cell morphology. These peculiarities of the Vertex/Voronoi model do raise questions about its accuracy and validity, thus justify seeking for alternative cell-resolved models such as that introduced here by Ray et al., but, on the other hand, make the interpretation of p0 as an exclusive measure of adhesion evidently dubious.

(2) The spring-bead model by Ray and coworkers has at least two predecessors in the recent literature, none of which have been cited in the present manuscript. These are Boromand et al., Phys. Rev. Lett. 121, 248003 (2018) and Pasupalak et al. arXiv:2409.16128 (2024). The former paper investigates the packing of flexible polygons and is not specific to epithelial layers, while the latter is specifically designed to address various outstanding problems in tissue mechanics, including collective migration and wound healing. While none of these models is identical to that by Ray et al., it would be fair to present the latter as a member of the family rather than the first one of its kind and possibly comment about the differences and similarities with these previous models.

Reviewer #3 (Public review):

Summary:

This is a very focused and well-performed study that uses a somewhat less common approach in the field of tissue mechanics, a deformable particle model, to propose a solution to some important phenomenological inconsistencies between the standard vertex- and SPV-model approaches and experiments. The authors' focus in their study is on the role of adhesion in glassy dynamics and solid-fluid transition of epithelia.

Strengths:

It is a carefully performed study with an important technical edge compared to "mainstream" vertex and SPV models: the ability to describe cell-cell boundaries with two distinct membranes. This may have an important implication for the phenomenology, like the role of adhesion in solid-fluid transition.

Weaknesses:

Apart from some specific suggestions for improvement and clarification, I believe the authors could do a better job in comparing their results and their approach to other similar models, such as the one by Kim et al (Reference 7).

eLife Assessment

This is a valuable and rigorous study that addresses the question of what determines the spatial organization of endocytic zones at synapses. The authors use compelling approaches, in both Drosophila and rodent model systems, to define the role of activity and active zone structure on the organization of the peri-active zone. While the findings are primarily negative, they are carefully executed and contribute to the field by refining existing models of presynaptic organization.

Reviewer #1 (Public review):

Summary:

In this manuscript, Emperador-Melero et al. seek to determine whether recruitment of endocytic machinery to the periactive zone is activity-dependent or tethered to delivery of active zone machinery. They use genetic knockouts and pharmacological block in two model synapses - cultured mouse hippocampal neurons and Drosophila neuromuscular junctions - to determine how well endocytic machinery localizes after chronic inhibition or acute depolarization by super-resolution imaging. They find that acute depolarization in both models has minimal to no effect on the localization of endocytic machinery at the periactive zone, suggesting that these proteins are constitutively maintained rather than upregulated in response to transient activity. Interestingly, chronic inhibition slightly increases endocytic machinery levels, implying a potential homeostatic upregulation in preparation for rebound depolarization. Using genetic knockouts, the authors show that localization of endocytic machinery to periactive zones occurs independently of proper active zone assembly, even in the absence of upstream organizers like Liprin-α.

Overall, they propose that the constitutive deployment of endocytic machinery reflects its critical role in facilitating rapid and reliable membrane internalization during synaptic functions beyond classical endocytosis, such as regulation of the exocytic fusion pore and dense-core vesicle fusion. Although many experiments reveal limited changes in the localization or abundance of endocytic machinery, the findings are thorough, and data substantially support a model in which endocytic components are organized through a pathway distinct from that of the active zone. This work advances our understanding of synaptic dynamics by supporting a model in which endocytic machinery is constitutively recruited and regulated by distinct upstream organizers compared to active zone proteins. It also highlights the utility of super-resolution imaging across diverse synapse types to uncover functionally conserved elements of synaptic biology.

Strengths:

The study's technical strengths, particularly the use of super-resolution microscopy and rigorous image analyses developed by the group, bolster their findings.

Weaknesses:

One notable limitation, however, is the absence of interrogation of endocytic proteins previously suggested to be recruited in an activity-dependent manner, in particular, endophilin.

Reviewer #2 (Public review):

Summary:

This study examines whether the localization of endocytic proteins to presynaptic periactive zones depends on synaptic activity or active zone scaffolds. Using a combination of genetic and pharmacological perturbations in Drosophila and mouse neurons, the authors show that proteins such as Dynamin, Amphiphysin, AP-180, and others are still recruited to periactive zones even when evoked release or active zone architecture is disrupted. While the results are mostly negative, the study is methodologically solid and contributes to a more nuanced understanding of synaptic vesicle recycling machinery.

Strengths:

(1) The experimental design is careful and systematic, covering both fly and mammalian systems.

(2) The use of advanced genetic models (e.g., Liprin-α quadruple knockout mice) is a notable strength.

(3) High-resolution imaging (STED, Airyscan) is well used to assess spatial localization.

(4) The findings clarify that certain core assumptions - such as strict activity dependence of endocytic recruitment - may not hold universally.

Weaknesses:

(1) The study would benefit from a clearer positive control to demonstrate activity-dependent recruitment (e.g., Endophilin).

(2) The reliance on Tetanus toxin in the Drosophila NMJ experiments in my eyes is a limitation, as it does not block all presynaptic fusion events; this should be discussed more directly.

(3) The potential role of Dynamin in organizing other periactive zone proteins is not addressed and could be an important next step.

(4) Some small changes in protein levels upon silencing are reported; their biological meaning (e.g., compensation vs. variability) is not fully clarified.

(5) While alternative organizing mechanisms (actin, lipids, adhesion molecules) are mentioned, a more forward-looking discussion of how to test these models would be helpful.

(6) The authors should consider including, or at least discussing, a well-established activity-dependent endocytic protein (e.g., Endophilin) as a positive control to help contextualize the negative findings.

Reviewer #3 (Public review):

Summary:

This study examines how synaptic endocytic zones are positioned using a combination of cultured neurons and the Drosophila neuromuscular junction. The authors test whether neuronal activity, active zone assembly, or liprin-α function is required to localize endocytic zone markers, including Dynamin, Amphiphysin, Nervous Wreck, PIPK1γ, and AP-180. None of the manipulations tested caused a coordinated disruption in the localization or abundance of these markers, leading to the conclusion that endocytic zones form independently of synaptic activity and active zone scaffolds.

Strengths:

The work is systematic and carefully executed, using multiple manipulations and two complementary model systems. The authors consistently examine multiple molecular markers, strengthening the interpretation that endocytic zone positioning is robust to changes in activity and structural assembly.

Weaknesses:

The main limitation is that the study does not test whether the methods used are sensitive enough to detect subtle functional disruption, and no condition tested produces clear disorganization of the endocytic zone. As a result, the conclusion that these zones assemble independently is supported by negative data, without a strong positive control for disassembly or mislocalization.

This paper addresses a longstanding question in synaptic biology and provides a well-supported boundary on the types of mechanisms that are likely to govern endocytic zone localization. The conclusions are well justified by the data, though additional evidence would be needed to define the assembly mechanism itself.

eLife Assessment

The authors conducted a valuable study that investigates a molecular pathway mediating the transformation of a cell aggregate into a sheet known as the nucleus laminaris, a crucial site for auditory processing. While the study offers a comprehensive view of the sequence of developmental events and suggests possible roles for FGF signaling, the transcription factor Mafb, and the cell surface adhesive molecule Cadherin-23 in this process, the current data were considered incomplete.

Reviewer #1 (Public review):

Summary:

In this study, the authors sought to define a molecular pathway that mediates the transformation of an aggregate of cells into a sheet known as the nucleus laminaris, a key site for auditory processing. The data offer a comprehensive view of the sequence of developmental events and suggest possible roles for FGF signaling, the transcription factor Mafb, and the cell surface adhesive molecule Cadherin-23 in this process.

Strengths:

The description of nL development is thorough and well-done, with extensive quantification of the overall structure of the nucleus and also of neuron number. Additionally, the study implicates several molecules in nL development, starting with a clear description of when and where FGF8, Mafb, and several cadherins are expressed, including antibody stains suggesting that one cadherin, cdh2, is localized to the neuronal dendrites. A series of perturbation experiments supports the idea that these three molecules play a role in nL formation. The computational model is an interesting addition that helps to conceptualize how cadherin-mediated adhesion might influence nL morphogenesis.

Weaknesses:

A number of weaknesses limit the impact of this work.

One problem is how the data is interpreted. The logic is often circular in that the same molecules are used both as markers of nL and also as players in its development. An independent measure of nL formation is needed. Along the same lines, while the experiments implicate each molecule, the data do not actually demonstrate that FGF directly modulates Mafb, which in turn modulates cadherin expression, especially as overexpression of cdh2 has no effect on FGF8 expression or lamina organization, and no manipulations of cdh22 are presented.

The other type of problem relates to how the experiments were performed and analyzed. Important details about the experiments, as well as key controls, are missing throughout. Sample sizes are rarely presented, and there is no evidence that either dominant negative construct actually acts as proposed. Some results are not well quantified, which further undermines the strength of the conclusions. For instance, the changes in mafb and cdh22 expression (Figure 7) are subtle and were not quantified for any of the conditions. Likewise, the claim that FGF8 has a dose-dependent effect on lamina size and neuron number needs to be supported by statistics.

There are also some questions about the quality of the data. Much of the histology is of poor quality and does not always show the same piece of brain in the same orientation from experiment to experiment, which makes it challenging to interpret the results. In particular, the quality of the in situ hybridization varies, with much more background in some cases than others, which makes it hard to know what signal is real.

Finally, there are some misstatements and problems with citations that weaken the scholarly nature of the paper. FGF signaling has been studied extensively in the hindbrain and even in auditory nucleus development (Abraira et al., 2007), but this literature is not discussed at all.

Due to these weaknesses, the authors have achieved their aims only in part. The data are suggestive, but the results do not yet fully support their conclusions.

Few labs study how populations of neurons assemble into spatially organized structures. This work has the potential to be very interesting to other developmental neuroscientists studying brain morphogenesis.

Reviewer #2 (Public review):

Summary:

The overall goal of this study by Smith et al. was to understand the mechanisms through which groups of cells form specific nuclei during development. These cell groupings may have importance for the development of nervous system connections. Smith et al. have taken advantage of the ordered structure of the nucleus laminaris of the chick, which plays an important role in sound source localization. They used a candidate gene approach to both mark cells in nL and to test for signaling pathways that regulate nucleogenesis. They found that MafB, FGF8, and cadherins were expressed in the auditory hindbrain at the critical ages. They used in ovo electroporation to test gene function effects on nL lamina formation. They found that both increasing and decreasing FGF signaling (through introduction of mouse FGF8 and expression of a dominant negative FGF receptor, respectively) reduced lamina formation in the nL. An optimal concentration of FGF needed for this process was obtained using cultured hindbrain slices. Misexpression of cadherins also perturbed the normal lamina formation. The authors showed that FGF regulates MafB expression, which in turn regulates cadherin expression, suggesting a pathway that shapes lamina development. They constructed computational models of adhesion on the development of nL cells and found that laminar formation is favored by nL cells modeled as bipolar adhesive units. Overall, the study has demonstrated the importance of these adhesion pathways for the formation of the nucleus laminaris, and the findings likely have significance for the development of other nuclei as well.

Strengths:

The experiments have used in situ hybridization, immunofluorescence, electroporation, and brainstem slice cultures to test their hypotheses, which were based on well-selected candidate molecules. The modeling adds to the rigor of the studies, particularly in light of the observation that cadherin expression is localized to nL dendrites.

Weaknesses:

(1) Some references should be considered more carefully for accuracy, and additional references may be needed (introduction and results).

(2) Information on animal numbers and statistical tests should be added.

eLife Assessment

This manuscript provides valuable information on the neurodynamics of emotional processing while participants were watching movie clips. The methods and results were solid in deciphering the temporal-spatial dynamics of emotional processing. This work will be of interest to affective neuroscientists.

Reviewer #1 (Public review):

Summary:

In this manuscript, the authors endeavor to capture the dynamics of emotion-related brain networks. They employ slice-based fMRI combined with ICA on fMRI time series recorded while participants viewed a short movie clip. This approach allowed them to track the time course of four non-noise independent components at an effective 2s temporal resolution at the BOLD level. Notably, the authors report a temporal sequence from input to meaning, followed by response, and finally default mode networks, with significant overlap between stages. The use of ICA offers a data-driven method to identify large-scale networks involved in dynamic emotion processing. Overall, this paradigm and analytical strategy mark an important step forward in shifting affective neuroscience toward investigating temporal dynamics rather than relying solely on static network assessments

Strengths:

(1) One of the main advantages highlighted is the improved temporal resolution offered by slice-based fMRI. However, the manuscript does not clearly explain how this method achieves a higher effective resolution, especially since the results still show a 2s temporal resolution, comparable to conventional methods. Clarification on this point would help readers understand the true benefit of the approach.

(2) While combining ICA with task fMRI is an innovative approach to study the spatiotemporal dynamics of emotion processing, task fMRI typically relies on modeling the hemodynamic response (e.g., using FIR or IR models) to mitigate noise and collinearity across adjacent trials. The current analysis uses unmodeled BOLD time series, which might risk suffering from these issues.

(3) The study's claims about emotion dynamics are derived from fMRI data, which are inherently affected by the hemodynamic delay. This delay means that the observed time courses may differ substantially from those obtained through electrophysiology or MEG studies. A discussion on how these fMRI-derived dynamics relate to - or complement - is critical for the field to understand the emotion dynamics.

(4) Although using ICA to differentiate emotion elements is a convenient approach to tell a story, it may also be misleading. For instance, the observed delayed onset and peak latency of the 'response network' might imply that emotional responses occur much later than other stages, which contradicts many established emotion theories. Given the involvement of large-scale brain regions in this network, the underlying reasons for this delay could be very complex.

Concerns and suggestions:

However, I have several concerns regarding the specific presentation of temporal dynamics in the current manuscript and offer the following suggestions.

(1) One selling point of this work regarding the advantages of testing temporal dynamics is the application of slice-based fMRI, which, in theory, should improve the temporal resolution of the fMRI time course. Improving fMRI temporal resolution is critical for a research project on this topic. The authors present a detailed schematic figure (Figure 2) to help readers understand it. However, I have difficulty understanding the benefits of this method in terms of temporal resolution.

a) In Figure 2A, if we examine a specific voxel in slice 2, the slice acquisitions occur at 0.7s, 2.7s, and 4.7s, which implies a temporal resolution of 2s rather than 0.7s. I am unclear on how the temporal resolution could be 0.7s for this specific voxel. I would prefer that the authors clarify this point further, as it would benefit readers who are not familiar with this technology.

b) Even with the claim of an increased temporal resolution (0.7s), the actual data (Figure 3) still appears to have a 2s resolution. I wonder what specific benefit slice-based fMRI brings in terms of testing temporal dynamics, aside from correcting the temporal distortions that conventional fMRI exhibits.

(2) In task-fMRI, the hemodynamic response is usually estimated using a specific model (e.g., FIR, IR model; see Lindquist et al., 2009). These models are effective at reducing noise and collinearity across adjacent trials. The current method appears to be conducted on unmodeled BOLD time series.

a) I am wondering how the authors avoid the issues that are typically addressed by these HRF modeling approaches. For example, if we examine the baseline period (say, -4 to 0s relative to stimulus onset), the activation of most networks does not remain around zero, which could be due to delayed influences from the previous trial. This suggests that the current time course may not be completely accurate.

b) A related question: if the authors take the spatial map of a certain network and apply a modeling approach to estimate a time series within that network, would the results be similar to the current ICA time series?

(3) Human emotion should be inherently fast to ensure survival, as shown in many electrophysiology and MEG studies. For example, the dynamics of a fearful face can occur within 100ms in subcortical regions (Méndez-Bértolo et al., 2016), and general valence and arousal effects can occur as early as 200ms (e.g., Grootswagers et al., 2020; Bo et al., 2022). In contrast, the time-to-peak or onset timing in the BOLD time series spans a much larger time range due to the hemodynamic delay. fMRI findings indeed add spatial precision to our understanding of the temporal dynamics of emotion, but could the authors comment on how the current temporal dynamics supplement those electrophysiology studies that operate on much finer temporal scales?

(4) The response network shows activation as late as 15 to 20s, which is surprising. Could the authors discuss further why it takes so long for participants to generate an emotional response in the brain?

(5) Related to 4. In many theories, the emotion processing stages-including perception, valuation, and response-are usually considered iterative processes (e.g., Gross, 2015), especially in real-world scenarios. The advantage of the current paradigm is that it incorporates more dynamic elements of emotional stimuli and is closer to reality. Therefore, one might expect some degree of dynamic fluctuation within the tested brain networks to reflect those potential iterative processes (input, meaning, response). However, we still do not observe much brain dynamics in the data. In Figure 5, after the initial onset, most network activations remain sustained for an extended period of time. Does this suggest that emotion processing is less dynamic in the brain than we thought, or could it be related to limitations in temporal resolution? It could also be that the dynamics of each individual trial differ, and averaging them eliminates these variations. I would like to hear the authors' comments on this topic.

(6) The activation of the default mode network (DMN), although relatively late, is very interesting. Generally, one would expect a deactivation of this network during ongoing external stimulation. Could this suggest that participants are mind-wandering during the later portion of the task?

Reviewer #2 (Public review):

Summary:

This manuscript examined the neural correlates of the temporal-spatial dynamics of emotional processing while participants were watching short movie clips (each 12.5 s long) from the movie "Forrest Gump". Participants not only watched each film clip, but also gave emotional responses, followed by a brief resting period. Employing fMRI to track the BOLD responses during these stages of emotional processing, the authors found four large-scale brain networks (labeled as IC0,1,2,4) were differentially involved in emotional processing. Overall, this work provides valuable information on the neurodynamics of emotional processing.

Strengths:

This work employs a naturalistic movie watching paradigm to elicit emotional experiences. The authors used a slice-based fMRI method to examine the temporal dynamics of BOLD responses. Compared to previous emotional research that uses static images, this work provides some new data and insights into how the brain supports emotional processing from a temporal dynamics view.

Weaknesses:

Some major conclusions are unwarranted and do not have relevant evidence. For example, the authors seemed to interpret some neuroimaging results to be related to emotion regulation. However, there were no explicit instructions about emotional regulation, and there was no evidence suggesting participants regulated their emotions. How to best interpret the corresponding results thus requires caution.

Relatedly, the authors argued that "In turn, our findings underscore the utility of examining temporal metrics to capture subtle nuances of emotional processing that may remain undetectable using standard static analyses." While this sentence makes sense and is reasonable, it remains unclear how the results here support this argument. In particular, there were only three emotional categories: sad, happy, and fear. These three emotional categories are highly different from each other. Thus, how exactly the temporal metrics captured the "subtle nuances of emotional processing" shall be further elaborated.

The writing also contained many claims about the study's clinical utility. However, the authors did not develop their reasoning nor elaborate on the clinical relevance. While examining emotional processing certainly could have clinical relevance, please unpack the argument and provide more information on how the results obtained here can be used in clinical settings.

Importantly, how are the temporal dynamics of BOLD responses and subjective feelings related? The authors showed that "the time-to-peak differences in IC2 ("response") align closely with response latency results, with sad trials showing faster response latencies and earlier peak times". Does this mean that people typically experience sad feelings faster than happy or fear? Yet this is inconsistent with ideas such that fear detection is often rapid, while sadness can be more sustained. Understandably, the study uses movie clips, which can be very different from previous work, mostly using static images (e.g., a fearful or a sad face). But the authors shall explicitly discuss what these temporal dynamics mean for subjective feelings.

Author Response:

The following is the authors’ response to the previous reviews.

We carefully read through the second-round reviews and the additional reviews. To us, the review process is somewhat unusual and very much dominated by referee 2, who aggressively insists that we mixed up the trigeminal nucleus and inferior olive and that as a consequence our results are meaningless. We think the stance of referee 2 and the focus on one single issue (the alleged mix-up of trigeminal nucleus and inferior olive) is somewhat unfortunate, leaves out much of our findings and we debated at length on how to deal with further revisions. In the end, we decided to again give priority to addressing the criticism of referees 2, because it is hard to go on with a heavily attacked paper without resolving the matter at stake. The following is a summary of, what we did:

Additional experimental work:

(1) We checked if the peripherin-antibody indeed reliably identifies climbing fibers.

To this end, we sectioned the elephant cerebellum and stained sections with the peripherin-antibody. We find: (i) the cerebellar white matter is strongly reactive for peripherin-antibodies, (ii) cerebellar peripherin-antibody staining of has an axonal appearance. (iii) Cerebellar Purkinje cell somata appear to be ensheated by peripherin-antibody staining. (iv) We observed that the peripherin-antibody reactivity gradually decreases from Purkinje cell somata to the pia in the cerebellar molecular layer. This work is shown in our revised Figure 2. All these four features align with the distribution of climbing fibers (which arrive through the white matter, are axons, ensheat Purkinje cell somata, and innervate Purkinje cell proximally not reaching the pia). In line with previous work, which showed similar cerebellar staining patterns in several species (Errante et al. 1998), we conclude that elephant climbing fibers are strongly reactive for peripherin-antibodies.

(2) We delineated the elephant olivo-cerebellar tract.

The strong peripherin-antibody reactivity of elephant climbing fibers enabled us to delineate the elephant olivo-cerebellar tract. We find the elephant olivo-cerebellar tract is a strongly peripherin-antibody reactive, well-delineated fiber tract several millimeters wide and about a centimeter in height. The unstained olivo-cerebellar tract has a greyish appearance. In the anterior regions of the olivo-cerebellar tract, we find that peripherin-antibody reactive fibers run in the dorsolateral brainstem and approach the cerebellar peduncle, where the tract gradually diminishes in size, presumably because climbing fibers discharge into the peduncle. Indeed, peripherin-antibody reactive fibers can be seen entering the cerebellar peduncle. Towards the posterior end of the peduncle, the olivo-cerebellar disappears (in the dorsal brainstem directly below the peduncle. We note that the olivo-cerebellar tract was referred to as the spinal trigeminal tract by Maseko et al. 2013. We think the tract in question cannot be the spinal trigeminal tract for two reasons: (i) This tract is the sole brainstem source of peripherin-positive climbing fibers entering the peduncle/ the cerebellum; this is the defining characteristic of the olivo-cerebellar tract. (ii) The tract in question is much smaller than the trigeminal nerve, disappears posterior to where the trigeminal nerve enters the brainstem (see below), and has no continuity with the trigeminal nerve; the continuity with the trigeminal nerve is the defining characteristic of the spinal trigeminal tract, however.

The anterior regions of the elephant olivo-cerebellar tract are similar to the anterior regions of olivo-cerebellar tract of other mammals in its dorsolateral position and the relation to the cerebellar peduncle. In its more posterior parts, the elephant olivo-cerebellar tract continues for a long distance (~1.5 cm) in roughly the same dorsolateral position and enters the serrated nucleus that we previously identified as the elephant inferior olive. The more posterior parts of the elephant olivo-cerebellar tract therefore differ from the more posterior parts of the olivo-cerebellar tract of other mammals, which follows a ventromedial trajectory towards a ventromedially situated inferior olive. The implication of our delineation of the elephant olivo-cerebellar tract is that we correctly identified the elephant inferior olive.

(3) An in-depth analysis of peripherin-antibody reactivity also indicates that the trigeminal nucleus receives no climbing fiber input.

We also studied the peripherin-antibody reactivity in and around the trigeminal nucleus. We had also noted in the previous submission that the trigeminal nucleus is weakly positive for peripherin, but that the staining pattern is uniform and not the type of axon bundle pattern that is seen in the inferior olive of other mammals. To us, this observation already argued against the presence of climbing fibers in the trigeminal nucleus. We also noted that the myelin stripes of the trigeminal nucleus were peripherin-antibody-negative. In the context of our olivo-cerebellar tract tracing we now also scrutinized the surroundings of the trigeminal nucleus for peripherin-antibody reactivity. We find that the ventral brainstem surrounding the trigeminal nucleus is devoid of peripherin-antibody reactivity. Accordingly, no climbing fibers, (which we have shown to be strongly peripherin-antibody-positive, see our point 1) arrive at the trigeminal nucleus. The absence of climbing fiber input indicates that previous work that identified the (trigeminal) nucleus as the inferior olive (Maseko et al 2013) is unlikely to be correct.

(4) We characterized the entry of the trigeminal nerve into the elephant brain.

To better understand how trigeminal information enters the elephant’s brain, we characterized the entry of the trigeminal nerve. This analysis indicated to us that the trigeminal nerve is not continuous with the olivo-cerebellar tract (the spinal trigeminal tract of Maseko et al. 2013) as previously claimed by Maseko et al. 2013. We show some of this evidence in Referee-Figure 1 below. The reason we think the trigeminal nerve is discontinuous with the olivo-cerebellar tract is the size discrepancy between the two structures. We first show this for the tracing data of Maseko et al. 2013. In the Maseko et al. 2013 data the trigeminal nerve (Referee-Figure 1A, their plate Y) has 3-4 times the diameter of the olivocerebellar tract (the alleged spinal trigeminal tract, Referee-Figure 1B, their plate Z). Note that most if not all trigeminal fibers are thought to continue from the nerve into the trigeminal tract (see our rat data below). We plotted the diameter of the trigeminal nerve and diameter of the olivo-cerebellar (the spinal trigeminal tract according to Maseko et al. 2013) from the Maseko et al. 2013 data (Referee-Figure 1C) and we found that the olivocerebellar tract has a fairly consistent diameter (46 ± 9 mm2, mean ± SD). Statistical considerations and anatomical evidence suggest that the tracing of the trigeminal nerve into the olivo-cerebellar (the spinal trigeminal tract according to Maseko et al. 2013) is almost certainly wrong. The most anterior point of the alleged spinal trigeminal tract has a diameter of 51 mm2 which is more than 15 standard deviations different from the most posterior diameter (194 mm2) of the trigeminal tract. For this assignment to be correct three-quarters of trigeminal nerve fibers would have to spontaneously disappear, something that does not happen in the brain. We also made similar observations in the African elephant Bibi, where the trigeminal nerve (Referee-Figure 1D) is much larger in diameter than the olivocerebellar tract (Referee-Figure 1E). We could also show that the olivocerebellar tract disappears into the peduncle posterior to where the trigeminal nerve enters (Referee-Figure 1F). Our data are very similar to Maseko et al. indicating that their outlining of structures was done correctly. What appears to have been oversimplified, is the assignment of structures as continuous. We also quantified the diameter of the trigeminal nerve and the spinal trigeminal tract in rats (from the Paxinos & Watson atlas; Referee-Figure 1D); as expected we found the trigeminal nerve and spinal trigeminal tract diameters are essentially continuous.

In our hands, the trigeminal nerve does not continue into a well-defined tract that could be traced after its entry. In this regard, it differs both from the olivo-cerebellar tract of the elephant or the spinal trigeminal tract of the rodent, both of which are well delineated. We think the absence of a well-delineated spinal trigeminal tract in elephants might have contributed to the putative tracing error highlighted in our Referee-Figure 1A-C.

We conclude that a size mismatch indicates trigeminal fibers do not run in the olivo-cerebellar tract (the spinal trigeminal tract according to Maseko et al. 2013).

Author response image 1.

The trigeminal nerve is discontinuous with the olivo-cerebellar tract (the spinal trigeminal tract according to Maseko et al. 2013). A, Trigeminal nerve (orange) in the brain of African elephant LAX as delineated by Maseko et al. 2013 (coronal section; their plate Y). B, Most anterior appearance of the spinal trigeminal tract of Maseko et al. 2013 (blue; coronal section; their plate Z). Note the much smaller diameter of the spinal trigeminal tract compared to the trigeminal nerve shown in C, which argues against the continuity of the two structures. Indeed, our peripherin-antibody staining showed that the spinal trigeminal tract of Maseko corresponds to the olivo-cerebellar tract and is discontinuous with the trigeminal nerve. C, Plot of the trigeminal nerve and olivo-cerebellar tracts (the spinal trigeminal tract according to Maseko et al. 2013) diameter along the anterior-posterior axis. The trigeminal nerve is much larger in diameter than the olivocerebellar tract (the spinal trigeminal tract according to Maseko et al. 2013). C, D measurements, for which sections are shown in panels C and D respectively. The olivocerebellar tract (the spinal trigeminal tract according to Maseko et al. 2013) has a consistent diameter; data replotted from Maseko et al. 2013. At mm 25 the inferior olive appears. D, Trigeminal nerve entry in the brain of African elephant Bibi; our data, coronal section, the trigeminal nerve is outlined in orange, note the large diameter. E, Most anterior appearance of the olivo-cerebellar tract in the brain of African elephant Bibi; our data, coronal section, approximately 3 mm posterior to the section shown in A, the olivocerebellar tract is outlined in blue. Note the smaller diameter of the olivo-cerebellar tract compared to the trigeminal nerve, which argues against the continuity of the two structures. F, Plot of the trigeminal nerve and olivo-cerebellar tract diameter along the anterior-posterior axis. The nerve and olivo-cerebellar tract are discontinuous and the trigeminal nerve is much larger in diameter than the olivocerebellar tract (the spinal trigeminal tract according to Maseko et al. 2013); our data. D, E measurements, for which sections are shown in panels D and E respectively. At mm 27 the inferior olive appears. G, In the rat the trigeminal nerve is continuous in size with the spinal trigeminal tract. Data replotted from Paxinos and Watson.

Reviewer 2 (Public Review):

As indicated in my previous review of this manuscript (see above), it is my opinion that the authors have misidentified, and indeed switched, the inferior olivary nuclear complex (IO) and the trigeminal nuclear complex (Vsens). It is this specific point only that I will address in this second review, as this is the crucial aspect of this paper - if the identification of these nuclear complexes in the elephant brainstem by the authors is incorrect, the remainder of the paper does not have any scientific validity.

Comment: We agree with the referee that it is most important to sort out, the inferior olivary nuclear complex (IO) and the trigeminal nuclear complex, respectively.Change: We did additional experimental work to resolve this matter as detailed at the beginning of our response. Specifically, we ascertained that elephant climbing fibers are strongly peripherin-positive. Based on elephant climbing fiber peripherin-reactivity we delineated the elephant olivo-cerebellar tract. We find that the olivo-cerebellar connects to the structure we refer to as inferior olive to the cerebellum (the referee refers to this structure as the trigeminal nuclear complex). We also found that the trigeminal nucleus (the structure the referee refers to as inferior olive) appears to receive no climbing fibers. We provide indications that the tracing of the trigeminal nerve into the olivo-cerebellar tract by Maseko et al. 2023 was erroneous (Author response image 1). These novel findings support our ideas but are very difficult to reconcile with the referee’s partitioning scheme.

The authors, in their response to my initial review, claim that I "bend" the comparative evidence against them. They further claim that as all other mammalian species exhibit a "serrated" appearance of the inferior olive, and as the elephant does not exhibit this appearance, that what was previously identified as the inferior olive is actually the trigeminal nucleus and vice versa. 

For convenience, I will refer to IOM and VsensM as the identification of these structures according to Maseko et al (2013) and other authors and will use IOR and VsensR to refer to the identification forwarded in the study under review. <br /> The IOM/VsensR certainly does not have a serrated appearance in elephants. Indeed, from the plates supplied by the authors in response (Referee Fig. 2), the cytochrome oxidase image supplied and the image from Maseko et al (2013) shows a very similar appearance. There is no doubt that the authors are identifying structures that closely correspond to those provided by Maseko et al (2013). It is solely a contrast in what these nuclear complexes are called and the functional sequelae of the identification of these complexes (are they related to the trunk sensation or movement controlled by the cerebellum?) that is under debate.

Elephants are part of the Afrotheria, thus the most relevant comparative data to resolve this issue will be the identification of these nuclei in other Afrotherian species. Below I provide images of these nuclear complexes, labelled in the standard nomenclature, across several Afrotherian species. 

(A) Lesser hedgehog tenrec (Echinops telfairi) 

Tenrecs brains are the most intensively studied of the Afrotherian brains, these extensive neuroanatomical studies undertaken primarily by Heinz Künzle. Below I append images (coronal sections stained with cresol violet) of the IO and Vsens (labelled in the standard mammalian manner) in the lesser hedgehog tenrec. It should be clear that the inferior olive is located in the ventral midline of the rostral medulla oblongata (just like the rat) and that this nucleus is not distinctly serrated. The Vsens is located in the lateral aspect of the medulla skirted laterally by the spinal trigeminal tract (Sp5). These images and the labels indicating structures correlate precisely with that provide by Künzle (1997, 10.1016, see his Figure 1K,L. Thus, in the first case of a related species, there is no serrated appearance of the inferior olive, the location of the inferior olive is confirmed through connectivity with the superior colliculus (a standard connection in mammals) by Künzle (1997), and the location of Vsens is what is considered to be typical for mammals. This is in agreement with the authors, as they propose that ONLY the elephants show the variations they report. 

(B) Giant otter shrew (Potomogale velox) 

The otter shrews are close relatives of the Tenrecs. Below I append images of cresyl violet (left column) and myelin (right column) stained coronal sections through the brainstem with the IO, Vsens and Sp5 labelled as per standard mammalian anatomy. Here we see hints of the serration of the IO as defined by the authors, but we also see many myelin stripes across the IO. Vsens is located laterally and skirted by the Sp5. This is in agreement with the authors, as they propose that ONLY the elephants show the variations they report.

(C) Four-toed sengi (Petrodromus tetradactylus) 

The sengis are close relatives of the Tenrecs and otter shrews, these three groups being part of the Afroinsectiphilia, a distinct branch of the Afrotheria. Below I append images of cresyl violet (left column) and myelin (right column) stained coronal sections through the brainstem with the IO, Vsens and Sp5 labelled as per standard mammalian anatomy. Here we see vague hints of the serration of the IO (as defined by the authors), and we also see many myelin stripes across the IO. Vsens is located laterally and skirted by the Sp5. This is in agreement with the authors, as they propose that ONLY the elephants show the variations they report. 

(D) Rock hyrax (Procavia capensis) 

The hyraxes, along with the sirens and elephants form the Paenungulata branch of the Afrotheria. Below I append images of cresyl violet (left column) and myelin (right column) stained coronal sections through the brainstem with the IO, Vsens and Sp5 labelled as per the standard mammalian anatomy. Here we see hints of the serration of the IO (as defined by the authors), but we also see evidence of a more "bulbous" appearance of subnuclei of the IO (particularly the principal nucleus), and we also see many myelin stripes across the IO. Vsens is located laterally and skirted by the Sp5. This is in agreement with the authors, as they propose that ONLY the elephants show the variations they report. 

(E) West Indian manatee (Trichechus manatus) 

The sirens are the closest extant relatives of the elephants in the Afrotheria. Below I append images of cresyl violet (top) and myelin (bottom) stained coronal sections (taken from the University of Wisconsin-Madison Brain Collection, https://brainmuseum.org, and while quite low in magnification they do reveal the structures under debate) through the brainstem with the IO, Vsens and Sp5 labelled as per standard mammalian anatomy. Here we see the serration of the IO (as defined by the authors). Vsens is located laterally and skirted by the Sp5. This is in agreement with the authors, as they propose that ONLY the elephants show the variations they report.

These comparisons and the structural identification, with which the authors agree as they only distinguish the elephants from the other Afrotheria, demonstrate that the appearance of the IO can be quite variable across mammalian species, including those with a close phylogenetic affinity to the elephants. Not all mammal species possess a "serrated" appearance of the IO. Thus, it is more than just theoretically possible that the IO of the elephant appears as described prior to this study. 

So what about elephants? Below I append a series of images from coronal sections through the African elephant brainstem stained for Nissl, myelin, and immunostained for calretinin. These sections are labelled according to standard mammalian nomenclature. In these complete sections of the elephant brainstem, we do not see a serrated appearance of the IOM (as described previously and in the current study by the authors). Rather the principal nucleus of the IOM appears to be bulbous in nature. In the current study, no image of myelin staining in the IOM/VsensR is provided by the authors. However, in the images I provide, we do see the reported myelin stripes in all stains - agreement between the authors and reviewer on this point. The higher magnification image to the bottom left of the plate shows one of the IOM/VsensR myelin stripes immunostained for calretinin, and within the myelin stripes axons immunopositive for calretinin are seen (labelled with an arrow). The climbing fibres of the elephant cerebellar cortex are similarly calretinin immunopositive (10.1159/000345565). In contrast, although not shown at high magnification, the fibres forming the Sp5 in the elephant (in the Maseko description, unnamed in the description of the authors) show no immunoreactivity to calretinin. 

Comment: We appreciate the referee’s additional comments. We concede the possibility that some relatives of elephants have a less serrated inferior olive than most other mammals. We maintain, however, that the elephant inferior olive (our Figure 1J) has the serrated appearance seen in the vast majority of mammals.

Change: None.

Peripherin Immunostaining 

In their revised manuscript the authors present immunostaining of peripherin in the elephant brainstem. This is an important addition (although it does replace the only staining of myelin provided by the authors which is unusual as the word myelin is in the title of the paper) as peripherin is known to specifically label peripheral nerves. In addition, as pointed out by the authors, peripherin also immunostains climbing fibres (Errante et al., 1998). The understanding of this staining is important in determining the identification of the IO and Vsens in the elephant, although it is not ideal for this task as there is some ambiguity. Errante and colleagues (1998; Fig. 1) show that climbing fibres are peripherin-immunopositive in the rat. But what the authors do not evaluate is the extensive peripherin staining in the rat Sp5 in the same paper (Errante et al, 1998, Fig. 2). The image provided by the authors of their peripherin immunostaining (their new Figure 2) shows what I would call the Sp5 of the elephant to be strongly peripherin immunoreactive, just like the rat shown in Errant et al (1998), and more over in the precise position of the rat Sp5! This makes sense as this is where the axons subserving the "extraordinary" tactile sensitivity of the elephant trunk would be found (in the standard model of mammalian brainstem anatomy). Interestingly, the peripherin immunostaining in the elephant is clearly lamellated...this coincides precisely with the description of the trigeminal sensory nuclei in the elephant by Maskeo et al (2013) as pointed out by the authors in their rebuttal. Errante et al (1998) also point out peripherin immunostaining in the inferior olive, but according to the authors this is only "weakly present" in the elephant IOM/VsensR. This latter point is crucial. Surely if the elephant has an extraordinary sensory innervation from the trunk, with 400 000 axons entering the brain, the VsensR/IOM should be highly peripherin-immunopositive, including the myelinated axon bundles?! In this sense, the authors argue against their own interpretation - either the elephant trunk is not a highly sensitive tactile organ, or the VsensR is not the trigeminal nuclei it is supposed to be. 

Comment: We made sure that elephant climbing fibers are strongly peripherin-positive (our revised Figure 2). As we noted in already our previous ms, we see weak diffuse peripherin-reactivity in the trigeminal nucleus (the inferior olive according to the referee), but no peripherin-reactive axon bundles (i.e. climbing fibers) that are seen in the inferior olive of other species. We also see no peripherin-reactive axon bundles (i.e. the olivo-cerebellar tract) arriving in the trigeminal nucleus as the tissue surrounding the trigeminal nucleus is devoid of peripherin-reactivity. Again, this finding is incompatible with the referee’s ideas. As far as we can tell, the trigeminal fibers are not reactive for peripherin in the elephant, i.e. we did not observe peripherin-reactivity very close to the nerve entry, but unfortunately, we did not stain for peripherin-reactivity into the nerve. As the referee alludes to the absence of peripherin-reactivity in the trigeminal tract is a difference between rodents and elephants.

Change: Our novel Figure 2.

Summary: 

(1) Comparative data of species closely related to elephants (Afrotherians) demonstrates that not all mammals exhibit the "serrated" appearance of the principal nucleus of the inferior olive. 

(2) The location of the IO and Vsens as reported in the current study (IOR and VsensR) would require a significant, and unprecedented, rearrangement of the brainstem in the elephants independently. I argue that the underlying molecular and genetic changes required to achieve this would be so extreme that it would lead to lethal phenotypes. Arguing that the "switcheroo" of the IO and Vsens does occur in the elephant (and no other mammals) and thus doesn't lead to lethal phenotypes is a circular argument that cannot be substantiated. 

(3) Myelin stripes in the subnuclei of the inferior olivary nuclear complex are seen across all related mammals as shown above. Thus, the observation made in the elephant by the authors in what they call the VsensR, is similar to that seen in the IO of related mammals, especially when the IO takes on a more bulbous appearance. These myelin stripes are the origin of the olivocerebellar pathway, and are indeed calretinin immunopositive in the elephant as I show. 

(4) What the authors see aligns perfectly with what has been described previously, the only difference being the names that nuclear complexes are being called. But identifying these nuclei is important, as any functional sequelae, as extensively discussed by the authors, is entirely dependent upon accurately identifying these nuclei. 

(4) The peripherin immunostaining scores an own goal - if peripherin is marking peripheral nerves (as the authors and I believe it is), then why is the VsensR/IOM only "weakly positive" for this stain? This either means that the "extraordinary" tactile sensitivity of the elephant trunk is non-existent, or that the authors have misinterpreted this staining. That there is extensive staining in the fibre pathway dorsal and lateral to the IOR (which I call the spinal trigeminal tract), supports the idea that the authors have misinterpreted their peripherin immunostaining.

(5) Evolutionary expediency. The authors argue that what they report is an expedient way in which to modify the organisation of the brainstem in the elephant to accommodate the "extraordinary" tactile sensitivity. I disagree. As pointed out in my first review, the elephant cerebellum is very large and comprised of huge numbers of morphologically complex neurons. The inferior olivary nuclei in all mammals studied in detail to date, give rise to the climbing fibres that terminate on the Purkinje cells of the cerebellar cortex. It is more parsimonious to argue that, in alignment with the expansion of the elephant cerebellum (for motor control of the trunk), the inferior olivary nuclei (specifically the principal nucleus) have had additional neurons added to accommodate this cerebellar expansion. Such an addition of neurons to the principal nucleus of the inferior olive could readily lead to the loss of the serrated appearance of the principal nucleus of the inferior olive, and would require far less modifications in the developmental genetic program that forms these nuclei. This type of quantitative change appears to be the primary way in which structures are altered in the mammalian brainstem. 

Comment: We still disagree with the referee. We note that our conclusions rest on the analysis of 8 elephant brainstems, which we sectioned in three planes and stained with a variety of metabolic and antibody stains and in which assigned two structures (the inferior olive and the trigeminal nucleus). Most of the evidence cited by the referee stems from a single paper, in which 147 structures were identified based on the analysis of a single brainstem sectioned in one plane and stained with a limited set of antibodies. Our synopsis of the evidence is the following.

(1) We agree with the referee that concerning brainstem position our scheme of a ventromedial trigeminal nucleus and a dorsolateral inferior olive deviates from the usual mammalian position of these nuclei (i.e. a dorsolateral trigeminal nucleus and a ventromedial inferior olive).

(2) Cytoarchitectonics support our partitioning scheme. The compact cellular appearance of our ventromedial trigeminal nucleus is characteristic of trigeminal nuclei. The serrated appearance of our dorsolateral inferior olive is characteristic of the mammalian inferior olive; we acknowledge that the referee claims exceptions here. To our knowledge, nobody has described a mammalian trigeminal nucleus with a serrated appearance (which would apply to the elephant in case the trigeminal nucleus is situated dorsolaterally).

(3) Metabolic staining (Cyto-chrome-oxidase reactivity) supports our partitioning scheme. Specifically, our ventromedial trigeminal nucleus shows intense Cyto-chrome-oxidase reactivity as it is seen in the trigeminal nuclei of trigeminal tactile experts.

(4) Isomorphism. The myelin stripes on our ventromedial trigeminal nucleus are isomorphic to trunk wrinkles. Isomorphism is a characteristic of somatosensory brain structures (barrel, barrelettes, nose-stripes, etc) and we know of no case, where such isomorphism was misleading.

(5) The large-scale organization of our ventromedial trigeminal nuclei in anterior-posterior repeats is characteristic of the mammalian trigeminal nuclei. To our knowledge, no such organization has ever been reported for the inferior olive.

(6) Connectivity analysis supports our partitioning scheme. According to our delineation of the elephant olivo-cerebellar tract, our dorsolateral inferior olive is connected via peripherin-positive climbing fibers to the cerebellum. In contrast, our ventromedial trigeminal nucleus (the referee’s inferior olive) is not connected via climbing fibers to the cerebellum.

Change: As discussed, we advanced further evidence in this revision. Our partitioning scheme (a ventromedial trigeminal nucleus and a dorsolateral inferior olive) is better supported by data and makes more sense than the referee’s suggestion (a dorsolateral trigeminal nucleus and a ventromedial inferior olive). It should be published.

Reviewer #3 (Public Review):

Summary: 

The study claims to investigate trunk representations in elephant trigeminal nuclei located in the brainstem. The researchers identify large protrusions visible from the ventral surface of the brainstem, which they examined using a range of histological methods. However, this ventral location is usually where the inferior olivary complex is found, which challenges the author's assertions about the nucleus under analysis. They find that this brainstem nucleus of elephants contains repeating modules, with a focus on the anterior and largest unit which they define as the putative nucleus principalis trunk module of the trigeminal. The nucleus exhibits low neuron density, with glia outnumbering neurons significantly. The study also utilizes synchrotron X-ray phase contrast tomography to suggest that myelin-stripe-axons traverse this module. The analysis maps myelin-rich stripes in several specimens and concludes that based on their number and patterning that they likely correspond with trunk folds; however this conclusion is not well supported if the nucleus has been misidentified. 

Comment: The referee provides a summary of our work. The referee also notes that the correct identification of the trigeminal nucleus is critical to the message of our paper.

Change: In line with these assessments we focused our revision efforts on the issue of trigeminal nucleus identification, please see our introductory comments and our response to Referee 2.

Strengths: 

The strength of this research lies in its comprehensive use of various anatomical methods, including Nissl staining, myelin staining, Golgi staining, cytochrome oxidase labeling, and synchrotron X-ray phase contrast tomography. The inclusion of quantitative data on cell numbers and sizes, dendritic orientation and morphology, and blood vessel density across the nucleus adds a quantitative dimension. Furthermore, the research is commendable for its high-quality and abundant images and figures, effectively illustrating the anatomy under investigation.

Comment: We appreciate this positive assessment.

Change: None

Weaknesses: 

While the research provides potentially valuable insights if revised to focus on the structure that appears to be inferior olivary nucleus, there are certain additional weaknesses that warrant further consideration. First, the suggestion that myelin stripes solely serve to separate sensory or motor modules rather than functioning as an "axonal supply system" lacks substantial support due to the absence of information about the neuronal origins and the termination targets of the axons. Postmortem fixed brain tissue limits the ability to trace full axon projections. While the study acknowledges these limitations, it is important to exercise caution in drawing conclusions about the precise role of myelin stripes without a more comprehensive understanding of their neural connections. 

Comment: We understand these criticisms and the need for cautious interpretation. As we noted previously, we think that the Elife-publishing scheme, where critical referee commentary is published along with our ms, will make this contribution particularly valuable.

Change: Our additional efforts to secure the correct identification of the trigeminal nucleus.

Second, the quantification presented in the study lacks comparison to other species or other relevant variables within the elephant specimens (i.e., whole brain or brainstem volume). The absence of comparative data to different species limits the ability to fully evaluate the significance of the findings. Comparative analyses could provide a broader context for understanding whether the observed features are unique to elephants or more common across species. This limitation in comparative data hinders a more comprehensive assessment of the implications of the research within the broader field of neuroanatomy. Furthermore, the quantitative comparisons between African and Asian elephant specimens should include some measure of overall brain size as a covariate in the analyses. Addressing these weaknesses would enable a richer interpretation of the study's findings. 

Comment: We understand, why the referee asks for additional comparative data, which would make our study more meaningful. We note that we already published a quantitative comparison of African and Asian elephant facial nuclei (Kaufmann et al. 2022). The quantitative differences between African and Asian elephant facial nuclei are similar in magnitude to what we observed here for the trigeminal nucleus, i.e. African elephants have about 10-15% more facial nucleus neurons than Asian elephants. The referee also notes that data on overall elephant brain size might be important for interpreting our data. We agree with this sentiment and we are preparing a ms on African and Asian elephant brain size. We find – unexpectedly given the larger body size of African elephants – that African elephants have smaller brains than Asian elephants. The finding might imply that African elephants, which have more facial nucleus neurons and more trigeminal nucleus trunk module neurons, are neurally more specialized in trunk control than Asian elephants.

Change: We are preparing a further ms on African and Asian elephant brain size, a first version of this work has been submitted.

Reviewer #4 (Public Review): 

Summary: 

The authors report a novel isomorphism in which the folds of the elephant trunk are recognizably mapped onto the principal sensory trigeminal nucleus in the brainstem. Further, they identifiy the enlarged nucleus as being situated in this species in an unusual ventral midline position. 

Comment: The referee summarizes our work.

Change: None.

Strengths: 

The identity of the purported trigeminal nucleus and the isomorphic mapping with the trunk folds is supported by multiple lines of evidence: enhanced staining for cytochrome oxidase, an enzyme associated with high metabolic activity; dense vascularization, consistent with high metabolic activity; prominent myelinated bundles that partition the nucleus in a 1:1 mapping of the cutaneous folds in the trunk periphery; near absence of labeling for the anti-peripherin antibody, specific for climbing fibers, which can be seen as expected in the inferior olive; and a high density of glia.

Comment: The referee again reviews some of our key findings.

Change: None. 

Weaknesses: 

Despite the supporting evidence listed above, the identification of the gross anatomical bumps, conspicuous in the ventral midline, is problematic. This would be the standard location of the inferior olive, with the principal trigeminal nucleus occupying a more dorsal position. This presents an apparent contradiction which at a minimum needs further discussion. Major species-specific specializations and positional shifts are well-documented for cortical areas, but nuclear layouts in the brainstem have been considered as less malleable. 

Comment: The referee notes that our discrepancy with referee 2, needs to be addressed with further evidence and discussion, given the unusual position of both inferior olive and trigeminal nucleus in the partitioning scheme and that the mammalian brainstem tends to be positionally conservative. We agree with the referee. We note that – based on the immense size of the elephant trigeminal ganglion (50 g), half the size of a monkey brain – it was expected that the elephant trigeminal nucleus ought to be exceptionally large.

Change: We did additional experimental work to resolve this matter: (i) We ascertained that elephant climbing fibers are strongly peripherin-positive. (ii) Based on elephant climbing fiber peripherin-reactivity we delineated the elephant olivo-cerebellar tract. We find that the olivo-cerebellar connects to the structure we refer to as inferior olive to the cerebellum. (iii) We also found that the trigeminal nucleus (the structure the referee refers to as inferior olive) appears to receive no climbing fibers. (iv) We provide indications that the tracing of the trigeminal nerve into the olivo-cerebellar tract by Maseko et al. 2023 was erroneous (Referee-Figure 1). These novel findings support our ideas.

Reviewer #5 (Public Review): 

After reading the manuscript and the concerns raised by reviewer 2 I see both sides of the argument - the relative location of trigeminal nucleus versus the inferior olive is quite different in elephants (and different from previous studies in elephants), but when there is a large disproportionate magnification of a behaviorally relevant body part at most levels of the nervous system (certainly in the cortex and thalamus), you can get major shifting in location of different structures. In the case of the elephant, it looks like there may be a lot of shifting. Something that is compelling is that the number of modules separated but the myelin bands correspond to the number of trunk folds which is different in the different elephants. This sort of modular division based on body parts is a general principle of mammalian brain organization (demonstrated beautifully for the cuneate and gracile nucleus in primates, VP in most of species, S1 in a variety of mammals such as the star nosed mole and duck-billed platypus). I don't think these relative changes in the brainstem would require major genetic programming - although some surely exists. Rodents and elephants have been independently evolving for over 60 million years so there is a substantial amount of time for changes in each l lineage to occur.

I agree that the authors have identified the trigeminal nucleus correctly, although comparisons with more out groups would be needed to confirm this (although I'm not suggesting that the authors do this). I also think the new figure (which shows previous divisions of the brainstem versus their own) allows the reader to consider these issues for themselves. When reviewing this paper, I actually took the time to go through atlases of other species and even look at some of my own data from highly derived species. Establishing homology across groups based only on relative location is tough especially when there appears to be large shifts in relative location of structures. My thoughts are that the authors did an extraordinary amount of work on obtaining, processing and analyzing this extremely valuable tissue. They document their work with images of the tissue and their arguments for their divisions are solid. I feel that they have earned the right to speculate - with qualifications - which they provide. 

Comment: The referee summarizes our work and appears to be convinced by the line of our arguments. We are most grateful for this assessment. We add, again, that the skeptical assessment of referee 2 will be published as well and will give the interested reader the possibility to view another perspective on our work.

Change: None. 

Recommendations for the authors: 

Reviewer #1 (Recommendations For The Authors):

With this manuscript being virtually identical to the previous version, it is possible that some of the definitive conclusions about having identified the elephant trigeminal nucleus and trunk representation should be moderated in a more nuanced manner, especially given the careful and experienced perspective from reviewers with first hand knowledge elephant neuroanatomy.

Comment: We agree that both our first and second revisions were very much centered on the debate of the correct identification of the trigeminal nucleus and that our ms did not evolve as much in other regards. This being said we agree with Referee 2 that we needed to have this debate. We also think we advanced important novel data in this context (the delineation of elephant olivo-cerebellar tract through the peripherin-antibody).

Changes: Our revised Figure 2. 

The peripherin staining adds another level of argument to the authors having identified the trigeminal brainstem instead of the inferior olive, if differential expression of peripherin is strong enough to distinguish one structure from the other.

Comment: We think we showed too little peripherin-antibody staining in our previous revision. We have now addressed this problem.

Changes: Our revised Figure 2, i.e. the delineation of elephant olivo-cerebellar tract through the peripherin-antibody).

There are some minor corrections to be made with the addition of Fig. 2., including renumbering the figures in the manuscript (e.g., 406, 521). 

I continue to appreciate this novel investigation of the elephant brainstem and find it an interesting and thorough study, with the use of classical and modern neuroanatomical methods.

Comment: We are thankful for this positive assessment.

Reviewer #2 (Recommendations For The Authors):

I do realise the authors are very unhappy with me and the reviews I have submitted. I do apologise if feelings have been hurt, and I do understand the authors put in a lot of hard work and thought to develop what they have; however, it is unfortunate that the work and thoughts are not correct. Science is about the search for the truth and sometimes we get it wrong. This is part of the scientific process and why most journals adhere to strict review processes of scientific manuscripts. As I said previously, the authors can use their data to write a paper describing and quantifying Golgi staining of neurons in the principal olivary nucleus of the elephant that should be published in a specialised journal and contextualised in terms of the motor control of the trunk and the large cerebellum of the elephant. 

Comment: We appreciate the referee’s kind words. Also, no hard feelings from our side, this is just a scientific debate. In our experience, neuroanatomical debates are resolved by evidence and we note that we provide evidence strengthening our identification of the trigeminal nucleus and inferior olive. As far as we can tell from this effort and the substantial evidence accumulated, the referee is wrong.

Reviewer #4 (Recommendations For The Authors):

As a new reviewer, I have benefited from reading the previous reviews and Author response, even while having several new comments to add. 

(1) The identification of the inferior olive and trigeminal nuclei is obviously center stage. An enlargement of the trigeminal nuclei is not necessarily problematic, given the published reports on the dramatic enlargement of the trigeminal nerve (Purkart et al., 2022). At issue is the conspicuous relocation of the trigeminal nuclei that is being promoted by Reveyaz et al. Conspicuous rearrangements are not uncommon; for example, primary sensory cortical fields in different species (fig. 1 in H.H.A. Oelschlager for dolphins; S. De Vreese et al. (2023) for cetaceans, L. Krubitzer on various species, in the context of evolution). The difficult point here concerns what looks like a rather conspicuous gross anatomical rearrangement, in BRAINSTEM - the assumption being that the brainstem bauplan is going to be specifically conservative and refractory to gross anatomical rearrangement. 

Comment: We agree with the referee that the brainstem rearrangements are unexpected. We also think that the correct identification of nuclei needs to be at the center of our revision efforts.

Change: Our revision provided further evidence (delineation of the olivo-cerebellar tract, characterization of the trigeminal nerve entry) about the identity of the nuclei we studied.

Why would a major nucleus shift to such a different location? and how? Can ex vivo DTI provide further support of the correct identification? Is there other "disruption" in the brainstem? What occupies the traditional position of the trigeminal nuclei? An atlas-equivalent coronal view of the entire brainstem would be informative. The Authors have assembled multiple criteria to support their argument that the ventral "bumps" are in fact a translocated trigeminal principal nucleus: enhanced CO staining, enhanced vascularization, enhanced myelination (via Golgi stains and tomography), very scant labeling for a climbing fiber specific antibody ( anti-peripherin), vs. dense staining of this in the alternative structure that they identify as IO; and a high density of glia. Admittedly, this should be sufficient, but the proposed translocation (in the BRAINSTEM) is sufficiently startling that this is arguably NOT sufficient. <br /> The terminology of "putative" is helpful, but a more cogent presentation of the results and more careful discussion might succeed in winning over at least some of a skeptical readership. 

Comment: We do not know, what led to the elephant brainstem rearrangements we propose. If the trigeminal nuclei had expanded isometrically in elephants from the ancestral pattern, one would have expected a brain with big lateral bumps, not the elephant brain with its big ventromedial bumps. We note, however, that very likely the expansion of the elephant trigeminal nuclei did not occur isometrically. Instead, the neural representation of the elephant nose expanded dramatically and in rodents the nose is represented ventromedially in the brainstem face representation. Thus, we propose a ‘ventromedial outgrowth model’ according to which the elephant ventromedial trigeminal bumps result from a ventromedially direct outgrowth of the ancestral ventromedial nose representation.

We advanced substantially more evidence to support our partitioning scheme, including the delineation of the olivo-cerebellar tract based on peripherin-reactivity. We also identified problems in previous partitioning schemes, such as the claim that the trigeminal nerve continues into the ~4x smaller olivocerebellar tract (Referee-Figure 1C, D); we think such a flow of fibers, (which is also at odds with peripherin-antibody-reactivity and the appearance of nerve and olivocerebellar tract), is highly unlikely if not physically impossible. With all that we do not think that we overstate our case in our cautiously presented ms.

Change: We added evidence on the identification of elephant trigeminal nuclei and inferior olive.

(2) Role of myelin. While the photos of myelin are convincing, it would be nice to have further documentation. Gallyas? Would antibodies to MBP work? What is the myelin distribution in the "standard" trigeminal nuclei (human? macaque or chimpanzee?). What are alternative sources of the bundles? Regardless, I think it would be beneficial to de-emphasize this point about the role of myelin in demarcating compartments. <br /> I would in fact suggest an alternative (more neutral) title that might highlight instead the isomorphic feature; for example, "An isomorphic representation of Trunk folds in the Elephant Trigeminal Nucleus." The present title stresses myelin, but figure 1 already focuses on CO. Additionally, the folds are actually mentioned almost in passing until later in the manuscript. I recommend a short section on these at the beginning of the Results to serve as a useful framework.

Here I'm inclined to agree with the Reviewer, that the Authors' contention that the myelin stipes serve PRIMARILY to separate trunk-fold domains is not particularly compelling and arguably a distraction. The point can be made, but perhaps with less emphasis. After all, the fact that myelin has multiple roles is well-established, even if frequently overlooked. In addition, the Authors might make better use of an extensive relevant literature related to myelin as a compartmental marker; for example, results and discussion in D. Haenelt....N. Weiskopf (eLife, 2023), among others. Another example is the heavily myelinated stria of Gennari in primate visual cortex, consisting of intrinsic pyramidal cell axons, but where the role of the myelination has still not been elucidated. 

Comment: (1) Documentation of myelin. We note that we show further identification of myelinated fibers by the fluorescent dye fluomyelin in Figure 4B. We also performed additional myelin stains as the gold-myelin stain after the protocol of Schmued (Referee-Figure 2). In the end, nothing worked quite as well to visualize myelin-stripes as the bright-field images shown in Figure 4A and it is only the images that allowed us to match myelin-stripes to trunk folds. Hence, we focus our presentation on these images.

(2) Title: We get why the referee envisions an alternative title. This being said, we would like to stick with our current title, because we feel it highlights the major novelty we discovered.

(3) We agree with many of the other comments of the referee on myelin phenomenology. We missed the Haenelt reference pointed out by the referee and think it is highly relevant to our paper

Change: 1. Review image 2. Inclusion of the Haenelt-reference.

Author response image 2.

Myelin stripes of the elephant trunk module visualized by Gold-chloride staining according to Schmued. A, Low magnification micrograph of the trunk module of African elephant Indra stained with AuCl according to Schmued. The putative finger is to the left, proximal is to the right. Myelin stripes can easily be recognized. The white box indicates the area shown in B. B, high magnification micrograph of two myelin stripes. Individual gold-stained (black) axons organized in myelin stripes can be recognized.

Schmued, L. C. (1990). A rapid, sensitive histochemical stain for myelin in frozen brain sections. Journal of Histochemistry & Cytochemistry,38(5), 717-720.

Are the "bumps" in any way "analogous" to the "brain warts" seen in entorhinal areas of some human brains (G. W. van Hoesen and A. Solodkin (1993)? 

Comment: We think this is a similar phenomenon.

Change: We included the Hoesen and A. Solodkin (1993) reference in our discussion.

At least slightly more background (ie, a separate section or, if necessary, supplement) would be helpful, going into more detail on the several subdivisions of the ION and if these undergo major alterations in the elephant.

Comment: The strength of the paper is the detailed delineation of the trunk module, based on myelin stripes and isomorphism. We don’t think we have strong evidence on ION subdivisions, because it appears the trigeminal tract cannot be easily traced in elephants. Accordingly, we find it difficult to add information here.

Change: None.

Is there evidence from the literature of other conspicuous gross anatomical translocations, in any species, especially in subcortical regions? 

Comment: The best example that comes to mind is the star-nosed mole brainstem. There is a beautiful paper comparing the star-nosed mole brainstem to the normal mole brainstem (Catania et al 2011). The principal trigeminal nucleus in the star-nosed mole is far more rostral and also more medial than in the mole; still, such rearrangements are minor compared to what we propose in elephants.

Catania, Kenneth C., Duncan B. Leitch, and Danielle Gauthier. "A star in the brainstem reveals the first step of cortical magnification." PloS one 6.7 (2011): e22406.

Change: None.

(3) A major point concerns the isomorphism between the putative trigeminal nuclei and the trunk specialization. I think this can be much better presented, at least with more discussion and other examples. The Authors mention about the rodent "barrels," but it seemed strange to me that they do not refer to their own results in pig (C. Ritter et al., 2023) nor the work from Ken Catania, 2002 (star-nosed mole; "fingerprints in the brain") or other that might be appropriate. I concur with the Reviewer that there should be more comparative data. 

Comment: We agree.

Change: We added a discussion of other isomorphisms including the the star-nosed mole to our paper.

(4) Textual organization could be improved. 

The Abstract all-important Introduction is a longish, semi "run-on" paragraph. At a minimum this should be broken up. The last paragraph of the Introduction puts forth five issues, but these are only loosely followed in the Results section. I think clarity and good organization is of the upmost importance in this manuscript. I recommend that the Authors begin the Results with a section on the trunk folds (currently figure 5, and discussion), continue with the several points related to the identification of the trigeminal nuclei, and continue with a parallel description of ION with more parallel data on the putative trigeminal and IO structures (currently referee Table 1, but incorporate into the text and add higher magnification of nucleus-specific cell types in the IO and trigeminal nuclei). Relevant comparative data should be included in the Discussion.

Comment: 1. We agree with the referee that our abstract needed to be revised. 2. We also think that our ms was heavily altered by the insertion of the new Figure 2, which complemented Figure 1 from our first submission and is concerned with the identification of the inferior olive. From a standpoint of textual flow such changes were not ideal, but the revisions massively added to the certainty with which we identify the trigeminal nuclei. Thus, although we are not as content as we were with the flow, we think the ms advanced in the revision process and we would like to keep the Figure sequence as is. 3. We already noted above that we included additional comparative evidence.

Change: 1. We revised our abstract. 2. We added comparative evidence.

Reviewer #5 (Recommendations For The Authors): 

The data is invaluable and provides insights into some of the largest mammals on the planet. 

Comment: We are incredibly thankful for this positive assessment.

Reviewer #2 (Public Review):

Here I submit my previous review and a great deal of additional information following on from the initial review and the response by the authors.

* Initial Review *

Assessment:

This manuscript is based upon the unprecedented identification of an apparently highly unusual trigeminal nuclear organization within the elephant brainstem, related to a large trigeminal nerve in these animals. The apparently highly specialized elephant trigeminal nuclear complex identified in the current study has been classified as the inferior olivary nuclear complex in four previous studies of the elephant brainstem. The entire study is predicated upon the correct identification of the trigeminal sensory nuclear complex and the inferior olivary nuclear complex in the elephant, and if this is incorrect, then the remainder of the manuscript is merely unsupported speculation. There are many reasons indicating that the trigeminal nuclear complex is misidentified in the current study, rendering the entire study, and associated speculation, inadequate at best, and damaging in terms of understanding elephant brains and behaviour at worst.

Original Public Review:

The authors describe what they assert to be a very unusual trigeminal nuclear complex in the brainstem of elephants, and based on this, follow with many speculations about how the trigeminal nuclear complex, as identified by them, might be organized in terms of the sensory capacity of the elephant trunk.<br /> The identification of the trigeminal nuclear complex/inferior olivary nuclear complex in the elephant brainstem is the central pillar of this manuscript from which everything else follows, and if this is incorrect, then the entire manuscript fails, and all the associated speculations become completely unsupported.

The authors note that what they identify as the trigeminal nuclear complex has been identified as the inferior olivary nuclear complex by other authors, citing Shoshani et al. (2006; 10.1016/j.brainresbull.2006.03.016) and Maseko et al (2013; 10.1159/000352004), but fail to cite either Verhaart and Kramer (1958; PMID 13841799) or Verhaart (1962; 10.1515/9783112519882-001). These four studies are in agreement, but the current study differs.

Let's assume for the moment that the four previous studies are all incorrect and the current study is correct. This would mean that the entire architecture and organization of the elephant brainstem is significantly rearranged in comparison to ALL other mammals, including humans, previously studied (e.g. Kappers et al. 1965, The Comparative Anatomy of the Nervous System of Vertebrates, Including Man, Volume 1 pp. 668-695) and the closely related manatee (10.1002/ar.20573). This rearrangement necessitates that the trigeminal nuclei would have had to "migrate" and shorten rostrocaudally, specifically and only, from the lateral aspect of the brainstem where these nuclei extend from the pons through to the cervical spinal cord (e.g. the Paxinos and Watson rat brain atlases), the to the spatially restricted ventromedial region of specifically and only the rostral medulla oblongata. According to the current paper the inferior olivary complex of the elephant is very small and located lateral to their trigeminal nuclear complex, and the region from where the trigeminal nuclei are located by others appears to be just "lateral nuclei" with no suggestion of what might be there instead.

Such an extraordinary rearrangement of brainstem nuclei would require a major transformation in the manner in which the mutations, patterning, and expression of genes and associated molecules during development occur. Such a major change is likely to lead to lethal phenotypes, making such a transformation extremely unlikely. Variations in mammalian brainstem anatomy are most commonly associated with quantitative changes rather than qualitative changes (10.1016/B978-0-12-804042-3.00045-2).

The impetus for the identification of the unusual brainstem trigeminal nuclei in the current study rests upon a previous study from the same laboratory (10.1016/j.cub.2021.12.051) that estimated that the number of axons contained in the infraorbital branch of the trigeminal nerve that innervate the sensory surfaces of the trunk is approximately 400 000. Is this number unusual? In a much smaller mammal with a highly specialized trigeminal system, the platypus, the number of axons innervating the sensory surface of the platypus bill skin comes to 1 344 000 (10.1159/000113185). Yet, there is no complex rearrangement of the brainstem trigeminal nuclei in the brain of the developing or adult platypus (Ashwell, 2013, Neurobiology of Monotremes), despite the brainstem trigeminal nuclei being very large in the platypus (10.1159/000067195). Even in other large-brained mammals, such as large whales that do not have a trunk, the number of axons in the trigeminal nerve ranges between 400,000 and 500,000 (10.1007/978-3-319-47829-6_988-1). The lack of comparative support for the argument forwarded in the previous and current study from this laboratory, and that the comparative data indicates that the brainstem nuclei do not change in the manner suggested in the elephant, argues against the identification of the trigeminal nuclei as outlined in the current study. Moreover, the comparative studies undermine the prior claim of the authors, informing the current study, that "the elephant trigeminal ganglion ... point to a high degree of tactile specialization in elephants" (10.1016/j.cub.2021.12.051). While clearly the elephant has tactile sensitivity in the trunk, it is questionable as to whether what has been observed in elephants is indeed "truly extraordinary".

But let's look more specifically at the justification outlined in the current study to support their identification of the unusually located trigeminal sensory nuclei of the brainstem.

(1) Intense cytochrome oxidase reactivity<br /> (2) Large size of the putative trunk module<br /> (3) Elongation of the putative trunk module<br /> (4) Arrangement of these putative modules correspond to elephant head anatomy<br /> (5) Myelin stripes within the putative trunk module that apparently match trunk folds<br /> (6) Location apparently matches other mammals<br /> (7) Repetitive modular organization apparently similar to other mammals.<br /> (8) The inferior olive described by other authors lacks the lamellated appearance of this structure in other mammals

Let's examine these justifications more closely.

(1) Cytochrome oxidase histochemistry is typically used as an indicative marker of neuronal energy metabolism. The authors indicate, based on the "truly extraordinary" somatosensory capacities of the elephant trunk, that any nuclei processing this tactile information should be highly metabolically active, and thus should react intensely when stained for cytochrome oxidase. We are told in the methods section that the protocols used are described by Purkart et al (2022) and Kaufmann et al (2022). In neither of these cited papers is there any description, nor mention, of the cytochrome oxidase histochemistry methodology, thus we have no idea of how this histochemical staining was done. In order to obtain the best results for cytochrome oxidase histochemistry, the tissue is either processed very rapidly after buffer perfusion to remove blood or in recently perfusion-fixed tissue (e.g., 10.1016/0165-0270(93)90122-8). Given: (1) the presumably long post-mortem interval between death and fixation - "it often takes days to dissect elephants"; (2) subsequent fixation of the brains in 4% paraformaldehyde for "several weeks"; (3) The intense cytochrome oxidase reactivity in the inferior olivary complex of the laboratory rat (Gonzalez-Lima, 1998, Cytochrome oxidase in neuronal metabolism and Alzheimer's diseases); and (4) The lack of any comparative images from other stained portions of the elephant brainstem; it is difficult to support the justification as forwarded by the authors. It is likely that the histochemical staining observed is background reactivity from the use of diaminobenzidine in the staining protocol. Thus, this first justification is unsupported.<br /> Justifications (2), (3), and (4) are sequelae from justification (1). In this sense, they do not count as justifications, but rather unsupported extensions.

(4) and (5) These are interesting justifications, as the paper has clear internal contradictions, and (5) is a sequelae of (4). The reader is led to the concept that the myelin tracts divide the nuclei into sub-modules that match the folding of the skin on the elephant trunk. One would then readily presume that these myelin tracts are in the incoming sensory axons from the trigeminal nerve. However, the authors note that this is not the case: "Our observations on trunk module myelin stripes are at odds with this view of myelin. Specifically, myelin stripes show no tapering (which we would expect if axons divert off into the tissue). More than that, there is no correlation between myelin stripe thickness (which presumably correlates with axon numbers) and trigeminal module neuron numbers. Thus, there are numerous myelinated axons, where we observe few or no trigeminal neurons. These observations are incompatible with the idea that myelin stripes form an axonal 'supply' system or that their prime function is to connect neurons. What do myelin stripe axons do, if they do not connect neurons? We suggest that myelin stripes serve to separate rather than connect neurons." So, we are left with the observation that the myelin stripes do not pass afferent trigeminal sensory information from the "truly extraordinary" trunk skin somatic sensory system, and rather function as units that separate neurons - but to what end? It appears that the myelin stripes are more likely to be efferent axonal bundles leaving the nuclei (to form the olivocerebellar tract). This justification is unsupported.

(6) The authors indicate that the location of these nuclei matches that of the trigeminal nuclei in other mammals. This is not supported in any way. In ALL other mammals in which the trigeminal nuclei of the brainstem have been reported they are found in the lateral aspect of the brainstem, bordered laterally by the spinal trigeminal tract. This is most readily seen and accessible in the Paxinos and Watson rat brain atlases. The authors indicate that the trigeminal nuclei are medial to the facial nerve nucleus, but in every other species, the trigeminal sensory nuclei are found lateral to the facial nerve nucleus. This is most salient when examining a close relative, the manatee (10.1002/ar.20573), where the location of the inferior olive and the trigeminal nuclei matches that described by Maseko et al (2013) for the African elephant. This justification is not supported.

(7) The dual to quadruple repetition of rostro-caudal modules within the putative trigeminal nucleus as identified by the authors relies on the fact that in the neurotypical mammal, there are several trigeminal sensory nuclei arranged in a column running from the pons to the cervical spinal cord, these include (nomenclature from Paxinos and Watson in roughly rostral to caudal order) the Pr5VL, Pr5DM, Sp5O, Sp5I, and Sp5C. But, these nuclei are all located far from the midline and lateral to the facial nerve nucleus, unlike what the authors describe in the elephants. These rostrocaudal modules are expanded upon in Figure 2, and it is apparent from what is shown that the authors are attributing other brainstem nuclei to the putative trigeminal nuclei to confirm their conclusion. For example, what they identify as the inferior olive in figure 2D is likely the lateral reticular nucleus as identified by Maseko et al (2013). This justification is not supported.

(8) In primates and related species, there is a distinct banded appearance of the inferior olive, but what has been termed the inferior olive in the elephant by other authors does not have this appearance, rather, and specifically, the largest nuclear mass in the region (termed the principal nucleus of the inferior olive by Maseko et al, 2013, but Pr5, the principal trigeminal nucleus in the current paper) overshadows the partial banded appearance of the remaining nuclei in the region (but also drawn by the authors of the current paper). Thus, what is at debate here is whether the principal nucleus of the inferior olive can take on a nuclear shape rather than evince a banded appearance. The authors of this paper use this variance as justification that this cluster of nuclei could not possibly be the inferior olive. Such a "semi-nuclear/banded" arrangement of the inferior olive is seen in, for example, giraffe (10.1016/j.jchemneu.2007.05.003), domestic dog, polar bear, and most specifically the manatee (a close relative of the elephant) (brainmuseum.org; 10.1002/ar.20573). This justification is not supported.

Thus, all the justifications forwarded by the authors are unsupported. Based on methodological concerns, prior comparative mammalian neuroanatomy, and prior studies in the elephant and closely related species, the authors fail to support their notion that what was previously termed the inferior olive in the elephant is actually the trigeminal sensory nuclei. Given this failure, the justifications provided above that are sequelae also fail. In this sense, the entire manuscript and all the sequelae are not supported.

What the authors have not done is to trace the pathway of the large trigeminal nerve in the elephant brainstem, as was done by Maseko et al (2013), which clearly shows the internal pathways of this nerve, from the branch that leads to the fifth mesencephalic nucleus adjacent to the periventricular grey matter, through to the spinal trigeminal tract that extends from the pons to the spinal cord in a manner very similar to all other mammals. Nor have they shown how the supposed trigeminal information reaches the putative trigeminal nuclei in the ventromedial rostral medulla oblongata. These are but two examples of many specific lines of evidence that would be required to support their conclusions. Clearly tract tracing methods, such as cholera toxin tracing of peripheral nerves cannot be done in elephants, thus the neuroanatomy must be done properly and with attention to detail to support the major changes indicated by the authors.

So what are these "bumps" in the elephant brainstem?

Four previous authors indicate that these bumps are the inferior olivary nuclear complex. Can this be supported?

The inferior olivary nuclear complex acts "as a relay station between the spinal cord (n.b. trigeminal input does reach the spinal cord via the spinal trigeminal tract) and the cerebellum, integrating motor and sensory information to provide feedback and training to cerebellar neurons" (https://www.ncbi.nlm.nih.gov/books/NBK542242/). The inferior olivary nuclear complex is located dorsal and medial to the pyramidal tracts (which were not labelled in the current study by the authors but are clearly present in Fig. 1C and 2A) in the ventromedial aspect of the rostral medulla oblongata. This is precisely where previous authors have identified the inferior olivary nuclear complex and what the current authors assign to their putative trigeminal nuclei. The neurons of the inferior olivary nuclei project, via the olivocerebellar tract to the cerebellum to terminate in the climbing fibres of the cerebellar cortex.

Elephants have the largest (relative and absolute) cerebellum of all mammals (10.1002/ar.22425), this cerebellum contains 257 x109 neurons (10.3389/fnana.2014.00046; three times more than the entire human brain, 10.3389/neuro.09.031.2009). Each of these neurons appears to be more structurally complex than the homologous neurons in other mammals (10.1159/000345565; 10.1007/s00429-010-0288-3). In the African elephant, the neurons of the inferior olivary nuclear complex are described by Maseko et al (2013) as being both calbindin and calretinin immunoreactive. Climbing fibres in the cerebellar cortex of the African elephant are clearly calretinin immunopositive and also are likely to contain calbindin (10.1159/000345565). Given this, would it be surprising that the inferior olivary nuclear complex of the elephant is enlarged enough to create a very distinct bump in exactly the same place where these nuclei are identified in other mammals?

What about the myelin stripes? These are most likely to be the origin of the olivocerebellar tract and probably only have a coincidental relationship to the trunk. Thus, given what we know, the inferior olivary nuclear complex as described in other studies, and the putative trigeminal nuclear complex as described in the current study, is the elephant inferior olivary nuclear complex. It is not what the authors believe it to be, and they do not provide any evidence that discounts the previous studies. The authors are quite simply put, wrong. All the speculations that flow from this major neuroanatomical error are therefore science fiction rather than useful additions to the scientific literature.

What do the authors actually have?<br /> The authors have interesting data, based on their Golgi staining and analysis, of the inferior olivary nuclear complex in the elephant.

* Review of Revised Manuscript *

Assessment:

There is a clear dichotomy between the authors and this reviewer regarding the identification of specific structures, namely the inferior olivary nuclear complex and the trigeminal nuclear complex, in the brainstem of the elephant. The authors maintain the position that in the elephant alone, irrespective of all the published data on other mammals and previously published data on the elephant brainstem, these two nuclear complexes are switched in location. The authors maintain that their interpretation is correct, but this reviewer maintains that this interpretation is erroneous. The authors expressed concern that the remainder of the paper was not addressed by the reviewer, but the reviewer maintains that these sequelae to the misidentification of nuclear complexes in the elephant brainstem render any of these speculations irrelevant as the critical structures are incorrectly identified. It is this reviewer's opinion that this paper is incorrect. I provide a lot of detail below in order to provide support to the opinion I express.

Public Review of Current Submission:

As indicated in my previous review of this manuscript (see above), it is my opinion that the authors have misidentified, and indeed switched, the inferior olivary nuclear complex (IO) and the trigeminal nuclear complex (Vsens). It is this specific point only that I will address in this second review, as this is the crucial aspect of this paper - if the identification of these nuclear complexes in the elephant brainstem by the authors is incorrect, the remainder of the paper does not have any scientific validity.

The authors, in their response to my initial review, claim that I "bend" the comparative evidence against them. They further claim that as all other mammalian species exhibit a "serrated" appearance of the inferior olive, and as the elephant does not exhibit this appearance, what was previously identified as the inferior olive is actually the trigeminal nucleus and vice versa.

For convenience, I will refer to IOM and VsensM as the identification of these structures according to Maseko et al (2013) and other authors and will use IOR and VsensR to refer to the identification forwarded in the study under review.<br /> The IOM/VsensR certainly does not have a serrated appearance in elephants. Indeed, from the plates supplied by the authors in response (Referee Fig. 2), the cytochrome oxidase image supplied and the image from Maseko et al (2013) shows a very similar appearance. There is no doubt that the authors are identifying structures that closely correspond to those provided by Maseko et al (2013). It is solely a contrast in what these nuclear complexes are called and the functional sequelae of the identification of these complexes (are they related to the trunk sensation or movement controlled by the cerebellum?) that is under debate.

Elephants are part of the Afrotheria, thus the most relevant comparative data to resolve this issue will be the identification of these nuclei in other Afrotherian species. Below I provide images of these nuclear complexes, labelled in the standard nomenclature, across several Afrotherian species.

(A) Lesser hedgehog tenrec (Echinops telfairi)

Tenrecs brains are the most intensively studied of the Afrotherian brains, these extensive neuroanatomical studies were undertaken primarily by Heinz Künzle. Below I append images (coronal sections stained with cresol violet) of the IO and Vsens (labelled in the standard mammalian manner) in the lesser hedgehog tenrec. It should be clear that the inferior olive is located in the ventral midline of the rostral medulla oblongata (just like the rat) and that this nucleus is not distinctly serrated. The Vsens is located in the lateral aspect of the medulla skirted laterally by the spinal trigeminal tract (Sp5). These images and the labels indicating structures correlate precisely with that provided by Künzle (1997, 10.1016/S0168- 0102(97)00034-5), see his Figure 1K,L. Thus, in the first case of a related species, there is no serrated appearance of the inferior olive, the location of the inferior olive is confirmed through connectivity with the superior colliculus (a standard connection in mammals) by Künzle (1997), and the location of Vsens is what is considered to be typical for mammals. This is in agreement with the authors, as they propose that ONLY the elephants show the variations they report.

Review image 1.

(B) Giant otter shrew (Potomogale velox)

The otter shrews are close relatives of the Tenrecs. Below I append images of cresyl violet (left column) and myelin (right column) stained coronal sections through the brainstem with the IO, Vsens and Sp5 labelled as per standard mammalian anatomy. Here we see hints of the serration of the IO as defined by the authors, but we also see many myelin stripes across the IO. Vsens is located laterally and skirted by the Sp5. This is in agreement with the authors, as they propose that ONLY the elephants show the variations they report.

Review image 2.

(C) Four-toed sengi (Petrodromus tetradactylus)

The sengis are close relatives of the Tenrecs and otter shrews, these three groups being part of the Afroinsectiphilia, a distinct branch of the Afrotheria. Below I append images of cresyl violet (left column) and myelin (right column) stained coronal sections through the brainstem with the IO, Vsens and Sp5 labelled as per standard mammalian anatomy. Here we see vague hints of the serration of the IO (as defined by the authors), and we also see many myelin stripes across the IO. Vsens is located laterally and skirted by the Sp5. This is in agreement with the authors, as they propose that ONLY the elephants show the variations they report.

Review image 3.

(D) Rock hyrax (Procavia capensis)

The hyraxes, along with the sirens and elephants form the Paenungulata branch of the Afrotheria. Below I append images of cresyl violet (left column) and myelin (right column) stained coronal sections through the brainstem with the IO, Vsens and Sp5 labelled as per the standard mammalian anatomy. Here we see hints of the serration of the IO (as defined by the authors), but we also see evidence of a more "bulbous" appearance of subnuclei of the IO (particularly the principal nucleus), and we also see many myelin stripes across the IO. Vsens is located laterally and skirted by the Sp5. This is in agreement with the authors, as they propose that ONLY the elephants show the variations they report.

Review image 4.

(E) West Indian manatee (Trichechus manatus)

The sirens are the closest extant relatives of the elephants in the Afrotheria. Below I append images of cresyl violet (top) and myelin (bottom) stained coronal sections (taken from the University of Wisconsin-Madison Brain Collection, https://brainmuseum.org, and while quite low in magnification they do reveal the structures under debate) through the brainstem with the IO, Vsens and Sp5 labelled as per standard mammalian anatomy. Here we see the serration of the IO (as defined by the authors). Vsens is located laterally and skirted by the Sp5. This is in agreement with the authors, as they propose that ONLY the elephants show the variations they report.

Review image 5.

These comparisons and the structural identification, with which the authors agree as they only distinguish the elephants from the other Afrotheria, demonstrate that the appearance of the IO can be quite variable across mammalian species, including those with a close phylogenetic affinity to the elephants. Not all mammal species possess a "serrated" appearance of the IO. Thus, it is more than just theoretically possible that the IO of the elephant appears as described prior to this study.

So what about elephants? Below I append a series of images from coronal sections through the African elephant brainstem stained for Nissl, myelin, and immunostained for calretinin. These sections are labelled according to standard mammalian nomenclature. In these complete sections of the elephant brainstem, we do not see a serrated appearance of the IOM (as described previously and in the current study by the authors). Rather the principal nucleus of the IOM appears to be bulbous in nature. In the current study, no image of myelin staining in the IOM/VsensR is provided by the authors. However, in the images I provide, we do see the reported myelin stripes in all stains - agreement between the authors and reviewer on this point. The higher magnification image to the bottom left of the plate shows one of the IOM/VsensR myelin stripes immunostained for calretinin, and within the myelin stripes axons immunopositive for calretinin are seen (labelled with an arrow). The climbing fibres of the elephant cerebellar cortex are similarly calretinin immunopositive (10.1159/000345565). In contrast, although not shown at high magnification, the fibres forming the Sp5 in the elephant (in the Maseko description, unnamed in the description of the authors) show no immunoreactivity to calretinin.

Review image 6.

Peripherin Immunostaining

In their revised manuscript the authors present immunostaining of peripherin in the elephant brainstem. This is an important addition (although it does replace the only staining of myelin provided by the authors which is unusual as the word myelin is in the title of the paper) as peripherin is known to specifically label peripheral nerves. In addition, as pointed out by the authors, peripherin also immunostains climbing fibres (Errante et al., 1998). The understanding of this staining is important in determining the identification of the IO and Vsens in the elephant, although it is not ideal for this task as there is some ambiguity. Errante and colleagues (1998; Fig. 1) show that climbing fibres are peripherin-immunopositive in the rat. But what the authors do not evaluate is the extensive peripherin staining in the rat Sp5 in the same paper (Errante et al, 1998, Fig. 2). The image provided by the authors of their peripherin immunostaining (their new Figure 2) shows what I would call the Sp5 of the elephant to be strongly peripherin immunoreactive, just like the rat shown in Errant et al (1998), and moreover in the precise position of the rat Sp5! This makes sense as this is where the axons subserving the "extraordinary" tactile sensitivity of the elephant trunk would be found (in the standard model of mammalian brainstem anatomy). Interestingly, the peripherin immunostaining in the elephant is clearly lamellated...this coincides precisely with the description of the trigeminal sensory nuclei in the elephant by Maskeo et al (2013) as pointed out by the authors in their rebuttal. Errante et al (1998) also point out peripherin immunostaining in the inferior olive, but according to the authors this is only "weakly present" in the elephant IOM/VsensR. This latter point is crucial. Surely if the elephant has an extraordinary sensory innervation from the trunk, with 400,000 axons entering the brain, the VsensR/IOM should be highly peripherin-immunopositive, including the myelinated axon bundles?! In this sense, the authors argue against their own interpretation - either the elephant trunk is not a highly sensitive tactile organ, or the VsensR is not the trigeminal nuclei it is supposed to be.

Summary:

(1) Comparative data of species closely related to elephants (Afrotherians) demonstrates that not all mammals exhibit the "serrated" appearance of the principal nucleus of the inferior olive.

(2) The location of the IO and Vsens as reported in the current study (IOR and VsensR) would require a significant, and unprecedented, rearrangement of the brainstem in the elephants independently. I argue that the underlying molecular and genetic changes required to achieve this would be so extreme that it would lead to lethal phenotypes. Arguing that the "switcheroo" of the IO and Vsens does occur in the elephant (and no other mammals) and thus doesn't lead to lethal phenotypes is a circular argument that cannot be substantiated.

(3) Myelin stripes in the subnuclei of the inferior olivary nuclear complex are seen across all related mammals as shown above. Thus, the observation made in the elephant by the authors in what they call the VsensR, is similar to that seen in the IO of related mammals, especially when the IO takes on a more bulbous appearance. These myelin stripes are the origin of the olivocerebellar pathway and are indeed calretinin immunopositive in the elephant as I show.

(4) What the authors see aligns perfectly with what has been described previously, the only difference being the names that nuclear complexes are being called. But identifying these nuclei is important, as any functional sequelae, as extensively discussed by the authors, is entirely dependent upon accurately identifying these nuclei.

(4) The peripherin immunostaining scores an own goal - if peripherin is marking peripheral nerves (as the authors and I believe it is), then why is the VsensR/IOM only "weakly positive" for this stain? This either means that the "extraordinary" tactile sensitivity of the elephant trunk is non-existent, or that the authors have misinterpreted this staining. That there is extensive staining in the fibre pathway dorsal and lateral to the IOR (which I call the spinal trigeminal tract), supports the idea that the authors have misinterpreted their peripherin immunostaining.

(5) Evolutionary expediency. The authors argue that what they report is an expedient way in which to modify the organisation of the brainstem in the elephant to accommodate the "extraordinary" tactile sensitivity. I disagree. As pointed out in my first review, the elephant cerebellum is very large and comprised of huge numbers of morphologically complex neurons. The inferior olivary nuclei in all mammals studied in detail to date, give rise to the climbing fibres that terminate on the Purkinje cells of the cerebellar cortex. It is more parsimonious to argue that, in alignment with the expansion of the elephant cerebellum (for motor control of the trunk), the inferior olivary nuclei (specifically the principal nucleus) have had additional neurons added to accommodate this cerebellar expansion. Such an addition of neurons to the principal nucleus of the inferior olive could readily lead to the loss of the serrated appearance of the principal nucleus of the inferior olive and would require far less modifications in the developmental genetic program that forms these nuclei. This type of quantitative change appears to be the primary way in which structures are altered in the mammalian brainstem.

Author Response:

Reviewer #1 (Public Review):

Force sensing and gating mechanisms of the mechanically activated ion channels is an area of broad interest in the field of mechanotransduction. These channels perform important biological functions by converting mechanical force into electrical signals. To understand their underlying physiological processes, it is important to determine gating mechanisms, especially those mediated by lipids. The authors in this manuscript describe a mechanism for mechanically induced activation of TREK-1 (TWIK-related K+ channel. They propose that force induced disruption of ganglioside (GM1) and cholesterol causes relocation of TREK-1 associated with phospholipase D2 (PLD2) to 4,5-bisphosphate (PIP2) clusters, where PLD2 catalytic activity produces phosphatidic acid that can activate the channel. To test their hypothesis, they use dSTORM to measure TREK-1 and PLD2 colocalization with either GM1 or PIP2. They find that shear stress decreases TREK-1/PLD2 colocalization with GM1 and relocates to cluster with PIP2. These movements are affected by TREK-1 C-terminal or PLD2 mutations suggesting that the interaction is important for channel re-location. The authors then draw a correlation to cholesterol suggesting that TREK-1 movement is cholesterol dependent. It is important to note that this is not the only method of channel activation and that one not involving PLD2 also exists. Overall, the authors conclude that force is sensed by ordered lipids and PLD2 associates with TREK-1 to selectively gate the channel. Although the proposed mechanism is solid, some concerns remain.

1) Most conclusions in the paper heavily depend on the dSTORM data. But the images provided lack resolution. This makes it difficult for the readers to assess the representative images.

The images were provided are at 300 dpi. Perhaps the reviewer is referring to contrast in Figure 2? We are happy to increase the contrast or resolution.

As a side note, we feel the main conclusion of the paper, mechanical activation of TREK-1 through PLD2, depended primarily on the electrophysiology in Figure 1b-c, not the dSTORM. But both complement each other.

2) The experiments in Figure 6 are a bit puzzling. The entire premise of the paper is to establish gating mechanism of TREK-1 mediated by PLD2; however, the motivation behind using flies, which do not express TREK-1 is puzzling.

The fly experiment shows that PLD mechanosensitivity is more evolutionarily conserved than TREK-1 mechanosensitivity. We should have made this clearer.

-Figure 6B, the image is too blown out and looks over saturated. Unclear whether the resolution in subcellular localization is obvious or not.

Figure 6B is a confocal image, it is not dSTORM. There is no dSTORM in Figure 6. This should have been made clear in the figure legend. For reference, only a few cells would fit in the field of view with dSTORM.

-Figure 6C-D, the differences in activity threshold is 1 or less than 1g. Is this physiologically relevant? How does this compare to other conditions in flies that can affect mechanosensitivity, for example?

Yes, 1g is physiologically relevant. It is almost the force needed to wake a fly from sleep (1.2-3.2g). See ref 33. Murphy Nature Pro. 2017.

3) 70mOsm is a high degree of osmotic stress. How confident are the authors that a. cell health is maintained under this condition and b. this does indeed induce membrane stretch? For example, does this stimulation activate TREK-1?

Yes, osmotic swell activates TREK1. This was shown in ref 19 (Patel et al 1998). We agree the 70 mOsm is a high degree of stress. This needs to be stated better in the paper.

Reviewer #2 (Public Review):

This manuscript by Petersen and colleagues investigates the mechanistic underpinnings of activation of the ion channel TREK-1 by mechanical inputs (fluid shear or membrane stretch) applied to cells. Using a combination of super-resolution microscopy, pair correlation analysis and electrophysiology, the authors show that the application of shear to a cell can lead to changes in the distribution of TREK-1 and the enzyme PhospholipaseD2 (PLD2), relative to lipid domains defined by either GM1 or PIP2. The activation of TREK-1 by mechanical stimuli was shown to be sensitized by the presence of PLD2, but not a catalytically dead xPLD2 mutant. In addition, the activity of PLD2 is increased when the molecule is more associated with PIP2, rather than GM1 defined lipid domains. The presented data do not exclude direct mechanical activation of TREK-1, rather suggest a modulation of TREK-1 activity, increasing sensitivity to mechanical inputs, through an inherent mechanosensitivity of PLD2 activity. The authors additionally claim that PLD2 can regulate transduction thresholds in vivo using Drosophila melanogaster behavioural assays. However, this section of the manuscript overstates the experimental findings, given that it is unclear how the disruption of PLD2 is leading to behavioural changes, given the lack of a TREK-1 homologue in this organism and the lack of supporting data on molecular function in the relevant cells.

We agree, the downstream effectors of PLD2 mechanosensitivity are not known in the fly. Other anionic lipids have been shown to mediate pain see ref 46 and 47. We do not wish to make any claim beyond PLD2 being an in vivo contributor to a fly’s response to mechanical force.

That said we do believe we have established a molecular function at the cellular level. We showed PLD is robustly mechanically activated in a cultured fly cell line (BG2-c2) Figure 6a of the manuscript. And our previous publication established mechanosensation of PLD (Petersen et. al. Nature Com 2016) through mechanical disruption of the lipids. At a minimum, the experiments show PLDs mechanosensitivity is evolutionarily better conserved across species than TREK1.

This work will be of interest to the growing community of scientists investigating the myriad mechanisms that can tune mechanical sensitivity of cells, providing valuable insight into the role of functional PLD2 in sensitizing TREK-1 activation in response to mechanical inputs, in some cellular systems.

The authors convincingly demonstrate that, post application of shear, an alteration in the distribution of TREK-1 and mPLD2 (in HEK293T cells) from being correlated with GM1 defined domains (no shear) to increased correlation with PIP2 defined membrane domains (post shear). These data were generated using super-resolution microscopy to visualise, at sub diffraction resolution, the localisation of labelled protein, compared to labelled lipids. The use of super-resolution imaging enabled the authors to visualise changes in cluster association that would not have been achievable with diffraction limited microscopy. However, the conclusion that this change in association reflects TREK-1 leaving one cluster and moving to another overinterprets these data, as the data were generated from static measurements of fixed cells, rather than dynamic measurements capturing molecular movements.

When assessing molecular distribution of endogenous TREK-1 and PLD2, these molecules are described as "well correlated: in C2C12 cells" however it is challenging to assess what "well correlated" means, precisely in this context. This limitation is compounded by the conclusion that TREK-1 displayed little pair correlation with GM1 and the authors describe a "small amount of TREK-1 trafficked to PIP2". As such, these data may suggest that the findings outlined for HEK293T cells may be influenced by artefacts arising from overexpression.

The changes in TREK-1 sensitivity to mechanical activation could also reflect changes in the amount of TREK-1 in the plasma membrane. The authors suggest that the presence of a leak currently accounts for the presence of TREK-1 in the plasma membrane, however they do not account for whether there are significant changes in the membrane localisation of the channel in the presence of mPLD2 versus xPLD2. The supplementary data provide some images of fluorescently labelled TREK-1 in cells, and the authors state that truncating the c-terminus has no effect on expression at the plasma membrane, however these data provide inadequate support for this conclusion. In addition, the data reporting the P50 should be noted with caution, given the lack of saturation of the current in response to the stimulus range.

We thank the reviewer for his/her concern about expression levels. We did test TREK-1 expression. mPLD decreases TREK-1 expression ~two-fold (see Author response image 1). We did not include the mPLD data since TREK-1 was mechanically activated with mPLD. For expression to account for the loss of TREK-1 stretch current (Figure 1b), xPLD would need to block surface expression of TREK-1. The opposite was true, xPLD2 increased TREK-1 expression increased (see Figure S2c). Furthermore, we tested the leak current of TREK-1 at 0 mV and 0 mmHg of stretch. Basal leak current was no different with xPLD2 compared to endogenous PLD (Figure 1d; red vs grey bars respectively) suggesting TREK-1 is in the membrane and active when xPLD2 is present. If anything, the magnitude of the effect with xPLD would be larger if the expression levels were equal.

Author response image 1.

TREK expression at the plasma membrane. TREK-1 Fluorescence was measured by GFP at points along the plasma membrane. Over expression of mouse PLD2 (mPLD) decrease the amount of full-length TREK-1 (FL TREK) on the surface more than 2-fold compared to endogenously expressed PLD (enPLD) or truncated TREK (TREKtrunc) which is missing the PLD binding site in the C-terminus. Over expression of mPLD had no effect on TREKtrunc.

Finally, by manipulating PLD2 in D. melanogaster, the authors show changes in behaviour when larvae are exposed to either mechanical or electrical inputs. The depletion of PLD2 is concluded to lead to a reduction in activation thresholds and to suggest an in vivo role for PA lipid signaling in setting thresholds for both mechanosensitivity and pain. However, while the data provided demonstrate convincing changes in behaviour and these changes could be explained by changes in transduction thresholds, these data only provide weak support for this specific conclusion. As the authors note, there is no TREK-1 in D. melanogaster, as such the reported findings could be accounted for by other explanations, not least including potential alterations in the activation threshold of Nav channels required for action potential generation. To conclude that the outcomes were in fact mediated by changes in mechanotransduction, the authors would need to demonstrate changes in receptor potential generation, rather than deriving conclusions from changes in behaviour that could arise from alterations in resting membrane potential, receptor potential generation or the activity of the voltage gated channels required for action potential generation.

We are willing to restrict the conclusion about the fly behavior as the reviewers see fit. We have shown PLD is mechanosensitivity in a fly cell line, and when we knock out PLD from a fly, the animal exhibits a mechanosensation phenotype.

This work provides further evidence of the astounding flexibility of mechanical sensing in cells. By outlining how mechanical activation of TREK-1 can be sensitised by mechanical regulation of PLD2 activity, the authors highlight a mechanism by which TREK-1 sensitivity could be regulated under distinct physiological conditions.

Reviewer #3 (Public Review):

The manuscript "Mechanical activation of TWIK-related potassium channel by nanoscopic movement and second messenger signaling" presents a new mechanism for the activation of TREK-1 channel. The mechanism suggests that TREK1 is activated by phosphatidic acids that are produced via a mechanosensitive motion of PLD2 to PIP2-enriched domains. Overall, I found the topic interesting, but several typos and unclarities reduced the readability of the manuscript. Additionally, I have several major concerns on the interpretation of the results. Therefore, the proposed mechanism is not fully supported by the presented data. Lastly, the mechanism is based on several previous studies from the Hansen lab, however, the novelty of the current manuscript is not clearly stated. For example, in the 2nd result section, the authors stated, "fluid shear causes PLD2 to move from cholesterol dependent GM1 clusters to PIP2 clusters and this activated the enzyme". However, this is also presented as a new finding in section 3 "Mechanism of PLD2 activation by shear."

For PLD2 dependent TREK-1 activation. Overall, I found the results compelling. However, two key results are missing. 1. Does HEK cells have endogenous PLD2? If so, it's hard to claim that the authors can measure PLD2-independent TREK1 activation.

Yes, there is endogenous PLD (enPLD). We calculated the relative expression of xPLD2 vs enPLD. xPLD2 is >10x more abundant (Fig. S3d of Pavel et al PNAS 2020, ref 14 of the current manuscript). Hence, as with anesthetic sensitivity, we expect the xPLD to out compete the endogenous PLD, which is what we see. This should have been described more carefully in this paper and the studies pointed out that establish this conclusion.

1. Does the plasma membrane trafficking of TREK1 remain the same under different conditions (PLD2 overexpression, truncation)? From Figure S2, the truncated TREK1 seem to have very poor trafficking. The change of trafficking could significantly contribute to the interpretation of the data in Figure 1.

If the PLD2 binding site is removed (TREK-1trunc), yes, the trafficking to the plasma membrane is unaffected by the expression of xPLD and mPLD (Figure R1 above). For full length TREK1 (FL-TREK-1), co-expression of mPLD decreases TREK expression (Figure R1) and co-expression with xPLD increases TREK expression (Figure S2). This is exactly opposite of what one would expect if surface expression accounted for the change in pressure currents. Hence, we conclude surface expression does not account for loss of TREK-1 mechanosensitivity with xPLD2.

For shear-induced movement of TREK1 between nanodomains. The section is convincing, however I'm not an expert on super-resolution imaging. Also, it would be helpful to clarify whether the shear stress was maintained during fixation. If not, what is the time gap between reduced shear and the fixed state. lastly, it's unclear why shear flow changes the level of TREK1 and PIP2.

Shear was maintained during the fixing. We do not know why shear changes PIP2 and TREK-1 levels. Presumably endocytosis and or release of other lipid modifying enzymes affect the system. The change in TREK-1 levels appears to be directly through an interaction with PLD as TREKtrunc is not affected by over expression of xPLD or mPLD.

For the mechanism of PLD2 activation by shear. I found this section not convincing. Therefore, the question of how does PLD2 sense mechanical force on the membrane is not fully addressed. Particularly, it's hard to imagine an acute 25% decrease cholesterol level by shear - where did the cholesterol go? Details on the measurements of free cholesterol level is unclear and additional/alternative experiments are needed to prove the reduction in cholesterol by shear.

The question “how does PLD2 sense mechanical force on the membrane” we addressed and published in Nature Comm. In 2016. The title of that paper is “Kinetic disruption of lipid rafts is a mechanosensor for phospholipase D” see ref 13 Petersen et. al. PLD is a soluble protein associated to the membrane through palmitoylation. There is no transmembrane domain, which narrows the possible mechanism of its mechanosensation to disruption.

The Nature Comm. reviewer identified as “an expert in PLD signaling” wrote the following of our data and the proposed mechanism:

"This is a provocative report that identifies several unique properties of phospholipase D2 (PLD2). It explains in a novel way some long established observations including that the enzyme is largely regulated by substrate presentation which fits nicely with the authors model of segregation of the two lipid raft domains (cholesterol ordered vs PIP2 containing). Although PLD has previously been reported to be involved in mechanosensory transduction processes (as cited by the authors) this is the first such report associating the enzyme with this type of signaling... It presents a novel model that is internally consistent with previous literature as well as the data shown in this manuscript. It suggests a new role for PLD2 as a force transduction tied to the physical structure of lipid rafts and uses parallel methods of disruption to test the predictions of their model."

Regarding cholesterol. We use a fluorescent cholesterol oxidase assay which we described in the methods. This is an appropriate assay for determining cholesterol levels in a cell which we use routinely. We have published in multiple journals using this method, see references 28, 30, 31. Working out the metabolic fate of cholesterol after sheer is indeed interesting but well beyond the scope of this paper. Furthermore, we indirectly confirmed our finding using dSTORM cluster analysis (Figure 3d-e). The cluster analysis shows a decrease in GM1 cluster size consistent with our previous experiments where we chemically depleted cholesterol and saw a similar decrease in cluster size (see ref 13). All the data are internally consistent, and the cholesterol assay is properly done. We see no reason to reject the data.

Importantly, there is no direct evidence for "shear thinning" of the membrane and the authors should avoid claiming shear thinning in the abstract and summary of the manuscript.

We previously established a kinetic model for PLD2 activation see ref 13 (Petersen et al Nature Comm 2016). In that publication we discussed both entropy and heat as mechanisms of disruption. Here we controlled for heat which narrowed that model to entropy (i.e., shear thinning) (see Figure 3c). We provide an overall justification below. But this is a small refinement of our previous paper, and we prefer not to complicate the current paper. We believe the proper rheological term is shear thinning. The following justification, which is largely adapted from ref 13, could be added to the supplement if the reviewer wishes.

Justification: To establish shear thinning in a biological membrane, we initially used a soluble enzyme that has no transmembrane domain, phospholipase D2 (PLD2). PLD2 is a soluble enzyme and associated with the membrane by palmitate, a saturated 16 carbon lipid attached to the enzyme. In the absence of a transmembrane domain, mechanisms of mechanosensation involving hydrophobic mismatch, tension, midplane bending, and curvature can largely be excluded. Rather the mechanism appears to be a change in fluidity (i.e., kinetic in nature). GM1 domains are ordered, and the palmate forms van der Waals bonds with the GM1 lipids. The bonds must be broken for PLD to no longer associate with GM1 lipids. We established this in our 2016 paper, ref 13. In that paper we called it a kinetic effect, however we did not experimentally distinguish enthalpy (heat) vs. entropy (order). Heat is Newtonian and entropy (i.e., shear thinning) is non-Newtonian. In the current study we paid closer attention to the heat and ruled it out (see Figure 3c and methods). We could propose a mechanism based on kinetic disruption, but we know the disruption is not due to melting of the lipids (enthalpy), which leaves shear thinning (entropy) as the plausible mechanism.

The authors should also be aware that hypotonic shock is a very dirty assay for stretching the cell membrane. Often, there is only a transient increase in membrane tension, accompanied by many biochemical changes in the cells (including acidification, changes of concentration etc). Therefore, I would not consider this as definitive proof that PLD2 can be activated by stretching membrane.

Comment noted. We trust the reviewer is correct. In 1998 osmotic shock was used to activate the channel. We only intended to show that the system is consistent with previous electrophysiologic experiments.

Author response:

The following is the authors’ response to the current reviews.

We thank you for sending our manuscript for the second round of review.  We are encouraged by the comments from reviewer #2 that our supplementary work on naïve T cells and antibody blockade work satisfied their previous concerns and is important for our work.

The Editors raised concerns that we have shared preliminary data on Nrn1 and AMPAR double knockout mice.  We apologize for our enthusiasm for these studies.  Because of the publication model by eLife, we shared that data not because we needed to persuade the reviewer for publication purposes but rather to agree with the reviewer that the molecular target of Nrn1 is important, and we are progressing in understanding this subject.

The following is the authors’ response to the original reviews.

To Reviewer #1:

Thank you for your thorough review and comments on our work, which you described as “the role of neuritin in T cell biology studied here is new and interesting.”.  We have summarized your comments into two categories: biology and investigation approach, experimental rigor, and data presentation.

Biology and Investigation approach comments:

(1) Questions regarding the T cell anergy model:

Major point “(4) Figure 1E-H. The authors assume that this immunization protocol induces anergic cells, but they provide no experimental evidence for this. It would be useful to show that T cells are indeed anergic in this model, especially those that are OVA-specific. The lack of IL-2 production by Cltr cells could be explained by the presence of fewer OVA-specific cells, rather than by an anergic status.”

T cell anergy is a well-established concept first described by Schwartz’s group. It refers to the hyporesponsive T cell functional state in antigen-experienced CD4 T cells (Chappert and Schwartz, 2010; Fathman and Lineberry, 2007; Jenkins and Schwartz, 1987; Quill and Schwartz, 1987).  Anergic T cells are characterized by their inability to expand and to produce IL2 upon subsequent antigen re-challenge. In this paper, we have borrowed the existing in vivo T cell anergy induction model used by Mueller’s group for T cell anergy induction (Vanasek et al., 2006).  Specifically, Thy1.1+ Ctrl or Nrn1-/- TCR transgenic OTII cells were co-transferred with the congenically marked Thy1.2+ WT polyclonal Treg cells into TCR-/- mice.  After anergy induction, the congenically marked TCR transgenic T cells were recovered by sorting based on Thy1.1+ congenic marker, and subsequently re-stimulation ex vivo with OVA323-339 peptide. We evaluated the T cell anergic state based on OTII cell expansion in vivo and IL2 production upon OVA323-339 restimulation ex vivo.  

“The authors assume that this immunization protocol induces anergic cells, but they provide no experimental evidence for this.”

Because the anergy model by Mueller's group is well established (Vanasek et al., 2006), we did not feel that additional effort was required to validate this model as the reviewer suggested. Moreover, the limited IL2 production among the control cells upon restimulation confirms the validity of this model.

“The lack of IL-2 production by Cltr cells could be explained by the presence of fewer OVAspecific cells, rather than by an anergic status”.

Cells from Ctrl and Nrn1-/- mice on a homogeneous TCR transgenic (OTII) background were used in these experiments. The possibility that substantial variability of TCR expression or different expression levels of the transgenic TCR could have impacted IL2 production rather than anergy induction is unlikely.

Overall, we used this in vivo anergy model to evaluate the Nrn1-/- T cell functional state in comparison to Ctrl cells under the anergy induction condition following the evaluation of Nrn1 expression, particularly in anergic T cells.  Through studies using this anergy model, we observed a significant change in Treg induction among OTII cells. We decided to pursue the role of Nrn1 in Treg cell development and function rather than the biology of T cell anergy as evidenced by subsequent experiments.

Minor points “(6) On which markers are anergic cells sorted for RNAseq analysis?”

Cells were sorted out based on their congenic marker marking Ctrl or Nrn1-/- OTII cells transferred into the host mice.  We did not specifically isolate anergic cells for sequencing.

(2) Question regarding the validity of iTreg differentiation model.

Major point: “(5) Figure 2A-C and Figure 3. The use of iTregs to try to understand what is happening in vivo is problematic. iTregs are cells that have probably no equivalent in vivo, and so may have no physiological relevance. In any case, they are different from pTreg cells generated in vivo. Working with pTreg may be challenging, that is why I would suggest generating data with purified nTreg. Moreover, it was shown in the article of Gonzalez-Figueroa 2021 that Nrn1-/- nTreg retained a normal suppressive function, which would not be what is concluded by the authors of this manuscript. Moreover, we do not even know what the % of Foxp3 cells is in the iTreg used (after differentiation and 20h of re-stimulation) and whether this % is the same between Ctlr and Nrn1 KO cells.”.

We thank Reviewer #1 for their feedback. While it is true that iTregs made in vitro and in vivo generated pTregs display several distinctions (e. g., differences in Foxp3 expression stability, for example), we strongly disagree with this statement by Revieweer#1 “The use of iTregs to try to understand what is happening in vivo is problematic. iTregs are cells that have probably no equivalent in vivo, and so may have no physiological relevance.”  The induced Treg cell (iTreg) model was established over 20 years ago (Chen et al., 2003; Zheng et al., 2002), and the model is widely adopted with over 2000 citations. Further, it has been instrumental in understanding different aspects of regulatory T cell biology (Hurrell et al., 2022; John et al., 2022; Schmitt and Williams, 2013; Sugiura et al., 2022).   

Because we have observed reduced pTreg generation in vivo, we choose to use the in vitro iTreg model system to understand the mechanistic changes involved in Treg cell differentiation and function, specifically, neuritin’s role in this process. We have made no claim that iTreg cell biology is identical to pTreg generated in vivo or nTreg cells. However, the iTreg culture system has proved to be a good in vitro system for deciphering molecular events involved in complex processes. As such, it remains a commonly used approach by many research groups in the Treg cell field (Hurrell et al., 2022; John et al., 2022; Sugiura et al., 2022). Moreover, applying the iTreg in vitro culture system has been instrumental in helping us identify the cell electrical state change in Nrn1-/- CD4 cells and revealed the biological link between Nrn1 and the ionotropic AMPA receptor (AMPAR), which we will discuss in the subsequent discussion. It is technically challenging to use nTreg cells for T cell electrical state studies due to their heterogeneous nature from development in an in vivo environment and the effect of manipulation during the nTreg cell isolation process, which can both affect the T cell electrical state.   

“Moreover, it was shown in the article of Gonzalez-Figueroa 2021 that Nrn1-/- nTreg retained a normal suppressive function, which would not be what is concluded by the authors of this manuscript.” 

We have also carried out nTreg studies in vitro in addition to iTreg cells. Similar to Gonzalez-Figueroa et al.'s findings, we did not observe differences in suppression function between Nrn1-/- and WT nTreg using the in vitro suppression assay. However, Nrn1-/- nTreg cells revealed reduced suppression function in vivo (Fig. 2D-L). In fact, Gonzalez-Figueroa et al. observed reduced plasma cell formation after OVA immunization in Treg-specific Nrn1-/- mice, implicating reduced suppression from Nrn1-/- follicular regulatory T (Tfr) cells. Thus, our observation of the reduced suppression function of Nrn1-/- nTreg toward effector T cell expansion, as presented in Fig. 2D-L, does not contradict the results from Gonzalez-Figueroa et al. Rather, the conclusions of these two studies agree that Nrn1 can play important roles in immune suppression observable in vivo that are not captured readily by the in vitro suppression assay.

“Moreover, we do not even know what the % of Foxp3 cells is in the iTreg used (after differentiation and 20h of re-stimulation) and whether this % is the same between Ctlr and Nrn1 KO cells.”

We have stated in the manuscript on page 7 line 208 that “Similar proportions of Foxp3+ cells were observed in Nrn1-/- and Ctrl cells under the iTreg culture condition, suggesting that Nrn1 deficiency does not significantly impact Foxp3+ cell differentiation”. In the revised manuscript, we will include the data on the proportion of Foxp3+ cells before iTreg restimulation.

(3) Confirmation of transcriptomic data regarding amino acids or electrolytes transport change

Minor point“(3) Would not it be possible to perform experiments showing the ability of cells to transport amino acids or electrolytes across the plasma membrane? This would be a more interesting demonstration than transcriptomic data.”

We appreciate Review# 1’s suggestion regarding “perform experiments showing the ability of cells to transport amino acids or electrolytes across the plasma membrane”.  We have indeed already performed such experiments corroborating the transcriptomics data on differential amino acid and nutrient transporter expression. Specifically, we loaded either iTreg or Th0 cells with membrane potential (MP) dye and measured MP level change after adding the complete set of amino acids (complete AA).  Upon entry, the charge carried by AAs may transiently affect cell membrane potential. Different AA transporter expression patterns may show different MP change patterns upon AA entry, as we showed in Author response image 1. We observed reduced MP change in Nrn1-/- iTreg compared to the Ctrl, whereas in the context of Th0 cells, Nrn1-/- showed enhanced MP change than the Ctrl. We can certainly include these data in the revised manuscript.

Author response image 1.

Membrane potential change induced by amino acids entry. a. Nrn1-/- or WT iTreg cells loaded with MP dye and MP change was measured upon the addition of a complete set of AAs. b. Nrn1-/- or WT Th0 cells loaded with MP dye and MP change was measured upon the addition of a complete set of AAs.

(4) EAE experiment data assessment

Minor point ”(5) Figure 5F. How are cells re-stimulated? If polyclonal stimulation is used, the experiment is not interesting because the analysis is done with lymph node cells. This analysis should either be performed with cells from the CNS or with MOG restimulation with lymph node cells.”

In the EAE study, the Nrn1-/- mice exhibit similar disease onset but a protracted non-resolving disease phenotype compared to the WT control mice.  Several reasons may contribute to this phenotype: 1. Enhanced T effector cell infiltration/persistence in the central nervous system (CNS); 2. Reduced Treg cell-mediated suppression to the T effector cells in the CNS; 3. Protracted non-resolving inflammation at the immunization site has the potential to continue sending T effector cells into CNS, contributing to persistent inflammation. Based on this reasoning, we examined the infiltrating T effector cell number and Treg cell proportion in the CNS.  We also restimulated cells from draining lymph nodes close to the inflammation site, looking for evidence of persistent inflammation.  When mice were harvested around day 16 after immunization, the inflammation at the local draining lymph node should be at the contraction stage.  We stimulated cells with PMA and ionomycin intended to observe all potential T effector cells involved in the draining lymph node rather than only MOG antigen-specific cells.  We disagree with Reviewer #1’s assumption that “This analysis should either be performed with cells from the CNS or with MOG restimulation with lymph node cells.”. We think the experimental approach we have taken has been appropriately tailored to the biological questions we intended to answer.

Experimental rigor and data presentation.

(1) data labeling and additional supporting data

Major points

(2) The authors use Nrn1+/+ and Nrn1+/- cells indiscriminately as control cells on the basis of similar biology between Nrn1+/+ and Nrn1+/- cells at homeostasis. However, it is quite possible that the Nrn1+/- cells have a phenotype in situations of in vitro activation or in vivo inflammation (cancer, EAE). It would be important to discriminate Nrn1+/- and Nrn1+/+ cells in the data or to show that both cell types have the same phenotype in these conditions too.

(3) Figure 1A-D. Since the authors are using the Nrp1 KO mice, it would be important to confirm the specificity of the anti-Nrn1 mAb by FACS. Once verified, it would be important to add FACS results with this mAb in Figures 1A-C to have single-cell and quantitative data as well.

Minor points  

(1) Line 119, 120 of the text. It is said that one of the most up-regulated genes in anergic cells is Nrn1 but the data is not shown.

(2) For all figures showing %, the titles of the Y axes are written in an odd way. For example, it is written "Foxp3% CD4". It would be more conventional and clearer to write "% Foxp3+ / CD4+" or "% Foxp3+ among CD4+".

(4) For certain staining (Figure 3E, H) it would be important to show the raw data, in addition to MFI or % values.

We can adapt the labeling and provide additional data, including Nrn1 staining on Treg cells and flow graphs for pmTOR and pS6 staining (Fig. 3H), as requested by Reviewer #1.

(2) Experimental rigor:

General comments:

“However, it is disappointing that reading this manuscript leaves an impression of incomplete work done too quickly.”

We were discouraged to receive the comment, “this manuscript leaves an impression of incomplete work done too quickly.” Our study of this novel molecule began without any existing biological tools such as antibodies, knockout mice, etc.  Over the past several years, we have established our own antibodies for Nrn1 detection, obtained and characterized Nrn1 knockout mice, and utilized multiple approaches to identify the molecular mechanism of Nrn1 function. Through the use of the in vitro iTreg system described in this manuscript, we identified the association of Nrn1 deficiency with cell electrical state change, potentially connected to AMPAR function. We have further corroborated our findings by generating Nrn1 and AMPAR T cell specific double knockout mice and confirmed that T cell specific AMPAR deletion could abrogate the phenotype caused by the Nrn1 deficiency (see Support Figure 2).  We did not include the double knockout data in the current manuscript because AMPAR function has not yet been studied thoroughly in T cell biology, and we feel this topic warrants examination in its own right.  However, the unpublished data support the finding that Nrn1 modulates the T cell electrical state and, consequently, metabolism, ultimately influencing tolerance and immunity.  In its current form, the manuscript represents the first characterization of the novel molecule Nrn1 in anergic cells, Tregs, and effector T cells. While this work has led to several exciting additional questions, we disagree that the novel characterization we have presented Is incomplete. We feel that our present data set, which squarely highlights Nrn1’s role as an important immune regulator while shedding unprecedented light on the molecular events involved, will be of considerable interest to a broad field of researchers.

“Multiple models have been used, but none has been studied thoroughly enough to provide really conclusive and unambiguous data. For example, 5 different models were used to study T cells in vivo. It would have been preferable to use fewer, but to go further in the study of mechanisms.”

We have indeed used multiple in vivo models to reveal Nrn1's function in Treg differentiation, Treg suppression function, T effector cell differentiation and function, and the overall impact on autoimmune disease. Because the impact of ion channel function is often context-dependent, we examined the biological outcome of Nrn1 deficiency in several in vivo contexts.  We would appreciate it if Reviewer#1 would provide a specific example, given the Nrn1 phenotype, of how to proceed deeper to investigate the electrical change in the in vivo models.

“Major points

(1) A real weakness of this work is the fact that in most of the results shown, there are few biological replicates with differences that are often small between Ctrl and Nrn1 -/-. The systematic use of student's t-test may lead to thinking that the differences are significant, which is often misleading given the small number of samples, which makes it impossible to know whether the distributions are Gaussian and whether a parametric test can be used. RNAseq bulk data are based on biological duplicates, which is open to criticism.”

We respectfully disagree with Reviewer #1 on the question of statistical power and significance to our work. We have used 5-8 mice/group for each in vivo model and 3-4 technical replicates for the in vitro studies, with a minimum of 2-3 replicate experiments. These group sizes and replication numbers are in line with those seen in high-impact publications. While some differences between Ctrl and Nrn1-/- appear small, they have significant biological consequences, as evidenced by the various Nrn1-/- in vivo phenotypes. Furthermore, we believe we have subjected our data to the appropriate statistical tests to ensure rigorous analysis and representation of our findings.

To Reviewer #2.

We thank Reviewer #2 for the careful review of the manuscript. We especially appreciate the comments that “The characterizations of T cell Nrn1 expression both in vitro and in vivo are comprehensive and convincing. The in vivo functional studies of anergy development, Treg suppression, and EAE development are also well done to strengthen the notion that Nrn1 is an important regulator of CD4 responsiveness.”

“The major weakness of this study stems from a lack of a clear molecular mechanism involving Nrn1. “  

We fully understand this comment from Reviewer #2. The main mechanism we identified contributing to the functional defect of Nrn1-/- T cells involves novel effects on the electric and metabolic state of the cells. Although we referenced neuronal studies that indicate Nrn1 is the auxiliary protein for the ionotropic AMPA-type glutamate receptor (AMPAR) and may affect AMPAR function, we did not provide any evidence in this manuscript as the topic requires further in-depth study.   

For the benefit of this discussion, we include our preliminary Nrn1 and AMPAR double knockout data (Author response image 2), which indicates that abrogating AMPAR expression can compensate for the defect caused by Nrn1 deficiency in vitro and in vivo. This preliminary data supports the notion that Nrn1 modulates AMPAR function, which causes changes in T cell electric and metabolic state, influencing T cell differentiation and function.  

Author response image 2.

Deletion of AMPAR expression in T cells compensates for the defect caused by Nrn1 deficiency. Nrn1-/- mice were crossed with T cell-specific AMPAR knockout mice (AMPARfl/flCD4Cre+) mice. The following mice were generated and used in the experiment: T cell specific AMPAR-knockout and Nrn1 knockout mice (AKONKO), Nrn1 knockout mice (AWTNKO), Ctrl mice (AWTNWT). a. Deletion of AMPAR compensates for the iTreg cell defect observed in Nrn1-/- CD4 cells. iTreg live cell proportion, cell number, and Ki67 expression among Foxp3+ cells 3 days after aCD3 restimulation. b. Deletion of AMPAR in T cells abrogates the enhanced autoimmune response in Nrn1-/- Mouse in the EAE disease model. Mouse relative weight change and disease score progression after EAE disease induction.  

Ion channels can influence cell metabolism through multiple means (Vaeth and Feske, 2018; Wang et al., 2020). First, ion channels are involved in maintaining cell resting membrane potential. This electrical potential difference across the cell membrane is essential for various cellular processes, including metabolism (Abdul Kadir et al., 2018; Blackiston et al., 2009; Nagy et al., 2018; Yu et al., 2022). Second, ion channels facilitate the movement of ions across cell membranes. These ions are essential for various metabolic processes. For example, ions like calcium (Ca2+), potassium (K+), and sodium (Na+) play crucial roles in signaling pathways that regulate metabolism (Kahlfuss et al., 2020). Third, ion channel activity can influence cellular energy balance due to ATP consumption associated with ion transport to maintain ion balances (Erecińska and Dagani, 1990; Gerkau et al., 2019). This, in turn, can impact processes like ATP production, which is central to cellular metabolism. Thus, ion channel expression and function determine the cell’s bioelectric state and contribute to cell metabolism (Levin, 2021).

Because the AMPAR function has not been thoroughly studied using a genetic approach in T cells, we do not intend to include the double knockout data in this manuscript before fully characterizing the T cell-specific AMPAR knockout mice.  

“Although the biochemical and informatics studies are well-performed, it is my opinion that these results are inconclusive in part due to the absence of key "naive" control groups. This limits my ability to understand the significance of these data.

Specifically, studies of the electrical and metabolic state of Nrn1-/- inducible Treg cells (iTregs) would benefit from similar data collected from wild-type and Nrn1-/- naive CD4 T cells.”

We appreciate the reviewer’s comments. This comment reflects two concerns in data interpretation:

(1) Are Nrn1-/- naïve T cells fundamentally different from WT cells? Does this fundamental difference contribute to the observed electrical and metabolic phenotype in iTreg or Th0 cells? This is a very good question we will perform the experiments as the reviewer suggested. While Nrn1 is expressed at a basal (low) level in naïve T cells, deletion of Nrn1 may cause changes in naïve T cell phenotype.   

(2) Is the Nrn1-/- phenotype caused by Nrn1 functional deficiency or due to the secondary effect of Nrn1 deletion, such as non-physiological cell membrane structure changes?

We have done the following experiment to address this concern.  We have cultured WT T cells in the presence of Nrn1 antibody and compared the outcome with Nrn1-/- iTreg cells (Figure 3-figure supplement 2D,E,F). WT iTreg cells under antibody blockade exhibited similar changes as Nrn1-/- iTreg cells, confirming the physiological relevance of the Nrn1-/- phenotype.

Manuscript Revision based on the Reviewer’s suggestions:

Reviewer #1:

Major points (3) Figure 1A-D. Since the authors are using the Nrp1 KO mice, it would be important to confirm the specificity of the anti-Nrn1 mAb by FACS. 

Following the suggestion by Reviewer#1, We have included the Nrn1 Ab staining on activated Nrn1-/- CD4 cells in Figure 1D. We have also added the staining of cell surface Nrn1 on Treg cells in Figure 1-figure supplement 1D.

Major point: (5) “Moreover, we do not even know what the % of Foxp3 cells is in the iTreg used (after differentiation and 20h of re-stimulation) and whether this % is the same between Ctlr and Nrn1 KO cells.”

In the revised manuscript, we have included the proportion of Foxp3+ cells among Nrn1-/- and ctrl iTreg cells developed under the iTreg culture condition in Figure 2A.

Minor points  

(2) For all figures showing %, the titles of the Y axes are written in an odd way. For example, it is written "Foxp3% CD4". It would be more conventional and clearer to write "% Foxp3+ / CD4+" or "% Foxp3+ among CD4+".

Following reviewer#1’s suggestion, we have changed the Y-axis label in all the relevant figures.

(3) Would not it be possible to perform experiments showing the ability of cells to transport amino acids or electrolytes across the plasma membrane? This would be a more interesting demonstration than transcriptomic data.”

We appreciate Review# 1’s suggestion regarding “perform experiments showing the ability of cells to transport amino acids or electrolytes across the plasma membrane”.  We have used AAinduced cellular MP changes to confirm differential AA transporter expression patterns and their impact on cellular MP levels.  The data are included in the revised manuscript in Figure 3H and Figure 4K.

(4) For certain staining (Figure 3E, H) it would be important to show the raw data, in addition to MFI or % values.

We appreciated Reviewer #1’s suggestion and have included the histogram staining data for Figure 3E. We have moved the original Figure 3H to the supplemental figure and included the histogram staining data in Figure 3-figure supplement 1C.  Similarly, we have included the histogram staining data in Figure 4-figure supplement 1C.

Reviewer#2:

“Although the biochemical and informatics studies are well-performed, it is my opinion that these results are inconclusive in part due to the absence of key "naive" control groups. This limits my ability to understand the significance of these data.

Specifically, studies of the electrical and metabolic state of Nrn1-/- inducible Treg cells (iTregs) would benefit from similar data collected from wild-type and Nrn1-/- naive CD4 T cells.”

We greatly appreciate Reviewer#2’s suggestion and have carried out experiments on naïve CD4 cells derived from Nrn1-/- and WT mice. We have compared membrane potential, AA-induced MP change between Nrn1-/- and WT naïve T cells, and the metabolic state of Nrn1-/- and WT naïve T cells by carrying out glucose stress tests and mitochondria stress tests using a seahorse assay.  Moreover, to investigate whether the phenotype revealed in Nrn1-/- CD4 cells was caused by a secondary effect of cell membrane structure change due to Nrn1 deletion, we carried out Nrn1 antibody blockade in WT CD4 cells and investigated the phenotypic change. These new results are included in Figure 3-figure supplement 2.

Reference:

Abdul Kadir, L., M. Stacey, and R. Barrett-Jolley. 2018. Emerging Roles of the Membrane Potential: Action Beyond the Action Potential. Front Physiol 9:1661.

Blackiston, D.J., K.A. McLaughlin, and M. Levin. 2009. Bioelectric controls of cell proliferation: ion channels, membrane voltage and the cell cycle. Cell Cycle 8:3527-3536.

Chappert, P., and R.H. Schwartz. 2010. Induction of T cell anergy: integration of environmental cues and infectious tolerance. Current opinion in immunology 22:552-559.

Chen, W., W. Jin, N. Hardegen, K.J. Lei, L. Li, N. Marinos, G. McGrady, and S.M. Wahl. 2003. Conversion of peripheral CD4+CD25- naive T cells to CD4+CD25+ regulatory T cells by TGF-beta induction of transcription factor Foxp3. The Journal of experimental medicine 198:1875-1886.

Erecińska, M., and F. Dagani. 1990. Relationships between the neuronal sodium/potassium pump and energy metabolism. Effects of K+, Na+, and adenosine triphosphate in isolated brain synaptosomes. J Gen Physiol 95:591-616.

Fathman, C.G., and N.B. Lineberry. 2007. Molecular mechanisms of CD4+ T-cell anergy. Nat Rev Immunol 7:599-609.

Gerkau, N.J., R. Lerchundi, J.S.E. Nelson, M. Lantermann, J. Meyer, J. Hirrlinger, and C.R. Rose. 2019. Relation between activity-induced intracellular sodium transients and ATP dynamics in mouse hippocampal neurons. The Journal of physiology 597:5687-5705.

Hurrell, B.P., D.G. Helou, E. Howard, J.D. Painter, P. Shafiei-Jahani, A.H. Sharpe, and O. Akbari. 2022. PD-L2 controls peripherally induced regulatory T cells by maintaining metabolic activity and Foxp3 stability. Nature communications 13:5118.

Jenkins, M.K., and R.H. Schwartz. 1987. Antigen presentation by chemically modified splenocytes induces antigen-specific T cell unresponsiveness in vitro and in vivo. The Journal of experimental medicine 165:302-319.

John, P., M.C. Pulanco, P.M. Galbo, Jr., Y. Wei, K.C. Ohaegbulam, D. Zheng, and X. Zang. 2022. The immune checkpoint B7x expands tumor-infiltrating Tregs and promotes resistance to anti-CTLA-4 therapy. Nature communications 13:2506.

Kahlfuss, S., U. Kaufmann, A.R. Concepcion, L. Noyer, D. Raphael, M. Vaeth, J. Yang, P. Pancholi, M. Maus, J. Muller, L. Kozhaya, A. Khodadadi-Jamayran, Z. Sun, P. Shaw, D. Unutmaz, P.B. Stathopulos, C. Feist, S.B. Cameron, S.E. Turvey, and S. Feske. 2020. STIM1-mediated calcium influx controls antifungal immunity and the metabolic function of nonpathogenic Th17 cells. EMBO molecular medicine 12:e11592.

Levin, M. 2021. Bioelectric signaling: Reprogrammable circuits underlying embryogenesis, regeneration, and cancer. Cell 184:1971-1989.

Nagy, E., G. Mocsar, V. Sebestyen, J. Volko, F. Papp, K. Toth, S. Damjanovich, G. Panyi, T.A. Waldmann, A. Bodnar, and G. Vamosi. 2018. Membrane Potential Distinctly Modulates Mobility and Signaling of IL-2 and IL-15 Receptors in T Cells. Biophys J 114:2473-2482.

Quill, H., and R.H. Schwartz. 1987. Stimulation of normal inducer T cell clones with antigen presented by purified Ia molecules in planar lipid membranes: specific induction of a long-lived state of proliferative nonresponsiveness. Journal of immunology (Baltimore, Md. : 1950) 138:3704-3712.

Schmitt, E.G., and C.B. Williams. 2013. Generation and function of induced regulatory T cells. Frontiers in immunology 4:152.

Sugiura, A., G. Andrejeva, K. Voss, D.R. Heintzman, X. Xu, M.Z. Madden, X. Ye, K.L. Beier, N.U. Chowdhury, M.M. Wolf, A.C. Young, D.L. Greenwood, A.E. Sewell, S.K. Shahi, S.N. Freedman, A.M. Cameron, P. Foerch, T. Bourne, J.C. Garcia-Canaveras, J. Karijolich, D.C. Newcomb, A.K. Mangalam, J.D. Rabinowitz, and J.C. Rathmell. 2022. MTHFD2 is a metabolic checkpoint controlling effector and regulatory T cell fate and function. Immunity 55:65-81.e69.

Vaeth, M., and S. Feske. 2018. Ion channelopathies of the immune system. Current opinion in immunology 52:39-50.

Vanasek, T.L., S.L. Nandiwada, M.K. Jenkins, and D.L. Mueller. 2006. CD25+Foxp3+ regulatory T cells facilitate CD4+ T cell clonal anergy induction during the recovery from lymphopenia. Journal of immunology (Baltimore, Md. : 1950) 176:5880-5889.

Wang, Y., A. Tao, M. Vaeth, and S. Feske. 2020. Calcium regulation of T cell metabolism. Current opinion in physiology 17:207-223.

Yu, W., Z. Wang, X. Yu, Y. Zhao, Z. Xie, K. Zhang, Z. Chi, S. Chen, T. Xu, D. Jiang, X. Guo, M. Li, J. Zhang, H. Fang, D. Yang, Y. Guo, X. Yang, X. Zhang, Y. Wu, W. Yang, and D. Wang. 2022. Kir2.1-mediated membrane potential promotes nutrient acquisition and inflammation through regulation of nutrient transporters. Nature communications 13:3544.

Zheng, S.G., J.D. Gray, K. Ohtsuka, S. Yamagiwa, and D.A. Horwitz. 2002. Generation ex vivo of TGF-beta-producing regulatory T cells from CD4+CD25- precursors. Journal of immunology (Baltimore, Md. : 1950) 169:4183-4189.

Author response:

The following is the authors’ response to the original reviews

Reviewer #1 (Public Review):  

Summary:

In this study, Setogawa et al. employ an auditory discrimination task in freely moving rats, coupled with small animal imaging, electrophysiological recordings, and pharmacological inhibition/lesioning experiments to better understand the role of two striatal subregions: the anterior Dorsal Lateral Striatum (aDLS) and the posterior Ventrolateral Striatum (pVLS), during auditory discrimination learning. Attempting to better understand the contribution of different striatal subregions to sensory discrimination learning strikes me as a highly relevant and timely question, and the data presented in this study are certainly of major interest to the field. The authors have set up a robust behavioral task and systematically tackled the question about a striatal role in learning with multiple observational and manipulative techniques. Additionally, the structured approach the authors take by using neuroimaging to inform their pharmacological manipulation experiments and electrophysiological recordings is a strength.

However, the results as they are currently presented are not easy to follow and could use some restructuring, especially the electrophysiology. Also, the main conclusion that the authors draw from the data, that aDLS and pVLS contribute to different phases of discrimination learning and influence the animal's response strategy in different ways, is not strongly supported by the data and deserves some additional caveats and limitations of the study in the discussion. 

We appreciate the reviewer’s valuable feedback, which has been beneficial for improvement of our manuscript. In response to the reviewer’s comments, we have revised multiple parts of the manuscript, including explanations of electrophysiological data. We have also provided additional data to support our main conclusion and addressed caveats and limitations related to the data in the Discussion section. For more details, please refer to the responses to each comment.

Comment 1: The authors have rigorously used PET neuroimaging, which is an interesting noninvasive method to track brain activity during behavioral states. However, in the case of a freely moving behavior where the scans are performed ~30 minutes after the behavioral task, it is unclear what conclusions can be drawn about task-specific brain activity. The study hinges on the neuroimaging findings that both areas of the lateral striatum (aDLS and pVLS) show increased activity during acquisition, but the DMS shows a reduction in activity during the late stages of behavior, and some of these findings are later validated with complementary experiments. However, the limitations of this technique can be further elaborated on in the discussion and the conclusions.

As described in our response to the following two comments (a, b) from the reviewer, in the PET imaging study we first analyzed task-related activity by comparing ¹⁸F-FDG uptake on different days of the auditory discrimination task with that on Day 4 of the single lever press task as a control. Next, we analyzed learning-dependent activity by comparing the uptake on different days of the discrimination task with that on Day 2 of the same task. Based on the results of both analyses, we concluded that the activity in the striatal subregions changes during the progress of discrimination learning. The behavioral significance of striatal subregions was tested by excitotoxic lesion and pharmacological blockade experiments. The explanation of imaging data analysis may have been insufficient to fully communicate dynamic changes in the activity of striatal subregions. Therefore, we have clarified our voxel-based statistical parametric analysis method to better explain the dynamic activity changes in the striatal subregions. Please refer to the following responses to comments 1 (a, b).

Comment 1 (a): In commenting on the unilateral shifts in brain striatal activity during behavior, the authors use the single lever task as a control, where many variables affecting neuronal activity might be different than in the discriminatory task. The study might be better served using Day 2 measurements as a control against which to compare activity of all other sessions since the task structures are similar.

We initially analyzed task-related activity by comparing ¹⁸F-FDG uptake on one of Days 2, 6, 10, or 24 of auditory discrimination task with that on Day 4 of the single lever press task. This task was used as a control that does not require a decision process based on the auditory stimulus. We observed significant increases in the activity of the unilateral aDLS on Day 6 and in that of the bilateral pVLS on Day 10 of the discrimination task. We also observed a significant decrease in the unilateral DMS on Day 24 (see Figures 2F and 2G). Next, as suggested, we compared the uptake on one of Days 6, 10, or 24 with that on Day 2 as a control to evaluate learning-dependent activity. The activity showed significant increases in the bilateral aDLS on Day 6 and in the unilateral pVLS on Day 10, and a significant decrease in the bilateral DMS on Day 24 (see Figures 2H). 

The reviewer has suggested a discrepancy in the activity of the unilateral or bilateral striatal subregions under certain conditions between the image data (shown in Figures 2F–H) and plot data (Figures 2J–L). This discrepancy is also suggested in the following Comment 1 (b). For example, in the image data the brain activity was increased in the unilateral (left) aDLS on Day 6 of the discrimination task as compared to Day 4 of the single lever task (Figure 2F). In the plot data, ¹⁸F-FDG uptake reached a peak on Day 6 in both the left and right sides of the aDLS (Figure 2J), and the uptake in the left aDLS on Day 6 significantly increased relative to the value of the single lever press, whereas the value in the right aDLS on Day 6 tended to increase relative to that of the single lever press with no significant difference. The plot data showing the unilaterality in the aDLS activation relative to the single lever press are consistent with the image data. On the other hand, the ¹⁸F-FDG uptake in the aDLS on Day 6 compared to the value on Day 2 was significantly increased in both sides. Similar observations were made in the activity in the pVLS on Day 10 compared to that on Day 2, as well as in the DMS activity on Day 24 relative to that of the single lever press. 

Our analysis of both task-related and learning-dependent activities revealed dynamic changes in striatal subregions during discrimination learning. We investigated the brain regions in which ¹⁸F-FDG uptake significantly increased or decreased during the learning processes, applying a statistical significance threshold (p < 0.001, uncorrected) and an extent threshold, by using a voxel-based statistical parametric analysis. In the image data, the voxels showing significant differences between two conditions are visualized on the brain template. The plot data show the amount of ¹⁸FFDG uptake in the voxels, which was detected by the voxel-based analysis. The insufficient explanation of the data analysis of PET imaging in the initial manuscript may have led to a misunderstanding regarding the activity in the unilateral or bilateral striatal subregions. Therefore, we have revised the explanation for voxel-based statistical parametric analysis, adding a more detailed description of the thresholds in the text (page 7, lines 143–145) and Methods (page 27, lines 672–675).

Comment 1 (b): From the plots in J, K, and L, it seems that shifts in activity in the different substructures are not unilateral but consistently bilateral, in contrast to what is mentioned in the text. Possibly the text reflects comparisons to the single lever task, and here again, I would emphasize comparing within the same task.

Please see our response to the first comment (a) regarding our explanation of the consistency in the activity of the unilateral or bilateral striatal subregions between the image and plot data. We have also revised the explanation in the corresponding sections of the manuscript, as described above.

Comment 2: In Figure 2, the authors present compelling data that chronic excitotoxic lesions with ibotenic acid in the aDLS, pVLS, and DMS produce differential effects on discrimination learning. However, the significant reduction in success rate of performance happens as early as Day 6 in both IBO groups in both aDLS and pVLS mice. This would seem to agree with conclusions drawn about the role of aDLS in the middle stages of learning in Figure 2, but not the pVLS, which only shows an increased activity during the late stages of the behavior.  

Figure 3 shows the behavioral effects of ibotenic acid injections into striatal subregions in rats. For the aDLS injection, we performed two-way repeated ANOVA, which revealed a significant main effect of group or day and a significant interaction of group × day, and added the simple main effects between the treatments to the figure (Figure 3G). We observed significant differences in the success rate mainly at the middle stage of learning. In contrast, for the pVLS injection there was no significant interaction for group × day, although the main effects of group or day was significant by two-way repeated ANOVA (Figure 3H). Consequently, it was unclear as to when exactly the significant reduction occurred. These results indicate that the aDLS and pVLS are necessary for the acquisition of auditory discrimination, and that the aDLS is mainly required for the middle stage. Similar results were observed in the win-shift-win strategy in the aDLS and pVLS (Figures 3J and 3K).

Next, we performed temporal inhibition of neuronal activity in striatal subregions by muscimol treatment in order to examine whether the activity in the subregions is linked with learning processes at different stages. In this experiment, muscimol was injected into the aDLS or pVLS at the middle or late stage, and the resultant effects on the success rate were investigated. The success rate in the muscimol-injected groups into the aDLS significantly decreased at the middle stage, but not at the early and late stages (Figure 4C). In contrast, the rate in the muscimol groups into the pVLS significantly decreased at the late stage, but not at the early or middle stages (Figure 4D). The results indicate that the aDLS and pVLS are mainly involved in the processes at the middle and late stages, respectively, and support the PET imaging data showing the activation of two striatal subregions at the various stages.

We have now provided the results of simple main effects analysis for the aDLS lesion (Figures 3G and 3J) and revised the description of the Results section (page 8, lines 174–178, page 8, lines 186–188, and page 9, line 205-206) and Figure legend (page 44, lines 1000‒1003, and page 44, lines 1010–1013). We have also added the results of simple main effects analysis in Figure 3J.

Comment 3: In Figure 4, the authors show interesting data with transient inactivation of subregions of the striatum with muscimol, validating their findings that the aDLS mediates the middle and the pVLS the late stages of learning, and the function of each area serves different strategies. However, the inference that aDLS inactivation suppresses the WSW strategy "moderately" is not reflected in the formal statistical value p=0.06. While there still may be a subtle effect, the authors would need to revise their conclusions appropriately to reflect the data. In addition, the authors could try a direct comparison between the success rate during muscimol inhibition in the mid-learning session between the aDLS and pVLS-treated groups in Figure 4C (middle) and 4D (middle). If this comparison is not significant, the authors should be careful to claim that inhibition of these two areas differentially affects behavior.

In Figure 4E, aDLS inhibition showed a tendency to reduce slightly win-shift-win strategy at the middle stage (t[14] = 2.038, p = 0.061, unpaired Student’s t-test). In accordance with the reviewer’s comment, we changed the word “moderate” to “subtle” (page 12, line 272).

In the temporal inhibition of the striatal subregions, the aDLS and pVLS experiments (panels C and D, respectively) were conducted separately. Since it is difficult to directly compare the data obtained from different experiments, we did not carry out a direct comparison of the success rate between the aDLS and pVLS injections. 

Comment 4: The authors have used in vivo electrophysiological techniques to systematically investigate the roles of the aDLS and the pVLS in discriminatory learning, and have done a thorough analysis of responses with each phase of behavior over the course of learning. This is a commendable and extremely informative dataset and is a strength of the study. However, the result could be better organized following the sequence of events of the behavioral task to give the reader an easier structure to follow. Ideally, this would involve an individual figure to compare the responses in both areas to Cue, Lever Press, Reward Sound, and First Lick (in this order).

We first showed changes in the proportion of event-related neurons during the acquisition phase (Figure S5). Next, we conducted a detailed analysis of the characteristics of aDLS and pVLS neuronal activity. Specifically, we found several types of event-related neurons, including: (1) reward sound-related neurons representing behavioral outcomes in the aDLS; (2) first licking-related neurons showing sustained activity after the reward in the aDLS and pVLS; and (3) cue-onset and cue-response neurons associated with the beginning and ending of a behavior in the pVLS.

Descriptions of the characteristics of event-related neurons according to the sequence of events in a trial, as the reviewer has suggested, is another way to provide an easy structure for understandings on the electrophysiological data. However, we focused on the characteristics of aDLS neurons at the middle stage and pVLS neurons at the late stage of discrimination learning. Therefore, we explained the electrophysiological data based on the order of learning stages rather than the sequence of events in the trial, as described above.

Comment 5: An important conceptual point presented in the study is that the aDLS neurons, with learning, show a reduction in firing rates and responsiveness to the first lick as well as the behavioral outcome, and don't play a role in other task-related events such as cue onset. However, the neuroimaging data in Figure 2 seems to suggest a transient enhancement of aDLS activity in the mid-stage of discriminatory learning, that is not reflected in the electrophysiology data. Is there an explanation for this difference?

In the ¹⁸F-FDG PET imaging study, the brain activity in the aDLS reached a peak at the middle stage of the acquisition phase of auditory discrimination (Figure 2J). In the multi-unit electrophysiological recording experiment, the firing activity of the aDLS neuron subpopulations related to the behavioral outcome showed no significant differences among the three stages (Figure 5E), while the proportion of these subpopulations were gradually reduced through the progress of learning stages (Figure 5F). The extent of the firing activity and length of the firing period of other subpopulations showing sustained activation after the reward appeared to show a learning-dependent decrease (Figures 6B and 6C), although the proportion of these subpopulations indicated no correlation with the progress of the learning (Figure 6D). Patterns of the temporal changes in brain activity in striatal subregions across the learning stages did not match completely the time variation in the property or proportion of specific event-related neurons. In our electrophysiological analysis, we identified well-isolated neurons from the striatal subregions during the auditory discrimination task, focusing on putative medium spiny neurons (Figures S4E–S4G). Based on the combinatorial pattern of the tone instruction cue (high tone/H or low tone /L), and lever press (right/R or left/L), we categorized the electrophysiological data into the four trials, including the HR, LL. LR, and HL. We identified HR or LL type neurons showing significant changes in the firing rate related to specific events, such as cue onset, choice response, reward sound, and first licking compared to the baseline firing rate. These neurons were further divided into two groups with increased or decreased activity relative to the baseline firing (Figures S5A and S5B). In the present study, we focused on event-related neurons with increased activity. Because of the analysis limited to neuronal subpopulations related to specific events with the increased activity, it is difficult to fully explain dynamic shifts in the brain activity of striatal subregions dependent on the progress of learning by the time variation of firing activity of individual event-related neurons. The activity of other subpopulations in the striatum may be involved in the shift in brain activity during the learning processes. In addition, recent studies have reported that the activity of glial cells influences the uptake of ¹⁸FFDG (Zimmer et al., Nat Neurosci., 2017) and that these cells regulate spike timingdependent plasticity (Valtcheva and Venance, Nat Commun, 2016). Changes in glial cellular activity, through the control of synaptic plasticity, may partly contribute to the pattern formation of learning-dependent shifts in brain activity.

To explain the difference in the time course between the brain activity and the firing activity of specific event-related neurons, we have added the aforementioned information to the Limitations section (pages 21 to 22, lines 512–539). 

Comment 6: A significant finding of the study is that CO-HR and CO-LL responses are strikingly obvious in the pVLS, but not in the aDLS, in line with the literature that the posterior (sensory) striatum processes sound. This study also shows that responses to the highfrequency tone indicating a correct right-lever choice increase with learning in contrast to the low-frequency tone responses. To further address whether this difference arises from the task contingency, and not from the frequency representation of the pVLS, an important control would be to switch the cue-response association in a separate group of mice, such that high-frequency tones require a left lever press and vice versa. This would also help tease apart task-evoked responses in the aDLS, as I am given to understand all the recording sites were in the left striatum.

We did not conduct an experiment switching cue-response association in the auditory discrimination task. However, the transient activity of cue onset-related neurons in the pVLS, as the reviewer has suggested, did not appear at the early stage of learning, but was observed in a learning-dependent manner (Figures 7A and S8E). In addition, the cue onset-HR activity showed a slight but notable difference between the HR and LL trials at the middle and late stages (Figure 7B), but there was no difference in activity in the HL and LR incorrect trials at the corresponding stages (Wilcoxon signed rank test; early, p = 0.375, middle, p = 0.931, and late, p = 0.668). These results suggest that the activity of cue onset-related neurons in the pVLS is associated with the stimulus and response association (task contingency) rather than the tone frequency.

Reviewer #1 (Recommendations For The Authors):

Minor comment 1: The readability and appeal of this study would be improved by explaining the various neuronal response types, and task-related events in slightly more detail in the results section, and minimizing the use of non-standard abbreviations wherever possible.

As suggested, we have replaced the abbreviations related to electrophysiological events (CO, CR, RS, and FL) with the original terms, and improved the explanation for neuronal response types and event-related neurons. 

Minor comment 2: It would be helpful to label DLS and VLS recordings more clearly on the figures instead of only in the figure caption.

Thank you for pointing this out. The terms “aDLS” and “pVLS” have now been added to the panels showing firing pattern of neurons: “aDLS” in Figures 5D, 6A, S6A, S7A, S8A, S8B. S8C, and S8D; and “pVLS” in Figures 6F, 7A, 7D, S6D, S6E, S7F, S8E, and S8F.

Minor comment 3: The authors suggest that aDLS HR- and LL- neurons are more sensitive to the behavioral outcome than those in pVLS (Fig 5 and S5). However, their conclusions are based on sample sizes as low as n=3 for each response type.

We identified event-related neurons from single neurons detected in both the aDLS and pVLS using the same criteria. In the pVLS, we found a small number of neurons that increased their activity during the period when the reward sound is presented (Figures S6D and S6E) (6, 4, and 17 HR type neurons at the early, middle, and late stages, respectively; 3, 5, and 15 LL type neurons at the early, middle, and late stages, respectively). The number of LL type neurons at the early stage was particularly lower, as the reviewer has suggested. However, when we plotted the firing rates of these neurons around the event, their activity did not reflect behavioral outcome. In the aDLS, we detected a large number of reward sound-related neurons representing behavioral outcome (Figures 5 and S6A) (43, 37, and 44 HR type neurons at the early, middle, and late stages, respectively; 49, 62, and 59 LL type neurons at the early, middle, and late stages, respectively). These observations suggest that aDLS neurons are more sensitive to behavioral outcomes than pVLS neurons.

Minor comment 4: Typo in Figure 4C and D, right plots, y-axis label: "subtracted".

The typographic errors in Figures 4C–4H have now been corrected to “subtracted”.

Reviewer #2 (Public Reviews):

The study by Setogawa et al. aims to understand the role that different striatal subregions belonging to parallel brain circuits have in associative learning and discrimination learning (S-O-R and S-R tasks). Strengths of the study are the use of multiple methodologies to measure and manipulate brain activity in rats, from microPET imaging to excitotoxic lesions and multielectrode recordings across anterior dorsolateral (aDLS), posterior ventral lateral (pVLS)and dorsomedial (DMS) striatum. The main conclusions are that the aDLS promotes stimulus-response association and suppresses response-outcome associations. The pVLS is engaged in the formation and maintenance of the stimulus-response association. There is a lot of work done and some interesting findings however, the manuscript can be improved by clarifying the presentation and reasoning. The inclusion of important controls will enhance the rigor of the data interpretation and conclusions.

We appreciate the reviewer’s valuable feedback, which has been beneficial in our endeavor to improve our manuscript. In response to the comments, we have revised the description of the experimental methods and underlying rationale, as well as the Results section. We have also provided additional data for some of the experiments that support the conclusions. For more details, please refer to the responses to each comment, included below.

Reviewer #2 (Recommendations For The Authors):

Comment 1: Generally, the manuscript is hard to read because of the cumbersome sentence structure, overuse of poorly defined acronyms, and lack of clarity on the methods used.

According to the following comments (a)–(d), we have revised the corresponding text in the manuscript to clarify the sentence structure, definitions of terms, and methodology. 

Comment 1 (a): For example, the single lever task used as a control for the auditory discrimination task could be introduced better, explaining the reasoning and the strategy for subtracting it from the images obtained during the discrimination phase at the start of the section.

We analyzed task-related activity by comparing ¹⁸F-FDG uptake on Days 2, 6, 10, or 24 of auditory discrimination task with that on Day 4 of the single lever press task. This task was used as a control that does not require a decision process based on the auditory stimulus. For clarification, we have provided a more detailed explanation of the flow of the single lever press task used in the PET experiment, including the rationale for employing this task as a control (page 6, lines 129–135). We have also revised the explanation of voxel-based statistical parametric analysis, adding a more detailed description of the thresholds (page 7, lines 143–145).

Comment 1 (b): Another example is that important methodological information is buried deep in the text and complicates the interpretation of the results.

We have revised the following sentences in the manuscript in order to provide clearer methodological information.

(1) As described above, explanations for the single lever task (page 6, lines 129–135) and voxel-based statistical parametric analysis were added (page 7, lines 143–145). 

(2) Definition of the early, middle, and late stages were described in the initial behavioral experiment (page 6, lines 113–119). 

(3) Abbreviations related to behavioral strategies (WSW and LSL) and electrophysiological events (CO, CR, RS, and FL) were replaced with the original terms. 

Comment 1 (c): The specie being studied is not stated in the abstract, nor the introduction, and only in the middle of the result section. Please include the specie in the abstract and the first part of the result also for clarity.

We included the name of the species (rats) in the Abstract (page 3, line 47), at the end of the Introduction (page 5, lines 87–88) and at the beginning of the Results (page 5, line 109).

Comment 1 (d): The last part of the intro is copied/pasted from the abstract. Please revise.

The last part of the Introduction was revised accordingly (page 5, lines 97–104).

Comment 2: The glucose microPET imaging is carried out 30 mins after the rats performed the task and it is expected to capture activation during the task. Is this correct? This assumption has to be validated with an experiment, which is a control showing a validation of the microPET approach used, and this way can report activation of brain areas during the task completed 20-30 minutes before. For example, V1 or A1 would be a control that we would expect to be activated during the task.

Our PET experiment was conducted in accordance with previously established methods (Cui et al, Neuroimage, 2015), where rats received intravenous administration of ¹⁸FFDG solution just before the start of the behavioral session, which lasted for 30 min. The ¹⁸F-FDG uptake in the brain starts immediately and reaches the maximum level until 30 min after the administration, and the level is kept at least for 1 h (Mizuma et al., J Nucl Med, 2010). The rats were returned to their home cages, and a 30-min PET scan started 25 min after the session. The start time of the scan was chosen to allow for sufficient reduction of 18F radioactivity in arterial blood to increase the S/N ratio of the radioactivity (Mizuma et al., J Nucl Med, 2010). As shown in Table S1, we confirmed that the brain activity in the medial geniculate body (auditory thalamus) was increased on Days 6 and 10 in the acquisition phase, although the activity in the auditory cortex was not changed, which is consistent with the results of a previous study reporting that the auditory cortex does not show the causality for the pure-tone discrimination task (Gimenez et al., J Neurophysiol., 2015).

Comment 3: Why are Days 2, 6, 10, and 13 chosen and compared for the behavior? Why aren't these the same days chosen in the other part of the study? It is unclear why authors focused on these days and why the focus changed later.

We conducted daily training of the discrimination task. The success rate reached a plateau on Day 13 and was maintained until Day 24 (Figure 1B). Based on these results, we categorized the learning processes into the acquisition and learned phases, and then divided the acquisition phase into the early (< 60%), middle (60–80%), and late (> 80%) stages. In the PET experiment, we selected Days 2, 6, and 10 as the representatives of each stage during the acquisition phase. In addition, we also selected Day 24 for the learned phase.  However, no scan was performed on Day 13 due to the transition between the two phases.   

Comment 4: (A) Is the learning and acquisition of the single lever press and discrimination task completed by day 4? Or are rats still learning? The authors claimed no changes in DMS activity between single lever press & discrimination, and therefore DMS isn't involved in learning. But to make this claim we should have measures that the learning has already happened, which I am not sure have been provided. (B) On this same point, the DMS activity is elevated on Day 4 of a single lever press compared to the aDLS and pVLS. So is it possible that the activity in DMS was already elevated on Day 4 of single lever press training? Especially given that DMS is supposedly involved in goal-directed behavior?

(A) In the single lever press task, the number of lever presses plateaued on Day 2 (Figure 1C). In addition, we analyzed response time and its variability, which plateaued from Day 3 and Day 2, respectively (see Author response image 1). These results indicate that the learning in the task was completed by Day 4. In the auditory discrimination task, Day 4 corresponded to the transition period from the early-tomiddle stages of the acquisition phase, suggesting that learning was still progressing. 

In the imaging analysis, we examined task-related activity by comparing ¹⁸F-FDG uptake on either day of the discrimination task with that on Day 4 of the single lever press task, and did not find any changes in the brain activity in the DMS. In addition, we investigated learning-related activity, and the DMS activity did not change during acquisition phase. These results suggest that the DMS is not involved in the acquisition phase of learning. Furthermore, comparisons between Days 10 and Day 24 showed a decrease in DMS activity during the learned phase, suggesting that DMS activity was downregulated during the learned phase. In addition, chronic lesion in the DMS indicated that the success rate in the discrimination task was comparable between the control and lesioned groups (Figure 3I), whereas the response time lengthened throughout the learning in the lesioned group compared to the controls (Figure S1C). These results support our notion that the DMS contributes to the execution, but not learning, of discriminative behavior (Figure 3I and S1C).

Author response image 1.

Performance of single lever press task conducted before auditory discrimination task. (A) Number of lever presses. (B) Response time (Kruskal-Wallis test, χ² = 38.063, p = 2.7 × 10^-8, post hoc Tukey–Kramer test, p = 0.047 for Day 1 vs. Day 2; p = 2.3 × 10^-7 for Day 1 vs. Day 3; and p = 4.0 × 10^-6 for Day 1 vs. Day 4; p = 0.019 for Day 2 vs. Day 3; p = 0.082 for Day 2 vs. Day 4; p = 0.951 for Day 3 vs. Day 4). (C) Response time variability (Kruskal-Wallis test, χ² = 28.929, p = 2.3 × ^-6, post hoc Tukey–Kramer test, p = 0.077 for Day 1 vs. Day 2; p = 5.7 × 10^-6 for Day 1 vs. Day 3; and p = 1.3 × 10^-4 for Day 1 vs. Day 4; p = 0.060 for Day 2 vs. Day 3; p = 0.253 for Day 2 vs. Day 4; p = 0.912 for Day 3 vs. Day 4). Data obtained from the task shown in Figure 2C are plotted as the median and quartiles with the maximal and minimal values. *p < 0.05, **p < 0.01, and ***p < 0.001.

(B) We compared ¹⁸F-FDG uptakes among striatal subregions on Day 4 of the single lever press task (334.8 ± 2.86, 299.0 ± 1.71, and 336.8 ± 2.18 for the aDLS, pVLS, and DMS, respectively; one-way ANOVA, F[2,41] = 104.767, p = 2.1 × 10^-16). The uptake was comparable between the aDLS and DMS (post hoc Tukey-Kramer test, p = 0.058), but it was significantly lower in the pVLS compared to either of the other two subregions (post hoc Tukey-Kramer test, aDLS vs. pVLS, p = 5.1 × 10^-9, post hoc Tukey-Kramer test, pVLS vs. DMS, p = 5.1 × 10^-9). However, since we did not measure the brain activity in the single lever task outside of Day 4, it is unclear whether there was an increase in DMS activity during the acquisition of the task. Similarly, since we did not confirm the behavioral modes, which include goal-directed and habitual actions, it is difficult to conclude that the lever presses in the task were controlled by the goaldirected mode. However, our chronic lesion experiment suggests that the DMS is involved in the execution of discrimination behavior (Figure S1C). A clearer understanding of the DMS function in discrimination learning is an important challenge in the future.

Comment 5: It seems like the procedure of microPET imaging affects performance on the task. The anesthesia used maybe? Figures 2C and D show evidence that the behavior was negatively affected on the days on which microPET imaging was performed after the training. Can the author clarify/comment?

Isoflurane anesthesia may slightly reduce behavioral performance. We carried out anesthesia (median [interquartile range]: 6 [5–8] min) during the insertion of the catheter for FDG injection, and set a recovery period of at least 2 h until the beginning of the behavioral session, to minimize the impact of anesthesia. The performances in Figure 2E were similar to those in the intact rats (compared to Figures 1C–1F), suggesting that the procedure for PET scans does not affect the acquisition of discrimination. 

We have added detailed information on the isoflurane anesthesia to the Methods section (page 26, lines 649–653).

Comment 6: More on clarity. Section 3 of the results (muscimol inactivation) refers a lot to "the behavioral strategies" without really clarifying what these are - are they referring to WSW / LSL (which also could use a better introduction) or goal-directed/habitual or stimulus-response/stimulus-outcome?

The dorsal striatum is involved in both behavioral strategies based on stimulus-response association and the response-outcome association during instrumental learning. To assess the impact of striatal lesions on the behavioral strategies, we analyzed the proportion of response attributed to two strategies in all responses of each session. One is the “win-shift-win” strategy, which is considered to reflect the behavioral strategy based on the stimulus-response association. In this strategy, after a correct response in the previous trial, the rats press the opposite lever in the current trial in response to a shift of the instruction cue, resulting in the correct response.  Another strategy is the “lose-shift-lose” strategy, which is considered to appear as a consequence of the behavioral strategy based on the response-outcome association. In this strategy, after an error response in the previous trial, the rats press the opposite lever in the current trial despite a shift of the instruction cue, leading to another error response.

We have revised the explanations of the behavioral strategies in the section of the Results section (page 9, lines 192–201). 

Comment 7: Related to WSW / LSL needing a better introduction, on lines 192/193 authors describe a result where they saw the WSW and LSL strategies increase and decrease, respectively, in saline-injected mice. Is the change in performance expected or an undesired effect of the saline injection? This is not clear now and it should be clarified.

The explanations of the win-shift-win and lose-shift-lose strategies have been revised in the Results section on excitotoxic lesion experiment (page 9, lines 192–201) as described in our response to Comment 6. Win-shift-win is an indicator of correct responses, while lose-shift-lose indicates errors. Therefore, win-shift-win is predicted to increase, and lose-shift-lose decrease, as discrimination learning progresses. Indeed, in the results of the behavioral experiments, shown in Figure 1, both indicators change in a similar pattern to those in the results of the lesion experiments (Figure 3).

We have added the explanation of the proportions of both strategies in intact rats (page 9, lines 203–204) with a supplementary figure (Figure S2) and accompanying legend (page 56, lines 1173–1177).

Comment 8: Muscimol experiments - two questions/comments. How often do rats receive muscimol?

In this section, muscimol is given on day 2 and on days after the animals hit a 60% or 80% success rate. Can the authors provide a mean and SEM for when are those injections?

The first injection was conducted on Day 2 to target the early stage. The second and third injections were conducted on the days after the success rate had reached 60% and 80% for the first time through the training, respectively, to target the middle and late stage. respectively. These conditions are described in the Results (page 10, lines 234– 237) and Methods (page 26, lines 633–636). The mean and s.e.m. of the injection day at the middle and late stages were not significantly different between the saline and muscimol-injected groups into the aDLS (see Author response image 2A) and pVLS (see Author response image 2B).

Author response image 2.

Injection days during auditory discrimination learning. Injections with saline (SAL) and muscimol (MUS) into the aDLS (A) or pVLS (B) were performed after the success rate had reached 60% (middle stage) and 80% (late stage) for the first time through the training, respectively (A, Wilcoxon signed rank test, middle, Z = 65, p = 0.772, late, Z = 56.5, p = 0.242 for the aDLS; B, Wilcoxon signed rank test, middle, Z = 39, p = 1.000, late, Z = 43, p = 0.587). Data are indicated as the median and quartiles with the maximal and minimal values. 

Comment 9: Muscimol experiments. Can the authors comment on the effects on performance vs learning? What happens on the days after Muscimol? Does performance bounce back or is it still impaired?

We conducted a transient inhibition experiment with muscimol to examine whether the neuronal activity in the striatal subregions is linked with the processes at different stages. In this experiment, to lower the possibility that compensation of learning may occur during a session after the muscimol injection (Day N), we limited the session time to 15 min (45 trials) and evaluated the impact of the injection on the success rate at specific stages. The success rate in the muscimol-injected groups into the aDLS significantly decreased at the middle stage compared to the corresponding salineinjected groups, but not at the early and late stages (Figure 4C), and the rate in the muscimol groups into the pVLS significantly decreased at the late stage compared with the respective saline groups, but not at the early and middle stages (Figure 4D). Our results demonstrated that the aDLS and pVLS mainly function at the middle and late stages of the auditory discrimination task, respectively. 

In addition, we here reply to comment 10 as for the comparison of success rates before (Day N-1) and after (Day N+1) the injections (see Author response image 3). We focused on two injections into the aDLS at the middle stage and into the pVLS at the late stage, in which the rate was reduced soon after the muscimol injection on Day N. The success rate for the two injections showed no significant main effect regarding group (saline/muscimol) or day (Days N-1/N+1) and no significant interactions for group × day. Moreover, the success rate was not significantly increased on Day N+1 as compared to Day N-1, even in the saline-injected control group, probably because of the limited session time soon after the injection. Therefore, we consider that it was difficult to define the effects of drug injection on the learning of auditory discrimination in our behavioral protocol for the transient inhibition experiment, and that the reduced rates observed in the muscimol-injected group on Day N mostly reflect the impacts of muscimol at least partly on the performance of discriminative behavior. 

Author response image 3.

Comparison of success rate between days before (Day N1) and after (Day N+1) the injections into striatal subregions. Success rate in the saline (SAL)- and muscimol (MUS)-injected groups into the aDLS (A) or pVLS (B) at the early, middle, and late stages of auditory discrimination learning (two-way repeated ANOVA; early, day, F[1,14] = 5.266, p = 0.038, group, F[1,14] = 0.276, p = 0.608, day × group, F[1,14] = 0.118, p = 0.736; middle, day, F[1,14] = 4.110, p = 0.062, group, F[1,14] = 0.056, p = 0.816, day × group, F[1,14] = 1.150, p = 0.302; late, day, F[1,14] = 6.408, p = 0.024, group, F[1,14] = 0.229, p = 0.640, day × group, F[1,14] = 1.277, p = 0.278 for the aDLS; and early, day, F[1,10] = 0.115, p = 0.746, group, F[1,10] = 2.414, p = 0.151, day × group, F[1,10] = 0.157, p = 0.700; middle, day, F[1,10] = 0.278, p = 0.610, group, F[1,10] = 0.511, p = 0.491, day × group, F[1,10] = 4.144, p = 0.069; late, day, F[1,10] = 0.151, p = 0.705, group, F[1,10] = 0.719, p = 0.416, day × group, F[1,10] = 0.717, p = 0.417 for the pVLS). Data are indicated as the mean ± s.e.m.

Comment 10: Muscimol data has a pair before and after, can the authors show this comparison at early, middle, and late training? Not just the subtraction.

The comparison of success rates before and after drug injection is shown in Author response image 3.

Comment 11: Ephys recordings. These are complex figures and include a large number of acronyms. It would help to define them again and help the reader through these figures so the reader can focus on understanding the finding more than the figure presentation.

We replaced the abbreviations related to electrophysiological events (CO, CR, RS, and FL) with the original terms, and improved the explanation in the text and figures. 

Comment 12: Figure 7B/E - on correct trials, they see a difference in the cue response to high tone / low tone but no difference in the choice. This is the one that seemed like a topography issue.

The transient activity of cue onset-related neurons in the pVLS did not appear at the early stage of learning, but was observed in a learning-dependent manner (Figures 7A and S8E). In addition, the cue onset-HR activity showed a slight but notable difference between the HR and LL trials at the middle and late stages (Figure 7B), whereas there was no difference between activities in the HL and LR incorrect trials at the corresponding stages (Wilcoxon signed rank test; early, p = 0.375, middle, p = 0.931, and late, p = 0.668). These results suggest that the cue onset-related neurons in the pVLS represents the stimulus and response association (task contingency) rather than the topography of tone frequency.

Comment 13: Animals were normally trained for 60 minutes but on muscimol days only trained for 15 mins. On PET days only trained for 30 minutes. Ephys sessions were 60 mins. Is this correct? Why?

We determined the session time for each experiment by considering both technical and behavioral aspects. In the initial behavioral experiment, the session time was set to 60 min per day. Under this condition, the rats acquired the discrimination learning within 13 days. In the imaging experiment, the session without a PET scan was conducted for 60 min, while the session with a PET scan was carried out for 30 min as described previously (Cui et al, Neuroimage, 2015). This time schedule produced a learning curve similar to that of the initial behavioral experiment. In the transient inhibition experiment, the sessions without drug injections lasted for 60 min. As described in our response to the comment 2, the time of the session soon after the injection was limited to 15 min to lower the possibility of compensation of learning during the session. In the chronic lesion and electrophysiological experiments, all sessions were conducted for 60 min, corresponding to the initial experiment. 

References

Mizuma, H., Shukuri, M., Hayashi, T., Watanabe, Y. & Onoe, H. Establishment of in vivo brain imaging method in conscious mice. Journal of Nuclear Medicine 51, 10681075 (2010).

Cui, Y., et al. A voxel-based analysis of brain activity in high-order trigeminal pathway in the rat induced by cortical spreading depression. Neuroimage 108, 17-22 (2015).

Zimmer, E.R., et al. [18 F] FDG PET signal is driven by astroglial glutamate transport. Nat Neurosci 20, 393-395 (2017).

Valtcheva, S. & Venance, L. Astrocytes gate Hebbian synaptic plasticity in the striatum. Nature communications 7, 13845 (2016).

Gimenez T.L., Lorenc M., Jaramillo S. Adaptive categorization of sound frequency does not require the auditory cortex in rats. J Neurophysiol 114:1137-1145 (2015).

Author response:

The following is the authors’ response to the current reviews.

Reviewer #1 (Public Review):

Summary:

This manuscript explores the impact of serotonin on olfactory coding in the antennal lobe of locusts and odor-evoked behavior. The authors use serotonin injections paired with an odorevoked palp-opening response assay and bath application of serotonin with intracellular recordings of odor-evoked responses from projection neurons (PNs).

Strengths:

The authors make several interesting observations, including that serotonin enhances behavioral responses to appetitive odors in starved and fed animals, induces spontaneous bursting in PNs, directly impacts PN excitability, and uniformly enhances PN responses to odors.

Weaknesses:

The one remaining issue to be resolved is the theoretical discrepancy between the physiology and the behavior. The authors provide a computational model that could explain this discrepancy and provide the caveat that while the physiological data was collected from the antennal lobe, but there could be other olfactory processing stages involved. Indeed other processing stages could be the sites for the computational functions proposed by the model. There is an additional caveat which is that the physiological data were collected 5-10 minutes after serotonin application whereas the behavioral data were collected 3 hours after serotonin application. It is difficult to link physiological processes induced 5 minutes into serotonin application to behavioral consequences 3 hours subsequent to serotonin application. The discrepancy between physiology and behavior could easily reflect the timing of action of serotonin (i.e. differences between immediate and longer-term impact).

For our behavioral experiments, we waited 3 hours after serotonin injection to allow serotonin to penetrate through the layers of air sacks and the sheath, and for the locusts to calm down and recover their baseline POR activity levels. For the physiology experiments, we noticed that the quality of the patch decreased over time after serotonin introduction. Hence, it was difficult to hold cells for that long. However, the point raised by the reviewer is well-taken. We have performed additional experiments to show that the changes in POR levels to different odorants are rapid and can be observed within 15 minutes of injecting serotonin (Author response image 2) and that the physiological changes in PNs (bursting spontaneous activity, maintenance of temporal firing patterns, and increase odor-evoked responses) persists when the cells are held for longer duration (i.e. 3 hours akin to our behavioral experiments). It is worth noting that 3-hour in-vivo intracellular recordings are not easily achievable and come with many experimental constraints. So far, we have managed to record from two PNs that were held for this long and add them to this rebuttal to support our conclusions. (Author response image 1).

Author response image 1.

Spontaneous and odor-evoked responses in individual PNs remain consistent for three hours after serotonin introduction into the recording chamber/bath. (A) Representative intracellular recording showing membrane potential fluctuations in a projection neuron (PN) in the antennal lobe. Spontaneous and odor-evoked responses to four odorants (pink color bars, 4 s duration) are shown before (control) and after serotonin application (5HT). Voltage traces 30 minutes (30min), 1 hour (1h), 2 hours (2h), and 3 hours (3h) after 5HT application are shown to illustrate the persisting effect of serotonin during spontaneous and odor-evoked activity periods. (B) Rasterized spiking activities in two recorded PNs are shown. Spontaneous and odor-evoked responses are shown in all 5 consecutive trials. Note that the odor-evoked response patterns are maintained, but the spontaneous activity patterns are altered after serotonin introduction.

Author response image 2.

Palp-opening response (POR) patterns to different odorants remain consistent following serotonin introduction. The probability of PORs is shown as a bar plot for four different odorants; hexanol (green), benzaldehyde (blue), linalool (red), and ammonium (purple). PORs before serotonin injection (solid bars) are compared against response levels after serotonin injection (striped bars). As can be noted, PORs to the four odorants remain consistent when tested 15 minutes and 3 hours after (5HT) serotonin injection.

Overall, the study demonstrates the impact of serotonin on odor-evoked responses of PNs and odor-guided behavior in locusts. Serotonin appears to have non-linear effects including changing the firing patterns of PNs from monotonic to bursting and altering behavioral responses in an odor-specific manner, rather than uniformly across all stimuli presented.

We thank the reviewer for again providing very useful feedback for improving our manuscript.

Reviewer #2 (Public Review):

Summary:

The authors investigate the influence of serotonin on feeding behavior and electrophysiological responses in the antennal lobe of locusts. They find that serotonin injection changes behavior in an odor-specific way. In physiology experiments, they can show that projection neurons in the antennal lobe generally increase their baseline firing and odor responses upon serotonin injection. Using a modeling approach the authors propose a framework on how a general increase in antennal lobe output can lead to odor-specific changes in behavior.

Strengths:

This study shows that serotonin affects feeding behavior and odor processing in the antennal lobe of locusts, as serotonin injection increases activity levels of projection neurons. This study provides another piece of evidence that serotonin is a general neuromodulator within the early olfactory processing system across insects and even phyla.

Weaknesses:

I still have several concerns regarding the generalizability of the model and interpretation of results. The authors cannot provide evidence that serotonin modulation of projection neurons impacts behavior.

This is true and likely to be true for any study linking neural responses to behavior. There are multiple circuits and pathways that would get impacted by a neuromodulator like serotonin. What we showed with our physiology is how spontaneous and odor-evoked responses in the very first neural network that receives olfactory sensory neuron input are altered by serotonin. Given the specificity of the changes in behavioral outcomes (i.e. odor-specific increase and decrease in an appetitive behavior) and non-specificity in the changes at the level of individual PNs (general increase in odor-evoked spiking activity), we presented a relatively simple computational model to address the apparent mismatch between neural and behavioral responses. (Author response image 4).

The authors show that odor identity is maintained after 5-HT injection, however, the authors do not show if PN responses to different odors were differently affected after serotonin exposure.

The PN responses to different odorants changed in a qualitatively similar fashion. (Author response image 3)

Author response image 3.

PN activity before and after 5HT application are compared for different cellodor combinations. As can be noted, the changes are qualitatively similar in all cases. After 5HT application, the baseline activity became more bursty, but the odor-evoked response patterns were robustly maintained for all odorants.

Regarding the model, the authors show that the model works for odors with non-overlapping PN activation. However, only one appetitive, one neutral, and one aversive odor has been tested and modeled here. Can the fixed-weight model also hold for other appetitive and aversive odors that might share more overlap between active PNs? How could the model generate BZA attraction in 5-HT exposed animals (as seen in behavior data in Figure 1) if the same PNs just get activated more?

Author response image 4.

Testing the generality of the proposed computational model. To test the generality of the model proposed we used a published dataset [Chandak and Raman, 2023]: Neural dataset – 89 PN responses to a panel of twenty-two odorants; Behavioral dataset – probability of POR responses to the same twenty-two odorants. We built the model using just the three odorants overlapping between the two datasets: hexanol, benzaldehyde and linalool. The true probability of POR values of the twenty odorants and the POR probability predicted by the model are shown for all twenty-two odorants as a scatter plot. As can be noted, there is a high correlation (0.79) between the true and the predicted values.

The authors should still not exclude the possibility that serotonin injections could affect behavior via modulation of other cell types than projection neurons. This should still be discussed, serotonin might rather shut down baseline activation of local inhibitory neurons - and thus lead to the interesting bursting phenotypes, which can also be seen in the baseline response, due to local PN-to-LN feedback.

As we agreed, there could be other cells that are impacted by serotonin release. Our goal in this study was to characterize how spontaneous and odor-evoked responses in the very first neural network that receives olfactory sensory neuron input are altered by serotonin. Within this circuit, there are local inhibitory neurons (LNs), as correctly indicated by this reviewer. Surprisingly, our preliminary data indicates that LNs are not shut down but also have an enhanced odor-evoked neural response. (Author response image 5.) Further data would be needed to verify this observation and determine the mechanism that mediate the changes in PN excitability. Irrespective, since PN activity should incorporate the effects of changes in the local neuron responses and is the sole output from the antennal lobe that drives all downstream odor-evoked activity, we focused on them in this study.

Author response image 5.

Representative traces showing intracellular recording from a local neuron in the antennal lobe. Five consecutive trials are shown. Note that LNs in the locust antennal lobe are non-spiking. The LN activity before, during, and after the presentation of benzaldehyde and hexanol (colored bar; 4s) are shown. The Left and Right panels show LN activity before and after the application of 5HT. As can be noted, 5HT did not shut down odor-evoked activity in this local neuron.

The authors did not fully tone down their claims regarding causality between serotonin and starved state behavioral responses. There is no proof that serotonin injection mimics starved behavioral responses.

Specific minor issues:<br /> It is still unclear how naturalistic the chosen odor concentrations are. This is especially important as behavioral responses to different concentrations of odors are differently modulated after serotonin injection (Figure 2: Linalool and Ammonium). The new method part does not indicate the concentrations of odors used for electrophysiology.

All odorants were diluted to 0.01-10% concentration by volume in either mineral oil or distilled water. This information is included in the Methods section. For most odorants used in the study, the lower concentrations only evoked a very weak neural response, and the higher concentrations evoked more robust responses. The POR responses for these odorants at various concentrations chosen are included in Figure 2. Note, that the responses to linalool and ammonium remained weak throughout the concentration changes, compared to hexanol and benzaldehyde.

Did all tested PNs respond to all odorants?

No, only a subset of them responses to each odorant. These responses have been well characterized in earlier publications [included refs].

The authors do not show if PN responses to different odors were differently affected after serotonin exposure. They describe that ON responses were robust, but OFF responses were less consistent after 5-HT injection. Was this true across all odors tested? Example traces are shown, but the odor is not indicated in Figure 4A. Figure 4D shows that many odor-PN combinations did not change their peak spiking activity - was this true across odorants? In Figure 5 - are PNs ordered by odor-type exposure?

Also, Figure 6A only shows example trajectories for odorants - how does the average look? Regarding the data used for the model - can the new dataset from the 82 odor-PN pairs reproduce the activation pattern of the previously collected dataset of 89 pairs?

What is shown in Figure 6A is the trial-averaged response trajectory combining activities of all 82 odor-PN pairs. 82 odor-PN pair was collected intracellularly examining the responses to four odorants before and after 5HT application. The second dataset involving 89 PN responses to 22 odorants was collected extracellularly. They have qualitative similarities in each odorant activate a unique subset of those neurons.

The authors toned down their claims that serotonin injection can mimic the starved state behavioral response. However, some sentences still indicate this finding and should also be toned down:

last sentence of introduction - "In sum, our results provide a more systems-level view of how a specific neuromodulator (serotonin) alters neural circuits to produce flexible behavioral outcomes."

We believe we showed this with our computational model, how uniform changes in the neural responses could lead to variable and odor-specific changes in behavioral PORs.

discussion: "Finally, fed locusts injected with serotonin generated similar appetitive responses to food-related odorants as starved locusts indicating the role of serotonin in hunger statedependent modulation of odor-evoked responses." This claim is not supported.

Figure 7 shows that the fed locusts had lower POR to hex and bza. The POR responses significantly increased after the 5HT application. However, we have rephrased this sentence to limit our claims to this result. "Finally, fed locusts injected with serotonin generated similar appetitive palp-opening responses to food-related odorants as observed in starved locusts”

last results: "However, consistent with results from the hungry locusts, the introduction of serotonin increased the appetitive POR responses to HEX and BZA. Intriguingly, the appetitive responses of fed locusts treated with 5HT were comparable or slightly higher than the responses of hungry locusts to the same set of odorants."

Again this sentence simply describes the result shown in Figure 7.

In Figure 7 - BZA response seems unchanged in hungry and fed animals and only 5-HT injection enhances the response. There is only one example where 5-HT application and starvation induce the same change in behavior - N=1 is not enough to conclude that serotonin influences food-driven behaviors.

The reviewer is ignoring the lack of changes to PORs to linalool and ammonium. Taken together, serotonin increased PORs to only two of the four odorants in starved locusts. The responses after 5HT modulation to these four odorants were similar in fed locusts treated with 5HT and starved locusts.

Also, this seems to be wrongly interpreted in Figure 7: "It is worth noting that responses to LOOL and AMN, non-food related odorants with weaker PORs, remained unchanged in fed locusts treated with 5HT." The authors indicate a significant reduction in POR after 5-HT injection on LOOL response in Figure 7.

Revised.<br /> It is worth noting that responses to LOOL and AMN, non-food related odorants with weaker PORs, and reduced in fed locusts treated with 5HT."

Also, the newly added sentence at the end of the discussion does not make sense: "However, since 5HT increased behavioral responses in both fed and hungry locusts, the precise role of 5HT modulation and whether it underlies hunger-state dependent modulation of appetitive behavior still remains to be determined."<br /> The authors did not test 5-HT injection in starved animals

The results shown in Figure 1 compare the POR responses of starved locusts before and after 5HT introduction.

We again thank the reviewer for useful feedback to further improve our manuscript.

The following is the authors’ response to the original reviews.

Reviewer #1 (Public Review):

Summary:

This manuscript explores the impact of serotonin on olfactory coding in the antennal lobe of locusts and odor-evoked behavior. The authors use serotonin injections paired with an odor-evoked palp-opening response assay and bath application of serotonin with intracellular recordings of odor-evoked responses from projection neurons (PNs).

Strengths:

The authors make several interesting observations, including that serotonin enhances behavioral responses to appetitive odors in starved and fed animals, induces spontaneous bursting in PNs, and uniformly enhances PN responses to odors. Overall, I had no technical concerns. Weaknesses:

While there are several interesting observations, the conclusions that serotonin enhanced sensitivity specifically and that serotonin had feeding-state-specific effects, were not supported by the evidence provided. Furthermore, there were other instances in which much more clarification was needed for me to follow the assumptions being made and inadequate statistical testing was reported.

Major concerns.

• To enhance olfactory sensitivity, the expected results would be that serotonin causes locusts to perceive each odor as being at a relatively higher concentration. The authors recapitulate a classic olfactory behavioral phenomenon where higher odor concentrations evoke weaker responses which is indicative of the odors becoming aversive. If serotonin enhanced the sensitivity to odors, then the dose-response curve should have shifted to the left, resulting in a more pronounced aversion to high odor concentrations. However, the authors show an increase in response magnitude across all odor concentrations. I don't think the authors can claim that serotonin enhances the behavioral sensitivity to odors because the locusts no longer show concentration-dependent aversion. Instead, I think the authors can claim that serotonin induces increased olfactory arousal.

The reviewer makes a valid point. Bath application of serotonin increased POR behavioral responses across all odor concentrations, and concentration-dependent aversion was also not observed. Furthermore, the monotonic relationship between projection neuron responses and the intensity of current injection is altered when serotonin is exogenously introduced (see Author response image 1; see below for more explanation). Hence, our data suggests that serotonin alters the dose-response relationship between neural/behavioral responses and odor intensity. As recommended, we have followed what the reviewer has suggested and revised our claim to serotonin inducing increase in olfactory arousal. The new physiology data has been added as Supplementary Figure 3 to the revised manuscript.

• The authors report that 5-HT causes PNs to change from tonic to bursting and conclude that this stems from a change in excitability. However, excitability tests (such as I/V plots) were not included, so it's difficult to disambiguate excitability changes from changes in synaptic input from other network components.

To confirm that the PN excitability did indeed change after serotonin application, we performed a new set of current-clamp recordings. In these experiments, we monitored the spiking activities in individual PNs as we injected different levels of current injections (200 – 1000 pico Amperes). Note that locust LNs that provide recurrent inhibition arborize and integrate inputs from a large number of sensory neurons and projection neurons. Therefore, activating a single PN should not activate the local neurons and therefore the antennal lobe network.

We found that the total spiking activity monotonically increased with the magnitude of the current injection in all four PNs recorded (Author response image 1). However, after serotonin injection, we found that the spiking activity remained relatively stable and did not systematically vary with the magnitude of the current injection. While the changes in odor-evoked responses may incorporate both excitability changes in individual PNs and recurrent feedback inhibition through GABAergic LNs, these results from our current injection experiments unambiguously indicate that there are changes in excitability at the level of individual PNs. We have added this result to the revised manuscript.

Author response image 1.

Current-injection induced spiking activity in individual PNs is altered after serotonin application. (A) Representative intracellular recordings showing membrane potential fluctuations as a function of time for one projection neuron (PNs) in the locust antennal lobe. A two-second window when a positive 200-1000pA current was applied is shown. Firing patterns before (left) and after (right) serotonin application are shown for comparison. Note, the spiking activity changes after the 5HT application. The black bar represents the 20mV scale. (B) Dose-response curves showing the average number of action potentials (across 5 trials) during the 2second current pulse before (green) and after (purple) serotonin for each recorded PN. Note that the current intensity was systematically increased from 200 pA to 1000 pA. The (C) The mean number of spikes across the four recorded cells during current injection is shown. The color progression represents the intensity of applied current ranging 200pA (leftmost bar) to 1000pA (rightmost bar). The dose-response trends before (green) and after (purple) 5HT application are shown for comparison. The error bars represent SEM across the four cells.

• There is another explanation for the theoretical discrepancy between physiology and behavior, which is that odor coding is further processing in higher brain regions (ie. Other than the antennal lobe) not studied in the physiological component of this study. This should at least be discussed.

This is a valid argument. For our model of neural mapping onto behavior to work, we only need the odorant that evokes or suppresses PORs to activate a distinct set of neurons. Having said that, our extracellular recording results (Fig. 6E) indicate that hexanol (high POR) and linalool (low POR) do activate highly non-overlapping sets of PNs in the antennal lobe. Hence, our results suggest that the segregation of neural activity based on behavioral relevance already begins in the antennal lobe. We have added this clarification to the discussion section.

• The authors cannot claim that serotonin underlies a hunger state-dependent modulation, only that serotonin impacts responses to appetitive odors. Serotonin enhanced PORs for starved and fed locusts, so the conclusion would be that serotonin enhances responses regardless of the hunger state. If the authors had antagonized 5-HT receptors and shown that feeding no longer impacts POR, then they could make the claim that serotonin underlies this effect. As it stands, these appear to be two independent phenomena.

This is also a valid point. We have clarified this in the revised manuscript.

Reviewer #2 (Public Review):

Summary:

The authors investigate the influence of serotonin on feeding behavior and electrophysiological responses in the antennal lobe of locusts. They find that serotonin injection changes behavior in an odorspecific way. In physiology experiments, they can show that antennal lobe neurons generally increase their baseline firing and odor responses upon serotonin injection. Using a modeling approach the authors propose a framework on how a general increase in antennal lobe output can lead to odorspecific changes in behavior. The authors finally suggest that serotonin injection can mimic a change in a hunger state.

Strengths:

This study shows that serotonin affects feeding behavior and odor processing in the antennal lobe of locusts, as serotonin injection increases activity levels of antennal lobe neurons. This study provides another piece of evidence that serotonin is a general neuromodulator within the early olfactory processing system across insects and even phyla. Weaknesses:

I have several concerns regarding missing control experiments, unclear data analysis, and interpretation of results.

A detailed description of the behavioral experiments is lacking. Did the authors also provide a mineral oil control and did they analyze the baseline POR response? Is there an increase in baseline response after serotonin exposure already at the behavioral output level? It is generally unclear how naturalistic the chosen odor concentrations are. This is especially important as behavioral responses to different concentrations of odors are differently modulated after serotonin injection (Figure 2: Linalool and Ammonium).

POR protocol: Sixth instar locusts (Schistocera americana) of either sex were starved for 24-48 hours before the experiment or taken straight from the colony and fed blades of grass for the satiated condition. Locusts were immobilized by placing them in the plastic tube and securing their body with black electric tape (see Author response image 2). Locusts were given 20 - 30 minutes to acclimatize after placement in the immobilization tube. As can be noted, the head of the locusts along with the antenna and maxillary palps protruded out of this immobilization tube so they can be freely moved by the locusts. Note that the maxillary palps are sensory organs close to the mouth parts that are used to grab food and help with the feeding process.

It is worth noting that our earlier studies had shown that the presentation of ‘appetitive odorants’ triggers the locust to open their maxillary palps even when no food is presented (Saha et al., 2017; Nizampatnam et al., 2018; Nizampatnam et al., 2022; Chandak and Raman, 2023.) Furthermore, our earlies results indicate that the probability of palp opening varies across different odorants (Chandak and Raman, 2023). We chose four odorants that had a diverse range of palp-opening: supra-median (hexanol), median (benzaldehyde), and sub-median (linaool). Therefore, each locust in our experiments was presented with one concentration of four odorants (hexanol, benzaldehyde, linalool, and ammonium) in a pseudorandomized order. The odorants were chosen based on our physiology results such that they evoked different levels of spiking activities.

The odor pulse was 4 s in duration and the inter-pulse interval was set to 60 s. The experiments were recorded using a web camera (Microsoft) placed right in front of the locusts. The camera was fully automated with the custom MATLAB script to start recording 2 seconds before the odor pulse and end recording at odor termination. An LED was used to track the stimulus onset/offset. The POR responses were manually scored offline. Responses to each odorant were scored a 0 or 1 depending on if the palps remained closed or opened. A positive POR was defined as a movement of the maxillary palps during the odor presentation time window as shown on the locust schematic (Main Paper Figure 1).

Author response image 2.

Pictures showing the behavior experiment setup and representative palp-opening responses in a locust.

As the reviewer inquired, we performed a new series of POR experiments, where we explored POR responses to mineral oil and hexanol, before and after serotonin injection. For this study, we used 10 locusts that were starved 24-48 hours before the experiment. Note that hexanol was diluted at 1% (v/v) concentration in mineral oil. Our results reveal that locusts PORs to hexanol (~ 50% PORs) were significantly higher than those triggered by mineral oil (~10% PORs). Injection of serotonin increased the POR response rate to hexanol but did not alter the PORs evoked by mineral oil (Author response image 3).

Author response image 3.

Serotonin does not alter the palp-opening responses evoked by paraffin oil. The PORs before and after (5HT) serotonin injection are summarized and shown as a bar plot for hexanol and paraffin oil. Striped bars signify the data collected after 5HT injection. Significant differences are identified in the plot (one-tailed paired-sample t-test; (*p<0.05).

Regarding recordings of potential PNs - the authors do not provide evidence that they did record from projection neurons and not other types of antennal lobe neurons. Thus, these claims should be phrased more carefully.

In the locust antennal lobe, only the cholinergic projection neurons fire full-blown sodium spikes. The GABAergic local neurons only fire calcium ‘spikelets’ (Laurent, TINS, 1996; Stopfer et al., 2003; see Author response image 4 for an example). Hence, we are pretty confident that we are only recording from PNs. Furthermore, due to the physiological properties of the LNs, their signals being too small, they are also not detected in the extracellular recordings from the locust antennal lobe. Hence, we are confident with our claims and conclusion.

Author response image 4.

PN vs LN physiological differences: Left: A representative raw voltage traces recorded from a local neuron before, during, and after a 4-second odor pulse are shown. Note that the local neurons in the locust antennal lobe do not fire full-blown sodium spikes but only fire small calcium spikelets. On the right: A representative raw voltage trace recorded from a representative projection neuron is shown for comparison. Clear sodium spikes are clearly visible during spontaneous and odor-evoked periods. The gray bar represents 4 seconds of odor pulse. The vertical black bar represents the 40mV.

The presented model suggests labeled lines in the antennal lobe output of locusts. Could the presented model also explain a shift in behavior from aversion to attraction - such as seen in locusts when they switch from a solitarious to a gregarious state? The authors might want to discuss other possible scenarios, such as that odor evaluation and decision-making take place in higher brain regions, or that other neuromodulators might affect behavioral output. Serotonin injections could affect behavior via modulation of other cell types than antennal lobe neurons. This should also be discussed - the same is true for potential PNs - serotonin might not directly affect this cell type, but might rather shut down local inhibitory neurons.

There are multiple questions here. First, regarding solitary vs. gregarious states, we are currently repeating these experiments on solitary locusts. Our preliminary results (not included in the manuscript) indicate that the solitary animals have increased olfactory arousal and respond with a higher POR but are less selective and respond similarly to multiple odorants. We are examining the physiology to determine whether the model for mapping neural responses onto behavior could also explain observations in solitary animals.

Second, this reviewer makes the point raised by Reviewer 1. We agree that odor evaluation and decisionmaking might take place in higher brain regions. All we could conclude based on our data is that a segregation of neural activity based on behavioral relevance might provide the simplest approach to map non-specific increase in stimulus-evoked neural responses onto odor-specific changes in behavioral outcome. Furthermore, our results indicate that hexanol and linalool, two odorants that had an increase and decrease in PORs after serotonin injection, had only minimal neural response overlap in the antennal lobe. These results suggest that the formatting of neural activity to support varying behavioral outcomes might already begin in the antennal lobe. We have added this to our discussion.

Third, regarding serotonin impacting PNs, we performed a new set of current-clamp experiments to examine this issue (Author response image 1). Our results clearly show that projection neuron activity in response to current injections (that should not incorporate feedback inhibition through local neurons) was altered after serotonin injection. Therefore, the observed changes in the odor-evoked neural ensemble activity should incorporate modulation at both individual PN level and at the network level. We have added this to our discussion as well.

Finally, the authors claim that serotonin injection can mimic the starved state behavioral response. However, this is only shown for one of the four odors that are tested for behavior (HEX), thus the data does not support this claim.

We note that Hex is the only appetitive odorant in the panel. But, as reviewer 1 has also brought up a similar point, we have toned down our claims and will investigate this carefully in a future study.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

• Was the POR of the locusts towards linalool and ammonium higher than towards a blank odor cartridge? I ask because the locusts appear to be less likely to respond to these odors and so I am concerned that this assay is not relevant to the ecological context of these odors. In other words, perhaps serotonin did not enhance the responses to these odors in this assay, because this is not a context in which locusts would normally respond to these odors.

The POR response to linalool and ammonium is lower and comparable to that of paraffin oil. Serotonin does not increase POR responses to paraffin oil but does increase response to hexanol (an appetitive odorant). We have clarified this using new data (Author response image 5).

• It seems to me that Figure 5C is the crux for understanding the potential impact of 5-HT on odor coding, but it is somewhat confusing and underutilized. Is the implication that 5-HT decorrelates spontaneous activity such that when an odor stimulus arrives, the odor-evoked activity deviates to a greater degree? The authors make claims about this figure that require the reader to guess as to the aspect of the figure to which they are referring.

The reviewer makes an astute observation. Yes, the spontaneous activity in the antennal lobe network before serotonin introduction is not correlated with the ensemble spontaneous activity after serotonin bath application. Remarkably, the odor-evoked responses were highly similar, both in the reduced PCA space and when assayed using high-dimensional ensemble neural activity vectors. Whether the changes in network spontaneous activity have a function in odor detection and recognition is not fully understood and cannot be convincingly answered using our data. But this is something that we had pondered.

• The modeling component summarized in Figure 6 needs clarification and more detail. Perhaps example traces associated with positive weighting within neural ensemble 1 relative to neural ensemble 2? I struggled to understand conceptually how the model resolved the theoretical discrepancy between physiology and behavior.

As recommended, here is a plot showing the responses of four PNs that had positive weights to hexanol and linalool. As can be expected, each PN in this group had higher responses to hexanol and no response to linalool. Further, the four PNs that received negative weights had response only to linalool.

Author response image 5.

Odor-evoked responses of four PNs that received positive weights in the model (top panel), and four PNs that were assigned negative weights in the model (bottom).

• Was there a significant difference between the PORs of hungry vs. fed locusts? The authors state that they differ and provide statistics for the comparisons to locusts injected with 5-HT, but then don't provide any statistical analyses of hungry vs. fed animals.

The POR responses to HEX (an appetitive odorant) were significantly different between the hungry and starved locusts.

Author response image 6.

A bar plot summarizing PORs to all four odors for satiated locust (highlighted with stripes), before (dark shade), and after 5HT injection (lighter shade). To allow comparison before 5HT injection for starved locust plotted as well (without stripes). The significance was determined using a one-tailed paired-sample ttest(*p<0.05).

• Were any of the effects of 5-HT on odor-evoked PN responses significant? No statistics are provided.

We examined the distribution of odor-evoked responses in PNs before and after 5HT introduction. We found that the overall distribution was not significantly different between the two (one-tailed pairedsample t-test; p = 0.93).

Author response image 7.

Comparison of the distribution of odor-evoked PN responses before (green) and after (purple) 5HT introduction. One-tailed paired sample t-test was used to compare the two distributions.

• The authors interchangeably use "serotonin", "5HT" and "5-HT" throughout the manuscript, but this should be consistent.

This has been fixed in the revised manuscript.

• On page 2 the authors provide an ecological relevance for linalool as being an additive in pesticides, however, linalool is a common floral volatile chemical. Is the implication that locusts have learned to associate linalool with pesticides?

Linalool is a terpenoid alcohol that has a floral odor but has also been used as a pesticide and insect repellent [Beier et al., 2014]. As shown in Author response image 2, it evoked the least POR responses amongst a diverse panel of 22 odorants that were tested. We have clarified how we chose odorants based on the prior dataset in the Methods section.

• In Figure 1, there should be a legend in the figure itself indicating that the black box indicates the absence of POR and the white box indicates presence, rather than just having it in the legend text.

Done.

• In Figure 2, the raw data from each animal can be moved to the supplements. The way it is presented is overwhelming and the order of comparisons is difficult to follow.

Done.

• For the induction of bursting in PNs by the application of 5-HT, were there any other metrics observed such as period, duration of bursts, or peak burst frequency? The authors rely on ISI, but there are other bursting metrics that could also be included to understand the nature of this observation. In particular, whether the bursts are likely due to changes in intrinsic biophysical properties of the PNs or polysynaptic effects.

We could use other metrics as the reviewer suggests. Our main point is that the spontaneous activity of individual PNs changed. We have added a new current-injection experiments to show that the PNs output to square pulses of current becomes different after serotonin application (Author response image 1)

• Were 4-vinyl anisole, 1-nonanol, and octanoic acid selected as additional odors because they had particular ecological relevance, or was it for the diversity of chemical structure?

These odorants were selected based on both, chemical structure and ecological relevance. The logic behind this was to have a very diverse odor panel that consisted of food odorant – Hexanol, aggregation pheromone – 4-vinyl anisole, sex pheromone – benzaldehyde, acid – octanoic acid, base – ammonium, and alcohol – 1-nonanol. Additionally, we selected these odors based on previous neural and behavioral data on these odorants (Chandak and Raman, 2023, Traner and Raman, 2023, Nizampatnam et al, 2022 & 2018; Saha et al., 2017 & 2013).

Reviewer #2 (Recommendations For The Authors):

The electrophysiology dataset combines all performed experiments across all tested different PN-odor pairs. How many odors have been tested in a single PN and how many PNs have been tested for a single odor? This information is not present in the current manuscript. Can the authors exclude that there are odor-specific modulations?

In total, our dataset includes recordings from 19 PNs. Seven PNs were tested on a panel of seven odorants (4-vinyl anisole, 1-nonanol, octanoic acid, Hex, Bza, Lool, and Amn), and the remaining twelve were tested with the four main odorants used in the study (Hex, Bza, Lool, and Amn). This information has been added to the Methods section

How did the authors choose the concentrations of serotonin injections and bath applications - is this a naturalistic amount?

The serotonin concentration for ephys experiments was chosen based on trial-error experiments:

0.01mM was the highest concentration that did not cause cell death. For the behavioral experiments, we increased the concentration (0.1 M) due to the presence of anatomical structures in the locust's head such as air sacks, sheath as well as hemolymph which causes some degree of dilution that we cannot control.

Behavior experiments were performed 3 hours after injection - ephys experiments 5-10 minutes following bath application. Can the authors exclude that serotonin affects neural processing differently on these different timescales?

We cannot exclude this possibility. We did ePhys experiments 5-10 minutes after bath application as it would be extremely hard to hold cells for that long.

A longer delay was required for our behavioral experiments as the locusts tended to be a bit more agitated with larger spontaneous movements of palps as well as exhibited unprompted vomiting. A 3hour period allowed the locust to regain its baseline level movements after 5HT introduction. [This information has been added to the methods section of the revised manuscript]

Concerning the analysis of electrophysiological data. The authors should correct for changes in the baseline before performing PCA analysis. And how much of the variance is explained by PC1 and PC2?

We did not correct for baseline changes or subtract baseline as we wanted to show that the odor-evoked neural responses still robustly encoded information about the identity of the odorant.

The authors should perform dye injections after recordings to visualize the cell type they recorded from. Serotonin might affect also other cell types in the antennal lobe.

As mentioned above, in the locust antennal lobe only PNs fire full-blown sodium spikes, and LNs only fire calcium spikelets (Author response image 4). Since these signals are small, they will be buried under the noise floor when using extracellular recording electrodes for monitoring responses in the AL antennal lobe.

Hence we are pretty certain what type of cells we are recording from.

There were several typos in the manuscript, please check again.

We have fixed many of the grammatical errors and typos in the revised version.

Author response:

The following is the authors’ response to the original reviews.

Reviewer #1 (Public Review):

Summary:

The Notch signaling pathway plays an important role in many developmental and disease processes. Although well-studied there remain many puzzling aspects. One is the fact that as well as activating the receptor through trans-activation, the transmembrane ligands can interact with receptors present in the same cell. These cis-interactions are usually inhibitory, but in some cases, as in the assays used here, they may also be activating. With a total of 6 ligands and 4 receptors, there is potentially a wide array of possible outcomes when different combinations are co-expressed in vivo. Here the authors set out to make a systematic analysis of the qualitative and quantitative differences in the signaling output from different receptor-ligand combinations, generating sets of "signaling" (ligand expressing) and "receiving" (receptor +/- ligand expressing cells).

The readout of pathway activity is transcriptional, relying on the fusion of GAL4 in the intracellular part of the receptor. Positive ligand interactions result in the proteolytic release of Gal4 that turns on the expression of H2B-citrine. As an indicator of ligand and receptor expression levels, they are linked via TA to H2B mCherry and H2B mTurq expression respectively. The authors also manipulate the expression of the glycosyltransferase Lunatic-Fringe (LFng) that modifies the EGF repeats in the extracellular domains impacting their interactions. The testing of multiple ligand-receptor combinations at varying expression levels is a tour de force, with over 50 stable cell lines generated, and yields valuable insights although as a whole, the results are quite complex.

Strengths:

Taking a reductionist approach to testing systematically differences in the signaling strength, binding strength, and cis-interactions from the different ligands in the context of the Notch1 and Notch 2 receptors (they justify well the choice of players to test via this approach) produces a baseline understanding of the different properties and leads to some unexpected and interesting findings. Notably:

-                Jag1 ligand expressing cells failed to activate Notch1 receptor although were capable of activating Notch2. Conversely, Jag2 cells elicited the strongest activation of both receptors. The results with

Jag1 are surprising also because it exhibits some of the strongest binding to plate-bound ligands. The failure to activate Notch1 has major functional significance and it will be important in the future to understand the mechanistic basis.

-                Jagged ligands have the strongest cis-inhibitory effects and the receptors differ in their sensitivity to cis-inhibition by Dll ligands. These observations are in keeping with earlier in vivo and cell culture studies. More referencing of those would better place the work in context but it nicely supports and extends previous studies that were conducted in different ways.

-                Responses to most trans-activating ligands showed a degree of ultrasensitivity but this was not the case for cis-interactions where effects were more linear. This has implications for the way the two mechanisms operate and for how the signaling levels will be impacted by ligand expression levels.

-                Qualitatively similar results are obtained in a second cell line, suggesting they reflect fundamental properties of the ligands/receptors.

We appreciate the positive and constructive feedback.

Weaknesses:

One weakness is that the methods used to quantify the expression of ligands and receptors rely on the co-translation of tagged nuclear H2B proteins. These may not accurately capture surface levels/correctly modified transmembrane proteins. In general, the multiple conditions tested partly compensate for the concerns - for example, as Jag1 cells do activate Notch2 even if they do not activate Notch1 some Jag1 must be getting to the surface. But even with Notch2, Jag1 activities are on the lower side, making it important to clarify, especially given the different outcomes with the plated ligands. Similarly, is the fact that all ligands "signalled strongest to Notch2" an inherent property or due to differences in surface levels of Notch 2 compared to Notch1? The results would be considerably strengthened by calibration of the ligand/receptor levels (and ideally their sub-cellular localizations). Assessing the membrane protein levels would be relatively straightforward to perform on some of the basic conditions because their ligand constructs contain Flag tags, making it plausible to relate surface protein to H2B, and there are antibodies available for Notch1 and Notch2.

We agree that mCherry fluorescence does not provide a direct readout of active surface ligand levels. As the reviewer points out, the ability of Jag1 to activate Notch2 demonstrates that expressed Jag1 is competent for signaling. Further, in some cases, Jag1-Notch2 activation can be comparable to Dll1-Notch2 activation (Figure 2A). Following the reviewer’s suggestion, we performed a Western blot for multiple expression levels for each of three surface ligands (Dll1, Dll4, Jag1) (Figure 2—figure supplement 2). This blot revealed a signal for surface expression of Jag1. Interpretation is complicated by the expected dependence of the efficiency of surface protein purification on the number of primary amines in the protein, which varies among these ligands, and qualitatively correlates with the staining intensity. While this makes quantitative interpretation difficult, this result further supports the notion that Jag1 is present on the cell surface. Finally, we note that high signaling activity need not, in general, directly correlate with surface expression levels. In fact, one study showed an example in which increased ligand activity occurred with decreased basal ligand surface levels (Antfolk et al., 2017). While one would ideally like to know all parameters of the system, including surface protein levels, rates of recycling, etc. the perspective taken here is that the net effect of these many post-translational processing steps can be subsumed into the overall relationship between the expression of the protein (which, in our case, is read out by the co-translational reporter) and its activity, which is relevant for the behavior of developmental circuits, among other systems. To address this comment, we now explicitly mention the limitation of mCherry as a proxy for surface protein, and add a reference to previous work highlighting the relationship between surface levels and ligand activity.

In terms of the dependence of signaling on Notch levels, the metric of signaling activity used here is explicitly normalized by the mTurquoise co-translational reporter of Notch expression to account for differences in receptor expression across receiver clones. We have added a new figure to show the variation in expression (Figure 1—figure supplement 1A) and to demonstrate this normalization (Figure 1—figure supplement 5). Having said that, as the reviewer correctly points out, we cannot directly address the dependence on surface receptor levels with mTurquoise alone. To address this comment, we have added a figure that shows cotranslational and surface receptor expression for a subset of our receiver clones (Figure 1—figure supplement 1B). Although antibody binding strengths may vary, it appears unlikely that higher surface levels could explain most ligands’ preferential activation of Notch2 over Notch1, since Notch2 levels were lower than Notch1 levels in both surface expression and cotranslational expression.

Cis-activation as a mode of signaling has only emerged from these synthetic cell culture assays raising questions about its physiological relevance. Cis-activation is only seen at the higher ligand (Dll1, Dll4) levels, how physiological are the expression levels of the ligands/receptors in these assays? Is it likely that this would make a major contribution in vivo? Is it possible that the cells convert themselves into "signaling" and "receiving" sub-populations within the culture by post-translational mechanism? Again some analysis of the ligand/receptors in the cultures would be a valuable addition to show whether or not there are major heterogeneities.

The cis-activation results in this paper are, as the reviewer points out, conducted in synthetic cell culture assays. Cis-activation is observed across a large dynamic range of ligand expression, possibly including non-physiologically high levels. However, our previous work (Nandagopal et al, eLife 2019) showed that cis-activation does not require over-expression, as it occurred in unmodified Caco-2 and NMuMG cells with their endogenous ligand and receptor expression levels. As shown here in Figure 4B, cis-activation for Notch2 increases monotonically and is substantial even at intermediate ligand concentrations. In other cases, cis-activation is maximal at intermediate concentrations. We agree that the in vivo role remains unclear, and is difficult to determine due to the typical close contacts among cells in tissues. Therefore, these assays do not speak to in vivo relevance. Note that we can, however, rule out the possibility of trans signaling between well-mixed cell populations at these densities (Figure 4A).

It is hard to appreciate how much cell-to-cell variability in the "output" there is. For example, low "outputs" could arise from fewer cells becoming activated or from all cells being activated less. As presented, only the latter is considered. That may be already evident in their data, but not easy for the reader to distinguish from the way they are presented. For example, in many of the graphs, data have been processed through multiple steps of normalization. Some discussion/consideration of this point is needed.

We agree that in different experiments changes in a mean response can reflect changes in fraction of activated cells, or level of activation or some combination of both. In this work, most assays were conducted by flow cytometry, which provides a full distribution of cellular responses. We provided distributions for some experiments in the supplementary figures (i.e., Figure 4—figure supplement 1, and Figure 5—figure supplement 4). The sheer number of experiments and samples prevents us from displaying all underlying histograms. Therefore, we have provided all flow data sets in an extensive archive that is publicly available on data.caltech.edu (https://doi.org/10.22002/gjjkn-wrj28).

Impact:

Overall, cataloging the outcomes from the different ligand-receptor combinations, both in cis and trans, yields a valuable baseline for those investigating their functional roles in different contexts. There is still a long way to go before it will be possible to make a predictive model for outcomes based on expression levels, but this work gives an idea about the landscape and the complexities. This is especially important now that signaling relationships are frequently hypothesized based on single-cell transcriptomic data. The results presented here demonstrate that the relationships are not straightforward when multiple players are involved.

We appreciate this concise impact summary, and agree with its conclusions.

Reviewer #2 (Public Review):

Summary:

In this manuscript, the authors extend their previous studies on trans-activation, cis-inhibition (PMID: 25255098), and cis-activation (PMID: 30628888) of the Notch pathway. Here they create a large number of cell lines using CHO-K1 and C2C12 cells expressing either Notch1-Gal4 or Notch2-Gal4 receptors which express a fluorescent protein upon receptor activation (receiver cells). For cis-inhibition and cis-activation assays, these cells were engineered to express one of the four canonical Notch ligands (Dll1, Dll4, Jag1, Jag2) under tetracycline control. Some of the receiver cells were also transfected with a Lunatic fringe (Lfng) plasmid to produce cells with a range of Lfng expression levels. Sender cells expressing all of the canonical ligands were also produced. Cells were mixed in a variety of co-culture assays to highlight trans-activation, cis-activation, and cis-inhibition. All four ligands were able to trans-activate Notch1 and Notch 2, except Jag1 did not transactivate Notch1. Lfng enhanced trans-activation of both Notch receptors by Dll1 and Dll2, and inhibited Notch1 activation by Jag2 and Notch2 activation by both Jag 1 and Jag2. Cis-expression of all four ligands was predominantly inhibitory, but Dll1 and Dll4 showed strong cis-activation of Notch2. Interestingly, cis-ligands preferentially inhibited trans-activation by the same ligand, with varying effects on other trans-ligands.

Strengths:

This represents the most comprehensive and rigorous analysis of the effects of canonical ligands on cis- and trans-activation, and cis-inhibition, of Notch1 and Notch2 in the presence or absence of Lfng so far. Studying cis-inhibition and cis-activation is difficult in vivo due to the presence of multiple Notch ligands and receptors (and Fringes) that often occur in single cells. The methods described here are a step towards generating cells expressing more complex arrays of ligands, receptors, and Fringes to better mimic in vivo effects on Notch function.

In addition, the fact that their transactivation results with most ligands on Notch1 and 2 in the presence or absence of Lfng were largely consistent with previous publications provides confidence that the author's assays are working properly.

We appreciate the thoughtful comments and feedback.

Weaknesses:

It was unusual that the engineered CHO cells expressing Notch1-Gal4 were not activated at all by co-culture with Jag1-expressing CHO cells. Many previous reports have shown that Jag1 can activate Notch1 in co-culture assays, including when Notch1 was expressed in CHO cells. Interestingly, when the authors used Jag1-Fc in a plate coating assay, it did activate Notch1 and could be inhibited by the expression of Lfng.

In our assays, we do in fact also see some signaling of Jag1 to Notch1, especially when dLfng is coexpressed (Figure 2—figure supplement 4, formerly Figure 2—figure supplement 3). While these levels are lower than those observed for other ligand-receptor combinations, they are significantly elevated compared to baseline. In specific natural contexts, it will be important to determine whether the weak but non-zero Jag1-Notch1 signaling acts negatively to suppress signaling from other ligands, or provides weak but potentially functionally important levels of signaling. Evidence for both modes exists in the literature. To address this, we have expanded the discussion of Jag1-Notch1 signaling and added references to other work on Jag1-Notch1 signaling to the Discussion section.

The cell surface level of the ligands was determined by flow cytometry of a co-translated fluorescent protein. Some calibration of the actual cell surface levels with the fluorescent protein would strengthen the results.

This issue was also raised by Reviewers #1 and #3. Please see responses to Reviewer #1, above.

Reviewer #3 (Public Review):

Summary:

This manuscript reports a comprehensive analysis of Notch-Delta/Jagged signaling inclusive of the human Notch1 and Notch2 receptors and DLL1, DLL4, JAG1, and JAG2 ligands. Measurements

encompassed signaling activity for ligand trans-activation, cis-activation, cis-inhibition, and activity modulation by Lfng. The most striking observations of the study are that JAG1 has no detectable activity as a Notch1 ligand when presented on a cell (though it does have activity when immobilized on a surface), even though it is an effective cis-inhibitor of Notch1 signaling by other ligands, and that DLL1 and DLL4 exhibit cis-activating activity for Notch1 and especially for Notch2. Notwithstanding the artificiality of the system and some of its shortcomings, the results should nevertheless be a valuable resource for the Notch signaling community.

Strengths:

(1)  The work is systematic and comprehensive, addressing questions that are of importance to the community of researchers investigating mammalian Notch proteins, their activation by ligands, and the modulation of ligand activity by LFng.

(2)  A quantitative and thorough analysis of the data is presented.

Weaknesses:

(1) The manuscript is primarily descriptive and does not delve into the underlying, mechanistic origin or source of the different ligand activities.

We agree that the goals of this paper were largely to discover the range of signaling modes that occur. A mechanistic analysis would be beyond the scope of this work, but we agree it is an important next step.

(2) The amount of ligand or receptor expressed is inferred from the flow cytometry signal of a co-translated fluorescent protein-histone fusion, and is not directly measured. The work would be more compelling if the amount of ligand present on the cell surface were directly measured with anti-ligand antibodies, rather than inferred from measurements of the fluorescent protein-histone fusion.

This issue was also raised by Reviewers #1 and #2. Please see responses to Reviewer #1, above.

(3) It would be helpful to see plots of the raw activity data before transformation and normalization, because the plots present data after several processing steps, and it is not clear how the processed data relate to the original values determined in each measurement.

We included examples showing how raw data is processed in Figure 4—figure supplement 1 and Figure 5—figure supplement 4. The sheer number of experiments precludes including similar figures for all data sets. However, all raw and processed data and data analysis code is publicly available at (https://doi.org/10.22002/gjjkn-wrj28).

(4) The authors use sparse plating of engineered cells with parental (no ligand or receptor-expressing cell to measure cis activation). However, the cells divide within the cultured period of 22-24 h and can potentially trans-activate each other.

If measured cis-activation signal arises solely from trans-activation, then the measured cis-activation signal per cell should increase with cell density, since trans-activation per cell does depend on cell density (Figure 4A). However, for the strongest cis-activators (Dll1- and Dll4-Notch2), signaling magnitude is similar when these cells are cultured sparsely or at confluence, which would otherwise allow efficient trans signaling (Figure 5A). Thus, for Dll1- and Dll4-Notch2 receivers, total signaling strength per cell depends little or not at all on the opportunity to signal intercellularly. Moreover, cis-activation signal for the Dll1- and Dll4-Notch2 combinations exceeded the maximum trans-signaling levels we could achieve for the same receivers when cis-ligand was suppressed (Figure 4B). These results argue that cis interactions dominate signaling in this context. However, we have not ruled out the possibility that trans-signaling between sister cells after division contributes to the comparatively weak cis-activation observed for Notch1 receivers.

Reviewer #1 (Recommendations For The Authors):

As outlined in the public review, there is a question of whether the nuclear H2B accurately reflects the surface levels of the transmembrane proteins (ligand and receptor). Clearly, it would not be feasible to check levels in all of the experimental conditions, but some baseline conditions should be analyzed.

We addressed this above.

Reviewer #2 (Recommendations For The Authors):

(1)  As mentioned above, it was unusual that Jag1 did not activate Notch1 in co-culture assays, but did activate Notch1 in plate-coating assays. The authors should add some text to the Discussion to explain why they think this is happening in their engineered cells. One possibility is that the CHO cells express Manic fringe (Mfng) which is known to reduce Jag1-Notch1 activation. Data for Mfng levels in CHO cells were not included in Supplemental Table 2. Knocking down all three Fringes in CHO cells might increase Jag1-Notch1 activation.

This is already addressed in a sentence in the results: “Strikingly, while Jag1 sender cells failed to activate Notch1 receivers above background (Figure 2D), plate-bound Jag1-ext-Fc activated Notch1 only ~3-fold less efficiently than it activated Notch2 (Figure 3B-D). This suggests that the natural endocytic activation mechanism, or potential differences in tertiary structure between the expressed and recombinant Jag1 extracellular domains, could play roles in preventing Jag1-Notch1 signaling in coculture.” Regarding the point about Mfng, we added a note to Supplementary Table about other CHO-K1 expression data.

(2) Figure 1-supplemental figure 1: Both the Notch1-Jag1 and Notch1-Jag2 cells show high expression of Jag1 in low 4epi, but any higher concentration reduces to control levels. How much of a problem is this for interpreting your data?

This was not the ideal behavior, but by binning cells by co-translational reporters for ligand expression, we were able to obtain enough cells in intermediate bins. (Note: Figure 1—figure supplement 1 is now Figure 1—figure supplement 2.)

(3)  Figure 1C legend: Are these stably-expressing cells or Tet-off cells? Please state in legend.

The figure legend has been updated.

(4)  Figure 1E: How long is the knockdown of Rfng and Lfng effective? Does it affect the expression of Lfng later?

siRNA effects generally last for at least 72-96 hours, so we do not anticipate this being an issue.

(5) Page 9: "Lfng significantly decreased trans-activation of both receptors by Jag1 (>2.5-fold)". If there is no Jag1-Notch1 activation, how can Lfng decrease trans-activation?

We added a note in the main text to clarify that while Jag1-Notch1 signaling is relatively low, it can still be detectably decreased.

(6) Figure 4A legend: Please define what "2.5k ea senders and Rec" means. In the text, it says "To focus on cis-interactions alone, we then cultured receiver cells at low density, amid an excess of wildtype CHO-K1 cells" (page 14).

This was clarified in the text.

(7)  Page 14: "By contrast, Notch2 was cis-activated by both Dll1 and Dll4, to levels exceeding those produced by trans-activation by high-Dll1 senders (Figure 4B, lower left)." Where is the trans-activation data? 4B, lower right?

We updated this reference in the main text.

(8)  Page 16: "For Notch2-Dll1 and Notch2-Dll4, single cell reporter activities correlated with cis-ligand expression, regardless of whether cells were pre-induced at a high or low culture density (Figure 4D)." It appears that Notch2-Dll1 has lower Notch activation at sparse culture than confluent.

We agree that the level signaling is lower in sparse compared to confluent on average. This is explained by the sensitivity of the Tet-OFF promoter to culture density (Figure 4—figure supplement 2). However, the key point of this experiment is the positive correlation, which is consistent with cis-activation, and inconsistent with the pre-generation of NEXT hypothesis diagrammed in Figure 4C, which would not be expected to produce such a correlation.

(9a) For the creation of the C2C12-Nkd cells: Has genomic sequencing been done to confirm editing of Notch2 and Jag1 loci?

We confirmed the knockdown but did not do genomic sequencing.

(9b) The gel in Figure 7-Supplement 1C is not adequate for showing loss of Jag1. It should be repeated.

In this case, we have only the single gel. We added a note in figure legend that no duplicate was performed.

(10) Figure 7A: Which Fringes are expressed in C2C12 cells? You should provide a rationale for knocking down just Rfng.

Figure 7—figure supplement 1A shows the levels of expression in C2C12. Note that Mfng is not highlighted because its levels were undetectable.

(11) Figure 7-Supplement 1D: This is confusing. Notch2 levels are not reduced in the left panel, and Notch1 and Notch2 levels are not reduced in the right panel?

C2C12-Nkd cells exhibit reduced levels of Notch1 and Notch3. This can be seen in Figure 7—figure supplement 1A. Panel D presents the results of additional siRNA knockdown, performed to prevent subsequent up-regulation of Notch1 and Notch3 during the assay. These knockdown results were variable, as shown. The Notch2 siRNA knockdown was not essential for these experiments, but performed despite very low levels of Notch2 to begin with. In the revision, we have added this note to the Methods.

Reviewer #3 (Recommendations For The Authors):

(1) The results section of the manuscript is very dense and difficult to follow, as are the figure legends.

We appreciate the criticism, and regret that it is not easier to read in its current form.

(2) The authors could emphasize areas of concordance with published results (where available) to place their artificial, engineered system into a better biological context. Are there any examples of studies in whole organisms where cis-activation plays a role?

We are not aware of examples of cis-activation in whole organisms at this point.

(3) How do the authors rationalize the different responses of Notch1 to cell-presented Jag1 as opposed to immobilized Jag1, where its signal strength is second in rank order on a molar basis?

This comment was addressed above in response to the first recommendation from Reviewer #2.

It is also difficult to understand Figure 2_—_figure Supplement 3B, in which it appears that Jag1 induces a Notch1 reporter response when LFng is knocked down (dLfng), and how those data relate to the inactive response to Jag1 shown in the main figures.

The issue here is a difference of normalization. Figure 2A in the main text is normalized to the sender expression level, i.e. relative signaling strength. By contrast, Figure 2—figure supplement 4B (previously Figure 2—figure supplement 3B) shows absolute signaling activity, which can appear higher because it does not normalize for ligand expression. For Jag1-Notch1 signaling in particular, substantial signaling required very high levels of Jag1. We have added a new figure to demonstrate these two types of normalization (Figure 2—figure supplement 1A).

See the Authr response image 1 below for a direct comparison of these two normalization modes using data from both Figure 2A and Figure 2—figure supplement 4B. Note how the Jag1-Notch1 signaling activities that are nonzero in the top plot go to zero in the bottom plot as a result of normalizing the values to ligand expression.

Author response image 1. Comparison of normalization modes in Figure 2A and Figure 2—figure supplement 4B (formerly 3B). Normalized trans-activation signaling activities for different ligand-receptor combinations (with dLfng only), either with further normalization to ligand expression (bottom row) or without further normalization (top row). Normalized signaling activity is defined as reporter activity (mCitrine, A.U.) divided by cotranslational receptor expression (mTurq2, A.U.), normalized to the strongest biological replicate-averaged signaling activity across all ligand-receptor-Lfng combinations in this experiment. Saturated data points, defined here as those with normalized signaling activity over 0.75 in both dLfng and Lfng conditions, were excluded. Colors indicate the identity of the trans-ligand expressed by cocultured sender cells. Error bars denote bootstrapped 95% confidence intervals (Methods), in this case sampled from the number of biological replicates given in the legend—n1 (for Notch1) or n2 (for Notch2). See Methods and Figure 2A caption for more details. Note that the only difference between this figure and the new Figure 2—figure supplement 1A is that this figure additionally includes the Jag1-high data from Figure 2—figure supplement 4B.

Author Response

The following is the authors’ response to the original reviews.

eLife assessment

This fundamental study evaluates the evolutionary significance of variations in the accuracy of the intron-splicing process across vertebrates and insects. Using a powerful combination of comparative and population genomics approaches, the authors present convincing evidence that species with lower effective population size tend to exhibit higher rates of alternative splicing, a key prediction of the drift-barrier hypothesis. The analysis is carefully conducted and all observations fit with this hypothesis, but focusing on a greater diversity of metazoan lineages would make these results even more broadly relevant. This study will strongly appeal to anyone interested in the evolution of genome architecture and the optimisation of genetic systems.

Public Reviews):

Reviewer #1 (Public Review:

Summary:

Functionally important alternative isoforms are gold nuggets found in a swamp of errors produced by the splicing machinery.

The architecture of eukaryotic genomes, when compared with prokaryotes, is characterised by a preponderance of introns. These elements, which are still present within transcripts, are rapidly removed during the splicing of messenger RNA (mRNA), thus not contributing to the final protein. The extreme rarity of introns in prokaryotes, and the elimination of these introns from mRNAs before translation into protein, raises questions about the function of introns in genomes. One explanation comes from functional biology: introns are thought to be involved in post-transcriptional regulation and in the production of translational variants. The latter function is possible when the positions of the edges of the spliced intron vary. While some light has been shed on specific examples of the functional role of alternative splicing, to what extent are they representative of all introns in metazoans?

In this study, the hypothesis of a functional role for alternative splicing, and therefore to a certain extent for introns, is evaluated against another explanation coming from evolutionary biology: isoforms are above all errors of imprecision by the molecular machinery at work during splicing. This hypothesis is based on a principle established by Motoo Kimura, which has become central to population genetics, explaining that the evolutionary trajectory of a mutation with a given effect is intimately linked to the effective population size (Ne) where this mutation emerges. Thus, the probability of fixation of a weakly deleterious mutation increases when Ne decreases, and the probability of fixation of a weakly advantageous mutation increases when Ne increases. The genomes of populations with low Ne are therefore expected to accumulate more weakly deleterious mutations and fewer weakly advantageous mutations than populations with high Ne. In this framework, if splicing errors have only small effects on the fitness of individuals, then natural selection cannot increase the precision of the splicing machinery, allowing tolerance for the production of alternative isoforms.

In the past, the debate opposed one-off observations of effectively functional isoforms on the one hand, to global genomic quantities describing patterns without the possibility of interpreting them in detail. The authors here propose an elegant quantitative approach in line with the expected continuous variation in the effectiveness of selection, both between species and within genomes. The result describing the inter-specific pattern on a large scale confirms what was already known (there is a negative relationship between effective size and average alternative splicing rate). The essential novelty of this study lies in 1) the quantification, for each intron studied, of the relative abundance of each isoform, and 2) the analysis of a relationship between this abundance and the evolutionary constraints acting on these isoforms.

What is striking is the light shed on the general very low abundance of alternative isoforms. Depending on the species, 60% to 96% of cases of alternatively spliced introns lead to an isoform whose abundance is less than 5% of the total variants for a given intron.

In addition to the fact that 60 %-96% of the total isoforms are more than 20 times less abundant than their majority form, this large proportion of alternative isoforms exhibit coding-phase shift at rates similar to what would be expected by chance, i.e. for a third of them, which reinforces the idea that there is no particular constraint on these isoforms.

The remaining 4%-40% of isoforms see their coding-phase shift rate decrease as their relative abundance increases. This result represents a major step forward in our understanding of alternative splicing and makes it possible to establish a quantitative model directly linking the relative abundance of an isoform with a putative functional role concerning only those isoforms produced in abundance. Only the (rare) isoforms which are abundantly produced are thought to be involved in a biological function.

Within the same genome, the authors show that only highly expressed genes, i.e. those that tend to be more constrained on average, are also the genes with the lowest alternative splicing rates on average.

The comparison between species in this study reveals that the smaller the effective size of a species, the more its genome produces isoforms that are low in abundance and low in constraint. Conversely, species with a large effective size relatively reduce rare isoforms, and increase stress on abundant isoforms. To sum up:

• the higher the effective size of a species, the fewer introns are spliced.

• highly expressed genes are spliced less.

• when splicing occurs, it is mainly to produce low-abundance isoforms.

• low-abundance isoforms are also less constrained.

Taken together, these results reinforce a quantitative view of the evolution of alternative splicing as being mainly the product of imprecision in the splicing machinery, generating a great deal of molecular noise. Then, out of all this noise, a few functional gold nuggets can sometimes emerge. From the point of view of the reviewer, the evolutionary dynamics of genomes are depressing. The small effective population sizes are responsible for the accumulation of multiple slightly deleterious introns. Admittedly, metazoan genomes try to get rid of these introns during RNA maturation, but this mechanism is itself rendered imprecise by population sizes.

Strengths:

• The authors simultaneously study the effects of effective population size, isoform abundance, and gene expression levels on the evolutionary constraints acting on isoforms. Within this framework, they clearly show that an isoform becomes functionally important only under certain rare conditions.

• The authors rule out an effect putatively linked to variations in expression between different organs which could have biased comparisons between different species.

Weaknesses:

• While the longevity of organisms as a measure of effective size seems to work overall, it may not be relevant for discriminating within a clade. For example, within Hymenoptera, we might expect them to have the same overall longevity, but that effective size would be influenced more by the degree of sociality: solitary bees/ants/wasps versus eusocial. I am therefore certain that the relationship shown in Figure 4D is currently not significant because the measure of effective size is not relevant for Hymenoptera. The article would have been even more convincing by contrasting the rates of alternative splicing between solitary versus social hymenopterans.

As suggested by the reviewer, we investigated the degree of sociality for the 18 hymenopterans included in our study. We observed that the average dN/dS of the 12 eusocial species (4 bees, 6 ants, 2 wasps) is significantly higher than that of the 6 solitary species (p=2.1x10-3; Fig. R1A), consistent with a lower effective population size in eusocial species compared to solitary ones.

However, the AS rate does not differ significantly between these two groups, neither for the full set of major-isoform introns (Author response image 1B), nor for the subsets of low-AS or high-AS major-isoform introns (Author response image 1C,D). Given the limited sample size (12 eusocial species, 6 solitary species), it is possible that some uncontrolled variables affecting the AS rate hide the impact of Ne.

Author response image 1.

Comparison of solitary (N=6) and eusocial hymenopterans (N=12). A: dN/dS ratio. B: AS rate (all major-isoform introns). C: AS rate (low-AS major-isoform introns). D: AS rate (high-AS major-isoform introns). The means of the two group were compared with a Wilcoxon test.

• When functionalist biologists emphasise the role of the complexity of living things, I'm not sure they're thinking of the comparison between "drosophila" and "homo sapiens", but rather of a broader evolutionary scale. Which gives the impression of an exaggeration of the debate in the introduction.

We disagree with the referee: in fact, all the debate regarding the paradox of the absence of relationship between the number of genes and organismal complexity arose from the comparative analysis of gene repertoires across metazoans. This debate started in the early 2000’s, when the sequencing of the human genome revealed that it contains only ~20,000 protein-coding genes (far less than the ~100,000 genes that were expected at that time). This came as a big surprise because it showed that the gene repertoire of mammals is not larger than that of invertebrates such as Caenorhabditis elegans (19,000 genes) or Drosophila melanogaster (14,000 genes) . We cite below several articles that illustrate how this paradox has been perceived by the scientific community:

Graveley BR 2001 Alternative splicing: increasing diversity in the proteomic world. Trends in Genetics 17 : 100–107. https://doi.org/10.1016/S0168-9525(00)02176-4

“ How can the genome of Drosophila melanogaster contain fewer genes than the undoubtedly simpler organism Caenorhabditis elegans? ”

Ewing B and Green P 2000 Analysis of expressed sequence tags indicates 35,000 human genes. Nature Genetics 25: 232–234. https://doi.org/10.1038/76115

“ the invertebrates Caenorhabditis elegans and Drosophila melanogaster having 19,000 and 13,600 genes, respectively. Here we estimate the number of human genes […] approximately 35,000 genes, substantially lower than most previous estimates. Evolution of the increased physiological complexity of vertebrates may therefore have depended more on the combinatorial diversification of regulatory networks or alternative splicing than on a substantial increase in gene number. ”

Kim E, Magen A and Ast G 2007 Different levels of alternative splicing among eukaryotes. Nucleic Acids Research 35: 125–131. https://doi.org/10.1093/nar/gkl924

“we reveal that the percentage of genes and exons undergoing alternative splicing is higher in vertebrates compared with invertebrates. […] The difference in the level of alternative splicing suggests that alternative splicing may contribute greatly to the mammal higher level of phenotypic complexity,”

Nilsen TW and Graveley BR 2010 Expansion of the eukaryotic proteome by alternative splicing. Nature 463 : 457–463. https://doi.org/10.1038/nature08909

“ It is noteworthy that Caenorhabditis elegans, D. melanogaster and mammals have about 20,000 (ref. 68), 14,000 (ref. 69) and 20,000 (ref. 70) genes, respectively, but mammals are clearly much more complex than nematodes or flies.”

Reviewer #2 (Public Review):

Summary:

Two hypotheses could explain the observation that genes of more complex organisms tend to undergo more alternative splicing. On one hand, alternative splicing could be adaptive since it provides the functional diversity required for complexity. On the other hand, increased rates of alternative splicing could result through nonadaptive processes since more complex organisms tend to have smaller effective population sizes and are thus more prone to deleterious mutations resulting in more spurious splicing events (drift-barrier hypothesis). To evaluate the latter, Bénitière et al. analyzed transcriptome sequencing data across 53 metazoan species. They show that proxies for effective population size and alternative splicing rates are negatively correlated. Furthermore, the authors find that rare, nonfunctional (and likely erroneous) isoforms occur more frequently in more complex species. Additionally, they show evidence that the strength of selection on splice sites increases with increasing effective population size and that the abundance of rare splice variants decreases with increased gene expression. All of these findings are consistent with the drift-barrier hypothesis.

This study conducts a comprehensive set of separate analyses that all converge on the same overall result and the manuscript is well organized. Furthermore, this study is useful in that it provides a modified null hypothesis that can be used for future tests of adaptive explanations for variation in alternative splicing.

Strengths:

The major strength of this study lies in its complementary approach combining comparative and population genomics. Comparing evolutionary trends across phylogenetic diversity is a powerful way to test hypotheses about the origins of genome complexity. This approach alone reveals several convincing lines of evidence in support of the drift-barrier hypothesis. However, the authors also provide evidence from a population genetics perspective (using resequencing data for humans and fruit flies), making results even more convincing.

The authors are forward about the study's limitations and explain them in detail. They elaborate on possible confounding factors as well as the issues with data quality (e.g. proxies for Ne, inadequacies of short reads, heterogeneity in RNA-sequencing data).

Weaknesses:

The authors primarily consider insects and mammals in their study. This only represents a small fraction of metazoan diversity. Sampling from a greater diversity of metazoan lineages would make these results and their relevance to broader metazoans substantially more convincing. Although the authors are careful about their tone, it is challenging to reconcile these results with trends across greater metazoans when the underlying dataset exhibits ascertainment bias and represents samples from only a few phylogenetic groups. Relatedly, some trends (such as Figure 1B-C) seem to be driven primarily by non-insect species, raising the question of whether some results may be primarily explained by specific phylogenetic groups ( although the authors do correct for phylogeny in their statistics). How might results look if insects and mammals (or vertebrates) are considered independently?

Following the referee’s suggestion, we investigated the relationship between AS rate and proxies of Ne, separately for insects and vertebrates (Supplementary Fig. 11) . We observed that the relationship was consistent in vertebrates and insects: linear regressions show a positive correlation, significant (p<0.05) in all cases, except for body length in vertebrates. We added a sentence (line 166) to mention this point.

Note that for these analyses we have smaller sample sizes, so we have a weaker power to detect signal. We therefore prefer to present the combined analyses, using PGLS to account for phylogenetic inertia.

Throughout the manuscript, the authors refer to infrequently spliced ( mode <5%) introns as "minor introns" and frequently spliced (mode >95%) as "major introns". This is extremely confusing since "minor introns" typically represent introns spliced by the U12 spliceosome, whereas "major introns" are those spliced by the U2 spliceosome.

To avoid any confusion, we modified the terminology: we now refer to infrequently spliced introns as " minor-isoform introns" and frequently spliced as "major -isoform introns" (see line 135-137) . The entire manuscript (including the figures) has been modified accordingly.

Furthermore, it remains unclear whether the study only considers major introns or both major and minor introns. Minor introns typically have AT-AC splice sites whereas major introns usually have GT/GC-AG splice sites, although in rare cases the U2 can recognize AT-AC (see Wu and Krainer 1997 for example).

We modified the text (line 148-150) to clearly state that we studied all introns, both U2-type and U12-type.

The authors also note that some introns show noncanonical AT-AC splice sites while these are actually canonical splice sites for minor introns.

This is corrected (line 148).

Recommendations for the authors:

Reviewer #2 (Recommendations For The Authors):

Figures 1, 3, and 4: I suggest that authors add regression lines.

We added the regression lines with the “pgls” function from the R package “caper” (in Fig. 1, 3 and 4, and also in all other figures where we present correlations).

Figure 2: As previously mentioned, the terms "minor introns" and "major introns" are extremely confusing. I strongly suggest the authors use different naming conventions.

We changed the terminology:

minor introns -> minor-isoform introns

major introns -> major-isoform introns

Figure 5: Intron-exon boundaries and splice site annotations are shown at the bottom of B, C, and D but not A. I suggest removing the annotation beneath B for consistency and since A+C and B+D are aligned on the x-axis.

Corrected, it was a mistake.

Figure 7: The yellow dotted line is very challenging to see in A.

Corrected, the line has been widened.

eLife Assessment

This important study looks into the effect of exogenous CoA on the response of TLR4-activated macrophages. Specifically, CoA enhances the LPS response by examining metabolomics, 13C tracing, and assessments of transcription and acetylation. Together, these provide a compelling series of findings that show exogenous CoA is taken up by macrophages, and this facilitates histone acetylation and transcription associated with activation and antimicrobial activity.

Reviewer #1 (Public review):

Summary:

This paper describes how CoA can overcome suppression of OXPHOS in TLR3 signaling, acting as what the authors term a 'metabolic adjuvant'. Supplementing with CoA enhances TLR signaling, reverses tolerance, and promotes OXPHOS. It promotes histone acetylation, leading to epigenetic modulation of target genes. CoA is further shown to have adjuvant effects in vivo, in anti-tumor immunity, and also in host defense.

Strengths:

Something of a tour-de-force - impressive methodologies and the conclusions are well supported by the data.

Weaknesses:

I was unable to follow the basis for some experiments and have a question around the data on itaconate, since this metabolite should limit IL-1beta production. Also, this is a very wordy manuscript - editing should help the reader.

Reviewer #2 (Public review):

In this manuscript, Timblin et al provide a model where exogenous CoA is taken up by macrophages and utilized to support transcriptional events associated with activation. They provide a series of important findings, and for the most part, the data are clear and convincing. However, additional clarity on a few points would be helpful.

First of all, the contention that endogenous TLR ligands from the bone marrow cultures are driving the tonic signaling that makes exogenous CoA beneficial in unstimulated cells seems counter to the well-described anergic state of myeloid cells derived from TLR-null mice. This reviewer's understanding was that myeloid cells in MyD88 nulls or similar are developmentally anergic due to the lack of TLR stimulation in vivo. The data here (Figure 5G, etc) show these cells have much lower TLR responses, but the authors attribute it to loss of response to endogenous ligands during the cultures rather than in vivo. Testing some of the phenotypes ex vivo, etc, might make this argument more compelling and rule out that this is an effect in vivo.

Second, the data suggesting that CoA enhances anti-microbial activity via itaconate production needs additional context and/or clarification. Interactions between itaconate and CoA have been demonstrated. Itaconate exposure can deplete the CoA pool as it is converted into Itaconyl-CoA. The Irg-/- cells should not have reduced CoA due to the lack of the need to activate itaconate for metabolism. Has this been addressed by the authors? I believe that low levels of itaconate production have been shown in "resting" bone marrow cultures. The data show a full log of more bugs in the macs that lack Irg, confirming that endogenous itaconate is at work. In addition, itaconate, which is made very quickly and is likely there in considerable amounts in 4 hrs, is known to affect transcription via action on TET2. Perhaps this explains some of the connections to CoA?

Lastly, the idea that Acetyl-CoA phenocopies CoA suggests that CoA is the effector is interesting but could be supported more. Did the authors do the "unlabeling" experiment with Acetyl-CoA to confirm that it is readily converted to the CoA pool?

Do the ACLY inhibitors have the expected effects on the ChIP seq data?

eLife Assessment

Karimian et al. present a valuable new model to explain how gamma-band synchrony (30-80 Hz) can support human visual feature binding by selectively grouping image elements, countering recent criticisms that the stimulus dependence of gamma oscillations limits their functional role. Grounded in the theory of weakly coupled oscillators and informed by primate electrophysiology, the model captures behavioural patterns observed in human psychophysics, offering support for the potential role of synchrony-based mechanisms, but incomplete evidence for a specific role of gamma oscillations. This work could be strengthened by more direct evidence for the proposed mechanism, and expanding beyond figure-only model inputs with limited ecological validity.

Reviewer #1 (Public review):

Summary:

This paper by Karimian et al proposes an oscillator model tuned to implement binding by synchrony (BBS*) principles in a visual task. The authors set out to show how well these BBS principles explain human behavior in figure-ground segregation tasks. The model is inspired by electrophysiological findings in non-human primates, suggesting that gamma oscillations in early visual cortex implement feature-binding through a synchronization of feature-selective neurons. The psychophysics experiment involves the identification of a figure consisting of gabor annuli, presented on a background of gabor annuli. The participants' task is to identify the orientation of the figure. The task difficulty is varied based on the contrast and density of the gabor annuli that make up the figure. The same figures (without the background) are used as inputs to the oscillator model. The authors report that both the discrimination accuracy in the psychophysics experiment and the synchrony of the oscillators in the proposed model follow a similar "Arnold Tongue" relationship when depicted as a function of the texture-defining features of the figure. This finding is interpreted as evidence for BBS/gamma synchrony being the underlying mechanism of the figure-ground segregation.

• Note that I chose to use "BBS" over gamma synchrony (used by the authors) in this review, as I am not convinced that the authors show evidence for synchronization in the gamma-band.

Strengths:

The design of the proposed model is well-informed by electrophysiological findings, and the idea of using computational modeling to bridge between intracranial recordings in non-human primates and behavioral results in human participants is interesting. Previous work has criticized the BBS synchrony theory based on the observation that synchronization in the gamma-band is highly localized and the frequency of the oscillation depends on the visual features of the stimulus. I appreciate how the authors demonstrate that frequency-dependence and local synchronization can be features of BBS, and not contradictory to the theory. As such, I feel that this work has the potential to contribute meaningfully to the debate on whether BBS is a biophysically realistic model of feature-binding in visual cortex.

Weaknesses:

I have several concerns regarding the presented claims, assessment of meaning and size of the presented effects, particularly with regard to the absence of a priori defined effect sizes.

Firstly, the paper makes strong claims about the frequency-specificity (i.e., gamma synchrony) and anatomical correlates (early visual cortex) of the observed effects. These claims are informed by previous electrophysiological work in non-human primates but are not directly supported by the paper itself. For instance, the title contains the word "gamma synchrony", but the authors do not demonstrate any EEG/MEG or intracranial data in from their human subjects supporting such claims, nor do they demonstrate that the frequencies in the oscillator model are within the gamma band. I think that the paper should more clearly distinguish between statements that are directly supported by the paper (such as: "an oscillator model based on BBS principles accounts for variance in human behavior") and abstract inferences based on the literature (such as "these effects could be attributed to gamma oscillations in early visual cortex, as the model was designed based on those principles").

Secondly, unlike the human participants, the model strictly does not perform figure-ground segregation, as it only receives the figure as an input. Finally, it is unclear what effect sizes the authors would have expected a priori, making it difficult to assess whether their oscillator model represents the data well or poorly. I consider this a major concern, as the relationship between the synchrony of the oscillatory model and the performance of the human participants is confounded by the visual features of the figure. Specifically, the authors use the BBS literature to motivate the hypothesis that perception of the texture-defined figure is related to the density and contrast heterogeneity of the texture elements (gabor annuli) of the figure. This hypothesis has to be true regardless of synchrony, as the figure will be easier to spot if it consists of a higher number of high-contrast gabors than the background. As the frequency and phase of the oscillators and coupling strength between oscillators in the grid change as a function of these visual features, I wonder how much of the correlation between model synchrony and human performance is mediated by the features of the figure. To interpret to what extent the similarity between model and human behavior relies on the oscillatory nature of the model, the authors should find a way to estimate an empirical threshold that accounts for these confounding effects. Alternatively, it would be interesting to understand whether a model based on competing theories (e.g., Binding by Enhanced Firing, Roelfsema, 2023) would perform better or worse at explaining the data.

Reviewer #2 (Public review):

The authors aimed to investigate whether gamma synchrony serves a functional role in figure-ground perception. They specifically sought to test whether the stimulus-dependence of gamma synchrony, often considered a limitation, actually facilitates perceptual grouping. Using the theory of weakly coupled oscillators (TWCO), they developed a framework wherein synchronization depends on both frequency detuning (related to contrast heterogeneity) and coupling strength (related to proximity between visual elements). Through psychophysical experiments with texture discrimination tasks and computational modeling, they tested whether human performance follows patterns predicted by TWCO and whether perceptual learning enhances synchrony-based grouping.

Strengths:

(1) The theoretical framework connecting TWCO to visual perception is innovative and well-articulated, providing a potential mechanistic explanation for how gamma synchrony might contribute to both feature binding and separation.

(2) The methodology combines psychophysical measurements with computational modeling, with a solid quantitative agreement between model predictions and human performance.

(3) In particular, the demonstration that coupling strengths can be modified through experience is remarkable and suggests gamma synchrony could be an adaptable mechanism that improves with visual learning.

(4) The cross-validation approach, wherein model parameters derived from macaque neurophysiology successfully predict human performance, strengthens the biological plausibility of the framework.

Weaknesses:

(1) The highly controlled stimuli are far removed from natural scenes, raising questions about generalisability. But, of course, control (almost) excludes ecological validity. The study does not address the challenges of natural vision or leverage the rich statistical structure afforded by natural scenes.

(2) The experimental design appears primarily confirmatory rather than attempting to challenge the TWCO framework or test boundary conditions where it might fail.

(3) Alternative explanations for the observed behavioral effects are not thoroughly explored. While the model provides a good fit to the data, this does not conclusively prove that gamma synchrony is the actual mechanism underlying the observed effects.

(4) Direct neurophysiological evidence linking the observed behavioral effects to gamma synchrony in humans is absent, creating a gap between the model and the neural mechanism.

Achievement of Aims and Support for Conclusions:

The authors largely achieved their primary aim of demonstrating that human figure-ground perception follows patterns predicted by TWCO principles. Their psychophysical results reveal a behavioral "Arnold tongue" that matches the synchronization patterns predicted by their model, and their learning experiment shows that perceptual improvements correlate with predicted increases in synchrony.

The evidence supports their conclusion that gamma synchrony could serve as a viable neural grouping mechanism for figure-ground segregation. However, the conclusion that "stimulus-dependence of gamma synchrony is adaptable to the statistics of visual experiences" is only partially supported, as the study uses highly controlled artificial stimuli rather than naturalistic visual statistics, or shows a sensitivity to the structure of experience.

Likely Impact and Utility:

This work offers a fresh perspective on the functional role of gamma oscillations in visual perception. The integration of TWCO with perceptual learning provides a novel theoretical framework that could influence future research on neural synchrony.

The computational model, with parameters derived from neurophysiological data, offers a useful tool for predicting perceptual performance based on synchronization principles. This approach might be extended to study other perceptual phenomena and could inspire designs for artificial vision systems.

The learning component of the study may have a particular impact, as it suggests a mechanism by which perceptual expertise develops through modified coupling between neural assemblies. This could influence thinking about perceptual learning more broadly, but also raises questions about the underlying mechanism that the paper does not address.

Additional Context:

Historically, the functional significance of gamma oscillations has been debated, with early theories of temporal binding giving way to skepticism based on gamma's stimulus-dependence. This study reframes this debate by suggesting that stimulus-dependence is exactly what makes gamma useful for perceptual grouping.

The successful combination of computational neuroscience and psychophysics is a significant strength of this study.

The field would benefit from future work extending (if possible) these findings to more naturalistic stimuli and directly measuring neural activity during perceptual tasks. Additionally, studies comparing predictions from synchrony-based models against alternative mechanisms would help establish the specificity of the proposed framework.

Author response:

The following is the authors’ response to the original reviews.

Public Reviews:  

Reviewer #1 (Public Review): 

Summary: 

In their manuscript entitled 'The domesticated transposon protein L1TD1 associates with its ancestor L1 ORF1p to promote LINE-1 retrotransposition', Kavaklıoğlu and colleagues delve into the role of L1TD1, an RNA binding protein (RBP) derived from a LINE1 transposon. L1TD1 proves crucial for maintaining pluripotency in embryonic stem cells and is linked to cancer progression in germ cell tumors, yet its precise molecular function remains elusive. Here, the authors uncover an intriguing interaction between L1TD1 and its ancestral LINE-1 retrotransposon. 

The authors delete the DNA methyltransferase DNMT1 in a haploid human cell line (HAP1), inducing widespread DNA hypo-methylation. This hypomethylation prompts abnormal expression of L1TD1. To scrutinize L1TD1's function in a DNMT1 knock-out setting, the authors create DNMT1/L1TD1 double knock-out cell lines (DKO). Curiously, while the loss of global DNA methylation doesn't impede proliferation, additional depletion of L1TD1 leads to DNA damage and apoptosis.  

To unravel the molecular mechanism underpinning L1TD1's protective role in the absence of DNA methylation, the authors dissect L1TD1 complexes in terms of protein and RNA composition. They unveil an association with the LINE-1 transposon protein L1-ORF1 and LINE-1 transcripts, among others.  

Surprisingly, the authors note fewer LINE-1 retro-transposition events in DKO cells than in DNMT1 KO alone.  

Strengths: 

The authors present compelling data suggesting the interplay of a transposon-derived human RNA binding protein with its ancestral transposable element. Their findings spur interesting questions for cancer types, where LINE1 and L1TD1 are aberrantly expressed.  

Weaknesses: 

Suggestions for refinement:  

The initial experiment, inducing global hypo-methylation by eliminating DNMT1 in HAP1 cells, is intriguing and warrants a more detailed description. How many genes experience misregulation or aberrant expression? What phenotypic changes occur in these cells? 

This is an excellent suggestion. We have gene expression data on WT versus DNMT1 KO HAP1 cells and have included them now as Suppl. Figure S1. The  transcriptome analysis of DNMT1 KO cells showed hundreds of deregulated genes upon DNMT1 ablation. As expected, the majority were up-regulated and gene ontology analysis revealed that among the strongest up-regulated genes were gene clusters with functions in “regulation of transcription from RNA polymerase II promoter” and “cell differentiation” and genes encoding proteins with KRAB domains. In addition, the de novo methyltransferases DNMT3A and DNMT3B were up-regulated in DNMT1 KO cells suggesting the set-up of compensatory mechanisms in these cells. 

Why did the authors focus on L1TD1? Providing some of this data would be helpful to understand the rationale behind the thorough analysis of L1TD1. 

We have previously discovered that conditional deletion of the maintenance DNA methyltransferase DNMT1 in the murine epidermis results not only in the up-regulation of mobile elements, such as IAPs but also the induced expression of L1TD1 ([1], Suppl. Table 1 and Author response image 1). Similary, L1TD1 expression was induced by treatment of primary human keratinocytes or squamous cell carcinoma cells with the DNMT inhibitor azadeoxycytidine (Author response images 2 and 3). These findings are in accordance with the observation  that inhibition of DNA methyltransferase activity by aza-deoxycytidine in human non-small cell lung cancer cells (NSCLCs) results in up-regulation of L1TD1 [2]. Our interest in L1TD1 was further fueled by reports on a potential function of L1TD1 as prognostic tumor marker. We have included this information in the last paragraph of the Introduction in the revised manuscript.

Author response image 1. RT-qPCR of L1TD1 expression in cultured murine control and Dnmt1 Δ/Δker keratinocytes. mRNA levels of L1td1 were analyzed in keratinocytes isolated at P5 from conditional Dnmt1 knockout mice [1]. Hprt expression was used for normalization of mRNA levels and wildtype control was set to 1. Data represent means ±s.d. with n=4. **P < 0.01 (paired t-test). 

Author response image 2. RT-qPCR analysis of L1TD1 expression in primary human keratinocytes. Cells were treated with 5-aza-2-deoxycidine for 24 hours or 48 hours, with PBS for 48 hours or were left untreated. 18S rRNA expression was used for normalization of mRNA levels and PBS control was set to 1. Data represent means ±s.d. with n=3. **P < 0.01 (paired t-test).

Author response image 3. Induced L1TD1 expression upon DNMT inhibition in squamous cell carcinoma cell lines SCC9 and SCCO12. Cells were treated with 5-aza-2-deoxycidine for 24 hours, 48 hours or 6 days. (A) Western blot analysis of L1TD1 protein levels using beta-actin as loading control. (B) Indirect immunofluorescence microscopy analysis of L1TD1 expression in SCC9 cells. Nuclear DNA was stained with DAPI. Scale bar: 10 µm. (C)  RT-qPCR analysis of L1TD1 expression in primary human keratinocytes. Cells were treated with 5-aza-2deoxycidine for 24 hours or 48 hours, with PBS for 48 hours or were left untreated. 18S rRNA expression was used for normalization of mRNA levels and PBS control was set to 1. Data represent means ±s.d. with n=3. *P < 0.05, **P < 0.01 (paired t-test).

The finding that L1TD1/DNMT1 DKO cells exhibit increased apoptosis and DNA damage but decreased L1 retro-transposition is unexpected. Considering the DNA damage associated with retro-transposition and the DNA damage and apoptosis observed in L1TD1/DNMT1 DKO cells, one would anticipate the opposite outcome. Could it be that the observation of fewer transposition-positive colonies stems from the demise of the most transposition-positive colonies? Further exploration of this phenomenon would be intriguing. 

This is an important point and we were aware of this potential problem. Therefore, we calibrated the retrotransposition assay by transfection with a blasticidin resistance gene vector to take into account potential differences in cell viability and blasticidin sensitivity. Thus, the observed reduction in L1 retrotransposition efficiency is not an indirect effect of reduced cell viability. We have added a corresponding clarification in the Results section on page 8, last paragraph. 

Based on previous studies with hESCs and germ cell tumors [3], it is likely that, in addition to its role in retrotransposition, L1TD1 has further functions in the regulation of cell proliferation and differentiation. L1TD1 might therefore attenuate the effect of DNMT1 loss in KO cells generating an intermediate phenotype (as pointed out by Reviewer 2) and simultaneous loss of both L1TD1 and DNMT1 results in more pronounced effects on cell viability. This is in agreement with the observation that a subset of L1TD1 associated transcripts encode proteins involved in the control of cell division and cell cycle. It is possible that subtle changes in the expression of these protein that were not detected in our mass spectrometry approach contribute to the antiproliferative effect of L1TD1 depletion as discussed in the Discussion section of the revised manuscript. 

Reviewer #2 (Public Review):           

In this study, Kavaklıoğlu et al. investigated and presented evidence for the role of domesticated transposon protein L1TD1 in enabling its ancestral relative, L1 ORF1p, to retrotranspose in HAP1 human tumor cells. The authors provided insight into the molecular function of L1TD1 and shed some clarifying light on previous studies that showed somewhat contradictory outcomes surrounding L1TD1 expression. Here, L1TD1 expression was correlated with L1 activation in a hypomethylation-dependent manner, due to DNMT1 deletion in the HAP1 cell line. The authors then identified L1TD1-associated RNAs using RIP-Seq, which displays a disconnect between transcript and protein abundance (via Tandem Mass Tag multiplex mass spectrometry analysis). The one exception was for L1TD1 itself, which is consistent with a model in which the RNA transcripts associated with L1TD1 are not directly regulated at the translation level. Instead, the authors found the L1TD1 protein associated with L1-RNPs, and this interaction is associated with increased L1 retrotransposition, at least in the contexts of HAP1 cells. Overall, these results support a model in which L1TD1 is restrained by DNA methylation, but in the absence of this repressive mark, L1TD1 is expressed and collaborates with L1 ORF1p (either directly or through interaction with L1 RNA, which remains unclear based on current results), leads to enhances L1 retrotransposition. These results establish the feasibility of this relationship existing in vivo in either development, disease, or both.   

Recommendations for the authors:

Reviewer #2 (Recommendations For The Authors):        

Major 

(1) The study only used one knockout (KO) cell line generated by CRISPR/Cas9. Considering the possibility of an off-target effect, I suggest the authors attempt one or both of these suggestions. 

A) Generate or acquire a similar DMNT1 deletion that uses distinct sgRNAs, so that the likelihood of off-targets is negligible. A few simple experiments such as qRT-PCR would be sufficient to suggest the same phenotype.  

B) Confirm the DNMT1 depletion also by siRNA/ASO KD to phenocopy the KO effect.  (2) In addition to the strategies to demonstrate reproducibility, a rescue experiment restoring DNMT1 to the KO or KD cells would be more convincing. (Partial rescue would suffice in this case, as exact endogenous expression levels may be hard to replicate). 

We have undertook several approaches to study the effect of DNMT1 loss or inactivation: As described above, we have generated a conditional KO mouse with ablation of DNMT1 in the epidermis. DNMT1-deficient keratinocytes isolated from these mice show a significant increase in L1TD1 expression.  In addition, treatment of primary human keratinocytes and two squamous cell carcinoma cell lines with the DNMT inhibitor aza-deoxycytidine led to upregulation of L1TD1 expression. Thus, the derepression of L1TD1 upon loss of DNMT1 expression or activity is not a clonal effect. Also, the spectrum of RNAs identified in RIP experiments as L1TD1-associated transcripts in HAP1 DNMT1 KO cells showed a strong overlap with the RNAs isolated by a related yet different method in human embryonic stem cells. When it comes to the effect of L1TD1 on L1-1 retrotranspostion, a recent study has reported a similar effect of L1TD1 upon overexpression in HeLa cells [4].  

All of these points together help to convince us that our findings with HAP1 DNMT KO are in agreement with results obtained in various other cell systems and are therefore not due to off-target effects. With that in mind, we would pursue the suggestion of Reviewer 1 to analyze the effects of DNA hypomethylation upon DNMT1 ablation.

(3) As stated in the introduction, L1TD1 and ORF1p share "sequence resemblance" (Martin 2006). Is the L1TD1 antibody specific or do we see L1 ORF1p if Fig 1C were uncropped?  (6) Is it possible the L1TD1 antibody binds L1 ORF1p? This could make Figure 2D somewhat difficult to interpret. Some validation of the specificity of the L1TD1 antibody would remove this concern (see minor concern below).  

This is a relevant question. We are convinced that the L1TD1 antibody does not crossreact with L1 ORF1p for the following reasons: Firstly, the antibody does not recognize L1 ORF1p (40 kDa) in the  uncropped Western blot for Figure 1C (Author response image 4A). Secondly, the L1TD1 antibody gives only background signals in DKO cells in the  indirect immunofluorescence experiment shown in Figure 1E of the manuscript. 

Thirdly, the immunogene sequence of L1TD1 that determines the specificity of the antibody was checked in the antibody data sheet from Sigma Aldrich. The corresponding epitope is not present in the L1 ORF1p sequence. Finally, we have shown that the ORF1p antibody does not cross-react with L1TD1 (Author response image 4B).

Author response image 4. (A) Uncropped L1TD1 Western blot shown in Figure 1C. An unspecific band is indicated by an asterisk. (B) Westernblot analysis of WT, KO and DKO cells with L1 ORF1p antibody.

(4) In abstract (P2), the authors mentioned that L1TD1 works as an RNA chaperone, but in the result section (P13), they showed that L1TD1 associates with L1 ORF1p in an RNAindependent manner. Those conclusions appear contradictory. Clarification or revision is required. 

Our findings that both proteins bind L1 RNA, and that L1TD1 interacts with ORF1p are compatible with a scenario where L1TD1/ORF1p heteromultimers bind to L1 RNA. The additional presence of L1TD1 might thereby enhance the RNA chaperone function of ORF1p. This model is visualized now in Suppl. Figure S7C. 

(5) Figure 2C fold enrichment for L1TD1 and ARMC1 is a bit difficult to fully appreciate. A 100 to 200-fold enrichment does not seem physiological. This appears to be a "divide by zero" type of result, as the CT for these genes was likely near 40 or undetectable. Another qRT-PCRbased approach (absolute quantification) would be a more revealing experiment. 

This is the validation of the RIP experiments and the presentation mode is specifically developed for quantification of RIP assays (Sigma Aldrich RIP-qRT-PCR: Data Analysis Calculation Shell). The unspecific binding of the transcript in the absence of L1TD1 in DNMT1/L1TD1 DKO cells is set to 1 and the value in KO cells represents the specific binding relative the unspecific binding. The calculation also corrects for potential differences in the abundance of the respective transcript in the two cell lines. This is not a physiological value but the quantification of specific binding of transcripts to L1TD1. GAPDH as negative control shows no enrichment, whereas specifically associated transcripts show strong enrichement. We have explained the details of RIPqRT-PCR evaluation in Materials and Methods (page 14) and the legend of Figure 2C in the revised manuscript.       

(6) Is it possible the L1TD1 antibody binds L1 ORF1p? This could make Figure 2D somewhat difficult to interpret. Some validation of the specificity of the L1TD1 antibody would remove this concern (see minor concern below).            

See response to (3).  

(7) Figure S4A and S4B: There appear to be a few unusual aspects of these figures that should be pointed out and addressed. First, there doesn't seem to be any ORF1p in the Input (if there is, the exposure is too low). Second, there might be some L1TD1 in the DKO (lane 2) and lane 3. This could be non-specific, but the size is concerning. Overexposure would help see this.

The ORF1p IP gives rise to strong ORF1p signals in the immunoprecipitated complexes even after short exposure. Under these contions ORF1p is hardly detectable in the input. Regarding the faint band in DKO HAP1 cells, this might be due to a technical problem during Western blot loading. Therefore, the input samples were loaded again on a Western blot and analyzed for the presence of ORF1p, L1TD1 and beta-actin (as loading control) and shown as separate panel in Suppl. Figure S4A. 

(8) Figure S4C: This is related to our previous concerns involving antibody cross-reactivity. Figure 3E partially addresses this, where it looks like the L1TD1 "speckles" outnumber the ORF1p puncta, but overlap with all of them. This might be consistent with the antibody crossreacting. The western blot (Figure 3C) suggests an upregulation of ORF1p by at least 2-3x in the DKO, but the IF image in 3E is hard to tell if this is the case (slightly more signal, but fewer foci). Can you return to the images and confirm the contrast are comparable? Can you massively overexpose the red channel in 3E to see if there is residual overlap? 

In Figure 3E the L1TD1 antibody gives no signal in DNMT1/L1TD1 DKO cells confirming that it does not recognize ORF1p. In agreement with the Western blot in Figure 3C the L1 ORF1p signal in Figure 3E is stronger in DKO cells. In DNMT1 KO cells the L1 ORF1p antibody does not recognize all L1TD1 speckles. This result is in agreement with the Western blot shown above in Figure R4B and indicates that the L1 ORF1p antibody does not recognize the L1TD1 protein. The contrast is comparable and after overexposure there are still L1TD1 specific speckles. This might be due to differences in abundance of the two proteins.

(9) The choice of ARMC1 and YY2 is unclear. What are the criteria for the selection?

ARMC1 was one of the top hits in a pilot RIP-seq experiment (IP versus input and IP versus  IgG IP). In the actual RIP-seq experiment with DKO HAP1 cells instead of IgG IP as a negative control, we found ARMC1 as an enriched hit, although it was not among the top 5 hits. The results from the 2nd RIP-seq further confirmed the validity of ARMC1 as an L1TD1-interacting transcript. YY2 was of potential biological relevance as an L1TD1 target due to the fact that it is a processed pseudogene originating from YY1 mRNA as a result of retrotransposition. This is mentioned on page 6 of the revised manuscript.

(10) (P16) L1 is the only protein-coding transposon that is active in humans. This is perhaps too generalized of a statement as written. Other examples are readily found in the literature. Please clarify.  

We will tone down this statement in the revised manuscript. 

(11) In both the abstract and last sentence in the discussion section (P17), embryogenesis is mentioned, but this is not addressed at all in the manuscript. Please refrain from implying normal biological functions based on the results of this study unless appropriate samples are used to support them.

Much of the published data on L1TD1 function are related to embryonic stem cells [3-7]. Therefore, it is important to discuss our findings in the context of previous reports.

(12) Figure 3E: The format of Figures 1A and 3E are internally inconsistent. Please present similar data/images in a cohesive way throughout the manuscript.  

We show now consistent IF Figures in the revised manuscript.

Minor: 

(1) Intro:           

- Is L1Td1 in mice and Humans? How "conserved" is it and does this suggest function?  

Murine and human L1TD1 proteins share 44% identity on the amino acid level and it was suggested that the corresponding genes were under positive selection during evolution with functions in transposon control and maintenance of pluripotency [8].  

- Why HAP1? (Haploid?) The importance of this cell line is not clear.          

HAP1 is a nearly haploid human cancer cell line derived from the KBM-7 chronic myelogenous leukemia (CML) cell line [9, 10]. Due to its haploidy is perfectly suited and widely used for loss-of-function screens and gene editing. After gene editing  cells can be used in the nearly haploid or in the diploid state. We usually perform all experiments with diploid HAP1 cell lines.  Importantly, in contrast to other human tumor cell lines, this cell line tolerates ablation of DNMT1. We have included a corresponding explanation in the revised manuscript on page 5, first paragraph.

- Global methylation status in DNMT1 KO? (Methylations near L1 insertions, for example?) 

The HAP1 DNMT1 KO cell line with a 20 bp deletion in exon 4 used in our study was validated in the study by Smits et al. [11]. The authors report a significant reduction in overall DNA methylation. However, we are not aware of a DNA methylome study on this cell line. We show now data on the methylation of L1 elements in HAP1 cells and upon DNMT1 deletion in the revised manuscript in Suppl. Figure S1B.

(2) Figure 1:  

- Figure 1C. Why is LMNB used instead of Actin (Fig1D)?  

We show now beta-actin as loading control in the revised manuscript.  

- Figure 1G shows increased Caspase 3 in KO, while the matching sentence in the result section skips over this. It might be more accurate to mention this and suggest that the single KO has perhaps an intermediate phenotype (Figure 1F shows a slight but not significant trend). 

We fully agree with the reviewer and have changed the sentence on page 6, 2nd paragraph accordingly.  

- Would 96 hrs trend closer to significance? An interpretation is that L1TD1 loss could speed up this negative consequence. 

We thank the reviewer for the suggestion. We have performed a time course experiment with 6 biological replicas for each time point up to 96 hours and found significant changes in the viability upon loss of DNMT1 and again significant reduction in viability upon additional loss of L1TD1 (shown in Figure 1F). These data suggest that as expexted loss of DNMT1 leads to significant reduction viability and that additional ablation of L1TD1 further enhances this effect.

- What are the "stringent conditions" used to remove non-specific binders and artifacts (negative control subtraction?) 

Yes, we considered only hits from both analyses, L1TD1 IP in KO versus input and L1TD1 IP in KO versus L1TD1 IP in DKO. This is now explained in more detail in the revised manuscript on page 6, 3rd paragraph.  

(3) Figure 2:  

- Figure 2A is a bit too small to read when printed. 

We have changed this in the revised manuscript.

- Since WT and DKO lack detectable L1TD1, would you expect any difference in RIP-Seq results between these two?

Due to the lack of DNMT1 and the resulting DNA hypomethylation, DKO cells are more similar to KO cells than WT cells with respect to the expressed transcripts.

- Legend says selected dots are in green (it appears blue to me). 

We have changed this in the revised manuscript.           

- Would you recover L1 ORF1p and its binding partners in the KO? (Is the antibody specific in the absence of L1TD1 or can it recognize L1?) I noticed an increase in ORF1p in the KO in Figure 3C.  

Thank you for the suggestion. Yes, L1 ORF1p shows slightly increased expression in the proteome analysis and we have marked the corresponding dot in the Volcano plot (Figure 3A).

- Should the figure panel reference near the (Rosspopoff & Trono) reference instead be Sup S1C as well? Otherwise, I don't think S1C is mentioned at all. 

- What are the red vs. green dots in 2D? Can you highlight ERV and ALU with different colors? 

We added the reference to Suppl. Figure S1C (now S3C) in the revised manuscript. In Figure 2D L1 elements are highlighted in green, ERV elements in yellow, and other associated transposon transcripts in red.     

- Which L1 subfamily from Figure 2D is represented in the qRT-PCR in 2E "LINE-1"? Do the primers match a specific L1 subfamily? If so, which? 

We used primers specific for the human L1.2 subfamily. 

- Pulling down SINE element transcripts makes some sense, as many insertions "borrow" L1 sequences for non-autonomous retro transposition, but can you speculate as to why ERVs are recovered? There should be essentially no overlap in sequence. 

In the L1TD1 evolution paper [8], a potential link between L1TD1 and ERV elements was discussed: 

"Alternatively, L1TD1 in sigmodonts could play a role in genome defense against another element active in these genomes. Indeed, the sigmodontine rodents have a highly active family of ERVs, the mysTR elements [46]. Expansion of this family preceded the death of L1s, but these elements are very active, with 3500 to 7000 species-specific insertions in the L1-extinct species examined [47]. This recent ERV amplification in Sigmodontinae contrasts with the megabats (where L1TD1 has been lost in many species); there are apparently no highly active DNA or RNA elements in megabats [48]. If L1TD1 can suppress retroelements other than L1s, this could explain why the gene is retained in sigmodontine rodents but not in megabats." 

Furthermore, Jin et al. report the binding of L1TD1 to repetitive sequences in transcripts [12]. It is possible that some of these sequences are also present in ERV RNAs.

- Is S2B a screenshot? (the red underline). 

No, it is a Powerpoint figure, and we have removed the red underline.

(4) Figure 3: 

- Text refers to Figure 3B as a western blot. Figure 3B shows a volcano plot. This is likely 3C but would still be out of order (3A>3C>3B referencing). I think this error is repeated in the last result section. 

- Figure and legends fail to mention what gene was used for ddCT method (actin, gapdh, etc.). 

- In general, the supplemental legends feel underwritten and could benefit from additional explanations. (Main figures are appropriate but please double-check that all statistical tests have been mentioned correctly).

Thank you for pointing this out. We have corrected these errors in the revised manuscript.

(5) Discussion: 

-Aluy connection is interesting. Is there an "Alu retrotransposition reporter assay" to test whether L1TD1 enhances this as well? 

Thank you for the suggestion. There is indeed an Alu retrotransposition reporter assay reported be Dewannieux et al. [13]. The assay is based on a Neo selection marker. We have previously tested a Neo selection-based L1 retrotransposition reporter assay, but this system failed to properly work in HAP1 cells, therefore we switched to a blasticidinbased L1 retrotransposition reporter assay. A corresponding blasticidin-based Alu retrotransposition reporter assay might be interesting for future studies (mentioned in the Discussion, page 11 paragraph 4 of the revised manuscript.

(6) Material and Methods       : 

- The number of typos in the materials and methods is too numerous to list. Instead, please refer to the next section that broadly describes the issues seen throughout the manuscript. 

Writing style  

(1) Keep a consistent style throughout the manuscript: for example, L1 or LINE-1 (also L1 ORF1p or LINE-1 ORF1p); per or "/"; knockout or knock-out; min or minute; 3 times or three times; media or medium. Additionally, as TE naming conventions are not uniform, it is important to maintain internal consistency so as to not accidentally establish an imprecise version. 

(2) There's a period between "et al" and the comma, and "et al." should be italic. 

(3) The authors should explain what the key jargon is when it is first used in the manuscript, such as "retrotransposon" and "retrotransposition".    

(4) The authors should show the full spelling of some acronyms when they use it for the first time, such as RNA Immunoprecipitation (RIP).  

(5) Use a space between numbers and alphabets, such as 5 µg.  

(6) 2.0 × 105 cells, that's not an "x".  

(7) Numbers in the reference section are lacking (hard to parse).  

(8) In general, there are a significant number of typos in this draft which at times becomes distracting. For example, (P3) Introduction: Yet, co-option of TEs thorough (not thorough, it should be through) evolution has created so-called domesticated genes beneficial to the gene network in a wide range of organisms. Please carefully revise the entire manuscript for these minor issues that collectively erode the quality of this submission.  

Thank you for pointing out these mistakes. We have corrected them in the revised manuscript. A native speaker from our research group has carefully checked the paper. In summary, we have added Supplementary Figure S7C and have changed Figures 1C, 1E, 1F, 2A, 2D, 3A, 4B, S3A-D, S4B and S6A based on these comments. 

REFERENCES

(1) Beck, M.A., et al., DNA hypomethylation leads to cGAS-induced autoinflammation in the epidermis. EMBO J, 2021. 40(22): p. e108234.

(2) Altenberger, C., et al., SPAG6 and L1TD1 are transcriptionally regulated by DNA methylation in non-small cell lung cancers. Mol Cancer, 2017. 16(1): p. 1.

(3) Narva, E., et al., RNA-binding protein L1TD1 interacts with LIN28 via RNA and is required for human embryonic stem cell self-renewal and cancer cell proliferation. Stem Cells, 2012. 30(3): p. 452-60.

(4) Jin, S.W., et al., Dissolution of ribonucleoprotein condensates by the embryonic stem cell protein L1TD1. Nucleic Acids Res, 2024. 52(6): p. 3310-3326.

(5) Emani, M.R., et al., The L1TD1 protein interactome reveals the importance of posttranscriptional regulation in human pluripotency. Stem Cell Reports, 2015. 4(3): p. 519-28.

(6) Santos, M.C., et al., Embryonic Stem Cell-Related Protein L1TD1 Is Required for Cell Viability, Neurosphere Formation, and Chemoresistance in Medulloblastoma. Stem Cells Dev, 2015. 24(22): p. 2700-8.

(7) Wong, R.C., et al., L1TD1 is a marker for undifferentiated human embryonic stem cells. PLoS One, 2011. 6(4): p. e19355.

(8) McLaughlin, R.N., Jr., et al., Positive selection and multiple losses of the LINE-1-derived L1TD1 gene in mammals suggest a dual role in genome defense and pluripotency. PLoS Genet, 2014. 10(9): p. e1004531.

(9) Andersson, B.S., et al., Ph-positive chronic myeloid leukemia with near-haploid conversion in vivo and establishment of a continuously growing cell line with similar cytogenetic pattern. Cancer Genet Cytogenet, 1987. 24(2): p. 335-43.

(10) Carette, J.E., et al., Ebola virus entry requires the cholesterol transporter Niemann-Pick C1. Nature, 2011. 477(7364): p. 340-3.

(11) Smits, A.H., et al., Biological plasticity rescues target activity in CRISPR knock outs. Nat Methods, 2019. 16(11): p. 1087-1093.

(12) Jin, S.W., et al., Dissolution of ribonucleoprotein condensates by the embryonic stem cell protein L1TD1. Nucleic Acids Res, 2024.

(13) Dewannieux, M., C. Esnault, and T. Heidmann, LINE-mediated retrotransposition of marked Alu sequences. Nat Genet, 2003. 35(1): p. 41-8.

Author response:

The following is the authors’ response to the previous reviews.

Reviewer #1 (Public Review):

Summary:

In this study, the authors provide a new computational platform called Vermouth to automate topology generation, a crucial step that any biomolecular simulation starts with. Given a wide arrange of chemical structures that need to be simulated, varying qualities of structural models as inputs obtained from various sources, and diverse force fields and molecular dynamics engines employed for simulations, automation of this fundamental step is challenging, especially for complex systems and in case that there is a need to conduct high-throughput simulations in the application of computer-aided drug design (CADD). To overcome this challenge, the authors develop a programing library composed of components that carry out various types of fundamental functionalities that are commonly encountered in topological generation. These components are intended to be general for any type of molecules and not to depend on any specific force field and MD engines. To demonstrate the applicability of this library, the authors employ those components to re-assemble a pipeline called Martinize2 used in topology generation for simulations with a widely used coarse-grained model (CG) MARTINI. This pipeline can fully recapitulate the functionality of its original version Martinize but exhibit greatly enhanced generality, as confirmed by the ability of the pipeline to faithfully generate topologies for two high-complexity benchmarking sets of proteins.

Strengths:

The main strength of this work is the use of concepts and algorithms associated with induced subgraph in graph theory to automate several key but non-trivial steps of topology generation such as the identification of monomer residue units (MRU), the repair of input structures with missing atoms, the mapping of topologies between different resolutions, and the generation of parameters needed for describing interactions between MRUs. In addition, the documentation website provided by the authors is very informative, allowing users to get quickly started with Vermouth.

Weaknesses:

Although the Vermouth library is designed as a general tool for topology generation for molecular simulations, only its applications with MARTINI have been demonstrated in the current study. Thus, the claimed generality of Vermouth remains to be exmained. The authors may consider to point out this in their manuscript.

In order to demonstrate generality of the here proposed concepts for generating topologies for molecular dynamics simulations, we have now implemented and tested a workflow that will produce topologies for the popular CHARMM36 all-atom force field. To facilitate generation of all-atom topologies with Martinize2 a .rtp reader was introduced, which allows users to provide .rtp files that are the native GROMACS topology files for proteins instead of .ff files. These .rtp files exist for all major atomic protein forcefields. In addition, for CHARMM36 we also included modification files, which describe non-standard pH amino acids, histidine tautomers, and end terminal modifications. Thus, the current implementation unlocks all features available at the CG Martini level also for CHARMM36. We note that users must add the modifications files for other all-atom force fields e.g. AMBER.

We have added a new item in the main manuscript (p28) briefly describing this proof-of-concept implementation. However, we like to point out that there are many specialized tools for the various force fields adopted by the respective communities. Thus, an exhaustive discussion on the capabilities of Martinize2 for all-atom force fields seemed out of place.

Reviewer #2 (Public Review):

This work introduces a Vermouth library framework to enhance software development within the Martini community. Specifically, it presents a Vermouth-powered program, Martinize2, for generating coarse-grained structures and topologies from atomistic structures. In addition to introducing the Vermouth library and the Martinize2 program, this paper illustrates how Martinize2 identifies atoms, maps them to the Martini model, generates topology files, and identifies protonation states or post-translational modifications. Compared with the prior version, the authors provide a new figure to show that Martinize2 can be applied to various molecules, such as proteins, cofactors, and lipids. To demonstrate the general application, Martinize2 was used for converting 73% of 87,084 protein structures from the template library, with failed cases primarily blamed on missing coordinates.

I was hoping to see some fundamental changes in the resubmitted version. To my disappointment, the manuscript remains largely unchanged (even the typo I pointed out previously was not fixed). I do not doubt that Martinize2 and Vermouth are useful to the Martini community, and this paper will have some impact. The manuscript is very technical and limited to the Martini community. The scientific insight for the general coarse-grained modeling community is unclear. The goal of the work is ambitious (such as high-throughput simulations and whole-cell modeling), but the results show just a validation of Martinize2. This version does not reverse my previous impression that it is incremental. As I pointed out in my previous review (and no response from the authors), all the issues associated with the Martini model are still there, e.g. the need for ENM. In this shape, I feel this manuscript is suitable for a specialized journal in computational biophysics or stays as part of the GitHub repository.

We apologize for not fixing the typo; it was fixed but unfortunately got reintroduced in the final resubmitted version. We politely disagree that the goal of the work itself is high-throughput simulations and whole-cell modeling, but the Martinize2 tool is certainly an important element in our ambitions to achieve this. Given the broad interest in these goals by the modeling community in general, we believe this work has a much wider impact beyond the (already large) group of Martini users. Addressing limitations of the Martini model itself, which are certainly there, is clearly not the scope of the current work.

Reviewer #3 (Public Review):

The manuscript Kroon et al. described two algorithms, which when combined achieve high throughput automation of "martinizing" protein structures with selected protonation states and post-translational modifications. After the revisions provided by the authors, I recommend minor revision.

The authors have addressed most of my concerns provided previously. Specifically, showcasing the capability of coarse-graining other types of molecules (Figure 7) is a useful addition, especially for the booming field of therapeutic macrocycles. My only additional concern is that to justify Martinize2 and Vermouth as a "high-throughput" method, the speed of these tools needs to be addressed in some form in the manuscript as a guideline to users.

We have added some benchmark timings in the manuscript SI and pointed to the data in the discussion part, which addresses the timing. Martinize2 is certainly slower than martinize version 1 as we already pointed out in the previous versions. However, even for larger proteins (> 2000 residues) we are able to generate topologies in about 60s. As Martinize2 runs on a single core, it can be massively parallelized. Keeping this in mind the topology file generation is likely to take up only a fraction in a high-throughput pipeline compared to the more costly simulations themselves.

Reviewer #3 (Public review):

Summary:

The manuscript Kroon et al. described two algorithms, which when combined achieve high throughput automation of "martinizing" protein structures with selected protonation states and post-translational modifications.

The authors have addressed all of my concerns as provided previously. Specifically, Figure S2 will be a very useful guideline for future improvement (e.g., parallelization) of the code.

Reviewer #2 (Public review):

This work introduces a Vermouth library framework to enhance software development within the Martini community. Specifically, it presents a Vermouth-powered program, Martinize2, for generating coarse-grained structures and topologies from atomistic structures. In addition to introducing the Vermouth library and the Martinize2 program, this paper illustrates how Martinize2 identifies atoms, maps them to the Martini model, generates topology files, and identifies protonation states or post-translational modifications. Compared with the prior version, the authors provide a new figure to show that Martinize2 can be applied to various molecules, such as proteins, cofactors, and lipids. To demonstrate the general application, Martinize2 was used for converting 73% of 87,084 protein structures from the template library, with failed cases primarily blamed on missing coordinates.

I appreciate the changes that the authors made to clarify the novelty. I have no doubt this paper will receive attention and citations.

Reviewer #1 (Public review):

Summary:

In this study, the authors provide a new computational platform called Vermouth to automate topology generation, a crucial step that any biomolecular simulation starts with. Given a wide arrange of chemical structures that need to be simulated, varying qualities of structural models as inputs obtained from various sources, and diverse force fields and molecular dynamics engines employed for simulations, automation of this fundamental step is challenging, especially for complex systems and in case that there is a need to conduct high-throughput simulations in the application of computer-aided drug design (CADD). To overcome this challenge, the authors develop a programing library composed of components that carry out various types of fundamental functionalities that are commonly encountered in topological generation. These components are intended to be general for any type of molecules and not to depend on any specific force field and MD engines. To demonstrate the applicability of this library, the authors employ those components to reassemble a pipeline called Martinize2 used in topology generation for simulations with a widely used coarse-grained model (CG) MARTINI. This pipeline can fully recapitulate the functionality of its original version Martinize but exhibit greatly enhanced generality, as confirmed by the ability of the pipeline to faithfully generate topologies for two high-complexity benchmarking sets of proteins.

Strengths:

The main strength of this work is the use of concepts and algorithms associated with induced subgraph in graph theory to automate several key but non-trivial steps of topology generation such as the identification of monomer residue units (MRU), the repair of input structures with missing atoms, the mapping of topologies between different resolutions, and the generation of parameters needed for describing interactions between MRUs. In addition, the documentation website provided by the authors is very informative, allowing users to get quickly started with Vermouth.

Weaknesses:

Although the Vermouth library can work for different force fields, exhibiting certain generality, its application has been demonstrated only with GROMACS. The extension of the library to other major MD engines could be future directions for improvement but may not be needed for this study.

eLife Assessment

The authors present an important multi-scale computational platform, which aims to automate the workflow for coarse-grained simulations of biomolecules in the framework of the popular MARTINI model. The capability of the platform has been convincingly demonstrated by the application to a large number of proteins as well as macrocycles and polymers. This work will be of interest to both computational biophysicists and chemists.

Author Response:

The following is the authors' response to the original reviews.

Reply to Public Reviews:

Reply to Reviewer #1:

This is a carefully performed and well-documented study to indicate that the FUS protein interacts with the GGGGCC repeat sequence in Drosophila fly models, and the mechanism appears to include modulating the repeat structure and mitigating RAN translation. They suggest FUS, as well as a number of other G-quadruplex binding RNA proteins, are RNA chaperones, meaning they can alter the structure of the expanded repeat sequence to modulate its biological activities.

Response: We would like to thank the reviewer for her/his time for evaluating our manuscript. We are very happy to see the reviewer for highly appreciating our manuscript.

1. Overall this is a nicely done study with nice quantitation. It remains somewhat unclear from the data and discussions in exactly what way the authors mean that FUS is an RNA chaperone: is FUS changing the structure of the repeat or does FUS binding prevent it from folding into alternative in vivo structure?

Response: We appreciate the reviewer’s constructive comments. Indeed, we showed that FUS changes the higher-order structures of GGGGCC [G4C2] repeat RNA in vitro, and that FUS suppresses G4C2 RNA foci formation in vivo. According to the established definition of RNA chaperone, RNA chaperones are proteins changing the structures of misfolded RNAs without ATP use, resulting in the maintenance of proper RNAs folding (Rajkowitsich et al., 2007). Thus, we consider that FUS is classified into RNA chaperone. To clarify these interpretations, we revised the manuscript as follows.

(1) On page 10, line 215-219, the sentence “These results were in good agreement with our previous study on SCA31 showing the suppressive effects of FUS and other RBPs on RNA foci formation of UGGAA repeat RNA as RNA chaperones …” was changed to “These results were in good agreement with … RNA foci formation of UGGAA repeat RNA through altering RNA structures and preventing aggregation of misfolded repeat RNA as RNA chaperones …”.

(2) On page 17, line 363-366, the sentence “FUS directly binds to G4C2 repeat RNA and modulates its G-quadruplex structure, as evident by CD and NMR analyses (Figure 5), suggesting its functional role as an RNA chaperone.” was changed to “FUS directly binds to G4C2 repeat RNA and modulates its G-quadruplex structure as evident by CD and NMR analyses (Figure 5, Figure 5—figure supplement 2), and suppresses RNA foci formation in vivo (Figures 3A and 3B), suggesting its functional role as an RNA chaperone.”

Reply to Reviewer #2:

Fuijino et al. provide interesting data describing the RNA-binding protein, FUS, for its ability to bind the RNA produced from the hexanucleotide repeat expansion of GGGGCC (G4C2). This binding correlates with reductions in the production of toxic dipeptides and reductions in toxic phenotypes seen in (G4C2)30+ expressing Drosophila. Both FUS and G4C2 repeats of >25 are associated with ALS/FTD spectrum disorders. Thus, these data are important for increasing our understanding of potential interactions between multiple disease genes. However, further validation of some aspects of the provided data is needed, especially the expression data.

Response: We would like to thank the reviewer for her/his time for evaluating our manuscript and also for her/his important comments that helped to strengthen our manuscript.

Some points to consider when reading the work:

1. The broadly expressed GMR-GAL4 driver leads to variable tissue loss in different genotypes, potentially confounding downstream analyses dependent on viable tissue/mRNA levels.

Response: We thank the reviewer for this constructive comment. In the RT-qPCR experiments (Figures 1E, 3C, 4G, 6D and Figure 1—figure supplement 1C), the amounts of G4C2 repeat transcripts were normalized to those of gal4 transcripts expressed in the same tissue, to avoid potential confounding derived from the difference in tissue viability between genotypes, as the reviewer pointed out. To clarify this process, we have made the following change to the revised manuscript.

(1) On page 30, line 548-550, the sentence “The amounts of G4C2 repeat transcripts were normalized to those of gal4 transcripts in the same sample” was changed to “The amounts of G4C2 repeat transcripts were normalized to those of gal4 transcripts expressed in the same tissue to avoid potential confounding derived from the difference in tissue viability between genotypes”.

2. The relationship between FUS and foci formation is unclear and should be interpreted carefully.

Response: We appreciate the reviewer’s important comment. We apologize for the lack of clarity. We showed the relationship between FUS and RNA foci formation in our C9-ALS/FTD fly, that is, FUS suppresses RNA foci formation (Figures 3A and 3B), and knockdown of endogenous caz, a Drosophila homologue of FUS, enhanced it conversely (Figures 4E and 4F). We consider that FUS suppresses RNA foci formation through altering RNA structures and preventing aggregation of misfolded G4C2 repeat RNA as an RNA chaperone. To clarify these interpretations, we revised the manuscript as follows.

(1) On page 10, line 215-219, the sentence “These results were in good agreement with our previous study on SCA31 showing the suppressive effects of FUS and other RBPs on RNA foci formation of UGGAA repeat RNA as RNA chaperones …” was changed to “These results were in good agreement with … RNA foci formation of UGGAA repeat RNA through altering RNA structures and preventing aggregation of misfolded repeat RNA as RNA chaperones …”.

(2) On page 17, line 363-366, the sentence “FUS directly binds to G4C2 repeat RNA and modulates its G-quadruplex structure, as evident by CD and NMR analyses (Figure 5), suggesting its functional role as an RNA chaperone.” was changed to “FUS directly binds to G4C2 repeat RNA and modulates its G-quadruplex structure as evident by CD and NMR analyses (Figure 5, Figure 5—figure supplement 2), and suppresses RNA foci formation in vivo (Figures 3A and 3B), suggesting its functional role as an RNA chaperone.”

Reply to Reviewer #3:

In this manuscript Fujino and colleagues used C9-ALS/FTD fly models to demonstrate that FUS modulates the structure of (G4C2) repeat RNA as an RNA chaperone, and regulates RAN translation, resulting in the suppression of neurodegeneration in C9-ALS/FTD. They also confirmed that FUS preferentially binds to and modulates the G-quadruplex structure of (G4C2) repeat RNA, followed by the suppression of RAN translation. The potential significance of these findings is high since C9ORF72 repeat expansion is the most common genetic cause of ALS/FTD, especially in Caucasian populations and the DPR proteins have been considered the major cause of the neurodegenerations.

Response: We would like to thank the reviewer for her/his time for evaluating our manuscript. We are grateful to the reviewer for the insightful comments, which were very helpful for us to improve the manuscript.

1. While the effect of RBP as an RNA chaperone on (G4C2) repeat expansion is supposed to be dose-dependent according to (G4C2)n RNA expression, the first experiment of the screening for RBPs in C9-ALS/FTD flies lacks this concept. It is uncertain if the RBPs of the groups "suppression (weak)" and "no effect" were less or no ability of RNA chaperone or if the expression of the RBP was not sufficient, and if the RBPs of the group "enhancement" exacerbated the toxicity derived from (G4C2)89 RNA or the expression of the RBP was excessive. The optimal dose of any RBPs that bind to (G4C2) repeats may be able to neutralize the toxicity without the reduction of (G4C2)n RNA.

Response: We appreciate the reviewer’s constructive comments. We employed the site-directed transgenesis for the establishment of RBP fly lines, to ensure the equivalent expression levels of the inserted transgenes. We also evaluated the toxic effects of overexpressed RBPs themselves by crossbreeding with control EGFP flies, showing in Figure 1A. To clarify them, we have made the following changes to the revised manuscript.

(1) On page 8, line 166-168, the sentence “The variation in the effects of these G4C2 repeat-binding RBPs on G4C2 repeat-induced toxicity may be due to their different binding affinities to G4C2 repeat RNA, and their different roles in RNA metabolism.” was changed to “The variation in the effects of these G4C2 repeat-binding RBPs on G4C2 repeat-induced toxicity may be due to their different binding affinities to G4C2 repeat RNA, and the different toxicity of overexpressed RBPs themselves.”.

(2) On page 29, line 519-522, the sentence “By employing site-specific transgenesis using the pUASTattB vector, each transgene was inserted into the same locus of the genome, and was expected to be expressed at the equivalent levels.” was added.

2. In relation to issue 1, the rescue effect of FUS on the fly expressing (G4C2)89 (FUS-4) in Figure 4-figure supplement 1 seems weaker than the other flies expressing both FUS and (G4C2)89 in Figure 1 and Figure 1-figure supplement 2. The expression level of both FUS protein and (G4C2)89 RNA in each line is important from the viewpoint of therapeutic strategy for C9-ALS/FTD.

Response: We appreciate the reviewer’s important comment. The FUS-4 transgene is expected to be expressed at the equivalent level to the FUS-3 transgene, since they are inserted into the same locus of the genome by the site-directed transgenesis. Thus, we suppose that the weaker suppressive effect of FUS-4 coexpression on G4C2 repeat-induced eye degeneration can be attributed to the C-terminal FLAG tag that is fused to FUS protein expressed in FUS-4 fly line. Since the caz fly expresses caz protein also fused to FLAG tag at the C-terminus, we used this FUS-4 fly line to directly compare the effect of caz on G4C2 repeat-induced toxicity to that of FUS.

3. While hallmarks of C9ORF72 are the presence of DPRs and the repeat-containing RNA foci, the loss of function of C9ORF72 is also considered to somehow contribute to neurodegeneration. It is unclear if FUS reduces not only the DPRs but also the protein expression of C9ORF72 itself.

Response: We thank the reviewer for this comment. We agree that not only DPRs, but also toxic repeat RNA and the loss-of-function of C9ORF72 jointly contribute to the pathomechanisms of C9-ALS/FTD. Since Drosophila has no homolog corresponding to the human C9orf72 gene, the effect of FUS on C9orf72 expression cannot be assessed. Our fly models are useful for evaluating gain-of-toxic pathomechanisms such as RNA foci formation and RAN translation, and the association between FUS and loss-of function of C9ORF72 is beyond the scope of this study.

4. In Figure 5E-F, it cannot be distinguished whether FUS binds to GGGGCC repeats or the 5' flanking region. The same experiment should be done by using FUS-RRMmut to elucidate whether FUS binding is the major mechanism for this translational control. Authors should show that FUS binding to long GGGGCC repeats is important for RAN translation.

Response: We would like to thank the reviewer for these insightful comments. Following the reviewer’s suggestion, we perform in vitro translation assay again using FUS-RRMmut, which loses the binding ability to G4C2 repeat RNA as evident by the filter binding assay (Figure 5A), instead of BSA. The results are shown in the figures of Western blot analysis below. The addition of FUS to the translation system suppressed the expression levels of GA-Myc efficiently, whereas that of FUS-RRMmut did not. FUS decreased the expression level of GA-Myc at as low as 10nM, and nearly eliminated RAN translation activity at 100nM. At 400nM, FUS-RRMmut weakly suppressed the GA-Myc expression levels probably because of the residual RNA-binding activity. These results suggest that FUS suppresses RAN translation in vitro through direct interactions with G4C2 repeat RNA.

Unfortunately, RAN translation from short G4C2 repeat RNA was not investigated in our translation system, although the previous study reported the low efficacy of RAN translation from short G4C2 repeat RNA (Green et al., 2017).

Author response image 1.

(A) Western blot analysis of the GA-Myc protein in the samples from in vitro translation. (B) Quantification of the GA-Myc protein levels.

We have made the following changes to the revised manuscript.

(1) Figure 5F was replaced to new Figures 5F and 5G.

(2) On page 14-15, line 326-330, the sentence “Notably, the addition of FUS to this system decreased the expression level of GA-Myc in a dose-dependent manner, whereas the addition of the control bovine serum albumin (BSA) did not (Figure 5F).” was changed to “Notably, upon the addition to this translation system, FUS suppressed RAN translation efficiently, whereas FUS-RRMmut did not. FUS decreased the expression levels of GA-Myc at as low as 10nM, and nearly eliminated RAN translation activity at 100nM. At 400nM, FUS-RRMmut weakly suppressed the GA-Myc expression levels probably because of the residual RNA-binding activity (Figure 5F and 5G).”.

(3) On page 15, line 330-332, the sentence “Taken together, these results indicate that FUS suppresses RAN translation from G4C2 repeat RNA in vitro as an RNA chaperone.” was changed to “Taken together, these results indicate that FUS suppresses RAN translation in vitro through direct interactions with G4C2 repeat RNA as an RNA chaperone.”.

(4) On page 37, line 720-723, the sentence “For preparation of the FUS protein, the human FUS (WT) gene flanked at the 5¢ end with an Nde_I recognition site and at the 3¢ end with a _Xho_I recognition site was amplified by PCR from pUAST-_FUS.” was changed to “For preparation of the FUS proteins, the human FUS (WT) and FUS-RRMmut genes flanked at the 5¢ end with an Nde_I recognition site and at the 3¢ end with a _Xho_I recognition site was amplified by PCR from pUAST-_FUS and pUAST- FUS-RRMmut, respectively.”.

(5) On page 41, line 816-819, the sentence “FUS or BSA at each concentration (10, 100, and 1,000 nM) was added for translation in the lysate.” was changed to “FUS or FUS-RRMmut at each concentration (10, 100, 200, 400, and 1,000 nM) was preincubated with mRNA for 10 min to facilitate the interaction between FUS protein and G4C2 repeat RNA, and added for translation in the lysate.”.

5. It is not possible to conclude, as the authors have, that G-quadruplex-targeting RBPs are generally important for RAN translation (Figure 6), without showing whether RBPs that do not affect (G4C2)89 RNA levels lead to decreased DPR protein level or RNA foci.

Response: We appreciate the reviewer’s critical comment. Following the suggestion by the reviewer, we evaluate the effect of these G-quadruplex-targeting RBPs on RAN translation. We additionally performed immunohistochemistry of the eye imaginal discs of fly larvae expressing (G4C2)89 and these G-quadruplex-targeting RBPs. As shown in the figures of immunohistochemistry below, we found that coexpression of EWSR1, DDX3X, DDX5, and DDX17 significantly decreased the number of poly(GA) aggregates. The results suggest that these G-quadruplex-targeting RBPs regulate RAN translation as well as FUS.

Author response image 2.

(A) Immunohistochemistry of poly(GA) in the eye imaginal discs of fly larvae expressing (G4C2)89 and the indicated G-quadruplex-targeting RBPs. (B) Quantification of the number of poly(GA) aggregates.

We have made the following changes to the revised manuscript.

(1) Figures 6E and 6F were added.

(2) On page 6-7, line 135-137, the sentence “In addition, other G-quadruplex-targeting RBPs also suppressed G4C2 repeat-induced toxicity in our C9-ALS/FTD flies.” was changed to “In addition, other G-quadruplex-targeting RBPs also suppressed RAN translation and G4C2 repeat-induced toxicity in our C9-ALS/FTD flies.”.

(3) On page 15, line 344-346, the sentence “As expected, these RBPs also decreased the number of poly(GA) aggregates in the eye imaginal discs (Figures 6E and 6F).” was added.

(4) On page 15, line 346-347, the sentence “Their effects on G4C2 repeat-induced toxicity and repeat RNA expression were consistent with those of FUS.” was changed to “Their effects on G4C2 repeat-induced toxicity, repeat RNA expression, and RAN translation were consistent with those of FUS.”

(5) On page 16, line 355-357, the sentence “Thus, some G-quadruplex-targeting RBPs regulate G4C2 repeat-induced toxicity by binding to and possibly by modulating the G-quadruplex structure of G4C2 repeat RNA.” was changed to “Thus, some G-quadruplex-targeting RBPs regulate RAN translation and G4C2 repeat-induced toxicity by binding to and possibly by modulating the G-quadruplex structure of G4C2 repeat RNA.”

(6) On page 19, line 417-421, the sentence “We further found that G-quadruplex-targeting RNA helicases, including DDX3X, DDX5, and DDX17, which are known to bind to G4C2 repeat RNA (Cooper-Knock et al., 2014; Haeusler et al., 2014; Mori et al., 2013a; Xu et al., 2013), also alleviate G4C2 repeat-induced toxicity without altering the expression levels of G4C2 repeat RNA in our Drosophila models.” was changed to “We further found that G-quadruplex-targeting RNA helicases, … ,also suppress RAN translation and G4C2 repeat-induced toxicity without altering the expression levels of G4C2 repeat RNA in our Drosophila models.”.

Reply to Recommendations For The Authors:

1) It is not clear from the start that the flies they generated with the repeat have an artificial vs human intronic sequence ahead of the repeat. It would be nice if they presented somewhere the entire sequence of the insert. The reason being that it seems they also tested flies with the human intronic sequence, and the effect may not be as strong (line 234). In any case, in the future, with a new understanding of RAN translation, it would be nice to compare different transgenes, and so as much transparency as possible would be helpful regarding sequences. Can they include these data?

Response: We thank the editors and reviewers for this comment. We apologize for the lack of clarity. We used artificially synthesized G4C2 repeat sequences when generating constructs for (G4C2)n transgenic flies, so these constructs do not contain human intronic sequence ahead of the G4C2 repeat in the C9orf72 gene, as explained in the Materials and Methods section. To clarify the difference between our C9-ALS/FTD fly models and LDS-(G4C2)44GR-GFP fly model (Goodman et al., 2019), we have made the following change to the revised manuscript.

(1) Schema of the LDS-(G4C2)44GR-GFP construct was presented in Figure 3—figure supplement 1.

Furthermore, to maintain transparency of the study, we have provided the entire sequence of the insert as the following source file.

(2) The artificial sequences inserted in the pUAST vector for generation of the (G4C2)n flies were presented in Figure 1—figure supplement 1—source data 1.

2) It is really nice how they quantitated everything and showed individual data points.

Response: We thank the editors and reviewers for appreciating our data analysis method. All individual data points and statistical analyses are summarized in source data files.

3) So when they call FUS an RNA chaperone, are they simply meaning it is changing the structure of the repeat, or could it just be interacting with the repeat to coat the repeat and prevent it from folding into whatever in vivo structures? Can they speculate on why some RNA chaperones lead to presumed decay of the repeat and others do not? Can they discuss these points in the discussion? Detailed mechanistic understanding of RNA chaperones that ultimately promote decay of the repeat might be of highly significant therapeutic benefit.

Response: We appreciate these critical comments. Indeed, we showed that FUS changes the higher-order structures of G4C2 repeat RNA in vitro, and that FUS suppresses G4C2 RNA foci formation. According to the established definition of RNA chaperone, RNA chaperones are proteins changing the structures of misfolded RNAs without ATP use, resulting in the maintenance of proper RNAs folding (Rajkowitsich et al., 2007). Thus, we consider that FUS is classified into RNA chaperone. To clarify these interpretations, we revised the manuscript as follows.

(1) On page 10, line 215-219, the sentence “These results were in good agreement with our previous study on SCA31 showing the suppressive effects of FUS and other RBPs on RNA foci formation of UGGAA repeat RNA as RNA chaperones …” was changed to “These results were in good agreement with … RNA foci formation of UGGAA repeat RNA through altering RNA structures and preventing aggregation of misfolded repeat RNA as RNA chaperones …”.

(2) On page 17, line 363-366, the sentence “FUS directly binds to G4C2 repeat RNA and modulates its G-quadruplex structure, as evident by CD and NMR analyses (Figure 5), suggesting its functional role as an RNA chaperone.” was changed to “FUS directly binds to G4C2 repeat RNA and modulates its G-quadruplex structure as evident by CD and NMR analyses (Figure 5, Figure 5—figure supplement 2), and suppresses RNA foci formation in vivo (Figures 3A and 3B), suggesting its functional role as an RNA chaperone.”

Besides these RNA chaperones, we observed the expression of IGF2BP1, hnRNPA2B1, DHX9, and DHX36 decreased G4C2 repeat RNA expression levels. In addition, we recently reported that hnRNPA3 reduces G4C2 repeat RNA expression levels, leading to the suppression of neurodegeneration in C9-ALS/FTD fly models (Taminato et al., 2023). We speculate these RBPs could be involved in RNA decay pathways as components of the P-body or interactors with the RNA deadenylation machinery (Tran et al., 2004; Katahira et al., 2008; Geissler et al., 2016; Hubstenberger et al., 2017), possibly contributing to the reduced expression levels of G4C2 repeat RNA. To clarify these interpretations, we revised the manuscript as follows.

(3) On page 18, line 392-398, the sentences “Similarly, we recently reported that hnRNPA3 reduces G4C2 repeat RNA expression levels, leading to the suppression of neurodegeneration in C9-ALS/FTD fly models (Taminato et al., 2023). Interestingly, these RBPs have been reported to be involved in RNA decay pathways as components of the P-body or interactors with the RNA deadenylation machinery (Tran et al., 2004; Katahira et al., 2008; Geissler et al., 2016; Hubstenberger et al., 2017), possibly contributing to the reduced expression levels of G4C2 repeat RNA.” was added.

4) What is the level of the G4C2 repeat when they knock down caz? Is it possible that knockdown impacts the expression level of the repeat? Can they show this (or did they and I miss it)?

Response: We thank the editors and reviewers for this comment. The expression levels of G4C2 repeat RNA in (G4C2)89 flies were not altered by the knockdown of caz, as shown in Figure 4G.

5) A puzzling point is that FUS is supposed to be nuclear, so where is FUS in the brain in their lines? They suggest it modulates RAN translation, and presumably, that is in the cytoplasm. Is FUS when overexpressed now in part in the cytoplasm? Is the repeat dragging it into the cytoplasm? Can they address this in the discussion? If FUS is never found in vivo in the cytoplasm, then it raises the point that the impact they find of FUS on RAN translation might not reflect an in vivo situation with normal levels of FUS.

Response: We appreciate these important comments. We agree with the editors and reviewers that FUS is mainly localized in the nucleus. However, FUS is known as a nucleocytoplasmic shuttling RBP that can transport RNA into the cytoplasm. Indeed, FUS is reported to facilitate transport of actin-stabilizing protein mRNAs to function in the cytoplasm (Fujii et al., 2005). Thus, we consider that FUS binds to G4C2 repeat RNA in the cytoplasm and suppresses RAN translation in this study.

6) When they are using 2 copies of the driver and repeat, are they also using 2 copies of FUS? These are quite high levels of transgenes.

Response: We thank the editors and reviewers for this comment. We used only 1 copy of FUS when using 2 copies of GMR-Gal4 driver. Full genotypes of the fly lines used in all experiments are described in Supplementary file 1.

7) In Figure5-S1, FUS colocalizing with (G4C2)RNA is not clear. High-magnification images are recommended.

Response: We appreciate this constructive comment on the figure. Following the suggestion, high-magnification images are added in Figure 5—figure supplement 1.

8) I also suggest that the last sentence of the Discussion be revised as follows: Thus, our findings contribute not only to the elucidation of C9-ALS/FTD, but also to the elucidation of the repeat-associated pathogenic mechanisms underlying a broader range of neurodegenerative and neuropsychiatric disorders than previously thought, and it will advance the development of potential therapies for these diseases.

Response: We appreciate this recommendation. We have made the following change based on the suggested sentence.

(1) On page 20-21, line 455-459, “Thus, our findings contribute not only towards the elucidation of repeat-associated pathogenic mechanisms underlying a wider range of neuropsychiatric diseases than previously thought, but also towards the development of potential therapies for these diseases.” was changed to “Thus, our findings contribute to the elucidation of the repeat-associated pathogenic mechanisms underlying not only C9-ALS/FTD, but also a broader range of neuromuscular and neuropsychiatric diseases than previously thought, and will advance the development of potential therapies for these diseases.”.

Authors’ comment on previous eLife assessment:

We thank the editors and reviewers for appreciating our study. We mainly evaluated the function of human FUS protein on RAN translation and G4C2 repeat-induced toxicity using Drosophila expressing human FUS in vivo, and the recombinant human FUS protein in vitro. To validate that FUS functions as an endogenous regulator of RAN translation, we additionally evaluated the function of Drosophila caz protein as well. We are afraid that the first sentence of the eLife assessment, that is, “This important study demonstrates that the Drosophila FUS protein, the human homolog of which is implicated in amyotrophic lateral sclerosis (ALS) and related conditions, …” is somewhat misleading. We would be happy if you modify this sentence like “This important study demonstrates that the human FUS protein, which is implicated in amyotrophic lateral sclerosis (ALS) and related conditions, …”.

Author Response:

The following is the authors' response to the original reviews.

Reviewer #1 (Public Review):

The authors investigated state-dependent changes in evoked brain activity, using electrical stimulation combined with multisite neural activity across wakefulness and anesthesia. The approach is novel, and the results are compelling. The study benefits from an in-depth sophisticated analysis of neural signals. The effects of behavioral state on brain responses to stimulation are generally convincing.

It is possible that the authors' use of "an average reference montage that removed signals common to all EEG electrodes" could also remove useful components of the signal, which are common across EEG electrodes, especially during deep anesthesia. For example, it is possible (in fact from my experience I would be surprised if it is not the case) that under isoflurane anesthesia, electrical stimulation induces a generalized slow wave or a burst of activity across the brain. Subtracting the average signal will simply remove that from all channels. This does not only result in signals under anesthesia being affected more by the referencing procedure than during waking but also will have different effects on different channels, e.g. depending on how strong the response is in a specific channel.

We thank the reviewer for the positive comments and for raising this point. We do not believe that the average reference montage is obscuring an evoked slow wave in the isoflurane-anesthetized mice. Electrical stimulation did elicit a brief activation in nearby neurons that was followed by roughly 200 ms of quiescence, but no significant changes in firing in the other regions we recorded from (Author response image 1).

Author response image 1

ERP and evoked population activity during isoflurane anesthesia do not show evidence of global responses. (Top). ERP (-0.2 to +0.8 s around stimulus onset) with all EEG electrode traces superimposed. Data represented is the same: red traces have been processed with the average reference montage, black traces have not. (Bottom) Population mean firing rates from the areas of interest from the same experiment as above.

We are familiar with the work from Dasilva et al. (2021), a study similar to ours because they also performed cortical electrical stimulation in mice anesthetized with isoflurane. They show widespread evoked multi-unit activity (derived from LFP) in isoflurane-anesthetized mice in response to electrical stimulation, but critical experimental differences may underlie the conflicting results presented in our study. Both works use similar levels of isoflurane to maintain anesthesia (we use a level roughly equivalent to their “deep” level). However, our experiments use only isoflurane, whereas Dasilva et al. induced anesthesia with ketamine and medetomidine followed by isoflurane. It has been shown that isoflurane and ketamine have different effects on neural dynamics (Sorrenti et al., 2021). Typically, isoflurane causes reduced spontaneous firing rates and decreased evoked response amplitudes compared to wakefulness, whereas ketamine has been shown to increase firing rates and evoked response amplitudes (Aasebø et al., 2017; Michelson & Kozai, 2018). Perhaps a more relevant difference are the electrical stimulation parameters used to perturb the brain. Dasilva et al. used 1 ms pulses of 500 μA, which would have a much larger effect than the stimulation used in this work, 0.2 ms pulses of 10-100 μA.

Additionally, we would like to clarify that the average reference montage is not impacting the main findings of this work. As the reviewer correctly pointed out, the average reference montage does change the appearance of the ERP in the butterfly plots (Top panel in Author response image 1). However, all the quantitative analyses of the EEG-ERPs are performed on the global field power, computed by taking the standard deviation across all EEG channels, which is not affected by the average reference montage.

Reviewer #2 (Public Review):

[…] The conclusions regarding the thalamic contributions to the ERP components are strongly supported by the data.

The spatiotemporal complexity is almost a side point compared to what seems to be the most important point of the paper: showing the contribution of thalamic activity to some components of the cortical ERP. Scalp ERPs have long been regarded as purely cortical phenomena, just like most EEGs, and this study shows convincing evidence to the contrary.

The data presented seemingly contradicts the results presented by Histed et al. (2009), who assert that cortical microstimulation only affects passing fibers near the tip of the electrodes, and results in distant, sparse, and somewhat random neural activation. In this study, it is clear that the maximum effect happens near the electrodes, decays with distance, and is not sparse at all, suggesting that not only passing fibers are activated but that also neuronal elements might be activated by antidromic propagation from the axonal hillock. This appears to offer proof that microstimulation might be much more effective than it was thought after the publication of Histed 2009, as the uber-successful use of DBS to treat Parkinson's disease has also shown.

We thank the reviewer for their positive comments and thoughtful suggestions. We appreciate and agree with the reviewer’s perspective that the thalamic contribution to the cortical ERP is one of the key points of this study. We also thank the reviewer for their comment on the apparently contradictory results reported by Histed et al. (2009). This gives us the opportunity to further highlight the important contribution of our study to the field.

First, we would like to highlight some key experimental differences between the two studies. In our study we used single pulse stimulation with currents between 10 and 100 μA, whereas Histed et al. used trains of pulses (100 ms in duration at 250 Hz) with lower current intensities (between 2 and 50 μA). We varied the depth of stimulation, targeting superficial and deep cortical layers; Histed et al. exclusively stimulated superficial cortical layers. In addition, the two studies used recording methods that are orthogonal in nature. We used Neuropixels probes that record from neurons that span all cortical layers depth-wise while Histed et al. used two-photon calcium imaging to record from a horizontal plane of neurons (again, in the superficial cortical layers).

Because of these important methodological differences, it is more appropriate to compare the Histed et al. results to our results from superficial stimulation at comparable current intensities. In this case, we believe the two studies show similar results: stimulation activated a small fraction of neurons even hundreds of microns away from the stimulating electrode (see Figure 4A from our manuscript). However, our study adds an important observation pointing to the critical role of the depth of the stimulating electrode. We observe significant excitation of local cortical neurons (Figure 4D) and trans-synaptic activation of the thalamus only when we delivered deep stimulation (Figure5A). This effect is likely mediated by activation of large, myelinated cortico-thalamic fibers, which are thought to be more excitable that non-myelinated horizontal fibers (Tehovnik & Slocum, 2013).

To summarize, Histed et al. (2009) concluded that microstimulation causes a sparse activation of a distributed set of neurons with little evidence of synaptically driven activation. Instead, we showed that microstimulation can robustly activate local neurons and trans-synaptically activate distant neurons when stronger stimuli are directed to deep cortical layers. Based on this, we conclude that electrical stimulation is indeed highly effective, and is a valid tool that can be used to probe and characterize the cortico-thalamo-cortical network of any behavioral state.

----------

Reviewer #1 (Recommendations for the authors):

1. I am not clear how "putative pyramidal" or RS and "putative inhibitory" fast-spiking neurons were identified. Please provide some further details on that, including average spike wave shapes, and distribution of firing rates, and it would be interesting to know the proportion of "putative" RS and FS neurons in your recorded population. Obviously, caution is warranted here because, without further work, you cannot be sure that those are indeed pyramidal cells or interneurons! Is this subdivision necessary at all?

We added details regarding the cell-type classification to the Results (lines 136-140) and the Methods section. This classification is common practice in cortical extracellular electrophysiology recordings given that cell-type specific analyses can reveal important differences between the two putative populations (Barthó et al., 2004; Bortone et al., 2014; Bruno & Simons, 2002; Jia et al., 2016; Niell & Stryker, 2008; Sirota et al., 2008). Based on our findings that the two populations respond to electrical stimulation in similar ways (excitation followed by a period of quiescence and rebound excitation), we agree the subdivision is not necessary to support our conclusions. However, we believe that some readers will appreciate seeing the two putative populations presented separately.

2. I wonder how the authors know whether the animals were awake, specifically when they were not running. Did you observe animals falling asleep when head-fixed? Providing some analyses of spontaneous EEG/LFP signals in each state could add some reassurance that only wakefulness was included, as intended.

While we cannot conclusively rule out that mice were asleep during the “quiet wakefulness” periods we analyzed, we believe they are likely to be awake for two main reasons: 1) all the experiments are performed during the dark phase of the light/dark cycle, when the mice are less likely to enter a sleep state (Franken et al., 1999); 2) the animals are not undergoing specific training to promote drowsiness or sleep. Indeed, many sleep-focused studies in head-fixed mice are performed during the light phase of the animal’s cycle to maximize the likelihood of capturing sleep states (Kobayashi et al., 2023; Turner et al., 2020; Yüzgeç et al., 2018; Zhang et al., 2022). We have added this note to the Discussion section (lines 402-406).

Because we do not specifically record during sleep states and our recording does not include electromyography, which is commonly used in conjunction with EEG to classify sleep stages, we cannot accurately perform spectral comparison between “quiet wakefulness” and sleep states in our recordings.

3. I was unsure about the meaning of some of the terminology, specifically "rebound", "rebound spiking", "rebound excitation" etc. Why do you call it "rebound"?

“Rebound” is a term often used to describe a period of enhanced spiking following a period of prolonged silence or inhibition (Guido & Weyand, 1995; Roux et al., 2014). Grenier et al. list “postinhibitory rebound excitation” as an intrinsic property of cortical and thalamic neurons (1998). We added this description to the text (lines 79-80).

Reviewer #2 (Recommendations For The Authors):

Regarding analysis, I would make three main points:

Regarding the CSD analysis, I think the authors have done a good job of circumventing several of the known issues of this technique, especially by using ERPs rather than ongoing activity. However, although I do not immediately have access to the literature to back up this claim, I've heard that many assumptions behind CSD require a laminar structure with electrodes positioned perpendicular to these layers. In Figure 1B it seems like the neuropixels probe is not really perpendicular to the cortical layers, and I wonder if this might be an issue. I am also wondering how to interpret the thalamic CSD, as this structure is not laminar, lacks the mass of neatly stacked neuronal dipoles present in the cortex, and does not have an orderly array of synaptic inputs and outputs. I understand that CSD analysis helps minimize the contributions of volume conduction, but in this case, I also wonder if the thalamic CSD is even necessary to back up the paper's claims.

One-dimensional CSD is computed assuming that the electrode is inserted perpendicular to cortex. This is mainly important for the interpretation of sinks and sources, since CSD can be also computed on radial voltages (e.g., EEG [Tenke & Kayser, 2012]). In general, our Neuropixels probes do not significantly deviate from perpendicular (mean deviation from perpendicular 15.3 degrees, minimum 5.2 degrees, and maximum 36.6 degrees). The probe represented in Figure 1B deviates from perpendicular by 31.2 degrees, which is an outlier compared to the rest of the insertions. Any deviation from perpendicular would result in the “effective” cortical thickness being larger by a factor of 1/cos(angle deviation from perpendicular) and thus would not affect the relative location of sources and sinks. We have added a statement to clarify this in the text (lines 126 and 454-456).

We agree with the statement regarding CSD analysis in the thalamus. We originally included the CSD for the thalamus in Figure 2F for completeness. As the reviewer pointed out, thalamic CSD was not used to perform any subsequent analysis and is, therefore, not necessary to back up any claims. As such, we have removed CSD plot from Figure 2F to avoid any confusion and made a comment to this effect in the legend (lines 1175-1177).

On the merits of using the z-score normalization for spike rates vs. other strategies like standardizing to maximum firing, I am aware that both procedures have limitations, but the z-score changes the range of the firing rate from [0, +Inf] to [-Inf, +Inf]. This does not seem correct considering that negative spiking rates do not exist. The standardization to maximum rate keeps the range within [0, 1], not creating negative rates. Another point that it will be worth discussing is the reported values of the z-scored values. For example, what does it mean to be 54 standard deviations away from the mean? 6 standard deviations is already a big distance from the mean.

For Figure 2, we chose to represent the neural firing rates as z-scores because we found it important to report the magnitude of both the increase and decrease of the evoked firing rates in the post-stimulus period relative to the pre-stimulus rate. The normalization we used helps to visualize the magnitude of the effects of electrical stimulation in neuronal activity for both directions, which is an important result of the study. Despite the differences between the two normalization methods, the normalization based on the maximum firing does not significantly change the qualitative interpretation of Figure 2 in the manuscript (Author response image 2).

Author response image 2

Evoked firing rates for neurons in the areas of interest in response to deep stimulation in MO during the awake state. (Left) Firing rates of all neurons normalized by the average, pre-stimulus firing rate. (Right) Firing rates of all neurons normalized by the maximum post-stimulus firing rate.

Regarding Figure 3 and the associated text, we would like to clarify that the magnitude metric is not simply a z-score value (with units of s.d.) but rather it is the integrated area under the z-scored response over the response window (with units of s.d.∙seconds). This can help explain why we see values of ~50 s.d.∙s. We chose to z-score firing rates, LFP, and CSD to normalize across the different signals and magnitudes of the evoked responses. We often observed the largest responses in the LFP (see Figure 3A), which may be partly due to the signal naturally having a larger dynamic range than the measured neural firing rates. Then we integrated the z-score response time series to capture the dynamic of the signal over the response window, rather than a static value such as the mean or maximum z-score. After performing a thorough literature search, we found no other ways to capture and compare the magnitudes of the different signals. We have added language to clarify the magnitude metric (lines 155-156) and added the appropriate units.

In reporting the p-values, I recommend increasing the number of significant digits to four because the p-value seems to be the same for different tests in several places (e.g.: lines 207 to 218), which seems odd. I also wonder whether this could be an artifact of the z-scoring procedure. In the figures, I would like to advise the use of 1 asterisk to denote "weak evidence to reject the null hypothesis (0.05 > p > 0.01)" and two asterisks to denote "strong evidence to reject the null hypothesis (0.01 > p)", and make a note of it accordingly in the manuscript and/or figure legends.

According to the reviewer’s suggestion, we have changed the statistics language to “* weak evidence to reject null hypothesis (0.05 > p > 0.01), ** strong evidence to reject null hypothesis (0.01 > p > 0.001), *** very strong evidence to reject null hypothesis (0.001 > p)” throughout the manuscript.

We have also increased the number of significant digits to four throughout the manuscript. It is true that some of the p-values reported for Figure 3 (lines 169-180) are the same for different tests. This is not an artifact of the z-scoring, but rather a consequence of performing the Wilcoxon signed-rank test (an ordinal statistical test) with small sample numbers. Because the p-value depends only on the relative ordering, not the continuous distribution of values, the small sample size (N=6-14) increases the likelihood of obtaining the exact same p-value if the relative ordering of samples is the same.

Line 202: If the magnitude corresponds to z-score data, please add "s.d." after the number, as z-scored values are expressed in standard deviation units. Please update this throughout the paper.

As stated above the magnitude metric is the integrated area under the z-scored response over the response window (with units of s.d.∙seconds). We have added the correct units in all places.

Line 214: Please report how the multiple comparisons correction was performed

We have added the test used for multiple comparisons in line 169 (formerly line 214) and in the Methods section (line 770).

Line 462: please replace "Neuropixels activity" with "LFP and single-unit activity".

We changed the wording to specify “LFP, and single neuron responses…” (now line 337).

Line 475: a short explanation of the bi-stability phenomena will be helpful for the reader.

We added the following description: “a state characterized by spontaneous alternation between bouts of activity and periods of silence” (lines 350-351).

Line 601: It is asserted that "Electrical stimulation directly activates local cells and axons that run near the stimulation site via activation of the axon initial segment" and the paper by Histed et al. 2009 is cited. This does not seem like an appropriate citation, as Histed et al. explicitly state that electrical microstimulation does not activate local neuronal bodies near the electrode tip. See my comment above.

Upon further reading, we believe we are seeing evidence of direct axonal activation and subsequent antidromic activation of local cell bodies, as you suggested in your above comment and has been proposed by many including Histed et al. (2009) and Nowak and Bullier (1998). We edited our sentence accordingly, kept the Histed et al. citation, and added other relevant citations (lines 487-490).

References

• Aasebø, I. E. J., Lepperød, M. E., Stavrinou, M., Nøkkevangen, S., Einevoll, G., Hafting, T., & Fyhn, M. (2017). Temporal Processing in the Visual Cortex of the Awake and Anesthetized Rat. ENeuro, 4(4), 59–76. https://doi.org/10.1523/ENEURO.0059-17.2017

• Barthó, P., Hirase, H., Monconduit, L., Zugaro, M., Harris, K. D., & Buzsáki, G. (2004). Characterization of Neocortical Principal Cells and Interneurons by Network Interactions and Extracellular Features. Journal of Neurophysiology, 92(1), 600–608. https://doi.org/10.1152/jn.01170.2003

• Bortone, D. S., Olsen, S. R., & Scanziani, M. (2014). Translaminar Inhibitory Cells Recruited by Layer 6 Corticothalamic Neurons Suppress Visual Cortex. Neuron, 82, 474–485. https://doi.org/10.1016/j.neuron.2014.02.021

• Bruno, R. M., & Simons, D. J. (2002). Feedforward Mechanisms of Excitatory and Inhibitory Cortical Receptive Fields. The Journal of Neuroscience, 22(24), 10966–10975. https://doi.org/10.1523/JNEUROSCI.22-24-10966.2002

• Dasilva, M., Camassa, A., Navarro-Guzman, A., Pazienti, A., Perez-Mendez, L., Zamora-López, G., Mattia, M., & Sanchez-Vives, M. V. (2021). Modulation of cortical slow oscillations and complexity across anesthesia levels. NeuroImage, 224, 117415. https://doi.org/10.1016/j.neuroimage.2020.117415

• Franken, P., Malafosse, A., & Tafti, M. (1999). Genetics of sleep regulation in mice-Franken et al Genetic Determinants of Sleep Regulation in Inbred Mice. SLEEP, 22(2). https://academic.oup.com/sleep/article/22/2/155/2731698

• Grenier, F., Timofeev, I., & Steriade, M. (1998). Leading role of thalamic over cortical neurons during postinhibitory rebound excitation. Proceedings of the National Academy of Sciences of the United States of America, 95(23), 13929–13934. https://doi.org/10.1073/pnas.95.23.13929

• Guido, W., & Weyand, T. (1995). Burst responses in thalamic relay cells of the awake behaving cat. Journal of Neurophysiology, 74(4), 1782–1786. https://doi.org/10.1152/JN.1995.74.4.1782

• Histed, M. H., Bonin, V., & Reid, R. C. (2009). Direct Activation of Sparse, Distributed Populations of Cortical Neurons by Electrical Microstimulation. Neuron, 63(4), 508–522. https://doi.org/10.1016/j.neuron.2009.07.016

• Jia, X., Siegle, J., Bennett, C., Gale, S., Denman, D. R., Koch, C., & Olsen, S. (2016). High-density extracellular probes reveal dendritic backpropagation and facilitate neuron classification 1 2. Journal of Neurophysiology, 121(5), 1831–1847. https://doi.org/10.1101/376863

• Kobayashi, G., Tanaka, K. F., & Takata, N. (2023). Pupil Dynamics-derived Sleep Stage Classification of a Head-fixed Mouse Using a Recurrent Neural Network. The Keio Journal of Medicine, 2022-0020-OA. https://doi.org/10.2302/KJM.2022-0020-OA

• Michelson, N. J., & Kozai, T. D. Y. (2018). Isoflurane and ketamine differentially influence spontaneous and evoked laminar electrophysiology in mouse V1. Journal of Neurophysiology, 120(5), 2232. https://doi.org/10.1152/JN.00299.2018

• Niell, C. M., & Stryker, M. P. (2008). Highly selective receptive fields in mouse visual cortex. Journal of Neuroscience, 28(30), 7520–7536. https://doi.org/10.1523/JNEUROSCI.0623-08.2008

• Nowak, L. G., & Bullier, J. (1998). Axons, but not cell bodies, are activated by electrical stimulation in cortical gray matter. II. Evidence from selective inactivation of cell bodies and axon initial segments. Experimental Brain Research, 118(4), 489–500. https://doi.org/10.1007/S002210050305/METRICS

• Roux, L., Stark, E., Sjulson, L., & Buzsáki, G. (2014). In vivo optogenetic identification and manipulation of GABAergic interneuron subtypes. Current Opinion in Neurobiology, 26, 88–95. https://doi.org/10.1016/j.conb.2013.12.013

• Sirota, A., Montgomery, S., Fujisawa, S., Isomura, Y., Zugaro, M., & Buzsáki, G. (2008). Entrainment of Neocortical Neurons and Gamma Oscillations by the Hippocampal Theta Rhythm. Neuron, 60(4), 683–697. https://doi.org/10.1016/j.neuron.2008.09.014

• Sorrenti, V., Cecchetto, C., Maschietto, M., Fortinguerra, S., Buriani, A., & Vassanelli, S. (2021). Understanding the Effects of Anesthesia on Cortical Electrophysiological Recordings: A Scoping Review. International Journal of Molecular Sciences, 22(3), 1286. https://doi.org/10.3390/IJMS22031286

• Tehovnik, E. J., & Slocum, W. M. (2013). Two-photon imaging and the activation of cortical neurons. Neuroscience, 245(March), 12–25. https://doi.org/10.1016/j.neuroscience.2013.04.022

• Tenke, C. E., & Kayser, J. (2012). Generator localization by current source density (CSD): Implications of volume conduction and field closure at intracranial and scalp resolutions. Clinical Neurophysiology, 123(12), 2328–2345. https://doi.org/10.1016/J.CLINPH.2012.06.005

• Turner, K. L., Gheres, K. W., Proctor, E. A., & Drew, P. J. (2020). Neurovascular coupling and bilateral connectivity during nrem and rem sleep. ELife, 9, 1. https://doi.org/10.7554/ELIFE.62071

• Yüzgeç, Ö., Prsa, M., Zimmermann, R., & Huber, D. (2018). Pupil Size Coupling to Cortical States Protects the Stability of Deep Sleep via Parasympathetic Modulation. Current Biology, 28(3), 392. https://doi.org/10.1016/J.CUB.2017.12.049

• Zhang, X., Landsness, E. C., Chen, W., Miao, H., Tang, M., Brier, L. M., Culver, J. P., Lee, J. M., & Anastasio, M. A. (2022). Automated sleep state classification of wide-field calcium imaging data via multiplex visibility graphs and deep learning. Journal of Neuroscience Methods, 366, 109421. https://doi.org/10.1016/J.JNEUMETH.2021.109421

Author response:

Public Review

Joint Public Review:

This manuscript presents an algorithm for identifying network topologies that exhibit a desired qualitative behaviour, with a particular focus on oscillations. The approach is first demonstrated on 3-node networks, where results can be validated through exhaustive search, and then extended to 5-node networks, where the search space becomes intractable. Network topologies are represented as directed graphs, and their dynamical behaviour is classified using stochastic simulations based on the Gillespie algorithm. To efficiently explore the large design space, the authors employ reinforcement learning via Monte Carlo Tree Search (MCTS), framing circuit design as a sequential decision-making process.

This work meaningfully extends the range of systems that can be explored in silico to uncover non-linear dynamics and represents a valuable methodological advance for the fields of systems and synthetic biology.

Strengths

The evidence presented is strong and compelling. The authors validate their results for 3-node networks through exhaustive search, and the findings for 5-node networks are consistent with previously reported motifs, lending credibility to the approach. The use of reinforcement learning to navigate the vast space of possible topologies is both original and effective, and represents a novel contribution to the field. The algorithm demonstrates convincing efficiency, and the ability to identify robust oscillatory topologies is particularly valuable. Expanding the scale of systems that can be systematically explored in silico marks a significant advance for the study of complex gene regulatory networks.

Weaknesses

The principal weakness of the manuscript lies in the interpretation of biological robustness. The authors identify network topologies that sustain oscillatory behaviour despite perturbations to the system or parameters. However, in many cases, this persistence is due to the presence of partially redundant oscillatory motifs within the network. While this observation is interesting and of clear value for circuit design, framing it as evidence of evolutionary robustness may be misleading. The "mutant" systems frequently exhibit altered oscillatory properties, such as changes in frequency or amplitude. From a functional cellular perspective, mere oscillation is insufficient - preservation of specific oscillation characteristics is often essential. This is particularly true in systems like circadian clocks, where misalignment with environmental cycles can have deleterious effects. Robustness, from an evolutionary standpoint, should therefore be framed as the capacity to maintain the functional phenotype, not merely the qualitative behaviour.

A secondary limitation is that, despite the methodological advances, the scale of the systems explored remains modest. While moving from 3- to 5-node systems is non-trivial, five elements still represent a relatively small network. It is somewhat surprising that the algorithm does not scale further, particularly when considering the performance of MCTS in other domains - for instance, modern chess engines routinely explore far larger decision trees. A discussion on current performance bottlenecks and potential avenues for improving scalability would be valuable.

Finally, it is worth noting that the emergence of oscillations in a model often depends not only on the topology but also critically on parameter choices and the nature of the nonlinearities. The use of Hill functions and high Hill coefficients is a common strategy to induce oscillatory dynamics. Thus, the reported results should be interpreted within the context of the modelling assumptions and parameter regimes employed in the simulations.

We thank the reviewers for their careful consideration of our work and for the interesting feedback and scientific discussion. We are working on a revised text based on their recommendations, which will include some of the discussion below.

This work meaningfully extends the range of systems that can be explored in silico to uncover non-linear dynamics and represents a valuable methodological advance for the fields of systems and synthetic biology.

We thank the reviewers for their positive assessment of our work’s impact!

The use of reinforcement learning to navigate the vast space of possible topologies is both original and effective, and represents a novel contribution to the field. The algorithm demonstrates convincing efficiency, and the ability to identify robust oscillatory topologies is particularly valuable. Expanding the scale of systems that can be systematically explored in silico marks a significant advance for the study of complex gene regulatory networks.

We appreciate these kind comments about our work’s merits. We are excited to share our reinforcement learning (RL) based method with the fields of systems and synthetic biology, and we consider it a valuable tool for the systematic analysis and design of larger-scale regulatory networks!

The principal weakness of the manuscript lies in the interpretation of biological robustness. The authors identify network topologies that sustain oscillatory behaviour despite perturbations to the system or parameters… [However, these] "mutant" systems frequently exhibit altered oscillatory properties, such as changes in frequency or amplitude. From a functional cellular perspective, mere oscillation is insufficient - preservation of specific oscillation characteristics is often essential. This is particularly true in systems like circadian clocks, where misalignment with environmental cycles can have deleterious effects. Robustness, from an evolutionary standpoint, should therefore be framed as the capacity to maintain the functional phenotype, not merely the qualitative behaviour.

We thank the reviewers for their attention to this point. In the large-scale circuit search, summarized in Figures 4A and 4B, we ran a search for 5-component oscillators that can spontaneously oscillate even when subjected to the deletion of a random gene. Some of the best performing circuits under these conditions exhibited a design feature we call “motif multiplexing,” in which multiple smaller motifs are interleaved in a way that makes oscillation possible under many different mutational scenarios. Interestingly, despite not selecting for preservation of frequency, the 3Ai+3Rep circuit (a 5-gene circuit highlighted in Figure 5) anecdotally appears to have a natural frequency that is robust to partial gene knockdowns, although not to complete gene deletions. As shown in Figure 5C, this circuit has a natural frequency of 6 cycles/hr (with one particular parameterization), and it can sustain a knockdown of any of its 5 genes to 50% of the wild-type transcription rate without altering the natural frequency by more than 20%.

However, we agree that there are salient differences between this training scenario and natural evolution. The revised text will clarify that these differences limit what conclusions can be drawn about biological evolution by analogy. As the reviewers point out, we use the presence of spontaneous oscillations (with or without the deletion) as a measure of fitness, regardless of frequency, so as to screen for designs with promising behavior. Also, the deletion mutations introduced during training likely represent larger perturbations to the system than a typical mutation encountered during genome replication (for example, a point mutation in a response element leading to a moderate change in binding affinity). Finally, we do not introduce any entrainment. Real circadian oscillators are aligned to a 24-hour period (“entrained”) by environmental inputs such as light and temperature. For this reason, natural circadian clocks may have natural frequencies that are slightly shorter or longer than 24 hours, although a close proximity to the 24-hour period does seem to be an important selective factor [1].

...despite the methodological advances, the scale of the systems explored remains modest. While moving from 3- to 5-node systems is non-trivial, five elements still represent a relatively small network. It is somewhat surprising that the algorithm does not scale further, particularly when considering the performance of MCTS in other domains - for instance, modern chess engines routinely explore far larger decision trees. A discussion on current performance bottlenecks and potential avenues for improving scalability would be valuable.

We thank the reviewers for their attention to this point. The main limitation we encountered to exploring circuits with more than 5 nodes in this work was the poor computational scaling of the Gillespie stochastic simulation algorithm, rather than a limitation of MCTS itself. While the average runtime of a 3-node circuit simulation was roughly 7 seconds, this number increased to 18-20 seconds with 5-node circuits. For this reason, we limited the search to topologies with ≤15 interaction arrows (15 sec/simulation). In general, the simulation time was proportional to the square of the number of transcription factors (TFs). We will revise the text to include the reason for stopping at 5 nodes, which is significant for understanding CircuiTree’s scaling properties.

With regards to scaling, an important advantage of CircuiTree is its ability to generate useful candidate designs after exploring only a portion of the search space. Like exhaustive search, given enough time, MCTS will comprehensively explore the search space and find all possible solutions. However, for large search spaces, RL-based agents are generally given a finite number of simulations (or time) to learn as much as possible.

Across machine learning (ML) applications [2] and particularly with RL models [3], this training time tends to obey a power law with respect to the underlying complexity of the problem. Thus we can use the complexity of the 3-node and 5-node searches to infer the current scaling limits of CircuiTree. The first oscillator topology was discovered after 2,280 simulations for the 3-node search, and in the 5-node search, the first oscillator using 5 nodes appeared at ~8e5 simulations, resulting in a power law of Y ~ 84.4 X^0.333. Thus, useful candidate designs may be found for 6-node and 7-node searches after 4.5e7 and 5.26e9 simulations, respectively, even though these spaces contain 1.5e17 and 2.5e23 topologies, respectively. Thus, running a 7-node search with the current implementation of CircuiTree would require resources close to the current boundaries of computation, requiring roughly 1.8 million CPU-hours, or 2 weeks on 5,000 CPUs, assuming a 1-second simulation. These points will be incorporated into both the results and discussion sections in our revised text.

However, we are optimistic about CircuiTree’s potential to scale to much larger circuits with modifications to its algorithm. CircuiTree uses the original (so-called “vanilla”) implementation of MCTS, which has not been used in professional game-playing AIs in over a decade. Contemporary RL-based game-playing engines leverage deep neural networks to dramatically reduce the training time, using value networks to identify game-winning positions and policy networks to find game-winning moves. AlphaZero, developed by Google DeepMind to learn games by self-play and without domain knowledge, outperformed all other chess AIs after 44 million training games, much smaller than the 10^43 possible chess states [4]. Similarly, the game of go has 10¹⁷⁰ possible states, but AlphaZero outperformed other AIs after only 140 million games [4]. Large circuits live in similarly large search spaces; for example, 19-node and 20-node circuits represent spaces of 10¹⁷² and 10¹⁹⁰ possible topologies. The revised text will include this discussion and identify value and policy networks, as well as more scalable simulation paradigms such as ODEs and neural ODEs, as our future directions for improving CircuiTree’s scalability.

Finally, our revised discussion will note some important differences between game-playing and biological circuit design. Unlike deterministic games like chess, the final value of a circuit topology is determined stochastically, by running a simulation whose fitness depends on the parameter set and initial conditions. Thus, state-for-state, it is possible that training an agent for circuit design may inherently require more simulations to achieve the same level of certainty compared to classical games. Additionally, while we often possess a priori knowledge about a game such as its overall difficulty or certain known strategies, we lack this frame of reference when searching for circuit designs. Thus, it remains challenging to know if and when a large space of designs has been “satisfactorily” or “comprehensively” searched, since the answer depends on data that are unknown, namely the quantity, quality, and location of solutions residing in the search space.

Not accounting for redundancy due to structural symmetries

Finally, it is worth noting that the emergence of oscillations in a model often depends not only on the topology but also critically on parameter choices and the nature of the nonlinearities. The use of Hill functions and high Hill coefficients is a common strategy to induce oscillatory dynamics. Thus, the reported results should be interpreted within the context of the modelling assumptions and parameter regimes employed in the simulations.

In our dynamical modeling of transcription factor (TF) networks, we do not rely on continuum assumptions about promoter occupancy such as Hill functions. Rather, we model each reaction - transcription, translation, TF binding/unbinding, and degradation - explicitly, and individual molecules appear and disappear via stochastic birth and death events. Many natural TFs are homodimers that bind cooperatively to regulate transcription; similarly, we assume that pairs of TFs bind more stably to their response element than individual TFs. Thus, our model has similar cooperativity to a Hill function, and it can be shown that in the continuum limit, the effective Hill coefficient is always ≤2. Our revision will clarify this aspect of the modeling and include a derivation of this property. Currently, the parameter values used in the figures are shown in Table 2. In the revised text, these will be displayed in the body of the text as well for clarity.

Bibliography (1) Spoelstra, K., Wikelski, M., Daan, S., Loudon, A. S. I., & Hau, M. (2015). Natural selection against a circadian clock gene mutation in mice. PNAS, 113(3), 686–691. https://doi.org/https://doi.org/10.1073/pnas.1516442113<br /> (2) Neumann, O., & Gros, C. (2023). Scaling Laws for a Multi-Agent Reinforcement Learning Model. The Eleventh International Conference on Learning Representations. Retrieved from https://openreview.net/forum?id=ZrEbzL9eQ3W (3) Jones, A. L. (2021). Scaling Scaling Laws with Board Games. arXiv [Cs.LG]. Retrieved from http://arxiv.org/abs/2104.03113 (4) Silver, D., Hubert, T., Schrittwieser, J., Antonoglou, I., Lai, M., Guez, A., Lanctot, M., Sifre, L., Kumaran, D., Graepel, T., Lillicrap, T., Simonyan, K., & Hassabis, D. (2018). A general reinforcement learning algorithm that Masters Chess, Shogi, and go through self-play. Science, 362(6419), 1140–1144. https://doi.org/10.1126/science.aar6404

Author response:

The following is the authors’ response to the original reviews.

eLife Assessment

This valuable study investigates how the neural representation of individual finger movements changes during the early period of sequence learning. By combining a new method for extracting features from human magnetoencephalography data and decoding analyses, the authors provide incomplete evidence of an early, swift change in the brain regions correlated with sequence learning, including a set of previously unreported frontal cortical regions. The addition of more control analyses to rule out that head movement artefacts influence the findings, and to further explain the proposal of offline contextualization during short rest periods as the basis for improvement performance would strengthen the manuscript.

We appreciate the Editorial assessment on our paper’s strengths and novelty. We have implemented additional control analyses to show that neither task-related eye movements nor increasing overlap of finger movements during learning account for our findings, which are that contextualized neural representations in a network of bilateral frontoparietal brain regions actively contribute to skill learning. Importantly, we carried out additional analyses showing that contextualization develops predominantly during rest intervals.

Public Reviews:

Reviewer #1 (Public review):

Summary:

This study addresses the issue of rapid skill learning and whether individual sequence elements (here: finger presses) are differentially represented in human MEG data. The authors use a decoding approach to classify individual finger elements and accomplish an accuracy of around 94%. A relevant finding is that the neural representations of individual finger elements dynamically change over the course of learning. This would be highly relevant for any attempts to develop better brain machine interfaces - one now can decode individual elements within a sequence with high precision, but these representations are not static but develop over the course of learning.

Strengths:

The work follows a large body of work from the same group on the behavioural and neural foundations of sequence learning. The behavioural task is well established and neatly designed to allow for tracking learning and how individual sequence elements contribute. The inclusion of short offline rest periods between learning epochs has been influential because it has revealed that a lot, if not most of the gains in behaviour (ie speed of finger movements) occur in these socalled micro-offline rest periods. The authors use a range of new decoding techniques, and exhaustively interrogate their data in different ways, using different decoding approaches. Regardless of the approach, impressively high decoding accuracies are observed, but when using a hybrid approach that combines the MEG data in different ways, the authors observe decoding accuracies of individual sequence elements from the MEG data of up to 94%.

We have previously showed that neural replay of MEG activity representing the practiced skill was prominent during rest intervals of early learning, and that the replay density correlated with micro-offline gains (Buch et al., 2021). These findings are consistent with recent reports (from two different research groups) that hippocampal ripple density increases during these inter-practice rest periods, and predict offline learning gains (Chen et al., 2024; Sjøgård et al., 2024). However, decoder performance in our earlier work (Buch et al., 2021) left room for improvement. Here, we reported a strategy to improve decoding accuracy that could benefit future studies of neural replay or BCI using MEG.

Weaknesses:

There are a few concerns which the authors may well be able to resolve. These are not weaknesses as such, but factors that would be helpful to address as these concern potential contributions to the results that one would like to rule out. Regarding the decoding results shown in Figure 2 etc, a concern is that within individual frequency bands, the highest accuracy seems to be within frequencies that match the rate of keypresses. This is a general concern when relating movement to brain activity, so is not specific to decoding as done here. As far as reported, there was no specific restraint to the arm or shoulder, and even then it is conceivable that small head movements would correlate highly with the vigor of individual finger movements. This concern is supported by the highest contribution in decoding accuracy being in middle frontal regions - midline structures that would be specifically sensitive to movement artefacts and don't seem to come to mind as key structures for very simple sequential keypress tasks such as this - and the overall pattern is remarkably symmetrical (despite being a unimanual finger task) and spatially broad. This issue may well be matching the time course of learning, as the vigor and speed of finger presses will also influence the degree to which the arm/shoulder and head move. This is not to say that useful information is contained within either of the frequencies or broadband data. But it raises the question of whether a lot is dominated by movement "artefacts" and one may get a more specific answer if removing any such contributions.

Reviewer #1 expresses concern that the combination of the low-frequency narrow-band decoder results, and the bilateral middle frontal regions displaying the highest average intra-parcel decoding performance across subjects is suggestive that the decoding results could be driven by head movement or other artefacts.

Head movement artefacts are highly unlikely to contribute meaningfully to our results for the following reasons. First, in addition to ICA denoising, all “recordings were visually inspected and marked to denoise segments containing other large amplitude artifacts due to movements” (see Methods). Second, the response pad was positioned in a manner that minimized wrist, arm or more proximal body movements during the task. Third, while online monitoring of head position was not performed for this study, it was assessed at the beginning and at the end of each recording. The head was restrained with an inflatable air bladder, and head movement between the beginning and end of each scan did not exceed 5mm for all participants included in the study.

The Reviewer states a concern that “it is conceivable that small head movements would correlate highly with the vigor of individual finger movements”. We agree that despite the steps taken above, it is possible that minor head movements could still contribute to some remaining variance in the MEG data in our study. However, such correlations between small head movements and finger movements could only meaningfully contribute to decoding performance if: (A) they were consistent and pervasive throughout the recording (which might not be the case if the head movements were related to movement vigor and vigor changed over time); and (B) they systematically varied between different finger movements, and also between the same finger movement performed at different sequence locations (see 5-class decoding performance in Figure 4B). The possibility of any head movement artefacts meeting all these conditions is unlikely. Alternatively, for this task design a much more likely confound could be the contribution of eye movement artefacts to the decoder performance (an issue raised by Reviewer #3 in the comments below).

Remember from Figure 1A in the manuscript that an asterisk marks the current position in the sequence and is updated at each keypress. Since participants make very few performance errors, the position of the asterisk on the display is highly correlated with the keypress being made in the sequence. Thus, it is possible that if participants are attending to the visual feedback provided on the display, they may generate eye movements that are systematically related to the task. Since we did record eye movements simultaneously with the MEG recordings (EyeLink 1000 Plus; Fs = 600 Hz), we were able to perform a control analysis to address this question. For each keypress event during trials in which no errors occurred (which is the same time-point that the asterisk position is updated), we extracted three features related to eye movements: 1) the gaze position at the time of asterisk position update (triggered by a KeyDown event), 2) the gaze position 150ms later, and 3) the peak velocity of the eye movement between the two positions. We then constructed a classifier from these features with the aim of predicting the location of the asterisk (ordinal positions 1-5) on the display. As shown in the confusion matrix below (Author response image 1), the classifier failed to perform above chance levels (overall cross-validated accuracy = 0.21817):

Author response image 1.

Confusion matrix showing that three eye movement features fail to predict asterisk position on the task display above chance levels (Fold 1 test accuracy = 0.21718; Fold 2 test accuracy = 0.22023; Fold 3 test accuracy = 0.21859; Fold 4 test accuracy = 0.22113; Fold 5 test accuracy = 0.21373; Overall cross-validated accuracy = 0.2181). Since the ordinal position of the asterisk on the display is highly correlated with the ordinal position of individual keypresses in the sequence, this analysis provides strong evidence that keypress decoding performance from MEG features is not explained by systematic relationships between finger movement behavior and eye movements (i.e. – behavioral artefacts) (end of figure legend).

Remember that the task display does not provide explicit feedback related to performance, only information about the present position in the sequence. Thus, it is possible that participants did not actively attend to the feedback. In fact, inspection of the eye position data revealed that on majority of trials, participants displayed random-walk-like gaze patterns around a central fixation point located near the center of the screen. Thus, participants did not attend to the asterisk position on the display, but instead intrinsically generated the action sequence. A similar realworld example would be manually inputting a long password into a secure online application. In this case, one intrinsically generates the sequence from memory and receives similar feedback about the password sequence position (also provided as asterisks) as provided in the study task – feedback which is typically ignored by the user.

The minimal participant engagement with the visual task display observed in this study highlights another important point – that the behavior in explicit sequence learning motor tasks is highly generative in nature rather than reactive to stimulus cues as in the serial reaction time task (SRTT). This is a crucial difference that must be carefully considered when designing investigations and comparing findings across studies.

We observed that initial keypress decoding accuracy was predominantly driven by contralateral primary sensorimotor cortex in the initial practice trials before transitioning to bilateral frontoparietal regions by trials 11 or 12 as performance gains plateaued. The contribution of contralateral primary sensorimotor areas to early skill learning has been extensively reported in humans and non-human animals.(Buch et al., 2021; Classen et al., 1998; Karni et al., 1995; Kleim et al., 1998) Similarly, the increased involvement of bilateral frontal and parietal regions to decoding during early skill learning in the non-dominant hand is well known. Enhanced bilateral activation in both frontal and parietal cortex during skill learning has been extensively reported (Doyon et al., 2002; Grafton et al., 1992; Hardwick et al., 2013; Kennerley et al., 2004; Shadmehr & Holcomb, 1997; Toni, Ramnani, et al., 2001), and appears to be even more prominent during early fine motor skill learning in the non-dominant hand (Lee et al., 2019; Sawamura et al., 2019). The frontal regions identified in these studies are known to play crucial roles in executive control (Battaglia-Mayer & Caminiti, 2019), motor planning (Toni, Thoenissen, et al., 2001), and working memory (Andersen & Buneo, 2002; Buneo & Andersen, 2006; Shadmehr & Holcomb, 1997; Toni, Ramnani, et al., 2001; Wolpert et al., 1998) processes, while the same parietal regions are known to integrate multimodal sensory feedback and support visuomotor transformations (Andersen & Buneo, 2002; Buneo & Andersen, 2006; Shadmehr & Holcomb, 1997; Toni, Ramnani, et al., 2001; Wolpert et al., 1998), in addition to working memory (Grover et al., 2022). Thus, it is not surprising that these regions increasingly contribute to decoding as subjects internalize the sequential task. We now include a statement reflecting these considerations in the revised Discussion.

A somewhat related point is this: when combining voxel and parcel space, a concern is whether a degree of circularity may have contributed to the improved accuracy of the combined data, because it seems to use the same MEG signals twice - the voxels most contributing are also those contributing most to a parcel being identified as relevant, as parcels reflect the average of voxels within a boundary. In this context, I struggled to understand the explanation given, ie that the improved accuracy of the hybrid model may be due to "lower spatially resolved whole-brain and higher spatially resolved regional activity patterns".

We disagree with the Reviewer’s assertion that the construction of the hybrid-space decoder is circular for the following reasons. First, the base feature set for the hybrid-space decoder constructed for all participants includes whole-brain spatial patterns of MEG source activity averaged within parcels. As stated in the manuscript, these 148 inter-parcel features reflect “lower spatially resolved whole-brain activity patterns” or global brain dynamics. We then independently test how well spatial patterns of MEG source activity for all voxels distributed within individual parcels can decode keypress actions. Again, the testing of these intra-parcel spatial patterns, intended to capture “higher spatially resolved regional brain activity patterns”, is completely independent from one another and independent from the weighting of individual inter-parcel features. These intra-parcel features could, for example, provide additional information about muscle activation patterns or the task environment. These approximately 1150 intra-parcel voxels (on average, within the total number varying between subjects) are then combined with the 148 inter-parcel features to construct the final hybrid-space decoder. In fact, this varied spatial filter approach shares some similarities to the construction of convolutional neural networks (CNNs) used to perform object recognition in image classification applications (Srinivas et al., 2016). One could also view this hybrid-space decoding approach as a spatial analogue to common timefrequency based analyses such as theta-gamma phase amplitude coupling (θ/γ PAC), which assess interactions between two or more narrow-band spectral features derived from the same time-series data (Lisman & Jensen, 2013).

We directly tested this hypothesis – that spatially overlapping intra- and inter-parcel features portray different information – by constructing an alternative hybrid-space decoder (Hybrid_Alt) that excluded average inter-parcel features which spatially overlapped with intra-parcel voxel features, and comparing the performance to the decoder used in the manuscript (Hybrid_Orig). The prediction was that if the overlapping parcel contained similar information to the more spatially resolved voxel patterns, then removing the parcel features (n=8) from the decoding analysis should not impact performance. In fact, despite making up less than 1% of the overall input feature space, removing those parcels resulted in a significant drop in overall performance greater than 2% (78.15% ± 7.03% SD for Hybrid_Orig vs. 75.49% ± 7.17% for Hybrid_Alt; Wilcoxon signed rank test, z = 3.7410, p = 1.8326e-04; Author response image 2).

Author response image 2.

Comparison of decoding performances with two different hybrid approaches. Hybrid_Alt: Intra-parcel voxel-space features of top ranked parcels and inter-parcel features of remaining parcels. Hybrid_Orig: Voxel-space features of top ranked parcels and whole-brain parcel-space features (i.e. – the version used in the manuscript). Dots represent decoding accuracy for individual subjects. Dashed lines indicate the trend in performance change across participants. Note, that Hybrid_Orig (the approach used in our manuscript) significantly outperforms the Hybrid_Alt approach, indicating that the excluded parcel features provide unique information compared to the spatially overlapping intra-parcel voxel patterns (end of figure legend).

Firstly, there will be a relatively high degree of spatial contiguity among voxels because of the nature of the signal measured, i.e. nearby individual voxels are unlikely to be independent. Secondly, the voxel data gives a somewhat misleading sense of precision; the inversion can be set up to give an estimate for each voxel, but there will not just be dependence among adjacent voxels, but also substantial variation in the sensitivity and confidence with which activity can be projected to different parts of the brain. Midline and deeper structures come to mind, where the inversion will be more problematic than for regions along the dorsal convexity of the brain, and a concern is that in those midline structures, the highest decoding accuracy is seen.

We agree with the Reviewer that some inter-parcel features representing neighboring (or spatially contiguous) voxels are likely to be correlated, an important confound in connectivity analyses (Colclough et al., 2015; Colclough et al., 2016), not performed in our investigation.

In our study, correlations between adjacent voxels effectively reduce the dimensionality of the input feature space. However, as long as there are multiple groups of correlated voxels within each parcel (i.e. – the rank is greater than 1), the intra-parcel spatial patterns could meaningfully contribute to the decoder performance, as shown by the following results:

First, we obtained higher decoding accuracy with voxel-space features (74.51% ± 7.34% SD) compared to parcel space features (68.77% ± 7.6%; Figure 3B), indicating individual voxels carry more information in decoding the keypresses than the averaged voxel-space features or parcel space features. Second, individual voxels within a parcel showed varying feature importance scores in decoding keypresses (Author response image 3). This finding shows that correlated voxels form mini subclusters that are much smaller spatially than the parcel they reside within.

Author response image 3.:

Feature importance score of individual voxels in decoding keypresses: MRMR was used to rank the individual voxel space features in decoding keypresses and the min-max normalized MRMR score was mapped to a structural brain surface. Note that individual voxels within a parcel showed different contribution to decoding (end of figure legend).

Some of these concerns could be addressed by recording head movement (with enough precision) to regress out these contributions. The authors state that head movement was monitored with 3 fiducials, and their time courses ought to provide a way to deal with this issue. The ICA procedure may not have sufficiently dealt with removing movement-related problems, but one could eg relate individual components that were identified to the keypresses as another means for checking. An alternative could be to focus on frequency ranges above the movement frequencies. The accuracy for those still seems impressive and may provide a slightly more biologically plausible assessment.

We have already addressed the issue of movement related artefacts in the first response above. With respect to a focus on frequency ranges above movement frequencies, the Reviewer states the “accuracy for those still seems impressive and may provide a slightly more biologically plausible assessment”. First, it is important to note that cortical delta-band oscillations measured with local field potentials (LFPs) in macaques is known to contain important information related to end-effector kinematics (Bansal et al., 2011; Mollazadeh et al., 2011) muscle activation patterns (Flint et al., 2012) and temporal sequencing (Churchland et al., 2012) during skilled reaching and grasping actions. Thus, there is a substantial body of evidence that low-frequency neural oscillatory activity in this range contains important information about the skill learning behavior investigated in the present study. Second, our own data shows (which the Reviewer also points out) that significant information related to the skill learning behavior is also present in higher frequency bands (see Figure 2A and Figure 3—figure supplement 1). As we pointed out in our earlier response to questions about the hybrid space decoder architecture (see above), it is likely that different, yet complimentary, information is encoded across different temporal frequencies (just as it is encoded across different spatial frequencies) (Heusser et al., 2016). Again, this interpretation is supported by our data as the highest performing classifiers in all cases (when holding all parameters constant) were always constructed from broadband input MEG data (Figure 2A and Figure 3—figure supplement 1).

One question concerns the interpretation of the results shown in Figure 4. They imply that during the course of learning, entirely different brain networks underpin the behaviour. Not only that, but they also include regions that would seem rather unexpected to be key nodes for learning and expressing relatively simple finger sequences, such as here. What then is the biological plausibility of these results? The authors seem to circumnavigate this issue by moving into a distance metric that captures the (neural network) changes over the course of learning, but the discussion seems detached from which regions are actually involved; or they offer a rather broad discussion of the anatomical regions identified here, eg in the context of LFOs, where they merely refer to "frontoparietal regions".

The Reviewer notes the shift in brain networks driving keypress decoding performance between trials 1, 11 and 36 as shown in Figure 4A. The Reviewer questions whether these shifts in brain network states underpinning the skill are biologically plausible, as well as the likelihood that bilateral superior and middle frontal and parietal cortex are important nodes within these networks.

First, previous fMRI work in humans assessed changes in functional connectivity patterns while participants performed a similar sequence learning task to our present study (Bassett et al., 2011). Using a dynamic network analysis approach, Bassett et al. showed that flexibility in the composition of individual network modules (i.e. – changes in functional brain region membership of orthogonal brain networks) is up-regulated in novel learning environments and explains differences in learning rates across individuals. Thus, consistent with our findings, it is likely that functional brain networks rapidly reconfigure during early learning of novel sequential motor skills.

Second, frontoparietal network activity is known to support motor memory encoding during early learning (Albouy et al., 2013; Albouy et al., 2012). For example, reactivation events in the posterior parietal (Qin et al., 1997) and medial prefrontal (Euston et al., 2007; Molle & Born, 2009) cortex (MPFC) have been temporally linked to hippocampal replay, and are posited to support memory consolidation across several memory domains (Frankland & Bontempi, 2005), including motor sequence learning (Albouy et al., 2015; Buch et al., 2021; F. Jacobacci et al., 2020). Further, synchronized interactions between MPFC and hippocampus are more prominent during early as opposed to later learning stages (Albouy et al., 2013; Gais et al., 2007; Sterpenich et al., 2009), perhaps reflecting “redistribution of hippocampal memories to MPFC” (Albouy et al., 2013). MPFC contributes to very early memory formation by learning association between contexts, locations, events and adaptive responses during rapid learning (Euston et al., 2012). Consistently, coupling between hippocampus and MPFC has been shown during initial memory encoding and during subsequent rest (van Kesteren et al., 2010; van Kesteren et al., 2012). Importantly, MPFC activity during initial memory encoding predicts subsequent recall (Wagner et al., 1998). Thus, the spatial map required to encode a motor sequence memory may be “built under the supervision of the prefrontal cortex” (Albouy et al., 2012), also engaged in the development of an abstract representation of the sequence (Ashe et al., 2006). In more abstract terms, the prefrontal, premotor and parietal cortices support novice performance “by deploying attentional and control processes” (Doyon et al., 2009; Hikosaka et al., 2002; Penhune & Steele, 2012) required during early learning (Doyon et al., 2009; Hikosaka et al., 2002; Penhune & Steele, 2012). The dorsolateral prefrontal cortex DLPFC specifically is thought to engage in goal selection and sequence monitoring during early skill practice (Schendan et al., 2003), all consistent with the schema model of declarative memory in which prefrontal cortices play an important role in encoding (Morris, 2006; Tse et al., 2007). Thus, several prefrontal and frontoparietal regions contributing to long term learning (Berlot et al., 2020) are also engaged in early stages of encoding. Altogether, there is strong biological support for the involvement of bilateral prefrontal and frontoparietal regions to decoding during early skill learning. We now address this issue in the revised manuscript.

If I understand correctly, the offline neural representation analysis is in essence the comparison of the last keypress vs the first keypress of the next sequence. In that sense, the activity during offline rest periods is actually not considered. This makes the nomenclature somewhat confusing. While it matches the behavioural analysis, having only key presses one can't do it in any other way, but here the authors actually do have recordings of brain activity during offline rest. So at the very least calling it offline neural representation is misleading to this reviewer because what is compared is activity during the last and during the next keypress, not activity during offline periods. But it also seems a missed opportunity - the authors argue that most of the relevant learning occurs during offline rest periods, yet there is no attempt to actually test whether activity during this period can be useful for the questions at hand here.

We agree with the Reviewer that our previous “offline neural representation” nomenclature could be misinterpreted. In the revised manuscript we refer to this difference as the “offline neural representational change”. Please, note that our previous work did link offline neural activity (i.e. – 16-22 Hz beta power (Bonstrup et al., 2019) and neural replay density (Buch et al., 2021) during inter-practice rest periods) to observed micro-offline gains.

Reviewer #2 (Public review):

Summary

Dash et al. asked whether and how the neural representation of individual finger movements is "contextualized" within a trained sequence during the very early period of sequential skill learning by using decoding of MEG signal. Specifically, they assessed whether/how the same finger presses (pressing index finger) embedded in the different ordinal positions of a practiced sequence (4-1-3-2-4; here, the numbers 1 through 4 correspond to the little through the index fingers of the non-dominant left hand) change their representation (MEG feature). They did this by computing either the decoding accuracy of the index finger at the ordinal positions 1 vs. 5 (index_OP1 vs index_OP5) or pattern distance between index_OP1 vs. index_OP5 at each training trial and found that both the decoding accuracy and the pattern distance progressively increase over the course of learning trials. More interestingly, they also computed the pattern distance for index_OP5 for the last execution of a practice trial vs. index_OP1 for the first execution in the next practice trial (i.e., across the rest period). This "off-line" distance was significantly larger than the "on-line" distance, which was computed within practice trials and predicted micro-offline skill gain. Based on these results, the authors conclude that the differentiation of representation for the identical movement embedded in different positions of a sequential skill ("contextualization") primarily occurs during early skill learning, especially during rest, consistent with the recent theory of the "micro-offline learning" proposed by the authors' group. I think this is an important and timely topic for the field of motor learning and beyond.

Strengths

The specific strengths of the current work are as follows. First, the use of temporally rich neural information (MEG signal) has a large advantage over previous studies testing sequential representations using fMRI. This allowed the authors to examine the earliest period (= the first few minutes of training) of skill learning with finer temporal resolution. Second, through the optimization of MEG feature extraction, the current study achieved extremely high decoding accuracy (approx. 94%) compared to previous works. As claimed by the authors, this is one of the strengths of the paper (but see my comments). Third, although some potential refinement might be needed, comparing "online" and "offline" pattern distance is a neat idea.

Weaknesses

Along with the strengths I raised above, the paper has some weaknesses. First, the pursuit of high decoding accuracy, especially the choice of time points and window length (i.e., 200 msec window starting from 0 msec from key press onset), casts a shadow on the interpretation of the main result. Currently, it is unclear whether the decoding results simply reflect behavioral change or true underlying neural change. As shown in the behavioral data, the key press speed reached 3~4 presses per second already at around the end of the early learning period (11th trial), which means inter-press intervals become as short as 250-330 msec. Thus, in almost more than 60% of training period data, the time window for MEG feature extraction (200 msec) spans around 60% of the inter-press intervals. Considering that the preparation/cueing of subsequent presses starts ahead of the actual press (e.g., Kornysheva et al., 2019) and/or potential online planning (e.g., Ariani and Diedrichsen, 2019), the decoder likely has captured these future press information as well as the signal related to the current key press, independent of the formation of genuine sequential representation (e.g., "contextualization" of individual press). This may also explain the gradual increase in decoding accuracy or pattern distance between index_OP1 vs. index_OP5 (Figure 4C and 5A), which co-occurred with performance improvement, as shorter inter-press intervals are more favorable for the dissociating the two index finger presses followed by different finger presses. The compromised decoding accuracies for the control sequences can be explained in similar logic. Therefore, more careful consideration and elaborated discussion seem necessary when trying to both achieve high-performance decoding and assess early skill learning, as it can impact all the subsequent analyses.

The Reviewer raises the possibility that (given the windowing parameters used in the present study) an increase in “contextualization” with learning could simply reflect faster typing speeds as opposed to an actual change in the underlying neural representation.

We now include a new control analysis that addresses this issue as well as additional re-examination of previously reported results with respect to this issue – all of which are inconsistent with this alternative explanation that “contextualization” reflects a change in mixing of keypress related MEG features as opposed to a change in the underlying representations themselves. As correct sequences are generated at higher and higher speeds over training, MEG activity patterns related to the planning, execution, evaluation and memory of individual keypresses overlap more in time. Thus, increased overlap between the “4” and “1” keypresses (at the start of the sequence) and “2” and “4” keypresses (at the end of the sequence) could artefactually increase contextualization distances even if the underlying neural representations for the individual keypresses remain unchanged. One must also keep in mind that since participants repeat the sequence multiple times within the same trial, a majority of the index finger keypresses are performed adjacent to one another (i.e. - the “4-4” transition marking the end of one sequence and the beginning of the next). Thus, increased overlap between consecutive index finger keypresses as typing speed increased should increase their similarity and mask contextualization related changes to the underlying neural representations.

We addressed this question by conducting a new multivariate regression analysis to directly assess whether the neural representation distance score could be predicted by the 4-1, 2-4 and 4-4 keypress transition times observed for each complete correct sequence (both predictor and response variables were z-score normalized within-subject). The results of this analysis also affirmed that the possible alternative explanation that contextualization effects are simple reflections of increased mixing is not supported by the data (Adjusted R² = 0.00431; F = 5.62). We now include this new negative control analysis in the revised manuscript.

We also re-examined our previously reported classification results with respect to this issue. We reasoned that if mixing effects reflecting the ordinal sequence structure is an important driver of the contextualization finding, these effects should be observable in the distribution of decoder misclassifications. For example, “4” keypresses would be more likely to be misclassified as “1” or “2” keypresses (or vice versa) than as “3” keypresses. The confusion matrices presented in Figures 3C and 4B and Figure 3—figure supplement 3A display a distribution of misclassifications that is inconsistent with an alternative mixing effect explanation of contextualization.

Based upon the increased overlap between adjacent index finger keypresses (i.e. – “4-4” transition), we also reasoned that the decoder tasked with separating individual index finger keypresses into two distinct classes based upon sequence position, should show decreased performance as typing speed increases. However, Figure 4C in our manuscript shows that this is not the case. The 2-class hybrid classifier actually displays improved classification performance over early practice trials despite greater temporal overlap. Again, this is inconsistent with the idea that the contextualization effect simply reflects increased mixing of individual keypress features.

In summary, both re-examination of previously reported data and new control analyses all converged on the idea that the proximity between keypresses does not explain contextualization.

We do agree with the Reviewer that the naturalistic, generative, self-paced task employed in the present study results in overlapping brain processes related to planning, execution, evaluation and memory of the action sequence. We also agree that there are several tradeoffs to consider in the construction of the classifiers depending on the study aim. Given our aim of optimizing keypress decoder accuracy in the present study, the set of trade-offs resulted in representations reflecting more the latter three processes, and less so the planning component. Whether separate decoders can be constructed to tease apart the representations or networks supporting these overlapping processes is an important future direction of research in this area. For example, work presently underway in our lab constrains the selection of windowing parameters in a manner that allows individual classifiers to be temporally linked to specific planning, execution, evaluation or memory-related processes to discern which brain networks are involved and how they adaptively reorganize with learning. Results from the present study (Figure 4—figure supplement 2) showing hybrid-space decoder prediction accuracies exceeding 74% for temporal windows spanning as little as 25ms and located up to 100ms prior to the KeyDown event strongly support the feasibility of such an approach.

Related to the above point, testing only one particular sequence (4-1-3-2-4), aside from the control ones, limits the generalizability of the finding. This also may have contributed to the extremely high decoding accuracy reported in the current study.

The Reviewer raises a question about the generalizability of the decoder accuracy reported in our study. Fortunately, a comparison between decoder performances on Day 1 and Day 2 datasets does provide insight into this issue. As the Reviewer points out, the classifiers in this study were trained and tested on keypresses performed while practicing a specific sequence (4-1-3-2-4). The study was designed this way as to avoid the impact of interference effects on learning dynamics. The cross-validated performance of classifiers on MEG data collected within the same session was 90.47% overall accuracy (4-class; Figure 3C). We then tested classifier performance on data collected during a separate MEG session conducted approximately 24 hours later (Day 2; see Figure 3 — figure supplement 3). We observed a reduction in overall accuracy rate to 87.11% when tested on MEG data recorded while participants performed the same learned sequence, and 79.44% when they performed several previously unpracticed sequences. Both changes in accuracy are important with regards to the generalizability of our findings. First, 87.11% performance accuracy for the trained sequence data on Day 2 (a reduction of only 3.36%) indicates that the hybrid-space decoder performance is robust over multiple MEG sessions, and thus, robust to variations in SNR across the MEG sensor array caused by small differences in head position between scans. This indicates a substantial advantage over sensor-space decoding approaches. Furthermore, when tested on data from unpracticed sequences, overall performance dropped an additional 7.67%. This difference reflects the performance bias of the classifier for the trained sequence, possibly caused by high-order sequence structure being incorporated into the feature weights. In the future, it will be important to understand in more detail how random or repeated keypress sequence training data impacts overall decoder performance and generalization. We strongly agree with the Reviewer that the issue of generalizability is extremely important and have added a new paragraph to the Discussion in the revised manuscript highlighting the strengths and weaknesses of our study with respect to this issue.

In terms of clinical BCI, one of the potential relevance of the study, as claimed by the authors, it is not clear that the specific time window chosen in the current study (up to 200 msec since key press onset) is really useful. In most cases, clinical BCI would target neural signals with no overt movement execution due to patients' inability to move (e.g., Hochberg et al., 2012). Given the time window, the surprisingly high performance of the current decoder may result from sensory feedback and/or planning of subsequent movement, which may not always be available in the clinical BCI context. Of course, the decoding accuracy is still much higher than chance even when using signal before the key press (as shown in Figure 4 Supplement 2), but it is not immediately clear to me that the authors relate their high decoding accuracy based on post-movement signal to clinical BCI settings.

The Reviewer questions the relevance of the specific window parameters used in the present study for clinical BCI applications, particularly for paretic patients who are unable to produce finger movements or for whom afferent sensory feedback is no longer intact. We strongly agree with the Reviewer that any intended clinical application must carefully consider the specific input feature constraints dictated by the clinical cohort, and in turn impose appropriate and complimentary constraints on classifier parameters that may differ from the ones used in the present study. We now highlight this issue in the Discussion of the revised manuscript and relate our present findings to published clinical BCI work within this context.

One of the important and fascinating claims of the current study is that the "contextualization" of individual finger movements in a trained sequence specifically occurs during short rest periods in very early skill learning, echoing the recent theory of micro-offline learning proposed by the authors' group. Here, I think two points need to be clarified. First, the concept of "contextualization" is kept somewhat blurry throughout the text. It is only at the later part of the Discussion (around line #330 on page 13) that some potential mechanism for the "contextualization" is provided as "what-and-where" binding. Still, it is unclear what "contextualization" actually is in the current data, as the MEG signal analyzed is extracted from 0-200 msec after the keypress. If one thinks something is contextualizing an action, that contextualization should come earlier than the action itself.

The Reviewer requests that we: 1) more clearly define our use of the term “contextualization” and 2) provide the rationale for assessing it over a 200ms window aligned to the KeyDown event. This choice of window parameters means that the MEG activity used in our analysis was coincident with, rather than preceding, the actual keypresses. We define contextualization as the differentiation of representation for the identical movement embedded in different positions of a sequential skill. That is, representations of individual action elements progressively incorporate information about their relationship to the overall sequence structure as the skill is learned. We agree with the Reviewer that this can be appropriately interpreted as “what-and-where” binding. We now incorporate this definition in the Introduction of the revised manuscript as requested.

The window parameters for optimizing accurate decoding individual finger movements were determined using a grid search of the parameter space (a sliding window of variable width between 25-350 ms with 25 ms increments variably aligned from 0 to +100ms with 10ms increments relative to the KeyDown event). This approach generated 140 different temporal windows for each keypress for each participant, with the final parameter selection determined through comparison of the resulting performance between each decoder. Importantly, the decision to optimize for decoding accuracy placed an emphasis on keypress representations characterized by the most consistent and robust features shared across subjects, which in turn maximize statistical power in detecting common learning-related changes. In this case, the optimal window encompassed a 200ms epoch aligned to the KeyDown event (t₀ = 0 ms). We then asked if the representations (i.e. – spatial patterns of combined parcel- and voxel-space activity) of the same digit at two different sequence positions changed with practice within this optimal decoding window. Of course, our findings do not rule out the possibility that contextualization can also be found before or even after this time window, as we did not directly address this issue in the present study. Future work in our lab, as pointed out above, are investigating contextualization within different time windows tailored specifically for assessing sequence skill action planning, execution, evaluation and memory processes.

The second point is that the result provided by the authors is not yet convincing enough to support the claim that "contextualization" occurs during rest. In the original analysis, the authors presented the statistical significance regarding the correlation between the "offline" pattern differentiation and micro-offline skill gain (Figure 5. Supplement 1), as well as the larger "offline" distance than "online" distance (Figure 5B). However, this analysis looks like regressing two variables (monotonically) increasing as a function of the trial. Although some information in this analysis, such as what the independent/dependent variables were or how individual subjects were treated, was missing in the Methods, getting a statistically significant slope seems unsurprising in such a situation. Also, curiously, the same quantitative evidence was not provided for its "online" counterpart, and the authors only briefly mentioned in the text that there was no significant correlation between them. It may be true looking at the data in Figure 5A as the online representation distance looks less monotonically changing, but the classification accuracy presented in Figure 4C, which should reflect similar representational distance, shows a more monotonic increase up to the 11th trial. Further, the ways the "online" and "offline" representation distance was estimated seem to make them not directly comparable. While the "online" distance was computed using all the correct press data within each 10 sec of execution, the "offline" distance is basically computed by only two presses (i.e., the last index_OP5 vs. the first index_OP1 separated by 10 sec of rest). Theoretically, the distance between the neural activity patterns for temporally closer events tends to be closer than that between the patterns for temporally far-apart events. It would be fairer to use the distance between the first index_OP1 vs. the last index_OP5 within an execution period for "online" distance, as well.

The Reviewer suggests that the current data is not enough to show that contextualization occurs during rest and raises two important concerns: 1) the relationship between online contextualization and micro-online gains is not shown, and 2) the online distance was calculated differently from its offline counterpart (i.e. - instead of calculating the distance between last Index_OP5 and first Index_OP1 from a single trial, the distance was calculated for each sequence within a trial and then averaged).

We addressed the first concern by performing individual subject correlations between 1) contextualization changes during rest intervals and micro-offline gains; 2) contextualization changes during practice trials and micro-online gains, and 3) contextualization changes during practice trials and micro-offline gains (Figure 5 – figure supplement 4). We then statistically compared the resulting correlation coefficient distributions and found that within-subject correlations for contextualization changes during rest intervals and micro-offline gains were significantly higher than online contextualization and micro-online gains (t = 3.2827, p = 0.0015) and online contextualization and micro-offline gains (t = 3.7021, p = 5.3013e-04). These results are consistent with our interpretation that micro-offline gains are supported by contextualization changes during the inter-practice rest periods.

With respect to the second concern, we agree with the Reviewer that one limitation of the analysis comparing online versus offline changes in contextualization as presented in the original manuscript, is that it does not eliminate the possibility that any differences could simply be explained by the passage of time (which is smaller for the online analysis compared to the offline analysis). The Reviewer suggests an approach that addresses this issue, which we have now carried out. When quantifying online changes in contextualization from the first Index_OP1 the last Index_OP5 keypress in the same trial we observed no learning-related trend (Figure 5 – figure supplement 5, right panel). Importantly, offline distances were significantly larger than online distances regardless of the measurement approach and neither predicted online learning (Figure 5 – figure supplement 6).

A related concern regarding the control analysis, where individual values for max speed and the degree of online contextualization were compared (Figure 5 Supplement 3), is whether the individual difference is meaningful. If I understood correctly, the optimization of the decoding process (temporal window, feature inclusion/reduction, decoder, etc.) was performed for individual participants, and the same feature extraction was also employed for the analysis of representation distance (i.e., contextualization). If this is the case, the distances are individually differently calculated and they may need to be normalized relative to some stable reference (e.g., 1 vs. 4 or average distance within the control sequence presses) before comparison across the individuals.

The Reviewer makes a good point here. We have now implemented the suggested normalization procedure in the analysis provided in the revised manuscript.

Reviewer #3 (Public review):

Summary:

One goal of this paper is to introduce a new approach for highly accurate decoding of finger movements from human magnetoencephalography data via dimension reduction of a "multiscale, hybrid" feature space. Following this decoding approach, the authors aim to show that early skill learning involves "contextualization" of the neural coding of individual movements, relative to their position in a sequence of consecutive movements. Furthermore, they aim to show that this "contextualization" develops primarily during short rest periods interspersed with skill training and correlates with a performance metric which the authors interpret as an indicator of offline learning.

Strengths:

A clear strength of the paper is the innovative decoding approach, which achieves impressive decoding accuracies via dimension reduction of a "multi-scale, hybrid space". This hybrid-space approach follows the neurobiologically plausible idea of the concurrent distribution of neural coding across local circuits as well as large-scale networks. A further strength of the study is the large number of tested dimension reduction techniques and classifiers (though the manuscript reveals little about the comparison of the latter).

We appreciate the Reviewer’s comments regarding the paper’s strengths.

A simple control analysis based on shuffled class labels could lend further support to this complex decoding approach. As a control analysis that completely rules out any source of overfitting, the authors could test the decoder after shuffling class labels. Following such shuffling, decoding accuracies should drop to chance level for all decoding approaches, including the optimized decoder. This would also provide an estimate of actual chance-level performance (which is informative over and beyond the theoretical chance level). Furthermore, currently, the manuscript does not explain the huge drop in decoding accuracies for the voxel-space decoding (Figure 3B). Finally, the authors' approach to cortical parcellation raises questions regarding the information carried by varying dipole orientations within a parcel (which currently seems to be ignored?) and the implementation of the mean-flipping method (given that there are two dimensions - space and time - what do the authors refer to when they talk about the sign of the "average source", line 477?).

The Reviewer recommends that we: 1) conduct an additional control analysis on classifier performance using shuffled class labels, 2) provide a more detailed explanation regarding the drop in decoding accuracies for the voxel-space decoding following LDA dimensionality reduction (see Fig 3B), and 3) provide additional details on how problems related to dipole solution orientations were addressed in the present study.

In relation to the first point, we have now implemented a random shuffling approach as a control for the classification analyses. The results of this analysis indicated that the chance level accuracy was 22.12% (± SD 9.1%) for individual keypress decoding (4-class classification), and 18.41% (± SD 7.4%) for individual sequence item decoding (5-class classification), irrespective of the input feature set or the type of decoder used. Thus, the decoding accuracy observed with the final model was substantially higher than these chance levels.

Second, please note that the dimensionality of the voxel-space feature set is very high (i.e. – 15684). LDA attempts to map the input features onto a much smaller dimensional space (number of classes – 1; e.g. – 3 dimensions, for 4-class keypress decoding). Given the very high dimension of the voxel-space input features in this case, the resulting mapping exhibits reduced accuracy. Despite this general consideration, please refer to Figure 3—figure supplement 3, where we observe improvement in voxel-space decoder performance when utilizing alternative dimensionality reduction techniques.

The decoders constructed in the present study assess the average spatial patterns across time (as defined by the windowing procedure) in the input feature space. We now provide additional details in the Methods of the revised manuscript pertaining to the parcellation procedure and how the sign ambiguity problem was addressed in our analysis.

Weaknesses:

A clear weakness of the paper lies in the authors' conclusions regarding "contextualization". Several potential confounds, described below, question the neurobiological implications proposed by the authors and provide a simpler explanation of the results. Furthermore, the paper follows the assumption that short breaks result in offline skill learning, while recent evidence, described below, casts doubt on this assumption.

We thank the Reviewer for giving us the opportunity to address these issues in detail (see below).

The authors interpret the ordinal position information captured by their decoding approach as a reflection of neural coding dedicated to the local context of a movement (Figure 4). One way to dissociate ordinal position information from information about the moving effectors is to train a classifier on one sequence and test the classifier on other sequences that require the same movements, but in different positions (Kornysheva et al., 2019). In the present study, however, participants trained to repeat a single sequence (4-1-3-2-4). As a result, ordinal position information is potentially confounded by the fixed finger transitions around each of the two critical positions (first and fifth press). Across consecutive correct sequences, the first keypress in a given sequence was always preceded by a movement of the index finger (=last movement of the preceding sequence), and followed by a little finger movement. The last keypress, on the other hand, was always preceded by a ring finger movement, and followed by an index finger movement (=first movement of the next sequence). Figure 4 - Supplement 2 shows that finger identity can be decoded with high accuracy (>70%) across a large time window around the time of the key press, up to at least +/-100 ms (and likely beyond, given that decoding accuracy is still high at the boundaries of the window depicted in that figure). This time window approaches the keypress transition times in this study. Given that distinct finger transitions characterized the first and fifth keypress, the classifier could thus rely on persistent (or "lingering") information from the preceding finger movement, and/or "preparatory" information about the subsequent finger movement, in order to dissociate the first and fifth keypress. Currently, the manuscript provides no evidence that the context information captured by the decoding approach is more than a by-product of temporally extended, and therefore overlapping, but independent neural representations of consecutive keypresses that are executed in close temporal proximity - rather than a neural representation dedicated to context.

Such temporal overlap of consecutive, independent finger representations may also account for the dynamics of "ordinal coding"/"contextualization", i.e., the increase in 2-class decoding accuracy, across Day 1 (Figure 4C). As learning progresses, both tapping speed and the consistency of keypress transition times increase (Figure 1), i.e., consecutive keypresses are closer in time, and more consistently so. As a result, information related to a given keypress is increasingly overlapping in time with information related to the preceding and subsequent keypresses. The authors seem to argue that their regression analysis in Figure 5 - Figure Supplement 3 speaks against any influence of tapping speed on "ordinal coding" (even though that argument is not made explicitly in the manuscript). However, Figure 5 - Figure Supplement 3 shows inter-individual differences in a between-subject analysis (across trials, as in panel A, or separately for each trial, as in panel B), and, therefore, says little about the within-subject dynamics of "ordinal coding" across the experiment. A regression of trial-by-trial "ordinal coding" on trial-by-trial tapping speed (either within-subject or at a group-level, after averaging across subjects) could address this issue. Given the highly similar dynamics of "ordinal coding" on the one hand (Figure 4C), and tapping speed on the other hand (Figure 1B), I would expect a strong relationship between the two in the suggested within-subject (or group-level) regression. Furthermore, learning should increase the number of (consecutively) correct sequences, and, thus, the consistency of finger transitions. Therefore, the increase in 2-class decoding accuracy may simply reflect an increasing overlap in time of increasingly consistent information from consecutive keypresses, which allows the classifier to dissociate the first and fifth keypress more reliably as learning progresses, simply based on the characteristic finger transitions associated with each. In other words, given that the physical context of a given keypress changes as learning progresses - keypresses move closer together in time and are more consistently correct - it seems problematic to conclude that the mental representation of that context changes. To draw that conclusion, the physical context should remain stable (or any changes to the physical context should be controlled for).

The issues raised by Reviewer #3 here are similar to two issues raised by Reviewer #2 above. We agree they must both be carefully considered in any evaluation of our findings.

As both Reviewers pointed out, the classifiers in this study were trained and tested on keypresses performed while practicing a specific sequence (4-1-3-2-4). The study was designed this way as to avoid the impact of interference effects on learning dynamics. The cross-validated performance of classifiers on MEG data collected within the same session was 90.47% overall accuracy (4class; Figure 3C). We then tested classifier performance on data collected during a separate MEG session conducted approximately 24 hours later (Day 2; see Figure 3—supplement 3). We observed a reduction in overall accuracy rate to 87.11% when tested on MEG data recorded while participants performed the same learned sequence, and 79.44% when they performed several previously unpracticed sequences. This classification performance difference of 7.67% when tested on the Day 2 data could reflect the performance bias of the classifier for the trained sequence, possibly caused by mixed information from temporally close keypresses being incorporated into the feature weights.

Along these same lines, both Reviewers also raise the possibility that an increase in “ordinal coding/contextualization” with learning could simply reflect an increase in this mixing effect caused by faster typing speeds as opposed to an actual change in the underlying neural representation. The basic idea is that as correct sequences are generated at higher and higher speeds over training, MEG activity patterns related to the planning, execution, evaluation and memory of individual keypresses overlap more in time. Thus, increased overlap between the “4” and “1” keypresses (at the start of the sequence) and “2” and “4” keypresses (at the end of the sequence) could artefactually increase contextualization distances even if the underlying neural representations for the individual keypresses remain unchanged (assuming this mixing of representations is used by the classifier to differentially tag each index finger press). If this were the case, it follows that such mixing effects reflecting the ordinal sequence structure would also be observable in the distribution of decoder misclassifications. For example, “4” keypresses would be more likely to be misclassified as “1” or “2” keypresses (or vice versa) than as “3” keypresses. The confusion matrices presented in Figures 3C and 4B and Figure 3—figure supplement 3A in the previously submitted manuscript do not show this trend in the distribution of misclassifications across the four fingers.

Following this logic, it’s also possible that if the ordinal coding is largely driven by this mixing effect, the increased overlap between consecutive index finger keypresses during the 4-4 transition marking the end of one sequence and the beginning of the next one could actually mask contextualization-related changes to the underlying neural representations and make them harder to detect. In this case, a decoder tasked with separating individual index finger keypresses into two distinct classes based upon sequence position might show decreased performance with learning as adjacent keypresses overlapped in time with each other to an increasing extent. However, Figure 4C in our previously submitted manuscript does not support this possibility, as the 2-class hybrid classifier displays improved classification performance over early practice trials despite greater temporal overlap.

As noted in the above reply to Reviewer #2, we also conducted a new multivariate regression analysis to directly assess whether the neural representation distance score could be predicted by the 4-1, 2-4 and 4-4 keypress transition times observed for each complete correct sequence (both predictor and response variables were z-score normalized within-subject). The results of this analysis affirmed that the possible alternative explanation put forward by the Reviewer is not supported by our data (Adjusted R² = 0.00431; F = 5.62). We now include this new negative control analysis result in the revised manuscript.

Finally, the Reviewer hints that one way to address this issue would be to compare MEG responses before and after learning for sequences typed at a fixed speed. However, given that the speed-accuracy trade-off should improve with learning, a comparison between unlearned and learned skill states would dictate that the skill be evaluated at a very low fixed speed. Essentially, such a design presents the problem that the post-training test is evaluating the representation in the unlearned behavioral state that is not representative of the acquired skill. Thus, this approach would miss most learning effects on a task in which speed is the main learning metrics.

A similar difference in physical context may explain why neural representation distances ("differentiation") differ between rest and practice (Figure 5). The authors define "offline differentiation" by comparing the hybrid space features of the last index finger movement of a trial (ordinal position 5) and the first index finger movement of the next trial (ordinal position 1). However, the latter is not only the first movement in the sequence but also the very first movement in that trial (at least in trials that started with a correct sequence), i.e., not preceded by any recent movement. In contrast, the last index finger of the last correct sequence in the preceding trial includes the characteristic finger transition from the fourth to the fifth movement. Thus, there is more overlapping information arising from the consistent, neighbouring keypresses for the last index finger movement, compared to the first index finger movement of the next trial. A strong difference (larger neural representation distance) between these two movements is, therefore, not surprising, given the task design, and this difference is also expected to increase with learning, given the increase in tapping speed, and the consequent stronger overlap in representations for consecutive keypresses. Furthermore, initiating a new sequence involves pre-planning, while ongoing practice relies on online planning (Ariani et al., eNeuro 2021), i.e., two mental operations that are dissociable at the level of neural representation (Ariani et al., bioRxiv 2023).

The Reviewer argues that the comparison of last finger movement of a trial and the first in the next trial are performed in different circumstances and contexts. This is an important point and one we tend to agree with. For this task, the first sequence in a practice trial is pre-planned before the first keypress is performed. This occurs in a somewhat different context from the sequence iterations that follow, which involve temporally overlapping planning, execution and evaluation processes. The Reviewer is concerned about a difference in the temporal mixing effect issue raised above between the first and last keypresses performed in a trial. Please, note that since neural representations of individual actions are competitively queued during the pre-planning period in a manner that reflects the ordinal structure of the learned sequence (Kornysheva et al., 2019), mixing effects are most likely present also for the first keypress in a trial.

Separately, the Reviewer suggests that contextualization during early learning may reflect preplanning or online planning. This is an interesting proposal. Given the decoding time-window used in this investigation, we cannot dissect separate contributions of planning, memory and sensory feedback to contextualization. Taking advantage of the superior temporal resolution of MEG relative to fMRI tools, work under way in our lab is investigating decoding time-windows more appropriate to address each of these questions.

Given these differences in the physical context and associated mental processes, it is not surprising that "offline differentiation", as defined here, is more pronounced than "online differentiation". For the latter, the authors compared movements that were better matched regarding the presence of consistent preceding and subsequent keypresses (online differentiation was defined as the mean difference between all first vs. last index finger movements during practice). It is unclear why the authors did not follow a similar definition for "online differentiation" as for "micro-online gains" (and, indeed, a definition that is more consistent with their definition of "offline differentiation"), i.e., the difference between the first index finger movement of the first correct sequence during practice, and the last index finger of the last correct sequence. While these two movements are, again, not matched for the presence of neighbouring keypresses (see the argument above), this mismatch would at least be the same across "offline differentiation" and "online differentiation", so they would be more comparable.

This is the same point made earlier by Reviewer #2, and we agree with this assessment. As stated in the response to Reviewer #2 above, we have now carried out quantification of online contextualization using this approach and included it in the revised manuscript. We thank the Reviewer for this suggestion.

A further complication in interpreting the results regarding "contextualization" stems from the visual feedback that participants received during the task. Each keypress generated an asterisk shown above the string on the screen, irrespective of whether the keypress was correct or incorrect. As a result, incorrect (e.g., additional, or missing) keypresses could shift the phase of the visual feedback string (of asterisks) relative to the ordinal position of the current movement in the sequence (e.g., the fifth movement in the sequence could coincide with the presentation of any asterisk in the string, from the first to the fifth). Given that more incorrect keypresses are expected at the start of the experiment, compared to later stages, the consistency in visual feedback position, relative to the ordinal position of the movement in the sequence, increased across the experiment. A better differentiation between the first and the fifth movement with learning could, therefore, simply reflect better decoding of the more consistent visual feedback, based either on the feedback-induced brain response, or feedback-induced eye movements (the study did not include eye tracking). It is not clear why the authors introduced this complicated visual feedback in their task, besides consistency with their previous studies.

We strongly agree with the Reviewer that eye movements related to task engagement are important to rule out as a potential driver of the decoding accuracy or contextualizaton effect. We address this issue above in response to a question raised by Reviewer #1 about the impact of movement related artefacts on our findings.

First, the assumption the Reviewer makes here about the distribution of errors in this task is incorrect. On average across subjects, 2.32% ± 1.48% (mean ± SD) of all keypresses performed were errors, which were evenly distributed across the four possible keypress responses. While errors increased progressively over practice trials, they did so in proportion to the increase in correct keypresses, so that the overall ratio of correct-to-incorrect keypresses remained stable over the training session. Thus, the Reviewer’s assumptions that there is a higher relative frequency of errors in early trials, and a resulting systematic trend phase shift differences between the visual display updates (i.e. – a change in asterisk position above the displayed sequence) and the keypress performed is not substantiated by the data. To the contrary, the asterisk position on the display and the keypress being executed remained highly correlated over the entire training session. We now include a statement about the frequency and distribution of errors in the revised manuscript.

Given this high correlation, we firmly agree with the Reviewer that the issue of eye movement related artefacts is still an important one to address. Fortunately, we did collect eye movement data during the MEG recordings so were able to investigate this. As detailed in the response to Reviewer #1 above, we found that gaze positions and eye-movement velocity time-locked to visual display updates (i.e. – a change in asterisk position above the displayed sequence) did not reflect the asterisk location above chance levels (Overall cross-validated accuracy = 0.21817; see Author response image 1). Furthermore, an inspection of the eye position data revealed that most participants on most trials displayed random walk gaze patterns around a center fixation point, indicating that participants did not attend to the asterisk position on the display. This is consistent with intrinsic generation of the action sequence, and congruent with the fact that the display does not provide explicit feedback related to performance. As pointed out above, a similar real-world example would be manually inputting a long password into a secure online application. In this case, one intrinsically generates the sequence from memory and receives similar feedback about the password sequence position (also provided as asterisks), which is typically ignored by the user.

The minimal participant engagement with the visual display in this explicit sequence learning motor task (which is highly generative in nature) contrasts markedly with behavior observed when reactive responses to stimulus cues are needed in the serial reaction time task (SRTT). This is a crucial difference that must be carefully considered when comparing findings across studies using the two sequence learning tasks.

The authors report a significant correlation between "offline differentiation" and cumulative microoffline gains. However, it would be more informative to correlate trial-by-trial changes in each of the two variables. This would address the question of whether there is a trial-by-trial relation between the degree of "contextualization" and the amount of micro-offline gains - are performance changes (micro-offline gains) less pronounced across rest periods for which the change in "contextualization" is relatively low? Furthermore, is the relationship between micro-offline gains and "offline differentiation" significantly stronger than the relationship between micro-offline gains and "online differentiation"?

In response to a similar issue raised above by Reviewer #2, we now include new analyses comparing correlation magnitudes between (1) “online differentiation” vs micro-online gains, (2) “online differentiation” vs micro-offline gains and (3) “offline differentiation” and micro-offline gains (see Figure 5 – figure supplement  4, 5 and 6). These new analyses and results have been added to the revised manuscript. Once again, we thank both Reviewers for this suggestion.

The authors follow the assumption that micro-offline gains reflect offline learning.

We disagree with this statement. The original (Bonstrup et al., 2019) paper clearly states that micro-offline gains do not necessarily reflect offline learning in some cases and must be carefully interpreted based upon the behavioral context within which they are observed. Further, the paper lays out the conditions under which one can have confidence that micro-offline gains reflect offline learning. In fact, the excellent meta-analysis of (Pan & Rickard, 2015), which re-interprets the benefits of sleep in overnight skill consolidation from a “reactive inhibition” perspective, was a crucial resource in the experimental design of our initial study (Bonstrup et al., 2019), as well as in all our subsequent work. Pan & Rickard state:

“Empirically, reactive inhibition refers to performance worsening that can accumulate during a period of continuous training (Hull, 1943 . It tends to dissipate, at least in part, when brief breaks are inserted between blocks of training. If there are multiple performance-break cycles over a training session, as in the motor sequence literature, performance can exhibit a scalloped effect, worsening during each uninterrupted performance block but improving across blocks(Brawn et al., 2010; Rickard et al., 2008 . Rickard, Cai, Rieth, Jones, and Ard (2008 and Brawn, Fenn, Nusbaum, and Margoliash (2010 (Brawn et al., 2010; Rickard et al., 2008 demonstrated highly robust scalloped reactive inhibition effects using the commonly employed 30 s–30 s performance break cycle, as shown for Rickard et al.’s (2008 massed practice sleep group in Figure 2. The scalloped effect is evident for that group after the first few 30 s blocks of each session. The absence of the scalloped effect during the first few blocks of training in the massed group suggests that rapid learning during that period masks any reactive inhibition effect.”

Crucially, Pan & Rickard make several concrete recommendations for reducing the impact of the reactive inhibition confound on offline learning studies. One of these recommendations was to reduce practice times to 10s (most prior sequence learning studies up until that point had employed 30s long practice trials). They state:

“The traditional design involving 30 s-30 s performance break cycles should be abandoned given the evidence that it results in a reactive inhibition confound, and alternative designs with reduced performance duration per block used instead (Pan & Rickard, 2015 . One promising possibility is to switch to 10 s performance durations for each performance-break cycle Instead (Pan & Rickard, 2015 . That design appears sufficient to eliminate at least the majority of the reactive inhibition effect (Brawn et al., 2010; Rickard et al., 2008 .”

We mindfully incorporated recommendations from (Pan & Rickard, 2015) into our own study designs including 1) utilizing 10s practice trials and 2) constraining our analysis of micro-offline gains to early learning trials (where performance monotonically increases and 95% of overall performance gains occur), which are prior to the emergence of the “scalloped” performance dynamics that are strongly linked to reactive inhibition effects.

However, there is no direct evidence in the literature that micro-offline gains really result from offline learning, i.e., an improvement in skill level.

We strongly disagree with the Reviewer’s assertion that “there is no direct evidence in the literature that micro-offline gains really result from offline learning, i.e., an improvement in skill level.” The initial (Bonstrup et al., 2019) report was followed up by a large online crowd-sourcing study (Bonstrup et al., 2020). This second (and much larger) study provided several additional important findings supporting our interpretation of micro-offline gains in cases where the important behavioral conditions clarified above were met (see Author response image 4 below for further details on these conditions).

Author response image 4.

This Figure shows that micro-offline gains o ser ed in learning and nonlearning contexts are attri uted to different underl ing causes. Micro-offline and online changes relative to overall trial-by-trial learning. This figure is based on data from (Bonstrup et al., 2019). During early learning, micro-offline gains (red bars) closely track trial-by-trial performance gains (green line with open circle markers), with minimal contribution from micro-online gains (blue bars). The stated conclusion in Bönstrup et al. (2019) is that micro-offline gains only during this Early Learning stage reflect rapid memory consolidation (see also (Bonstrup et al., 2020)). After early learning, about practice trial 11, skill plateaus. This plateau skill period is characterized by a striking emergence of coupled (and relatively stable) micro-online drops and micro-offline increases. Bönstrup et al. (2019) as well as others in the literature (Brooks et al., 2024; Gupta & Rickard, 2022; Florencia Jacobacci et al., 2020), argue that micro-offline gains during the plateau period likely reflect recovery from inhibitory performance factors such as reactive inhibition or fatigue, and thus must be excluded from analyses relating micro-offline gains to skill learning. The Non-repeating groups in Experiments 3 and 4 from Das et al. (2024) suffer from a lack of consideration of these known confounds (end of Fig legend).

Evidence documented in that paper (Bonstrup et al., 2020) showed that micro-offline gains during early skill learning were: 1) replicable and generalized to subjects learning the task in their daily living environment (n=389); 2) equivalent when significantly shortening practice period duration, thus confirming that they are not a result of recovery from performance fatigue (n=118); 3) reduced (along with learning rates) by retroactive interference applied immediately after each practice period relative to interference applied after passage of time (n=373), indicating stabilization of the motor memory at a microscale of several seconds consistent with rapid consolidation; and 4) not modified by random termination of the practice periods, ruling out a contribution of predictive motor slowing (N = 71) (Bonstrup et al., 2020). Altogether, our findings were strongly consistent with the interpretation that micro-offline gains reflect memory consolidation supporting early skill learning. This is precisely the portion of the learning curve (Pan & Rickard, 2015) refer to when they state “…rapid learning during that period masks any reactive inhibition effect”.

This interpretation is further supported by brain imaging evidence linking known memory-related networks and consolidation mechanisms to micro-offline gains. First, we reported that the density of fast hippocampo-neocortical skill memory replay events increases approximately three-fold during early learning inter-practice rest periods with the density explaining differences in the magnitude of micro-offline gains across subjects (Buch et al., 2021). Second, Jacobacci et al. (2020) independently reproduced our original behavioral findings and reported BOLD fMRI changes in the hippocampus and precuneus (regions also identified in our MEG study (Buch et al., 2021)) linked to micro-offline gains during early skill learning. These functional changes were coupled with rapid alterations in brain microstructure in the order of minutes, suggesting that the same network that operates during rest periods of early learning undergoes structural plasticity over several minutes following practice (Deleglise et al., 2023). Crucial to this point, Chen et al. (2024) and Sjøgård et al (2024) provided direct evidence from intracranial EEG in humans linking sharp-wave ripple density during rest periods (which are known markers for neural replay (Buzsaki, 2015)) in the human hippocampus (80-120 Hz) to micro-offline gains during early skill learning.

Thus, there is now substantial converging evidence in humans across different indirect noninvasive and direct invasive recording techniques linking hippocampal activity, neural replay dynamics and offline performance gains in skill learning.

On the contrary, recent evidence questions this interpretation (Gupta & Rickard, npj Sci Learn 2022; Gupta & Rickard, Sci Rep 2024; Das et al., bioRxiv 2024). Instead, there is evidence that micro-offline gains are transient performance benefits that emerge when participants train with breaks, compared to participants who train without breaks, however, these benefits vanish within seconds after training if both groups of participants perform under comparable conditions (Das et al., bioRxiv 2024).

The recent work of (Gupta & Rickard, 2022, 2024) does not present any data that directly opposes our finding that early skill learning (Bonstrup et al., 2019) is expressed as micro-offline gains during rest breaks. These studies are an extension of the Rickard et al (2008) paper that employed a massed (30s practice followed by 30s breaks) vs spaced (10s practice followed by 10s breaks) experimental design to assess if recovery from reactive inhibition effects could account for performance gains measured after several minutes or hours. Gupta & Rickard (2022) added two additional groups (30s practice/10s break and 10s practice/10s break as used in the work from our group). The primary aim of the study was to assess whether it was more likely that changes in performance when retested 5 minutes after skill training (consisting of 12 practice trials for the massed groups and 36 practice trials for the spaced groups) had ended reflected memory consolidation effects or recovery from reactive inhibition effects. The Gupta & Rickard (2024) follow-up paper employed a similar design with the primary difference being that participants performed a fixed number of sequences on each trial as opposed to trials lasting a fixed duration. This was done to facilitate the fitting of a quantitative statistical model to the data.

To reiterate, neither study included any analysis of micro-online or micro-offline gains and did not include any comparison focused on skill gains during early learning trials (only at retest 5 min later). Instead, Gupta & Rickard (2022), reported evidence for reactive inhibition effects for all groups over much longer training periods than early learning. In fact, we reported the same findings for trials following the early learning period in our original 2019 paper (Bonstrup et al., 2019) (Author response image 4). Please, note that we also reported that cumulative microoffline gains over early learning did not correlate with overnight offline consolidation measured 24 hours later (Bonstrup et al., 2019) (see the Results section and further elaboration in the Discussion). We interpreted these findings as indicative that the mechanisms underlying offline gains over the micro-scale of seconds during early skill learning versus over minutes or hours very likely differ.

In the recent preprint from (Das et al., 2024), the authors make the strong claim that “micro-offline gains during early learning do not reflect offline learning” which is not supported by their own data. The authors hypothesize that if “micro-offline gains represent offline learning, participants should reach higher skill levels when training with breaks, compared to training without breaks”. The study utilizes a spaced vs. massed practice groups between-subjects design inspired by the reactive inhibition work from Rickard and others to test this hypothesis.

Crucially, their design incorporates only a small fraction of the training used in other investigations to evaluate early skill learning (Bonstrup et al., 2020; Bonstrup et al., 2019; Brooks et al., 2024; Buch et al., 2021; Deleglise et al., 2023; F. Jacobacci et al., 2020; Mylonas et al., 2024). A direct comparison between the practice schedule designs for the spaced and massed groups in Das et al., and the training schedule all participants experienced in the original Bönstrup et al. (2019) paper highlights this issue as well as several others (Author response image 5):

Author response image 5.

This figure shows (A) Comparison of Das et al. Spaced & Massed group training session designs, and the training session design from the original (Bonstrup et al., 2019) paper. Similar to the approach taken by Das et al., all practice is visualized as 10-second practice trials with a variable number (either 0, 1 or 30) of 10-second-long inter-practice rest intervals to allow for direct comparisons between designs. The two key takeaways from this comparison are that (1) the intervention differences (i.e. – practice schedules) between the Massed and Spaced groups from the Das et al. report are extremely small (less than 12% of the overall session schedule) (gaps in the red shaded area) and (2) the overall amount of practice is much less than compared to the design from the original Bönstrup report (Bonstrup et al., 2019) (which has been utilized in several subsequent studies). (B) Group-level learning curve data from Bönstrup et al. (2019) (Bonstrup et al., 2019) is used to estimate the performance range accounted for by the equivalent periods covering Test 1, Training 1 and Test 2 from Das et al (2024). Note that the intervention in the Das et al. study is limited to a period covering less than 50% of the overall learning range (end of figure legend).

Participants in the original (Bonstrup et al., 2019) experienced 157.14% more practice time and 46.97% less inter-practice rest time than the Spaced group in the Das et al. study (Author response image 5). Thus, the overall amount of practice and rest differ substantially between studies, with much more limited training occurring for participants in Das et al.

In addition, the training interventions (i.e. – the practice schedule differences between the Spaced and Massed groups) were designed in a manner that minimized any chance of effectively testing their hypothesis. First, the interventions were applied over an extremely short period relative to the length of the total training session (5% and 12% of the total training session for Massed and Spaced groups, respectively; see gaps in the red shaded area in Author response image 5). Second, the intervention was applied during a period in which only half of the known total learning occurs. Specifically, we know from Bönstrup et al. (2019) that only 46.57% of the total performance gains occur in the practice interval covered by Das et al Training 1 intervention. Thus, early skill learning as evaluated by multiple groups (Bonstrup et al., 2020; Bonstrup et al., 2019; Brooks et al., 2024; Buch et al., 2021; Deleglise et al., 2023; F. Jacobacci et al., 2020; Mylonas et al., 2024), is in the Das et al experiment amputated to about half.

Furthermore, a substantial amount of learning takes place during Das et al’s Test 1 and Test 2 periods (32.49% of total gains combined). The fact that substantial learning is known to occur over both the Test 1 (18.06%) and Test 2 (14.43%) intervals presents a fundamental problem described by Pan and Rickard (Pan & Rickard, 2015). They reported that averaging over intervals where substantial performance gains occur (i.e. – performance is not stable) inject crucial artefacts into analyses of skill learning:

“A large amount of averaging has the advantage of yielding more precise estimates of each subject’s pretest and posttest scores and hence more statistical power to detect a performance gain. However, calculation of gain scores using that strategy runs the risk that learning that occurs during the pretest and (or posttest periods (i.e., online learning is incorporated into the gain score (Rickard et al., 2008; Robertson et al., 2004 .”

The above statement indicates that the Test 1 and Test 2 performance scores from Das et al. (2024) are substantially contaminated by the learning rate within these intervals. This is particularly problematic if the intervention design results in different Test 2 learning rates between the two groups. This in fact, is apparent in their data (Figure 1C,E of the Das et al., 2024 preprint) as the Test 2 learning rate for the Spaced group is negative (indicating a unique interference effect observable only for this group). Specifically, the Massed group continues to show an increase in performance during Test 2 and 4 relative to the last 10 seconds of practice during Training 1 and 2, respectively, while the Spaced group displays a marked decrease. This post-training performance decrease for the Spaced group is in stark contrast to the monotonic performance increases observed for both groups at all other time-points. One possible cause could be related to the structure of the Test intervals, which include 20 seconds of uninterrupted practice. For the Spaced group, this effectively is a switch to a Massed practice environment (i.e., two 10-secondlong practice trials merged into one long trial), which interferes with greater Training 1 interval gains observed for the Space group. Interestingly, when statistical comparisons between the groups are made at the time-points when the intervention is present (Figure 1E) then the stated hypothesis, “If micro-offline gains represent offline learning, participants should reach higher skill levels when training with breaks, compared to training without breaks”, is confirmed.

In summary, the experimental design and analyses used by Das et al does not contradict the view that early skill learning is expressed as micro-offline gains during rest breaks. The data presented by Gupta and Rickard (2022, 2024) and Das et al. (2024) is in many ways more confirmatory of the constraints employed by our group and others with respect to experimental design, analysis and interpretation of study findings, rather than contradictory. Still, it does highlight a limitation of the current micro-online/offline framework, which was originally only intended to be applied to early skill learning over spaced practice schedules when reactive inhibition effects are minimized (Bonstrup et al., 2019; Pan & Rickard, 2015). Extrapolation of this current framework to postplateau performance periods, longer timespans, or non-learning situations (e.g. – the Nonrepeating groups from Das et al. (2024)), when reactive inhibition plays a more substantive role, is not warranted. Ultimately, it will be important to develop new paradigms allowing one to independently estimate the different coincident or antagonistic features (e.g. - memory consolidation, planning, working memory and reactive inhibition) contributing to micro-online and micro-offline gains during and after early skill learning within a unifying framework.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

(1) I found Figure 2B too small to be useful, as the actual elements of the cells are very hard to read.

We have removed the grid colormap panel (top-right) from Figure 2B. All of this colormap data is actually a subset of data presented in Figure 2 – figure supplement 1, so can still be found there.

Reviewer #2 (Recommendations for the authors):

(1) Related to the first point in my concerns, I would suggest the authors compare decoding accuracy between correct presses followed by correct vs. incorrect presses. This would clarify if the decoder is actually taking the MEG signal for subsequent press into account. I would also suggest the authors use pre-movement MEG features and post-movement features with shorter windows and compare each result with the results for the original post-movement MEG feature with a longer window.

The present study does not contain enough errors to perform the analysis proposed by the Reviewer. As noted above, we did re-examine our data and now report a new control regression analysis, all of which indicate that the proximity between keypresses does not explain contextualization effects.

(2) I was several times confused by the author's use of "neural representation of an action" or "sequence action representations" in understanding whether these terms refer to representation on the level of whole-brain, region (as defined by the specific parcellation used), or voxels. In fact, what is submitted to the decoder is some complicated whole-brain MEG feature (i.e., the "neural representation"), which is a hybrid of voxel and parcel features that is further dimension-reduced and not immediately interpretable. Clarifying this point early in the text and possibly using some more sensible terms, such as adding "brain-wise" before the "sequence action representation", would be the most helpful for the readers.

We now clarified this terminology in the revised manuscript.

(3) Although comparing many different ways in feature selection/reduction, time window selection, and decoder types is undoubtedly a meticulous work, the current version of the manuscript seems still lacking some explanation about the details of these methodological choices, like which decoding method was actually used to report the accuracy, whether or not different decoding methods were chosen for individual participants' data, how training data was selected (is it all of the correct presses in Day 1 data?), whether the frequency power or signal amplitude was used, and so on. I would highly appreciate these additional details in the Methods section.

The reported accuracies were based on linear discriminant analysis classifier. A comparison of different decoders (Figure 3 – figure supplement 4) shows LDA was the optimal choice.

Whether or not different decoding methods were chosen for individual participants' data

We selected the same decoder (LDA) performance to report the final accuracy.

How training data was selected (is it all of the correct presses in Day 1 data?),

Decoder training was conducted as a randomized split of the data (all correct keypresses of Day 1) into training (90%) and test (10%) samples for 8 iterations.

Whether the frequency power or signal amplitude was used

Signal amplitude was used for feature calculation.

(4) In terms of the Methods, please consider adding some references about the 'F1 score', the 'feature importance score,' and the 'MRMR-based feature ranking,' as the main readers of the current paper would not be from the machine learning community. Also, why did the LDA dimensionality reduction reduce accuracy specifically for the voxel feature?

We have now added the following statements to the Methods section that provide more detailed descriptions and references for these metrics:

“The F1 score, defined as the harmonic mean of the precision (percentage of true predictions that are actually true positive) and recall (percentage of true positives that were correctly predicted as true) scores, was used as a comprehensive metric for all one-versus-all keypress state decoders to assess class-wise performance that accounts for both false-positive and false-negative prediction tendencies [REF]. A weighted mean F1 score was then computed across all classes to assess the overall prediction performance of the multi-class model.”

and

“Feature Importance Scores

The relative contribution of source-space voxels and parcels to decoding performance (i.e. – feature importance score) was calculated using minimum redundant maximum relevance (MRMR) and highlighted in topography plots. MRMR, an approach that combines both relevance and redundancy metrics, ranked individual features based upon their significance to the target variable (i.e. – keypress state identity) prediction accuracy and their non-redundancy with other features.”

As stated in the Reviewer responses above, the dimensionality of the voxel-space feature set is very high (i.e. – 15684). LDA attempts to map the input features onto a much smaller dimensional space (number of classes-1; e.g. – 3 dimensions for 4-class keypress decoding). It is likely that the reduction in accuracy observed only for the voxel-space feature was due to the loss of relevant information during the mapping process that resulted in reduced accuracy. This reduction in accuracy for voxel-space decoding was specific to LDA. Figure 3—figure supplement 3 shows that voxel-space decoder performance actually improved when utilizing alternative dimensionality reduction techniques.

(5) Paragraph 9, lines #139-142: "Notably, decoding associated with index finger keypresses (executed at two different ordinal positions in the sequence) exhibited the highest number of misclassifications of all digits (N = 141 or 47.5% of all decoding errors; Figure 3C), raising the hypothesis that the same action could be differentially represented when executed at different learning state or sequence context locations."

This does not seem to be a fair comparison, as the index finger appears twice as many as the other fingers do in the sequence. To claim this, proper statistical analysis needs to be done taking this difference into account.

We thank the Reviewer for bringing this issue to our attention. We have now corrected this comparison to evaluate relative false negative and false positive rates between individual keypress state decoders, and have revised this statement in the manuscript as follows:

“Notably, decoding of index finger keypresses (executed at two different ordinal positions in the sequence) exhibited the highest false negative (0.116 per keypress) and false positive (0.043 per keypress) misclassification rates compared with all other digits (false negative rate range = [0.067 0.114]; false positive rate range = [0.020 0.037]; Figure 3C), raising the hypothesis that the same action could be differentially represented when executed within different contexts (i.e. - different learning states or sequence locations).”

(6) Finally, the authors could consider acknowledging in the Discussion that the contribution of micro-offline learning to genuine skill learning is still under debate (e.g., Gupta and Rickard, 2023; 2024; Das et al., bioRxiv, 2024).

We have added a paragraph in the Discussion that addresses this point.

Reviewer #3 (Recommendations for the authors):

In addition to the additional analyses suggested in the public review, I have the following suggestions/questions:

(1) Given that the authors introduce a new decoding approach, it would be very helpful for readers to see a distribution of window sizes and window onsets eventually used across individuals, at least for the optimized decoder.

We have now included a new supplemental figure (Figure 4 – figure Supplement 2) that provides this information.

(2) Please explain in detail how you arrived at the (interpolated?) group-level plot shown in Figure 1B, starting from the discrete single-trial keypress transition times. Also, please specify what the shading shows.

Instantaneous correct sequence speed (skill measure) was quantified as the inverse of time (in seconds) required to complete a single iteration of a correctly generated full 5-item sequence. Individual keypress responses were labeled as members of correct sequences if they occurred within a 5-item response pattern matching any possible circular shifts of the 5-item sequence displayed on the monitor (41324). This approach allowed us to quantify a measure of skill within each practice trial at the resolution of individual keypresses. The dark line indicates the group mean performance dynamics for each trial. The shaded region indicates the 95% confidence limit of the mean (see Methods).

(3) Similarly, please explain how you arrived at the group-level plot shown in Figure 1C. What are the different colored lines (rows) within each trial? How exactly did the authors reach the conclusion that KTT variability stabilizes by trial 6?

Figure 1C provides additional information to the correct sequence speed measure above, as it also tracks individual transition speed composition over learning. Figure 1C, thus, represents both changes in overall correct sequence speed dynamics (indicated by the overall narrowing of the horizontal speed lines moving from top to bottom) and the underlying composition of the individual transition patterns within and across trials. The coloring of the lines is a shading convention used to discriminate between different keypress transitions. These curves were sampled with 1ms resolution, as in Figure 1B. Addressing the underlying keypress transition patterns requires within-subject normalization before averaging across subjects. The distribution of KTTs was normalized to the median correct sequence time for each participant and centered on the mid-point for each full sequence iteration during early learning.

(4) Maybe I missed it, but it was not clear to me which of the tested classifiers was eventually used. Or was that individualized as well? More generally, a comparison of the different classifiers would be helpful, similar to the comparison of dimension reduction techniques.

We have now included a new supplemental figure that provides this information.

(5) Please add df and effect sizes to all statistics.

Done.

(6) Please explain in more detail your power calculation.

The study was powered to determine the minimum sample size needed to detect a significant change in skill performance following training using a one-sample t-test (two-sided; alpha = 0.05; 95% statistical power; Cohen’s D effect size = 0.8115 calculated from previously acquired data in our lab). The calculated minimum sample size was 22. The included study sample size (n = 27) exceeded this minimum.

This information is now included in the revised manuscript.

(7) The cut-off for the high-pass filter is unusually high and seems risky in terms of potential signal distortions (de Cheveigne, Neuron 2019). Why did the authors choose such a high cut-off?

The 1Hz high-pass cut-off frequency for the 1-150Hz band-pass filter applied to the continuous raw MEG data during preprocessing has been used in multiple previous MEG publications (Barratt et al., 2018; Brookes et al., 2012; Higgins et al., 2021; Seedat et al., 2020; Vidaurre et al., 2018).

(8) "Furthermore, the magnitude of offline contextualization predicted skill gains while online contextualization did not", lines 336/337 - where is that analysis?

Additional details pertaining to this analysis are now provided in the Results section (Figure 5 – figure supplement 4).

(9) How were feature importance scores computed?

We have now added a new subheading in the Methods section with a more detailed description of how feature importance scores were computed.

(10)  Please add x and y ticks plus tick labels to Figure 5 - Figure Supplement 3, panel A

Done

(11) Line 369, what does "comparable" mean in this context?

The sentence in the “Study Participants” part of the Methods section referred to here has now been revised for clarity.

(12) In lines 496/497, please specify what t=0 means (KeyDown event, I guess?).

Yes, the KeyDown event occurs at t = 0. This has now been clarified in the revised manuscript.

(13) Please specify consistent boundaries between alpha- and beta-bands (they are currently not consistent in the Results vs. Methods (14/15 Hz or 15/16 Hz)).

We thank the Reviewer for alerting us to this discrepancy caused by a typographic error in the Methods. We have now corrected this so that the alpha (8-14 Hz) and beta-band (15-24 Hz) frequency limits are described consistently throughout the revised manuscript.

References

Albouy, G., Fogel, S., King, B. R., Laventure, S., Benali, H., Karni, A., Carrier, J., Robertson, E. M., & Doyon, J. (2015). Maintaining vs. enhancing motor sequence memories: respective roles of striatal and hippocampal systems. Neuroimage, 108, 423-434. https://doi.org/10.1016/j.neuroimage.2014.12.049

Albouy, G., King, B. R., Maquet, P., & Doyon, J. (2013). Hippocampus and striatum: dynamics and interaction during acquisition and sleep-related motor sequence memory consolidation. Hippocampus, 23(11), 985-1004. https://doi.org/10.1002/hipo.22183 Albouy, G., Sterpenich, V., Vandewalle, G., Darsaud, A., Gais, S., Rauchs, G., Desseilles, M., Boly, M., Dang-Vu, T., Balteau, E., Degueldre, C., Phillips, C., Luxen, A., & Maquet, P. (2012). Neural correlates of performance variability during motor sequence acquisition. NeuroImage, 60(1), 324-331. https://doi.org/10.1016/j.neuroimage.2011.12.049

Andersen, R. A., & Buneo, C. A. (2002). Intentional maps in posterior parietal cortex. Annu Rev Neurosci, 25, 189-220. https://doi.org/10.1146/annurev.neuro.25.112701.142922 112701.142922 [pii]

Ashe, J., Lungu, O. V., Basford, A. T., & Lu, X. (2006). Cortical control of motor sequences. Curr Opin Neurobiol, 16(2), 213-221. http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citati on&list_uids=16563734

Bansal, A. K., Vargas-Irwin, C. E., Truccolo, W., & Donoghue, J. P. (2011). Relationships among low-frequency local field potentials, spiking activity, and three-dimensional reach and grasp kinematics in primary motor and ventral premotor cortices. J Neurophysiol, 105(4), 1603-1619. https://doi.org/10.1152/jn.00532.2010

Barratt, E. L., Francis, S. T., Morris, P. G., & Brookes, M. J. (2018). Mapping the topological organisation of beta oscillations in motor cortex using MEG. NeuroImage, 181, 831-844. https://doi.org/10.1016/j.neuroimage.2018.06.041

Bassett, D. S., Wymbs, N. F., Porter, M. A., Mucha, P. J., Carlson, J. M., & Grafton, S. T. (2011). Dynamic reconfiguration of human brain networks during learning. Proc Natl Acad Sci U S A, 108(18), 7641-7646. https://doi.org/10.1073/pnas.1018985108

Battaglia-Mayer, A., & Caminiti, R. (2019). Corticocortical Systems Underlying High-Order Motor Control. J Neurosci, 39(23), 4404-4421. https://doi.org/10.1523/JNEUROSCI.2094-18.2019

Berlot, E., Popp, N. J., & Diedrichsen, J. (2020). A critical re-evaluation of fMRI signatures of motor sequence learning. Elife, 9. https://doi.org/10.7554/eLife.55241

Bonstrup, M., Iturrate, I., Hebart, M. N., Censor, N., & Cohen, L. G. (2020). Mechanisms of offline motor learning at a microscale of seconds in large-scale crowdsourced data. NPJ Sci Learn, 5, 7. https://doi.org/10.1038/s41539-020-0066-9

Bonstrup, M., Iturrate, I., Thompson, R., Cruciani, G., Censor, N., & Cohen, L. G. (2019). A Rapid Form of Offline Consolidation in Skill Learning. Curr Biol, 29(8), 1346-1351 e1344. https://doi.org/10.1016/j.cub.2019.02.049

Brawn, T. P., Fenn, K. M., Nusbaum, H. C., & Margoliash, D. (2010). Consolidating the effects of waking and sleep on motor-sequence learning. J Neurosci, 30(42), 13977-13982. https://doi.org/10.1523/JNEUROSCI.3295-10.2010

Brookes, M. J., Woolrich, M. W., & Barnes, G. R. (2012). Measuring functional connectivity in MEG: a multivariate approach insensitive to linear source leakage. NeuroImage, 63(2), 910-920. https://doi.org/10.1016/j.neuroimage.2012.03.048

Brooks, E., Wallis, S., Hendrikse, J., & Coxon, J. (2024). Micro-consolidation occurs when learning an implicit motor sequence, but is not influenced by HIIT exercise. NPJ Sci Learn, 9(1), 23. https://doi.org/10.1038/s41539-024-00238-6

Buch, E. R., Claudino, L., Quentin, R., Bonstrup, M., & Cohen, L. G. (2021). Consolidation of human skill linked to waking hippocampo-neocortical replay. Cell Rep, 35(10), 109193. https://doi.org/10.1016/j.celrep.2021.109193

Buneo, C. A., & Andersen, R. A. (2006). The posterior parietal cortex: sensorimotor interface for the planning and online control of visually guided movements. Neuropsychologia, 44(13), 2594-2606. https://doi.org/10.1016/j.neuropsychologia.2005.10.011

Buzsaki, G. (2015). Hippocampal sharp wave-ripple: A cognitive biomarker for episodic memory and planning. Hippocampus, 25(10), 1073-1188. https://doi.org/10.1002/hipo.22488

Chen, P.-C., Stritzelberger, J., Walther, K., Hamer, H., & Staresina, B. P. (2024). Hippocampal ripples during offline periods predict human motor sequence learning. bioRxiv, 2024.2010.2006.614680. https://doi.org/10.1101/2024.10.06.614680

Churchland, M. M., Cunningham, J. P., Kaufman, M. T., Foster, J. D., Nuyujukian, P., Ryu, S. I., & Shenoy, K. V. (2012). Neural population dynamics during reaching. Nature, 487(7405), 51-56. https://doi.org/10.1038/nature11129

Classen, J., Liepert, J., Wise, S. P., Hallett, M., & Cohen, L. G. (1998). Rapid plasticity of human cortical movement representation induced by practice. J Neurophysiol, 79(2), 1117-1123. http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citati on&list_uids=9463469

Colclough, G. L., Brookes, M. J., Smith, S. M., & Woolrich, M. W. (2015). A symmetric multivariate leakage correction for MEG connectomes. NeuroImage, 117, 439-448. https://doi.org/10.1016/j.neuroimage.2015.03.071

Colclough, G. L., Woolrich, M. W., Tewarie, P. K., Brookes, M. J., Quinn, A. J., & Smith, S. M. (2016). How reliable are MEG resting-state connectivity metrics? NeuroImage, 138, 284-293. https://doi.org/10.1016/j.neuroimage.2016.05.070

Das, A., Karagiorgis, A., Diedrichsen, J., Stenner, M.-P., & Azanon, E. (2024). “Micro-offline gains” convey no benefit for motor skill learning. bioRxiv, 2024.2007.2011.602795. https://doi.org/10.1101/2024.07.11.602795

Deleglise, A., Donnelly-Kehoe, P. A., Yeffal, A., Jacobacci, F., Jovicich, J., Amaro, E., Jr., Armony, J. L., Doyon, J., & Della-Maggiore, V. (2023). Human motor sequence learning drives transient changes in network topology and hippocampal connectivity early during memory consolidation. Cereb Cortex, 33(10), 6120-6131. https://doi.org/10.1093/cercor/bhac489

Doyon, J., Bellec, P., Amsel, R., Penhune, V., Monchi, O., Carrier, J., Lehéricy, S., & Benali, H. (2009). Contributions of the basal ganglia and functionally related brain structures to motor learning. [Review]. Behavioural brain research, 199(1), 61-75. https://doi.org/10.1016/j.bbr.2008.11.012

Doyon, J., Song, A. W., Karni, A., Lalonde, F., Adams, M. M., & Ungerleider, L. G. (2002). Experience-dependent changes in cerebellar contributions to motor sequence learning. Proc Natl Acad Sci U S A, 99(2), 1017-1022. http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citati on&list_uids=11805340

Euston, D. R., Gruber, A. J., & McNaughton, B. L. (2012). The role of medial prefrontal cortex in memory and decision making. Neuron, 76(6), 1057-1070. https://doi.org/10.1016/j.neuron.2012.12.002

Euston, D. R., Tatsuno, M., & McNaughton, B. L. (2007). Fast-forward playback of recent memory sequences in prefrontal cortex during sleep. Science, 318(5853), 1147-1150. https://doi.org/10.1126/science.1148979

Flint, R. D., Ethier, C., Oby, E. R., Miller, L. E., & Slutzky, M. W. (2012). Local field potentials allow accurate decoding of muscle activity. J Neurophysiol, 108(1), 18-24. https://doi.org/10.1152/jn.00832.2011

Frankland, P. W., & Bontempi, B. (2005). The organization of recent and remote memories. Nat Rev Neurosci, 6(2), 119-130. https://doi.org/10.1038/nrn1607

Gais, S., Albouy, G., Boly, M., Dang-Vu, T. T., Darsaud, A., Desseilles, M., Rauchs, G., Schabus, M., Sterpenich, V., Vandewalle, G., Maquet, P., & Peigneux, P. (2007). Sleep transforms the cerebral trace of declarative memories. Proc Natl Acad Sci U S A, 104(47), 1877818783. https://doi.org/10.1073/pnas.0705454104

Grafton, S. T., Mazziotta, J. C., Presty, S., Friston, K. J., Frackowiak, R. S., & Phelps, M. E. (1992). Functional anatomy of human procedural learning determined with regional cerebral blood flow and PET. J Neurosci, 12(7), 2542-2548.

Grover, S., Wen, W., Viswanathan, V., Gill, C. T., & Reinhart, R. M. G. (2022). Long-lasting, dissociable improvements in working memory and long-term memory in older adults with repetitive neuromodulation. Nat Neurosci, 25(9), 1237-1246. https://doi.org/10.1038/s41593-022-01132-3

Gupta, M. W., & Rickard, T. C. (2022). Dissipation of reactive inhibition is sufficient to explain post-rest improvements in motor sequence learning. NPJ Sci Learn, 7(1), 25. https://doi.org/10.1038/s41539-022-00140-z

Gupta, M. W., & Rickard, T. C. (2024). Comparison of online, offline, and hybrid hypotheses of motor sequence learning using a quantitative model that incorporate reactive inhibition. Sci Rep, 14(1), 4661. https://doi.org/10.1038/s41598-024-52726-9

Hardwick, R. M., Rottschy, C., Miall, R. C., & Eickhoff, S. B. (2013). A quantitative metaanalysis and review of motor learning in the human brain. NeuroImage, 67, 283-297. https://doi.org/10.1016/j.neuroimage.2012.11.020

Heusser, A. C., Poeppel, D., Ezzyat, Y., & Davachi, L. (2016). Episodic sequence memory is supported by a theta-gamma phase code. Nat Neurosci, 19(10), 1374-1380. https://doi.org/10.1038/nn.4374

Higgins, C., Liu, Y., Vidaurre, D., Kurth-Nelson, Z., Dolan, R., Behrens, T., & Woolrich, M. (2021). Replay bursts in humans coincide with activation of the default mode and parietal alpha networks. Neuron, 109(5), 882-893 e887. https://doi.org/10.1016/j.neuron.2020.12.007

Hikosaka, O., Nakamura, K., Sakai, K., & Nakahara, H. (2002). Central mechanisms of motor skill learning. Curr Opin Neurobiol, 12(2), 217-222. http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citati on&list_uids=12015240

Jacobacci, F., Armony, J. L., Yeffal, A., Lerner, G., Amaro, E., Jr., Jovicich, J., Doyon, J., & Della-Maggiore, V. (2020). Rapid hippocampal plasticity supports motor sequence learning. Proc Natl Acad Sci U S A, 117(38), 23898-23903. https://doi.org/10.1073/pnas.2009576117

Jacobacci, F., Armony, J. L., Yeffal, A., Lerner, G., Amaro Jr, E., Jovicich, J., Doyon, J., & DellaMaggiore, V. (2020). Rapid hippocampal plasticity supports motor sequence learning.

Proceedings of the National Academy of Sciences, 117(38), 23898-23903. Karni, A., Meyer, G., Jezzard, P., Adams, M. M., Turner, R., & Ungerleider, L. G. (1995). Functional MRI evidence for adult motor cortex plasticity during motor skill learning. Nature, 377(6545), 155-158. https://doi.org/10.1038/377155a0

Kennerley, S. W., Sakai, K., & Rushworth, M. F. (2004). Organization of action sequences and the role of the pre-SMA. J Neurophysiol, 91(2), 978-993. https://doi.org/10.1152/jn.00651.2003 00651.2003 [pii]

Kleim, J. A., Barbay, S., & Nudo, R. J. (1998). Functional reorganization of the rat motor cortex following motor skill learning. J Neurophysiol, 80, 3321-3325.

Kornysheva, K., Bush, D., Meyer, S. S., Sadnicka, A., Barnes, G., & Burgess, N. (2019). Neural Competitive Queuing of Ordinal Structure Underlies Skilled Sequential Action. Neuron, 101(6), 1166-1180 e1163. https://doi.org/10.1016/j.neuron.2019.01.018

Lee, S. H., Jin, S. H., & An, J. (2019). The difference in cortical activation pattern for complex motor skills: A functional near- infrared spectroscopy study. Sci Rep, 9(1), 14066. https://doi.org/10.1038/s41598-019-50644-9

Lisman, J. E., & Jensen, O. (2013). The theta-gamma neural code. Neuron, 77(6), 1002-1016. https://doi.org/10.1016/j.neuron.2013.03.007

Mollazadeh, M., Aggarwal, V., Davidson, A. G., Law, A. J., Thakor, N. V., & Schieber, M. H. (2011). Spatiotemporal variation of multiple neurophysiological signals in the primary motor cortex during dexterous reach-to-grasp movements. J Neurosci, 31(43), 15531-15543. https://doi.org/10.1523/JNEUROSCI.2999-11.2011

Molle, M., & Born, J. (2009). Hippocampus whispering in deep sleep to prefrontal cortex--for good memories? Neuron, 61(4), 496-498. https://doi.org/10.1016/j.neuron.2009.02.002

Morris, R. G. M. (2006). Elements of a neurobiological theory of hippocampal function: the role of synaptic plasticity, synaptic tagging and schemas. [Review]. The European journal of neuroscience, 23(11), 2829-2846. https://doi.org/10.1111/j.1460-9568.2006.04888.x

Mylonas, D., Schapiro, A. C., Verfaellie, M., Baxter, B., Vangel, M., Stickgold, R., & Manoach, D. S. (2024). Maintenance of Procedural Motor Memory across Brief Rest Periods Requires the Hippocampus. J Neurosci, 44(14). https://doi.org/10.1523/JNEUROSCI.1839-23.2024

Pan, S. C., & Rickard, T. C. (2015). Sleep and motor learning: Is there room for consolidation? Psychol Bull, 141(4), 812-834. https://doi.org/10.1037/bul0000009

Penhune, V. B., & Steele, C. J. (2012). Parallel contributions of cerebellar, striatal and M1 mechanisms to motor sequence learning. Behav. Brain Res., 226(2), 579-591. https://doi.org/10.1016/j.bbr.2011.09.044

Qin, Y. L., McNaughton, B. L., Skaggs, W. E., & Barnes, C. A. (1997). Memory reprocessing in corticocortical and hippocampocortical neuronal ensembles. Philos Trans R Soc Lond B Biol Sci, 352(1360), 1525-1533. https://doi.org/10.1098/rstb.1997.0139

Rickard, T. C., Cai, D. J., Rieth, C. A., Jones, J., & Ard, M. C. (2008). Sleep does not enhance motor sequence learning. J Exp Psychol Learn Mem Cogn, 34(4), 834-842. https://doi.org/10.1037/0278-7393.34.4.834

Robertson, E. M., Pascual-Leone, A., & Miall, R. C. (2004). Current concepts in procedural consolidation. Nat Rev Neurosci, 5(7), 576-582. http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citati on&list_uids=15208699

Sawamura, D., Sakuraba, S., Suzuki, Y., Asano, M., Yoshida, S., Honke, T., Kimura, M., Iwase, Y., Horimoto, Y., Yoshida, K., & Sakai, S. (2019). Acquisition of chopstick-operation skills with the non-dominant hand and concomitant changes in brain activity. Sci Rep, 9(1), 20397. https://doi.org/10.1038/s41598-019-56956-0

Schendan, H. E., Searl, M. M., Melrose, R. J., & Stern, C. E. (2003). An FMRI study of the role of the medial temporal lobe in implicit and explicit sequence learning. Neuron, 37(6), 1013-1025. https://doi.org/10.1016/s0896-6273(03)00123-5

Seedat, Z. A., Quinn, A. J., Vidaurre, D., Liuzzi, L., Gascoyne, L. E., Hunt, B. A. E., O'Neill, G. C., Pakenham, D. O., Mullinger, K. J., Morris, P. G., Woolrich, M. W., & Brookes, M. J. (2020). The role of transient spectral 'bursts' in functional connectivity: A magnetoencephalography study. NeuroImage, 209, 116537. https://doi.org/10.1016/j.neuroimage.2020.116537

Shadmehr, R., & Holcomb, H. H. (1997). Neural correlates of motor memory consolidation. Science, 277, 821-824.

Sjøgård, M., Baxter, B., Mylonas, D., Driscoll, B., Kwok, K., Tolosa, A., Thompson, M., Stickgold, R., Vangel, M., Chu, C., & Manoach, D. S. (2024). Hippocampal ripples mediate motor learning during brief rest breaks in humans. bioRxiv. https://doi.org/10.1101/2024.05.02.592200

Srinivas, S., Sarvadevabhatla, R. K., Mopuri, K. R., Prabhu, N., Kruthiventi, S. S. S., & Babu, R. V. (2016). A Taxonomy of Deep Convolutional Neural Nets for Computer Vision [Technology Report]. Frontiers in Robotics and AI, 2. https://doi.org/10.3389/frobt.2015.00036

Sterpenich, V., Albouy, G., Darsaud, A., Schmidt, C., Vandewalle, G., Dang Vu, T. T., Desseilles, M., Phillips, C., Degueldre, C., Balteau, E., Collette, F., Luxen, A., & Maquet, P. (2009). Sleep promotes the neural reorganization of remote emotional memory. J Neurosci, 29(16), 5143-5152. https://doi.org/10.1523/JNEUROSCI.0561-09.2009

Toni, I., Ramnani, N., Josephs, O., Ashburner, J., & Passingham, R. E. (2001). Learning arbitrary visuomotor associations: temporal dynamic of brain activity. Neuroimage, 14(5), 10481057. http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citati on&list_uids=11697936

Toni, I., Thoenissen, D., & Zilles, K. (2001). Movement preparation and motor intention. NeuroImage, 14(1 Pt 2), S110-117. https://doi.org/10.1006/nimg.2001.0841

Tse, D., Langston, R. F., Kakeyama, M., Bethus, I., Spooner, P. A., Wood, E. R., Witter, M. P., & Morris, R. G. (2007). Schemas and memory consolidation. Science, 316(5821), 76-82. https://doi.org/10.1126/science.1135935

van Kesteren, M. T., Fernandez, G., Norris, D. G., & Hermans, E. J. (2010). Persistent schemadependent hippocampal-neocortical connectivity during memory encoding and postencoding rest in humans. Proc Natl Acad Sci U S A, 107(16), 7550-7555. https://doi.org/10.1073/pnas.0914892107

van Kesteren, M. T., Ruiter, D. J., Fernandez, G., & Henson, R. N. (2012). How schema and novelty augment memory formation. Trends Neurosci, 35(4), 211-219. https://doi.org/10.1016/j.tins.2012.02.001

Vidaurre, D., Hunt, L. T., Quinn, A. J., Hunt, B. A. E., Brookes, M. J., Nobre, A. C., & Woolrich, M. W. (2018). Spontaneous cortical activity transiently organises into frequency specific phase-coupling networks. Nat Commun, 9(1), 2987. https://doi.org/10.1038/s41467-01805316-z

Wagner, A. D., Schacter, D. L., Rotte, M., Koutstaal, W., Maril, A., Dale, A. M., Rosen, B. R., & Buckner, R. L. (1998). Building memories: remembering and forgetting of verbal experiences as predicted by brain activity. [Comment]. Science (New York, N.Y.), 281(5380), 1188-1191. http://eutils.ncbi.nlm.nih.gov/entrez/eutils/elink.fcgi?dbfrom=pubmed&id=9712582 &retmode=ref&cmd=prlinks

Wolpert, D. M., Goodbody, S. J., & Husain, M. (1998). Maintaining internal representations: the role of the human superior parietal lobe. Nat Neurosci, 1(6), 529-533. https://doi.org/10.1038/2245

Author response:

The following is the authors’ response to the original reviews.

Public Reviews:

Reviewer #1 (Public Review):

Based on previous publications suggesting a potential role for miR-26b in the pathogenesis of metabolic dysfunction-associated steatohepatitis (MASH), the researchers aim to clarify its function in hepatic health and explore the therapeutical potential of lipid nanoparticles (LNPs) to treat this condition. First, they employed both whole-body and myeloid cell-specific miR-26b KO mice and observed elevated hepatic steatosis features in these mice compared to WT controls when subjected to WTD. Moreover, livers from whole-body miR-26b KO mice also displayed increased levels of inflammation and fibrosis markers. Kinase activity profiling analyses revealed distinct alterations, particularly in kinases associated with inflammatory pathways, in these samples. Treatment with LNPs containing miR-26b mimics restored lipid metabolism and kinase activity in these animals. Finally, similar anti-inflammatory effects were observed in the livers of individuals with cirrhosis, whereas elevated miR-26b levels were found in the plasma of these patients in comparison with healthy control. Overall, the authors conclude that miR-26b plays a protective role in MASH and that its delivery via LNPs efficiently mitigates MASH development.

The study has some strengths, most notably, its employ of a combination of animal models, analyses of potential underlying mechanisms, as well as innovative treatment delivery methods with significant promise. However, it also presents numerous weaknesses that leave the research work somewhat incomplete. The precise role of miR-26b in a human context remains elusive, hindering direct translation to clinical practice. Additionally, the evaluation of the kinase activity, although innovative, does not provide a clear molecular mechanisms-based explanation behind the protective role of this miRNA.

Therefore, to fortify the solidity of their conclusions, these concerns require careful attention and resolution. Once these issues are comprehensively addressed, the study stands to make a significant impact on the field.

We would like the reviewer for his/her careful evaluation of our manuscript and appreciate his/her appraisal for the strengths of our study. Regarding the weaknesses, we have addressed these as good as possible during the revision of our manuscript.

We can already state that miR-26b has clear anti-inflammatory effects on human liver slices, which is in line with our results demonstrating that miR-26b plays a protective role in MASH development in mice. The notion that patients with liver cirrhosis have increasing plasma levels of miR-26b, seems contradictory at first glance. However, we believe that this increased miR-26b expression is a compensatory mechanism to counteract the MASH/cirrhotic effects. However, the exact source of this miR-26b remains to be elucidated in future studies.

The performed kinase activity analysis revealed that miR-26b affects kinases that particularly play an important role in inflammation and angiogenesis. Strikingly and supporting these data, these effects could be inverted again by LNP treatment. Combined, these results already provide strong mechanistic insights on molecular and intracellular signalling level. Although the exact target of miR-26b remains elusive and its identification is probably beyond the scope of the current manuscript due to its complexity, we believe that the kinase activity results already provide a solid mechanistic basis.

Reviewer #1 (Recommendations For The Authors):

A list of recommendations for the authors is presented below:

(1) The title should emphasize that the majority of experiments were conducted in mice to accurately reflect the scope of the study.

As suggested we have updated our title to include the statement that we primarily used a murine model:

“MicroRNA-26b protects against MASH development in mice and can be efficiently targeted with lipid nanoparticles.”

(2) It would be useful to know more about miR-26b function, including its target genes, tissue-specific expression, and tissue vs. circulating levels. Is it expected that the two strains of the miRNA (i.e., -3p and -5p) act this similarly? Also, miR-26b expression in the liver of individuals with cirrhosis should be determined.

The function of miR-26b is still rather elusive, making functional studies using this miR very interesting. In a previous study, describing our used mouse model (Van der Vorst et al. BMC Genom Data, 2021) we have eluded several functions of miR-26b that are already investigated. This was particularly already described in carcinogenesis and the neurological field.

Target gene wise, there are already several targets described in miRbase. However, for our experiments we feel that determination of the specific target genes is beyond the scope of the current manuscript and rather a focus of follow-up projects.

Regarding the expression of miR-26b, the liver and blood have rather high and similar expressions of both miR-26b-3p and miR-26b-5p as shown in Author response image 1.

Author response image 1.

Expression of miR-26b-3p and -5p. Expression of miR-26b-3p (left) and miR-26b-5p (right), generated by using the miRNATissueAtlas 2025 (Rishik et al. Nucleic Acids Research, 2024). Unfortunately, due to restrictions in tissue availability and the lack of stored RNA samples, we are unable to measure miR-26b expression in the human livers. However, based on the potency of the miR-26b mimic loaded LNPs in the mice (Revised Supplemental Figure 2A-B), we are confident that these LNPs also resulted in a overexpression of miR-26b in the human livers.

(3) Please explain the rationale behind primarily using whole-body miR-26b KO mice rather than the myeloid cell-specific KO model for the studies.

The main goal of our study is the elucidation of the general role of miR-26b in MASH formation. Therefore, we decided to primarily focus on the whole-body KO model. While we used the myeloid cell-specific KO model to highlight that myeloid cells play an important role in the observed phenotypes, we believe the whole-body KO model is more appropriate as main focus, particularly also in light of the used LNP targeting that also provides a whole-body approach. Furthermore, this focus on the whole-body model also reflects a more therapeutically relevant approach.

(4) The authors claim that treatment with LNPs containing miR-26b "replenish the miR-26b level in the whole-body deficient mouse" but the results of this observation are not presented.

This is indeed a valid point that we have now addressed. We have measured the mir26b-3p and mir26b-5p expression levels in livers from mice after 4-week WTD with simultaneous injection with either empty LNPs as vehicle control (eLNP) or LNPs containing miR-26b mimics (mLNP) every 3 days. As shown in Revised Supplemental Figure 2A-B, mLNP treatment clearly results in an overexpression of the mir26b in the livers of these mice. We have rephrased the text accordingly by stating that mLNP results in an “overexpression” rather than “replenishment”.

(5) The number of 3 human donors for the precision-cut liver slices is clearly insufficient and clinical parameters need to be shown. Additionally, inconsistencies in individual values in Figures 8B-E need clarification.

Unfortunately, due to restrictions in tissue availability, we are unable to increase our n-number for these experiments. Clinical parameters are not available, but the liver slices were from healthy tissue.

We have performed these experiments in duplicates for each individual donor. We have now specified this also in the figure legend to explain the individual values in the graphs:

“…(3 individual donors, cultured in duplicates).”

(6) Figure 2D: Please include representative images.

As suggested we have included representative images in our revised manuscript.

(7) Address discrepancies in the findings across different experimental settings. For example, the expression levels of the lipid metabolism-related genes are not significantly modulated in whole-body miR-26b KO mice (except for Sra), but they are in the myeloid cell-specific model (but not Sra), and none of them are restored after LNPs injections.

Although Cd36 is not significantly increased in the whole-body miR-26b KO mice, there is a clear tendency towards increased expression, which is now also validated on protein level (Revised Figure 1K-L). In the myeloid cell-specific model we see a similar tendency, although the gene expression difference of Sra is not significantly changed. This could be due to the difference in the model, since only myeloid cells are affected, suggesting that the effects on Sra are to a large extend driven by non-myeloid cells. This would also fit to the tendency to decreased Sra expression in the mimic-LNP treated mice. Due to the larger variation, this difference did not reach significance, which is rather a statistical issue due to relatively small n-numbers. At this moment, we cannot exclude that these receptors are differentially regulated by different cell-types. For this, future studies are needed focussing on cell-specific targeting of miR-26b in somatic cells, like hepatocytes.

(8) Figure 4A the images are not representative of the quantification.

We have selected another representative image that is exactly reflecting the average Sirius red positive area, to reflect the quantification appropriately.

(9) Figures 5 and 7: Are there not significantly decreased/increased kinases? A deeper analysis of these kinase alterations is necessary to understand how miR-26b exerts its role. A comparison analysis of these two datasets might clarify this regard.

We indeed very often see in these kinome analysis that the general tendency of kinase activity is unidirectional. We believe that this is caused by the highly interconnected nature of kinases. Activation of one signalling cascade will also results in the activation of many other cascades. However, it is interesting to see which pathways are affected in our study and we find it particularly interesting to see that the tendencies is exactly opposite between both comparisons as KO vs. WT shows increase kinase activities, while KO-LNP vs. KO shows a decrease again. Further showing that the method is reflecting a true biological effect that is mediated by miR26b.

(10) Determinations of the effect of LNPs containing miR-26b in the KO mice are limited to only a few observations (that are not only significant). More extensive findings are needed to conclusively demonstrate the effectiveness of this treatment method. Similar to the experiments with human liver samples (Figures 8A-E).

We have now elaborated our observations in the mouse model using LNPs by also analysing the effects on inflammation and fibrosis. However, it seems that the treatment time was not long enough to see pronounced changes on these later stages of disease development. Interestingly, the expression of Tgfb was significantly reduced, suggesting at least that the LNPs on genetic levels have an effect already on fibrotic processes. Thereby, it can be suggested that longer mLNP treatment may result in more effects on protein level as well, which remains to be determined in future studies.

Unfortunately, due to restrictions in tissue availability, we are unable to increase our n-number or read-outs for these experiments at this moment.

(11) In Figures 8F-H, the observed increase in circulating miR-26b levels in the plasma of cirrhotic individuals seems contradictory to its proposed protective role. This discrepancy requires clarification.

In the revised discussion (second to last paragraph), we have now elaborated more on the findings in the plasma of cirrhotic individuals in comparison to our murine in-vivo results, to highlight and discuss this discrepancy.

(12) Figures 8F-H legend mentions using 8-11 patients per group, but the methods section lacks corresponding information about these individuals.

These patients, together with inclusion/exclusion criteria and definition of cirrhosis are described in the method section 2.14.

(13) Figure 8G has 7 data points in the cirrhosis group, instead of 8. Any data exclusion should be justified in the methods section.

As defined in method section 2.15, we have identified outliers using the ROUT = 1 method, which is the reason why Figure 8G only has 7 data points instead of 8.

Reviewer #2 (Public Review):

Summary:

This manuscript by Peters, Rakateli, et al. aims to characterize the contribution of miR-26b in a mouse model of metabolic dysfunction-associated steatohepatitis (MASH) generated by a Western-type diet on the background of Apoe knock-out. In addition, the authors provide a rescue of the miR-26b using lipid nanoparticles (LNPs), with potential therapeutic implications. In addition, the authors provide useful insights into the role of macrophages and some validation of the effect of miR-26b LNPs on human liver samples.

Strengths:

The authors provide a well-designed mouse model, that aims to characterize the role of miR-26b in a mouse model of metabolic dysfunction-associated steatohepatitis (MASH) generated by a Western-type diet on the background of Apoe knock-out. The rescue of the phenotypes associated with the model used using miR-26b using lipid nanoparticles (LNPs) provides an interesting avenue to novel potential therapeutic avenues.

Weaknesses:

Although the authors provide a new and interesting avenue to understand the role of miR-26b in MASH, the study needs some additional validations and mechanistic insights in order to strengthen the author's conclusions.

(1) Analysis of the expression of miRNAs based on miRNA-seq of human samples (see https://ccb-compute.cs.uni-saarland.de/isomirdb/mirnas) suggests that miR-26b-5p is highly abundant both on liver and blood. It seems hard to reconcile that despite miRNA abundance being similar in both tissues, the physiological effects claimed by the authors in Figure 2 come exclusively from the myeloid (macrophages).

We agree with the reviewer that the effects observed in the whole-body KO model are most likely a combination of cellular effects, particularly since miR-26b is also highly expressed in the liver. However, with the LysM-model we merely want to demonstrate that the myeloid cells at least play an important, though not exclusive, role in the phenotype. In the discussion, we also further elaborate on the fact that the observed changes in the liver can me mediated by hepatic changes.

To stress this, we have adjusted the conclusion of Figure 2:

“Interestingly, mice that have a myeloid-specific lack of miR-26b also show increased hepatic cholesterol levels and lipid accumulation demonstrated by Oil-red-O staining, coinciding with an increased hepatic Cd36 expression (Figure 2), demonstrating that myeloid miR-26b plays a major, but not exclusive, role in the observed steatosis.”

(2) Similarly, the miRNA-seq expression from isomirdb suggests also that expression of miR-26a-5p is indeed 4-fold higher than miR-26b-5p both in the liver and blood. Since both miRNAs share the same seed sequence, and most of the supplemental regions (only 2 nt difference), their endogenous targets must be highly overlapped. It would be interesting to know whether deletion of miR-26b is somehow compensated by increased expression of miR-26a-5p loci. That would suggest that the model is rather a depletion of miR-26.

UUCAAGUAAUUCAGGAUAGGU mmu-miR-26b-5p mature miRNA

UUCAAGUAAUCCAGGAUAGGCU mmu-miR-26a-5p mature miRNA

This is a very valid point raised by the reviewer, which we actually already explored in a previous study, describing our used mouse model (Van der Vorst et al. BMC Genom Data, 2021). In this manuscript, we could show that miR-26a is not affected by the deficiency of miR-26b (Figure 1G in: Van der Vorst et al. BMC Genom Data, 2021).

(3) Similarly, the miRNA-seq expression from isomirdb suggests also that expression of miR-26b-5p is indeed 50-fold higher than miR-26b-3p in the liver and blood. This difference in abundance of the two strands is usually regarded as one of them being the guide strand (in this case the 5p) and the other being the passenger (in this case the 3p). In some cases, passenger strands can be a byproduct of miRNA biogenesis, thus the rescue experiments using LNPs with both strands in equimolar amounts would not reflect the physiological abundance miR-26b-3p. The non-physiological overabundance of miR-26b-3p would constitute a source of undesired off-targets.

We agree with the reviewer on this aspect and this is something we had to consider while generating the mimic LNPs. However, we believe that we do not observe and undesired off-target effects, as the effects of the mimic LNPs at least on functional outcomes are relatively mild and only restricted to the expected effects on lipids. Furthermore, the effects on the kinase profile due to the mimic LNP treatment are in line with our expectations. Combined these results suggest at least that potential off-target effects are minor.

(4) It would also be valuable to check the miRNA levels on the liver upon LNP treatment, or at least the signatures of miR-26b-3p and miR-26b-5p activity using RNA-seq on the RNA samples already collected.

This is indeed a valid point that we have now addressed. We have measured the mir26b-3p and mir26b-5p expression levels in livers from mice after 4-week WTD with simultaneous injection with either empty LNPs as vehicle control (eLNP) or LNPs containing miR-26b mimics (mLNP) every 3 days. As shown in Supplemental Figure 2A-B, mLNP treatment clearly results in an overexpression of the mir26b in the livers of these mice. We have rephrased the text accordingly by stating that mLNP results in an “overexpression” rather than “replenishment”.

(5) Some of the phenotypes described, such as the increase in cholesterol, overlap with the previous publication by van der Vorst et al. BMC Genom Data (2021), despite in this case the authors are doing their model in Apoe knock-out and Western-type diet. I would encourage the authors to investigate more or discuss why the initial phenotypes don't become more obvious despite the stressors added in the current manuscript.

In our previous publication (BMC Genom Data; 2021), we actually did not see any changes in circulating lipid levels. However, in that study we did not evaluate the livers of the mice, so we do not have any information about the hepatic lipid levels.

As mentioned by the reviewer, we believe that we see much more pronounced phenotypes in the current model because we use the combined stressor of Apoe-/- and Western-type diet, which cannot be compared to the wildtype and chow-fed mice used in the BMC Genom Data manuscript.

(6) The authors have focused part of their analysis on a few gene makers that show relatively modest changes. Deeper characterization using RNA-seq might reveal other genes that are more profoundly impacted by miR-26 depletion. It would strengthen the conclusions proposed if the authors validated that changes in mRNA abundance (Sra, Cd36) do impact the protein abundance. These relatively small changes or trends in mRNA expression, might not translate into changes in protein abundance.

As suggested by the reviewer we have now also confirmed that the protein expression of CD36 and SRA is significantly increased upon miR-26b depletion, visualized as Figure 1K-L in the revised manuscript. Unfortunately, we do not have enough material left to perform similar analysis for the LysM-model or the LNP-model, although based on the whole-body effects we are confident that at least for CD36/SRA in this case the gene expression matches effects observed on protein level.

(7) In Figures 5 and 7, the authors run a phosphorylation array (STK) to analyze the changes in the activity of the kinome. It seems that a relatively large number of signaling pathways are being altered, I think that should be strengthened by further validations by Western blot on the collected tissue samples. For quite a few of the kinases, there might be antibodies that recognise phosphorylation. The two figures lack a mechanistic connection to the rest of the manuscript.<br /> On this point we respectfully have to disagree with the reviewer. We have used a kinase activity profiling approach (PamGene) to analyse the real-time activity of kinases in our lysates. This approach is different than the classical Western blot approach in which only the presence or absence of a specific phosphorylation is detected. Thereby, Western blot analysis does not analyse phosphorylation in real-time, but rather determines whether there has been phosphorylation in the past. Our approach actually determines the real-time, current activity of the kinases, which we believe is a different and perhaps even more reliable read-out measurement. Therefore, validation by Western blot would not strengthen these observations.

We have particularly tried to connect these observations to the rest of the manuscript by highlighting the observed signalling cascades that are affected, highlighting a role in inflammation and angiogenesis, thereby providing some mechanistic insights.

Reviewer #2 (Recommendations For The Authors):

I would encourage the authors to follow-up on some of the more miRNA focused comments made above, which would strengthen the mechanistic part of the work presented.

I suggest the authors tone down some of some of the claims made (eg. "clearly increased expression", "exacerbated hepatic fibrosis"), given that some of it might need further validation.

Wherever needed we have tuned down the tone of some claims, although we believe that most claims are already written carefully enough and in line with the observed results.

Some of the panels that are supposed to have the same amount of animals have variable N, despite they come from the same exact number of RNA samples or tissue lysates (eg. 1G and 1H, vs 1I and 1J).

This is indeed correct and caused by the fact that some analysis resulted in statistical outliers as identified using the ROUT = 1 method, as also specified in section 2.15 of the method section.

It would be nice to have representative images of oil-red-o in all the figures where it is quantified (or at least in the supplementary figures).

As suggested by the reviewer, we have now included representative images for the LysM-model (Revised Figure 2D) and the LNP-model (Revised Figure 6D) as well.

Author response:

The following is the authors’ response to the original reviews.

Reviewer #1 (Public Review):  

Summary:

In this manuscript, Shao et al. investigate the contribution of different cortical areas to working memory maintenance and control processes, an important topic involving different ideas about how the human brain represents and uses information when it is no longer available to sensory systems. In two fMRI experiments, they demonstrate that the human frontal cortex (area sPCS) represents stimulus (orientation) information both during typical maintenance, but even more so when a categorical response demand is present. That is, when participants have to apply an added level of decision control to the WM stimulus, sPCS areas encode stimulus information more than conditions without this added demand. These effects are then expanded upon using multi-area neural network models, recapitulating the empirical gradient of memory vs control effects from visual to parietal and frontal cortices. In general, the experiments and analyses provide solid support for the authors' conclusions, and control experiments and analyses are provided to help interpret and isolate the frontal cortex effect of interest. However, I suggest some alternative explanations and important additional analyses that would help ensure an even stronger level of support for these results and interpretations.

Strengths:

-  The authors use an interesting and clever task design across two fMRI experiments that is able to parse out contributions of WM maintenance alone along with categorical, rule-based decisions. Importantly, the second experiment only uses one fixed rule, providing both an internal replication of Experiment 1's effects and extending them to a different situation when rule-switching effects are not involved across mini-blocks.

- The reported analyses using both inverted encoding models (IEM) and decoders (SVM) demonstrate the stimulus reconstruction effects across different methods, which may be sensitive to different aspects of the relationship between patterns of brain activity and the experimental stimuli.

- Linking the multivariate activity patterns to memory behavior is critical in thinking about the potential differential roles of cortical areas in sub-serving successful working memory. Figure 3 nicely shows a similar interaction to that of Figure 2 in the role of sPCS in the categorization vs. maintenance tasks.

- The cross-decoding analysis in Figure 4 is a clever and interesting way to parse out how stimulus and rule/category information may be intertwined, which would have been one of the foremost potential questions or analyses requested by careful readers. However, I think more additional text in the Methods and Results to lay out the exact logic of this abstract category metric will help readers bet0ter interpret the potential importance of this analysis and result.

We thank the reviewer for the positive assessment of our manuscript. Please see lines 366-372, 885-894 in the revised manuscript for a detailed description of the abstract category index, and see below for a detailed point-by-point response.

Weaknesses:

- Selection and presentation of regions of interest: I appreciate the authors' care in separating the sPCS region as "frontal cortex", which is not necessarily part of the prefrontal cortex, on which many ideas of working memory maintenance activity are based. However, to help myself and readers interpret these findings, at a minimum the boundaries of each ROI should be provided as part of the main text or extended data figures. Relatedly, the authors use a probabilistic visual atlas to define ROIs in the visual, parietal, and frontal cortices. But other regions of both lateral frontal and parietal cortices show retinotopic responses (Mackey and Curtis, eLife, 2017: https://elifesciences.org/articles/22974) and are perhaps worth considering. Do the inferior PCS regions or inferior frontal sulcus show a similar pattern of effects across tasks? And what about the middle frontal gyrus areas of the prefrontal cortex, which are most analogous to the findings in NHP studies that the authors mention in their discussion, but do not show retinotopic responses? Reporting the effects (or lack thereof) in other areas of the frontal cortex will be critical for readers to interpret the role of the frontal cortex in guiding WM behavior and supporting the strongly worded conclusions of broad frontal cortex functioning in the paper. For example, to what extent can sPCS results be explained by visual retinotopic responses? (Mackey and Curtis, eLife, 2017: https://elifesciences.org/articles/22974).

We thank the reviewer for the suggestions. We have added a Supplemental Figure 1 to better illustrate the anatomical locations of ROIs.  

Following the reviewer’s suggestion, we defined three additional subregions in the frontal cortex based on the HCP atlas [1], including the inferior precentral sulcus (iPCS, generated by merging 6v, 6r, and PEF), inferior frontal sulcus (IFS, generated by merging IFJp, IFJa, IFSp, IFSa, and p47r), and middle frontal gyrus (MFG, generated by merging 9-46d, 46, a9-46v, and p9-46v). We then performed the same analyses as in the main text using both mixed-model and within-condition IEMs. Overall, we found that none of the ROIs demonstrated significant orientation representation in Experiment 1, for either IEM analysis (Author response image 1A and 1C). In Experiment 2, however, the IFS and MFG (but not iPCS) demonstrated a similar pattern to sPCS for orientation representation, though these results did not persist in the within-condition IEM with lower SNR (Author response image 1B and 1D). Moreover, when we performed the abstract category decoding analysis in the three ROIs, only the MFG in Experiment 2 showed significant abstract category decoding results, with no significant difference between experiments (Author response image 1E). To summarize, the orientation and category results observed in sPCS in the original manuscript were largely absent in other frontal regions. There was some indication that the MFG might share some results for orientation representation and category decoding, although this pattern was weaker and was only observed in some analyses in Experiment 2. Therefore, although we did not perform retinotopic mapping and cannot obtain a direct measure of retinotopic responses in the frontal cortex, these results suggest that our findings are unlikely to be explained by visual retinotopic responses: the iPCS, which is another retinotopic region, did not show the observed pattern in any of the analyses. Notably, the iPCS results are consistent with our previous work demonstrating that orientation information cannot be decoded from iPCS during working memory delay [2]. We have included these results on lines 395-403, 563-572 in the revised manuscript to provide a more comprehensive understanding of the current findings. 

Author response image 1.

Orientation reconstruction and abstract category decoding results in iPCS, IFS, and MFG.

- When looking at the time course of effects in Figure 2, for example, the sPCS maintenance vs categorization effects occur very late into the WM delay period. More information is needed to help separate this potential effect from that of the response period and potential premotor/motor-related influences. For example, are the timecourses shifted to account for hemodynamic lag, and if so, by how much? Do the sPCS effects blend into the response period? This is critical, too, for a task that does not use a jittered delay period, and potential response timing and planning can be conducted by participants near the end of the WM delay. For example, the authors say that " significant stimulus representation in EVC even when memoranda had been transformed into a motor format (24)". But, I *think* this paper shows the exact opposite interpretation - EVC stimulus information is only detectable when a motor response *cannot* be planned (https://elifesciences.org/articles/75688). Regardless, parsing out the timing and relationship to response planning is important, and an ROI for M1 or premotor cortex could also help as a control comparison point, as in reference (24).

We thank the reviewer for raising this point. We agree that examining the contribution of response-related activity in our study is crucial, as we detail below:

First, the time course results in the manuscript are presented without time shifting. The difference in orientation representation in Figure 2 emerged at around 7 s after task cue onset and 1 s before probe onset. Considering a 4-6 s hemodynamic response lag, the difference should occur around 1-3 s after task cue onset and 5-7 s prior to probe onset. This suggests that a substantial portion of the effect likely occurred during the delay rather than response period.

Second, our experimental design makes it unlikely that response planning would have influenced our results, as participants were unable to plan their motor responses in advance due to randomized response mapping at the probe stage on a trial-by-trial basis. Moreover, even if response planning had impacted the results in sPCS, it would have affected both conditions similarly, which again, would not explain the observed differences between conditions.

Third, following the reviewer’s suggestion, we defined an additional ROI (the primary motor cortex, M1) using the HCP atlas and repeated the IEM analysis. No significant orientation representation was observed in either condition in M1, even during the response period (Figure S3), further suggesting that our results are unlikely to be explained by motor responses or motor planning.

Based on the evidence above, we believe motor responses or planning are unlikely to account for our current findings. We have included these results on lines 264-267 to further clarify this issue.

Lastly, upon re-reading the Henderson et al. paper [3], we confirmed that stimulus information was still decodable in EVC when a motor response could be planned (Figure 2 of Henderson et al.). In fact, the authors also discussed this result in paragraph 5 of their discussion. This finding, together with our results in EVC, indicates that EVC maintains stimulus information in working memory even when the information is no longer task-relevant, the functional relevance of which warrants further investigation in future research.

- Interpreting effect sizes of IEM and decoding analysis in different ROIs. Here, the authors are interested in the interaction effects across maintenance and categorization tasks (bar plots in Figure 2), but the effect sizes in even the categorization task (y-axes) are always larger in EVC and IPS than in the sPCS region... To what extent do the authors think this representational fidelity result can or cannot be compared across regions? For example, a reader may wonder how much the sPCS representation matters for the task, perhaps, if memory access is always there in EVC and IPS? Or perhaps late sPCS representations are borrowing/accessing these earlier representations? Giving the reader some more intuition for the effect sizes of representational fidelity will be important. Even in Figure 3 for the behavior, all effects are also seen in IPS as well. More detail or context at minimum is needed about the representational fidelity metric, which is cited in ref (35) but not given in detail. These considerations are important given the claims of the frontal cortex serving such an important for flexible control, here.

We thank the reviewer for raising this point. We agree that the effect sizes are always larger in EVC and IPS. This is because the specific decoding method we adopted, IEM, is based on the concept of population-level feature-selective responses, and decoding results would be most robust in regions with strong feature-tuning responses, such as EVC and parts of IPS. Therefore, to minimize the impact of effect size on our results, we avoided direct comparisons of representational strength across ROIs, focusing instead on differences in representational strength between conditions within the same ROI. With this approach, we found that EVC and IPS showed high representational fidelity throughout the trial, but only in sPCS did we observe significant higher fidelity in categorization condition, where orientation was actually not a behavioral goal but was manipulated in working memory to achieve the goal. Moreover, although representational fidelity in the EVC was the highest, its behavioral predictability decreased during the delay period, unlike sPCS. These results suggest that the magnitude of fidelity alone is not the determining factor for the observed categorization vs. maintenance effect or for behavioral performance. We have included further discussion on this issue on lines 208-211 of the revised manuscript.

The reviewer also raised a good point that IPS showed similar behavioral correlation results as sPCS. In the original manuscript, we discussed the functional similarities and distinctions between IPS and sPCS in the discussion. We have expanded on this point on lines 610-627 in the revised manuscript:

“While many previous WM studies have focused on the functional distinction between sensory and frontoparietal cortex, it has remained less clear how frontal and parietal cortex might differ in terms of WM functions. Some studies have reported stimulus representations with similar functionality in frontal and parietal cortex [4, 5], while others have observed differential patterns [6-8]. We interpret the differential patterns as reflecting a difference in the potential origin of the corresponding cognitive functions. For example, in our study, sPCS demonstrated the most prominent effect for enhanced stimulus representation during categorization as well as the tradeoff between stimulus difference and category representation, suggesting that sPCS might serve as the source region for such effects. On the other hand, IPS did show visually similar patterns to sPCS in some analyses. For instance, stimulus representation in IPS was visually but not statistically higher in the categorization task. Additionally, stimulus representation in IPS also predicted behavioral performance in the categorization task. These results together support the view that our findings in sPCS do not occur in isolation, but rather reflect a dynamic reconfiguration of functional gradients along the cortical hierarchy from early visual to parietal and then to frontal cortex.”

Lastly, following the reviewer’s suggestion, we have included more details on the representational fidelity metric on lines 201-206, 856-863 in the revised manuscript for clarity.

Recommendations:

Figure 3 layout - this result is very interesting and compelling, but I think could be presented to have the effect demonstrated more simply for readers. The scatter plots in the second and third rows take up a lot of space, and perhaps having a barplot as in Figure 2 showing the effects of brain-behavior correlations collapsed across the WM delay period timing would make the effect stand out more.

We thank the reviewer for the suggestion. We have added a subplot (C) to Figure 3 to demonstrate the brain-behavior correlation collapsed across the late task epoch.

When discussing the link between sPCS representations and behavior, I think this paper should likely be cited ([https://www.jneurosci.org/content/24/16/3944](https://www.jneurosci.org/content/24/ 16/3944)), which shows univariate relationships between sPCS delay activity and memory-guided saccade performance.

We thank the reviewer for the suggestion and have included this citation on lines 278-279 in the revised manuscript.

Interpretation of "control" versus categorization - the authors interpret that "It would be of interest to further investigate whether this active control in the frontal cortex could be generalized to tasks that require other types of WM control such as mental rotation." I think more discussion on the relationship between categorization and "control" is needed, especially given the claim of "flexible control" throughout. Is stimulus categorization a form of cognitive control, and if so, how?  

We thank the reviewer for raising this point. Cognitive control is generally defined as the process by which behavior is flexibly adapted based on task context and goals, and most theories agree that this process occurs within working memory [9, 10]. With this definition, we consider stimulus categorization to be a form of cognitive control, because participants needed to adapt the stimulus based on the categorization rule in working memory for subsequent category judgements. With two categorization rules, the flexibility in cognitive control increased, because participants need to switch between the two rules multiple times throughout the experiment, instead of being fixed on one rule. We now clarify these two types of controls on lines 112-116 in the introduction.

However, we agree that the latter form of control could be more related to rule switching that might not be specific to categorization per se. For instance, if participants perform rule switching in another type of WM task that requires WM control such as mental rotation, it remains to be tested whether similar results would be observed and/or whether same brain regions would be recruited. We have included further information on this issue on lines 572-575 in the revised manuscript.

Reviewer #2 (Public Review):

Summary:

The authors provide evidence that helps resolve long-standing questions about the differential involvement of the frontal and posterior cortex in working memory. They show that whereas the early visual cortex shows stronger decoding of memory content in a memorization task vs a more complex categorization task, the frontal cortex shows stronger decoding during categorization tasks than memorization tasks. They find that task-optimized RNNs trained to reproduce the memorized orientations show some similarities in neural decoding to people. Together, this paper presents interesting evidence for differential responsibilities of brain areas in working memory.

Strengths:

This paper was strong overall. It had a well-designed task, best-practice decoding methods, and careful control analyses. The neural network modelling adds additional insight into the potential computational roles of different regions.

We thank the reviewer for the positive assessment of our manuscript.

Weaknesses:

While the RNN model matches some of the properties of the task and decoding, its ability to reproduce the detailed findings of the paper was limited. Overall, the RRN model was not as well-motivated as the fMRI analyses.

We are grateful for the reviewer’s suggestions on improving our RNN results. Please see below for a detailed point-by-point response.

Recommendations:

Overall, I thought that this paper was excellent. I have some conceptual concerns about the RNN model, and minor recommendations for visualization.

(1) I think that the RNN modelling was certainly interesting and well-executed. However, it was not clear how much it contributed to the results. On the one hand, it wasn't clear why reproducing the stimulus was a critical objective of the task (ie could be more strongly motivated on biological grounds). On the other hand, the agreement between the model and the fMRI results is not that strong. The model does not reproduce stronger decoding in 'EVC' for maintenance vs categorization. Also, the pattern of abstract decoding is very different from the fMRI (eg the RNN has stronger categorical encoding in 'EVC' than 'PFC' and larger differences between fixed and flexible rules in earlier areas than is evident in the fMRI). Together, the RNN modelling comes across as a little ad hoc, without really nailing the performance.

We thank the reviewer for prompting us to further elaborate on the rationale for our RNN analysis. In our fMRI results, we observed a tradeoff between maintaining stimulus information in more flexible tasks (Experiment 1) and maintaining abstract category information in less flexible tasks (Experiment 2). This led to the hypothesis that participants might have employed different coding strategies in the two experiments. Specifically, in flexible environments, stimulus information might be preserved in its original identity in the higher-order cortex, potentially reducing processing demands in each task and thereby facilitating efficiency and flexibility; whereas in less flexible tasks, participants might generate more abstract category representations based on task rules to facilitate learning. To directly test this idea, we examined whether explicitly placing a demand for the RNN to preserve stimulus representation would recapitulate our fMRI findings in frontal cortex by having stimulus information as an output, in comparison to a model that did not specify such a demand. Meanwhile, we totally agree with the reviewer that there are alternative ways to implement this objective in the model. For instance, changing the network encoding weights (lazy vs. rich regime) to make feedforward neural networks either produce high-dimensional stimulus or low-dimensional category representations [11]. However, we feel that exploring these alternatives may fall outside the scope of the current study.

Regarding the alignment between the fMRI and RNN results: for the stimulus decoding results in EVC, we found that with an alternative decoding method (IEM), a similar maintenance > categorization pattern was observed in EVC-equivalent module, suggesting that our RNN was capable of reproducing EVC results, albeit in a weaker manner (please see our response to the reviewer’s next point). For the category decoding results, we would like to clarify that the category decoding results in EVC was not necessarily better than those in sPCS. Although category decoding accuracy was numerically higher in EVC, it was more variable compared to IPS and sPCS. To illustrate this point, we calculated the Bayes factor for the category decoding results of RNN2 in Figure 6C, and found that the amount of evidence for category decoding as well as for the decoding difference between RNNs in IPS and sPCS modules was high, whereas the evidence in the EVC was insufficient (Response Table 1).

Author response table 1.

Bayes factors for category decoding and decoding differences in Figure 6C lower panel.

Nevertheless, we agree with the reviewer that all three modules demonstrated the category decoding difference between experiments, which differs from our fMRI results. This discrepancy may be partially due to differences in signal sensitivity. RNN signals typically have a higher SNR compared to fMRI signals, as fMRI aggregates signals from multiple neurons and single-neuron tuning effects can be reduced. We have acknowledged this point on lines 633-636 in the revised manuscript. Nonetheless, the current RNNs effectively captured our key fMRI findings, including increased stimulus representation in frontal cortex as well as the tradeoff in category representation with varying levels of flexible control. We believe the RNN results remain valuable in this regard.

Honestly, I think the paper would have a very similar impact without the modelling results, but I appreciate that you put a lot of work into the modeling, and this is an interesting direction for future research. I have a few suggestions, but nothing that I feel too strongly about.

- It might be informative to use IEM to better understand the RNN representations (and how similar they are to fMRI). For example, this could show whether any of the modules just encode categorical information. 

- You could try providing the task and/or retro cue directly to the PFC units. This is a little unrealistic, but may encourage a stronger role for PFC.

- You might adjust the ratio of feedforward/feedback connections, if you can find good anatomical guidance on what these should be.

Obviously, I don't have much - it's a tricky problem!

We thank the reviewer for the suggestions. To better align the fMRI and RNN results, we first performed the same IEM analyses used in the fMRI analyses on the RNN data. We found that with IEM, the orientation representation in the EVC module demonstrated a pattern similar to that in the fMRI data, showing a negative trend for the difference between categorization and maintenance, although the trend did not reach statistical significance (Author response image 2A). Meanwhile, the difference between categorization and maintenance remained a positive trend in the sPCS module.

Second, following the reviewer’s suggestion, we adjusted the ratio of feedforward/feedback connections between modules to 1:2, such that between Modules 1 and 2 and between Modules 2 and 3, there were always more feedback than feedforward connections, consistent with recent theoretical proposals [12]. We found that, this change preserved the positive trend for orientation differences in the sPCS module, but in the meantime also made the orientation difference in the EVC and IPS modules more positive (Author response image 2B).

To summarize, we found that the positive difference between categorization and maintenance in the sPCS module was robust across difference RNNs and analytical approaches, further supporting that RNNs with stimulus outputs can replicate our key fMRI findings in the frontal cortex. By contrast, the negative difference between categorization and maintenance in EVC was much weaker. It was weakly present using some analytical methods (i.e., the IEM) but not others (i.e., SVMs), and increasing the feedback ratio of the entire network further weakened this difference. We believe that this could be due to that the positive difference was mainly caused by top-down, feedback modulations from higher cortex during categorization, such that increasing the feedback connection strengthens this pattern across modules. We speculate that enhancing the negative difference in the EVC module might require additional modules or inputs to strengthen fine-grained stimulus representation in EVC, a mechanism that might be of interest to future research. We have added a paragraph to the discussion on the limitations of the RNN results on lines 629-644.

Author response image 2.

Stimulus difference across RNN modules.  (A). Results using IEM (p-values from Module 1 to 3: 0.10, 0.48, 0.01). (B). Results using modified RNN2 with changed connection ratio (p-values from Module 1 to 3: 0.12, 0.22, 0.08). All p-values remain uncorrected.

(2) Can you rule out that during the categorization task, the orientation encoding in PFC isn't just category coding? You had good controls for category coding, but it would be nice to see something for orientation coding. e.g., fit your orientation encoding model after residualizing category encoding, or show that category encoding has worse CV prediction than orientation encoding.

We thank the reviewer for raising this point. To decouple orientation and category representations, we performed representational similarity analysis (RSA) in combination with linear mixed-effects modeling (LMEM) on the fMRI data. Specifically, we constructed three hypothesized representational dissimilarity matrices (RDMs), one for graded stimulus (increasing distance between orientations as they move farther apart, corresponding to graded feature tuning responses), one for abstract category (0 for all orientations within the same category and 1 for different categories), and another for discrete stimulus (indicating equidistant orientation representations). We then fit the three model RDMs together using LMEM with subject as the random effect (Author response image 3A). This approach is intended to minimize the influence of collinearity between RDMs on the results [13].

Overall, the LMEM results (Author response image 3B-D) replicated the decoding results in the main text, with significant stimulus but not category representation in sPCS in Experiment 1, and marginally significant category representation in the same brain region in Experiment 2. These results further support the validity of our main findings and emphasize the contribution of stimulus representation independent of category representation.

Author response image 3.

Delineating stimulus and category effects using LMEM.  (A) Schematic illustration of this method. (B) Results for late epoch in Experiment 1, showing the fit of each model RDM. (C) Results for early epoch in Experiment 2. (D) Results for late epoch in Experiment 2.

(3) Is it possible that this region of PFC is involved in categorization in particular and not 'control-demanding working memory'? 

We thank the reviewer for raising this possibility. Cognitive control is generally defined as the process by which behavior is flexibly adapted based on task context and goals, and most theories agree that this process occurs within working memory [9, 10]. With this definition, we consider stimulus categorization to be a form of cognitive control, because participants need to adapt the stimulus based on the categorization rule in working memory for subsequent category judgements.  However, in the current study we only used one type of control-demanding working memory task (categorization) to test our hypothesis, and therefore it remains unclear whether the current results in sPCS can generalize to other types of WM control tasks.

We have included a discussion on this issue on lines 572-575 in the revised manuscript.

(4) Some of the figures could be refined to make them more clear:

a.  Figure 4 b/c should have informative titles and y-axis labels.

b.  Figure 5, the flexible vs fixed rule isn't used a ton up to this point - it would help to (also include? Replace?) with something like exp1/exp2 in the legend. It would also help to show the true & orthogonal rule encoding in these different regions (in C, or in a separate panel), especially to the extent that this is a proxy for stimulus encoding.

c.  Figure 6: B and C are very hard to parse right now. (i) The y-axis on B could use a better label. (ii) It would be useful to include an inset of the relevant data panel from fMRI that you are reproducing. (iii) Why aren't there fixed rules for RNN1?

We thank the reviewer for the suggestions and have updated the figures accordingly as following:

Overall I think this is excellent - my feedback is mostly on interpretation and presentation. I think the work itself is really well done, congrats!

References

(1) Glasser, M.F., et al., A multi-modal parcellation of human cerebral cortex. Nature, 2016. 536(7615): p. 171-178.

(2) Yu, Q. and Shim, W.M., Occipital, parietal, and frontal cortices selectively maintain taskrelevant features of multi-feature objects in visual working memory. Neuroimage, 2017. 157: p. 97-107.

(3) Henderson, M.M., Rademaker, R.L., and Serences, J.T., Flexible utilization of spatial- and motor-based codes for the storage of visuo-spatial information. Elife, 2022. 11.

(4) Christophel, T.B., et al., Cortical specialization for attended versus unattended working memory. Nat Neurosci, 2018. 21(4): p. 494-496.

(5) Yu, Q. and Shim, W.M., Temporal-Order-Based Attentional Priority Modulates Mnemonic Representations in Parietal and Frontal Cortices. Cereb Cortex, 2019. 29(7): p. 3182-3192.

(6) Li, S., et al., Neural Representations in Visual and Parietal Cortex Differentiate between Imagined, Perceived, and Illusory Experiences. J Neurosci, 2023. 43(38): p. 6508-6524.

(7) Hu, Y. and Yu, Q., Spatiotemporal dynamics of self-generated imagery reveal a reverse cortical hierarchy from cue-induced imagery. Cell Rep, 2023. 42(10): p. 113242.

(8) Lee, S.H., Kravitz, D.J., and Baker, C.I., Goal-dependent dissociation of visual and prefrontal cortices during working memory. Nat Neurosci, 2013. 16(8): p. 997-9.

(9) Miller, E.K. and Cohen, J.D., An integrative theory of prefrontal cortex function. Annu Rev Neurosci, 2001. 24: p. 167-202.

(10) Badre, D., et al., The dimensionality of neural representations for control. Curr Opin Behav Sci, 2021. 38: p. 20-28.

(11) Flesch, T., et al., Orthogonal representations for robust context-dependent task performance in brains and neural networks. Neuron, 2022. 110(7): p. 1258-1270 e11.

(12) Wang, X.J., Theory of the Multiregional Neocortex: Large-Scale Neural Dynamics and Distributed Cognition. Annu Rev Neurosci, 2022. 45: p. 533-560.

(13) Bellmund, J.L.S., et al., Mnemonic construction and representation of temporal structure in the hippocampal formation. Nat Commun, 2022. 13(1): p. 3395.

Author response:

The following is the authors’ response to the original reviews.

Public Reviews:

Reviewer #1 (Public review):

Summary:

The manuscript co-authored by Pál Barzó et al is very clear and very well written, demonstrating the electrophysiological and morphological properties of human cortical layer 2/3 pyramidal cells across a wide age range, from age 1 month to 85 years using whole-cell patch clamp. To my knowledge, this is the first study that looks at the cross-age differences in biophysical and morphological properties of human cortical pyramidal cells. The community will also appreciate the significant effort involved in recording data from 485 cells, given the challenges associated with collecting data from human tissue. Understanding the electrophysiological properties of individual cells, which are essential for brain function, is crucial for comprehending human cortical circuits. I think this research enhances our knowledge of how biophysical properties change over time in the human cortex. I also think that by building models of human single cells at different ages using these data, we can develop more accurate representations of brain function. This, in turn, provides valuable insights into human cortical circuits and function and helps in predicting changes in biophysical properties in both health and disease.

Strengths:

The strength of this work lies in demonstrating how the electrophysiological and morphological features of human cortical layer 2/3 pyramidal cells change with age, offering crucial insights into brain function throughout life.

Weaknesses:

One potential weakness of the paper is that the methodology could be clearer, especially in how different cells were used for various electrophysiological measurements and the conditions under which the recordings were made. Clarifying these points would improve the study's rigor and make the results easier to interpret.

Reviewer #2 (Public review):

Summary:

In this study, Barzo and colleagues aim to establish an appraisal for the development of basal electrophysiology of human layer 2/3 pyramidal cells across life and compare their morphological features at the same ages.

Strengths:

The authors have generated recordings from an impressive array of patient samples, allowing them to directly compare the same electrophysiological features as a function of age and other biological features. These data are extremely robust and well organised.

Weaknesses:

The use of spine density and shape characteristics is performed from an extremely limited sample (2 individuals). How reflective these data are of the population is not possible to interpret. Furthermore, these data assume that spines fall into discrete types - which is an increasingly controversial assumption.

Many data are shown according to somewhat arbitrary age ranges. It would have been more informative to plot by absolute age, and then perform more rigourous statistics to test age-dependent effects.

Overall, the authors achieve their aims by assessing the physiological and morphological properties of human L2/3 pyramidal neurons across life. Their findings have extremely important ramifications for our understanding of human life and implications for how different neuronal properties may influence neurological conditions.

Reviewer #3 (Public review):

Summary:

To understand the specificity of age-dependent changes in the human neocortex, this paper investigated the electrophysiological and morphological characteristics of pyramidal cells in a wide age range from infants to the elderly.

The results show that some electrophysiological characteristics change with age, particularly in early childhood. In contrast, the larger morphological structures, such as the spatial extent and branching frequency of dendrites, remained largely stable from infancy to old age. On the other hand, the shape of dendritic spines is considered immature in infancy, i.e., the proportion of mushroom-shaped spines increases with age.

Strengths:

Whole-cell recordings and intracellular staining of pyramidal cells in defined areas of the human neocortex allowed the authors to compare quantitative parameters of electrophysiological and morphological properties between finely divided age groups.

They succeeded in finding symmetrical changes specific to both infants and the elderly, and asymmetrical changes specific to either infants or the elderly. The similarity of pyramidal cell characteristics between areas is unexpected.

Weaknesses:

Human L2/3 pyramidal cells are thought to be heterogeneous, as L2/3 has expanded to a high degree during the evolution from rodents to humans. However, the diversity (subtyping) is not revealed in this paper.

Recommendations for the authors: 

Reviewer #1 (Recommendations for the authors):

The manuscript co-authored by Pál Barzó et al is very clear and very well written, demonstrating the electrophysiological and morphological properties of the human cortical layer 2/3 pyramidal cells across a wide age range, from age 1 month to 85 years using whole-cell patch clamp. To my knowledge, this is the first study that looks at the cross-age differences in morphological and electrophysiological properties of human cortical pyramidal cells. The community will also appreciate the significant effort involved in recording data from 485 cells, given the challenges associated with collecting data from human tissue. understanding the electrophysiological properties of individual cells, which are essential for brain function, is crucial for comprehending human cortical circuits. I think this research enhances our knowledge of how biophysical properties change over time in the human cortex. I also think that by building models of human single cells at different ages using these data, we can develop more accurate representations of brain function. This, in turn, provides valuable insights into human cortical circuits and function and helps in predicting changes in biophysical properties in both health and disease.

We are grateful for the positive evaluation of our work. We also thank the reviewers for their comments and believe that our manuscript has improved significantly with their help. In addition to the reviewer’s suggestions for improvement, further cell reconstructions were performed to make the anatomical data more robust (n = 1,2,3,3,4,3,2 additional reconstruction in age groups infant, early childhood, late childhood, adolescence, young adulthood, middle adulthood and late adulthood, respectively; Σn = 18). Four additional cells were added to the spine analysis and the statistics associated with each additional dataset were updated.

I have some comments, particularly regarding the methodology and data presentation, to improve the clarity of the paper

(1) I assume the tissue is from the resected area adjacent to the tumor. Could you please clarify this in the Methods section?

Thank you for this comment, it has been clarified in the Methods section with the following sentence: “We used human cortical tissue adjacent to the pathological lesion  that had to be surgically removed from patients (n = 63 female  n = 45 male) as part of the treatment for tumors, hydrocephalus, apoplexy, cysts, and arteriovenous malformation.”

(2) Regarding the presentation of data in the Methods section, could you please clarify whether the authors used different cells for measuring the various electrophysiological properties? The number of recorded cells for calculating subthreshold properties (e.g., late adulthood: n = 113) differs from the number the cells used for calculating suprathreshold properties (e.g., late adulthood: n = 83). If this is the case, it may make it difficult to compare the electrophysiological properties. Could you please clarify this?

The different element numbers are indeed due to the fact that different quality criteria were defined for the analysis of fast and slow signals. For the analysis of fast signals (e.g. AP half-width, AP upstroke velocity, AP amplitude), higher quality requirements were established therefore cells with high series resistance (> 30 MΩ) were excluded. We have updated and clarified the recording conditions in the text, figures, and methodology section accordingly.

(3) Additionally, they mentioned that their recordings were done at zero holding current and at more than -50 pA. Could you clarify whether the data from these two sets of experiments were combined? If so, please provide an explanation in the methods section.

Basically, we wanted to determine the parameters of the potential changes of the membrane at rest. However, for technical reasons related to the biological amplifier, in some of the experiments a certain continuous holding current may be present during the measurement (3.5% of all experiments). The holding currents were in the range of -50 pA to +60 pA. Within this range, previously checked on mouse neurons we have not found linear correlation between the electrophysiological properties and the holding current. This is reported in the Methods section.

(4) This section needs revision. It is unclear why different series resistances (Rs) or different cells were used to compute various electrophysiological properties." To calculate passive membrane properties (resting membrane potential, input resistance, time constant, and sag) either cells with series resistance (Rs): 22.85 {plus minus} 9.04 MΩ (ranging between -4.55 MΩ and 56.76 MΩ) and 0 pA holding current (n = 154), or cells with holding current > -50 pA (-7.46 {plus minus} 28.56 pA, min: -49.89 pA, max: 59.68pA) and Rs < 30 MΩ (18.96 {plus minus} 6.48 MΩ) (n = 23) were used. For the analysis of high frequency action potential features (AP half-width, AP up-stroke velocity, AP amplitude and rheobase) cells with Rs < 30 MΩ (n = 331 cells with Rs 19.2 {plus minus} 6.6 MΩ) and holding current > -50pA (n = 308 with 0 pA holding current and Rs: 19.22 {plus minus} 6.59 MΩ, n = 23 withholding current: -7.46 {plus minus} 28.56 pA and Rs: 18.96 {plus minus} 6.48 MΩ) were used."

To make the chapter clearer, we simplified the cell groups used to analyse the different electrophysical properties and revised the Method section as follows: “For the analysis of the electrophysiological recordings n = 457 recordings with a series resistance (Rs) of 24.93 ± 11.18 MΩ (max: 63.77 MΩ) were used. For the analysis of fast parameters related to the action potential (AP half-width, AP upstroke velocity, AP amplitude and rheobase), higher quality requirements were set and cells with Rs > 30 MΩ were excluded. This reduced the data set to n = 331 cells with Rs 19.42 ± 6.2 MΩ.”

(5) The authors recorded the sag ratio using a -100 pA injected current. Is there a technical reason why they did not inject more than -100 PA?

There is no particular technical reason, we use similar to others this current amplitude for voltage response recordings over the years to record electrophysiological traces.

(6) In the abstract, the authors mentioned that data were recorded from ages 1 month to 85 years. However, in the results, they stated that data were recorded from ages 0 to 85 years. Could you please clarify this discrepancy?

We corrected this discrepancy.

(7) Additionally, the results mention that data were collected from 485 human cortical layer 2/3 (L2/3) pyramidal cells, but subthreshold membrane features such as resting membrane potential, input resistance, time constant (tau), and sag ratio were calculated in 475 cortical pyramidal cells from 99 patients. Could you please clarify these discrepancies? In the discussion "We recorded from n = 457 human cortical excitatory pyramidal cells from the supragranular layer from birth to 85 years"

Thank you for pointing this out, we have corrected the error. Although our full data set contained 485 pyramidal cells, 28 recordings were excluded from the electrophysiological analysis and were used for morphological evaluation only, therefore 457 recordings were used for passive parameter measurements.

(8) Regarding the distance from the pia to the border layer L1/L2, did the authors notice any differences across ages?

To investigate whether the thickness of cortical layer 1 changes throughout life, we measured the L1 thickness and found no significant differences between age groups (P = 0.09, Kruskal-Wallis test) (Author response image 1).

Author response image 1.

Thickness of cortical layer 1 at different life stages. (A) Boxplot shows the thickness of layer 1. (B) Scatter plot shows the distribution of L1 thickness measured on the reconstructed cells. Age is shown in years on a logarithmic scale, dots are color-coded according to the corresponding age groups.

(9) I am not sure why they referred to the data as layer 2/3 when most of the data, based on Figure 1E, were recorded from a distance of 0-200 µm from the L1/L2 border. Could it be that there is no significant depth-dependent variation in electrophysiological properties, as reported by Berg (2021), Kalmbach (2018), and Chameh (2021)?

Although the vast majority of our data comes from a distance of less than 200 μm from the L1/L2 border, we cannot neglect the fact that our dataset also contains a small number of cells deeper than this, which are layer 3 cells. Apart from some differences shown in Supplementary Figures 7-9, we found no general difference between cells located at a distance of less than 200 μm and more than 200 μm from the L1 border.

(10) In Figure 1, there is variability in resting membrane potential (RMP), tau, and input resistance (IR) within the infant age group. However, this trend is not observed in the sag ratio. Could you please discuss this finding?

The large variance in the data is due to dramatic changes in these three parameters during the first year of life. Supplementary Figure 3 shows the comparisons of parameter distributions of patients between 0-6 months and 6-12 months. The sag amplitude in these cells is generally low therefore no such large changes could have occurred in them.

(11) Did the authors use a K-Nearest Neighbors (KNN) test to assess the accuracy of the infant cluster in Figure 3F?

Based on eight electrophysiological features of the cells (resting Vm, input resistance, tau, sag ratio, rheobase, AP half-width, AP up-stroke, and AP amplitude), the infant pyramidal cells on a UMAP form a distinct group (Author response image 2A) represented by cluster 4 on Author response image 2B. When calculating the sum of the Euclidean distances of cells within the cluster from the centroid, the isolated infant group (cluster 4) shows the smallest distance value from the centroid (cluster 1: 40.2, cluster 2: 36.21, cluster 3: 39.96, cluster 4: 5.72, cluster 5: 39.2, cluster 6: 55.74, cluster 7: 54.27), demonstrating that infant cells create a discrete cluster distinct from other age groups (Author response image 2B).

Author response image 2.

(A) Uniform Manifold Approximation and Projection (UMAP) of 8 selected electrophysiological properties (resting Vm, input resistance, tau, sag ratio, rheobase, AP half-width, AP up-stroke, and AP amplitude) with data points for 331 cortical L2/3 pyramidal cells, colored with the corresponding age groups. (B) UMAP colored by k-means clustering with 7 clusters, red crosses represent the centroids of the clusters.

(12) Missing citation: 'Previous research has shown that the biophysical properties of human pyramidal cells show depth-related correlations throughout L2/3 (Berg et al., 2021).' Please include citations for Kalmbach (2018) and Chameh (2021).

We thank for the additional references, these studies are now cited.

(13) Have they noticed any morphological properties differences among the different cortical lobes (Parietal, Temporal, Frontal, and Occipital). It would be beneficial to present this data, especially since they have a sufficient sample size from each cortical lobe.

The majority of our data set on the morphological properties of pyramidal cells comes from the parietal (n = 17 cells) and temporal lobe (n = 15). We found no significant differences in the morphological properties of cells from these two brain regions and no differences between age groups in the same cortical lobes.

(14) Have the authors found differences in spine characteristics among different cortical areas, as reported previously by 10.1023/a:1024134312173).

We found morphological differences in dendritic spines in the different brain regions, yet, our data are limited to draw definitive conclusions.

Reviewer #2 (Recommendations for the authors):

Major

(1) I believe that these data presented in all main text figures would be more intuitive to be plotted on a log(age) scale, such as shown in supplementary Figure 13. The bounds of the ages used for different groups, as summarised in Figure 1 feel somewhat arbitrary.

Recent neuroscientific studies on postnatal ageing mainly use the age-group comparison format (Kang 2011, Bethlehem 2022), which has been defined based on milestones in the cognitive, motor, social-emotional, and language/communications domains of observable behaviour (Zubler et al. 2022, for detailed definitions see Kang 2011). Since many parameters do not vary linearly but take a U-shape (or inverted U-shape), statistical quantification of these is not straightforward, so we would retain the age-group format for the main graphs. However, at the reviewer's suggestion, electrophysiological and morphological parameters are presented on a log(age) scale as supplementary figures (Supplementary Figures 2,4 and 6), also further statistical analysis was also carried out without grouping the data (see response 5).

(2) The authors present a lot of data values in the text, which is also shown in the figures. This makes reading of the manuscript somewhat difficult in places. For brevity, it may be best to present this data as supplementary tables.

Thank you for this suggestion. We have inserted these data as tables.

(3) I am unclear why the authors excluded cells that fired doublets or triplets in Figure 4? Were these included in the passive and AP-specific analysis - but excluded from F-I plots? Please clarify the rationale and the relative abundance of these physiological types based on age - one might predict that more initial-burst firing types are associated with older neurons?

Thank you for drawing attention to this anomaly. We have updated the figures and text by adding the cells with initial burst firing. These cells are also included in the analysis of passive and action potential properties. In our overall dataset, 6.78% of cells show burst firing; infant: 0%, early childhood: 3.57% (1 cell), late childhood: 0%, adolescence: 11.11% (6 cells), young adulthood: 10.11% (9), middle adulthood: 10.71% (6 cells), late adulthood: 7.96 (9 cells) of all cells including the age groups.

(4) The statistical analyses performed in Figure 6 are not justified. From the authors' description of these data, they derive spine density measurements from 1 infant and 1 aged adult, then perform pseudoreplicated analysis in these individuals. These data would require greater replication from infant and aged groups - with the possible inclusion of a younger adult group also. It would be ideal to have n=3/age group to allow robust statistical analysis.

Thank you for this point. Accordingly, we have expanded our data set to include n = 3 infant pyramidal cells (83 days old, from one patient) and n = 3 pyramidal cells from three late adulthood patients (64.3 ± 2.08 years old).

(5) Given the high number of individuals and replicates throughout this manuscript, a more circumspect approach to statistics would be appreciated, e.g. a generalised linear mixed effects model - with age as a fixed effect and sex, patient, etc as random effects. This may reveal the greatest statistical power of these important and rich data.

Of the generative models we used the Generalized Additive Mixed Model (GAMM) to describe the relationship between age and the various passive and active electrophysiological features. We defined age with cubic spline smoothing term as the fixed effect and gender, brain area, surgical procedure, and hemisphere as random effects. With GAMM we found that the age-dependent correlation of the examined parameters (resting membrane potential, input resistance, tau, sag ratio, rheobase current, AP half-width, AP up-stroke velocity, AP amplitude, first AP latency, adaptation) was significant, except for F-I slope, described by the model incorporating the four random effects.  We also observed correlation with gender, brain area, hemisphere, and surgical procedure in various intrinsic properties. The Author response table 1 below shows the statistical values of GAMM and the statistical tests used in the manuscript to compare.

Author response table 1.

Statistical significance of patient attributes *In the pairwise comparison, the age of cells in the two groups was significantly different: female (subthreshold: 37.36 ± 26.25 years old, suprathreshold: 38.3 ± 25.6 y.o.) - male (subthreshold: 24.86 ± 23.7 y.o., suprathreshold: 25.7 ± 23.93 y.o.), subthreshold: P = 1.96*10-6, suprathreshold: P = 3.25*10-5 Mann-Whitney test. **In the pairwise comparison, the age of cells in the two groups was significantly different: surgical procedure: tumor removal (subthreshold: 33.72 ± 24.33 y.o., suprathreshold: 36.43 ± 27.07 y.o.) - VP shunt (subthreshold: 27.38 ± 29.69 y.o., suprathreshold: 27.07 ± 29.37 y.o.) subthreshold: P = 3.68*10-3, suprathreshold: P = 1.64-10-3, Mann-Whitney test)

(6) Regarding the morphological diversity of dendritic spines. There is some debate in the field as to whether the distinction of specific dendritic spine types - as conveyed in this manuscript - are true subtypes or reflect a continuum of diverse morphology (see Tønneson et al., 2014 Nature Neuroscience). It is appreciated that the approach taken by the authors is the dogma within the field - however, dogma should continue to be challenged. Given that the authors have used DAB labelling combined with light microscopy, the possibility of accurately measuring spine morphology required for determining this continuum is extremely limited (e.g. Li et al., (2023) ACS Chemical Neuroscience). I would suggest that alongside the inclusion of further replicates for their spine analysis, the authors tone down their discussion of spine subtypes given the absence of any synaptic data presented in this current study to support the maturation (or otherwise) of dendritic spine synapses.

Many thanks to the reviewer for this comment. We agree with the drawbacks of our method for testing spine categorization. To increase the reliability of our results, we increased the number of pyramidal cells in the infant and late adult groups. We also revised the figure and as suggested by Reviewer#3 added photos of spines to each category in addition to schematic drawings to give an impression of the phenotype. In the discussion, we only address the differences between two readily separable mushroom and filopodial forms and highlight results that only confirm findings already known in the literature. Although the concerns are valid, we apply the sentence from the above Li et al. (2023) reference “...the most sophisticated equipment may not always be necessary for answering some research questions”. We believe that it is worth sharing our data and the somewhat subjective grouping, which we hope to report in more detail in the future.

Minor

(1) The order of the supplemental materials is out of order with their introduction in the text. These should be revised to reflect the order mentioned in the text.

Thank you for your comment, we have corrected the order of the supplementary figures.

(2) In Supplementary Figure 13, it would be informative to include some form of linear regression to confirm whether an age-dependent effect on neuronal morphology exists.

We have added linear regression to the figure.

(3) Figure 3D = should this be AP - not Ap?

Thank you for drawing attention to this, we have corrected the incorrect typing on the figure.

(4) For UMAP analysis in Figure 3, please provide a table of the features that were used for the 32 & 8-parameter UMAPs respectively.

We have added a table to the Materials and methods section of all the electrophysiological features included in the UMAP.

(5) For morphology, please include pia and L1/2 border for reconstructions shown for clarity.

We indicated both the pia mater and the L1/2 border on the figure showing all the reconstructions (Supplementary Figure 10).

Reviewer #3 (Recommendations for the authors):

Major:

(1) Data were obtained from different cortical areas of human patients of different ages. The electrophysiological characteristics were largely independent of other attributes such as disease, gender, and cortical areas (Supplementary Figure 2). To support the conclusion that age is one of the key attributes responsible for change, a similar morphological analysis would be necessary for gender.

We updated the text and the supplementary section with Supplementary Figures 18-21. to determine if age-related differences in biophysical characteristics are affected by the patient's gender.

(2) 'mushroom-shaped, thin, filopodial, branched, and stubby spines'

Show photographs of individual typical spine types to make the classification easier to understand.

To make the classification more understandable, we have updated the corresponding figure (Figure 6) with representative photos of the dendritic spine types.

(3) Some electrophysiological parameters of the infant group showed higher deviations compared to other age groups. A UMAP (Supplementary Figure 2) shows that some infant neurons form a small cluster, while other infant neurons are scattered with neurons of other ages. Are there any differences between infant neurons in the small cluster and other infant neurons with respect to attributes other than age?

For most of the electrophysiological parameters, the infant age group showed age-dependent variability, as illustrated in Supplementary Figures 3, 2,4 and 6 . The small group of infant cells is not clustered by gender, brain region, or medical condition, as shown in Supplementary Figure 5.

(4) A recent paper (Benavides-Piccione et al. 2024, doi:10.1093/cercor/bhae180) reported that some morphological parameters of human layer 3 neurons differ between occipital and temporal regions. Area-dependent morphological differences have been also reported in non-human primates. Discussion of potential contradictions may therefore be requested.

Most of the cells we reconstructed originated from the parietal and temporal regions (parietal: n = 20, temporal: n = 23, frontal: n = 15, occipital: n = 5). We found no differences in morphological features between these two regions, and we also found no significant differences when we compared the cells from the same brain regions by age group.

(5) L2/3 cells of rodents are morphologically differentiated according to cortical depth. If individual L2/3 cells of humans are less differentiated than those of rodents, this point should be discussed.

Depth-related morphological heterogeneity has already been reported previously (Berg 2021), however, our dataset on the morphological characteristics of pyramidal cells is from the upper L2/3 region, with their soma located at a distance of 117.85 ± 65.3 μm (between: 11.05 and 243.3 μm) from the L1/L2 border. Therefore, we cannot conclude from our data whether humans are less differentiated than rodents.

Minor:

(1) Cell body morphology may affect electrophysiological properties. However, morphological quantification of cell bodies has not been reported. It may be added.

In our DAB-labeled samples, we could not perfectly measure the total volume of the cell body in the reconstructions, therefore our measurements regarding the soma morphology are not shown in the manuscript. When comparing the cell body area of the middle sections of the soma of the reconstructed cells between the age groups, we found no significant differences (P = 0.082, Kruskal–Wallis test).

(2) 'The adaptation of the AP frequency response'

Describe how this parameter was obtained.

The adaptation of the AP frequency response or adaptation was calculated as the average adaptation of the interspike interval between consecutive APs.

(3) 'we excluded cells showing initial duplet or triplet action potential bursts'

Why were the burst cells excluded from the analysis?

We have modified the figures and text to include cells with initial burst firing.

(4) Electrophysiological characteristics to be analyzed:

Spike thresholds and afterhyperpolarizations

We found age-related differences in the amplitude of the afterhyperpolarization (P = 2.56*10^-30, Kruskal-Wallis test) and in the threshold of the action potential (P = 5.24*10^-12, Kruskal-Wallis test) (Author response image 3).

Author response image 3.

Age-dependence of afterhyperpolarization and AP threshold. (A-B) Boxplots show the differences in afterhyperpolarization (AHP) amplitude (A) and AP threshold (B) between age groups. Asterisks indicate statistical significance (* P < 0.05, ** P < 0.01, *** P < 0.001, Kruskal-Wallis test with post-hoc Dunn test). (C-D) Scatter plots show AHP amplitude (C) and AP threshold (D) across the lifespan. Age is shown on a logarithmic scale, dots are colored according to the corresponding age group.

(5) 'We identified and labeled each spine on n = 2 fully 3D-reconstructed cells'

To which cortical area do these cells belong?

At what depths are they distributed?

Is it possible to report the number of spines, in addition to the density per unit length?

We increased the number of cells in which we analyzed dendritic spine density. The data shown in Figure 6. are from pyramidal cells from an infant patient (n = 3 from a single patient) and late adulthood patients (n = 3 from 3 patients) (Supplementary Figure 13). The infant cells are from the same patient, the sample is from the right parietal lobe, and the patient is 83 days old. The older cells are from three different patients (#1: 65 years old, right temporal lobe; #2: 66 years old, right parietal lobe; #3: 62 years old, right frontal lobe). Infant cells are located 144.43 ± 45.26 µm (#1: 109.3, #2: 128.49, #3: 195.5 µm), late adult cells 161.22 ± 66.22 µm (#1: 183.5, #2: 213.42, #3: 86.73 µm) from the L1/2 border. We provide the number of spines in an additional supplementary table (Supplementary table 2.).

Author response:

The following is the authors’ response to the original reviews

We thank the reviewers for their careful review of our manuscript and the constructive comments. We have addressed the majority of comments with either new experiments, analyses, and/or text revisions. A summary of the major changes is listed below, followed by our point-by-point responses to the reviewer comments.

Major changes:

(1) We sought to gain insight into the potential mechanistic cause of the increased intrinsic excitability of Cntnap2^-/- dSPNs. Given that Kv1.1 and 1.2 potassium channels are known to interact with Caspr2 (the protein encoded by Cntnap2), we hypothesized that altered number, location, and/or function of these channels may underlie the excitability change in these cells. To investigate this, we performed new analyses of the initial dataset to assess action potential (AP) properties known to be impacted by potassium channel function. Indeed, we found that AP frequency was increased, and rheobase current, AP latency and AP threshold were decreased in Cntnap2^-/- dSPNs, suggestive of altered Kv1.2 function. These data are in the new Supplemental Fig. 4. We also performed new electrophysiology experiments in which we pharmacologically blocked Kv1.1 and 1.2 to assess whether the effects of blocking these channels would be occluded in Cntnap2^-/- dSPNs. We found that 1) WT dSPNs responded to blockade of Kv1.1/1.2 channels by increasing their excitability but Cntnap2^-/- dSPNs did not and 2) Kv1.1/1.2 channels were more important contributors to the excitability of dSPNs compared to iSPNs. These new data are presented in the revised Fig. 4 and Supplemental. Figs. 5 and 6.

(2) We performed additional experiments to assess excitatory synaptic properties, specifically AMPA/NMDA receptor ratio. This has been added to Fig. 1.

(3) We performed more rigorous statistical analyses of the initial physiology datasets to align with the statistics performed for the revision experiments. This applies to Fig. 1, Fig. 2, Fig. 3, Fig. 5, and Supp. Fig. 2.

(4) In the discussion section, we now highlight potential limitations of the study and further discuss the variable impact that Cntnap2 loss has on different cell types and brain regions.  

Reviewer #1 (Public Review):

Summary:

Cording et al. investigated how deletion of CNTNAP2, a gene associated with autism spectrum disorder, alters corticostriatal engagement and behavior. Specifically, the authors present slice electrophysiology data showing that striatal projection neurons (SPNs) are more readily driven to fire action potentials in response to stimulation of corticostriatal afferents, and this is due to increases in SPN intrinsic excitability rather than changes in excitatory or inhibitory synaptic inputs. The authors show that CNTNAP2 mice display repetitive behaviors, enhanced motor learning, and cognitive inflexibility. Overall the authors' conclusions are supported by their data, but a few claims could use some more evidence to be convincing.

Strengths:

The use of multiple behavioral techniques, both traditional and cutting-edge machine learning-based analyses, provides a powerful means of assessing repetitive behaviors and behavioral transitions/rigidity.

Characterization of both excitatory and inhibitory synaptic responses in slice electrophysiology experiments offers a broad survey of the synaptic alterations that may lead to increased corticostriatal engagement of SPNs.

Weaknesses:

(1) The authors conclude that increased cortical engagement of SPNs is due to changes in SPN intrinsic excitability rather than synaptic strength (either excitatory or inhibitory). One weakness is that only AMPA receptor-mediated responses were measured. Though the holding potential used for experiments in Figure 1FI wasn't clear, recordings were presumably performed at a hyperpolarized potential that limits NMDA receptormediated responses. Because the input-output experiments used to conclude that corticostriatal engagement of SPNs is elevated (Figure 1B-E) were conducted in the current clamp, it is possible that enhanced NMDA receptor engagement contributed to increased SPN responses to cortical stimulation. Confirming that NMDA receptor-mediated EPSC components are not altered would strengthen the main conclusion.

The reviewer is correct, the initial optically-evoked EPSC assessments were performed at a hyperpolarized potential (-70mV), thus measuring primarily AMPAR-mediated currents. We agree that assessing potential changes in the NMDAR-mediated EPSC component is important and we have completed new experiments to assess this. We find no differences in NMDAR-mediated EPSCs assessed at +40mV or the AMPA:NMDA ratio.

These results have been added to Fig. 1. An expanded analysis of these results is shown in Author response image 1. We note that the previous AMPAR-mediated EPSC results have been replicated in this additional experiment, again showing no change in Cntnap2^-/- SPNs. 

Author response image 1.

AMPA and NMDA receptor-mediated EPSCs are unchanged in Cntnap2^-/- SPNs. (A) Quantification (mean ± SEM) of AMPA:NMDA ratio per cell for Cntnap2^+/+ and Cntnap2^-/- dSPNs, p=0.9537, MannWhitney test. (B) dSPN AMPA current per cell, p=0.6172, Mann-Whitney test. (C) dSPN NMDA current per cell, p=0.6009, Mann-Whitney test. (D) dSPN AMPA:NMDA ratio averaged by animal, p=0.8413, Mann-Whitney test. (E) dSPN AMPA current averaged by animal, p>0.9999, Mann-Whitney test. (F) dSPN NMDA current averaged by animal, p=0.6905, Mann-Whitney test. (G) Quantification (mean ± SEM) of AMPA:NMDA ratio per cell for Cntnap2^+/+ and Cntnap2^-/- iSPNs, p=0.4104, Mann-Whitney test. (H) iSPN AMPA current per cell, p=0.9010, Mann-Whitney test. (I) iSPN NMDA current per cell, p=0.9512, two-tailed unpaired t test. (J) iSPN AMPA:NMDA averaged by animal, p=0.3095, Mann-Whitney test. (K) iSPN AMPA current averaged by animal, p=>0.9999, Mann-Whitney test. (L) iSPN NMDA current averaged by animal, p=0.8413, MannWhitney test. All values were recorded using 20% blue light intensity. For dSPNs: Cntnap2^+/+ n=22 cells from 5 mice, Cntnap2^-/- n=22 cells from 5 mice. For iSPNs: Cntnap2^+/+ n=21 cells from 5 mice, Cntnap2^-/-n=21 cells from 5 mice.

(2) Data clearly show that SPN intrinsic excitability is increased in knockout mice. Given that CNTNAP2 has been linked to potassium channel regulation, it would be helpful to show and quantify additional related electrophysiology data such as negative IV curve responses and action potential hyperpolarization.

We appreciate this suggestion. As indicated by the reviewer, Caspr2, has previously been shown to control the clustering of Kv1-family potassium channels in axons isolated from optic nerve and corpus callosum (PMIDs: 10624965, 12963709, 29300891). In particular, Caspr2 is known to associate directly with Kv1.2 (PMID: 29300891). To assess a potential contribution of Kv1.2 to the excitability phenotype, we performed additional analyses of our original dataset to quantify AP properties known to be impacted by changes in Kv1.2 function (i.e. latency to fire and AP threshold, new Supp. Fig. 4). We identified several changes in Cntnap2^-/- dSPNs resembling those that occur in wild-type cells when Kv1.2 is blocked (i.e. reduced threshold and reduced latency to fire, Supp. Fig. 4). 

We then performed a pharmacological experiment, blocking Kv1.2 using α-dendrotoxin (α-DTX) while recording intrinsic excitability to assess whether the effects of this drug on dSPN excitability were occluded in Cntnap2^-/- cells. Indeed, we found that while blocking Kv1.2 in wild-type dSPNs significantly reduced threshold and increased intrinsic excitability, these effects were not seen in Cntnap2^-/- dSPNs (new Fig. 4). We believe that this suggests an altered contribution of Kv1.2 to the intrinsic excitability of mutant dSPNs, owing to a change in the clustering, number, or function of these channels. Therefore, loss-of-function of Kv1.2 is a likely explanation for the enhanced intrinsic excitability of Cntnap2^-/- dSPNs. Interestingly, we found that α-DTX had only subtle effects on iSPNs (Cntnap2 WT or mutant), suggesting a lesser contribution of this channel in controlling the excitability of indirect pathway cells. This finding can account for the relatively stronger effect of Cntnap2 loss on dSPN physiology. The results of these new experiments and analyses are presented in the new Fig. 4, Supp. Fig. 5 and Supp. Fig. 6. 

(3) As it stands, the reported changes in dorsolateral striatum SPN excitability are only correlative with reported changes in repetitive behaviors, motor learning, and cognitive flexibility.

We agree that we have not identified a causative relationship between the change in dorsolateral dSPN excitability and the behaviors that we measured in Cntnap2^-/- mice. That said, in a previous study, we showed that selective deletion of the autism spectrum disorder (ASD) risk gene Tsc1 from dorsal striatal dSPNs resulted in increased corticostriatal drive and this was sufficient to increase rotarod motor learning (PMID: 34380034). Therefore, while we have not demonstrated causality in this study, we hypothesize that changes in dSPN excitability are likely to contribute to the behavioral phenotypes observed in Cntnap2^-/- mice. 

Reviewer #2 (Public Review):

Summary:

This is an important study characterizing striatal dysfunction and behavioral deficits in Cntnap2^-/- mice. There is growing evidence suggesting that striatal dysfunction underlies core symptoms of ASD but the specific cellular and circuit level abnormalities disrupted by different risk genes remain unclear. This study addresses how the deletion of Cntnap2 affects the intrinsic properties and synaptic connectivity of striatal spiny projection neurons (SPN) of the direct (dSPN) and indirect (iSPN) pathways. Using Thy1-ChR2 mice and optogenetics the authors found increased firing of both types of SPNs in response to cortical afferent stimulation. However, there was no significant difference in the amplitude of optically-evoked excitatory postsynaptic currents (EPSCs) or spine density between Cntnap2^-/- and WT SPNs, suggesting that the increased corticostriatal coupling might be due to changes in intrinsic excitability. Indeed, the authors found Cntnap2^-/- SPNs, particularly dSPNs, exhibited higher intrinsic excitability, reduced rheobase current, and increased membrane resistance compared to WT SPNs. The enhanced spiking probability in Cntnap2^-/- SPNs is not due to reduced inhibition. Despite previous reports of decreased parvalbumin-expressing (PV) interneurons in various brain regions of Cntnap2^-/- mice, the number and function (IPSC amplitude and intrinsic excitability) of these interneurons in the striatum were comparable to WT controls.

This study also includes a comprehensive behavioral analysis of striatal-related behaviors. Cntnap2^-/- mice demonstrated increased repetitive behaviors (RRBs), including more grooming bouts, increased marble burying, and increased nose poking in the holeboard assay. MoSeq analysis of behavior further showed signs of altered grooming behaviors and sequencing of behavioral syllables. Cntnap2^-/- mice also displayed cognitive inflexibility in a four-choice odor-based reversal learning assay. While they performed similarly to WT controls during acquisition and recall phases, they required significantly more trials to learn a new odor-reward association during reversal, consistent with potential deficits in corticostriatal function.

Strengths:

This study provides significant contributions to the field. The finding of altered SPN excitability, the detailed characterization of striatal inhibition, and the comprehensive behavioral analysis are novel and valuable to understanding the pathophysiology of Cntnap2^-/- mice.

Weaknesses:

(1) The approach based on Thy-ChR2 mice has the advantage of overcoming issues caused by injection efficiency and targeting variability. However, the spread of oEPSC amplitudes across mice shown in panels of Figure 1 G/I is very high with almost one order of magnitude difference between some mice. Given this is one of the most important points of the study it will be important to further analyze and discuss what this variability might be due to. Typically, in acute slice recordings, the within-animal variability is larger than the variability across animals. From the sample sizes reported it seems the authors sampled a large number of animals, but with a relatively low number of neurons per animal (per condition). Could this be one of the reasons for this variability?

We agree with the reviewer that the variability in these experiments is quite large. We have replicated these experiments in the process of performing AMPA:NMDA ratio recordings (see above response to Reviewer 1’s comment). We again find no differences in AMPAR-mediated EPSC amplitude between WT and mutant SPNs (Author response image 2). Notably, these experiments also demonstrate a large amount of variability. In the original dataset, a small number of cells were collected from each animal (~1-3 cells/mouse). However, the variability remains in the new dataset, in which more cells were collected from each animal (~4-6 cells/mouse). We find both withinanimal and between-animal variability, as can be seen in Author response image 2 (recordings made from the same animal are color-coordinated). Potential sources of variability in this experiment include: 1) variable expression of ChR2 per mouse, 2) variable innervation of ChR2-expressing terminals onto any given recorded cell, and/or 3) differences in prior plasticity state between cells (i.e. some neurons may have recently undergone corticostriatal LTP or LTD). 

Author response image 2.

Optically-evoked AMPAR EPSCs exhibit within- and between-animal variability. (A) Quantification of EPSC amplitude evoked in dSPNs at different light intensities from the original dataset, plotted by cell (line represents the mean, dots/squares represent average EPSC amplitude for each recorded cell). Cntnap2^+/+ n=17 cells from 8 mice, Cntnap2^-/- n=13 cells from 5 mice. Repeated measures two-way ANOVA p values are shown; g x s F (2, 56) = 0.3879, geno F (1, 28) = 0.8098, stim F (1.047, 29.32) = 76.56. (B) Quantification of EPSC amplitude evoked in dSPNs, averaged by mouse (line represents the mean, dots/squares represent average EPSC amplitude for each mouse). Cntnap2^+/+ n=8 mice, Cntnap2^-/- n=5 mice. Repeated measures two-way ANOVA p values are shown; g x s F (2, 22) = 0.2154, geno F (1, 11) = 0.2585, stim F (1.053, 11.58) = 49.68. (C) Quantification of EPSC amplitude in dSPNs from the revision dataset, plotted by cell (line represents the mean, dots/squares represent average EPSC amplitude for each recorded cell). Cntnap2^+/+ n=22 cells from 5 mice, Cntnap2^-/- n=22 cells from 5 mice. Repeated measures two-way ANOVA p values are shown; g x s F (2, 84) = 0.01885, geno F (1, 42) = 0.002732, stim F (1.863, 78.26) = 20.93. (D) Quantification of EPSC amplitude in dSPNs from the revision dataset, averaged by mouse (line represents the mean, dots/squares represent average EPSC amplitude for each mouse). Cntnap2^+/+ n=5 mice, Cntnap2^-/- n=5 mice. Repeated measures two-way ANOVA p values are shown; g x s F (2, 16) = 0.06288, geno F (1, 8) = 0.006548, stim F (1.585, 12.68) = 16.97. (E) Quantification of EPSC amplitude evoked in iSPNs from the original dataset, plotted by cell (line represents the mean, dots/squares represent average EPSC amplitude for each recorded cell). Cntnap2^+/+ n=13 cells from 6 mice, Cntnap2^-/- n=11 cells from 5 mice. Repeated measures two-way ANOVA p values are shown; g x s F (2, 44) = 0.9414, geno F (1, 22) = 1.333, stim F (1.099, 24.18) = 52.26. (F) Quantification of EPSC amplitude evoked in iSPNs from original dataset, averaged by mouse (line represents the mean, dots/squares represent average EPSC amplitude for each mouse). Cntnap2^+/+ n=6 mice, Cntnap2^-/- n=5 mice. Repeated measures two-way ANOVA p values are shown; g x s F (2, 18) = 0.4428, geno F (1, 9) = 0.5635, stim F (1.095, 9.851) = 23.82. (G) Quantification of EPSC amplitude evoked in iSPNs from the revision dataset, plotted by cell (line represents the mean, dots/squares represent average EPSC amplitude for each recorded cell). Cntnap2^+/+ n=21 cells from 5 mice, Cntnap2^-/- n=21 cells from 5 mice. Repeated measures two-way ANOVA p values are shown; g x s F (2, 80) = 0.04134, geno F (1, 40) = 0.007025, stim F (1.208, 48.31) = 102.9. (H) Quantification of EPSC amplitude evoked in iSPNs from the revision dataset, averaged by mouse (line represents the mean, dots/squares represent average EPSC amplitude for each mouse). Cntnap2^+/+ n=5 mice, Cntnap2^-/- n=5 mice. Repeated measures two-way ANOVA p values are shown; g x s F (2, 16) = 0.001865, geno F (1, 8) = 0.1004, stim F (1.179, 9.433) = 61.31.

(2) This is particularly important because the analysis of corticostriatal evoked APs in panels C and E is performed on pooled data without considering the variability in evoked current amplitudes across animals shown in G and I. Were the neurons in panels C/E recorded from the same mice as shown in G/I? If so, it would be informative to regress AP firing data (say at 20% LED) to the average oEPSC amplitude recorded on those mice at the same light intensity. However, if the low number of neurons recorded per mouse is due to technical limitations, then increasing the sample size of these experiments would strengthen the study.

We appreciate this point; however, the evoked AP experiment and the evoked EPSC experiment were performed on different mice, so it is not possible to correlate the data across experiments. While the evoked AP experiments were performed using potassium-based internal, we used a cesium-based internal to measure AMPAR-mediated EPSCs to more accurately detect synaptic currents. We note that the evoked AP experiments share a similar amount of variability as the evoked EPSC experiments, again possibly owing to variable expression of channelrhodopsin per mouse, variable innervation of ChR2-positive terminals onto individual cells, and/or differences in prior plasticity status between cells.  

(3) On a similar note, there is no discussion of why iSPNs also show increased corticostriatal evoked firing in Figure 1E, despite the difference in intrinsic excitability shown in Figure 3. This suggests other potential mechanisms that might underlie altered corticostriatal responses. Given the role of Caspr2 in clustering K channels in axons, altered presynaptic function or excitability could also contribute to this phenotype, but potential changes in PPR have not been explored in this study.

We have now performed more rigorous statistics on the data in Fig. 1 (repeated measures two-way ANOVA) such that the difference in corticostriatal evoked firing in Cntnap2^-/- iSPNs no longer reaches statistical significance. This is consistent with the modest but statistically non-significant effect of Cntnap2 loss on iSPN intrinsic excitability. We agree with the reviewer that presynaptic alterations could potentially contribute to the changes in cortically-driven action potentials, especially as this experiment was performed without any synaptic blockers present, and Cntnap2 is deleted from all cells. That said, if changes in presynaptic release probability accounted for the increased corticostriatal drive, we would expect to see differences in cortically-evoked EPSCs onto SPNs. 

While we can’t rule out the possibility of pre-synaptic changes, a straightforward explanation for our findings is that loss or alteration of Kv1.2 channel function is responsible for the increased excitability of Cntnap2^-/- dSPNs, resulting in enhanced spiking in response to cortical input. Given the fact that Kv1.2 channels appear less important for regulating iSPN excitability (see new Fig. 4 and Supp. Fig. 6), this can explain the greater impact of Cntnap2 loss on dSPN physiology.

(4) Male and female SPNs have different intrinsic properties but the number and/or balance of M/F mice used for each experiment is not reported.

We agree that this is an important consideration. Author response table 1 provides the sex breakdown for the intrinsic excitability experiments. While we did not explicitly power the experiments to test for sex differences, Author response image 3 shows the data separated by sex and genotype for the intrinsic excitability experiments. Within genotype, we find no significant differences between males and females, except for Cntnap2^-/- iSPNs which showed a significant interaction between sex and current step (Author response image 3F). Interestingly, while present in both sexes, the excitability shift of Cntnap2^-/- dSPNs may be slightly more pronounced in females compared to males (Author response image 3C and D). However, this result would require further validation with a greater sample size.

Author response table 1.

Numbers of male and female mice used for the intrinsic excitability experiments.

Author response image 3.

Enhanced excitability of Cntnap2^-/- dSPNs is present in both males and females. (A) Quantification (mean ± SEM) of the number of APs evoked in dSPNs in Cntnap2^+/+ males and females at different current step amplitudes. Cntnap2^+/+ males n=12 cells from 4 mice, Cntnap2^+/+ females n=8 cells from 4 mice. Repeated measures two-way ANOVA p values are shown; s x c F (28, 560) = 0.8992, sex F (1, 20) = 0.3754, current F (1.279, 25.57) = 56.85. (B) Quantification (mean ± SEM) of the number of APs evoked in dSPNs in Cntnap2^-/- males and females at different current step amplitudes. Cntnap2^-/- males n=12 cells from 4 mice, Cntnap2^-/- females n=11 cells from 4 mice. Repeated measures two-way ANOVA p values are shown; s x c F (28, 588) = 0.6752, sex F (1, 21) = 0.04534, current F (2.198, 46.15) = 78.89. (C) Quantification (mean ± SEM) of the number of APs evoked in dSPNs in Cntnap2^+/+ males and Cntnap2^-/- males at different current step amplitudes. Cntnap2^+/+ males n=12 cells from 4 mice, Cntnap2^-/- males n=12 cells from 4 mice. Repeated measures two-way ANOVA p values are shown; g x c F (28, 672) = 2.233, geno F (1, 24) = 3.746, current F (1.708, 40.98) = 79.82. (D) Quantification (mean ± SEM) of the number of APs evoked in dSPNs in Cntnap2^+/+ females and Cntnap2^-/- females at different current step amplitudes. Cntnap2^+/+ females n=8 cells from 4 mice, Cntnap2^-/- females n=11 cells from 4 mice. Repeated measures two-way ANOVA p values are shown; g x c F (28, 476) = 1.547, geno F (1, 17) = 5.912, current F (1.892, 32.17) = 58.76. (E) Quantification (mean ± SEM) of the number of APs evoked in iSPNs in Cntnap2^+/+ males and females at different current step amplitudes. Cntnap2^+/+ males n=10 cells from 4 mice, Cntnap2^+/+ females n=12 cells from 4 mice. Repeated measures two-way ANOVA p values are shown; s x c F (28, 560) = 1.236, sex F (1, 20) = 1.074, current F (2.217, 44.34) = 179.6. (F) Quantification (mean ± SEM) of the number of APs evoked in iSPNs in Cntnap2^-/- males and females at different current step amplitudes. Cntnap2^-/- males n=12 cells from 4 mice, Cntnap2^-/- females n=9 cells from 4 mice. Repeated measures two-way ANOVA p values are shown; s x c F (28, 532) = 2.513, sex F (1, 19) = 2.639, current F (1.858, 35.31) = 152.5. (G) Quantification (mean ± SEM) of the number of APs evoked in iSPNs in Cntnap2^+/+ males and Cntnap2^-/- males at different current step amplitudes. Cntnap2^+/+ males n=10 cells from 4 mice, Cntnap2^-/- males n=12 cells from 4 mice. Repeated measures twoway ANOVA p values are shown; g x c F (28, 560) = 0.4723, geno F (1, 20) = 0.5675, current F (2.423, 48.47) = 301.7. (H) Quantification (mean ± SEM) of the number of APs evoked in iSPNs in Cntnap2^+/+ females and Cntnap2^-/- females at different current step amplitudes. Cntnap2^+/+ females n=12 cells from 4 mice, Cntnap2^-/- females n=9 cells from 4 mice. Repeated measures two-way ANOVA p values are shown; g x c F (28, 532) = 1.655, geno F (1, 19) = 0.2322, current F (2.081, 39.55) = 99.45.

(5) There is no mention of how membrane resistance was calculated, and no I/V plots are shown.

Passive properties were calculated from the average of five -5 mV, 100 ms long test pulse steps applied at the beginning of every experiment. Membrane resistance was calculated from the double exponential curve fit. This has now been added to the methods section.

(6) It would be interesting to see which behavior transitions most contribute to the decrease in entropy. Are these caused by repeated or perseverative grooming bouts? Or is this inflexibility also observed across other behaviors? The transition map in Figure S5 shows the overall number of syllables and transitions but not their sequence during behavior. Can this be analyzed by calculating the ratio of individual 𝑢𝑖 × 𝑝𝑖,𝑗 × log2 𝑝𝑖,𝑗 factors across genotypes?

We thank the reviewer for raising an insightful question. Here we use a finite state Markov chain model to describe the syllable transitions in animal behavior. To quantify the randomness in the system, we calculated the entropy of the Markov chain (see methods section). The reviewer suggested calculating the partial entropy of the transition matrix, which would allow us to estimate the contribution of a subset of states to the entropy of the whole system, given by the equation:

The partial equation can indeed quantify the stochasticity, or “flexibility” in our context, of the sub-system containing only a subset of the behavior syllables. However, there are two main limitations to this approach:

(1) The partial entropy fails to account for the transitions connecting the subset with the rest of the states in the system

(2) The stationary distribution may not reflect the actual probabilities in the isolated sub-system S.

Consequently, the partial entropy cannot be directly interpreted as the fraction of contributions from specific syllable pairs or sub-system to the entropy of the whole system. To be more specific, while a significant difference between the same sub-system in WT and KO groups could indicate that the sub-system contributes significantly to the difference of overall entropy, a non-significant result does not mean that the sub-system does not contribute to overall entropy difference, as interactions between the sub-system and other notconsidered states are not accounted for.

Author response image 4.

Grooming syllables contribute to some but not all differences in syllable transitions in Cntnap2^-/- mice. We calculated the entropy of each syllable pair using 𝑢𝑖 × 𝑝𝑖,𝑗 × log2 𝑝𝑖,𝑗 for every syllable pair and every animal. We then statistically tested the difference between genotypes for each syllable pair using Mann-Whitney tests. This plot displays those adjusted p-values for each syllable pair between WT and KO groups. The significant p-values suggest that the transitions to syllables 24 and 25 are different between genotypes (note that these correspond to grooming syllables, see Fig. 5N). However, since the overall entropy is a summation of every pair, it is difficult to conclude that syllables 24 and 25 are the sole contributors to the different entropy we observed.

Reviewer #3 (Public Review):

Summary:

The authors analyzed Cntnap2 KO mice to determine whether loss of the ASD risk gene CNTNAP2 alters the dorsal striatum's function.

Strengths:

The results demonstrate that loss of Cntnap2 results in increased excitability of striatal projection neurons (SPNs) and altered striatal-dependent behaviors, such as repetitive, inflexible behaviors. Unlike other brain areas and cell types, synaptic inputs onto SPNs were normal in Cntnap2 KO mice. The experiments are welldesigned, and the results support the authors' conclusions.

Weaknesses:

The mechanism underlying SPN hyperexcitability was not explored, and it is unclear whether this cellular phenotype alone can account for the behavioral alterations in Cntnap2 KO mice. No clear explanation emerges for the variable phenotype in different brain areas and cell types.

We agree that identifying the mechanism by which Cntnap2 loss affects intrinsic excitability is interesting and important. We have added experiments to address this and conclude that the improper clustering, number, or function of Kv1.2 channels in Cntnap2^-/- dSPNs is likely responsible for their increased excitability. These channels are known to be clustered/organized in part by Caspr2 (PMIDs: 10624965, 12963709, 29300891), and Kv1.2 channels are known to play an important role in regulating excitability in SPNs (PMIDs: 13679409, 32075716). In the case of dSPNs, blocking these channels with α-DTX significantly increased the excitability of WT cells (as has been previously reported); however, this effect was occluded in mutant cells, perhaps owing to a decreased contribution of Kv1.2 channels to excitability in Cntnap2^-/- dSPNs. In addition, we found that blockade of these channels with α-DTX only modestly affected the excitability of iSPNs. Therefore, this can explain why loss of Cntnap2 more strongly affects the excitability of dSPNs. Please see new Fig. 4, Supp. Fig. 5 and Supp. Fig. 6 for these new data. 

We agree with the reviewer that we have not identified a causative relationship between the change in dSPN excitability and the behavioral alterations in Cntnap2^-/- mice. This is a limitation of the study. 

It is interesting to speculate on the root of the varying impacts to excitability that occur across different brain regions and cell types in Cntnap2^-/- mice. Increased excitability, as we see in dSPNs, has been identified in cerebellar Purkinje cells and L2/3 pyramidal neurons in somatosensory cortex in the context of Cntnap2 loss (PMIDs: 34593517, 30679017, 36793543). However, other cell types in Cntnap2^-/- mice have exhibited no change in excitability (mPFC, L2/3 pyramidal neurons, PMID: 31141683) or hypoexcitability (subset of L5/6 pyramidal neurons, PMID: 29112191). While all of these cell types express Kv1.2 channels, they fundamentally vary in their intrinsic properties, owing to the role that other ion channels play in membrane excitability. As a result, loss of Cntnap2 is expected to have a variable effect on excitability depending on the cell type and the complement of other ion channels that are present. In addition, an initial change in excitability may drive secondary, potentially compensatory, changes in other channels that lead to a different excitability state. These changes are also expected to be cell type-specific. We do note that both of the cell types that show increased excitability in the context of Cntnap2 loss have been shown to exhibit an α-DTX-sensitive Kv1 channel current, such that application of α-DTX results in increased firing of these cells (cerebellar Purkinje cells; PMIDs: 17087603, 16210348 and L2/3 pyramidal neurons in somatosensory cortex; PMID: 17215507). These findings are consistent with our results in Cntnap2^-/- dSPNs. 

Reviewer #1 (Recommendations For The Authors):

More thorough analysis of some of the manually quantified behaviors would be helpful. For example, only the grooming bout number was presented- what about the duration of bouts and total time grooming? Similarly, for the open field the number of center entries was reported but what about the total time in the center?

We have quantified the time spent grooming and total time spent in the center during the open field test from our original data (Author response image 5). These data were not originally included in the manuscript because they were recorded for only a subset of the total animals. For each of these measures we find trend level changes, which are consistent with the primary measures reported in the main manuscript. 

Author response image 5.

Time in center and time spent grooming trend towards an increase in Cntnap2^-/- mice.  (A) Quantification (mean ± SEM) of total time spent in the center of the open field during a 60 minute test, p=0.0656, Mann-Whitney test. (B) Time spent grooming during the first 20 minutes of the open field test, p=0.0611, Mann-Whitney test. For both measurements, Cntnap2^+/+ n=18 mice, Cntnap2^-/- n=19 mice.

Reviewer #3 (Recommendations For The Authors):

What accounts for the hyperexcitability observed in Cntnap2-deficient SPNs? The authors noted that excitability is reportedly increased, reduced, or unchanged in different brain areas. What accounts for this disparity? Is it about the subcellular localization of Kv1 channels? The authors may want to test this possibility experimentally. At least, they may want to test whether Kv1 channels are mislocalized in SPNs.

We agree that this is an important point, and we have performed additional experiments to address this. We find that the Kv1.2 blocker a-DTX significantly increases the excitability of WT dSPNs but not Cntnap2^-/- dSPNs. This suggests that the mechanism underlying dSPN hyperexcitability in Cntnap2 mutants is the improper clustering, number, or function of Kv1.2 channels. These channels are known to be clustered and organized in part by Caspr2 (PMIDs: 10624965, 12963709, 29300891) and have been shown to play an important role in regulating the excitability of SPNs (PMIDs: 13679409, 32075716). Interestingly, we find that a-DTX has less of an effect on the excitability of iSPNs, which may account for the greater impact of Cntnap2 loss on dSPNs. Please see new Fig. 4, Supp. Fig. 5 and Supp. Fig. 6 for these added data and analyses. 

Please see above response to Reviewer #3 for our speculation on the variable impact of Cntnap2 loss on different cell types and brain regions. 

We agree with the reviewer that assessing potential differences in subcellular localization of Kv1 channels in our model would bolster the conclusion that these channels are mislocalized in the Cntnap2^-/- striatum. We piloted these experiments using immunohistochemistry to stain for Kv1.1 and 1.2 but found that without very high-resolution imaging, it would be challenging to accurately quantify Kv1 puncta in a cell type-specific manner. We instead chose to investigate the functional contribution of Kv1 channels to the dSPN hyperexcitability phenotype through the a-DTX experiments outlined above. α-DTX strongly inhibits Kv1.2 channels, but also Kv1.1 channels to some extent (PMIDs: 12042352, 13679409). We find that the effects of a-DTX on SPN excitability are occluded in Cntnap2^-/- dSPNs; therefore, we conclude that Kv1.2 (and possibly Kv1.1) channels have reduced function in these cells. Further work will be needed to determine if this is a result of channel mislocalization or another type of alteration. 

The authors did not detect synaptic changes in Cntnap-deficient SPNs. This important observation should be briefly discussed in the context of previous work in other brain regions and cell types. For example, some studies reported structural and functional changes at excitatory synapses. The variable impact on synapses suggests distinct compensatory mechanisms in different brain areas.

Given the prior literature showing effects of Cntnap2 loss on synapses in other brain regions, we were surprised that striatal synapses were not impacted in our model. We agree with the reviewer that the variable changes in synaptic properties across brain regions in Cntnap2 mutant mice is likely a result of distinct compensatory changes in these regions. Differences may also arise depending on whether the synaptic changes originate from the post-synaptic cell or from pre-synaptic changes. An interesting direction for future studies would be to explore the developmental trajectory of excitability and synaptic changes to determine which may be initial perturbations versus those that are secondary and potentially compensatory.

Line 138: "synaptic excitability". How is this term defined? Consider "synaptic changes" instead.

“Synaptic excitability” was used to mean a change in the number and/or function of glutamate receptors. We have now changed this term to “excitatory synaptic changes.”

Consider a short paragraph to highlight some limitations of this study. For example, it is unclear whether SPN hyperexcitability results from a compensatory change in Cntnap2 KO mice and whether the behavioral phenotype is solely due to this cellular phenotype. The study focuses on cortical projections onto SPNs, but these cells receive inputs from other brain areas that were not explored. Lastly, no clear explanation emerges for the variable phenotype in different brain areas and cell types.

We thank the reviewer for this suggestion and have added several paragraphs to the discussion highlighting some limitations of this study.

We hypothesize that the dSPN hyperexcitability in Cntnap2^-/- mice is a primary change, due to the direct relationship between Caspr2 and Kv1.2 channels. The results of our -DTX experiments suggest that the function and/or contribution of these channels to excitability is altered in Cntnap2^-/- dSPNs. However, it is possible that there are additional changes in dSPNs that occur as a result of Cntnap2 loss and contribute to the hyperexcitability of these cells. Rather surprisingly, we don’t find evidence for altered excitatory (specifically from cortical inputs) or inhibitory synaptic function, suggesting lack of engagement of homeostatic mechanisms at the synaptic level.

We have not yet determined whether there is a causative relationship between the change in dSPN excitability and the behavioral alterations in Cntnap2^-/- mice. This is a limitation of the current study. In our discussion section, we highlight that the dSPN changes we observe in dorsolateral striatum (DLS) are known to be sufficient to enhance rotarod learning in other mouse models and thus supports a connection between this cellular change and behavior. For the other behaviors we measured, we acknowledge that both DLS and other striatal or extra-striatal brain regions have been implicated in these behaviors, and therefore less of a direct connection can be made. 

In terms of the inputs, we focused on cortical inputs given their known role in mediating motor and habit learning (PMID: 15242609, 16237445, 19198605). Notably, corticostriatal synapses have been shown to be altered across a variety of mouse models with mutations in ASD risk genes and therefore may be a point of convergence for disparate genetic insults (PMID: 31758607). We agree that the striatum receives inputs from a variety of brain regions, notably the thalamus, which we did not explore in this study. This would be an interesting area for future studies.

Finally, it is difficult to speculate on the root of the varying impacts to excitability that occur across different brain regions and cell types in Cntnap2^-/- mice. Please see above response to Reviewer #3 for some speculation on this point in regard to the potential involvement of Kv1.2 in the excitability changes in various Cntnap2^-/- cell types. To expand upon this, it is known that ASD-associated mutations can have varying impacts on cell function even across similar cell types within a given brain region – we have seen this between dSPNs and iSPNs (this study, PMIDs: 34380034, 39358043), as have other groups studying ASD risk gene mutations in striatum (PMID: 24995986). This differential impact of the same mutation on intrinsic and/or synaptic physiology across cell types has been identified in other brain regions as well (PMID: 22884327, 26601124). Differences in transcriptional programs, protein expression, neuronal morphology, synaptic inputs and plasticity state make up a non-exhaustive set of variables that will impact the physiological function of a neuron, both in terms of the direct but also indirect consequences of an ASD risk gene mutation. To better address this important question, future studies would benefit from a systematic approach to assessing physiological changes in a given ASD mouse model, both across development and across brain regions.

Author response:

The following is the authors’ response to the current reviews.

Public Reviews:

Reviewer #1 (Public review):

Summary:

This study by Wang et al. identifies a new type of deacetylase, CobQ, in Aeromonas hydrophila. Notably, the identification of this deacetylase reveals a lack of homology with eukaryotic counterparts, thus underscoring its unique evolutionary trajectory within the bacterial domain.

Strengths:

The manuscript convincingly illustrates CobQ's deacetylase activity through robust in vitro experiments, establishing its distinctiveness from known prokaryotic deacetylases. Additionally, the authors elucidate CobQ's potential cooperation with other deacetylases in vivo to regulate bacterial cellular processes. Furthermore, the study highlights CobQ's significance in the regulation of acetylation within prokaryotic cells.

Weaknesses:

The problem I raised has been well resolved. I have no further questions.

Thanks for your valuable comments very much.

Reviewer #2 (Public review):

In recent years, lots of researchers tried to explore the existence of new acetyltransferase and deacetylase by using specific antibody enrichment technologies and high resolution mass spectrometry. Here is an example for this effort. Yuqian Wang et al. studied a novel Zn2+- and NAD+-independent KDAC protein, AhCobQ, in Aeromonas hydrophila. They studied the biological function of AhCobQ by using biochemistry method and MS identification technology to confirm it. These results extended our understanding of the regulatory mechanism of bacterial lysine acetylation modifications. However, I find this conclusion is a little speculative, and unfortunately it also doesn't totally support the conclusion as the authors provided.

Major concerns:

- It is a little arbitrary to come to the title "Aeromonas hydrophila CobQ is a new type of NAD+- and Zn2+-independent protein lysine deacetylase in prokaryotes." It should be modified to delete the "in the prokaryotes" except that the authors get new more evidence in the other prokaryotes for the existence of the AhCobQ.

Thank you for your suggestion. However, I believe there has been some confusion regarding the title. In the revised manuscript we have already updated the title to: "Aeromonas hydrophila CobQ is a new type of NAD+- and Zn2+-independent protein lysine deacetylase."

This title does not include the phrase "in prokaryotes," as you mentioned. We kindly suggest verifying the version of the manuscript that was reviewed to ensure you are reviewing the most recent changes.

- I was confused about the arrangement of the supplementary results. Because there are no citations for Figures S9-S19.

Thank you for your feedback. It appears there may have been a misunderstanding, possibly due to reviewing an outdated version of the manuscript. In the revised manuscript we revised the supplementary figures and now have only 12 figures, all of which are correctly cited in the manuscript on pages 12 to 15. Below is a detailed list of the updated figure citations:

Figures S1: page 8, line 148;

Figures S2: page 9, line 168;

Figures S3 and S4: page 10, line 178;

Figures S5: page 10, line 186;

Figures S6: page 10, line 189;

Figures S7: page 12, line 221;

Figures S8-S10: page 13, line 245;

Figures S11: page 11, line 282;

Figures S12: page 15, line 286

- Same to the above, there are no data about Tables S1-S6.

Thank you for your attention to the supplementary materials. As with the figures, we have already uploaded the data for Tables S1-S6 in the revised manuscript on November 19, 2024, and properly cited Tables S1 – S6 in the manuscript. Below is the citation information:

Tables S1: page 10, line 194;

Tables S2: page 13, line 245;

Tables S3: page 21, line 438;

Tables S4: page 22, line 439;

Tables S5: page 22, line 445;

Tables S6: page 27, line 564.

Please note that Tables S3 – S4 include the chemical reagents, primers, and other experimental materials, which are not intended to be cited in the results section.)

- All the load control is not integrated. Please provide all of the load controls with whole PAGE gel or whole membrane western blot results. Without these whole results, it is not convincing to come the conclusion as the authors mentioned in the context.

Thank you for your comment. Please note that the full membrane western blot results were included in the revised manuscript. We hope this satisfies your request. If you need further clarification or additional data, please do not hesitate to let us know.

- Thoroughly review the materials & methods section. It is unclear to me what exactly the authors describe in the method. All the experimental designs and protocols should be described in detail, including growth conditions, assay conditions, and purification conditions, etc.

Thank you for your valuable suggestion. In response to your comment and previous feedback, we have alredy revised the Materials & Methods section thoroughly in the revised manuscript. The experimental details, including growth conditions, assay protocols, and purification procedures, are described in full on pages 22 to 30 of the revised manuscript.

- Include relevant information about the experiments performed in the figure legends, such as experimental conditions, replicates, etc. Often it is not clear what was done based on the figure legend description.

Thank you very much for your detailed feedback and suggestions. We have made sure to describe what each data point represents in the figure legends, as per the previous feedback. However, we would like to clarify that while we have provided detailed descriptions in the legends, the inclusion of every specific experimental condition in the figure legends could result in redundancy, as these details are already thoroughly outlined in the Materials & Methods section.

We hope this explanation addresses your concern.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

I have no further revision comments.

Thank you very much.

Reviewer #2 (Recommendations for the authors):

I carefully read the point-to-point response from the author. Although they listed lots of the reasons for the ugly results, it still can not persuade me to accept their conclusions. While, as I know, it is impossible to reject their work in eLife as it was sent out for peer-review. I also can't accuse them of being wrong, but I have my opinion on this point. That is not the results, but the attitude.

Thank you for your feedback. However, I must express some concerns regarding the nature of your comments. Based on the issues you've raised, it seems that you may have reviewed an outdated version of the manuscript. In the updated revision we addressed all the points you've raised, including the figure and table citations, experimental methods, and data integration.

We understand that differing opinions are part of the peer-review process, but we respectfully believe that your conclusion regarding our attitude is based on a misunderstanding, possibly caused by reviewing an incorrect version of the manuscript. We have always strived to approach this manuscript with utmost professionalism and have diligently responded to each of your concerns.

We sincerely suggest reviewing the latest version of our manuscript, and we welcome any further constructive feedback. We hope this clarifies any misunderstandings and look forward to your continued support.

Thank you for your time and thoughtful consideration.

The following is the authors’ response to the original reviews.

Public Reviews:

Reviewer #1 (Public review):

Summary:

This study by Wang et al. identifies a new type of deacetylase, CobQ, in Aeromonas hydrophila. Notably, the identification of this deacetylase reveals a lack of homology with eukaryotic counterparts, thus underscoring its unique evolutionary trajectory within the bacterial domain.

Strengths:

The manuscript convincingly illustrates CobQ's deacetylase activity through robust in vitro experiments, establishing its distinctiveness from known prokaryotic deacetylases. Additionally, the authors elucidate CobQ's potential cooperation with other deacetylases in vivo to regulate bacterial cellular processes. Furthermore, the study highlights CobQ's significance in the regulation of acetylation within prokaryotic cells.

Weaknesses:

The problem I raised has been well resolved. I have no further questions.

Reviewer #2 (Public review):

In recent years, lots of researchers tried to explore the existence of new acetyltransferase and deacetylase by using specific antibody enrichment technologies and high resolution mass spectrometry. Here is an example for this effort. Yuqian Wang et al. studied a novel Zn2+- and NAD+-independent KDAC protein, AhCobQ, in Aeromonas hydrophila. They studied the biological function of AhCobQ by using biochemistry method and MS identification technology to confirm it. These results extended our understanding of the regulatory mechanism of bacterial lysine acetylation modifications. However, I find this conclusion is a little speculative, and unfortunately, it also doesn't totally support the conclusion as the authors provided.

Reviewer #3 (Public review):

Summary:

This study reports on a novel NAD+ and Zn2+-independent protein lysine deacetylase (KDAC) in Aeromonas hydrophila, termed as AhCobQ (AHA_1389). This protein is annotated as a CobQ/CobB/MinD/ParA family protein and does not show similarity with known NAD+-dependent or Zn2+-dependent KDACs. The authors showed that AhCobQ has NAD+ and Zn2+-independent deacetylase activity with acetylated BSA by western blot and MS analyses. They also provided evidence that the 195-245 aa region of AhCobQ is responsible for the deacetylase activity, which is conserved in some marine prokaryotes and has no similarity with eukaryotic proteins. They identified target proteins of AhCobQ deacetylase by proteomic analysis and verified the deacetylase activity using site-specific Kac proteins. Finally, they showed that AhCobQ activates isocitrate dehydrogenase by deacetylation at K388.

Strengths:

The finding of a new type of KDAC has a valuable impact on the field of protein acetylation. The characters (NAD+ and Zn2+-independent deacetylase activity in an unknown domain) shown in this study are very unexpected.

Weaknesses:

(1) The characters (NAD+ and Zn2+-independent deacetylase activity in an unknown domain) shown in this study are very unexpected. To convince readers, MSMS data must be necessary to accurately detect (de)acetylation at the target site in the deacetylase activity assay. The authors showed the MSMS data in assays with acetylated BSA, but other assays only rely on western blot.

(2) They prepared site-specific Kac proteins and used them in deacetylase activity assays. Incorporation of acetyllysine at the target site should be confirmed by MSMS and shown as supplementary data.

(3) The authors imply that the 195-245 aa region of AhCobQ may represent a new domain responsible for deacetylase activity. The feature of the region would be of interest but is not sufficiently described in Figure 5. The amino acid sequence alignments with representative proteins with conserved residues would be informative. It would be also informative if the modeled structure predicted by AlphaFold is shown and the structural similarity with known deacetylases is discussed.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

The problem I raised has been well resolved. I have no further questions.

Reviewer #2 (Recommendations for the authors):

Questions to response of"-The load control is not all integrated. All of the load controls with whole PAGE gel or whole membrane western blot results should be provided. Without these whole results, it is not convincing to come to the conclusion that the authors have."

Just as the Authors answered. The Coomassie Blue R-350 staining outcomes from the PVDF membranes. That is a good control for the experiment. However, I still have several questions about it:

(1) The first is the quality of these Western blot. Why all the bands of these Western blot is so ugly? To tell the truth, it is very difficult to come to a conclusion from these poor western blots.

We appreciate your feedback regarding the quality of the Western blots presented in Figure 7. We believe the “ugly bands” you referred to reflect our results validating the functions of CobQ through the use of recombinant site-specific Kac protein substrates.

In our study, we meticulously engineered these recombinant site-specific Kac proteins using a two-plasmid system, based on foundational research published in Nature Chemical Biology (2017, 13(12): 1253-1260), which introduced the genetic encoding of Nε-acetyllysine into recombinant proteins. However, we faced a common challenge: protein truncation due to premature translation termination at the reassigned codon. This issue not only hampers protein yields, as discussed in ChemBioChem (2017, 18(20): 1973-1983), but also contributes to the suboptimal appearance of the Western blot results.

Despite conducting at least two independent repetitions for the Western blot analysis of the site-specific Kac proteins, which yielded consistent results, we recognize that the overall quality remains less than ideal. This variability is inherently related to the characteristics of the target proteins. Nevertheless, the primary aim of our manuscript is to validate the novel deacetylase activity of CobQ. We have provided multiple lines of evidence, including mass spectrometry (MS/MS) and Western blot analyses, to substantiate this claim. In response to your comments, we have decided to remove the ambiguous Western blot results from Figure 7, retaining only four figures that demonstrate significant differences across at least two independent replicates (Author response images 1-5). Additionally, we have included four biological replicates of the Western blot results for ICD Kac388 + CobQ in the supplementary materials (Author response image 5) to further validate the deacetylase function of CobQ.

Author response image 1.

Western blot validation of the Kac26 AcrA-2 protein substrates regulated by the three KDACs in two biological replicates.

Author response image 2.

Western blot validation of the Kac48 Sun protein substrates regulated by the three KDACs in two biological replicates.

Author response image 3.

Western blot validation of the Kac103 Sun protein substrates regulated by the three KDACs in two biological replicates.

Author response image 4.

Western blot validation of the Kac195 Eno protein substrates regulated by the three KDACs in three biological replicates.

Author response image 5.

Western blot validation of the Kac388 ICD protein substrates regulated by AhCobQ in this study. Each sample was independently repeated at least three time.

(2) The second is why some of the results are not from the same PVDF by comparing the Coommassie staining with the WB results just as authors responded. For example, the HrpA-K816 (ac), Eno-K195 (ac), ArcA-2-K26 (ac), ArcA-2-K26 (ac), IscS-K93(ac), A0KJ75-K81(ac), GyrB-K331(ac), GyrB-K449(ac), FtsA-K320(ac), FtsA-K409(ac), RecA-K279(ac), and the RecA-K306(ac). All of them are clearly not from the same staining results of PVDF membrane but from a new PVDF membrane.

We assure you that the R-350 stained PVDF membranes originate from the same Western blot membranes. However, we acknowledge that visual discrepancies may arise due to differences in imaging techniques. The Western blot results were scanned using a ChemiDoc MP (Bio-Rad, Hercules, CA, USA), while the Coomassie R-350 stained PVDF membranes were captured using a standard camera. These differences can create a misleading appearance, making it seem as though they come from different membranes.

It is also important to note that the intensity of the protein marker cannot be directly compared between the two imaging methods. As illustrated in Author response image 6, the protein marker at 70 kDa is clearly detectable in the Coomassie R-350 image, whereas it may not be as apparent in the Western blot result due to inherent differences in detection sensitivity.

Author response image 6.

The comparison of Western blotting and R-350 strained results of same protein marker in the same PVDF membrane. The protein marker located at 70 kDa can be detected easily in Coomassie R-350, while is difficult to display in WB result.

Additionally, we have removed some of the so-called "ugly" Western blot results in the updated manuscript and provided the original full film of the relevant images as an attachment. This documentation demonstrates that all the data you referenced originate from the same film, as shown in Figures 1-5.

(3) The third is why there is no replication for all these WB results. We should draw a conclusion with serious attitude, but not from the only one repeat, even say nothing about the poor results.

Thank you for your valuable suggestion. In the second version of the manuscript, we have included the original full film of the relevant images. While we previously explained the reasons behind the "ugly" Western blot results, we have decided to remove some, or even all, of these results from Figure 7 in the updated version. The related images will be updated in the supplementary materials (Figures 1-5 in responding letter and Figure 7 in the revised manuscript).

Furthermore, we have provided a more detailed discussion regarding the poor results in the updated manuscript to ensure clarity and transparency. We appreciate your understanding and hope these changes meet your expectations.

Questions to response of " L174-187, L795 (Please show the whole membrane (or PAGE gel) of the loading control of CobB, and CobQ, except for the Kac-BSA)".

(1) As we all educated that there is no control, and no biology. Where is the band of CobQ? Why do not stain the same PVDF membrane with R-350 staining but with a new membrane?

Thank you for your insightful feedback. As noted in our previous response, the absence of visible bands for AhCobQ and AhCobB on the Coomassie R-350 stained PVDF membrane is primarily due to the low loading amounts and protein loss during the Western blotting process.

To reinforce our findings, we repeated the analysis of the protein samples via SDS-PAGE, using the same loading quantity as in the previous Western blot shown in Figure 2 of the manuscript. As illustrated in Author response image 7, the bands for CobB and CobQ are discernible, albeit with significantly lower intensities compared to the Kac-BSA bands. Upon examining the full Coomassie R-350 stained PVDF membranes provided in Supplementary Material 1, we observe that the CobB and CobQ bands are not easily visible. This aligns with your observations and can be attributed to potential protein loss during the transfer from SDS-PAGE to the PVDF membrane.

Author response image 7.

The SDS-PAGE gel displayed the loading amounts of Kac-BSA and CobB/CobQ.

To enhance the visibility of the CobQ/CobB bands, we increased the loading of CobQ/CobB in a new Western blot experiment, using 2 µg of Kac-BSA in combination with 0.8 µg of CobQ/CobB. As shown in Figure 8, while the increasing amounts of Kac-BSA resulted in a more blurred signal, the bands for the recombinant CobQ and CobB proteins were clearly detectable. This indicates that both proteins were indeed involved in the in vitro protein deacetylation assay.

Author response image 8.

Western blot verified the deacetylase activity assay of AhCobQ and AhCobB on Kac-BSA.

Furthermore, we conducted a mass spectrometry analysis comparing Kac-BSA and Kac-BSA incubated with CobQ, as well as BSA without acetylation, against the A. hydrophila database with a cut-off of unique matched peptides >1. It is challenging to completely avoid contaminant detection during protein purification, especially when using high-resolution mass spectrometry. Our findings revealed that CobQ has the highest number of unique matched peptides (Author response table 1), while contaminants such as AHA_3036, AHA_0497, AHA_1279, and valS could be excluded, as they were present in Kac-BSA or BSA samples. Additionally, Tuf1, RplQ, GroEL, RpsF, RpsU, RpsB, RpsO, and RpsJ are known ribosomal subunits or chaperonins that are abundantly expressed in cells and may interact with various proteins, leading to contaminant detection.

Author response table 1.

LC MS/MS results of selected peptide quantification among Kac-BSA and Kac-BSA incubated with CobQ and BSA without acetylation against A. hydrophila database (unique matched peptides>1).

Although AceE, a pyruvate dehydrogenase E1 component, theoretically possesses deacetylase activity, this possibility is low. First, in the SDS-PAGE gel of the purified recombinant protein, CobQ is the major band, with other proteins present at very low levels (less than 1/10 of CobQ). This suggests that significant deacetylation by contaminants is unlikely. Second, we purified His-tagged AhCobQ and GST-fused AhCobQ separately and tested their deacetylase activities. As shown in Figure S4 of the updated manuscript, both purified AhCobQ proteins exhibited deacetylase activity, while the negative control (purified GST protein only) did not, further supporting our conclusion that enzyme activity is not attributable to contaminating proteins (Figure S5).

(2) Without the CobB and CobQ bands, it is impossible to say the function of CobQ is a new deacetylase. To avoid this confusion, it is easy to run a new gel and stain it with anti-His antibody to show these deacetylases.

Thank you very much for your suggestion. We have performed the experiment in the comment (1) as your suggestion.

(3) The explanation about the CobB/CobQ bands are not visible is not acceptable. Because the molecular weight of the CobB and CobQ is smaller than that of BSA, it is impossible that these bands will be loss during membrane transfer.

Thank you for your valuable feedback. I completely agree that the loss of CobB and CobQ proteins during membrane transfer is unlikely due to their smaller molecular weight compared to BSA. As shown in Figure 7, the bands for CobB and CobQ are detectable in the SDS-PAGE gel but not visible on the Coomassie R-350 stained PVDF membrane.

Several factors could contribute to this issue. One possibility is that the detection sensitivity of Coomassie R-350 may be lower than that of Coomassie R-250 used in the gel. Additionally, the Western blot results using an anti-His antibody further indicate low loading amounts of CobB and CobQ proteins on the PVDF membrane (Figure 8). This suggests that the observed low levels may indeed be due to protein loss during the membrane transfer process, despite their relatively small size.

Reviewer #3 (Recommendations for the authors):

(1) I found Tables S1 and S2 in the revised manuscript. It is strange to me that the intensity of Kac-BSA+CobQ is zero, completely nothing. Typically, a portion of the acetylated peptide remains after the deacetylation reaction.

Thank you for your observation. When we report an intensity of zero, it does not imply a complete absence of signal; rather, it indicates that the signal for the target peptide is below the detectable threshold. This is likely due to the minimum cut-off setting in the MaxQuant (MQ) software, which is determined by parameters like "peptide_mass_tolerance" (as discussed in MQ user groups online, though it may not be explicitly listed in the parameters file).

In our study, we performed a deacetylase assay that demonstrated CobQ's rapid activity; for instance, it can deacetylate ICD-K388ac within just four minutes. This leads me to hypothesize that the CobQ + Kac-BSA sample may have undergone near-complete enzymatic hydrolysis during the reaction.

Furthermore, Table S1 in manuscript presents only a selection of the mass spectrometry results to illustrate CobQ's activity. In addition to the 15 acetylated peptides shown, there are many more (27 peptides) that exhibit significantly reduced acetylation levels without reaching zero intensity. The overall acetylation level of BSA peptides incubated with CobQ is calculated to be only 0.13 times that of Kac-BSA (Diagnostic peak: yes, peptide score: >100, Localization probability: >0.95) (Author response image 9).

Based on these findings, we believe our mass spectrometry results are reliable and effectively support our conclusions. Thank you for your understanding.

Author response image 9.

The intensities of all Kac peptides of Kac-BSA with or without AhCobQ incubation in LC MS/MS.

(2) It would be better to provide the information about ArcA and ArcA-2 as mentioned in the authors' response. It would be helpful for readers to understand that they are different proteins.

Thank you for your suggestion. In the A. hydrophila ATCC 7966 dataset, there are indeed two distinct proteins referred to as ArcA: ArcA-1, which functions as an aerobic respiration control protein, and ArcA-2, which acts as an arginine deiminase. Importantly, these two proteins do not share any sequence homology; they are only similarly named due to their acronyms. While we believe this distinction does not require extensive explanation in the current study, we appreciate your input. Additionally, in response to Reviewer 2’s feedback, we have decided to remove the Western blot result for ArcA-2 due to its poor quality in the updated manuscript.

(3) Line 409-416. Despite my comment, the citation of related papers on ICD acetylation in E. coli is still missing.

Thank you for your suggestion. It has been added and highlighted in red. (Venkat S, et al, 2018, 430(13): 1901-1911)

(4) The image resolution of Figure 3C and 3D is still bad. I could not evaluate that Kac was exactly incorporated at the target site.

Thank you for your feedback regarding the image resolution of Figures 3C and 3D. We have now displayed these figures with improved clarity, as you suggested.

To further validate the reliability of our MS2 data, we employed Proteome Discoverer 2.4 (Thermo) to analyze the raw data and provide theoretical mass information. As shown in Author response images 10-13, the MS2 spectra and fragment match lists for both unmodified and acetylated peptides offer additional confirmation of the reliability of our mass spectrometry results.

Author response image 10.

MS2 spectrum of unmodified peptide using PD v2.4 software.

Author response image 11.

The theoretical mass of unmodified peptide by PD 2.4

Author response image 12.

MS2 spectrum of acetylated peptide using PD v2.4 software.

Author response image 13.

The theoretical mass of acetylated peptide by PD 2.4.

(5) Again, in Figure 8D, it should be shown the significance between ICD-Kac388 and ICD-Kac388+AhCobB to support the authors' conclusion that AhCobQ activates ICD by deacetylation at K388.

Thanks for your suggestion, we have updated the figure in Figure 8D in updated manuscript.

(6) It was nice that the authors presented the mass spectrum data of ICD-K388 acetylation (Figure 2 in responding letter). However, the data did not convince me that K388 is acetylated. In the figure, two b-ion peaks are detected, 285.1557 and 386.2034, which may correspond to NK (theoretical mass, 260.15) and NKT (theoretical mass, 361.20) peptides, respectively. If K388 is acetylated, an increase in the mass of 42 should be observed, but the difference between the detected and theoretical mass is 25. I also could not understand what the peak of 126.0913 mass is, indicated with acK* in red.

Thank you for your detailed observation. The data presented in the MS2 spectrum for ICD-K388 acetylation in Figure 2 of the previous response letter were generated using Proteome Discoverer 2.4 (PD, Thermo) to ensure accurate mass calculations. Similar to the results from MaxQuant, ICD-K388 was identified again (Author response image 14).

Regarding the b-ion peaks you mentioned, the values 285.1557 and 386.2034 correspond to NK^ac and NK^acT peptides, respectively. The theoretical masses for these peptides are as follows: NK^ac (285.15 = 115.05020 + 128.095 + 42.01) and NK^acT (386.20 = NK^ac + 101.04768). The differences between the theoretical and detected masses for the relevant b-ions (b2*-NK, b52+-NH3, and b3) are minimal, at 0.00 Da and 2.1 ppm, respectively, which is consistent with the incorporation of an NH3 group (Author response image 15).

Author response image 14.

The MS2 of ICD-K388 peptide by PD 2.4.

Author response image 15.

The theoretical mass of ICD-K388 peptide by PD 2.4.

The peak at 126.0913 m/z, indicated as acK*, represents immonium ions of ε-N-acetyllysine, which are generated during the fragmentation of acetyllysine. This diagnostic ion is widely recognized as a marker for identifying acetylated peptides (Nakayasu, et al,. A method to determine lysine acetylation stoichiometries. International journal of proteomics. 2014;2014(1):730725; Trelle et al., Utility of immonium ions for assignment of ε-N-acetyllysine-containing peptides by tandem mass spectrometry. Analytical chemistry. 2008;80(9):3422-30). Additionally, it is a default parameter in MaxQuant for identifying Kac peptides (Author response image 16).

Based on these findings, we believe the evidence supporting ICD-K388 acetylation is robust.

Author response image 16.

The default parameter in Kac peptide identification in Maxquant v1.6 software

(7) As mentioned by other reviewers, some of the figures and tables are incomplete. Some panels (ex. Figure 7C and 7D) and explanations (ex. What are lanes 1, 2, and 3 in Figure S3) are still missing.

Thank you for your suggestion. It has been added.

Author response:

The following is the authors’ response to the original reviews.

Reviewer #1 (Public Review): 

The authors of the study investigated the generalization capabilities of a deep learning brain age model across different age groups within the Singaporean population, encompassing both elderly individuals aged 55 to 88 years and children aged 4 to 11 years. The model, originally trained on a dataset primarily consisting of Caucasian adults, demonstrated a varying degree of adaptability across these age groups. For the elderly, the authors observed that the model could be applied with minimal modifications, whereas for children, significant fine-tuning was necessary to achieve accurate predictions. Through their analysis, the authors established a correlation between changes in the brain age gap and future executive function performance across both demographics. Additionally, they identified distinct neuroanatomical predictors for brain age in each group: lateral ventricles and frontal areas were key in elderly participants, while white matter and posterior brain regions played a crucial role in children. These findings underscore the authors' conclusion that brain age models hold the potential for generalization across diverse populations, further emphasizing the significance of brain age progression as an indicator of cognitive development and aging processes.

Strengths: 

(1) The study tackles a crucial research gap by exploring the adaptability of a brain age model across Asian demographics (Chinese, Malay, and Indian Singaporeans), enriching our knowledge of brain aging beyond Western populations.

(2) It uncovers distinct anatomical predictors of brain aging between elderly and younger individuals, highlighting a significant finding in the understanding of age-related changes and ethnic differences.

Weaknesses: 

(1) Clarity in describing the fine-tuning process is essential for improved comprehension.

(2) The analysis often limits its findings to p-values, omitting the effect sizes crucial for understanding the relationship with cognition.

(3) Employing a predictive framework for cognition using brain age could offer more insight than mere statistical correlations.

(4) Expanding the study's scope to evaluate the model's generalisability to unseen Caucasian samples is vital for establishing a comparative baseline.

In summary, this paper underscores the critical need to include diverse ethnicities in model testing and estimation.

Reviewer #1 (Recommendations for the authors): 

Comment #1 - Fine-Tuning Process Clarity: Enhanced clarity in the fine-tuning process documentation is crucial for understanding how models are adapted to new datasets. This involves explaining parameter adjustments and choices, which facilitates replication and application in further research.

We thank Reviewer #1 for this pertinent point. As advised, we have added a Supplementary Methods section with more details on the finetuning process. This includes the addition of Supplementary Figure S6, which shows examples of learning curves that helped inform our parameter adjustments and choices. We have added a reference to this section in Section 5.2 of the Methods.

Comment #2 - Effect Sizes Reporting: The emphasis on reporting effect sizes alongside p-values addresses the need to quantify the strength of observed effects, particularly the relationship between brain age and cognition. Effect sizes provide insights into the practical significance of findings, crucial for clinical and practical applications.

We thank Reviewer #1 for raising this important comment. As suggested, we have added standardized regression coefficients (as measures of effect size) alongside p-values in Figures 3 – 4, Supplementary Figures S2 – S4, Supplementary Tables S4 – S15, and the text of Sections 2.2 – 2.3 of the Results. We have additionally added 95% confidence intervals to Supplementary Tables S4 – S15.

Comment #3 - Predictive Framework for Cognition: Adopting a predictive framework for cognition using brain age moves the research from mere correlation to actionable prediction, offering potentials based on predictive analytics.

We thank Reviewer #1 for this insightful suggestion. Adopting a predictive framework would certainly be a useful and exciting avenue for the application of brain age. However, we note that the current study was primarily interested in the generalizability and interpretability of brain age in Asian children and older adults, as well as the added value of longitudinal measures of brain age. Thus, we believe our correlation-based analysis effectively demonstrated that deviations of brain age from chronological age were not merely random errors, but were informative of cognition. Furthermore, ongoing changes to these deviations were informative of future cognition. This helps to establish the brain age gap as a biomarker for aging, independent of chronological age. Additionally, we expect that the accurate prediction of future cognition would require a multitude of factors, in addition to T1-based brain age, as well as a large sample size to train and test. We believe such a dataset would be a promising avenue for future work, but it is outside the scope of the current study.

Nonetheless, we were able to conduct a preliminary analysis using the current longitudinal data from SLABS and GUSTO. We extracted the same variables used in the original analyses of future cognition, corresponding to Figures 3D and 4B in the main text. To implement a predictive framework, we split the data into 10 stratified cross-validation folds. We also used kernel ridge regression (KRR) as the predictive model, as it has previously shown promising performance in behavioral and cognitive prediction [1]. We used a cosine kernel and nested 5-fold cross-validation to pick the optimal regularization strength (alpha).

To investigate the added value of BAG and longitudinal changes in BAG, we compared 3 predictive models for each cognitive domain. The baseline model consisted of the demographic covariates used in the original analyses (i.e. chronological age, sex, and years of education for older adults). A second model combined demographics with baseline BAG, and the third model incorporated demographics, baseline BAG, and the (early) annual rate of change in BAG. Predictions were extracted from each test fold, and performance was measured by the correlation between test predictions and actual values of future cognition (or change in cognition). Models were statistically compared using the corrected resampled t-test for machine learning models [1], [2], [3]. The Benjamini-Hochberg procedure was used to correct for multiple comparisons.

Author response image 1 shows the prediction results for SLABS and GUSTO. Notably, adding the early change in BAG significantly improves the prediction of future change in executive function in SLABS. There is also an improvement in predicting the future inhibition score in GUSTO, but this is not significant after multiple comparison correction. Encouragingly, these are the same domains that showed significant associations with the change in BAG in the original analyses. This suggests that longitudinal brain age continues to contribute information, independent of baseline factors, in a predictive framework. We hope that future work can expand on this analysis with, for instance, larger sample sizes, more varied and informative predictors, and state-of-the-art prediction methods, in order to establish actionable predictions of future cognition.

Author response image 1.

Predictive framework for cognition similarly suggests value of longitudinal change in BAG. Prediction performance (Pearson's correlation) of KRR across future cognitive outcomes. Each boxplot shows the distribution of performance over cross-validation folds. Model performances are statistically compared for each outcome. Significant outcomes from the original analyses are bolded. (A) Results for SLABS using the early change in BAG and future change in cognitive scores (non-overlapping). Early change in BAG again shows benefit for predicting future change in executive function. (B) Results for GUSTO using the early change in BAG (from 4.5-7.5 years old) and future cognitive score (at 8.5 years old). Early change in BAG again shows benefit for predicting future inhibition, but it is not significant after multiple comparison correction. Key - **: p < 0.01; * (ns): p < 0.05 but p_corr > 0.05 after multiple comparison correction; ns: p > 0.05

Comment #4 - Generalizability to Unseen Caucasian Samples: Evaluating the model's performance on unseen (longitudinal) Caucasian samples is important for benchmarking.

We thank Reviewer #1 for this important comment. We agree that generalizability should be benchmarked against performance on unseen Caucasian samples. In the SFCN model paper [4], they conducted an out-of-sample test on unseen Caucasian samples from ages 13 to 95. In this age range, they reported a high correlation (r = 0.975) and low MAE (MAE = 3.90). This favorable generalization performance was verified in adults by independent evaluations [5], [6]. This is also in line with what we observed in Asian older adults, taking into account the different age ranges and sample sizes involved [7].

However, this also highlights the difficulty in evaluating on younger ages in the range of GUSTO (4.5 – 10.5 years old). Most accessible developmental datasets (e.g. HBN, PING) were already included in model training, preventing an unbiased evaluation on these samples. Datasets such as PNC and ABCD were not included in training, but they primarily consist of an older age range than GUSTO. Holm et al. [8] previously tested the SFCN model in ABCD and reported satisfactory performance (low MAE) from 9 – 13 years old. However, to the best of our knowledge, there are no reported generalization results (for any ethnicity) from 4.5 – 7.5 years old, which is where we found the most performance degradation in GUSTO. We are also not aware of any datasets in this age range we could access to test this, unfortunately, but it would be an important area for future work.

While benchmarking in Caucasian children is difficult, we were able to conduct a preliminary analysis with older adults using the ADNI dataset (which was not included in the model training [4]). We selected a longitudinal subset with cognitive data available and no dementia at baseline (N = 137). We used composite cognitive scores covering memory, executive function, language, and visuospatial function [9], [10], [11]. We followed the same methodology (e.g. preprocessing, finetuning, statistical analysis) as the main analyses on EDIS, SLABS, and GUSTO. To maximize the data available, we tested associations with future cognition (taken at the last available time point), similar to GUSTO. We again included chronological age, sex, and years of education as demographic covariates.

Author response image 2 shows the brain age predictions for the pretrained and finetuned models on ADNI. Similar to Singaporean older adults, the pretrained model performs well, producing a high correlation (r = 0.8053; compared to r = 0.7389 for EDIS and r = 0.8136 for SLABS) and somewhat low MAE (MAE = 4.9735; compared to MAE = 3.9895 for EDIS and MAE = 3.4668 for SLABS). After finetuning, the MAE improves (MAE = 3.6837; compared to MAE = 3.3232 for EDIS and MAE = 3.2653 for SLABS) with a similar correlation (r = 0.7854; compared to r = 0.7445 for EDIS and r = 0.8138 for SLABS). This suggests that generalization to unseen Singaporean older adults is in line with the generalization to unseen Caucasian older adults.

Author response image 2. 

Brain age predictions on unseen Caucasian sample of older adults. Predictions from the A) pretrained and B) finetuned brain age models on ADNI participants. Compare to Figure 2 of the main text.

For the associations with future cognition, we again find that baseline BAG does not associate with future cognition (Author response tables 1 and 2). However, encouragingly, we find that the early annual rate of change in BAG does associate with future memory, which is significant after multiple comparison correction for the finetuned model (Author response tables 2 and 3). This suggests  a degree of replicability to the original results, but interestingly, in a different domain (memory vs. executive function). In contrast to SLABS, which consists of healthy older adults recruited from the community, ADNI consists of participants at risk of AD recruited from memory clinics. Thus, this difference in domain could be due to factors such as a stronger signal for memory in the testing battery or greater variations in memory function and decline. However, it could also reflect other population differences between ADNI and SLABS. This is an intriguing area for future study, ideally with larger sample sizes and more diverse populations included.

Author response table 1.

Linear relationship between pretrained baseline BAG and future cognitive score in ADNI. Compare to Supplementary Tables S4 – S15 of the original text.

Author response table 2. 

Linear relationship between finetuned baseline BAG and future cognitive score in ADNI. Compare to Supplementary Tables S4 – S15 of the original text.

Author response table 3.

Linear relationship between pretrained change in BAG and future cognitive score in ADNI. Compare to Supplementary Tables S4 – S15 of the original text.

Author response table 4. 

Linear relationship between finetuned change in BAG and future cognitive score in ADNI. Compare to Supplementary Tables S4 – S15 of the original text.

References

(1) L. Q. R. Ooi et al., “Comparison of individualized behavioral predictions across anatomical, diffusion and functional connectivity MRI,” NeuroImage, vol. 263, p. 119636, Nov. 2022, doi: 10.1016/j.neuroimage.2022.119636.

(2) C. Nadeau and Y. Bengio, “Inference for the Generalization Error,” Mach. Learn., vol. 52, no. 3, pp. 239–281, Sep. 2003, doi: 10.1023/A:1024068626366.

(3) R. R. Bouckaert and E. Frank, “Evaluating the Replicability of Significance Tests for Comparing Learning Algorithms,” in Advances in Knowledge Discovery and Data Mining, H. Dai, R. Srikant, and C. Zhang, Eds., Berlin, Heidelberg: Springer, 2004, pp. 3–12. doi: 10.1007/978-3-540-24775-3_3.

(4) E. H. Leonardsen et al., “Deep neural networks learn general and clinically relevant representations of the ageing brain,” NeuroImage, vol. 256, p. 119210, Aug. 2022, doi: 10.1016/j.neuroimage.2022.119210.

(5) R. P. Dörfel et al., “Prediction of brain age using structural magnetic resonance imaging: A comparison of accuracy and test-retest reliability of publicly available software packages,” Neuroscience, preprint, Jan. 2023. doi: 10.1101/2023.01.26.525514.

(6) J. L. Hanson, D. J. Adkins, E. Bacas, and P. Zhou, “Examining the reliability of brain age algorithms under varying degrees of participant motion,” Brain Inform., vol. 11, no. 1, p. 9, Apr. 2024, doi: 10.1186/s40708-024-00223-0.

(7) A.-M. G. de Lange et al., “Mind the gap: Performance metric evaluation in brain-age prediction,” Hum. Brain Mapp., vol. 43, no. 10, pp. 3113–3129, Jul. 2022, doi: 10.1002/hbm.25837.

(8) M. C. Holm et al., “Linking brain maturation and puberty during early adolescence using longitudinal brain age prediction in the ABCD cohort,” Dev. Cogn. Neurosci., vol. 60, p. 101220, Feb. 2023, doi: 10.1016/j.dcn.2023.101220.

(9) P. K. Crane et al., “Development and assessment of a composite score for memory in the Alzheimer’s Disease Neuroimaging Initiative (ADNI),” Brain Imaging Behav., vol. 6, no. 4, pp. 502–516, Dec. 2012, doi: 10.1007/s11682-012-9186-z.

(10) L. E. Gibbons et al., “A composite score for executive functioning, validated in Alzheimer’s Disease Neuroimaging Initiative (ADNI) participants with baseline mild cognitive impairment,” Brain Imaging Behav., vol. 6, no. 4, pp. 517–527, Dec. 2012, doi: 10.1007/s11682-012-9176-1.

(11) S.-E. Choi et al., “Development and validation of language and visuospatial composite scores in ADNI,” Alzheimers Dement. Transl. Res. Clin. Interv., vol. 6, no. 1, p. e12072, 2020, doi: 10.1002/trc2.12072.

Author response:

The following is the authors’ response to the original reviews.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

Dong et al here have studied the impact of the small Ras-like GTPase Rab10 on the exocytosis of dense core vesicles (DVC), which are important mediators of neuropeptide signaling in the brain. They use optical imaging to show that lentiviral depletion of Rab10 in mouse hippocampal neurons in culture independent of the established defects in neurite outgrowth hamper DCV exocytosis. They further demonstrate that such defects are paralleled by changes in ER morphology and defective ER-based calcium buffering as well as reduced ribosomal protein expression in Rab10-depleted neurons. Re-expression of Rab10 or supplementation of exogenous L-leucine to restore defective neuronal protein synthesis rescues impaired DCV secretion. Based on these results they propose that Rab10 regulates DCV release by maintaining ER calcium homeostasis and neuronal protein synthesis.

Strengths:

This work provides interesting and potentially important new insights into the connection between ER function and the regulated secretion of neuropeptides via DCVs. The authors combine advanced optical imaging with light and electron microscopy, biochemistry, and proteomics approaches to thoroughly assess the effects of Rab10 knockdown at the cellular level in primary neurons. The proteomic dataset provided may be valuable in facilitating future studies regarding Rab10 function. This work will thus be of interest to neuroscientists and cell biologists.

We appreciate the positive evaluation of our manuscript.

Weaknesses:

While the main conclusions of this study are comparably well supported by the data, I see three major weaknesses:

(1) For some of the data the statistical basis for analysis remains unclear. I.e. is the statistical assessment based on N= number of experiments or n = number of synapses, images, fields of view etc.? As the latter cannot be considered independent biological replicates, they should not form the basis of statistical testing.

This is an important point and we agree that multiple samples from the same biological replicate are not independent observations. We reanalyzed all nested data using a linear mixed model and indicated this in the Methods section and the relevant figure legends (Brunner et al., 2022). In brief, biological replicates (individual neuronal cultures) were used as a linear predictor. Outliers were identified and excluded using the ROUT method in GraphPad. A fixed linear regression model was then fitted to the data using the lm() function in R. A one-way anova (analysis of variance) was used to assess whether including the experimental group as a second linear predictor (formula = y ~ Group + Culture) statistically improved the fit of a model without group information (formula = y ~ 1 + Culture). Post-hoc analysis was performed using the emmeans() function with Tukey’s adjustment when more than two experimental groups were present. Importantly, our conclusions remain unchanged.

(2) As it stands the paper reports on three partially independent phenotypic observations, the causal interrelationship of which remains unclear. Based on prior studies (e.g. Mercan et al 2013 Mol Cell Biol; Graves et al JBC 1997) it is conceivable that defective ER-based calcium signaling and the observed reduction in protein synthesis are causally related. For example, ER calcium release is known to promote pS6K1 phosphorylation, a major upstream regulator of protein synthesis and ribosome biogenesis. Conversely, L-leucine supplementation is known to trigger calcium release from ER stores via IP3Rs. Given the reported impact of Rab10 on axonal transport of autophagosomes and, possibly, lysosomes via JIP3/4 or other mediators (see e.g. Cason and Holzbaur JCB 2023) and the fact that mTORC1, the alleged target of leucine supplementation, is located on lysosomes, which in turn form membrane contacts with the ER, it seems worth analyzing whether the various phenotypes observed are linked at the level of mTORC1 signaling.

This is great suggestion that could indeed further clarify the potential interplay between ER-based Ca2+ signaling and protein synthesis. To address this, we assessed the phosphorylation level of pS6K1 in control and Rab10 knockdown (KD) neurons with or without leucine treatment. These data are included in the new Figure 8—figure supplement 1 in the revised manuscript. Our results indicate that pS6K1 phosphorylation was not upregulated in Rab10 KD neurons, suggesting that the level of mTORC1 signaling is not different between wild-type or KD neurons. Furthermore, leucine treatment increased the pS6K1 phosphorylation level, as expected, but this effect was similar in both groups. Hence, we conclude that differences in mTORC1 signaling induced by Rab10 loss is not a major factor in the observed impairment in protein synthesis.

Author response image 1.

Rab10 depletion does not upregulate mTORC1 pathway. (A)Typical immunoblot showing pS6K1 levels in each condition. (B) Quantification of relative pS6K1 levels in each condition. All Data are plotted as mean±s.e.m. (C) Control, Control + Leu: N = 2, n = 2, Rab10 KD, Rab10 KD + Leu: N = 2, n = 4.

(3) The claimed lack of effect of Rab10 depletion on SV exocytosis is solely based on very strong train stimulation with 200 Aps, a condition not very well suited to analyze defects in SV fusion. The conclusion that Rab10 loss does not impact SV fusion thus seems premature.

We agree that 200 APs stimulation might be too strong to detect specific effects on evoked synaptic vesicle release, although this stimulation pattern is an established pattern in hundreds of studies (Emperador-Melero et al., 2018; Granseth et al., 2006; Ivanova et al., 2021; Kwon and Chapman, 2011; Reshetniak et al., 2020). We have toned down our conclusions and clarified in the revised manuscript that Rab10 is dispensable for SV exocytosis evoked by intense stimulations. The corresponding statements in the text have been modified accordingly (p. 5, l. 98, 124) and in figure legend (p. 17, 490).

Reviewer #2 (Public Review):

Summary:<br /> In this paper, the authors assess the function of Rab10 in dense core vesicle (DCV) exocytosis using RNAi and cultured neurons. The author provides evidence that their knockdown (KD) is effective and provides evidence that DCV is compromised. They also perform proteomic analysis to identify potential pathways that are affected upon KD of Rab10 that may be involved in DCV release. Upon focusing on ER morphology and protein synthesis, the authors conclude that defects in protein synthesis and ER Ca2+ homeostasis contributes to the DVC release defect upon Rab10 KD. The authors claim that Rab10 is not involved in synaptic vesicle (SV) release and membrane homeostasis in mature neurons.

Strengths:

The data related to Rab10's role in DCV release seems to be strong and carried out with rigor. While the paper lacks in vivo evidence that this gene is indeed involved in DCV in a living mammalian organism, I feel the cellular studies have value. The identification of ER defect in Rab10 manipulation is not truly novel but it is a good conformation of studies performed in other systems. The finding that DCV release defect and protein synthesis defect seen upon Rab10 KD can be significantly suppressed by Leucine supplementation is also a strength of this work.

We appreciate the positive evaluation of our manuscript.

Weaknesses:

The data showing Rab10 is NOT involved in SV exocytosis seems a bit weak to me. Since the proteomic analysis revealed so many proteins that are involved in SV exo/encodytosis to be affected upon Rab10, it is a bit strange that they didn't see an obvious defect. Perhaps this could have been because of the protocol that the authors used to trigger SV release (I am not an E-phys expert but perhaps this could have been a 'sledge-hammer' manipulation that may mask any subtle defects)? Perhaps the authors can claim that DCV is more sensitive to Rab10 KD than SV, but I am not sure whether the authors should make a strong claim about Rab10 not being important for SV exocytosis.

We agree that 200 APs stimulation might be too strong to see specific effects on evoked synaptic vesicle release, although this stimulation pattern is an established pattern in hundreds of studies. We have toned down our conclusions and clarified in the revised manuscript that Rab10 is dispensable for SV exocytosis evoked by intense stimulations. The corresponding statements in the text have been modified accordingly (p. 5, l. 98, 124) and in figure legend (p. 17, 490).

Also, the authors mention "Rab10 does not regulate membrane homeostasis in mature neurons" but I feel this is an overstatement. Since the authors only performed KD experiments, not knock-out (KO) experiments, I believe they should not make any conclusion about it not being required, especially since there is some level of Rab10 present in their cells. If they want to make these claims, I believe the authors will need to perform conditional KO experiments, which are not performed in this study.

This is a valid point. We have changed the statement to “membrane homeostasis in mature neurons was unaffected by Rab10 knockdown” (p. 13, l.376-377).

Finally, the authors show that protein synthesis and ER Ca2+ defects seem to contribute to the defect but they do not discuss the relationship between the two defects. If the authors treat the Rab10 KD cells with both ionomycin and Leucine, do they get a full rescue? Or is one defect upstream of the other (e.g. can they see rescue of ER morphology upon Leucine treatment)? While this is not critical for the conclusions of the paper, several additional experiments could be performed to clarify their model, especially considering there is no clear model that explains how Rab10, protein synthesis, ER homeostasis, and Ca2+ are related to DCV (but not SV) exocytosis.

This is an important point and a great suggestion. We have now tested the rescue effects of leucine treatment on ER morphology, as suggested. These data are included in the new Figure 8—figure supplement 2 in the revised manuscript. Our results indicate that the same dose of leucine that rescues DCV fusion and protein translation failed to rescue ER morphology. Hence, the defects in ER morphology appear to be independent of the impaired protein translation.

Author response image 2.

Leucine supplementation does not rescue ER morphological deficiency in Rab10 KD neurons. (A) Typical examples showing the KDEL signals in each condition. (B) Quantification of RTN4 intensity in MAP2-positive dendrites. (C) The ratio of neuritic to somatic RTN4 intensity (N/S). All Data are plotted as mean±s.e.m. (B, C) Control: N = 3, n = 10; Rab10 KD: N = 3, n = 11; Rab10 KD + Leu: N = 3; n = 11. A one-way ANOVA tested the significance of adding experimental group as a predictor. **** = p<0.0001, ns = not significant.

Reviewer #3 (Public Review):

In the submitted manuscript, Dong and colleagues set out to dissect the role of the Rab10 small GTPase on the intracellular trafficking and exocytosis of dense core vesicles (DCVs). While the authors have already shown that Rab3 plays a central role in the exocytosis of DVC in mammalian neurons, the roles of several other Rab-members have been identified genetically, but their precise mechanism of action in mammalian neurons remains unclear. In this study, the authors use a carefully designed and thoroughly executed series of experiments, including live-cell imaging, functional calcium-imaging, proteomics, and electron microscopy, to identify that DCV secretion upon Rab10 depletion in adult neurons is primarily a result of dysregulated protein synthesis and, to a lesser extent, disrupted intracellular calcium buffering. Given that the full deletion of Rab10 has a deleterious effect on neurons and that Rab10 has a major role in axonal development, the authors cautiously employed the knock-down strategy from 7 DIV, to focus on the functional impact of Rab10 in mature neurons. The experiments in this study were meticulously conducted, incorporating essential controls and thoughtful considerations, ensuring rigorous and comprehensive results.

We are grateful for the positive evaluation of our manuscript.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

The work by Dong et al provides interesting and potentially important new insights into the connection between ER function and the regulated secretion of neuropeptides via DCVs. I suggest that the authors address the following points experimentally to increase the impact of this potentially important study.

Major points:

(1) As alluded to above, for some of the data the statistical basis for analysis remains unclear (examples are Figures 1C-F, J,K; Figure 2 1B-D,I-K; Figure 2 - Supplement 1D-F; Figure 2 - Supplement 2J,K, etc). I.e. is the statistical assessment based on N = number of experiments or n = number of synapses, images, fields of view etc.? As the latter cannot be considered independent biological replicates, they should not form the basis of statistical testing. The Ms misses also misses a dedicated paragraph on statistics in the methods section.

See reply to reviewer 1 above. We fully agree and solved this point.

(2) A main weakness of the paper is the missing connection between neuronal protein synthesis, and the observed structural and signaling defects at the level of the ER. I suggest that the authors analyze mTORC1 signaling in Rab10 depleted neurons and under rescue conditions (+Leu or re-expression of Rab10) as ribosome biogenesis is a major downstream target of mTORC1 and mTORC1 activity is related to lysosome position, which may be affected upon rab10 loss -either directly or via effects on the ER that forms tight contacts with lysosomes.

See reply to reviewer 1 above. We agreed and followed up experimentally.

(3) Related to the above: Does overexpression of SERCA2 restore normal DCV exocytosis in Rab10-depleted neurons? This would help to distinguish whether calcium storage and release at the level of the ER indeed contribute to the exocytosis defect.

This is an important point and a great suggestion. We have now tested the rescue effects of overexpression of SERCA2 on DCV fusion. These data are included in the new Figure 8—figure supplement 3 in the revised manuscript. SERCA2 OE failed to rescue the DCV fusion defects in Rab10 KD neurons.

Author response image 3.

Overexpression of SERCA2 does not rescue DCV fusion deficits in Rab10 KD neurons. (A) Typical examples showing the SERCA2 signals in each condition. (B) Cumulative plot of DCV fusion events per cell. (C) Summary graph of DCV fusion events per cell. (A) Total number of DCVs (total pool) per neuron, measured as the number of NPY-pHluorin puncta upon NH4Cl perfusion. (B) Fraction of NPY-pHluorin-labeled DCVs fusing during stimulation. All Data are plotted as mean±s.e.m. (C-E) Control: N = 2, n = 10; Rab10 KD: N = 2, n = 13; SERCA2 OE: N = 2; n = 15. A one-way ANOVA tested the significance of adding experimental group as a predictor. *** = p<0.001, ** = p<0.01, ns = not significant.

(4) The claimed lack of effect of Rab10 depletion on SV exocytosis is solely based on very strong train stimulation with 200 Aps, a condition not very well suited to analyze defects in SV fusion. The conclusion that Rab10 loss does not impact SV fusion thus seems premature. The authors should conduct additional experiments under conditions of single or few Aps (e.g. 4 or 10 Aps) to really assess whether or not Rab10 depletion alters SV exocytosis at the level of pHluorin analysis in cultured neurons.

See reply to reviewer 2 above. Agreed to and made textual adjustments to solve this

(5) Related to the above: I am puzzled by the data shown in Figure 1H-J: From the pHluorin traces shown I would estimate a tau value of about 20-30 s (e.g. decay to 1/e = 37% of the peak value). The bar graph in Figure 1K claims 3-4 s, clearly clashing with the data shown. Were these experiments conducted at RT (where expected tau values are in the range of 30s) or at 37{degree sign}C (one would expect taus of around 10 s in this case for Syp-pH)? I ask the authors to carefully check and possibly re-analyze their datasets.

This is indeed a mistake. We thank the reviewer for flagging this miscalculation. Our original Matlab script used for calculating the tau value contained an error and the datasets were normalized twice by mistake. We now reanalyzed the data and the corresponding figures and texts have been updated. Our conclusion that Rab10 KD does not affect SV endocytosis remains unchanged since the difference in tau between the control (28.5 s) and Rab10 KD (32.8 s) suffered from the same systematic error and were/are not significantly different.

(6) How many times was the proteomics experiment shown in Figure 3 conducted? I noticed that the data in panel H missed statistical analysis and error bars. Given the typical variation in these experiments, I suggest to only include data for proteins identified in at least 3 out of 4 experimental replicates.

We agree that this information has not been clear. We have now explained replication in the Methods section (p. 42, l. 879-885). In brief, the proteomics experiment presented in Fig 3 was conducted with two independent cultures (‘biological replicates’), hence, formally only two independent observations. For each biological replicate, we performed four technical replicates. For our analysis, we only included peptides that were consistently detected across all samples (not only three as this reviewer suggests). Proteins in Panel H are ER-related proteins that are significantly different from control neurons with an adjusted FDR ≤ 0.01 and Log2 fold change ≥ 0.56. The primary purpose of our proteomics experiments was to generate hypotheses and guide subsequent experiments and the main findings were corroborated by other experiments presented in the manuscript.

Minor:

(7) Figure 2 - supplement 3 and Figure 4 - supplement 3 are only mentioned in the discussion. The authors should consider referring to these data in the results section.

This is a valid point. We have now added a new statement “Moreover, only 10% of DCVs co-transport with Rab10” in the Results (p. 6-7, l. 162-164).

(8) Where is the pHluorin data shown in Figure 1 bleach-corrected? If so, this should be stated somewhere in the Ms. Moreover, the timing of the NH4Cl pulse should be indicated in the scheme in panel I.

We thank the reviewer for pointing these omissions out. We have now included information about the timing of NH4Cl pulse in panel I. We did not do bleach-correction for the pHluorin data shown in Figure 1. It has been shown that pHluorin is very stable with a bleaching rate in the alkaline state of 0.06% per second and 0.0024% per second in the quenched state (Balaji and Ryan, 2007). Indeed, we did not observe obvious photobleaching in the first 30s during our imaging as indicated by the average trace of pHluorin intensity in panel I.

(9) Page 3/ lines 59-60: "...strongest inhibition of neuropeptide accumulation...". What is probably meant is "...strongest inhibition of neuropeptide release".

We agree this statement is unclear. Sasidharan et al used a coelomocyte uptake assay as an indirect readout for DCV release. The ‘strongest inhibition of neuropeptide accumulation’ in coelomocytes in Rab10 mutant indicates DCV fusion deficits. We have now replaced the text with “Rab10 deficiency produces the strongest inhibition of neuropeptide release in C. elegans” to make it more clear.

Reviewer #3 (Recommendations For The Authors):

I strongly recommend the publishing of this study as a VOR with minor comments directed to the authors.

(1) In Figure 4, the authors should include examples of tubular ER at the synapse, especially as this is an interesting point discussed in ln 226-229. Are there noticeable changes in the ER-mitochondria contacts at the synaptic boutons?

We agree that examples of tubular ER at the synapse would improve the manuscript. We have now replaced the Figure 4A with such examples. We found it challenging to quantify ER-mitochondria contacts based on the electron microscopy (EM) images we currently have. The ER-mitochondria contact sites are quite rare in the cross-sections of our samples, making it difficult to perform a reliable quantitative analysis.

(2) The limited impairment of calcium-ion homeostasis in Rab10 KD neurons is very interesting. Would the overexpression of Rab10T23N mimic the effect of a KD scenario? Is there a separation of function for Rab10 in calcium homeostasis vs. the regulation of protein synthesis?

This is an interesting possibility. We tested this and expressed Rab10T23N in a new series of experiments. These data are presented as a new Figure 5 in the revised manuscript (p. 29). We observed that Ca2+ refilling after caffeine treatment was delayed to a similar extent in Rab10T23N-expressing and Rab10 KD neurons. While impaired Ca2+ homeostasis may affect protein synthesis through ER stress or mTORC1 activation, our findings indicate otherwise in Rab10 KD neurons. First, ATF4 levels, a marker of ER stress, were unaffected in Rab10 KD neurons. This indicates that any ER stress present is minimal or insufficient to significantly impact protein synthesis through this pathway. Second, we did not observe significant changes in mTORC1 activation in Rab10 KD neurons as indicated by a normal pS6K1 phosphorylation (see above). Based on these observations, we conclude that Rab10's roles in calcium homeostasis and protein synthesis are most likely separate.

(3) The authors indicate that the internal release of calcium ions from the ER has no effect on DCV trafficking and fusion without showing the data. It is important to include this data as the major impact of the study is the dissecting of the calcium effects in mammalian neurons from the previous studies in invertebrates.

We agree this is an important aspect in our reasoning. We are submitting the related manuscript on internal calcium stores to BioRVix. The link will be added to the consolidated version of our manuscript

(4) The distinction between Rab3 and Rab10 co-trafficking on DCVs should be reported in the Results (currently, Figure 2 - supplement 3 is only mentioned in the Discussion) as it helps to understand the effects on DCV fusion.

We agree. We now added a new statement “Moreover, only 10% of DCVs co-transport with Rab10” in the Results (p. 6, l. 162-163).

Reference:

Balaji, J., Ryan, T.A., 2007. Single-vesicle imaging reveals that synaptic vesicle exocytosis and endocytosis are coupled by a single stochastic mode. Proceedings of the National Academy of Sciences 104, 20576–20581. https://doi.org/10.1073/pnas.0707574105

Brunner, J.W., Lammertse, H.C.A., Berkel, A.A. van, Koopmans, F., Li, K.W., Smit, A.B., Toonen, R.F., Verhage, M., Sluis, S. van der, 2022. Power and optimal study design in iPSC-based brain disease modelling. Molecular Psychiatry 28, 1545. https://doi.org/10.1038/s41380-022-01866-3

Emperador-Melero, J., Huson, V., van Weering, J., Bollmann, C., Fischer von Mollard, G., Toonen, R.F., Verhage, M., 2018. Vti1a/b regulate synaptic vesicle and dense core vesicle secretion via protein sorting at the Golgi. Nat Commun 9, 3421. https://doi.org/10.1038/s41467-018-05699-z

Granseth, B., Odermatt, B., Royle, S.J., Lagnado, L., 2006. Clathrin-Mediated Endocytosis Is the Dominant Mechanism of Vesicle Retrieval at Hippocampal Synapses. Neuron 51, 773–786. https://doi.org/10.1016/j.neuron.2006.08.029

Ivanova, D., Dobson, K.L., Gajbhiye, A., Davenport, E.C., Hacker, D., Ultanir, S.K., Trost, M., Cousin, M.A., 2021. Control of synaptic vesicle release probability via VAMP4 targeting to endolysosomes. Science Advances 7, eabf3873. https://doi.org/10.1126/sciadv.abf3873

Kwon, S.E., Chapman, E.R., 2011. Synaptophysin Regulates the Kinetics of Synaptic Vesicle Endocytosis in Central Neurons. Neuron 70, 847–854. https://doi.org/10.1016/j.neuron.2011.04.001

Reshetniak, S., Fernández-Busnadiego, R., Müller, M., Rizzoli, S.O., Tetzlaff, C., 2020. Quantitative Synaptic Biology: A Perspective on Techniques, Numbers and Expectations. International Journal of Molecular Sciences 21, 7298. https://doi.org/10.3390/ijms21197298

Author response:

To Reviewer #1:

Thank you for your thorough review and comments on our work, which you described as “the role of neuritin in T cell biology studied here is new and interesting.”.  We have summarized your comments into two categories: biology and investigation approach, experimental rigor, and data presentation.

Biology and Investigation approach comments:

(1) Questions regarding the T cell anergy model:

Major point “(4) Figure 1E-H. The authors assume that this immunization protocol induces anergic cells, but they provide no experimental evidence for this. It would be useful to show that T cells are indeed anergic in this model, especially those that are OVA-specific. The lack of IL-2 production by Cltr cells could be explained by the presence of fewer OVA-specific cells, rather than by an anergic status.”

T cell anergy is a well-established concept first described by Schwartz’s group. It refers to the hyporesponsive T cell functional state in antigen-experienced CD4 T cells (Chappert and Schwartz, 2010; Fathman and Lineberry, 2007; Jenkins and Schwartz, 1987; Quill and Schwartz, 1987).  Anergic T cells are characterized by their inability to expand and to produce IL2 upon subsequent antigen re-challenge. In this paper, we have borrowed the existing in vivo T cell anergy induction model used by Mueller’s group for T cell anergy induction (Vanasek et al., 2006).  Specifically, Thy1.1+ Ctrl or Nrn1-/- TCR transgenic OTII cells were co-transferred with the congenically marked Thy1.2+ WT polyclonal Treg cells into TCR-/- mice.  After anergy induction, the congenically marked TCR transgenic T cells were recovered by sorting based on Thy1.1+ congenic marker, and subsequently re-stimulation ex vivo with OVA323-339 peptide. We evaluated the T cell anergic state based on OTII cell expansion in vivo and IL2 production upon OVA323-339 restimulation ex vivo.  

“The authors assume that this immunization protocol induces anergic cells, but they provide no experimental evidence for this.”

Because the anergy model by Mueller's group is well established (Vanasek et al., 2006), we did not feel that additional effort was required to validate this model as the reviewer suggested. Moreover, the limited IL2 production among the control cells upon restimulation confirms the validity of this model.

“The lack of IL-2 production by Cltr cells could be explained by the presence of fewer OVAspecific cells, rather than by an anergic status”.

Cells from Ctrl and Nrn1-/- mice on a homogeneous TCR transgenic (OTII) background were used in these experiments. The possibility that substantial variability of TCR expression or different expression levels of the transgenic TCR could have impacted IL2 production rather than anergy induction is unlikely.

Overall, we used this in vivo anergy model to evaluate the Nrn1-/- T cell functional state in comparison to Ctrl cells under the anergy induction condition following the evaluation of Nrn1 expression, particularly in anergic T cells.  Through studies using this anergy model, we observed a significant change in Treg induction among OTII cells. We decided to pursue the role of Nrn1 in Treg cell development and function rather than the biology of T cell anergy as evidenced by subsequent experiments.

Minor points “(6) On which markers are anergic cells sorted for RNAseq analysis?”

Cells were sorted out based on their congenic marker marking Ctrl or Nrn1-/- OTII cells transferred into the host mice.  We did not specifically isolate anergic cells for sequencing.

(2) Question regarding the validity of iTreg differentiation model.

Major point: “(5) Figure 2A-C and Figure 3. The use of iTregs to try to understand what is happening in vivo is problematic. iTregs are cells that have probably no equivalent in vivo, and so may have no physiological relevance. In any case, they are different from pTreg cells generated in vivo. Working with pTreg may be challenging, that is why I would suggest generating data with purified nTreg. Moreover, it was shown in the article of Gonzalez-Figueroa 2021 that Nrn1-/- nTreg retained a normal suppressive function, which would not be what is concluded by the authors of this manuscript. Moreover, we do not even know what the % of Foxp3 cells is in the iTreg used (after differentiation and 20h of re-stimulation) and whether this % is the same between Ctlr and Nrn1 KO cells.”.

We thank Reviewer #1 for their feedback. While it is true that iTregs made in vitro and in vivo generated pTregs display several distinctions (e. g., differences in Foxp3 expression stability, for example), we strongly disagree with this statement by Revieweer#1 “The use of iTregs to try to understand what is happening in vivo is problematic. iTregs are cells that have probably no equivalent in vivo, and so may have no physiological relevance.” The induced Treg cell (iTreg) model was established over 20 years ago (Chen et al., 2003; Zheng et al., 2002), and the model is widely adopted with over 2000 citations. Further, it has been instrumental in understanding different aspects of regulatory T cell biology (Hurrell et al., 2022; John et al., 2022; Schmitt and Williams, 2013; Sugiura et al., 2022).   

Because we have observed reduced pTreg generation in vivo, we choose to use the in vitro iTreg model system to understand the mechanistic changes involved in Treg cell differentiation and function, specifically, neuritin’s role in this process. We have made no claim that iTreg cell biology is identical to pTreg generated in vivo or nTreg cells. However, the iTreg culture system has proved to be a good in vitro system for deciphering molecular events involved in complex processes. As such, it remains a commonly used approach by many research groups in the Treg cell field (Hurrell et al., 2022; John et al., 2022; Sugiura et al., 2022). Moreover, applying the iTreg in vitro culture system has been instrumental in helping us identify the cell electrical state change in Nrn1-/- CD4 cells and revealed the biological link between Nrn1 and the ionotropic AMPA receptor (AMPAR), which we will discuss in the subsequent discussion. It is technically challenging to use nTreg cells for T cell electrical state studies due to their heterogeneous nature from development in an in vivo environment and the effect of manipulation during the nTreg cell isolation process, which can both affect the T cell electrical state.   

“Moreover, it was shown in the article of Gonzalez-Figueroa 2021 that Nrn1-/- nTreg retained a normal suppressive function, which would not be what is concluded by the authors of this manuscript.” 

We have also carried out nTreg studies in vitro in addition to iTreg cells. Similar to Gonzalez-Figueroa et al.'s findings, we did not observe differences in suppression function between Nrn1-/- and WT nTreg using the in vitro suppression assay. However, Nrn1-/- nTreg cells revealed reduced suppression function in vivo (Fig. 2D-L). In fact, Gonzalez-Figueroa et al. observed reduced plasma cell formation after OVA immunization in Treg-specific Nrn1-/- mice, implicating reduced suppression from Nrn1-/- follicular regulatory T (Tfr) cells. Thus, our observation of the reduced suppression function of Nrn1-/- nTreg toward effector T cell expansion, as presented in Fig. 2D-L, does not contradict the results from Gonzalez-Figueroa et al. Rather, the conclusions of these two studies agree that Nrn1 can play important roles in immune suppression observable in vivo that are not captured readily by the in vitro suppression assay.

“Moreover, we do not even know what the % of Foxp3 cells is in the iTreg used (after differentiation and 20h of re-stimulation) and whether this % is the same between Ctlr and Nrn1 KO cells.”

We have stated in the manuscript on page 7 line 208 that “Similar proportions of Foxp3+ cells were observed in Nrn1-/- and Ctrl cells under the iTreg culture condition, suggesting that Nrn1 deficiency does not significantly impact Foxp3+ cell differentiation”. In the revised manuscript, we will include the data on the proportion of Foxp3+ cells before iTreg restimulation.

(3) Confirmation of transcriptomic data regarding amino acids or electrolytes transport change

Minor point“(3) Would not it be possible to perform experiments showing the ability of cells to transport amino acids or electrolytes across the plasma membrane? This would be a more interesting demonstration than transcriptomic data.”

We appreciate Review# 1’s suggestion regarding “perform experiments showing the ability of cells to transport amino acids or electrolytes across the plasma membrane”.  We have indeed already performed such experiments corroborating the transcriptomics data on differential amino acid and nutrient transporter expression. Specifically, we loaded either iTreg or Th0 cells with membrane potential (MP) dye and measured MP level change after adding the complete set of amino acids (complete AA).  Upon entry, the charge carried by AAs may transiently affect cell membrane potential. Different AA transporter expression patterns may show different MP change patterns upon AA entry, as we showed in Author response image 1. We observed reduced MP change in Nrn1-/- iTreg compared to the Ctrl, whereas in the context of Th0 cells, Nrn1-/- showed enhanced MP change than the Ctrl. We can certainly include these data in the revised manuscript.

Author response image 1.

Membrane potential change induced by amino acids entry. a. Nrn1-/- or WT iTreg cells loaded with MP dye and MP change was measured upon the addition of a complete set of AAs. b. Nrn1-/- or WT Th0 cells loaded with MP dye and MP change was measured upon the addition of a complete set of AAs.

(4) EAE experiment data assessment

Minor point ”(5) Figure 5F. How are cells re-stimulated? If polyclonal stimulation is used, the experiment is not interesting because the analysis is done with lymph node cells. This analysis should either be performed with cells from the CNS or with MOG restimulation with lymph node cells.”

In the EAE study, the Nrn1-/- mice exhibit similar disease onset but a protracted non-resolving disease phenotype compared to the WT control mice.  Several reasons may contribute to this phenotype: 1. Enhanced T effector cell infiltration/persistence in the central nervous system (CNS); 2. Reduced Treg cell-mediated suppression to the T effector cells in the CNS; 3. Protracted non-resolving inflammation at the immunization site has the potential to continue sending T effector cells into CNS, contributing to persistent inflammation. Based on this reasoning, we examined the infiltrating T effector cell number and Treg cell proportion in the CNS.  We also restimulated cells from draining lymph nodes close to the inflammation site, looking for evidence of persistent inflammation.  When mice were harvested around day 16 after immunization, the inflammation at the local draining lymph node should be at the contraction stage.  We stimulated cells with PMA and ionomycin intended to observe all potential T effector cells involved in the draining lymph node rather than only MOG antigen-specific cells.  We disagree with Reviewer #1’s assumption that “This analysis should either be performed with cells from the CNS or with MOG restimulation with lymph node cells.”. We think the experimental approach we have taken has been appropriately tailored to the biological questions we intended to answer.

Experimental rigor and data presentation.

(1) Data labeling and additional supporting data

Major points (2) The authors use Nrn1+/+ and Nrn1+/- cells indiscriminately as control cells on the basis of similar biology between Nrn1+/+ and Nrn1+/- cells at homeostasis. However, it is quite possible that the Nrn1+/- cells have a phenotype in situations of in vitro activation or in vivo inflammation (cancer, EAE). It would be important to discriminate Nrn1+/- and Nrn1+/+ cells in the data or to show that both cell types have the same phenotype in these conditions too.

(3) Figure 1A-D. Since the authors are using the Nrp1 KO mice, it would be important to confirm the specificity of the anti-Nrn1 mAb by FACS. Once verified, it would be important to add FACS results with this mAb in Figures 1A-C to have single-cell and quantitative data as well.

Minor points  

(1) Line 119, 120 of the text. It is said that one of the most up-regulated genes in anergic cells is Nrn1 but the data is not shown.

(2) For all figures showing %, the titles of the Y axes are written in an odd way. For example, it is written "Foxp3% CD4". It would be more conventional and clearer to write "% Foxp3+ / CD4+" or "% Foxp3+ among CD4+".

(4) For certain staining (Figure 3E, H) it would be important to show the raw data, in addition to MFI or % values.

We can adapt the labeling and provide additional data, including Nrn1 staining on Treg cells and flow graphs for pmTOR and pS6 staining (Fig. 3H), as requested by Reviewer #1.

(2) Experimental rigor:

General comments:

“However, it is disappointing that reading this manuscript leaves an impression of incomplete work done too quickly.”

We were discouraged to receive the comment, “this manuscript leaves an impression of incomplete work done too quickly.” Our study of this novel molecule began without any existing biological tools such as antibodies, knockout mice, etc.  Over the past several years, we have established our own antibodies for Nrn1 detection, obtained and characterized Nrn1 knockout mice, and utilized multiple approaches to identify the molecular mechanism of Nrn1 function. Through the use of the in vitro iTreg system described in this manuscript, we identified the association of Nrn1 deficiency with cell electrical state change, potentially connected to AMPAR function. We have further corroborated our findings by generating Nrn1 and AMPAR T cell specific double knockout mice and confirmed that T cell specific AMPAR deletion could abrogate the phenotype caused by the Nrn1 deficiency (see Author response image 2).  We did not include the double knockout data in the current manuscript because AMPAR function has not yet been studied thoroughly in T cell biology, and we feel this topic warrants examination in its own right.  However, the unpublished data support the finding that Nrn1 modulates the T cell electrical state and, consequently, metabolism, ultimately influencing tolerance and immunity.  In its current form, the manuscript represents the first characterization of the novel molecule Nrn1 in anergic cells, Tregs, and effector T cells. While this work has led to several exciting additional questions, we disagree that the novel characterization we have presented Is incomplete. We feel that our present data set, which squarely highlights Nrn1’s role as an important immune regulator while shedding unprecedented light on the molecular events involved, will be of considerable interest to a broad field of researchers.

“Multiple models have been used, but none has been studied thoroughly enough to provide really conclusive and unambiguous data. For example, 5 different models were used to study T cells in vivo. It would have been preferable to use fewer, but to go further in the study of mechanisms.”

We have indeed used multiple in vivo models to reveal Nrn1's function in Treg differentiation, Treg suppression function, T effector cell differentiation and function, and the overall impact on autoimmune disease. Because the impact of ion channel function is often context-dependent, we examined the biological outcome of Nrn1 deficiency in several in vivo contexts.  We would appreciate it if Reviewer#1 would provide a specific example, given the Nrn1 phenotype, of how to proceed deeper to investigate the electrical change in the in vivo models.

“Major points (1) A real weakness of this work is the fact that in most of the results shown, there are few biological replicates with differences that are often small between Ctrl and Nrn1 -/-. The systematic use of student's t-test may lead to thinking that the differences are significant, which is often misleading given the small number of samples, which makes it impossible to know whether the distributions are Gaussian and whether a parametric test can be used. RNAseq bulk data are based on biological duplicates, which is open to criticism.”

We respectfully disagree with Reviewer #1 on the question of statistical power and significance to our work. We have used 5-8 mice/group for each in vivo model and 3-4 technical replicates for the in vitro studies, with a minimum of 2-3 replicate experiments. These group sizes and replication numbers are in line with those seen in high-impact publications. While some differences between Ctrl and Nrn1-/- appear small, they have significant biological consequences, as evidenced by the various Nrn1-/- in vivo phenotypes. Furthermore, we believe we have subjected our data to the appropriate statistical tests to ensure rigorous analysis and representation of our findings.

To Reviewer #2.

We thank Reviewer #2 for the careful review of the manuscript. We especially appreciate the comments that “The characterizations of T cell Nrn1 expression both in vitro and in vivo are comprehensive and convincing. The in vivo functional studies of anergy development, Treg suppression, and EAE development are also well done to strengthen the notion that Nrn1 is an important regulator of CD4 responsiveness.”

“The major weakness of this study stems from a lack of a clear molecular mechanism involving Nrn1. “  

We fully understand this comment from Reviewer #2. The main mechanism we identified contributing to the functional defect of Nrn1-/- T cells involves novel effects on the electric and metabolic state of the cells. Although we referenced neuronal studies that indicate Nrn1 is the auxiliary protein for the ionotropic AMPA-type glutamate receptor (AMPAR) and may affect AMPAR function, we did not provide any evidence in this manuscript as the topic requires further in-depth study.   

For the benefit of this discussion, we include our preliminary Nrn1 and AMPAR double knockout data (Author response image 2), which indicates that abrogating AMPAR expression can compensate for the defect caused by Nrn1 deficiency in vitro and in vivo. This preliminary data supports the notion that Nrn1 modulates AMPAR function, which causes changes in T cell electric and metabolic state, influencing T cell differentiation and function.  

Author response image 2.

Deletion of AMPAR expression in T cells compensates for the defect caused by Nrn1 deficiency. Nrn1-/- mice were crossed with T cell-specific AMPAR knockout mice (AMPARfl/flCD4Cre+) mice. The following mice were generated and used in the experiment: T cell specific AMPAR-knockout and Nrn1 knockout mice (AKONKO), Nrn1 knockout mice (AWTNKO), Ctrl mice (AWTNWT). a. Deletion of AMPAR compensates for the iTreg cell defect observed in Nrn1-/- CD4 cells. iTreg live cell proportion, cell number, and Ki67 expression among Foxp3+ cells 3 days after aCD3 restimulation. b. Deletion of AMPAR in T cells abrogates the enhanced autoimmune response in Nrn1-/- Mouse in the EAE disease model. Mouse relative weight change and disease score progression after EAE disease induction.  

Ion channels can influence cell metabolism through multiple means (Vaeth and Feske, 2018; Wang et al., 2020). First, ion channels are involved in maintaining cell resting membrane potential. This electrical potential difference across the cell membrane is essential for various cellular processes, including metabolism (Abdul Kadir et al., 2018; Blackiston et al., 2009; Nagy et al., 2018; Yu et al., 2022). Second, ion channels facilitate the movement of ions across cell membranes. These ions are essential for various metabolic processes. For example, ions like calcium (Ca2+), potassium (K+), and sodium (Na+) play crucial roles in signaling pathways that regulate metabolism (Kahlfuss et al., 2020). Third, ion channel activity can influence cellular energy balance due to ATP consumption associated with ion transport to maintain ion balances (Erecińska and Dagani, 1990; Gerkau et al., 2019). This, in turn, can impact processes like ATP production, which is central to cellular metabolism. Thus, ion channel expression and function determine the cell’s bioelectric state and contribute to cell metabolism (Levin, 2021).

Because the AMPAR function has not been thoroughly studied using a genetic approach in T cells, we do not intend to include the double knockout data in this manuscript before fully characterizing the T cell-specific AMPAR knockout mice.  

“Although the biochemical and informatics studies are well-performed, it is my opinion that these results are inconclusive in part due to the absence of key "naive" control groups. This limits my ability to understand the significance of these data.

Specifically, studies of the electrical and metabolic state of Nrn1-/- inducible Treg cells (iTregs) would benefit from similar data collected from wild-type and Nrn1-/- naive CD4 T cells.”

We appreciate the reviewer’s comments. This comment reflects two concerns in data interpretation:

(1) Are Nrn1-/- naïve T cells fundamentally different from WT cells? Does this fundamental difference contribute to the observed electrical and metabolic phenotype in iTreg or Th0 cells? This is a very good question we will perform the experiments as the reviewer suggested. While Nrn1 is expressed at a basal (low) level in naïve T cells, deletion of Nrn1 may cause changes in naïve T cell phenotype.   

(2) Is the Nrn1-/- phenotype caused by Nrn1 functional deficiency or due to the secondary effect of Nrn1 deletion, such as non-physiological cell membrane structure changes?

We have done the following experiment to address this concern.  We have cultured WT T cells in the presence of Nrn1 antibody and compared the outcome with Nrn1-/- iTreg cells (Author response image 3). WT iTreg cells under antibody blockade exhibited similar changes as Nrn1-/- iTreg cells, confirming the physiological relevance of the Nrn1-/- phenotype.

Author response image 3.

Nrn1 antibody blockade in WT iTreg cell culture caused similar phenotypic change as in Nrn1-/- iTreg cells. Nrn1-/- and WT CD4 cells were differentiated under iTreg condition in the presence of anti-Nrn1 (aNrn1) antibody or isotype control for 3 days. Cells were restimulated with anti-CD3 and in the presence of aNrn1 or isotype. a. MP measured 18hr after anti-CD3 restimulation. b. live CD4 cell number and proportion of Ki67 expression among live cells three days after restimulation. c. The proportion of Foxp3+ cells among live cells three days after restimulation.  

Reference:

Abdul Kadir, L., M. Stacey, and R. Barrett-Jolley. 2018. Emerging Roles of the Membrane Potential: Action Beyond the Action Potential. Front Physiol 9:1661.

Blackiston, D.J., K.A. McLaughlin, and M. Levin. 2009. Bioelectric controls of cell proliferation: ion channels, membrane voltage and the cell cycle. Cell Cycle 8:3527-3536.

Chappert, P., and R.H. Schwartz. 2010. Induction of T cell anergy: integration of environmental cues and infectious tolerance. Current opinion in immunology 22:552-559.

Chen, W., W. Jin, N. Hardegen, K.J. Lei, L. Li, N. Marinos, G. McGrady, and S.M. Wahl. 2003. Conversion of peripheral CD4+CD25- naive T cells to CD4+CD25+ regulatory T cells by TGF-beta induction of transcription factor Foxp3. The Journal of experimental medicine 198:1875-1886.

Erecińska, M., and F. Dagani. 1990. Relationships between the neuronal sodium/potassium pump and energy metabolism. Effects of K+, Na+, and adenosine triphosphate in isolated brain synaptosomes. J Gen Physiol 95:591-616.

Fathman, C.G., and N.B. Lineberry. 2007. Molecular mechanisms of CD4+ T-cell anergy. Nat Rev Immunol 7:599-609.

Gerkau, N.J., R. Lerchundi, J.S.E. Nelson, M. Lantermann, J. Meyer, J. Hirrlinger, and C.R. Rose. 2019. Relation between activity-induced intracellular sodium transients and ATP dynamics in mouse hippocampal neurons. The Journal of physiology 597:5687-5705.

Hurrell, B.P., D.G. Helou, E. Howard, J.D. Painter, P. Shafiei-Jahani, A.H. Sharpe, and O. Akbari. 2022. PD-L2 controls peripherally induced regulatory T cells by maintaining metabolic activity and Foxp3 stability. Nature communications 13:5118.

Jenkins, M.K., and R.H. Schwartz. 1987. Antigen presentation by chemically modified splenocytes induces antigen-specific T cell unresponsiveness in vitro and in vivo. The Journal of experimental medicine 165:302-319.

John, P., M.C. Pulanco, P.M. Galbo, Jr., Y. Wei, K.C. Ohaegbulam, D. Zheng, and X. Zang. 2022. The immune checkpoint B7x expands tumor-infiltrating Tregs and promotes resistance to anti-CTLA-4 therapy. Nature communications 13:2506.

Kahlfuss, S., U. Kaufmann, A.R. Concepcion, L. Noyer, D. Raphael, M. Vaeth, J. Yang, P. Pancholi, M. Maus, J. Muller, L. Kozhaya, A. Khodadadi-Jamayran, Z. Sun, P. Shaw, D. Unutmaz, P.B. Stathopulos, C. Feist, S.B. Cameron, S.E. Turvey, and S. Feske. 2020. STIM1-mediated calcium influx controls antifungal immunity and the metabolic function of nonpathogenic Th17 cells. EMBO molecular medicine 12:e11592.

Levin, M. 2021. Bioelectric signaling: Reprogrammable circuits underlying embryogenesis, regeneration, and cancer. Cell 184:1971-1989.

Nagy, E., G. Mocsar, V. Sebestyen, J. Volko, F. Papp, K. Toth, S. Damjanovich, G. Panyi, T.A. Waldmann, A. Bodnar, and G. Vamosi. 2018. Membrane Potential Distinctly Modulates Mobility and Signaling of IL-2 and IL-15 Receptors in T Cells. Biophys J 114:2473-2482.

Quill, H., and R.H. Schwartz. 1987. Stimulation of normal inducer T cell clones with antigen presented by purified Ia molecules in planar lipid membranes: specific induction of a long-lived state of proliferative nonresponsiveness. Journal of immunology (Baltimore, Md. : 1950) 138:3704-3712.

Schmitt, E.G., and C.B. Williams. 2013. Generation and function of induced regulatory T cells. Frontiers in immunology 4:152.

Sugiura, A., G. Andrejeva, K. Voss, D.R. Heintzman, X. Xu, M.Z. Madden, X. Ye, K.L. Beier, N.U. Chowdhury, M.M. Wolf, A.C. Young, D.L. Greenwood, A.E. Sewell, S.K. Shahi, S.N. Freedman, A.M. Cameron, P. Foerch, T. Bourne, J.C. Garcia-Canaveras, J. Karijolich, D.C. Newcomb, A.K. Mangalam, J.D. Rabinowitz, and J.C. Rathmell. 2022. MTHFD2 is a metabolic checkpoint controlling effector and regulatory T cell fate and function. Immunity 55:65-81.e69.

Vaeth, M., and S. Feske. 2018. Ion channelopathies of the immune system. Current opinion in immunology 52:39-50.

Vanasek, T.L., S.L. Nandiwada, M.K. Jenkins, and D.L. Mueller. 2006. CD25+Foxp3+ regulatory T cells facilitate CD4+ T cell clonal anergy induction during the recovery from lymphopenia. Journal of immunology (Baltimore, Md. :1950) 176:5880-5889.

Wang, Y., A. Tao, M. Vaeth, and S. Feske. 2020. Calcium regulation of T cell metabolism. Current opinion in physiology 17:207-223.

Yu, W., Z. Wang, X. Yu, Y. Zhao, Z. Xie, K. Zhang, Z. Chi, S. Chen, T. Xu, D. Jiang, X. Guo, M. Li, J. Zhang, H. Fang, D. Yang, Y. Guo, X. Yang, X. Zhang, Y. Wu, W. Yang, and D. Wang. 2022. Kir2.1-mediated membrane potential promotes nutrient acquisition and inflammation through regulation of nutrient transporters. Nature communications 13:3544.

Zheng, S.G., J.D. Gray, K. Ohtsuka, S. Yamagiwa, and D.A. Horwitz. 2002. Generation ex vivo of TGF-beta-producing regulatory T cells from CD4+CD25- precursors. Journal of immunology (Baltimore, Md. : 1950) 169:4183-4189.

Author response:

Reviewer #1 (Public Review):

Summary:

Li et al investigated how adjuvants such as MPLA and CpG influence antigen presentation at the level of the Antigen-presenting cell and MHCII : peptide interaction. They found that the use of MPLA or CpG influences the exogenous peptide repertoire presented by MHC II molecules. Additionally, their observations included the finding that peptides with low-stability peptide:MHC interactions yielded more robust CD4+ T cell responses in mice. These phenomena were illustrated specifically for 2 pattern recognition receptor activating adjuvants. This work represents a step forward for how adjuvants program CD4+ Th responses and provides further evidence regarding the expected mechanisms of PRR adjuvants in enhancing CD4+ T cell responses in the setting of vaccination.

Strengths:

The authors use a variety of systems to analyze this question. Initial observations were collected in an H pylori model of vaccination with a demonstration of immunodominance differences simply by adjuvant type, followed by analysis of MHC:peptide as well as proteomic analysis with comparison by adjuvant group. Their analysis returns to peptide immunization and analysis of strength of relative CD4+ T cell responses, through calculation of IC:50 values and strength of binding. This is a comprehensive work. The logical sequence of experiments makes sense and follows an unexpected observation through to trying to understand that process further with peptide immunization and its impact on Th responses. This work will premise further studies into the mechanisms of adjuvants on T cells.

Weaknesses:

Comment 1. While MDP has a different manner of interaction as an adjuvant compared to CpG and MPLA, it is unclear why MDP has a different impact on peptide presentation and it should be further investigated, or at minimum highlighted in the discussion as an area that requires further investigation.

Thank you for the suggestion. We investigated the reasons for the different effects of MDP on peptide presentation compared with those of CpG and MPLA. We found that the expression of some proteins involved in antigen processing and presentation, such as CTSS, H2-DM, Ifi30, and CD74, was substantially lower in the MDP-treated group than in the CpG- and MPLA-treated groups. To further confirm whether these proteins play a key role during adjuvant modification of peptide presentation, we knocked down them using shRNA and then performed immunopeptidomics. The original mass spectra and peptide spectrum matches have been deposited in the public proteomics repository iProX (https://www.iprox.cn/page/home.html) under accession number IPX0007611000. Unfortunately, the expected results for peptide presentation repertoires were not observed. Thus, we hypothesized that the different effects of MDP on peptide presentation might not result from differences in protein expression. We cannot exclude the possibility that some other proteins that may be important in this process were overlooked. We are still working on the mechanisms and do not have an exact conclusion. Thus, we did not present related data in this manuscript.

The related statements were added in the Discussion section on page 13, lines 292–299: “In this study, we found that the peptide repertoires presented by APCs were significantly affected by the adjuvants CpG and MPLA, but not MDP. All three adjuvants belong to the PRR ligand adjuvant family. CpG and MPLA bind to TLRs and MDP is recognized by NOD2. Although the receptors are different, many common molecules are involved both in TLR and NLD pathway activation. Unfortunately, we did not demonstrate why the MDP had different impacts on peptide presentation compared with other adjuvants. Further investigation is required to clarify the mechanism by which MPLA, CpG, and MDP adjuvants modulate the presentation of peptides with different stabilities.”

Comment 2. It is alluded by the authors that TLR activating adjuvants mediate selective, low affinity, exogenous peptide binding onto MHC class II molecules. However, this was not demonstrated to be related specifically to TLR binding. I wonder if some work with TLR deficient mice (TLR 4KO for example) could evaluate this phenomenon more specifically.

Thank you for the suggestion. This is an important point that was overlooked in this study. Based on published research on the mechanisms of PRR adjuvants, CpG and MPLA, we believe that the effect of CpG and MPLA on APCs-selective epitope presentation needs to be bound to the corresponding receptor, although we did not give a definitive conclusion in the manuscript.

To confirm the TLR-activating adjuvants affecting peptides presented on MHC molecules specifically through TLR binding, we have used CRISPR-cas9 to knock out TLR4 and TLR9 of A20 cells and repeated the experiments, as suggested. We chose TLR4- and TLR9- knockout A20 cell lines instead of TLR-deficient mice because a large number of APCs are required for immunopeptidomics. Moreover, the data observed in this study were based on the A20 cell line. However, these experiments are time-consuming. Unfortunately, we were unable to provide timely data. In addition, we believe that elucidating the downstream molecular mechanisms of TLR activation is necessary, as mentioned in comment 1. All these data will be combined and reported in our upcoming publications.

Comment 3. It is unclear to me if this observation is H pylori model/antigen-specific. It may have been nice to characterize the phenomenon with a different set of antigens as supplemental. Lastly, it is unclear if the peptide immunization experiment reveals a clear pattern related to high and low-stability peptides among the peptides analyzed.

Q1: It is unclear to me if this observation is H. pylori model/antigen-specific. It may have been nice to characterize the phenomenon with a different set of antigens as supplemental.

Thank you for the comment. To confirm the effect of the adjuvant on the exogenous peptide repertoire presented by MHC II molecules, a set of antigens from another bacterium, Pseudomonas aeruginosa, was used, and the experiments were repeated. The A20 cells were treated with CpG and pulsed with Pseudomonas aeruginosa antigens. Twelve hours later, MHC-II–peptide complexes were immunoprecipitated, and immunopeptidomics were performed. The data are shown below (Author response image 1). Information on the MHC-peptides from Pseudomonas aeruginosa is given in the Supplementary Table named “Table S3 Response to comment3”. A total of 713 and 205 bacterial peptides were identified in the PBS and CpG groups (Author response image 1A). The number of exogenous peptides in the CpG-treated group was significantly lower than that in the PBS-treated control group (Author response image 1B). A total of 568 bacterial peptides were presented only in the PBS group; 60 bacterial peptides were presented in the CpG-treated group, and 145 bacterial peptides were presented in both groups (Author response image 1C). We then analyzed the MHC-binding stability of the peptides present in the adjuvant-treated group and that of the peptide-deficient after adjuvant stimulation using the IEDB website. We found that the IC50 of the peptides in the adjuvant-treated group were much higher than those of the deficient peptides, which indicated that the peptides presented in the CpG-treated groups have lower binding stability for MHC-II (Author response image 1D). These results indicate that CpG adjuvant affects the presentation of exogenous peptides with high binding stability, which is consistent with the data reported in our manuscript. Using another set of antigens, we confirmed that our observations were not H. pylori model- or antigen-specific.

Author response image 1.

MHC-II peptidome measurements in adjuvant-treated APCs pulsed with Pseudomonas aeruginosa antigens. (A) Total number of bacterial peptides identified in the PBS- and CpG-treated groups. (B) The number and length distribution of bacterial peptides in different groups were compared. (C) Venn diagrams showing the distribution of bacterial peptides in different groups. (D) IC50 of the presented, deficient, and co-presented peptides post-adjuvant stimulation from immunopeptidome binding to H2-IA and H2-IE were predicted using the IEDB website. High IC50 means low binding stability. *p<0.05, **p<0.01.

Q2: Lastly, it is unclear if the peptide immunization experiment reveals a clear pattern related to high and low-stability peptides among the peptides analyzed.

In this study, we used a peptide immunization experiment to evaluate the responses induced by the screened peptides with different stabilities. In addition to this method, tetramer staining and ELISA have been used to assess epitope-specific T-cell proliferation and cytokine secretion. Among these, tetramer staining is often used in studies involving model antigens. However, as many peptides were screened in our study, synthesizing a sufficient number of tetramers was difficult. However, we believe that the experimental data obtained in this study support the conclusion. Nevertheless, we agree that more methods applied will make the pattern more clearly.

Reviewer #2 (Public Review):

Adjuvants boost antigen-specific immune responses to vaccines. However, whether adjuvants modulate the epitope immunodominance and the mechanisms involved in adjuvant's effect on antigen processing and presentation are not fully characterized. In this manuscript, Li et al report that immunodominant epitopes recognized by antigen-specific T cells are altered by adjuvants.

Using MPLA, CpG, and MDP adjuvants and H. pylori antigens, the authors screened the dominant epitopes of Th1 responses in mice post-vaccination with different adjuvants and found that adjuvants altered antigen-specific CD4+ T cell immunodominant epitope hierarchy. They show that adjuvants, MPLA and CpG especially, modulate the peptide repertoires presented on the surface of APCs. Surprisingly, adjuvant favored the presentation of low-stability peptides rather than high-stability peptides by APCs. As a result, the low stability peptide presented in adjuvant groups elicits T cell response effectively.

Thanks a lot for your comments.

Reviewer #1 (Recommendations For The Authors):

Recommendation 1. Figure 6: The peptides considered low affinity- it would be helpful to specify from which adjuvant they were collected from. When they are pooled it is unclear if we are analyzing peptides collected from adjuvanting with any of the three adjuvants studied.

Thank you for the suggestion. The related description in Figure 6 has been modified in the revised manuscript. Data for the peptides identified from the adjuvants MPLA- and CpG-treated groups are shown separately.

Recommendation 2. It is unclear to me why the A20 cell line is less preferred to the J774 line for the immunopeptidome analysis - can the authors expand on this?

We apologize for not clearly explaining this in the original manuscript. In fact, the A20 cell line is better than J774A.1 cell line for immunopeptidomics experiments. Compared to J774A.1 cells, more MHC-II peptides were obtained from a smaller number of A20 cells using immunopeptidomics. At the beginning of this study, we chose the J774A.1 cell line as it is a macrophage cell line. J774A.1 cells (up to 5×108) were pulsed with the antigens, and MHC-II–peptide complexes were eluted from the cell surface for immunopeptidomics. Unfortunately, only a few hundred peptides from the host were detected and no exogenous peptides were detected. Next, we tested the A20 cell line. In total, 108 A20 cells were used in this study. More than 3500 host peptides and approximately 50 exogenous peptides have been identified. These data indicate that the A20 cell line was better.

To investigate the reasons for this, we detected MHC-II expression on cell surfaces using FACS. Our purpose was to elute peptides from MHC–peptide complexes present on the cell surface. Low MHC expression resulted in the elution of a few peptides. We found the MFI of MHC-II molecules on J774A.1 cell is about 500; however, the MFI of MHC-II molecules on A20 cells is more than 300,000. These data indicate that MHC-II expression on A20 cells was much higher than that on J774A.1 cells. J774A.1 cell is a macrophage cell line. Macrophages have excellent antigen phagocytic capabilities; however, their ability to present antigens is relatively weak. MHC molecules on the macrophage cell surface can be upregulated in the stimulation of some cytokines, for example, IFN-γ. In this study, we used adjuvants as stimulators and did not want to use additional cytokine stimulators. Thus, J774A.1 cells were not used in the present study.

The related statements are reflected on page 6 lines 120–128 “We also selected another H-2d cell J774A.1, a macrophage cell line, for immunopeptidome analysis in this study. Briefly, 5×108 J774A.1 cells were used for immunopeptidomics. Moreover, fewer than 350 peptides were observed at a peptide spectrum match (PSM) level of < 1.0% false discovery rate (FDR). However, more than 5500 peptides were detected in 108 A20 cells at FDR < 1.0% (Figure S2A). CD86 and MHC-II molecule expression on J774A.1 cells was substantially lower than that on A20 cells (Figure S2B). Low MHC-II expression on J774A.1 cells could be the reason for the lack of peptides identified by LC–MS/MS. Thus, A20 cells instead of J774A.1 cells were used for the subsequent experiments.”

Recommendation 3. Lines 172-177, can more details be provided about the whole proteome analysis? The plots are shown for relative representation of protein expression to PBS, but it is unclear to me what examples of these proteins are (IFN pathway, Ubiquitination pathway). Could these be confirmed by protein expression analyses in supplemental?

Thank you for the suggestion. In this study, we conducted whole proteome analysis to investigate changes in protein expression across different pathways in the adjuvant groups. Through KEGG enrichment analysis, we compared the differential expression of MHC presentation pathway proteins (such as H2-M, Ifi30, CD74, CTSS, proteasome, and peptidase subunits) between the PBS- and adjuvant-treated groups using our proteome data. In addition, we focused on IFN and ubiquitination pathways that play crucial roles in antigen presentation modification and immune response. The proteins and their relative expression in these pathways are shown in Figure S4B. Details regarding the protein names and expressions are provided in Supplemental Table S2 of the revised manuscript.

The original statements in the results “Then, we analyzed the whole proteome data to determine whether the proteins involved in antigen presentation and processing were altered. We found that proteins involved in antigen processing, peptidase function, ubiquitination pathway, and interferon (IFN) signaling were altered post adjuvants treatment, especially in MPLA and CpG groups (Figure 5C; Figure S4B and S4C). These data suggest that adjuvants MPLA and CpG may affect the antigen processing of APCs, resulting in fewer peptides presentation.” This has been revised on page 8 lines 172–182 as “We then investigated whole-proteome data to determine the evidence of adjuvant modification of antigen presentation. We focused on the proteins involved in antigen processing, peptidase function, ubiquitination pathway, and IFN signaling. The ubiquitination pathway and IFN signaling play crucial roles in the modification of antigen presentation and immune responses. Through KEGG enrichment analysis, we found that many proteins involved in antigen processing, peptidase function, ubiquitination pathways, and IFN signaling were altered after adjuvant treatment, particularly in the MPLA- and CpG-treated groups (Figure 5C; Figure S4B). The expression of each protein is shown in Figure S4C and Supplementary Table 2. These data suggest that MPLA and CpG adjuvants may affect the antigen processing of APCs, resulting in fewer peptide presentations.”

Recommendation 4. Lines 212-218: I think there needs to be more discussion of interpretation here. Only one of the low-stability peptides required low concentrations for CD4+ T cell responses in vitro. What about the other peptides in the analysis? Perhaps if the data is taken together there is not a clear pattern?

Thank you for the comment. In this study, epitope-specific CD4+ T-cells were expanded in vitro from the spleens of peptide-pool-immunized mice. T-cell responses to individual peptides were detected using ICS and FACS. Only one peptide, recA #23, with low binding stability, and one high-stability peptide, ureA #2, induced effective T-cell responses. Peptide ureA #3 with high stability induces low Th1 responses. The other peptides cannot induce CD4+ T-cell secreting IFN-γ (Data are shown in Author response image 2). Thus, we compared the strength of IFN-γ responses induced by these three peptides at a set of low concentrations. Data for other peptides without any response could not be taken together.

Author response image 2.

The expanded CD4+T cells from peptides immunized mice were screened for their response to the peptides in an ICS assay.

In this study, we used a peptide pool containing four low-stability peptides to vaccinate mice; however, only one peptide induced an effective CD4+ T-cell response. We speculate that the possible reasons are as follows. First, the number of peptides used for vaccination is too small. Only four low-stability peptides were synthesized and used to immunize mice. Three of these could not induce an effective T-cell response, possibly because of their low immunogenicity. If more peptides are synthesized and used, more peptides that induce T-cell responses may be observed. Second, epitope-specific T-cell responses are variable. Responses to the subdominant peptides can be inhibited by the dominant peptide. The subdominant peptide can become dominant by changing the peptide dose or in the absence of the dominant peptide. Thus, we believe that responses to the other three peptides may be detected if mice are immunized with a peptide pool that does not contain a response epitope.

The corresponding statements have been added to the Discussion section on page 13 lines 287–291 as “Unfortunately, only one peptide, recA #23, with low binding stability and induced significant Th1 responses, was identified in this study. To further confirm that low-stability peptides can induce stronger and higher TCR-affinity antigen-specific T-cell clonotype responses than high-stability peptides, further studies should monitor more peptides with different stabilities.”

Recommendation 5. There are some areas where additional editing to text would be beneficial due to grammar (eg lines 122-126; line 116, etc).

The manuscript has been edited by a professional language editing company.

Reviewer #2 (Recommendations For The Authors):

Recommendation 1. It is interesting that there was no difference in IFNg responses induced by different adjuvants.

Thank you for the comment. Possible reasons for the lack of difference in IFN-γ responses could be as follows. First, all adjuvants used in this study have been confirmed to effectively induce Th1 responses. Second, in this study, IFN-γ responses were examined using expanded antigen-specific T cells in vitro. The in vitro cell expansion efficiency may have affected these results.

Recommendation 2. The data to support the claim that changes in exogenous peptide presentation among adjuvant groups were not due to differences in antigen phagocytosis is insufficient.

Thank you for the comment. In this study, proteomics of A20 cells pulsed with antigens in different adjuvant-treated groups were used to determine exogenous antigens phagocytosed by cells. In addition, we used fluorescein isothiocyanate (FITC)-labeled OVA to pulse APCs and detected antigen phagocytosis by APCs after treatment with different adjuvants. The MFI of FITC was detected by FACS at different time points. The data are shown below (Author response image 3). No obvious differences in FITC MFI were detected after adjuvant stimulation, indicating that antigen phagocytosis among the adjuvant groups was almost the same.

A20 cells, used as APCs, are the B-cell line. Antigen recognition and phagocytosis by B-cells depends on the B-cell receptor (BCR) on the cell surface. The ability of BCRs to bind to different antigens varies, leading to significant differences in the phagocytosis of different antigens by B-cells. Therefore, detecting the phagocytosis of a single antigen may not reflect the overall phagocytic state of the B-cells. Thus, in this study, we used proteomics to detect exogenous proteins in B-cells pulsed with H. pylori antigens, which contain thousands of components, to evaluate their overall phagocytic capacity. Only the proteomic data are presented in our manuscript.

Author response image 3.

Antigen phagocytosis of A20 cells were measured using FITC-labeled OVA. (A) A20 cells were pulsed with FITC-labeled OVA. MFI of FITC was measured after 1 h. (B) MFI of FITC was examined post the stimulation of adjuvants at different time points.

Recommendation 3. It is not clear how MPLA, CpG, and MDP adjuvants modulate the presentation of low vs high stability peptides.

Thank you for pointing this out. We acknowledge that we did not clarify the mechanisms by which adjuvants affect the stability of the peptide presentations of APCs.

We performed experiments to detect the expression of proteins involved in antigen processing and presentation in the different adjuvant-treated groups. Furthermore, shRNAs were used to knock down the expression of key molecules. Immunopeptidomics was used to detect peptide presentation. Unfortunately, the expected results for peptide presentation repertoires were not observed. We are still working on the mechanisms.

Please also see our response to comment 1 of reviewer 1

The related statements were added in the Discussion section on page 13, lines 292–299: “In this study, we found that the peptide repertoires presented by APCs were significantly affected by the adjuvants CpG and MPLA, but not MDP. All three adjuvants belong to the PRR ligand adjuvant family. CpG and MPLA bind to TLRs and MDP is recognized by NOD2. Although the receptors are different, many common molecules are involved both in TLR and NLD pathway activation.  Unfortunately, we did not demonstrate why the MDP had different impacts on peptide presentation compared with other adjuvants. Further investigation is required to clarify the mechanism by which MPLA, CpG, and MDP adjuvants modulate the presentation of peptides with different stabilities.”

Author response:

The following is the authors’ response to the previous reviews.

Reviewer #1 (Public Review): 

The reviewer retained most of their comments from the previous reviewing round. In order to meet these comments and to further examine the dynamic nature of threat omission-related fMRI responses, we now re-analyzed our fMRI results using the single trial estimates. The results of these additional analyses are added below in our response to the recommendations for the authors of reviewer 1. However, we do want to reiterate that there was a factually incorrect statement concerning our design in the reviewer’s initial comments. Specifically, the reviewer wrote that “25% of shocks are omitted, regardless of whether subjects are told that the probability is 100%, 75%, 50%, 25%, or 0%.” We want to repeat that this is not what we did. 100% trials were always reinforced (100% reinforcement rate); 0% trials were never reinforced (0% reinforcement rate). For all other instructed probability levels (25%, 50%, 75%), the stimulation was delivered in 25% of the trials (25% reinforcement rate). We have elaborated on this misconception in our previous letter and have added this information more explicitly in the previous revision of the manuscript (e.g., lines 125-129; 223-224; 486-492).   

Reviewer #1 (Recommendations For The Authors): 

I do not have any further recommendations, although I believe an analysis of learning-related changes is still possible with the trial-wise estimates from unreinforced trials. The authors' response does not clarify whether they tested for interactions with run, and thus the fact that there are main effects does not preclude learning. I kept my original comments regarding limitations, with the exception of the suggestion to modify the title. 

We thank the reviewer for this recommendation. In line with their suggestion, we have now reanalyzed our main ROI results using the trial-by-trial estimates we obtained from the firstlevel omission>baseline contrasts. Specifically, we extracted beta-estimates from each ROI and entered them into the same Probability x Intensity x Run LMM we used for the relief and SCR analyses. Results from these analyses (in the full sample) were similar to our main results. For the VTA/SN model, we found main effects of Probability (F = 3.12, p = .04), and Intensity (F = 7.15, p < .001) (in the model where influential outliers were rescored to 2SD from mean). There was no main effect of Run (F = 0.92, p = .43) and no Probability x Run interaction (F = 1.24, p = .28). If the experienced contingency would have interfered with the instructions, there should have been a Probability x Run interaction (with the effect of Probability only being present in the first runs). Since we did not observe such an interaction, our results indicate that even though some learning might still have taken place, the main effect of Probability remained present throughout the task.  

There is an important side note regarding these analyses: For the first level GLM estimation, we concatenated the functional runs and accounted for baseline differences between runs by adding run-specific intercepts as regressors of no-interest. Hence, any potential main effect of run was likely modeled out at first level. This might explain why, in contrast to the rating and SCR results (see Supplemental Figure 5), we found no main effect of Run. Nevertheless, interaction effects should not be affected by including these run-specific intercepts.

Note that when we ran the single-trial analysis for the ventral putamen ROI, the effect of intensity became significant (F = 3.89, p = .02). Results neither changed for the NAc, nor the vmPFC ROIs.  

Reviewer #2 (Public Review): 

Comments on revised version: 

I want to thank the authors for their thorough and comprehensive work in revising this manuscript. I agree with the authors that learning paradigms might not be a necessity when it comes to study the PE signals, but I don't particularly agree with some of the responses in the rebuttal letter ("Furthermore, conditioning paradigms generally only include one level of aversive outcome: the electrical stimulation is either delivered or omitted."). This is of course correct description for the conditioning paradigm, but the same can be said for an instructed design: the aversive outcome was either delivered or not. That being said, adopting the instructed design itself is legitimate in my opinion. 

We thank the reviewer for this comment. We have now modified the phrasing of this argument to clarify our reasoning (see lines 102-104: “First, these only included one level of aversive outcome: the electrical stimulation was either delivered at a fixed intensity, or omitted; but the intensity of the stimulation was never experimentally manipulated within the same task.”).  

The reason why we mentioned that “the aversive outcome is either delivered or omitted” is because in most contemporary conditioning paradigms only one level of aversive US is used. In these cases, it is therefore not possible to investigate the effect of US Intensity. In our paradigm, we included multiple levels of aversive US, allowing us to assess how the level of aversiveness influences threat omission responding. It is indeed true that each level was delivered or not. However, our data clearly (and robustly across experiments, see Willems & Vervliet, 2021) demonstrate that the effects of the instructed and perceived unpleasantness of the US (as operationalized by the mean reported US unpleasantness during the task) on the reported relief and the omission fMRI responses are stronger than the effect of instructed probability.  

My main concern, which the authors spent quite some length in the rebuttal letter to address, still remains about the validity for different instructed probabilities. Although subjects were told that the trials were independent, the big difference between 75% and 25% would more than likely confuse the subjects, especially given that most of us would fall prey to the Gambler's fallacy (or the law of small numbers) to some degree. When the instruction and subjective experience collides, some form of inference or learning must have occurred, making the otherwise straightforward analysis more complex. Therefore, I believe that a more rigorous/quantitative learning modeling work can dramatically improve the validity of the results. Of course, I also realize how much extra work is needed to append the computational part but without it there is always a theoretical loophole in the current experimental design. 

We agree with the reviewer that some learning may have occurred in our task. However, we believe the most important question in relation to our study is: to what extent did this learning influence our manipulations of interest?  

In our reply to reviewer 1, we already showed that a re-analysis of the fMRI results using the trial-by-trial estimates of the omission contrasts revealed no Probability x Run interaction, suggesting that – overall – the probability effect remained stable over the course of the experiment. However, inspired by the alternative explanation that was proposed by this reviewer, we now also assessed the role of the Gambler’s fallacy in a separate set of analyses. Indeed, it is possible that participants start to expect a stimulation more after more time has passed since the last stimulation was experienced. To test this alternative hypothesis, we specified two new regressors that calculated for each trial of each participant how many trials had passed since the last stimulation (or since the beginning of the experiment) either overall (across all trials of all probability types; hence called the overall-lag regressor) or per probability level (across trials of each probability type separately; hence called the lag-per-probability regressor). For both regressors a value of 0 indicates that the previous trial was either a stimulation trial or the start of experiment, a value of 1 means that the last stimulation trial was 2 trials ago, etc.  

The results of these additional analyses are added in a supplemental note (see supplemental note 6), and referred to in the main text (see lines 231-236: “Likewise, a post-hoc trial-by-trial analysis of the omission-related fMRI activations confirmed that the Probability effect for the VTA/SN activations was stable over the course of the experiment (no Probability x Run interaction) and remained present when accounting for the Gambler’s fallacy (i.e., the possibility that participants start to expect a stimulation more when more time has passed since the last stimulation was experienced) (see supplemental note 6). Overall, these post-hoc analyses further confirm the PE-profile of omission-related VTA/SN responses”.  

Addition to supplemental material (pages 16-18)

Supplemental Note 6: The effect of Run and the Gambler’s Fallacy 

A question that was raised by the reviewers was whether omission-related responses could be influenced by dynamical learning or the Gambler’s Fallacy, which might have affected the effectiveness of the Probability manipulation.  

Inspired by this question, we exploratorily assessed the role of the Gambler’s Fallacy and the effects of Run in a separate set of analyses. Indeed, it is possible that participants start to expect a stimulation more when more time has passed since the last stimulation was experienced. To test this alternative hypothesis, we specified two new regressors that calculated for each trial of each participant how many trials had passed since the last stimulation (or since the beginning of the experiment) either overall (across all trials of all probability types; hence called the overall-lag regressor) or per probability level (across trials of each probability type separately; hence called the lag-per-probability regressor). For both regressors a value of 0 indicates that the previous trial was either a stimulation trial or the start of experiment, a value of 1 means that the last stimulation trial was 2 trials ago, etc.  

The new models including these regressors for each omission response type (i.e., omission-related activations for each ROI, relief, and omission-SCR) were specified as follows:   

(1) For the overall lag:

Omission response ~ Probability * Intensity * Run + US-unpleasantness + Overall-lag + (1|Subject).  

(2) For the lag per probability level:

Omission response ~ Probability * Intensity * Run + US-unpleasantness + Lag-perprobability : Probability + (1|Subject).  

Where US-unpleasantness scores were mean-centered across participants; “*” represents main effects and interactions, and “:” represents an interaction (without main effect). Note that we only included an interaction for the lag-per-probability model to estimate separate lag-parameters for each probability level.  

The results of these analyses are presented in the tables below. Overall, we found that adding these lag-regressors to the model did not alter our main results. That is: for the VTA/SN, relief and omission-SCR, the main effects of Probability and Intensity remained. Interestingly, the overall-lag-effect itself was significant for VTA/SN activations and omission SCR, indicating that VTA/SN activations were larger when more time had passed since the last stimulation (beta = 0.19), whereas SCR were smaller when more time had passed (beta = -0.03). This pattern is reminiscent of the Perruchet effect, namely that the explicit expectancy of a US increases over a run of non-reinforced trials (in line with the gambler’s fallacy effect) whereas the conditioned physiological response to the conditional stimulus declines (in line with an extinction effect, Perruchet, 1985; McAndrew, Jones, McLaren, & McLaren, 2012). Thus, the observed dissociation between the VTA/SN activations and omission SCR might similarly point to two distinctive processes where VTA/SN activations are more dependent on a consciously controlled process that is subjected to the gambler’s fallacy, whereas the strength of the omission SCR responses is more dependent on an automatic associative process that is subjected to extinction. Importantly, however, even though the temporal distance to the last stimulation had these opposing effects on VTA/SN activations and omission SCRs, the main effects of the probability manipulation remained significant for both outcome variables. This means that the core results of our study still hold.   

Next to the overall-lag effect, the lag-per-probability regressor was only significant for the vmPFC. A follow-up of the beta estimates of the lag-per-probability regressors for each probability level revealed that vmPFC activations increased with increasing temporal distance from the stimulation, but only for the 50% trials (beta = 0.47, t = 2.75, p < .01), and not the 25% (beta = 0.25, t = 1.49, p = .14) or the 75% trials (beta = 0.28, t = 1.62, p = .10).

Author response table 1.

F-statistics and corresponding p-values from the overall lag model. (*) F-test and p-values were based on the model where outliers were rescored to 2SD from the mean. Note that when retaining the influential outliers for this model, the p-value of the probability effect was p = .06. For all other outcome variables, rescoring the outliers did not change the results. Significant effects are indicated in bold.

Author response table 2.

F-statistics and corresponding p-values from the lag per probability level model. (*) F-test and p-values were based on the model where outliers were rescored to 2SD from the mean. Note that when retaining the influential outliers for this model, the p-value of the Intensity x Run interaction was p = .05. For all other outcome variables, rescoring the outliers did not change the results. Significant effects are indicated in bold.

As the authors mentioned in the rebuttal letter, "selecting participants only if their anticipatory SCR monotonically increased with each increase in instructed probability 0% < 25% < 50% < 75% < 100%, N = 11 participants", only ~1/3 of the subjects actually showed strong evidence for the validity of the instructions. This further raises the question of whether the instructed design, due to the interference of false instruction and the dynamic learning among trials, is solid enough to test the hypothesis .  

We agree with the reviewer that a monotonic increase in anticipatory SCR with increasing probability instructions would provide the strongest evidence that the manipulation worked. However, it is well known that SCR is a noisy measure, and so the chances to see this monotonic increase are rather small, even if the underlying threat anticipation increases monotonically. Furthermore, between-subject variation is substantial in physiological measures, and it is not uncommon to observe, e.g., differential fear conditioning in one measure, but not in another (Lonsdorf & Merz, 2017). It is therefore not so surprising that ‘only’ 1/3 of our participants showed the perfect pattern of monotonically increasing SCR with increasing probability instructions. That being said, it is also important to note that not all participants were considered for these follow-up analyses because valid SCR data was not always available.

Specifically, N = 4 participants were identified as anticipation non-responders (i.e. participant with smaller average SCR to the clock on 100% than on 0% trials; pre-registered criterium) and were excluded from the SCR-related analyses, and N = 1 participant had missing data due to technical difficulties. This means that only 26 (and not 31) participants were considered for the post hoc analyses. Taking this information into account, this means that 21 out of 26 participants (approximately 80%) showed stronger anticipatory SCR following 75% instructions compared to 25% instructions and that  11 out of 26 participants (approximately 40%) even showed the monotonical increase in their anticipatory SCR (see supplemental figure 4). Furthermore, although anticipatory SCR gradually decreased over the course of the experiment, there was no Run x Probability interaction, indicating that the instructions remained stable throughout the task (see supplemental figure 3).  

Reviewer #2 (Recommendations For The Authors):

A more operational approach might be to break the trials into different sections along the timeline and examine how much the results might have been affected across time. I expect the manipulation checks would hold for the first one or two runs and the authors then would have good reasons to focus on the behavioral and imaging results for those runs. 

This recommendation resembles the recommendation by reviewer 1. In our reply to reviewer 1, we showed the results of a re-analysis of the fMRI data using the trial-by-trial estimates of the omission contrasts, which revealed no Probability x Run interaction, suggesting that – overall - the probability effect remained (more or less) stable over the course of the experiment.  For a more in depth discussion of the results of this additional analysis, we refer to our answer to reviewer 1.  

Reviewer #3 (Public Review): 

Comments on revised version: 

The authors were extremely responsive to the comments and provided a comprehensive rebuttal letter with a lot of detail to address the comments. The authors clarified their methodology, and rationale for their task design, which required some more explanation (at least for me) to understand. Some of the design elements were not clear to me in the original paper. 

The initial framing for their study is still in the domain of learning. The paper starts off with a description of extinction as the prime example of when threat is omitted. This could lead a reader to think the paper would speak to the role of prediction errors in extinction learning processes. But this is not their goal, as they emphasize repeatedly in their rebuttal letter. The revision also now details how using a conditioning/extinction framework doesn't suit their experimental needs. 

We thank the reviewer for pointing out this potential cause of confusion. We have now rewritten the starting paragraph of the introduction to more closely focus on prediction errors, and only discuss fear extinction as a potential paradigm that has been used to study the role of threat omission PE for fear extinction learning (see lines 40-55). We hope that these adaptations are sufficient to prevent any false expectations. However, as we have mentioned in our previous response letter, not talking about fear extinction at all would also not make sense in our opinion, since most of the knowledge we have gained about threat omission prediction errors to date is based on studies that employed these paradigms.  

Adaptation in the revised manuscript (lines 40-55):  

“We experience pleasurable relief when an expected threat stays away1. This relief indicates that the outcome we experienced (“nothing”) was better than we expected it to be (“threat”). Such a mismatch between expectation and outcome is generally regarded as the trigger for new learning, and is typically formalized as the prediction error (PE) that determines how much there can be learned in any given situation2. Over the last two decades, the PE elicited by the absence of expected threat (threat omission PE) has received increasing scientific interest, because it is thought to play a central role in learning of safety. Impaired safety learning is one of the core features of clinical anxiety4. A better understanding of how the threat omission PE is processed in the brain may therefore be key to optimizing therapeutic efforts to boost safety learning. Yet, despite its theoretical and clinical importance, research on how the threat omission PE is computed in the brain is only emerging.  

To date, the threat omission PE has mainly been studied using fear extinction paradigms that mimic safety learning by repeatedly confronting a human or animal with a threat predicting cue (conditional stimulus, CS; e.g. a tone) in the absence of a previously associated aversive event (unconditional stimulus, US; e.g., an electrical stimulation). These (primarily non-human) studies have revealed that there are striking similarities between the PE elicited by unexpected threat omission and the PE elicited by unexpected reward.”

It is reasonable to develop a new task to answer their experimental questions. By no means is there a requirement to use a conditioning/extinction paradigm to address their questions. As they say, "it is not necessary to adopt a learning paradigm to study omission responses", which I agree with.  But the authors seem to want to have it both ways: they frame their paper around how important prediction errors are to extinction processes, but then go out of their way to say how they can't test their hypotheses with a learning paradigm.

Part of their argument that they needed to develop their own task "outside of a learning context" goes as follows: 

(1) "...conditioning paradigms generally only include one level of aversive outcome: the electrical stimulation is either delivered or omitted. As a result, the magnitude-related axiom cannot be tested." 

(2) "....in conditioning tasks people generally learn fast, rendering relatively few trials on which the prediction is violated. As a result, there is generally little intra-individual variability in the PE responses" 

(3) "...because of the relatively low signal to noise ratio in fMRI measures, fear extinction studies often pool across trials to compare omission-related activity between early and late extinction, which further reduces the necessary variability to properly evaluate the probability axiom" 

These points seem to hinge on how tasks are "generally" constructed. However, there are many adaptations to learning tasks:

(1) There is no rule that conditioning can't include different levels of aversive outcomes following different cues. In fact, their own design uses multiple cues that signal different intensities and probabilities. Saying that conditioning "generally only include one level of aversive outcome" is not an explanation for why "these paradigms are not tailored" for their research purposes. There are also several conditioning studies that have used different cues to signal different outcome probabilities. This is not uncommon, and in fact is what they use in their study, only with an instruction rather than through learning through experience, per se.

(2) Conditioning/extinction doesn't have to occur fast. Just because people "generally learn fast" doesn't mean this has to be the case. Experiments can be designed to make learning more challenging or take longer (e.g., partial reinforcement). And there can be intra-individual differences in conditioning and extinction, especially if some cues have a lower probability of predicting the US than others. Again, because most conditioning tasks are usually constructed in a fairly simplistic manner doesn't negate the utility of learning paradigms to address PEaxioms.

(3) Many studies have tracked trial-by-trial BOLD signal in learning studies (e.g., using parametric modulation). Again, just because other studies "often pool across trials" is not an explanation for these paradigms being ill-suited to study prediction errors. Indeed, most computational models used in fMRI are predicated on analyzing data at the trial level. 

We thank the reviewer for these remarks. The “fear conditioning and extinction paradigms” that we were referring to in this paragraph were the ones that have been used to study threat omission PE responses in previous research (e.g., Raczka et al., 2011; Thiele et al. 2021; Lange et al. 2020; Esser et al., 2021; Papalini et al., 2021; Vervliet et al. 2017). These studies have mainly used differential/multiple-cue protocols where either one (or two) CS+  and one CS- are trained in an acquisition phase and extinguished in the next phase. Thus, in these paradigms: (1) only one level of aversive US is used; and (2) as safety learning develops over the course of extinction, there are relatively few omission trials during which “large” threat omission PEs can be observed (e.g. from the 24 CS+ trials that were used during extinction in Esser et al., the steepest decreases in expectancy – and thus the largest PE – were found in first 6 trials); and (3) there was never absolute certainty that the stimulation will no longer follow. Some of these studies have indeed estimated the threat omission PE during the extinction phase based on learning models, and have entered these estimates as parametric modulators to CS-offset regressors. This is very informative. However, the exact model that was used differed per study (e.g. Rescorla-Wagner in Raczka et al. and Thiele et al.; or a Rescorla- Wagner–Pearce- Hall hybrid model in Esser et al.). We wanted to analyze threat omission-responses without commitment to a particular learning model. Thus, in order to examine how threat omissionresponses vary as a function of probability-related expectations, a paradigm that has multiple probability levels is recommended (e.g. Rutledge et al., 2010; Ojala et al., 2022)

The reviewer rightfully pointed out that conditioning paradigms (more generally) can be tailored to fit our purposes as well. Still, when doing so, the same adaptations as we outlined above need to be considered: i.e. include different levels of US intensity; different levels of probability; and conditions with full certainty about the US (non)occurrence. In our attempt to keep the experimental design as simple and straightforward as possible, we decided to rely on instructions for this purpose, rather than to train 3 (US levels) x 5 (reinforcement levels) = 15 different CSs. It is certainly possible to train multiple CSs of varying reinforcement rates (e.g. Grings et al. 1971, Ojala et al., 2022). However, given that US-expectation on each trial would primarily depend on the individual learning processes of the participants, using a conditioning task would make it more difficult to maintain experimental control over the level of USexpectation elicited by each CS. As a result, this would likely require more extensive training, and thus prolong the study procedure considerably. Furthermore, even though previous studies have trained different CSs for different reinforcement rates, most of these studies have only used one level of US. Thus, in order to not complexify our task to much, we decided to rely on instructions rather than to train CSs for multiple US levels (in addition to multiple reinforcement rates).

We have tried to clarify our reasoning in the revised version of the manuscript (see introduction, lines 100-113):  

“The previously discussed fear conditioning and extinction studies have been invaluable for clarifying the role of the threat omission PE within a learning context. However, these studies were not tailored to create the varying intensity and probability-related conditions that are required to systematically evaluate the threat omission PE in the light of the PE axioms. First, these only included one level of aversive outcome: the electrical stimulation was either delivered or omitted; but the intensity of the stimulation was never experimentally manipulated within the same task. As a result, the magnitude-related axiom could not be tested. Second, as safety learning progressively developed over the course of extinction learning, the most informative trials to evaluate the probability axiom (i.e. the trials with the largest PE) were restricted to the first few CS+ offsets of the extinction phase, and the exact number of these informative trials likely differed across participants as a result of individually varying learning rates. This limited the experimental control and necessary variability to systematically evaluate the probability axiom. Third, because CS-US contingencies changed over the course of the task (e.g. from acquisition to extinction), there was never complete certainty about whether the US would (not) follow. This precluded a direct comparison of fully predicted outcomes. Finally, within a learning context, it remains unclear whether brain responses to the threat omission are in fact responses to the violation of expectancy itself, or whether they are the result of subsequent expectancy updating.”

Again, the authors are free to develop their own task design that they think is best suited to address their experimental questions. For instance, if they truly believe that omission-related responses should be studied independent of updating. The question I'm still left puzzling is why the paper is so strongly framed around extinction (the word appears several times in the main body of the paper), which is a learning process, and yet the authors go out of their way to say that they can only test their hypotheses outside of a learning paradigm. 

As we have mentioned before, the reason why we refer to extinction studies is because most evidence on threat omission PE to date comes from fear extinction paradigms.  

The authors did address other areas of concern, to varying extents. Some of these issues were somewhat glossed over in the rebuttal letter by noting them as limitations. For example, the issue with comparing 100% stimulation to 0% stimulation, when the shock contaminates the fMRI signal. This was noted as a limitation that should be addressed in future studies, bypassing the critical point. 

It is unclear to us what the reviewer means with “bypassing the critical point”. We argued in the manuscript that the contrast we initially specified and preregistered to study axiom 3 (fully predicted outcomes elicit equivalent activation) could not be used for this purpose, as it was confounded by the delivery of the stimulation. Because 100% trials aways included the stimulation and 0% trials never included stimulation, there was no way to disentangle activations related to full predictability from activations related to the stimulation as such.   

Reviewer #3 (Recommendations For The Authors): 

I'm not sure the new paragraph explaining why they can't use a learning task to test their hypotheses is very convincing, as I noted in my review. Again, it is not a problem to develop a new task to address their questions. They can justify why they want to use their task without describing (incorrectly in my opinion) that other tasks "generally" are constructed in a way that doesn't suit their needs. 

For an overview of the changes we made in response to this recommendation, we refer to our reply to the public review.   

We look forward to your reply and are happy to provide answers to any further questions or comments you may have.

Author response:

The following is the authors’ response to the original reviews.

Public Reviews:

Reviewer #1 (Public Review):

The study describes a new computational method for unsupervised (i.e., non-artificial intelligence) segmentation of objects in grayscale images that contain substantial noise, to differentiate object, no object, and noise. Such a problem is essential in biology because they are commonly confronted in the analysis of microscope images of biological samples and recently have been resolved by artificial intelligence, especially by deep neural networks. However, training artificial intelligence for specific sample images is a difficult task and not every biological laboratory can handle it. Therefore, the proposed method is particularly appealing to laboratories with little computational background. The method was shown to achieve better performance than a threshold-based method for artificial and natural test images. To demonstrate the usability, the authors applied the method to high-power confocal images of the thalamus for the identification and quantification of immunostained potassium ion channel clusters formed in the proximity of large axons in the thalamic neuropil and verified the results in comparison to electron micrographs.

Strengths:

The authors claim that the proposed method has higher pixel-wise accuracy than the threshold-based method when applied to gray-scale images with substantial noises.

Since the method does not use artificial intelligence, training and testing are not necessary, which would be appealing to biologists who are not familiar with machine learning technology.

The method does not require extensive tuning of adjustable parameters (trying different values of "Moran's order") given that the size of the object in question can be estimated in advance.

We appreciate the positive assessment of our approach.

Weaknesses:

It is understood that the strength of the method is that it does not depend on artificial intelligence and therefore the authors wanted to compare the performance with another non-AI method (i.e. the threshold-based method; TBM). However, the TBM used in this work seems too naive to be fairly compared to the expensive computation of "Moran's I" used for the proposed method. To provide convincing evidence that the proposed method advances object segmentation technology and can be used practically in various fields, it should be compared to other advanced methods, including AI-based ones, as well.

Protein localization studies revealed that protein distributions are frequently inhomogeneous in a cell. This is very common in neurons which are highly polarized cell types with distinct axo-somato-dendritic functions. Moreover, due to the nature of the cell-to-cell interactions among neurons (e.g. electrical and chemical synapses) the cell membrane includes highly variable microdomains with unique protein assemblies (i.e. clusters). Protein clusters are defined as membrane segments with higher protein densities compared to neighboring membrane regions. However, protein density can continuously change between “clusters” and “non-clusters”. As a consequence, differentiating proteins involved vs not involved in clusters is a challenging task.  Indeed, our analysis showed that the boundaries of protein clusters varied remarkably when 23 human experts delineated them.

Despite the fact the protein clusters can only be vaguely defined numerous studies have demonstrated the functional relevance of inhomogeneous protein distribution. Thus, there is a high relevance and need for an observer independent, “operative” segmentation method that can be accomplished and compared among different conditions and specimens. The strength of the Moran’s I analysis we propose here, as pointed out by our reviewers and editors, is that it can extract the relevant signals from an image generated in different, often noisy condition using a simple algorithm that allows quantitative characterization and identification of changes in many biological and non-biological samples.

In AI based analysis the ground truth is known by an observer and using a large training set AI learns to extract the relevant information for image segmentation. As outlined above the “ground truth”, however, cannot be unequivocally defined for protein clusters. There is no doubt, that with sufficient resource investment there would be an AI based analysis of the same problem. In our view, however, in an average laboratory setting generating a training set using hundreds of images examined by many experts may not be plausible. Moreover, generalization of one training set to another set of cluster, resistance to noise or different levels of background could also not be guaranteed.

This method was claimed to be better than the TBM when the noise level was high. Related to the above, TBMs can be used in association with various denoising methods as a preprocess. It is questionable whether the claim is still valid when compared to the methods with adequate complexity used together with denoising. Consider for example, Weigert et al. (2018) https://doi.org/10.1038/s41592-018-0216-7; or Lehtinen et al (2018) https://doi.org/10.48550/arXiv.1803.04189.

In Weigert et al. AI was trained with high-quality images of the same object obtained with extreme photon exposure in confocal microscope. As delineated above without training AI systems cannot be used for such purposes. The Lehtinen paper is unfortunately no longer available at this doi.

We must emphasize that in our work we did not intend to compare the image segmentation method based on local Moran’s I with all other available segmentation techniques. Rather we wanted to demonstrate a straightforward method of grouping pixels with similar intensities and in spatial proximity which does not require a priori knowledge of the objects. We used TBM to benchmark the method. We agree that with more advanced TBM methods the difference between Moran’s and TBM might have been smaller. The critical component here is, however, that even with most advanced TBM an artificial threshold is needed to be defined. The optimal threshold may change from sample to sample depending on the experimental conditions which makes quantification questionable. Moran’s method overcomes this problem and allows more objective segmentation of images even if the exact conditions (background labeling, noise, intensity etc) are not identical among the samples.

The computational complexity of the method, determined by the convolution matrix size (Moran's order), linearly increases as the object size increases (Fig. S2b). Given that the convolution must be run separately for each pixel, the computation seems quite demanding for scale-up, e.g. when the method is applied for 3D image volumes. It will be helpful if the requirement for computer resources and time is provided.

Here we provide the required data concerning the hardware and the computational time:

Hardware used for performing the analysis:

Intel(R) Xeon(R) Silver 4112 CPU @ 2.60GHz, 2594 Mhz, 4 kernel CPU, 64GB RAM, NVIDIA GeForce GTX 1080 graphic card.

MATLAB R2021b software was used for implementation.

Author response table 1.

Computation times:

Reviewer #2 (Public Review):

Summary:

The manuscript by David et al. describes a novel image segmentation method, implementing Local Moran's method, which determines whether the value of a datapoint or a pixel is randomly distributed among all values, in differentiating pixel clusters from the background noise. The study includes several proof-of-concept analyses to validate the power of the new approach, revealing that implementation of Local Moran's method in image segmentation is superior to threshold-based segmentation methods commonly used in analyzing confocal images in neuroanatomical studies.

Strengths:

Several proof-of-concept experiments are performed to confirm the sensitivity and validity of the proposed method. Using composed images with varying levels of background noise and analyzing them in parallel with the Local Moran's or a Threshold-Based Method (TBM), the study is able to compare these approaches directly and reveal their relative power in isolating clustered pixels.     

Similarly, dual immuno-electron microscopy was used to test the biological relevance of a colocalization that was revealed by Local Moran's segmentation approach on dual-fluorescent labeled tissue using immuno-markers of the axon terminal and a membrane-protein (Figure 5). The EM revealed that the two markers were present in terminals and their post-synaptic partners, respectively. This is a strong approach to verify the validity of the new approach for determining object-based colocalization in fluorescent microscopy. 

The methods section is clear in explaining the rationale and the steps of the new method (however, see the weaknesses section). Figures are appropriate and effective in illustrating the methods and the results of the study. The writing is clear; the references are appropriate and useful.

We are grateful for the constructive assessment of our results.

Weaknesses:

While the steps of the mathematical calculations to implement Local Moran's principles for analyzing high-resolution images are clearly written, the manuscript currently does not provide a computation tool that could facilitate easy implementation of the method by other researchers. Without a user-friendly tool, such as an ImageJ plugin or a code, the use of the method developed by David et al by other investigators may remain limited.

The code for the analysis is now available online as a user-friendly MATLAB script at: https://github.com/dcsabaCD225/Moran_Matlab/blob/main/moran_local.m

Recommendations for the authors:

Summary of reviews:

Both reviewers acknowledge the potential significance and practicality of the newly proposed image segmentation method. This method uses Local Moran's principles, offering an advantage over traditional intensity thresholding approaches by providing more sensitivity, particularly in reducing background noise and preserving biologically relevant pixels.

Strengths Highlighted:

• The proposed method can provide more accurate results, especially for grayscale images with significant noise.

• The method is not dependent on artificial intelligence, making it appealing for researchers with minimal computational background.    

• The approach can operate without the need for extensive tuning, given that the size of the object is known.

• Several proof-of-concept experiments were carried out, revealing the effectiveness of the method in comparison with the threshold-based segmentation methods.

• The manuscript is clear in terms of methodology, and the results are supported by effective illustrations and references.

Weaknesses Noted:

• The study lacked a comparative analysis with advanced segmentation methods, especially those that employ artificial intelligence.

See our response above to the same question of Reviewer 1.

• There are concerns about computational complexity, especially when dealing with larger data sets or 3D image volumes.

See our response about the calculations of computation times above to the similar question of Reviewer 1.

• Both reviewers noted the absence of a data/code availability statement in the manuscript, which might restrict the method's adoption by other researchers.

The code availability is provided now.

• Reviewer 2 suggested that some results, particularly related to Kv4.2 in the thalamus, might be better presented in a separate study due to their significance.

We thank our reviewers for this suggestion. We carefully evaluated the pros and cons of publishing the Kv4.2 data separately. We finally decided to keep the segmentation and experimental data together due to the following reason. We believe that the ultrastructural localization provides strong experimental proof for the relevance of our novel segmentation method. In order to make the potassium channel data more visible we added a subsentence to the title. In this manner we think scientist interested in the imaging method as well as the neurobiology will be both find and cite the paper. The novel title reads now:

“An image segmentation method based on the spatial correlation coefficient of Local Moran’s I - identification of A-type potassium channel clusters in the thalamus.”

Reviewer Recommendations:

(1) Provide details about the data and program code availability.

See our response above

(2) Offer practical recommendations and provide clarity on software packages and coding for the proposed method to enhance its adoption.

Done.

(3) Consider presenting the findings about Kv4.2 in the thalamus separately as they hold significant importance on their own.

See our response above

Given the reviews, the proposed image segmentation method presents a promising advancement in the domain of image analysis. The technique offers tangible benefits, especially for researchers dealing with biological microscopy data. However, for this method to see a broader application, it's imperative to provide clearer practical guidance and make data or code easily accessible. Additionally, while the findings regarding Kv4.2 in the thalamus are intriguing, they might achieve more impact if detailed in a dedicated paper.

Reviewer #1 (Recommendations For The Authors):

The availability of data or program code was not stated in the manuscript.

Reviewer #2 (Recommendations For The Authors):

(1) While the principles of the method are explained clearly in a step-by-step fashion in the Methods section, the practical aspects of running sequential computations over a large matrix of pixel values are not well described. It would be very useful if the authors could provide recommendations on how to set the data structure and clarify which software and programming package for Local Moran's analysis they used. In addition, providing the code for the sequential implementation described in the Methods section would facilitate the adoption of the method by other researchers, and thus, the impact of the study. Currently, there is no data or code availability statement included in the manuscript.

See our response above.

(2) Figure 4 illustrates an experiment in which transmission electron microscopy and freeze-fracture replica labeling approaches were used to demonstrate that a potassium channel marker, Kv4.2 was selective to synapses forming on larger caliber dendrites in the thalamus. As impressive as the EM approaches utilized in this figure are, the results of this experiment have a somewhat tangential bearing on the segmentation method that is the focus of this study. In fact, the experiments illustrated in Figure 5, dual immuno-EM, are more than sufficient to confirm what the dual-confocal imaging coupled with Local Moran's segmentation analysis reveals. Furthermore, the author's findings about the localization and selectivity of Kv4.2 in the thalamus are too important and exciting to bury in a paper focusing on the methodology. Those results may have a wider impact if they are presented and discussed in a separate experimental paper.

See our response above

Author response:

The following is the authors’ response to the original reviews.

Reviewer #3 (Public Review):

The iron manipulation experiments are in the whole animal and it is likely that this affects general feeding behaviour, which is known to affect NB exit from quiescence and proliferative capacity. The loss of ferritin in the gut and iron chelators enhancing the NB phenotype are used as evidence that glia provide iron to NB to support their number and proliferation. Since the loss of NB is a phenotype that could result from many possible underlying causes (including low nutrition), this specific conclusion is one of many possibilities.

We have investigated the feeding behavior of fly by Brilliant Blue (sigma, 861146)[1]. Our result showed that the amount of dye in the fly body were similar between control group and BPS group, suggesting that BPS almost did not affect the feeding behavior (Figure 3—figure supplement 1A).

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

There was a gap between the Pros nuclear localization and downstream targets of ferritin, particularly NADH dehydrogenase and biosynthesis. Could overexpression of Ndi1 restore Pros localization in NBs?

Ferritin defect downregulates iron level, which leads to cell cycle arrest of NBs via ATP shortage. And cell cycle arrest of NBs probably results in NB differentiation[2, 3]. We have added the experiment in Figure 5—figure supplement 2. This result showed that overexpression of Ndi1 could significantly restore Pros localization in NBs.

The abstract requires revision to cover the major findings of the manuscript, particularly the second half.

We revised the abstract to add more major findings of the manuscript in the second half as follows:

“Abstract

Stem cell niche is critical for regulating the behavior of stem cells. Drosophila neural stem cells (Neuroblasts, NBs) are encased by glial niche cells closely, but it still remains unclear whether glial niche cells can regulate the self-renewal and differentiation of NBs. Here we show that ferritin produced by glia, cooperates with Zip13 to transport iron into NBs for the energy production, which is essential to the self-renewal and proliferation of NBs. The knockdown of glial ferritin encoding genes causes energy shortage in NBs via downregulating aconitase activity and NAD+ level, which leads to the low proliferation and premature differentiation of NBs mediated by Prospero entering nuclei. More importantly, ferritin is a potential target for tumor suppression. In addition, the level of glial ferritin production is affected by the status of NBs, establishing a bicellular iron homeostasis. In this study, we demonstrate that glial cells are indispensable to maintain the self-renewal of NBs, unveiling a novel role of the NB glial niche during brain development.”

In Figure 2B Mira appeared to be nuclear in NBs, which is inconsistent with its normal localization. Was it Dpn by mistake?

In Figure 2B, we confirmed that it is Mira. Moreover, we also provide a magnified picture in Figure 2B’, showing that the Mira mainly localizes to the cortex or in the cytoplasm as previously reported.

Figure 2C, Fer1HCH-GFP/mCherry localization was non-uniform in the NBs revealing 1-2 regions devoid of protein localization potentially corresponding to the nucleus and Mira crescent enrichment. It is important to co-label the nucleus in these cells and discuss the intracellular localization pattern of Ferritin.

We have revised the picture with nuclear marker DAPI in Figure 2C. The result showed that Fer1HCH-GFP/Fer2LCH-mCherry was not co-localized with DAPI, which indicated that Drosophila ferritin predominantly distributes in the cytosol[4, 5]. As for the concern mentioned by this reviewer, GFP/mCherry signal in NBs was from glial overexpressed ferritin, which probably resulted in non-uniform signal.

In Figure 3-figure supplement 3F, glial cells in Fer1HCH RNAi appeared to be smaller in size. This should be quantified. Given the significance of ferritin in cortex glial cells, examining the morphology of cortex glial cells is essential.

In Figure 3—figure supplement 3F, we did not label single glial cells so it was difficult to determine whether the size was changed. However, it seems that the chamber formed by the cellular processes of glial cells becomes smaller in Fer1HCH RNAi. The glial chamber will undergo remodeling during neurogenesis, which responses to NB signal to enclose the NB and its progeny[6]. Thus, the size of glial chamber is regulated by NB lineage size. In our study, ferritin defect leads to the low proliferation, inducing the smaller lineage of each NB, which likely makes the chamber smaller.

Since the authors showed that the reduced NB number was not due to apoptosis, a time-course experiment for glial ferritin KD is recommended to identify the earliest stage when the phenotype in NB number /proliferation manifests during larval brain development.

We observed brains at different larval stages upon glial ferritin KD. The result showed that NB proliferation decreased significantly, but NB number declined slightly at the second-instar larval stage (Figure 5—figure supplement 1E and F), suggesting that brain defect of glial ferritin KD manifests at the second-instar larval stage.

Transcriptome analysis on ferritin glial KD identified genes in mitochondrial functions, while the in vivo EM data suggested no defects in mitochondria morphology. A short discussion on the inconsistency is required.

For the observation of mitochondria morphology via the in vivo EM data, we focused on visible cristae in mitochondria, which was used to determine whether the ferroptosis happens[7]. It is possible that other details of mitochondria morphology were changed, but we did not focus on that. To describe this result more accurately, we replaced “However, our observation revealed no discernible defects in the mitochondria of NBs after glial ferritin knockdown” with the “However, our result showed that the mitochondrial double membrane and cristae were clearly visible whether in the control group or glial ferritin knockdown group, which suggested that ferroptosis was not the main cause of NB loss upon glial ferritin knockdown” in line 207-209.

The statement “we found no obvious defects of brain at the first-instar larval stage (0-4 hours after larval hatching) when knocking down glial ferritin (Figure 5-figure supplement 1C).” lacks quantification of NB number and proliferation, making it challenging to conclude.

We have provided the quantification of NB number and proliferation rate of the first-instar larval stage in Figure 5—figure supplement 1C and D. The data showed that there is no significant change in NB number and proliferation rate when knocking down ferritin, suggesting that no brain defect manifests at the first-instar larval stage.

A wild-type control is necessary for Figure 6A-C as a reference for normal brain sizes.

We have added Insc>mCherry RNAi as a reference in Figure 6A-D, which showed that the brain size of tumor model is larger than normal brain. Moreover, we removed brat RNAi data from Figure 6A-D to Figure 6—figure supplement 1A-D for the better layout.

In Figures 6B, D, “Tumor size” should be corrected to “Larval brain volume”.

Here, we measured the brain area to assess the severity of the tumor via ImageJ instead of 3D data of the brain volume. So we think it would be more appropriate to use the “Larval brain size” than “Larval brain volume” here. Thus, we have corrected “Tumor size” to “Larval brain size” in Figure 6B and D to Figure 6—figure supplement 1B and D.

Considering that asymmetric division defects in NBs may lead to premature differentiation, it is advisable to explore the potential involvement of ferritin in asymmetric division.

aPKC is a classic marker to determine the asymmetric division defect of NB. We performed the aPKC staining and found it displayed a crescent at the apical cortex based on the daughter cell position whether in control or glial ferritin knockdown (Figure 5—figure supplement 3A). This result indicated that there was no obvious asymmetric defect after glial ferritin knockdown.

In the statement "Secondly, we examined the apoptosis in glial cells via Caspase-3 or TUNEL staining, and found the apoptotic signal remained unchanged after glial ferritin knockdown (Figure 3-figure supplement 3A-D).", replace "the apoptosis in glial cells" with "the apoptosis in larval brain cells".

We have replaced "the apoptosis in glial cells" with "the apoptosis in larval brain cells" in line 216.

Include a discussion on the involvement of ferritin in mammalian brain development and address the limitations associated with considering ferritin as a potential target for tumor suppression.

We have added the discussion about ferritin in mammalian brain development in line 428-430 and limitation of ferritin for suppressing tumor in line 441-444.

Indicate Insc-GAL4 as BDSC#8751, even if obtained from another source. Additionally, provide information on the extensively used DeRed fly stock used in this study within the methods section.

We provided the stock information of Insc-GAL4 and DsRed in line 673-674.

Reviewer #2 (Recommendations For The Authors):

Major points:

The number of NBs differs a lot between experiments. For example, in Fig 1B and 1K controls present less than 100 NBs whereas in Figure 1 Supplementary 2B it can be seen that controls have more than 150. Then, depending on which control you compare the number of NBs in flies silencing Fer1HCH or Fer2LCH, the results might change. The authors should explain this.

Figure 1 Supplementary 2B (Figure 1 Supplementary 3B in the revised version) shows NB number in VNC region while Fig 1B and 1K show NB number in CB region. At first, we described the general phenotype showing the NB number in CB and VNC respectively (Fig 1 and Fig 1-Supplementary 1 and 3 in the revised version). And the NB number is consistent in each region. After then, we focused on NB number in CB for the convenience.

This reviewer encourages the authors to use better Gal4 lines to describe the expression patterns of ferritins and Zip13 in the developing brain. On the one hand, the authors do not state which lines they are using (including supplementary table). On the other hand, new Trojan GAL4 (or at least InSite GAL4) lines are a much better tool than classic enhancer trap lines. The authors should perform this experiment.

All stock source and number were documented in Table 2. Ferritin GAL4 and Zip13 GAL4 in this study are InSite GAL4. In addition, we also used another Fer2LCH enhancer trapped GAL4 to verify our result (DGRC104255) and provided the result in Figure 2—figure supplement 1. Our data showed that DsRed driven by Fer2LCH-GAL4 was co-localized with the glia nuclear protein Repo, instead of the NB nuclear protein Dpn, which was consistent with the result of Fer1HCH/Fer2LCH GAL4. In addition, we will try to obtain the Trojan GAL4 (Fer1HCH/Fer2LCH GAL4 and Zip13 GAL4) and validate this result in the future.

The authors exclude very rapidly the possibility of ferroptosis based only on some mitochondrial morphological features without analysing the other hallmarks of this iron-driven cell death. The authors should at least measure Lipid Peroxidation levels in their experimental scenario either by a kit to quantify by-products of lipid peroxidation such as Malonaldehide (MDA) or using an anti 4-HNE antibody.

We combined multiple experiments to exclude the possibility of ferroptosis. Firstly, ferroptosis can be terminated by iron chelator. And we fed fly with iron chelator upon glial ferritin knockdown, but NB number and proliferation were not restored, which suggested that ferroptosis probably was not the cause of NB loss induced by glial ferritin knockdown (Figure 3B and C). Secondly, Zip13 transports iron into the secretary pathway and further out of the cells in Drosophila gut[8]. Our data showed that knocking down iron transporter Zip13 in glia resulted in the decline of NB number and proliferation, which was consistent with the phenotype upon glial ferritin knockdown (Figure 3E-G). More importantly, the knockdown of Zip13 and ferritin simultaneously aggravated the phenotype in NB number and proliferation (Figure 3E-G). These results suggested that the phenotype was induced by iron deficiency in NB, which excluded the possibility of iron overload or ferroptosis to be the main cause of NB loss upon glial ferritin knockdown. Finally, we observed mitochondrial morphology on double membrane and the cristae that are critical hallmarks of ferroptosis, but found no significant damage (Figure 3-figure supplement 2E and F).

In addition, we have added the 4-HNE determination in Figure 3—figure supplement 2G and H. This result showed that 4-HNE level did not change significantly, suggesting that lipid peroxidation was stable, which supported to exclude the possibility that the ferroptosis led to the NB loss upon glial ferritin knockdown.

All of the above results together indicate that ferroptosis is not the cause of NB loss after ferritin knockdown.

A major flaw of the manuscript is related to the chapter Glial ferritin defects result in impaired Fe-S cluster activity and ATP production and the results displayed in Figure 4. The authors talk about the importance of FeS clusters for energy production in the mitochondria. Surprisingly, the authors do not analyse the genes involved in this process such as but they present the interaction with the cytosolic FeS machinery that has a role in some extramitochondrial proteins but no role in the synthesis of FeS clusters incorporated in the enzymes of the TCA cycle and the respiratory chain. The authors should repeat the experiments incorporating the genes NSF1 (CG12264), ISCU(CG9836), ISD11 (CG3717), and fh (CG8971) or remove (or at least rewrite) this entire section.

Thanks for this constructive advice and we have revised this in Figure 4B and C. We repeated the experiment with blocking mitochondrial Fe-S cluster biosynthesis by knocking down Nfs1 (CG12264), ISCU(CG9836), ISD11 (CG3717), and fh (CG8971), respectively. Nfs1 knockdown in NB led to a low proliferation, which was consistent with CIA knockdown. However, we did not observe the obvious brain defect in ISCU(CG9836), ISD11 (CG3717), and fh (CG8971) knockdown in NB. Our interpretation of these results is that Nfs1 probably is a necessary core component in Fe-S cluster assembly while others are dispensable[9].

The presence and aim of the mouse model Is unclear to this reviewer. On the one hand, It Is not used to corroborate the fly findings regarding iron needs from neuroblasts. On the other hand, and without further explanation, authors migrate from a fly tumor model based on modifying all neuroblasts to a mammalian model based exclusively on a glioma. The authors should clarify those issues.

Although iron transporter probably is different in Drosophila and mammal, iron function is conserved as an essential nutrient for cell growth and proliferation from Drosophila to mammal. The data of fly suggested that iron is critical for brain tumor growth and thus we verified this in mammalian model. Glioma is the most common form of central nervous system neoplasm that originates from neuroglial stem or progenitor cells[10]. Therefore, we validated the effect of iron chelator DFP on glioma in mice and found that DFP could suppress the glioma growth and further prolong the survival of tumor-bearing mice.

Minor points

Although referred to adult flies, the authors did not include either in the introduction or in the discussion existing literature about expression of ferritins in glia or alterations of iron metabolism in fly glia cells (PMID: 21440626 and 25841783, respectively) or usage of the iron chelator DFP in drosophila (PMID: 23542074). The author should check these manuscripts and consider the possibility of incorporating them into their manuscript.

Thanks for your remind. We have incorporated all recommended papers into our manuscript line 65-67 and 168.

The number of experiments in each figure is missing.

All experiments were repeated at least three times. And we revised this in Quantifications and Statistical Analysis of Materials and methods.

If graphs are expressed as mean +/- sem, it is difficult to understand the significance stated by the authors in Figure 2E.

We apologize for this mistake and have revised this in Quantifications and Statistical Analysis. All statistical results were presented as means ± SD.

When authors measure aconitase activity, are they measuring all (cytosolic and mitochondrial) or only one of them? This is important to better understand the experiments done by the authors to describe any mitochondrial contribution (see above in major points).

In this experiment, we were measuring the total aconitase activity. We also tried to determine mitochondrial aconitase but it failed, which was possibly ascribed to low biomass of tissue sample.

In this line, why do controls in aconitase and atp lack an error bar? Are the statistical tests applied the correct ones? It is not the same to have paired or unpaired observations.

It is the normalization. We repeated these experiments at least three times in different weeks respectively, because the whole process was time-consuming and energy-consuming including the collection of brains, protein determination and ATP or aconitase determination. And the efficiency of aconitase or ATP kit changed with time. We cannot control the experiment condition identically in different batches. Therefore, we performed normalization every time to present the more accurate result. The control group was normalized as 1 via dividing into itself and other groups were divided by the control. This normalized process was repeated three times. Therefore, there is no error bar in the control group. We think it is appropriate to apply ANOVA with a Bonferroni test in the three groups.

In some cases, further rescue experiments would be appreciated. For example, expression of Ndi restores control NAD+ levels or number of NBs, it would be interesting to know if this is accompanied by restoring mitochondrial integrity and its ability to produce ATP.

We have determined ATP production after overexpressing Ndi1 and provided this result in Figure 4—figure supplement 1B. The data showed that expression of Ndi1 could restore ATP production upon glial Fer2LCH knockdown, which was consistent with our conclusion.

Lines 293-299 on page 7 are difficult to understand.

According to our above results, the decrease of NB number and proliferation upon glial ferritin knockdown (KD) was caused by energy deficiency. As shown in the schematic diagram (Author response image 1), “T” represented the total energy which was used for NB maintenance and proliferation. “N” indicated the energy for maintaining NB number. “P” indicated the energy for NB proliferation. “T” is equal to “N” plus “P”. When ferritin was knocked down in glia, “T”, “N” and “P” declined in “Ferritin KD” compared to “wildtype (WT)”. Knockdown of pros can prevent the differentiation of NB, but it cannot supply the energy for NB, which probably results in the rescue of NB number but not proliferation. Specifically, NB number increased significantly in “Ferritin KD Pros KD” compared to “Ferritin KD”, which resulted in consuming more energy for NB maintenance in “Ferritin KD Pros KD”. As shown in the schematic diagram, “T” was not changed between “Ferritin KD Pros KD” and “Ferritin KD”, whereas ”N” was increased in “Ferritin KD Pros KD” compared to “Ferritin KD”. Thus, “P” was decreased, which suggested that less energy was remained for proliferation, leading to the failure of rescue in NB proliferation. It seemed that the level of proliferation in “Ferritin KD Pros KD” was even lower than “Ferritin KD”.

Author response image 1.

The schematic diagram of relationship between energy and NB function in different groups. “T” represents total energy for NB maintenance and proliferation. “N” represents the energy for NB maintenance. “P” represents the energy for NB proliferation. T=N+P 

Line 601 should indicate that Tables 2 and 3 are part of the supplementary material.

We have revised this in line 678.

Figure 4-supplement 1. Only validation of 2 genes from a RNAseq seems too little.

We dissected hundreds of brains for sorting NBs because of low biomass of fly brain. This is a difficult and energy-consuming work. Most NBs were used for RNA-seq, so we can only use a small amount of sample left for validation which is not enough for more genes.

Figure 6E, the authors indicate that 10 mg/ml DFP injection could significantly prolong the survival time. Which increase in % is produced by DFP?

We have provided the bar graph in Author response image 2. The increase is about 16.67% by DFP injection.

Author response image 2.

The bar graph of survival time of mice treated with DFP. (The unpaired two-sided Student’s t test was employed to assess statistical significance. Statistical results were presented as means ± SD. n=7,6; *: p<0.05)

Reviewer #3 (Recommendations For The Authors):

As I read the initial results that built the story (glia make ferritin>release it> NBs take them up>use it for TCA and ETC) I kept thinking about what it meant for NBs to be 'lost'. This led me to consider alternate possibilities that the results might point to, other than the ones the authors were suggesting. It was only in Figure 5 that the authors ruled out some of those possibilities. I would suggest that they first illustrate how NBs are lost upon glial ferritin loss of function before they delve into the mechanism. This would also be a place to similarly address that glial numbers and general morphology are unchanged upon ferritin loss.

This recommendation provides a valuable guideline to build this story especially for researchers who are interested in neural stem cell studies. Actually, we tried this logic to present our study but found that there are several gaps in the middle of the manuscript, such as the relationship between glial ferritin and Pros localization in NB, so that the whole story cannot be fluently presented. Therefore, we decided to present this study in the current way.

More details of the screen would be useful to know. How many lines did they screen, what was the assay? This is not mentioned anywhere in the text.

We have added this in Screen of Materials and methods. We screened about 200 lines which are components of classical signaling pathways, highly expressed genes in glial cells or secretory protein encoding genes. UAS-RNAi lines were crossed with repo-Gal4, and then third-instar larvae of F1 were dissected. We got the brains from F1 larvae and performed immunostaining with Dpn and PH3. Finally, we observed the brain in Confocal Microscope.

Many graphs seem to be repeated in the main figures and the supplementary data. This is unnecessary, or at least should be mentioned.

We appreciate your kind reminder. However, we carefully went through all the figures and did not find the repeated graphs, though some of them look similar.

The authors mention that they tested which glial subtypes ferritin is needed in, but don't show the data. Could they please show the data? Same with the other iron transport/storage/regulation. Also, in both this and later sections, the authors could mention which Gal4 was used to label what cell types. The assumption is that the reader will know this information.

We have added the result of ferritin knockdown in glial subpopulations in Figure 1—figure supplement 2. However, considering that the quantity of iron-related genes, we did not take the picture, but we recorded this in Table 3.

For all their images showing colocalisation, magnified, single-colour images shown in grayscale will be useful. For example, without the magnification, it is not possible to see the NB expression of the protein trap line in Figure 2B. A magnified crop of a few NBs (not a single one like in 2C) would be more useful.

We have provided Figure 2A’, B’, D’ and Figure 3D’ as suggested.

There are a lot of very specific assays used to detect ROS, NAD, aconitase activity, among others. It would be nice to have a brief but clear description of how they work in the main text. I found myself having to refer to other sources to understand them. (I believe SoNAR should be attributed to Zhao et al 206 and not Bonnay et al 2020.)

We have added a brief description about ROS, aconitase activity, NAD in line 198-199, 229-231, and 269 as suggested.

I did not understand the normalisation done with respect to SoNAR. Is this standard practice? Is the assumption that 'overall protein levels will be higher in slowly proliferating NBs' reasonable? This is why they state the need to normalise.

The SoNAR normalization is not a standard practice. However, we think that our normalization of SoNar is reasonable. According to our results, the expression level of Dpn and Mira seemed higher in glial ferritin knockdown, so we speculated that some proteins accumulated in slowly proliferating NBs. Thus, we used Insc-GAL4 to drive DsRed for indicating the expression level of Insc and found that DsRed rose after glial ferritin knockdown, suggesting that Insc expression was increased indeed. Therefore, we have to normalize SoNar driven by Insc-GAL4 based on DsRed driven by Insc-Gal4, which eliminates the effect of increased Insc upon glial ferritin knockdown.

FAC is mentioned as a chelator? But the authors seem to use it oppositely. Is there an error?

FAC is a type of iron salt, which is used to supply iron. We have also indicated that in line 156 according to your advice. 

The lack of any cell death in the L3 brain surprised me. There should be plenty of hemilineages that die, as do many NBs, particularly in the abdominal segments. Is the stain working? Related to this, P35 is not the best method for rescuing cell death. H99 might be a better way to go.

We were also surprised to see this result and repeated this experiment for several times with both negative and positive controls. Moreover, we also used TUNEL to validate this result, which led to the same result. We will try to use H99 to rescue NB loss in the future, because it needs to be integrated and recombined with our current genetic tools.

It would be nice to see the aconitase activity signal as opposed to just the quantification.

This method can only determine the absorbance for indicating aconitase activity, so our result is just the quantification.

Glia are born after NBs are specified. In fact, they arise from NBs (and glioblasts). So, it's unlikely that the knockdown of ferritin in glia can at all affect initial NB specification.

We completely agree with this statement.

The section on tumor suppression seems out of place. The fly data on which the authors base this as an angle to chase is weak. Dividing cells will be impaired if they have inadequate energy production. As a therapeutic, this will affect every cell in the body. I'm not sure that cancer therapeutics is pursuing such broadly acting lines of therapies anymore.

Our data suggested that iron/ferritin is more critical for high proliferative cells. Tumor cells have a high expression of TfR (Transferrin Receptor)[11], which can bind to Transferrin and ferritin[12]. And ferritin specifically targets on the tumor cells[11]. Thus, we think iron/ferritin is extremely essential for tumor cells. If we can find the appropriate dose of iron/ferritin inhibitor, suppressing tumor growth but maintaining normal cell growth, iron/ferritin might be an effective target of tumor treatment.

The feedback from NB to glial ferritin is also weak data. The increased cell numbers (of unknown identity) could well be contributing to the increase in ferritin. I would omit the last two sections from the MS.

In brat RNAi and numb RNAi, increased cells are NB-like cells, which cannot undergo further differentiation and are not expected to produce ferritin. More importantly, we used Repo (glia marker) as the reference and quantified the ratio of ferritin level to Repo level, which can exclude the possibility that increased glial cells lead to the increase in ferritin.

References

(1) Tanimura T, Isono K, Takamura T, et al. Genetic Dimorphism in the Taste Sensitivity to Trehalose in Drosophila-Melanogaster. J Comp Physiol, 1982,147(4):433-7

(2) Myster DL, Duronio RJ. Cell cycle: To differentiate or not to differentiate? Current Biology, 2000,10(8):R302-R4

(3) Dalton S. Linking the Cell Cycle to Cell Fate Decisions. Trends in Cell Biology, 2015,25(10):592-600

(4) Nichol H, Law JH, Winzerling JJ. Iron metabolism in insects. Annu Rev Entomol, 2002,47:535-59

(5) Pham DQ, Winzerling JJ. Insect ferritins: Typical or atypical? Biochim Biophys Acta, 2010,1800(8):824-33

(6) Speder P, Brand AH. Systemic and local cues drive neural stem cell niche remodelling during neurogenesis in Drosophila. Elife, 2018,7

(7) Mumbauer S, Pascual J, Kolotuev I, et al. Ferritin heavy chain protects the developing wing from reactive oxygen species and ferroptosis. PLoS Genet, 2019,15(9):e1008396

(8) Xiao G, Wan Z, Fan Q, et al. The metal transporter ZIP13 supplies iron into the secretory pathway in Drosophila melanogaster. Elife, 2014,3:e03191

(9) Marelja Z, Leimkühler S, Missirlis F. Iron Sulfur and Molybdenum Cofactor Enzymes Regulate the  Life Cycle by Controlling Cell Metabolism. Front Physiol, 2018,9

(10) Morgan LL. The epidemiology of glioma in adults: a "state of the science" review. Neuro-Oncology, 2015,17(4):623-4

(11) Fan K, Cao C, Pan Y, et al. Magnetoferritin nanoparticles for targeting and visualizing tumour tissues. Nat Nanotechnol, 2012,7(7):459-64

(12) Li L, Fang CJ, Ryan JC, et al. Binding and uptake of H-ferritin are mediated by human transferrin receptor-1. Proc Natl Acad Sci U S A, 2010,107(8):3505-10

Author response:

Public Reviews:

Reviewer #1 (Public Review):

Summary:

This study sought to reveal the potential roles of m6A RNA methylation in gene dosage regulatory mechanisms, particularly in the context of aneuploid genomes in Drosophila. Specifically, this work looked at the relationships between the expression of m6A regulatory factors, RNA methylation status, classical and inverse dosage effects, and dosage compensation. Using RNA sequencing and m6A mapping experiments, an in-depth analysis was performed to reveal changes in m6A status and expression changes across multiple aneuploid Drosophila models. The authors propose that m6A methylation regulates MOF and, in turn, deposition of H4K16Ac, critical regulators of gene dosage in the context of genomic imbalance.

Strengths:

This study seeks to address an interesting question with respect to gene dosage regulation and the possible roles of m6A in that process. Previous work has linked m6A to X-inactivation in humans through the Xist lncRNA, and to the regulation of the Sxl in flies. This study seeks to broaden that understanding beyond these specific contexts to more broadly understand how m6A impacts imbalanced genomes in other contexts.

Weaknesses:

The methods being used particularly for analysis of m6A at both the bulk and transcript-specific level are not sufficiently specific or quantitative to be able to confidently draw the conclusions the authors seek to make. MeRIP m6A mapping experiments can be very valuable, but differential methylation is difficult to assess when changes are small (as they often are, in this study but also m6A studies more broadly). For instance, based on the data presented and the methods described, it is not clear that the statement that "expression levels at m6A sites in aneuploidies are significantly higher than that in wildtype" is supported. MeRIP experiments are not quantitative, and since there are far fewer peaks in aneuploidies, it stands to reason that more antibody binding sites may be available to enrich those fewer peaks to a larger extent. But based on the data as presented (figure 2D) this conclusion was drawn from RPKM in IP samples, which may not fully account for changing transcript abundances in absolute (expression level changes) and relative (proportion of transcripts in input RNA sample) terms.

Methylated RNA immunoprecipitation followed by sequencing (MeRIP-seq) is a commonly used strategy of genome-wide mapping of m6A modification. This method uses anti-m6A antibody to immunoprecipitate RNA fragments, which results in selective enrichment of methylated RNA. Then the RNA fragments were subjected to deep sequencing, and the regions enriched in the immunoprecipitate relative to input samples are identified as m6A peaks using the peak calling algorithm. We identified m6A peaks in different samples by the exomePeak2 program and determined common m6A peaks for each genotype based on the intersection of biological replicates. Figure 2D shows the RPM values of m6A peaks in MeRIP samples for each genotype, indicating that the levels of reads in the m6A peak regions were significantly higher in the aneuploid IP samples than in wildtypes. When the enrichment of IP samples relative to Input samples (RPM.IP/RPM.Input) was taken into account, the statistics for all three aneuploidies were still significantly higher than those of the wildtypes (Mann Whitney U test p-values < 0.001). This analysis is not about changes in the abundance of transcripts, but from the MeRIP perspective, showing that there are relatively more m6A-modified reads mapped to the m6A peaks in aneuploidies than that in wildtypes. In addition, we have added the results of IP/Input in the main text, and revised the description in the manuscript to make it more precise to reduce possible misunderstandings.

The bulk-level m6A measurements as performed here also cannot effectively support these conclusions, as they are measured in total RNA. The focus of the work is mRNA m6A regulators, but m6A levels measured from total RNA samples will not reflect mRNA m6A levels as there are other abundance RNAs that contain m6A (including rRNA). As a result, conclusions about mRNA m6A levels from these measurements are not supported.

According to some published articles, m6A levels of purified mRNA or total RNA can be detected by different methods (such as mass spectrometry, 2D thin-layer chromatography, etc.) in Drosophila cells or tissues [1-3].

Here, we used the EpiQuik m6A RNA Methylation Quantification Kit (Colorimetric) (Epigentek, NY, USA, Cat # P-9005), which is suitable for detecting m6A methylation status directly using total RNA isolated from any species such as mammals, plants, fungi, bacteria, and viruses. This kit has previously been used by researchers to detect the m6A/A ratio in total RNA [4, 5] or purified mRNA [6] from different species.

In order to compare the m6A levels between the total RNA and mRNA, it was shown that the enrichment of mRNA from total RNA using Dynabeads™mRNA Purification Kit (Invitrogen Cat # 61006) did not show any significantly differences comparing with the results of total RNA (Figure 1). That’s the reason why most of the results of m6A levels in the manuscript were detected in total RNA.

Author response image 1.

The m6A levels of total RNA and mRNA

As suggested, we will try to extract and purify mRNA from different genotypes to verify our conclusion based on the m6A levels of total RNA if necessary. In addition, m6A modification in other types of RNA other than mRNA (e.g., lncRNA, rRNA) is not necessarily meaningless. We will also add discussions of this issue in the manuscript.

(1) Lence T, et al. (2016) m6A modulates neuronal functions and sex determination in Drosophila. Nature 540(7632):242-247.

(2) Haussmann IU, et al. (2016) m(6)A potentiates Sxl alternative pre-mRNA splicing for robust Drosophila sex determination. Nature 540(7632):301-304.

(3) Kan L, et al. (2017) The m(6)A pathway facilitates sex determination in Drosophila. Nat Commun 8:15737.

(4) Zhu C, et al. (2023) RNA Methylome Reveals the m(6)A-mediated Regulation of Flavor Metabolites in Tea Leaves under Solar-withering. Genomics Proteomics Bioinformatics 21(4):769-787.

(5) Song H, et al. (2021) METTL3-mediated m(6)A RNA methylation promotes the anti-tumour immunity of natural killer cells. Nat Commun 12(1):5522.

(6) Yin H, et al. (2021) RNA m6A methylation orchestrates cancer growth and metastasis via macrophage reprogramming. Nat Commun 12(1):1394.

Reviewer #2 (Public Review):

Summary:

The authors have tested the effects of partial- or whole-chromosome aneuploidy on the m6A RNA modification in Drosophila. The data reveal that overall m6A levels trend up but that the number of sites found by meRIP-seq trend down, which seems to suggest that aneuploidy causes a subset of sites to become hyper-methylated. Subsequent bioinformatic analysis of other published datasets establish correlations between the activity of the H4K16 acetyltransferase dosage compensation complex (DCC) and the expression of m6A components and m6A abundance, suggesting that DCC and m6A can act in a feedback loop on each other. Overall, this paper uses bioinformatic trends to generate a candidate model of feedback between DCC and m6A. It would be improved by functional studies that validate the effect in vivo.

Strengths:

• Thorough bioinformatic analysis of their data.

• Incorporation of other published datasets that enhance scope and rigor.

• Finds trends that suggest that a chromosome counting mechanism can control m6A, as fits with pub data that the Sxl mRNA is m6A modified in XX females and not XY males.

• Suggests this counting mechanism may be due to the effect of chromatin-dependent effects on the expression of m6A components.

Weaknesses:

• The linkage between H4K16 machinery and m6A is indirect and based on bioinformatic trends with little follow-up to test the mechanistic bases of these trends.

We found a set of ChIP-seq data (GSE109901) of H4K16ac in female and male Drosophila larvae from the public database, and analyzed whether H4K16ac is directly associated with m6A regulator genes. ChIP-seq is a standard method to study transcription factor binding and histone modification by using efficient and specific antibodies for immunoprecipitation. The results showed that there were H4K16ac peaks at the 5' region in gene of m6A reader Ythdc1 in both males and females. In addition, most of the genome sites where the other m6A regulator genes located are acetylated at H4K16 in both sexes, except that Ime4 shows sexual dimorphism and only contains H4K16ac peak in females. These results indicate that the m6A regulator gene itself is acetylated at H4K16, so there is a direct relationship between H4K16ac and m6A regulators. We have added these contents to the text.

Besides the above conclusion from the seq data, we are also going to do some experiments to test the linkage between H4K16 and m6A in the next, such as how about the m6A levels when MOF is over expressed with the increased levels of H4K16Ac, the H4K16 levels when YT521B is knocked down or over expressed and the relative expression levels of important regulatory genes in there.

• The paper lacks sufficient in vivo validation of the effects of DCC alleles on m6A and vice versa. For example, Is the Ythdc1 genomic locus a direct target of the DCC component Msl-2 ? (see Figure 7).

In order to study whether Ythdc1 genomic locus is a direct target of DCC component, we first analyzed a published MSL2 ChIP-seq data of Drosophila (GSE58768). Since MSL2 is only expressed in males under normal conditions, this set of data is from male Drosophila. According to the results, the majority (99.1%) of MSL2 peaks are located on the X chromosome, while the MSL2 peaks on other chromosomes are few. This is consistent with the fact that MSL2 is enriched on the X chromosome in male Drosophila [1, 2]. Ythdc1 gene is located on chromosome 3L, and there is no MSL2 peak near it. Similarly, other m6A regulator genes are not X-linked, and there is no MSL2 peak. Then we analyzed the MOF ChIP-seq data (GSE58768) of male Drosophila. It was found that 61.6% of MOF peaks were located on the X chromosome, which was also expected [3, 4]. Although there are more MOF peaks on autosomes than MSL2 peaks, MOF peaks are absent on m6A regulator genes on autosomes. Therefore, at present, there is no evidence that the gene locus of m6A regulators are the direct targets of DCC component MSL2 and MOF, which may be due to the fact that most MSL2 and MOF are tethered to the X chromosome by MSL complex under physiological conditions. Whether there are other direct or indirect interactions between Ythdc1 and MSL2 is an issue worthy of further study in the future.

(1) Bashaw GJ & Baker BS (1995) The msl-2 dosage compensation gene of Drosophila encodes a putative DNA-binding protein whose expression is sex specifically regulated by Sex-lethal. Development 121(10):3245-3258.

(2) Kelley RL, et al. (1995) Expression of msl-2 causes assembly of dosage compensation regulators on the X chromosomes and female lethality in Drosophila. Cell 81(6):867-877.

(3) Kind J, et al. (2008) Genome-wide analysis reveals MOF as a key regulator of dosage compensation and gene expression in Drosophila. Cell 133(5):813-828.

(4) Conrad T, et al. (2012) The MOF chromobarrel domain controls genome-wide H4K16 acetylation and spreading of the MSL complex. Dev Cell 22(3):610-624.

Quite a bit of technical detail is omitted from the main text, making it difficult for the reader to interpret outcomes.

(1) Please add the tissues to the labels in Figure 1D.

Figure 1D shows the subcellular localization of FISH probe signals in Drosophila embryos. Arrowheads indicate the foci of probe signals. The corresponding tissue types are (1) blastoderm nuclei; (2) yolk plasm and pole cells; (3) brain and midgut; (4) salivary gland and midgut; (5) blastoderm nuclei and yolk cortex; (6) blastoderm nuclei and pole cells; (7) blastoderm nuclei and yolk cortex; (8) germ band. We have added these to the manuscript.

(2) In the main text, please provide detail on the source tissues used for meRIP; was it whole larvae? adult heads? Most published datasets are from S2 cells or adult heads and comparing m6A across tissues and developmental stages could introduce quite a bit of variability, even in wt samples. This issue seems to be what the authors discuss in lines 197-199.

In this article, the material used to perform MeRIP-seq was the whole third instar larvae. Because trisomy 2L and metafemale Drosophila died before developing into adults, it was not possible to use the heads of adults for MeRIP-seq detection of aneuploidy. For other experiments described here, the m6A abundance was measured using whole larvae or adult heads; material used for RT-qPCR analysis was whole larvae, larval brains, or adult heads; Drosophila embryos at different developmental stages were used for fluorescence in situ hybridization (FISH) experiments. We provide a detailed description of the experimental material for each assay in the manuscript.

(3) In the main text, please identify the technique used to measure "total m6A/A" in Fig 2A. I assume it is mass spec.

We used the EpiQuik m6A RNA Methylation Quantification Kit (Colorimetric) (Epigentek, NY, USA, Cat # P-9005) to measure the m6A/A ratio in RNA samples. This kit is commercially available for quantification of m6A RNA methylation, which used colorimetric assay with easy-to-follow steps for convenience and speed, and is suitable for detecting m6A methylation status directly using total RNA isolated from any species such as mammals, plants, fungi, bacteria, and viruses.

(4) Line 190-191: the text describes annotating m6A sites by "nearest gene" which is confusing. The sites are mapped in RNAs, so the authors must unambiguously know the identity of the gene/transcript, right?

When the m6A peaks were annotated using the R package ChIPseeker, it will include two items: "genomic annotation" and "nearest gene annotation". "Genomic annotation" tells us which genomic features the peak is annotated to, such as 5’UTR, 3’UTR, exon, etc. "Nearest gene annotation" indicates which specific gene/transcript the peak is matched to. We modified the description in the main text to make it easier to understand.

Author response:

The following is the authors’ response to the previous reviews.

Reviewer #3:

Comments on current version:

As mentioned in my first review, this work is significantly underpowered for the following reasons: 1) n=4 for each treatment group.; 2) no randomization of the surgical sites receiving treatments; 3) implants surgically inserted without precision/guided surgery. The authors have not addressed these concerns.

On a minor note: not sure why the authors present a methodology to evaluate the dynamic bone formation (line 272) but do not present results (i.e. by means of histomorphometrical analyses) utilizing this methodology.

We sincerely appreciate your thorough review and valuable feedback. We have carefully considered your comments and would like to address them as follows:

As mentioned in my first review, this work is significantly underpowered for the following reasons:

(1) n=4 for each treatment group.;

We acknowledge your concern regarding the limited sample size (n=4 per group). While we understand this may affect statistical power, our choice was influenced by ethical considerations in animal experimentation and resource constraints. Increasing the sample size would undoubtedly strengthen the statistical power of our study. However, the logistical and ethical constraints associated with using a larger number of animals in such invasive procedures were significant limiting factors. Specifically, increasing the number of medium to large experimental animals could raise ethical issues, so we used the minimum number possible. Additionally, our study design was reviewed and approved by the animal IRB, which dictated the minimum number of animals we could use. Nevertheless, we conducted power analysis to ensure that our sample size, although limited, was sufficient to detect significant differences given the high variability typically observed in biological responses. The results obtained from our n=4 samples showed consistent trends and significant differences between groups, indicating the robustness of our findings. I will include this point in the limitations section of the discussion. Thank you.

(2) no randomization of the surgical sites receiving treatments;

Thank you for pointing out this issue. We agree that randomization is essential when considering individual differences and the anatomical variations of the jawbone, such as those found in humans. However, this study is an animal experiment where other conditions were controlled, and the interventions were applied after complete bone healing following tooth extraction. Therefore, the impact of randomization of surgical sites was likely minimal, and it is challenging to determine whether it significantly influenced the experimental results. Of course, twelve female OVX beagles were randomly designated into three groups. (Methods section, line 298) However regarding your concern, we would like to present the robustness of histological results from different surgical sites as shown below. Also we will include this point in the limitations section of the discussion.

Histologic analysis of the different surgical sites showed significant differences in bone formation and osseointegration among the three treatment groups: vehicle control, rhPTH(1-34), and dimeric Cys25PTH(1-34). Goldner trichrome staining (Figure A-C) showed enhanced bone formation in both the rhPTH(1-34) and dimeric Cys25PTH(1-34) groups compared to the vehicle control group. The rhPTH(1-34) group showed the most pronounced bone mass gain around the implant. Both treatment groups showed improved bone-to-implant contact compared to the control group, as indicated by the red arrows.

Masson trichrome staining (Figure D-F) further confirmed these results, showing an increase in bone matrix (blue staining) in the rhPTH(1-34) and dimeric Cys25PTH(1-34) groups, with the dimeric rhPTH(1-34) group showing the most extensive and dense bone formation.

TRAP staining (Figure G-I and G'-I') was used to assess osteoclast activity. Interestingly, both the rhPTH(1-34) and dimeric Cys25PTH(1-34) groups showed an increase in TRAP-positive cells compared to the vehicle control, suggesting enhanced bone remodeling activity. The highest number of TRAP-positive cells was observed in the rhPTH(1-34) group and the highest trabecular number, indicating the most active bone remodeling.

To summarize the results, histological analyses revealed that both rhPTH(1-34) and dimeric Cys25PTH(1-34) treatments significantly enhanced osseointegration and bone formation around titanium implants in a postmenopausal osteoporosis model compared to the control. The rhPTH(1-34) group demonstrated superior outcomes, exhibiting the most substantial increase in bone volume, bone-to-implant contact, and osteoclastic activity, indicating its greater efficacy in promoting bone regeneration and implant integration in this experimental context.

Author response image 1.

Histological analysis using Goldner trichrome, Masson trichrome, and TRAP staining

(3) implants surgically inserted without precision/guided surgery. The authors have not addressed these concerns.

The primary purpose of precision guides is to prevent damage to various anatomical structures and to ensure perfect placement at the desired location. Even disregarding the potential inaccuracies of precision guides in actual clinical settings, the primary goal of this animal experiment was not to achieve perfect placement or prevent damage to anatomical structures. Instead, the objective was to histologically measure the integrity of the bone surrounding titanium fixture's platform after pharmacological intervention, ensuring it was fully seated in the alveolar bone. To this end, we secured sufficient visibility through periosteal dissection to confirm the perfect placement of the implant and adhered to the principle of maintaining sufficient mesiodistal distance between each fixture. Using such precision guides in this animal experiment, which is not an evaluation of 'implant precision guides,' could potentially introduce inaccuracies and contradict the experimental objectives. Furthermore, since this experiment was conducted on an edentulous ridge where all teeth had been extracted, achieving the same placement as in the presurgical simulation would be impossible, even with the use of precision guides. Thank you once again for your constructive feedback. We will include this point in the limitations section of the discussion.

On a minor note: not sure why the authors present a methodology to evaluate the dynamic bone formation (line 272) but do not present results (i.e. by means of histomorphometrical analyses) utilizing this methodology.

As the reviewer mentioned, we confirmed that the sentence was included in the Methods section despite the analysis not actually being performed. We sincerely apologize for this oversight and will make the necessary corrections immediately. Thank you very much for your keen observation.

Author response:

The following is the authors’ response to the original reviews.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

This paper presents a compelling and comprehensive study of decision-making under uncertainty. It addresses a fundamental distinction between belief-based (cognitive neuroscience) formulations of choice behaviour with reward-based (behavioural psychology) accounts. Specifically, it asks whether active inference provides a better account of planning and decision-making, relative to reinforcement learning. To do this, the authors use a simple but elegant paradigm that includes choices about whether to seek both information and rewards. They then assess the evidence for active inference and reinforcement learning models of choice behaviour, respectively. After demonstrating that active inference provides a better explanation of behavioural responses, the neuronal correlates of epistemic and instrumental value (under an optimised active inference model) are characterised using EEG. Significant neuronal correlates of both kinds of value were found in sensor and source space. The source space correlates are then discussed sensibly, in relation to the existing literature on the functional anatomy of perceptual and instrumental decision-making under uncertainty.

Strengths:

The strengths of this work rest upon the theoretical underpinnings and careful deconstruction of the various determinants of choice behaviour using active inference. A particular strength here is that the experimental paradigm is designed carefully to elicit both information-seeking and reward-seeking behaviour; where the information-seeking is itself separated into resolving uncertainty about the context (i.e., latent states) and the contingencies (i.e., latent parameters), under which choices are made. In other words, the paradigm - and its subsequent modelling - addresses both inference and learning as necessary belief and knowledge-updating processes that underwrite decisions.

The authors were then able to model belief updating using active inference and then look for the neuronal correlates of the implicit planning or policy selection. This speaks to a further strength of this study; it provides some construct validity for the modelling of belief updating and decision-making; in terms of the functional anatomy as revealed by EEG. Empirically, the source space analysis of the neuronal correlates licences some discussion of functional specialisation and integration at various stages in the choices and decision-making.

In short, the strengths of this work rest upon a (first) principles account of decision-making under uncertainty in terms of belief updating that allows them to model or fit choice behaviour in terms of Bayesian belief updating - and then use relatively state-of-the-art source reconstruction to examine the neuronal correlates of the implicit cognitive processing.

Response: We are deeply grateful for your careful review of our work and for the thoughtful feedback you have provided. Your dedication to ensuring the quality and clarity of the work is truly admirable. Your comments have been invaluable in guiding us towards improving the paper, and We appreciate your time and effort in not just offering suggestions but also providing specific revisions that I can implement. Your insights have helped us identify areas where I can strengthen the arguments and clarify the methodology.

Comment 1:

The main weaknesses of this report lies in the communication of the ideas and procedures. Although the language is generally excellent, there are some grammatical lapses that make the text difficult to read. More importantly, the authors are not consistent in their use of some terms; for example, uncertainty and information gain are sometimes conflated in a way that might confuse readers. Furthermore, the descriptions of the modelling and data analysis are incomplete. These shortcomings could be addressed in the following way.

First, it would be useful to unpack the various interpretations of information and goal-seeking offered in the (active inference) framework examined in this study. For example, it will be good to include the following paragraph:

"In contrast to behaviourist approaches to planning and decision-making, active inference formulates the requisite cognitive processing in terms of belief updating in which choices are made based upon their expected free energy. Expected free energy can be regarded as a universal objective function, specifying the relative likelihood of alternative choices. In brief, expected free energy can be regarded as the surprise expected following some action, where the expected surprise comes in two flavours. First, the expected surprise is uncertainty, which means that policies with a low expected free energy resolve uncertainty and promote information seeking. However, one can also minimise expected surprise by avoiding surprising, aversive outcomes. This leads to goal-seeking behaviour, where the goals can be regarded as prior preferences or rewarding outcomes.

Technically, expected free energy can be expressed in terms of risk plus ambiguity - or rearranged to be expressed in terms of expected information gain plus expected value, where value corresponds to (log) prior preferences. We will refer to both decompositions in what follows; noting that both decompositions accommodate information and goal-seeking imperatives. That is, resolving ambiguity and maximising information gain have epistemic value, while minimising risk or maximising expected value have pragmatic or instrumental value. These two kinds of values are sometimes referred to in terms of intrinsic and extrinsic value, respectively [1-4]."

Response 1: We deeply thank you for your comments and corresponding suggestions about our interpretations of active inference. In response to your identified weaknesses and suggestions, we have added corresponding paragraphs in the Methods section (The free energy principle and active inference, line 95-106):

“Active inference formulates the necessary cognitive processing as a process of belief updating, where choices depend on agents' expected free energy. Expected free energy serves as a universal objective function, guiding both perception and action. In brief, expected free energy can be seen as the expected surprise following some policies. The expected surprise can be reduced by resolving uncertainty, and one can select policies with lower expected free energy which can encourage information-seeking and resolve uncertainty. Additionally, one can minimize expected surprise by avoiding surprising or aversive outcomes (oudeyer et al., 2007; Schmidhuber et al., 2010). This leads to goal-seeking behavior, where goals can be viewed as prior preferences or rewarding outcomes.

Technically, expected free energy can also be expressed as expected information gain plus expected value, where the value corresponds to (log) prior preferences. We will refer to both formulations in what follows. Resolving ambiguity, minimizing risk, and maximizing information gain has epistemic value while maximizing expected value have pragmatic or instrumental value. These two types of values can be referred to in terms of intrinsic and extrinsic value, respectively (Barto et al., 2013; Schwartenbeck et al., 2019).”

Oudeyer, P. Y., & Kaplan, F. (2007). What is intrinsic motivation? A typology of computational approaches. Frontiers in neurorobotics, 1, 108.

Schmidhuber, J. (2010). Formal theory of creativity, fun, and intrinsic motivation (1990–2010). IEEE transactions on autonomous mental development, 2(3), 230-247.

Barto, A., Mirolli, M., & Baldassarre, G. (2013). Novelty or surprise?. Frontiers in psychology, 4, 61898.

Schwartenbeck, P., Passecker, J., Hauser, T. U., FitzGerald, T. H., Kronbichler, M., & Friston, K. J. (2019). Computational mechanisms of curiosity and goal-directed exploration. elife, 8, e41703.

Comment 2:

The description of the modelling of choice behaviour needs to be unpacked and motivated more carefully. Perhaps along the following lines:

"To assess the evidence for active inference over reinforcement learning, we fit active inference and reinforcement learning models to the choice behaviour of each subject. Effectively, this involved optimising the free parameters of active inference and reinforcement learning models to maximise the likelihood of empirical choices. The resulting (marginal) likelihood was then used as the evidence for each model. The free parameters for the active inference model scaled the contribution of the three terms that constitute the expected free energy (in Equation 6). These coefficients can be regarded as precisions that characterise each subjects' prior beliefs about contingencies and rewards. For example, increasing the precision or the epistemic value associated with model parameters means the subject would update her beliefs about reward contingencies more quickly than a subject who has precise prior beliefs about reward distributions. Similarly, subjects with a high precision over prior preferences or extrinsic value can be read as having more precise beliefs that she will be rewarded. The free parameters for the reinforcement learning model included..."

Response 2: We deeply thank you for your comments and corresponding suggestions about our description of the behavioral modelling. In response to your identified weaknesses and suggestions, we have added corresponding content in the Results section (Behavioral results, line 279-293):

“To assess the evidence for active inference over reinforcement learning, we fit active inference (Eq.9), model-free reinforcement learning, and model-based reinforcement learning models to the behavioral data of each participant. This involved optimizing the free parameters of active inference and reinforcement learning models. The resulting likelihood was used to calculate the Bayesian Information Criterion (BIC) (Vrieze 2012) as the evidence for each model. The free parameters for the active inference model (AL, AI, EX, prior, and α) scaled the contribution of the three terms that constitute the expected free energy in Eq.9. These coefficients can be regarded as precisions that characterize each participant's prior beliefs about contingencies and rewards. For example, increasing α means participants would update their beliefs about reward contingencies more quickly, increasing AL means participants would like to reduce ambiguity more, and increasing AI means participants would like to learn the hidden state of the environment and avoid risk more. The free parameters for the model-free reinforcement learning model are the learning rate α and the temperature parameter γ and the free parameters for the model-based are the learning rate α, the temperature parameter γ and prior (the details for the model-free reinforcement learning model can be seen in Eq.S1-11 and the details for the model-based reinforcement learning model can be seen Eq.S12-23 in the Supplementary Method). The parameter fitting for these three models was conducted using the `BayesianOptimization' package in Python (Frazire 2018), first randomly sampling 1000 times and then iterating for an additional 1000 times.”

Vrieze, S. I. (2012). Model selection and psychological theory: a discussion of the differences between the Akaike information criterion (AIC) and the Bayesian information criterion (BIC). Psychological methods, 17(2), 228.

Frazier, P. I. (2018). A tutorial on Bayesian optimization. arXiv preprint arXiv:1807.02811.

Comment 3:

In terms of the time-dependent correlations with expected free energy - and its constituent terms - I think the report would benefit from overviewing these analyses with something like the following:

"In the final analysis of the neuronal correlates of belief updating - as quantified by the epistemic and intrinsic values of expected free energy - we present a series of analyses in source space. These analyses tested for correlations between constituent terms in expected free energy and neuronal responses in source space. These correlations were over trials (and subjects). Because we were dealing with two-second timeseries, we were able to identify the periods of time during decision-making when the correlates were expressed.

In these analyses, we focused on the induced power of neuronal activity at each point in time, at each brain source. To illustrate the functional specialisation of these neuronal correlates, we present whole-brain maps of correlation coefficients and pick out the most significant correlation for reporting fluctuations in selected correlations over two-second periods. These analyses are presented in a descriptive fashion to highlight the nature and variety of the neuronal correlates, which we unpack in relation to the existing EEG literature in the discussion. Note that we did not attempt to correct for multiple comparisons; largely, because the correlations observed were sustained over considerable time periods, which would be almost impossible under the null hypothesis of no correlations."

Response 3: We deeply thank you for your comments and corresponding suggestions about our description of the regression analysis in the source space. In response to your suggestions, we have added corresponding content in the Results section (EEG results at source level, line 331-347):

“In the final analysis of the neural correlates of the decision-making process, as quantified by the epistemic and intrinsic values of expected free energy, we presented a series of linear regressions in source space. These analyses tested for correlations over trials between constituent terms in expected free energy (the value of avoiding risk, the value of reducing ambiguity, extrinsic value, and expected free energy itself) and neural responses in source space. Additionally, we also investigated the neural correlate of (the degree of) risk, (the degree of) ambiguity, and prediction error. Because we were dealing with a two-second time series, we were able to identify the periods of time during decision-making when the correlates were expressed. The linear regression was run by the "mne.stats.linear regression" function in the MNE package (Activity ~ Regressor + Intercept). Activity is the activity amplitude of the EEG signal in the source space and regressor is one of the regressors that we mentioned (e.g., expected free energy, the value of reducing ambiguity, etc.).

In these analyses, we focused on the induced power of neural activity at each time point, in the brain source space. To illustrate the functional specialization of these neural correlates, we presented whole-brain maps of correlation coefficients and picked out the brain region with the most significant correlation for reporting fluctuations in selected correlations over two-second periods. These analyses were presented in a descriptive fashion to highlight the nature and variety of the neural correlates, which we unpacked in relation to the existing EEG literature in the discussion. Note that we did not attempt to correct for multiple comparisons; largely, because the correlations observed were sustained over considerable time periods, which would be almost impossible under the null hypothesis of no correlations.”

Comment 4:

There was a slight misdirection in the discussion of priors in the active inference framework. The notion that active inference requires a pre-specification of priors is a common misconception. Furthermore, it misses the point that the utility of Bayesian modelling is to identify the priors that each subject brings to the table. This could be easily addressed with something like the following in the discussion:

"It is a common misconception that Bayesian approaches to choice behaviour (including active inference) are limited by a particular choice of priors. As illustrated in our fitting of choice behaviour above, priors are a strength of Bayesian approaches in the following sense: under the complete class theorem [5, 6], any pair of choice behaviours and reward functions can be described in terms of ideal Bayesian decision-making with particular priors. In other words, there always exists a description of choice behaviour in terms of some priors. This means that one can, in principle, characterise any given behaviour in terms of the priors that explain that behaviour. In our example, these were effectively priors over the precision of various preferences or beliefs about contingencies that underwrite expected free energy."

Response 4: We deeply thank you for your comments and corresponding suggestions about the prior of Bayesian methods. In response to your suggestions, we have added corresponding content in the Discussion section (The strength of the active inference framework in decision-making, line 447-453):

“However, it may be the opposite. As illustrated in our fitting results, priors can be a strength of Bayesian approaches. Under the complete class theorem (Wald 1947; Brown 1981), any pair of behavioral data and reward functions can be described in terms of ideal Bayesian decision-making with particular priors. In other words, there always exists a description of behavioral data in terms of some priors. This means that one can, in principle, characterize any given behavioral data in terms of the priors that explain that behavior. In our example, these were effectively priors over the precision of various preferences or beliefs about contingencies that underwrite expected free energy.”

Wald, A. (1947). An essentially complete class of admissible decision functions. The Annals of Mathematical Statistics, 549-555.

Brown, L. D. (1981). A complete class theorem for statistical problems with finite sample spaces. The Annals of Statistics, 1289-1300.

Reviewer #2 (Public Review):

Summary:

Zhang and colleagues use a combination of behavioral, neural, and computational analyses to test an active inference model of exploration in a novel reinforcement learning task.

Strengths:

The paper addresses an important question (validation of active inference models of exploration). The combination of behavior, neuroimaging, and modeling is potentially powerful for answering this question.

Response: We want to express our sincere gratitude for your thorough review of our work and for the valuable comments you have provided. Your attention to detail and dedication to improving the quality of the work are truly commendable. Your feedback has been invaluable in guiding us towards revisions that will strengthen the work. We have made targeted modifications based on most of the comments. However, due to factors such as time and energy constraints, we have not added corresponding analyses for several comments.

Comment 1:

The paper does not discuss relevant work on contextual bandits by Schulz, Collins, and others. It also does not mention the neuroimaging study of Tomov et al. (2020) using a risky/safe bandit task.

Response 1:

We deeply thank you for your suggestions about the relevant work. We now discussion and cite these representative papers in the Introduction section (line 42-55):

“The decision-making process frequently involves grappling with varying forms of uncertainty, such as ambiguity - the kind of uncertainty that can be reduced through sampling, and risk - the inherent uncertainty (variance) presented by a stable environment. Studies have investigated these different forms of uncertainty in decision-making, focusing on their neural correlates (Daw et al., 2006; Badre et al., 2012; Cavanagh et al., 2012).

These studies utilized different forms of multi-armed bandit tasks, e.g the restless multi-armed bandit tasks (Daw et al., 2006; Guha et al., 2010), risky/safe bandit tasks (Tomov et al., 2020; Fan et al., 2022; Payzan et al., 2013), contextual multi-armed bandit tasks (Schulz et al., 2015; Schulz et al., 2015; Molinaro et al., 2023). However, these tasks either separate risk from ambiguity in uncertainty, or separate action from state (perception). In our work, we develop a contextual multi-armed bandit task to enable participants to actively reduce ambiguity, avoid risk, and maximize rewards using various policies (see Section 2.2) and Figure 4(a)). Our task makes it possible to study whether the brain represents these different types of uncertainty distinctly (Levy et al., 2010) and whether the brain represents both the value of reducing uncertainty and the degree of uncertainty. The active inference framework presents a theoretical approach to investigate these questions. Within this framework, uncertainties can be reduced to ambiguity and risk. Ambiguity is represented by the uncertainty about model parameters associated with choosing a particular action, while risk is signified by the variance of the environment's hidden states. The value of reducing ambiguity, the value of avoiding risk, and extrinsic value together constitute expected free energy (see Section 2.1).”

Daw, N. D., O'doherty, J. P., Dayan, P., Seymour, B., & Dolan, R. J. (2006). Cortical substrates for exploratory decisions in humans. Nature, 441(7095), 876-879.

Badre, D., Doll, B. B., Long, N. M., & Frank, M. J. (2012). Rostrolateral prefrontal cortex and individual differences in uncertainty-driven exploration. Neuron, 73(3), 595-607.

Cavanagh, J. F., Figueroa, C. M., Cohen, M. X., & Frank, M. J. (2012). Frontal theta reflects uncertainty and unexpectedness during exploration and exploitation. Cerebral cortex, 22(11), 2575-2586.

Guha, S., Munagala, K., & Shi, P. (2010). Approximation algorithms for restless bandit problems. Journal of the ACM (JACM), 58(1), 1-50.

Tomov, M. S., Truong, V. Q., Hundia, R. A., & Gershman, S. J. (2020). Dissociable neural correlates of uncertainty underlie different exploration strategies. Nature communications, 11(1), 2371.

Fan, H., Gershman, S. J., & Phelps, E. A. (2023). Trait somatic anxiety is associated with reduced directed exploration and underestimation of uncertainty. Nature Human Behaviour, 7(1), 102-113.

Payzan-LeNestour, E., Dunne, S., Bossaerts, P., & O’Doherty, J. P. (2013). The neural representation of unexpected uncertainty during value-based decision making. Neuron, 79(1), 191-201.

Schulz, E., Konstantinidis, E., & Speekenbrink, M. (2015, April). Exploration-exploitation in a contextual multi-armed bandit task. In International conference on cognitive modeling (pp. 118-123).

Schulz, E., Konstantinidis, E., & Speekenbrink, M. (2015, November). Learning and decisions in contextual multi-armed bandit tasks. In CogSci.

Molinaro, G., & Collins, A. G. (2023). Intrinsic rewards explain context-sensitive valuation in reinforcement learning. PLoS Biology, 21(7), e3002201.

Levy, I., Snell, J., Nelson, A. J., Rustichini, A., & Glimcher, P. W. (2010). Neural representation of subjective value under risk and ambiguity. Journal of neurophysiology, 103(2), 1036-1047.

Comment 2:

The statistical reporting is inadequate. In most cases, only p-values are reported, not the relevant statistics, degrees of freedom, etc. It was also not clear if any corrections for multiple comparisons were applied. Many of the EEG results are described as "strong" or "robust" with significance levels of p<0.05; I am skeptical in the absence of more details, particularly given the fact that the corresponding plots do not seem particularly strong to me.

Response 2: We deeply thank you for your comments about our statistical reporting. We have optimized the fitting model and rerun all the statistical analyses. As can be seen (Figure 6, 7, 8, S3, S4, S5), the new regression results are significantly improved compared to the previous ones. Due to the limitation of space, we place the other relevant statistical results, including t-values, std err, etc., on our GitHub (https://github.com/andlab-um/FreeEnergyEEG). Currently, we have not conducted multiple comparison corrections based on Reviewer 1’s comments (Comments 3) “Note that we did not attempt to correct for multiple comparisons; largely, because the correlations observed were sustained over considerable time periods, which would be almost impossible under the null hypothesis of no correlations”.

Author response image 1.

Comment 3:

The authors compare their active inference model to a "model-free RL" model. This model is not described anywhere, as far as I can tell. Thus, I have no idea how it was fit, how many parameters it has, etc. The active inference model fitting is also not described anywhere. Moreover, you cannot compare models based on log-likelihood, unless you are talking about held-out data. You need to penalize for model complexity. Finally, even if active inference outperforms a model-free RL model (doubtful given the error bars in Fig. 4c), I don't see how this is strong evidence for active inference per se. I would want to see a much more extensive model comparison, including model-based RL algorithms which are not based on active inference, as well as model recovery analyses confirming that the models can actually be distinguished on the basis of the experimental data.

Response 3: We deeply thank you for your comments about the model comparison details. We previously omitted some information about the comparison model, as classical reinforcement learning is not the focus of our work, so we put the specific details in the supplementary materials. Now we have placed relevant information in the main text (see the part we have highlighted in yellow). We have now added the relevant information regarding the model comparison in the Results section (Behavioral results, line 279-293):

“To assess the evidence for active inference over reinforcement learning, we fit active inference (Eq.9), model-free reinforcement learning, and model-based reinforcement learning models to the behavioral data of each participant. This involved optimizing the free parameters of active inference and reinforcement learning models. The resulting likelihood was used to calculate the Bayesian Information Criterion (BIC) as the evidence for each model. The free parameters for the active inference model (AL, AI, EX, prior, and α) scaled the contribution of the three terms that constitute the expected free energy in Eq.9. These coefficients can be regarded as precisions that characterize each participant's prior beliefs about contingencies and rewards. For example, increasing α means participants would update their beliefs about reward contingencies more quickly, increasing AL means participants would like to reduce ambiguity more, and increasing AI means participants would like to learn the hidden state of the environment and avoid risk more. The free parameters for the model-free reinforcement learning model are the learning rate α and the temperature parameter γ and the free parameters for the model-based are the learning rate α, the temperature parameter γ and prior (the details for the model-free reinforcement learning model can be found in Eq.S1-11 and the details for the model-based reinforcement learning model can be found in Eq.S12-23 in the Supplementary Method). The parameter fitting for these three models was conducted using the `BayesianOptimization' package in Python, first randomly sampling 1000 times and then iterating for an additional 1000 times.”

We have now incorporated model-based reinforcement learning into our comparison models and placed the descriptions of both model-free and model-based reinforcement learning algorithms in the supplementary materials. We have also changed the criterion for model comparison to Bayesian Information Criterion. As indicated by the results, the performance of the active inference model significantly outperforms both comparison models.

Sorry, we didn't do model recovery before, but now we have placed the relevant results in the supplementary materials. From the result figures, we can see that each model fits its own generated simulated data well:

“To demonstrate how reliable our models are (the active inference model, model-free reinforcement learning model, and model-based reinforcement learning model), we run some simulation experiments for model recovery. We use these three models, with their own fitting parameters, to generate some simulated data. Then we will fit all three sets of data using these three models.

The model recovery results are shown in Fig.S6. This is the confusion matrix of models: the percentage of all subjects simulated based on a certain model that is fitted best by a certain model. The goodness-of-fit was compared using the Bayesian Information Criterion. We can see that the result of model recovery is very good, and the simulated data generated by a model can be best explained by this model.”

Author response image 2.

Comment 4:

Another aspect of the behavioral modeling that's missing is a direct descriptive comparison between model and human behavior, beyond just plotting log-likelihoods (which are a very impoverished measure of what's going on).

Response 4: We deeply thank you for your comments about the comparison between the model and human behavior. Due to the slight differences between our simulation experiments and real behavioral experiments (the "you can ask" stage), we cannot directly compare the model and participants' behaviors. However, we can observe that in the main text's simulation experiment (Figure 3), the active inference agent's behavior is highly consistent with humans (Figure 4), exhibiting an effective exploration strategy and a desire to reduce uncertainty. Moreover, we have included two additional simulation experiments in the supplementary materials, which demonstrate that active inference may potentially fit a wide range of participants' behavioral strategies.

Author response image 3.

(An active inference agent with AL=AI=EX=0. It can accomplish tasks efficiently like a human being, reducing the uncertainty of the environment and maximizing the reward.)

Author response image 4.

(An active inference agent with AL=AI=0, EX=10. It will only pursue immediate rewards (not choosing the "Cue" option due to additional costs), but it can also gradually optimize its strategy due to random effects.)

Author response image 5.

(An active inference agent with EX=0, AI=AL=10. It will only pursue environmental information to reduce the uncertainty of the environment. Even in "Context 2" where immediate rewards are scarce, it will continue to explore.) (a) shows the decision-making of active inference agents in the Stay-Cue choice. Blue corresponds to agents choosing the "Cue" option and acquiring "Context 1"; orange corresponds to agents choosing the "Cue" option and acquiring "Context 2"; purple corresponds to agents choosing the "Stay" option and not knowing the information about the hidden state of the environment. The shaded areas below correspond to the probability of the agents making the respective choices. (b) shows the decision-making of active inference agents in the Stay-Cue choice. The shaded areas below correspond to the probability of the agents making the respective choices. (c) shows the rewards obtained by active inference agents. (d) shows the reward prediction errors of active inference agents. (e) shows the reward predictions of active inference agents for the "Risky" path in "Context 1" and "Context 2".

Comment 5:

The EEG results are intriguing, but it wasn't clear that these provide strong evidence specifically for the active inference model. No alternative models of the EEG data are evaluated.

Overall, the central claim in the Discussion ("we demonstrated that the active inference model framework effectively describes real-world decision-making") remains unvalidated in my opinion.

Response 5: We deeply thank you for your comments. We applied the active inference model to analyze EEG results because it best fit the participants' behavioral data among our models, including the new added results. Further, our EEG results serve only to verify that the active inference model can be used to analyze the neural mechanisms of decision-making in uncertain environments (if possible, we could certainly design a more excellent reinforcement learning model with a similar exploration strategy). We aim to emphasize the consistency between active inference and human decision-making in uncertain environments, as we have discussed in the article. Active inference emphasizes both perception and action, which is also what we wish to highlight: during the decision-making process, participants not only passively receive information, but also actively adopt different strategies to reduce uncertainty and maximize rewards.

Reviewer #3 (Public Review):

Summary:

This paper aims to investigate how the human brain represents different forms of value and uncertainty that participate in active inference within a free-energy framework, in a two-stage decision task involving contextual information sampling, and choices between safe and risky rewards, which promotes a shift from exploration to exploitation. They examine neural correlates by recording EEG and comparing activity in the first vs second half of trials and between trials in which subjects did and did not sample contextual information, and perform a regression with free-energy-related regressors against data "mapped to source space." Their results show effects in various regions, which they take to indicate that the brain does perform this task through the theorised active inference scheme.

Strengths:

This is an interesting two-stage paradigm that incorporates several interesting processes of learning, exploration/exploitation, and information sampling. Although scalp/brain regions showing sensitivity to the active-inference-related quantities do not necessarily suggest what role they play, it can be illuminating and useful to search for such effects as candidates for further investigation. The aims are ambitious, and methodologically it is impressive to include extensive free-energy theory, behavioural modelling, and EEG source-level analysis in one paper.

Response: We would like to express our heartfelt thanks to you for carefully reviewing our work and offering insightful feedback. Your attention to detail and commitment to enhancing the overall quality of our work are deeply admirable. Your input has been extremely helpful in guiding us through the necessary revisions to enhance the work. We have implemented focused changes based on a majority of your comments. Nevertheless, owing to limitations such as time and resources, we have not included corresponding analyses for a few comments.

Comment 1:

Though I could surmise the above general aims, I could not follow the important details of what quantities were being distinguished and sought in the EEG and why. Some of this is down to theoretical complexity - the dizzying array of constructs and terms with complex interrelationships, which may simply be part and parcel of free-energy-based theories of active inference - but much of it is down to missing or ambiguous details.

Response 1: We deeply thank you for your comments about our work’s readability. We have significantly revised the descriptions of active inference, models, research questions, etc. Focusing on active inference and the free energy principle, we have added relevant basic descriptions and unified the terminology. We have added information related to model comparison in the main text and supplementary materials. We presented our regression results in clearer language. Our research focused on the brain's representation of decision-making in uncertain environments, including expected free energy, the value of reducing ambiguity, the value of avoiding risk, extrinsic value, ambiguity, and risk.

Comment 2:

In general, an insufficient effort has been made to make the paper accessible to readers not steeped in the free energy principle and active inference. There are critical inconsistencies in key terminology; for example, the introduction states that aim 1 is to distinguish the EEG correlates of three different types of uncertainty: ambiguity, risk, and unexpected uncertainty. But the abstract instead highlights distinctions in EEG correlates between "uncertainty... and... risk" and between "expected free energy .. and ... uncertainty." There are also inconsistencies in mathematical labelling (e.g. in one place 'p(s|o)' and 'q(s)' swap their meanings from one sentence to the very next).

Response 2: We deeply thank you for your comments about the problem of inconsistent terminology. First, we have unified the symbols and letters (P, Q, s, o, etc.) that appeared in the article and described their respective meanings more clearly. We have also revised the relevant expressions of "uncertainty" throughout the text. In our work, uncertainty refers to ambiguity and risk. Ambiguity can be reduced through continuous sampling and is referred to as uncertainty about model parameters in our work. Risk, on the other hand, is the inherent variance of the environment and cannot be reduced through sampling, which is referred to as uncertainty about hidden states in our work. In the analysis of the results, we focused on how the brain encodes the value of reducing ambiguity (Figure 8), the value of avoiding risk (Figure 6), and (the degree of) ambiguity (Figure S5) during action selection. We also analyzed how the brain encodes reducing ambiguity and avoiding risk during belief update (Figure 7).

Comment 3:

Some basic but important task information is missing, and makes a huge difference to how decision quantities can be decoded from EEG. For example:

- How do the subjects press the left/right buttons - with different hands or different fingers on the same hand?

Response 3: We deeply thank you for your comments about the missing task information. We have added the relevant content in the Methods section (Contextual two-armed bandit task and Data collection, line 251-253):

“Each stage was separated by a jitter ranging from 0.6 to 1.0 seconds. The entire experiment consists of a single block with a total of 120 trials. The participants are required to use any two fingers of one hand to press the buttons (left arrow and right arrow on the keyboard).”

Comment 4:

- Was the presentation of the Stay/cue and safe/risky options on the left/right sides counterbalanced? If not, decisions can be formed well in advance especially once a policy is in place.

Response 4: The presentation of the Stay/cue and safe/risky options on the left/right sides was not counterbalanced. It is true that participants may have made decisions ahead of time. However, to better study the state of participants during decision-making, our choice stages consist of two parts. In the first two seconds, we ask participants to consider which option they would choose, and after these two seconds, participants are allowed to make their choice (by pressing the button).

We also updated the figure of the experiment procedure as below (We circled the time that the participants spent on making decisions).

Author response image 6.

Comment 5:

- What were the actual reward distributions ("magnitude X with probability p, magnitude y with probability 1-p") in the risky option?

Response 5: We deeply thank you for your comments about the missing task information. We have placed the relevant content in the Methods section (Contextual two-armed bandit task and Data collection, line 188-191):

“The actual reward distribution of the risky path in "Context 1" was [+12 (55%), +9 (25%), +6 (10%), +3 (5%), +0 (5%)] and the actual reward distribution of the risky path in "Context 2" was [+12 (5%), +9 (5%), +6 (10%), +3 (25%), +0 (55%)].”

Comment 6:

The EEG analysis is not sufficiently detailed and motivated.

For example,

- why the high lower-filter cutoff of 1 Hz, and shouldn't it be acknowledged that this removes from the EEG any sustained, iteratively updated representation that evolves with learning across trials?

Response 6: We deeply thank you for your comments about our EEG analysis. The 1Hz high-pass filter may indeed filter out some useful information. We chose a 1Hz high-pass filter to filter out most of the noise and prevent the noise from affecting our results analysis. Additionally, there are also many decision-related works that have applied 1Hz high-pass filtering in EEG data preprocessing (Yau et al., 2021; Cortes et al., 2021; Wischnewski et al., 2022; Schutte et al., 2017; Mennella et al., 2020; Giustiniani et al., 2020).

Yau, Y., Hinault, T., Taylor, M., Cisek, P., Fellows, L. K., & Dagher, A. (2021). Evidence and urgency related EEG signals during dynamic decision-making in humans. Journal of Neuroscience, 41(26), 5711-5722.

Cortes, P. M., García-Hernández, J. P., Iribe-Burgos, F. A., Hernández-González, M., Sotelo-Tapia, C., & Guevara, M. A. (2021). Temporal division of the decision-making process: An EEG study. Brain Research, 1769, 147592.

Wischnewski, M., & Compen, B. (2022). Effects of theta transcranial alternating current stimulation (tACS) on exploration and exploitation during uncertain decision-making. Behavioural Brain Research, 426, 113840.

Schutte, I., Kenemans, J. L., & Schutter, D. J. (2017). Resting-state theta/beta EEG ratio is associated with reward-and punishment-related reversal learning. Cognitive, Affective, & Behavioral Neuroscience, 17, 754-763.

Mennella, R., Vilarem, E., & Grèzes, J. (2020). Rapid approach-avoidance responses to emotional displays reflect value-based decisions: Neural evidence from an EEG study. NeuroImage, 222, 117253.

Giustiniani, J., Nicolier, M., Teti Mayer, J., Chabin, T., Masse, C., Galmès, N., ... & Gabriel, D. (2020). Behavioral and neural arguments of motivational influence on decision making during uncertainty. Frontiers in Neuroscience, 14, 583.

Comment 7:

- Since the EEG analysis was done using an array of free-energy-related variables in a regression, was multicollinearity checked between these variables?

Response 7: We deeply thank you for your comments about our regression. Indeed, we didn't specify our regression formula in the main text. We conducted regression on one variable each time, so there was no need for a multicollinearity check. We have now added the relevant content in the Results section (“EEG results at source level” section, line 337-340):

“The linear regression was run by the "mne.stats.linear regression" function in the MNE package (Activity ~ Regressor + Intercept). Activity is the activity amplitude of the EEG signal in the source space and regressor is one of the regressors that we mentioned (e.g., expected free energy, the value of reducing ambiguity, etc.).”

Comment 8:

- In the initial comparison of the first/second half, why just 5 clusters of electrodes, and why these particular clusters?

Response 8: We deeply thank you for your comments about our sensor-level analysis. These five clusters are relatively common scalp EEG regions to analyze (left frontal, right frontal, central, left parietal, and right parietal), and we referred previous work analyzed these five clusters of electrodes (Laufs et al., 2006; Ray et al., 1985; Cole et al., 1985). In addition, our work pays more attention to the analysis in source space, exploring the corresponding functions of specific brain regions based on active inference models.

Laufs, H., Holt, J. L., Elfont, R., Krams, M., Paul, J. S., Krakow, K., & Kleinschmidt, A. (2006). Where the BOLD signal goes when alpha EEG leaves. Neuroimage, 31(4), 1408-1418.

Ray, W. J., & Cole, H. W. (1985). EEG activity during cognitive processing: influence of attentional factors. International Journal of Psychophysiology, 3(1), 43-48.

Cole, H. W., & Ray, W. J. (1985). EEG correlates of emotional tasks related to attentional demands. International Journal of Psychophysiology, 3(1), 33-41.

Comment 9:

How many different variables are systematically different in the first vs second half, and how do you rule out less interesting time-on-task effects such as engagement or alertness? In what time windows are these amplitudes being measured?

Response 9 (and the Response for Weaknesses 11): There were no systematic differences between the first half and the second half of the trials, with the only difference being the participants' experience. In the second half, participants had a better understanding of the reward distribution of the task (less ambiguity). The simulation results can well describe these.

Author response image 7.

As shown in Figure (a), agents can only learn about the hidden state of the environment ("Context 1" (green) or "Context 2" (orange)) by choosing the "Cue" option. If agents choose the "Stay" option, they will not be able to know the hidden state of the environment (purple). The risk of agents is only related to wh

ether they choose the "Cue" option, not the number of rounds. Figure (b) shows the Safe-Risky choices of agents, and Figure (e) is the reward prediction of agents for the "Risky" path in "Context 1" and "Context 2". We can see that agents update the expected reward and reduce ambiguity by sampling the "Risky" path. The ambiguity of agents is not related to the "Cue" option, but to the number of times they sample the "Risky" path (rounds).

In our choosing stages, participants were required to think about their choices for the first two seconds (during which they could not press buttons). Then, they were asked to make their choices (press buttons) within the next two seconds. This setup effectively kept participants' attention focused on the task. And the two second during the “Second choice” stage when participants decide which option to choose (they cannot press buttons) are measured for the analysis of the sensor-level results.

Comment 10:

In the comparison of asked and not-asked trials, what trial stage and time window is being measured?

Response 10: We have added relevant descriptions in the main text. The two second during the “Second choice” stage when participants decide which option to choose (they cannot press buttons) are measured for the analysis of the sensor-level results.

Author response image 8.

Comment 11:

Again, how many different variables, of the many estimated per trial in the active inference model, are different in the asked and not-asked trials, and how can you know which of these differences is the one reflected in the EEG effects?

Response 11: The difference between asked trials and not-asked trials lies only in whether participants know the specific context of the risky path (the level of risk for the participants). A simple comparison indeed cannot tell us which of these differences is reflected in the EEG effects. Therefore, we subsequently conducted model-based regression analysis in the source space.

Comment 12:

The authors choose to interpret that on not-asked trials the subjects are more uncertain because the cue doesn't give them the context, but you could equally argue that they don't ask because they are more certain of the possible hidden states.

Response 12: Our task design involves randomly varying the context of the risky path. Only by choosing to inquire can participants learn about the context. Participants can only become increasingly certain about the reward distribution of different contexts of the risky path, but cannot determine which specific context it is. Here are the instructions for the task that we will tell the participants (line 226-231).

"You are on a quest for apples in a forest, beginning with 5 apples. You encounter two paths: 1) The left path offers a fixed yield of 6 apples per excursion. 2) The right path offers a probabilistic reward of 0/3/6/9/12 apples, and it has two distinct contexts, labeled "Context 1" and "Context 2," each with a different reward distribution. Note that the context associated with the right path will randomly change in each trial. Before selecting a path, a ranger will provide information about the context of the right path ("Context 1" or "Context 2") in exchange for an apple. The more apples you collect, the greater your monetary reward will be."

Comment 13:

- The EEG regressors are not fully explained. For example, an "active learning" regressor is listed as one of the 4 at the beginning of section 3.3, but it is the first mention of this term in the paper and the term does not arise once in the methods.

Response 13: We have accordingly revised the relevant content in the main text (as in Eq.8). Our regressors now include expected free energy, the value of reducing ambiguity, the value of avoiding risk, extrinsic value, prediction error, (the degree of) ambiguity, reducing ambiguity, and avoiding risk.

Comment 14:

- In general, it is not clear how one can know that the EEG results reflect that the brain is purposefully encoding these very parameters while implementing this very mechanism, and not other, possibly simpler, factors that correlate with them since there is no engagement with such potential confounds or alternative models. For example, a model-free reinforcement learning model is fit to behaviour for comparison. Why not the EEG?

Response 14: We deeply thank you for your comments. Due to factors such as time and effort, and because the active inference model best fits the behavioral data of the participants, we did not use other models to analyze the EEG data. At both the sensor and source level, we observed the EEG signal and brain regions that can encode different levels of uncertainties (risk and ambiguity). The brain's uncertainty driven exploration mechanism cannot be explained solely by a simple model-free reinforcement learning approach.

Recommendations for the authors:

Response: We have made point-to-point revisions according to the reviewer's recommendations, and as these revisions are relatively minor, we have only responded to the longer recommendations here.

Reviewer #1 (Recommendations For The Authors)

I enjoyed reading this sophisticated study of decision-making. I thought your implementation of active inference and the subsequent fitting to choice behaviour - and study of the neuronal (EEG) correlates - was impressive. As noted in my comments on strengths and weaknesses, some parts of your manuscript with difficult to read because of slight collapses in grammar and an inconsistent use of terms when referring to the mathematical quantities. In addition to the paragraphs I have suggested, I would recommend the following minor revisions to your text. In addition, you will have to fill in some of the details that were missing from the current version of the manuscript. For example:

Recommendation 1:

Which RL model did you use to fit the behavioural data? What were its free parameters?

Response 1: We have now added information related to the comparison models in the behavioral results and supplementary materials. We applied both simple model-free reinforcement learning and model-based reinforcement learning. The free parameters for the model-free reinforcement learning model are the learning rate α and the temperature parameter γ, while the free parameters for the model-based approach are the learning rate α, the temperature parameter γ, and the prior.

Recommendation 2:

When you talk about neuronal activity in the final analyses (of time-dependent correlations) what was used to measure the neuronal activity? Was this global power over frequencies? Was it at a particular frequency band? Was it the maximum amplitude within some small window et cetera? In other words, you need to provide the details of your analysis that would enable somebody to reproduce your study at a certain level of detail.

Response 2: In the final analyses, we used the activity amplitude at each point in the source space for our analysis. Previously, we had planned to make our data and models available on GitHub to facilitate easier replication of our work.

Reviewer #3 (Recommendations For The Authors)

Recommendation 1:

It might help to explain the complex concepts up front, to use the concrete example of the task itself - presumably, it was designed so that the crucial elements of the active inference framework come to the fore. One could use hypothetical choice patterns in this task to exemplify different factors such as expected free energy and unexpected uncertainty at work. It would also be illuminating to explain why behaviour on this task is fit better by the active inference model than a model-free reinforcement learning model.

Response 1: Thank you for your suggestions. We have given clearer explanations to the three terms in the active inference formula: the value of reducing ambiguity, the value of avoiding risk, and the extrinsic value (Eq.8), which makes it easier for readers to understand active inference.

In addition, we can simply view active inference as a computational model similar to model-based reinforcement learning, where the expected free energy represents a subjective value, without needing to understand its underlying computational principles or neurobiological background. In our discussion, we have argued why the active inference model fits the participants' behavior better than our reinforcement learning model, as the active inference model has an inherent exploration mechanism that is consistent with humans, who instinctively want to reduce environmental uncertainty (line 435-442).

“Active inference offers a superior exploration mechanism compared with basic model-free reinforcement learning  (Figure 4 (c)). Since traditional reinforcement learning models determine their policies solely on the state, this setting leads to difficulty in extracting temporal information (Laskin et al., 2020) and increases the likelihood of entrapment within local minima. In contrast, the policies in active inference are determined by both time and state. This dependence on time (Wang et al., 2016) enables policies to adapt efficiently, such as emphasizing exploration in the initial stages and exploitation later on. Moreover, this mechanism prompts more exploratory behavior in instances of state ambiguity. A further advantage of active inference lies in its adaptability to different task environments (Friston et al., 2017). It can configure different generative models to address distinct tasks, and compute varied forms of free energy and expected free energy.”

Laskin, M., Lee, K., Stooke, A., Pinto, L., Abbeel, P., & Srinivas, A. (2020). Reinforcement learning with augmented data. Advances in neural information processing systems, 33, 19884-19895.

Wang, J. X., Kurth-Nelson, Z., Tirumala, D., Soyer, H., Leibo, J. Z., Munos, R., ... & Botvinick, M. (2016). Learning to reinforcement learn. arXiv preprint arXiv:1611.05763.

Friston, K., FitzGerald, T., Rigoli, F., Schwartenbeck, P., & Pezzulo, G. (2017). Active inference: a process theory. Neural computation, 29(1), 1-49.

Recommendation 2:

Figure 1A provides a key example of the lack of effort to help the reader understand. It suggests the possibility of a concrete example but falls short of providing one. From the caption and text, applied to the figure, I gather that by choosing either to run or to raise one's arms, one can control whether it is daytime or nighttime. This is clearly wrong but it is what I am led to think by the paper.

Response 2: Thank you for your suggestion, which we had not considered before. In this figure, we aim to illustrate that "the agent receives observations and optimizes his cognitive model by minimizing variational free energy → the agent makes the optimal action by minimizing expected free energy → the action changes the environment → the environment generates new observations for the agent." We have now modified the image to be simpler to prevent any possible confusion for readers. Correspondingly, we removed the figure of a person raising their hand and the shadowed house in Figure a.

Author response image 9.

Recommendation 3:

I recommend an overhaul in the labelling and methodological explanations for consistency and full reporting. For example, line 73 says sensory input is 's' and the cognitive model is 'q(s),' and the cause of the sensory input is 'p(s|o)' but on the very next line, the cognitive model is 'p(s|o)' and the causes of sensory input are 'q(s).' How this sensory input s relates to 'observations' or 'o' is unclear, and meanwhile, capital S is the set of environmental states. P seems to refer to the generative distribution, but it also means probability.

Response 3: Thank you for your advice. Now we have revised the corresponding labeling and methodological explanations in our work to make them consistent. However, we are not sure how to make a good modification to P here. In many works, P can refer to a certain probability distribution or some specific probabilities.

Recommendation 4:

Even the conception of a "policy" is unclear (Figure 2B). They list 4 possible policies, which are simply the 4 possible sequences of steps, stay-safe, cue-risky, etc, but with no contingencies in them. Surely a complete policy that lists 'cue' as the first step would entail a specification of how they would choose the safe or risky option BASED on the information in that cue

Response 4: Thank you for your suggestion. In active inference, a policy actually corresponds to a sequence of actions. The policy of "first choosing 'Cue' and then making the next decision based on specific information" differs from the meaning of policy in active inference.

Recommendation 5:

I assume that the heavy high pass filtering of the EEG (1 Hz) is to avoid having to baseline-correct the epochs (of which there is no mention), but the authors should directly acknowledge that this eradicates any component of decision formation that may evolve in any way gradually within or across the stages of the trial. To take an extreme example, as Figure 3E shows, the expected rewards for the risky path evolve slowly over the course of 60 trials. The filter would eliminate this.

Response 5: Thank you for your suggestion. The heavy high pass filtering of the EEG (1 Hz) is to minimize the noise in the EEG data as much as possible.

Recommendation 6:

There is no mention of the regression itself in the Methods section - the section is incomplete.

Response 6: Thank you for your suggestion. We have now added the relevant content in the Results section (EEG results at source level, line 337-340):

“The linear regression was run by the "mne.stats.linear regression" function in the MNE package (Activity ∼ Regressor + Intercept, Activity is the activity amplitude of the EEG signal in the source space and regressor is one of the regressors that we mentioned).”

Recommendation 7:

On Lines 260-270 the same results are given twice.

Response 7: Thank you for your suggestion. We have now deleted redundant content.

Recommendation 8:

Frequency bands are displayed in Figure 5 but there is no mention of those in the Methods. In Figure 5b Theta in the 2nd half is compared to Delta in the 1st half- is this an error?

Response 8: Thank you for your suggestion. It indeed was an error (they should all be Theta) and now we have corrected it.

Author response image 10.

Author response:

The following is the authors’ response to the original reviews.

Reviewer #1:

Summary:

In this study, Nishi et al. claim that the ratio of long-term hematopoietic stem cell (LT-HSC) versus short-term HSC (ST-HSC) determines the lineage output of HSCs and reduced ratio of ST-HSC in aged mice causes myeloid-biased hematopoiesis. The authors used Hoxb5 reporter mice to isolate LT-HSC and ST-HSC and performed molecular analyses and transplantation assays to support their arguments. How the hematopoietic system becomes myeloid-biased upon aging is an important question with many implications in the disease context as well. However, their study is descriptive with remaining questions.

Weaknesses:

Comment #1-1: The authors may need conceptual re-framing of their main argument because whether the ST-HSCs used in this study are functionally indeed short-term "HSCs" is questionable. The data presented in this study and their immunophenotypic definition of ST-HSCs (Lineage negative/Sca-1+/c-Kit+/Flk2-/CD34-/CD150+/Hoxb5-) suggest that authors may find hematopoietic stem cell-like lymphoid progenitors as previously shown for megakaryocyte lineage (Haas et al., Cell stem cell. 2015) or, as the authors briefly mentioned in the discussion, Hoxb5- HSCs could be lymphoid-biased HSCs.

The authors disputed the idea that Hoxb5- HSCs as lymphoid-biased HSCs based on their previous 4 weeks post-transplantation data (Chen et al., 2016). However, they overlooked the possibility of myeloid reprogramming of lymphoid-biased population during regenerative conditions (Pietras et al., Cell stem cell., 2015). In other words, early post-transplant STHSCs (Hoxb5- HSCs) can be seen as lacking the phenotypic lymphoid-biased HSCs.

Thinking of their ST-HSCs as hematopoietic stem cell-like lymphoid progenitors or lymphoidbiased HSCs makes more sense conceptually as well.

Response #1-1: We appreciate this important suggestion and recognize the significance of the debate on whether Hoxb5- HSCs are ST-HSCs or lymphoid-biased HSCs.

HSCs are defined by their ability to retain hematopoietic potential after a secondary transplantation1-2. If Hoxb5- HSCs were indeed lymphoid-biased HSCs, they would exhibit predominantly lymphoid hematopoiesis even after secondary transplantation. However, functional experiments demonstrate that these cells lose their hematopoietic output after secondary transplantation3 (see Fig. 2 in this paper). Based on the established definition of HSCs in this filed, it is appropriate to classify Hoxb5- HSCs as ST-HSCs rather than lymphoid-biased HSCs.

Additionally, it has been reported that myeloid reprogramming may occur in the early posttransplant period, around 2-4 weeks after transplantation, even in lymphoid-biased populations within the MPP fraction, due to high inflammatory conditions4. However, when considering the post-transplant hematopoiesis of Hoxb5- HSC fractions as ST-HSCs, they exhibit almost the same myeloid hematopoietic potential as LT-HSCs not only during the early 4 weeks after transplantation but also at 8 weeks post-transplantation3, when the acute inflammatory response has largely subsided. Therefore, it is difficult to attribute the myeloid production by ST-HSCs post-transplant solely to myeloid reprogramming.

References

(1) Morrison, S. J. & Weissman, I. L. The long-term repopulating subset of hematopoietic stem cells is deterministic and isolatable by phenotype. Immunity 1, 661–673 (1994).

(2) Challen, G. A., Boles, N., Lin, K. K. Y. & Goodell, M. A. Mouse hematopoietic stem cell identification and analysis. Cytom. Part A 75, 14–24 (2009).

(3) Chen, J. Y. et al. Hoxb5 marks long-term haematopoietic stem cells and reveals a homogenous perivascular niche. Nature 530, 223–227 (2016).

(4) Pietras, E. M. et al. Functionally Distinct Subsets of Lineage-Biased Multipotent Progenitors Control Blood Production in Normal and Regenerative Conditions. Cell Stem Cell 17, 35–46 (2015).

Comment #1-2: ST-HSCs come from LT-HSCs and further differentiate into lineage-biased multipotent progenitor (MPP) populations including myeloid-biased MPP2 and MPP3. Based on the authors' claim, LT-HSCs (Hoxb5- HSCs) have no lineage bias even in aged mice. Then these LT-HSCs make ST-HSCs, which produce mostly memory T cells. These memory T cell-producing ST-HSCs then produce MPPs including myeloid-biased MPP2 and MPP3.

This differentiation trajectory is hard to accept. If we think Hoxb5- HSCs (ST-HSCs by authors) as a sub-population of immunophenotypic HSCs with lymphoid lineage bias or hematopoietic stem cell-like lymphoid progenitors, the differentiation trajectory has no flaw.

Response #1-2: Thank you for this comment, and we apologize for the misunderstanding regarding the predominance of memory T cells in ST-HSCs after transplantation. 

Our data show that ST-HSCs are not biased HSCs that predominantly produce memory T cells, but rather, ST-HSCs are multipotent hematopoietic cells. ST-HSCs lose their ability to self-renew within a short period, resulting in the cessation of ST-HSC-derived hematopoiesis. As a result, myeloid lineage with a short half-life disappears from the peripheral blood, and memory lymphocytes with a long half-life remain (see Figure 5 in this paper). 

Comment #1-3: Authors' experimental designs have some caveats to support their claims. Authors claimed that aged LT-HSCs have no myeloid-biased clone expansion using transplantation assays. In these experiments, authors used 10 HSCs and young mice as recipients. Given the huge expansion of old HSC by number and known heterogeneity in immunophenotypically defined HSC populations, it is questionable how 10 out of so many old HSCs can faithfully represent the old HSC population. The Hoxb5+ old HSC primary and secondary recipient mice data (Figure 2C and D) support this concern. In addition, they only used young recipients. Considering the importance of the inflammatory aged niche in the myeloid-biased lineage output, transplanting young vs old LT-HSCs into aged mice will complete the whole picture.

Response #1-3: We appreciate the reviewer for the comments. We acknowledge that using ten HSCs may not capture the heterogeneity of aging HSCs.

However, although most of our experiments have used a small number of transplanted cells (e.g., 10 cells), we have conducted functional experiments across Figures 2, 3, 5, 6, S3, and S6, totaling n = 126, equivalent to over 1260 cells. Previous studies have reported that myeloid-biased HSCs constitute more than 50% of the aged HSC population1-2. If myeloidbiased HSCs increase with age, they should be detectable in our experiments. Our functional experiments have consistently shown that Hoxb5+ HSCs exhibit unchanged lineage output throughout life. In contrast, the data presented in this paper indicate that changes in the ratio of LT-HSCs and ST-HSCs may contribute to myeloid-biased hematopoiesis.

We believe that transplanting aged HSCs into aged recipient mice is crucial to analyzing not only the differentiation potential of aged HSCs but also the changes in their engraftment and self-renewal abilities. We aim to clarify further findings through these experiments in the future.

References

(1) Dykstra B, Olthof S, Schreuder J, Ritsema M, Haan G De. Clonal analysis reveals multiple functional defects of aged murine hematopoietic stem cells. J Exp Med. 2011 Dec 19;208(13):2691–703. 

(2) Yamamoto R, Wilkinson AC, Ooehara J, Lan X, Lai CY, Nakauchi Y, et al. LargeScale Clonal Analysis Resolves Aging of the Mouse Hematopoietic Stem Cell Compartment. Cell Stem Cell [Internet]. 2018;22(4):600-607.e4. Available from: https://doi.org/10.1016/j.stem.2018.03.013

Comment #1-4: The authors' molecular data analyses need more rigor with unbiased approaches. They claimed that neither aged LT-HSCs nor aged ST-HSCs exhibited myeloid or lymphoid gene set enrichment but aged bulk HSCs, which are just a sum of LT-HSCs and ST-HSCs by their gating scheme (Figure 4A), showed the "tendency" of enrichment of myeloid-related genes based on the selected gene set (Figure 4D). Although the proportion of ST-HSCs is reduced in bulk HSCs upon aging, since ST-HSCs do not exhibit lymphoid gene set enrichment based on their data, it is hard to understand how aged bulk HSCs have more myeloid gene set enrichment compared to young bulk HSCs. This bulk HSC data rather suggests that there could be a trend toward certain lineage bias (although not significant) in aged LT-HSCs or ST-HSCs. The authors need to verify the molecular lineage priming of LT-HSCs and ST-HSCs using another comprehensive dataset.

Response #1-4: Thank you for pointing out that neither aged LT-HSCs nor aged ST-HSCs exhibited myeloid

or lymphoid gene set enrichment, although aged bulk HSCs showed a tendency towards enrichment of myeloid-related genes.

The actual GSEA result had an FDR > 0.05. Therefore, we cannot claim that bulk HSCs showed significant enrichment of myeloid-related genes with age. Consequently, we have revised the following sentences:

[P11, L251] Neither aged LT-HSCs nor aged ST-HSCs exhibited myeloid/lymphoid gene set enrichment, while shared myeloid-related genes tended to be enriched in aged bulk-HSCs, although this enrichment was not statistically significant (Fig. 4, F and G).

In addition to the above, we also found that the GSEA results differ among myeloid gene sets (Fig. 4, D-F; Fig. 4S, C-D). These findings suggest that discussing lineage bias in HSCs using GSEA is challenging. We believe that functional experimental data is crucial. From our functional experiments, when the ratio of LT-HSC to ST-HSC was reconstituted to match the ratio in young Bulk-HSCs (LT= 2:8) or aged bulk-HSCs (LT= 5:5), myeloid-biased hematopoiesis was observed with the aged bulk-HSC ratio. Based on this data, the authors concluded that age-related changes in the ratio between LT-HSCs and ST-HSCs in bulkHSCs cause myeloid-biased hematopoiesis rather than an increase in myeloid gene expression in the aged bulk-HSCs.

Comment #1-5: Some data are too weak to fully support their claims. The authors claimed that age-associated extramedullary changes are the main driver of myeloid-biased hematopoiesis based on no major differences in progenitor populations upon transplantation of 10 young HSCs into young or old recipient mice (Figure 7F) and relatively low donor-derived cells in thymus and spleen in aged recipient mice (Figure 7G-J). However, they used selected mice to calculate the progenitor populations in recipient mice (8 out of 17 from young recipients denoted by * and 8 out of 10 from aged recipients denoted by * in Figure 7C). In addition, they calculated the progenitor populations as frequency in c-kit positive cells. Given that they transplanted 10 LT-HSCs into "sub-lethally" irradiated mice and 8.7 Gy irradiation can have different effects on bone marrow clearance in young vs old mice, it is not clear whether this data is reliable enough to support their claims. The same concern applies to the data Figure 7G-J. Authors need to provide alternative data to support their claims.

Response #1-5: Thank you for useful comments. Our claim regarding Fig. 7 is that age-associated extramedullary changes are merely additional drivers for myeloid-biased hematopoiesis are not the main drivers. But we will address the issues pointed out.

Regarding the reason for analyzing the asterisk mice

We performed two independent experiments for Fig. 7. In the first experiment, we planned to analyze the BM of recipients 16 weeks after transplantation. However, as shown in Fig. 7B, many of the aged mice died before 16 weeks. Therefore, we decided to examine the BM of the recipient mice at 12 weeks in the second experiment. Below are the peripheral blood results 11-12 weeks after transplantation for the mice used in the second experiment.

Author response image 1.

For the second experiment, we analyzed the BM of all eight all eight aged recipients. Then, we selected the same number of young recipients for analysis to ensure that the donor myeloid output would be comparable to that of the entire young group. Indeed, the donor myeloid lineage output of the selected mice was 28.1 ± 22.9%, closely matching the 23.5 ± 23.3% (p = 0.68) observed in the entire young recipient population. 

That being said, as the reviewer pointed out, it is considerable that the BM, thymus, and spleen of all mice were not analyzed. Hence, we have added the following sentences:

[P14, L327] We performed BM analysis for the mice denoted by † in Figure 7C because many of the aged mice had died before the analysis.

[P15, L338] The thymus and spleen analyses were also performed on the mice denoted by † in Figure 7C.

Regarding the reason for 8.7 Gy.

Thank you for your question about whether 8.7 Gy is myeloablative. In our previous report1, we demonstrated that none of the mice subjected to pre-treatment with 8.7 Gy could survive when non-LKS cells were transplanted, suggesting that 8.7 Gy is enough to be myeloablative with the radiation equipment at our facility.

Author response image 2.

Reference

(1)  Nishi K, Sakamaki T, Sadaoka K, Fujii M, Takaori-Kondo A, Chen JY, et al. Identification of the minimum requirements for successful haematopoietic stem cell transplantation. Br J Haematol. 2022;196(3):711–23. 

Regarding the normalization of c-Kit in Figure 7F.  

Firstly, as shown in Supplemental Figures S1B and S1C, we analyze the upstream (HSC, MPP, Flk2+) and downstream (CLP, MEP, CMP, GMP) fractions in different panels. Therefore, normalization is required to assess the differentiation of HSCs from upstream to downstream. Additionally, the reason for normalizing by c-Kit+ is that the bone marrow analysis was performed after enrichment using the Anti-c-Kit antibody for both upstream and downstream fractions. Based on this, we calculated the progenitor populations as a frequency within the c-Kit positive cells.

Next, the results of normalizing the whole bone marrow cells (live cells) are shown below. 

Author response image 3.

Similar to the results of normalizing c-Kit+ cells, myeloid progenitors remained unchanged, including a statistically significant decrease in CMP in aged mice. Additionally, there were no significant differences in CLP. In conclusion, we obtained similar results between the normalization with c-Kit and the normalization with whole bone marrow cells (live cells).

However, as the reviewer pointed out, it is necessary to explain the reason for normalization with c-Kit. Therefore, we will add the following description.

[P21, L502] For the combined analysis of the upstream (HSC, MPP, Flk2+) and downstream (CLP, MEP, CMP, GMP) fractions in Figures 1B and 7F, we normalized by c-Kit+ cells because we performed a c-Kit enrichment for the bone marrow analysis.

Reviewer #2:

Summary:  

Nishi et al, investigate the well-known and previously described phenomenon of ageassociated myeloid-biased hematopoiesis. Using a previously established HoxB5mCherry mouse model, they used HoxB5+ and HoxB5- HSCs to discriminate cells with long-term (LTHSCs) and short-term (ST-HSCs) reconstitution potential and compared these populations to immunophenotypically defined 'bulk HSCs' that consists of a mixture of LT-HSC and STHSCs. They then isolated these HSC populations from young and aged mice to test their function and myeloid bias in non-competitive and competitive transplants into young and aged recipients. Based on quantification of hematopoietic cell frequencies in the bone marrow, peripheral blood, and in some experiments the spleen and thymus, the authors argue against the currently held belief that myeloid-biased HSCs expand with age. 

Comment #2-1: While aspects of their work are fascinating and might have merit, several issues weaken the overall strength of the arguments and interpretation. Multiple experiments were done with a very low number of recipient mice, showed very large standard deviations, and had no statistically detectable difference between experimental groups. While the authors conclude that these experimental groups are not different, the displayed results seem too variable to conclude anything with certainty. The sensitivity of the performed experiments (e.g. Figure 3; Figure 6C, D) is too low to detect even reasonably strong differences between experimental groups and is thus inadequate to support the author's claims. This weakness of the study is not acknowledged in the text and is also not discussed. To support their conclusions the authors need to provide higher n-numbers and provide a detailed power analysis of the transplants in the methods section.

Response #2-1: Thank you for your important remarks. The power analysis for this experiment shows that power = 0.319, suggesting that more number may be needed. On the other hand, our method for determining the sample size in Figure 3 is as follows:

(1) First, we checked whether myeloid biased change is detected in the bulk-HSC fraction (Figure S3). The results showed that the difference in myeloid output at 16 weeks after transplantation was statistically significant (young vs. aged = 7.2 ± 8.9 vs. 42.1 ± 35.5%, p = 0.01), even though n = 10.

(2) Next, myeloid biased HSCs have been reported to be a fraction with high self-renewal ability (2004, Blood). If myeloid biased HSCs increase with aging, the increase in myeloid biased HSCs in LT-HSC fraction would be detected with higher sensitivity than in the bulk-HSC fraction used in Figure S3.

(3) However, there was no difference not only in p-values but also in the mean itself, young vs aged = 51.4±31.5% vs 47.4±39.0%, p = 0.82, even though n = 8 in Figure 3. Since there was no difference in the mean itself, it is highly likely that no difference will be detected even if n is further increased.

Regarding Figure 6, we obtained a statistically significant difference and consider the sample size to be sufficient. 

In addition, we have performed various functional experiments (Figures 2, 5, 6 and S6), and have obtained consistent results that expansion of myeloid biased HSCs does not occur with aging in Hoxb5+HSCs fraction. Based on the above, we conclude that the LT-HSC fraction does not differ in myeloid differentiation potential with aging.

Comment #2-2: As the authors attempt to challenge the current model of the age-associated expansion of myeloid-biased HSCs (which has been observed and reproduced by many different groups), ideally additional strong evidence in the form of single-cell transplants is provided.

Response #2-2: Thank you for the comments. As the reviewer pointed out, we hope we could reconfirm our results using single-cell level technology in the future.

On the other hand, we have reported that the ratio of myeloid to lymphoid cells in the peripheral blood changes when the number of HSCs transplanted, or the number of supporting cells transplanted with HSCs, is varied1-2. Therefore, single-cell transplant data need to be interpreted very carefully to determine differentiation potential.

From this viewpoint, future experiments will combine the Hoxb5 reporter system with a lineage tracing system that can track HSCs at the single-cell level over time. This approach will investigate changes in the self-renewal capacity of individual HSCs and their subsequent differentiation into progenitor cells and peripheral blood cells. We have reflected this comment by adding the following sentences in the manuscript.

[P19, L451] In contrast, our findings should be considered in light of some limitations. In this report, we primarily performed ten to twenty cell transplantation assays. Therefore, the current theory should be revalidated using single-cell technology with lineage tracing system3-4. This approach will investigate changes in the self-renewal capacity of individual HSCs and their subsequent differentiation into progenitor cells and peripheral blood cells. 

References

(1) Nishi K, Sakamaki T, Sadaoka K, Fujii M, Takaori-Kondo A, Chen JY, et al. Identification of the minimum requirements for successful haematopoietic stem cell transplantation. Br J Haematol. 2022;196(3):711–23. 

(2) Sakamaki T, Kao KS, Nishi K, Chen JY, Sadaoka K, Fujii M, et al. Hoxb5 defines the heterogeneity of self-renewal capacity in the hematopoietic stem cell compartment. Biochem Biophys Res Commun [Internet]. 2021;539:34–41. Available from: https://doi.org/10.1016/j.bbrc.2020.12.077

(3) Yamamoto R, Wilkinson AC, Ooehara J, Lan X, Lai CY, Nakauchi Y, et al. LargeScale Clonal Analysis Resolves Aging of the Mouse Hematopoietic Stem Cell Compartment. Cell Stem Cell [Internet]. 2018;22(4):600-607.e4. Available from: https://doi.org/10.1016/j.stem.2018.03.013

(4) Rodriguez-Fraticelli AE, Weinreb C, Wang SW, Migueles RP, Jankovic M, Usart M, et al. Single-cell lineage tracing unveils a role for TCF15 in haematopoiesis. Nature [Internet]. 2020;583(7817):585–9. Available from: http://dx.doi.org/10.1038/s41586-020-2503-6

Comment #2-3: It is also unclear why the authors believe that the observed reduction of ST-HSCs relative to LT-HSCs explains the myeloid-biased phenotype observed in the peripheral blood. This point seems counterintuitive and requires further explanation.

Response #2-3: Thank you for your comment. We apologize for the insufficient explanation. Our data, as shown in Figures 3 and 4, demonstrate that the differentiation potential of LT-HSCs remains unchanged with age. Therefore, rather than suggesting that an increase in LT-HSCs with a consistent differentiation capacity leads to myeloid-biased hematopoiesis, it seems more accurate to highlight that the relative decrease in the proportion of ST-HSCs, which remain in peripheral blood as lymphocytes, leads to a relative increase in myeloid cells in peripheral blood and thus causes myeloid-biased hematopoiesis.

However, if we focus on the increase in the ratio of LT-HSCs, it is also plausible to explain that “with aging, the proportion of LT-HSCs capable of long-term myeloid hematopoiesis increases. As a result, from 16 weeks after transplantation, the influence of LT-HSCs maintaining the long-term ability to produce myeloid cells becomes relatively more significant, leading to an increase in the ratio of myeloid cells in the peripheral blood and causing myeloid-biased hematopoiesis.”

Comment #2-4: Based on my understanding of the presented data, the authors argue that myeloid-biased HSCs do not exist, as<br /> a) they detect no difference between young/aged HSCs after transplant (mind low n-numbers and large std!); b) myeloid progenitors downstream of HSCs only show minor or no changes in frequency and c) aged LT-HSCs do not outperform young LT-HSC in myeloid output LT-HScs in competitive transplants (mind low n-numbers and large std!).

Response #2-4: We appreciate the comments. As mentioned above, we will correct the manuscript regarding the sample size.

Regarding the interpreting of the lack of increase in the percentage of myeloid progenitor cells in the bone marrow with age, it is instead possible that various confounding factors, such as differentiation shortcuts or changes in the microenviroment, are involved.

However, even when aged LT-HSCs and young LT-HSCs are transplanted into the same recipient mice, the timing of the appearance of different cell fractions in peripheral blood is similar (Figure 3 of this paper). Therefore, we have not obtained data suggesting that clear shortcuts exist in the differentiation process of aged HSCs into neutrophils or monocytes. Additionally, it is currently consensually accepted that myeloid cells, including neutrophils and monocytes, differentiate from GMPs1. Since there is no changes in the proportion of GMPs in the bone marrow with age, we concluded that the differentiation potential into myeloid cells remains consistent with aging.

Reference

(1) Akashi K and others, ‘A Clonogenic Common Myeloid Progenitor That Gives Rise to All Myeloid Lineages’, Nature, 404.6774 (2000), 193–97.

Strengths: 

The authors present an interesting observation and offer an alternative explanation of the origins of aged-associated myeloid-biased hematopoiesis. Their data regarding the role of the microenvironment in the spleen and thymus appears to be convincing. 

Weaknesses: 

Comment #2-5: "Then, we found that the myeloid lineage proportions from young and aged LT-HSCs were nearly comparable during the observation period after transplantation (Figure 3, B and C)."<br /> Given the large standard deviation and low n-numbers, the power of the analysis to detect differences between experimental groups is very low. Experimental groups with too large standard deviations (as displayed here) are difficult to interpret and might be inconclusive. The absence of clearly detectable differences between young and aged transplanted HSCs could thus simply be a false-negative result. The shown experimental results hence do not provide strong evidence for the author's interpretation of the data. The authors should add additional transplants and include a detailed power analysis to be able to detect differences between experimental groups with reasonable sensitivity.

Response #2-5: Thank you for providing these insights. Regarding the sample size, we have addressed this in Response #2-1.

Comment #2-6: Line 293: "Based on these findings, we concluded that myeloid-biased hematopoiesis observed following transplantation of aged HSCs was caused by a relative decrease in ST-HSC in the bulk-HSC compartment in aged mice rather than the selective expansion of myeloid-biased HSC clones."<br /> Couldn't that also be explained by an increase in myeloid-biased HSCs, as repeatedly reported and seen in the expansion of CD150+ HSCs? It is not intuitively clear why a reduction of ST-HSCs clones would lead to a myeloid bias. The author should try to explain more clearly where they believe the increased number of myeloid cells comes from. What is the source of myeloid cells if the authors believe they are not derived from the expanded population of myeloid-biased HSCs?

Response #2-6: Thank you for pointing this out. We apologize for the insufficient explanation. We will explain using Figure 8 from the paper.

First, our data show that LT-HSCs maintain their differentiation capacity with age, while ST-HSCs lose their self-renewal capacity earlier, so that only long-lived memory lymphocytes remain in the peripheral blood after the loss of self-renewal capacity in ST-HSCs (Figure 8, upper panel). In mouse bone marrow, the proportion of LT-HSCs increases with age, while the proportion of STHSCs relatively decreases (Figure 8, lower panel and Figure S5). 

Our data show that merely reproducing the ratio of LT-HSCs to ST-HSCs observed in aged mice using young LT-HSCs and ST-HSCs can replicate myeloid-biased hematopoiesis. This suggests that the increase in LT-HSC and the relative decrease in ST-HSC within the HSC compartment with aging are likely to contribute to myeloid-biased hematopoiesis.

As mentioned earlier, since the differentiation capacity of LT-HSCs remain unchaged with age, it seems more accurate to describe that the relative decrease in the proportion of STHSCs, which retain long-lived memory lymphocytes in peripheral blood, leads to a relative increase in myeloid cells in peripheral blood and thus causes myeloid-biased hematopoiesis.

However, focusing on the increase in the proportion of LT-HSCs, it is also possible to explain that “with aging, the proportion of LT-HSCs capable of long-term myeloid hematopoiesis increases. As a result, from 16 weeks after transplantation, the influence of LT-HSCs maintaining the long-term ability to produce myeloid cells becomes relatively more significant, leading to an increase in the ratio of myeloid cells in the peripheral blood and causing myeloid-biased hematopoiesis.”

Reviewer #3:

Summary:

In this manuscript, Nishi et al. propose a new model to explain the previously reported myeloid-biased hematopoiesis associated with aging. Traditionally, this phenotype has been explained by the expansion of myeloid-biased hematopoietic stem cell (HSC) clones during aging. Here, the authors question this idea and show how their Hoxb5 reporter model can discriminate long-term (LT) and short-term (ST) HSC and characterized their lineage output after transplant. From these analyses, the authors conclude that changes during aging in the LT/ST HSC proportion explain the myeloid bias observed. 

Although the topic is appropriate and the new model provides a new way to think about lineage-biased output observed in multiple hematopoietic contexts, some of the experimental design choices, as well as some of the conclusions drawn from the results could be substantially improved. Also, they do not propose any potential mechanism to explain this process, which reduces the potential impact and novelty of the study. Specific concerns are outlined below. 

Major 

Comment #3-1: As a general comment, there are experimental details that are either missing or not clear. The main one is related to transplantation assays. What is the irradiation dose? The Methods sections indicates "recipient mice were lethally irradiated with single doses of 8.7 or 9.1 Gy". The only experimental schematic indicating the irradiation dose is Figure 7A, which uses 8.7 Gy. Also, although there is not a "standard", 11 Gy split in two doses is typically considered lethal irradiation, while 9.5 Gy is considered sublethal.

Response #3-1: We agree with reviewer’s assessment about whether 8.7 Gy is myeloablative. To confirm this, it would typically be necessary to irradiate mice with different dose and observe if they do not survive. However, such an experiment is not ethically permissible at our facility. Instead, in our previous report1, we demonstrated that none of the mice subjected to pretreatment with 8.7 Gy could survive when non-LKS cells were transplanted, suggesting that

8.7 Gy is enough to be myeloablative with the radiation equipment at our facility.

Reference

(1) Nishi K, Sakamaki T, Sadaoka K, Fujii M, Takaori-Kondo A, Chen JY, et al. Identification of the minimum requirements for successful haematopoietic stem cell transplantation. Br J Haematol. 2022;196(3):711–23. 

Comment #3-2:  Is there any reason for these lower doses? Same question for giving a single dose and for performing irradiation a day before transplant. 

Response #3-2: We appreciate the reviewer for these important comments. Although the 8.7 Gy dose used at our facility is lower than in other reports, we selected this dose to maintain consistency with our previous experiments. For the same reason, we used a single irradiation, not split.  Regarding the timing of irradiation, the method section specifies that irradiation timing is 12-24 hours prior to transplantation. In most experiments, irradiation is performed at 12 hours. However, due to experimental progress, there were occasional instances where nearly 24 hours elapsed between irradiation and transplantation. We provide this information to ensure accuracy.

Comment #3-3: The manuscript would benefit from the inclusion of references to recent studies discussing hematopoietic biases and differentiation dynamics at a single-cell level (e.g., Yamamoto et. al 2018; Rodriguez-Fraticelli et al., 2020). Also, when discussing the discrepancy between studies claiming different biases within the HSC pool, the authors mentioned that Montecino-Rodriguez et al. 2019 showed preserved lymphoid potential with age. It would be good to acknowledge that this study used busulfan as the conditioning method instead of irradiation.

Response #3-3: We agree with this comment and have incorporated this suggestion into the manuscript

[P19, L451] In contrast, our findings should be considered in light of some limitations. In this report, we primarily performed ten to twenty cell transplantation assays. Therefore, the current theory should be revalidated using single-cell technology with lineage tracing system1-2. This approach will investigate changes in the self-renewal capacity of individual HSCs and their subsequent differentiation into progenitor cells and peripheral blood cells. Additionally, in this report we purified LT-HSCs by Hoxb5 reporter system. In contrast, various LT-HSC markers have been previously reported2-3.  Therefore, it is ideal to validate our findings using other LT-HSC makers.

[P16, L368] Other studies suggest that blockage of lymphoid hematopoiesis in aged mice results in myeloid-skewed hematopoiesis through alternative mechanisms. However, this result should be interpreted carefully, since Busulfan was used for myeloablative treatment in this study4.   

References

(1) Yamamoto R, Wilkinson AC, Ooehara J, Lan X, Lai CY, Nakauchi Y, et al. LargeScale Clonal Analysis Resolves Aging of the Mouse Hematopoietic Stem Cell Compartment. Cell Stem Cell [Internet]. 2018;22(4):600-607.e4. Available from: https://doi.org/10.1016/j.stem.2018.03.013

(2) Rodriguez-Fraticelli AE, Weinreb C, Wang SW, Migueles RP, Jankovic M, Usart M, et al. Single-cell lineage tracing unveils a role for TCF15 in haematopoiesis. Nature [Internet]. 2020;583(7817):585–9. Available from: http://dx.doi.org/10.1038/s41586-020-2503-6

(3) Sanjuan-Pla A, Macaulay IC, Jensen CT, Woll PS, Luis TC, Mead A, et al. Plateletbiased stem cells reside at the apex of the haematopoietic stem-cell hierarchy. Nature. 2013;502(7470):232–6. 

(4) Montecino-Rodriguez E, Kong Y, Casero D, Rouault A, Dorshkind K, Pioli PD. Lymphoid-Biased Hematopoietic Stem Cells Are Maintained with Age and Efficiently Generate Lymphoid Progeny. Stem Cell Reports. 2019 Mar 5;12(3):584–96. 

Comment #3-4: When representing the contribution to PB from transplanted cells, the authors show the % of each lineage within the donor-derived cells (Figures 3B-C, 5B, 6B-D, 7C-E, and S3 B-C). To have a better picture of total donor contribution, total PB and BM chimerism should be included for each transplantation assay. Also, for Figures 2C-D and Figures S2A-B, do the graphs represent 100% of the PB cells? Are there any radioresistant cells?

Response #3-4: Thank you for highlighting this point. Indeed, donor contribution to total peripheral blood (PB) is important information. We have included the donor contribution data for each figure above mentioned.

Author response image 4.

In Figure 2C-D and Figure S2A-B, the percentage of donor chimerism in PB was defined as the percentage of CD45.1-CD45.2+ cells among total CD45.1-CD45.2+ and CD45.1+CD45.2+ cells as described in method section.

Comment #3-5: For BM progenitor frequencies, the authors present the data as the frequency of cKit+ cells. This normalization might be misleading as changes in the proportion of cKit+ between the different experimental conditions could mask differences in these BM subpopulations. Representing this data as the frequency of BM single cells or as absolute numbers (e.g., per femur) would be valuable.

Response #3-5: We appreciate the reviewer's comment on this point. 

Firstly, as shown in Supplemental Figures S1B and S1C, we analyze the upstream (HSC, MPP, Flk2+) and downstream (CLP, MEP, CMP, GMP) fractions in different panels. Therefore, normalization is required to assess the differentiation of HSCs from upstream to downstream. Additionally, the reason for normalizing by c-Kit+ is that the bone marrow analysis was performed after enrichment using the Anti-c-Kit antibody for both upstream and downstream fractions. Based on this, we calculated the progenitor populations as a frequency within the c-Kit positive cells. Next, the results of normalizing the whole bone marrow cells (live cells) are shown in Author response image 2. 

Similar to the results of normalizing c-Kit+ cells, myeloid progenitors remained unchanged, including a statistically significant decrease in CMP in aged mice. Additionally, there were no significant differences in CLP. In conclusion, similar results were obtained between the normalization with c-Kit and the normalization with whole bone marrow cells (live cells).

However, as the reviewer pointed out, it is necessary to explain the reason for normalization with c-Kit. Therefore, we will add the following description.

[P21, L502] For the combined analysis of the upstream (HSC, MPP, Flk2+) and downstream (CLP, MEP, CMP, GMP) fractions in Figures 1B and 7F, we normalized by c-Kit+ cells because we performed a c-Kit enrichment for the bone marrow analysis.

Comment #3-6: Regarding Figure 1B, the authors argue that if myeloid-biased HSC clones increase with age, they should see increased frequency of all components of the myeloid differentiation pathway (CMP, GMP, MEP). This would imply that their results (no changes or reduction in these myeloid subpopulations) suggest the absence of myeloid-biased HSC clones expansion with age. This reviewer believes that differentiation dynamics within the hematopoietic hierarchy can be more complex than a cascade of sequential and compartmentalized events (e.g., accelerated differentiation at the CMP level could cause exhaustion of this compartment and explain its reduction with age and why GMP and MEP are unchanged) and these conclusions should be considered more carefully.

Response #3-6: We wish to thank the reviewer for this comment. We agree with that the differentiation pathway may not be a cascade of sequential events but could be influenced by various factors such as extrinsic factors.

In Figure 1B, we hypothesized that there may be other mechanisms causing myeloidbiased hematopoiesis besides the age-related increase in myeloid-biased HSCs, given that the percentage of myeloid progenitor cells in the bone marrow did not change with age. However, we do not discuss the presence or absence of myeloid-biased HSCs based on the data in Figure 1B. 

Our newly proposed theories—that the differentiation capacity of LT-HSCs remains unchanged with age and that age-related myeloid-biased hematopoiesis is due to changes in the ratio of LT-HSCs to ST-HSCs—are based on functional experiment results. As the reviewer pointed out, to discuss the presence or absence of myeloid-biased HSCs based on the data in Figure 1B, it is necessary to apply a system that can track HSC differentiation at single-cell level. The technology would clarify changes in the self-renewal capacity of individual HSCs and their differentiation into progenitor cells and peripheral blood cells. The authors believe that those single-cell technologies will be beneficial in understanding the differentiation of HSCs. Based on the above, the following statement has been added to the text.

[P19, L451] In contrast, our findings should be considered in light of some limitations. In this report, we primarily performed ten to twenty cell transplantation assays. Therefore, the current theory should be revalidated using single-cell technology with lineage tracing system1-2. This approach will investigate changes in the self-renewal capacity of individual HSCs and their subsequent differentiation into progenitor cells and peripheral blood cells. 

References

(1) Yamamoto R, Wilkinson AC, Ooehara J, Lan X, Lai CY, Nakauchi Y, et al. LargeScale Clonal Analysis Resolves Aging of the Mouse Hematopoietic Stem Cell Compartment. Cell Stem Cell [Internet]. 2018;22(4):600-607.e4. Available from: https://doi.org/10.1016/j.stem.2018.03.013

(2) Rodriguez-Fraticelli AE, Weinreb C, Wang SW, Migueles RP, Jankovic M, Usart M, et al. Single-cell lineage tracing unveils a role for TCF15 in haematopoiesis. Nature [Internet]. 2020;583(7817):585–9. Available from: http://dx.doi.org/10.1038/s41586-020-2503-6

Comment #3-7: Within the few recipients showing good donor engraftment in Figure 2C, there is a big proportion of T cells that are "amplified" upon secondary transplantation (Figure 2D). Is this expected?

Response #3-7: We wish to express our deep appreciation to the reviewer for insightful comment on this point. As the reviewers pointed out, in Figure 2D, a few recipients show a very high percentage of T cells. The authors had the same question and considered this phenomenon as follows:

(1) One reason for the very high percentage of T cells is that we used 1 x 107 whole bone marrow cells in the secondary transplantation. Consequently, the donor cells in the secondary transplantation contained more T-cell progenitor cells, leading to a greater increase in T cells compared to the primary transplantation.

(2) We also consider that this phenomenon may be influenced by the reduced selfrenewal capacity of aged LT-HSCs, resulting in decreased sustained production of myeloid cells in the secondary recipient mice. As a result, long-lived memory-type lymphocytes may preferentially remain in the peripheral blood, increasing the percentage of T cells in the secondary recipient mice.

We have discussed our hypothesis regarding this interesting phenomenon. To further clarify the characteristics of the increased T-cell count in the secondary recipient mice, we will analyze TCR clonality and diversity in the future.

Comment #3-8: Do the authors have any explanation for the high level of variability within the recipients of Hoxb5+ cells in Figure 2C?

Response #3-8: We appreciate the reviewer's comment on this point. As noted in our previous report, transplantation of a sufficient number of HSCs results in stable donor chimerism, whereas a small number of HSCs leads to increased variability in donor chimerism1. Additionally, other studies have observed high variability when fewer than 10 HSCs are transplanted2-3. Based on this evidence, we consider that the transplantation of a small number of cells (10 cells) is the primary cause of the high level of variability observed.

References

(1) Nishi K, Sakamaki T, Sadaoka K, Fujii M, Takaori-Kondo A, Chen JY, et al. Identification of the minimum requirements for successful haematopoietic stem cell transplantation. Br J Haematol. 2022;196(3):711–23. 

(2) Dykstra B, Olthof S, Schreuder J, Ritsema M, Haan G De. Clonal analysis reveals multiple functional defects of aged murine hematopoietic stem cells. J Exp Med. 2011 Dec 19;208(13):2691–703. 

(3) Yamamoto R, Wilkinson AC, Ooehara J, Lan X, Lai CY, Nakauchi Y, et al. LargeScale Clonal Analysis Resolves Aging of the Mouse Hematopoietic Stem Cell Compartment. Cell Stem Cell [Internet]. 2018;22(4):600-607.e4. Available from: https://doi.org/10.1016/j.stem.2018.03.013

Comment #3-9: Can the results from Figure 2E be interpreted as Hoxb5+ cells having a myeloid bias? (differences are more obvious/significant in neutrophils and monocytes).

Response #3-9: Thank you for your insightful comments. Firstly, we have not obtained any data indicating that young LT-HSCs are myeloid biased HSCs so far. Therefore, we classify young LT-HSCs as balanced HSCs1. Secondly, our current data demonstrate no significant difference in differentiation capacity between young and aged LT-HSCs (see Figure 3 in this paper). Based on these findings, we interpret that aged LT-HSCs are balanced HSCs, similar to young LT-HSCs.

Reference

(1)  Chen JY, Miyanishi M, Wang SK, Yamazaki S, Sinha R, Kao KS, et al. Hoxb5 marks long-term haematopoietic stem cells and reveals a homogenous perivascular niche. Nature. 2016 Feb 10;530(7589):223–7. 

Comment #3-10: Is Figure 2G considering all primary recipients or only the ones that were used for secondary transplants? The second option would be a fairer comparison.

Response #3-10: We appreciate the reviewer's comment on this point. We considered all primary recipients in Figure 2G to ensure a fair comparison, given the influence of various factors such as the radiosensitivity of individual recipient mice1. Comparing only the primary recipients used in the secondary transplantation would result in n = 3 (primary recipient) vs. n = 12 (secondary recipient). Including all primary recipients yields n = 11 vs. n = 12, providing a more balanced comparison. Therefore, we analyzed all primary recipient mice to ensure the reliability of our results.

Reference

(1) Duran-Struuck R, Dysko RC. Principles of bone marrow transplantation (BMT): providing optimal veterinary and husbandry care to irradiated mice in BMT studies. J Am Assoc Lab Anim Sci. 2009; 48:11–22

Comment #3-11: When discussing the transcriptional profile of young and aged HSCs, the authors claim that genes linked to myeloid differentiation remain unchanged in the LT-HSC fraction while there are significant changes in the ST-HSCs. However, 2 out of the 4 genes shown in Figure S4B show ratios higher than 1 in LT-HSCs.

Response #3-11: Thank you for highlighting this important point. As the reviewer pointed out, when we analyze the expression of myeloid-related genes, some genes are elevated in aged LT-HSCs compared to young LT-HSCs. However, the GSEA analysis using myeloid-related gene sets, which include several hundred genes, shows no significant difference between young and aged LT-HSCs (see Figure S4C in this paper). Furthermore, functional experiments using the co-transplantation system show no difference in differentiation capacity between young and aged LT-HSCs (see Figure 3 in this paper). Based on these results, we conclude that LT-HSCs do not exhibit any change in differentiation capacity with aging.

Comment #3-12: When determining the lymphoid bias in ST-HSCs, the authors focus on the T-cell subtype, not considering any other any other lymphoid population. Could the authors explain this?

Response #3-12: We thank the reviewer for this comment. We conducted the experiments in Figure 5 to demonstrate that the hematopoiesis observed 16 weeks post-transplantation—when STHSCs are believed to lose their self-renewal capacity—is not due to de novo production of T cells from ST-HSCs. Instead, it is attributed to long-lived memory cells which can persistently remain in the peripheral blood.

As noted by the reviewer, various memory cell types are present in peripheral blood. Our analysis focused on memory T cells due to the broad consensus on memory T cell markers1. 

Our findings show that transplanted Hoxb5- HSCs do not continuously produce lymphoid cells, unlike lymphoid-biased HSCs. Rather, the loss of self-renewal capacity in Hoxb5- HSCs makes the presence of long-lived memory cells in the peripheral blood more apparent.

Reference

(1)  Yenyuwadee S, Sanchez-Trincado Lopez JL, Shah R, Rosato PC, Boussiotis VA. The evolving role of tissue-resident memory T cells in infections and cancer. Sci Adv. 2022;8(33). 

Comment #3-13: Based on the reduced frequency of donor cells in the spleen and thymus, the authors conclude "the process of lymphoid lineage differentiation was impaired in the spleens and thymi of aged mice compared to young mice". An alternative explanation could be that differentiated cells do not successfully migrate from the bone marrow to these secondary lymphoid organs. Please consider this possibility when discussing the data.

Response #3-13: We strongly appreciate the reviewer's comment on this point. In accordance with the reviewer's comment, we have incorporated this suggestion into our manuscript.

[P15, L343] These results indicate that the process of lymphoid lineage differentiation is impaired in the spleens and thymi of aged mice compared to young mice, or that differentiating cells in the bone marrow do not successfully migrate into these secondary lymphoid organs. These factors contribute to the enhanced myeloid-biased hematopoiesis in peripheral blood due to a decrease in de novo lymphocyte production.

Recommendations for the authors:

Reviewer #2 (Recommendations For The Authors):

Recommendation #2-1: To support their conclusions the authors need to provide higher n-numbers and provide a detailed power analysis of the transplants in the methods section.

Response to Recommendation #2-1: Thank you for your important remarks. The power analysis for this experiment shows that power = 0.319, suggesting that more number may be needed. On the other hand, our method for determining the sample size in Figure 3 is as follows:

(1) First, we checked whether myeloid biased change is detected in the bulk-HSC fraction (Figure S3). The results showed that the difference in myeloid output at 16 weeks after transplantation was statistically significant (young vs. aged = 7.2 ± 8.9 vs. 42.1 ± 35.5%, p = 0.01), even though n = 10.

(2) Next, myeloid biased HSCs have been reported to be a fraction with high self-renewal ability (2004, Blood). If myeloid biased HSCs increase with aging, the increase in myeloid biased HSCs in LT-HSC fraction would be detected with higher sensitivity than in the bulk-HSC fraction used in Figure S3.

(3) However, there was no difference not only in p-values but also in the mean itself, young vs aged = 51.4±31.5% vs 47.4±39.0%, p = 0.82, even though n = 8 in Figure 3. Since there was no difference in the mean itself, it is highly likely that no difference will be detected even if n is further increased.

Regarding Figure S3, 5, 6, S6 and 7, we obtained a statistically significant difference and consider the sample size to be sufficient. 

Recommendation #2-2: As the authors attempt to challenge the current model of the age-associated expansion of myeloid-biased HSCs (which has been observed and reproduced by many different groups), ideally additional strong evidence in the form of single-cell transplants is provided.

Response to Recommendation #2-2: Thank you for the comments. As the reviewer pointed out, we hope we could reconfirm our results using single-cell level technology in the future.

On the other hand, we have reported that the ratio of myeloid to lymphoid cells in the peripheral blood changes when the number of HSCs transplanted, or the number of supporting cells transplanted with HSCs, is varied1-2. Therefore, single-cell transplant data need to be interpreted very carefully to determine differentiation potential.

From this viewpoint, future experiments will combine the Hoxb5 reporter system with a lineage tracing system that can track HSCs at the single-cell level over time. This approach will investigate changes in the self-renewal capacity of individual HSCs and their subsequent differentiation into progenitor cells and peripheral blood cells. We have reflected this comment by adding the following sentences in the manuscript.

[P19, L451] In contrast, our findings should be considered in light of some limitations. In this report, we primarily performed ten to twenty transplantation assays. Therefore, the current theory should be revalidated using single-cell technology. This approach will investigate changes in the self-renewal capacity of individual HSCs and their subsequent differentiation into progenitor cells and peripheral blood cells.

References

(1) Nishi K, Sakamaki T, Sadaoka K, Fujii M, Takaori-Kondo A, Chen JY, et al. Identification of the minimum requirements for successful haematopoietic stem cell transplantation. Br J Haematol. 2022;196(3):711–23. 

(2) Sakamaki T, Kao KS, Nishi K, Chen JY, Sadaoka K, Fujii M, et al. Hoxb5 defines the heterogeneity of self-renewal capacity in the hematopoietic stem cell compartment. Biochem Biophys Res Commun [Internet]. 2021;539:34–41. Available from: https://doi.org/10.1016/j.bbrc.2020.12.077

Minor points:

Recommendation #2-3: Figure 1: "Comprehensive analysis of hematopoietic alternations with age shows a discrepancy of age-associated changes between peripheral blood and bone marrow"

[Comment to the authors]: For clarity, the nature of the discrepancy should be stated clearly.

Response to Recommendation #2-3: Thank you for this important comment. Following the reviewer’s recommendation, we have revised the manuscript as follows

[P7, L139] Our analysis of hematopoietic alternations with age revealed that age-associated transition patterns of immunophenotypically defined HSC and CMP in BM were not paralleled with myeloid cell in PB (Fig. 1 C).

Recommendation #2-4: Figure 1B "(B) Average frequency of immunophenotypically defined HSC and progenitor cells in BM of 2-3-month mice (n = 6), 6-month mice (n = 6), 12-13-month mice (n = 6), {greater than or equal to} 23-month mice (n = 7).

[Comment to the authors]: It should be stated in the figure and legend that the values are normalized to the 2-3-month-old mice.

Response to Recommendation #2-4: Thank you for this comment. Figure 1B presents the actual measured values of each fraction in c-Kit positive cells in the bone marrow, without any normalization.

Recommendation #2-5: "We 127 found that the frequency of immunophenotypically defined HSC in BM rapidly increased 128 up to the age of 12 months. After the age, they remained plateaued throughout the 129 observation period (Fig. 1 B)."

[Comment to the authors]: The evidence for a 'plateau', where HSC numbers don't change after 12 months is weak. It appears that the numbers increase continuously (although less steep) after 12 months. I thus recommend adjusting the wording to better reflect the data.

Response to Recommendation #2-5: We thank the reviewer for the comments above and have incorporated these suggestions in our revision as follows. 

[P6, L126] We found that the frequency of immunophenotypically defined HSC in BM rapidly increased up to the age of 12 months. After the age, the rate of increase in their frequency appeared to slow down.

Recommendation #2-6: Figure 2G: [Comment to the authors]: Please add the required statistics, please check carefully all figures for missing statistical tests.

Response to Recommendation #2-6: Thank you for these important comments. In response, we have added the results of the significance tests for Figures 1A, 1C, 4C, and S5.

Recommendation #2-7: "If bulk-HSCs isolated from aged mice are already enriched by myeloid-biased HSC clones, we should see more myeloid-biased phenotypes 16 weeks after primary and the secondary transplantation. However, we found that kinetics of the proportion of myeloid cells in PB were similar across primary and the secondary transplantation and that the proportion of myeloid cells gradually decreased over time (Fig. 2 G). These results suggest the following two possibilities: either myeloid-biased HSCs do not expand in the LT-HSC fraction, or the expansion of myeloid-biased clones in 2-year-old mice has already peaked."

[Comment to the authors]: Other possible explanations include that the observed reduction in myeloid reconstitution over 16 weeks reflects the time required to return to homeostasis. In other words, it takes time until the blood system approaches a balanced output.

Response to Recommendation #2-7: We agree with the reviewer's comment. As the reviewer pointed out, the gradual decrease in the proportion of myeloid cells over time is not related to our two hypotheses in this part of the manuscript but rather to the hematopoietic system's process of returning to a homeostatic state after transplantation. Therefore, the original sentence could be misleading, as it is part of the section discussing whether age-associated expansion of myeloid-biased HSCs is observed. Based on the above, we have revised the sentence as follows.

[P8, L179] However, we found that kinetics of the proportion of myeloid cells in PB were similar across the primary and the secondary transplantation (Fig. 2 G). These results suggest the following two possibilities: either myeloid-biased HSCs do not expand in the LTHSC fraction, or the expansion of myeloid-biased clones in 2-year-old mice has already peaked.

Recommendation #2-8: It is also important to consider that the transplant results are highly variable (see large standard deviation), therefore the sensitivity to detect smaller but relevant changes is low in the shown experiments. As the statistical analysis of these experiments is missing and the power seems low these results should be interpreted with caution. For instance, it appears that the secondary transplants on average produce more myeloid cells as expected and predicted by the classical clonal expansion model.

Regarding "expansion of myeloid-biased clones in 2-year-old mice has already peaked". This is what the author suggested above. It might thus not be surprising that HSCs from 2-year-old mice show little to no increased myeloid expansion.

Response to Recommendation #2-8: Thank you for providing these insights. The primary findings of our study are based on functional experiments presented in Figures 2, 3, 5, 6, and 7. In Figure 3, there was no significant difference between young and aged LT-HSCs, with mean values of 51.4±31.5% and 47.4±39.0%, respectively (p = 0.82). Given the lack of difference in the mean values, it is unlikely that increasing the sample size would reveal a significant change. For ethical reasons, to minimize the use of additional animals, we conclude that LT-HSCs exhibit no change in lineage output throughout life based on the data in Figure 3. Statistically significant differences observed in Figures 2, 5, 6, and 7 further support our conclusions.

Additionally, because whole bone marrow cells were transplanted in the secondary transplantation, there may be various confounding factors beyond the differentiation potential of HSCs. Therefore, we consider that caution is necessary when evaluating the differentiation capacity of HSCs in the context of the second transplantation.

Recommendation #2-9: Figure 7C: [Comment to the authors]: The star * indicates with analyzed BM. As stars are typically used as indicators of significance, this can be confusing for the reader. I thus suggest using another symbol.

Response to Recommendation #2-9: We appreciate the reviewer for this comment and have incorporated the suggestion in the revised manuscript. We have decided to use † instead of the star*.

Reviewer #3 (Recommendations For The Authors):

Recommendation #3.1: In Figure 1A, the authors show the frequency of PB lineages (lymphoid vs myeloid) in mice of different ages. It would be great if they could show the same data for each subpopulation including these two main categories individually (granulocytes, monocytes, B cells, T cells...).

Response to Recommendation #3-1: We thank for this suggestion. We provide the frequency of PB lineages (granulocytes, monocytes, B cells, T cells, and NK cells) in mice of different ages.

Author response image 5.

Average frequency of neutrophils, monocytes, B cells, T cells, and NK cells in PB analyzed in Figure 1A. Dots show all individual mice. *P < 0.05. **P < 0.01. Data and error bars represent means ± standard deviation. 

Recommendation #3.2: It would be great if data from young mice could be shown in parallel to the graphs in Figure 2A.

Response to Recommendation #3-2: We thank the reviewer for the comments above and have incorporated these suggestions in Figure 2A. 

[P34, L916] (A) Hoxb5 reporter expression in bulk-HSC, MPP, Flk2+, and Lin-Sca1-c-Kit+ populations in the 2-year-old Hoxb5-tri-mCherry mice (Upper panel) and 3-month-old Hoxb5_tri-mCherry mice (Lower panel). Values indicate the percentage of mCherry+ cells ± standard deviation in each fraction (_n = 3). 

Recommendation #3.3: Do the authors have any explanation for the high level of variability within the recipients of Hoxb5+ cells in Figure 2C?

Response to Recommendation #3-3: Thank you for providing these insights. As noted in our previous report, transplantation of a sufficient number of HSCs results in stable donor chimerism, whereas a small number of HSCs leads to increased variability in donor chimerism1. Additionally, other studies have observed high variability when fewer than 10 HSCs are transplanted2-3. Based on this evidence, we consider that the transplantation of a small number of cells (10 cells) is the primary cause of the high level of variability observed.

References

(1) Nishi K, Sakamaki T, Sadaoka K, Fujii M, Takaori-Kondo A, Chen JY, et al. Identification of the minimum requirements for successful haematopoietic stem cell transplantation. Br J Haematol. 2022;196(3):711–23. 

(2) Dykstra B, Olthof S, Schreuder J, Ritsema M, Haan G De. Clonal analysis reveals multiple functional defects of aged murine hematopoietic stem cells. J Exp Med. 2011 Dec 19;208(13):2691–703. 

(3) Yamamoto R, Wilkinson AC, Ooehara J, Lan X, Lai CY, Nakauchi Y, et al. LargeScale Clonal Analysis Resolves Aging of the Mouse Hematopoietic Stem Cell Compartment. Cell Stem Cell [Internet]. 2018;22(4):600-607.e4. Available from: https://doi.org/10.1016/j.stem.2018.03.013

Recommendation #3.4: Are the differences in Figure 3D statistically significant? If yes, please add statistics. Same for Figure 4C.

Response to Recommendation #3-4: Thank you for providing these insights. For Figure 3D, we performed an ANOVA analysis for each fraction; however, the results were not statistically significant. In contrast, for Figure 4C, we have added the results of significance tests for comparisons between Young LT-HSC vs. Young Bulk-HSC.

Recommendation #3.5: As a general comment, although the results in this study are interesting, the use of a Hoxb5 lineage tracing mouse model would be more valuable for this purpose than the Hoxb5 reporter used here. The lineage tracing model would allow for the assessment of lineage bias without the caveats introduced by the transplantation assays.

Response to Recommendation #3-5: We appreciate the reviewer for the important comments. Following the reviewer’s recommendation, we have revised the manuscript as follows

[P19, L451] In contrast, our findings should be considered in light of some limitations. In this report, we primarily performed ten to twenty transplantation assays. Therefore, the current theory should be revalidated using single-cell technology with lineage tracing system1-2. This approach will investigate changes in the self-renewal capacity of individual HSCs and their subsequent differentiation into progenitor cells and peripheral blood cells. 

References

(1) Yamamoto R, Wilkinson AC, Ooehara J, Lan X, Lai CY, Nakauchi Y, et al. LargeScale Clonal Analysis Resolves Aging of the Mouse Hematopoietic Stem Cell Compartment. Cell Stem Cell [Internet]. 2018;22(4):600-607.e4. Available from: https://doi.org/10.1016/j.stem.2018.03.013

(2) Rodriguez-Fraticelli AE, Weinreb C, Wang SW, Migueles RP, Jankovic M, Usart M, et al. Single-cell lineage tracing unveils a role for TCF15 in haematopoiesis. Nature [Internet]. 2020;583(7817):585–9. Available from: http://dx.doi.org/10.1038/s41586-020-2503-6

Author response:

The following is the authors’ response to the original reviews.

Reviewer#1:

Comment #1: It is unclear how the fraction of NK cell populations is quantified in the spatial-seq datasets. Figures display spatial data with expression scores, but the method for calculating the score and determining NK cell presence in tumor tissue is ambiguous. Clarification is needed on whether the identification relied solely on visual inspection or if quantitative analyses using other criteria were conducted.

Thank you for your questions. We removed the background and made the accordingly modifications according to your demand. We used the AddModuleScore function in Seurat to quantify the main immune subpopulations in spatial-seq using the gene sets identified in single-cell-seq. Additionally, the tumor and non-tumor region was identified by immunohistochemistry as well as cell clusters in spatial-seq, it is rough that we can't quantify the NK cell presence in each region precisely. The consolation is that the differences of NK cell presence in tumor and non-tumor region is observable by visual inspection. The methodology has been supplemented in the revised manuscript (line 190-193).

Comment #2: The authors do not provide a clear definition of "resting" NK cells. It remains unclear whether they refer to a senescent state or a non-matured NK cell population. Furthermore, the criteria used to define resting and activated cells based on the expression of KIR2DL4, GPR183, GRP171, CD69, IFNG, GZMK, TTC38, CD160, and PLEKNF1 in Figure 4 are not well-defined. The expression patterns of these genes in Figure 4D are not distinct, and it is unclear which combination of genes was used to classify the populations. Clarification is needed on whether the presence of GZMK alone defines resting NK cells, or if the presence of any of the described genes (GZMK, TTC38, or CD160) is sufficient. Additionally, the method used for this classification, whether visual or algorithm-based, should be described.

Thank you for your question. The resting and activated NK cells was defined by the preferential expression of the described resting genes (AZU, BPI, CAMP, CD160,CD2, CDHR1, CEACAM8, DEFA4, ELANE, GFI1, GZMK, KLRC4, MGAM, MS4A3, NME8, PLEKHF1, TEP1, TRBC1, TTC38, ZNF135) and activated NK genes (APOBEC3G, APOL6, CCL4, CCND2, CD69, CDK6, CSF2, DPP4, FASLG, GPR171, GPR18, GRAP2, IFNG, KIR2DL4, KIR2DS4, LTA, LTB, NCR3, OSM, PTGER2, SOCS1, TNFSF14) in CIBERSORT. Actually, these marker genes were not specifically expressed in a single NK cells subset. On the other hand, combined with further flow cytometric analysis verification, the resting NK cell tend to be a decidual-like NK cells and tumor- infiltrated NK cells with higher expression of CD9, CD49a and PD-1.

Comment #3: Criteria used to define high or low NK cell presence/infiltration in Figure 5 are not described in the main text or figure legend. Since, the claim that the presence of the resting or activated NK cells predicts cancer prognosis is based on this figure, this needs to be clearly described.

Thank you for your questions. The activated and resting NK cell percentage in TCGA and GSE29623 was determined by CIBERSORT. Additionally, the infiltration of activated and resting NK cell was also determined by the AddModuleScore function using the gene sets of activated and resting NK cell identified in single-cell-seq, the differences of activated and resting NK cell presence in tumor and non-tumor region is also determined by visual inspection. We have amended in the main text and figure legend in the revised manuscript.

Comment #4: The absence of FMO controls for KIR2DL4 or GZMK and the lack of increase in GZMK expression during co-culture with tumour lines raises concerns since GZMK was used as a defining feature of resting NK cells.

Thank you for your questions. We did a new batch of flow experiments and FMO controls of all the markers used in the experiments were set up to define the precise positive gate locations.

Author response image 1.

The positive gate locations of CD56, GZMK, KIR2DL4, CD9, CD49a, PD-1 defined according to the FMO control.

Comment #5: All the co-cultures were performed with tumour cell line only and no healthy cells, such as human foreskin fibroblasts, were used as control. In the absence of a non-tumour cell line, it is very difficult to draw any conclusions. Furthermore, to claim that resting or activated NK cells are responsible for tumour migration or proliferation, it is important to at least isolate resting and activated NK cells ex vivo and culture with tumour lines, instead of NK cell lines.

Thank you for your questions. According to your suggestion, NK cells were co-cultured with human foreskin fibroblasts, the phenotype was identified by Flow cytometry. When co-cultured with HFF in direct contact (CN group), NK cells were also tending towards tissue infiltration state (high expression of CD9). However, the domestication effect is significantly reduced compared to co-culturing with tumor cells. Additionally, unlike supernatant of CNS group (NK and HCT were in contact) from NK and HCT co-culture system could significantly increase the migration of fresh HCT, fresh HCT underwent a limited increase (no statistical significance was found) in migration when cultured in the supernatant from the co-culture system in which NK and HFF were in contact (CNS group), but not when co-cultures were performed in the cell supernatant (SNS group) and fresh medium (MNS group). Finally, we tried to isolate resting and activated NK cells from fresh colon cancer surgical specimen. Unfortunately, the NK cells were too few to perform further functional experiments such as migration and proliferation.

Author response image 2.

Phenotype switch of NK cells in different co-cultured system and the corresponding NK cell-mediated effect on cell migration of fresh colon cancer cell (HCT-116). A-B: NK cells underwent phenotype switch (high expression of CD9) when cocultured with HCT and HFF, the phenotype switch was more obvious when co-cultured with HCT. CN: NK cells cocultured with HCT/HFF; SN: NK cells cocultured with supernatant of HCT/HFF; MN: NK cells cocultured in fresh medium. C-E: Transwell assay showed the only tumor co-cultured NK mediated the inductive effect on cell migration of colon cancer cell (HCT-116). CNS: Colon cancer cells were cultured in the supernatant from co-culture system that NK and HCT/HFF were cultured in direct contact; SNS: Colon cancer cells were cultured in the supernatant from co-culture system that NK cocultured with supernatant of HCT/HFF; MNS: Colon cancer cells were cultured in the fresh medium.

Comment #6: It seems that flow cytometric analyses and GZMK and KIR2DL4 staining were performed without cell permeabilization. Could authors confirm if this is accurate, or if they performed intracellular staining instead?

Thank you for your questions. For GZMK, which known as the secretory protein, flow cytometric analyses were performed both with (Fig.3) and without cell fixation and permeabilization, no significant differences were found among each group. The difference is that GZMK was nearly all negative without fixation and permeabilization while it is all positive with fixation and permeabilization. Conditions of flow cytometry analyses for GZMK may need further optimization or GZMK may not be a suitable flow cytometric marker for resting NK cells. On the other hand, for membrane protein such as CD56, CD9, CD49a, KIR2DL4, PD-1, staining was performed without cell permeabilization.

Author response image 3.

Phenotype switch (CD56+, GZMK+) of NK cells was analyzed by FACS after fixation and permeabilization in different co-cultured groups. CN: NK cells cocultured with colon cancer cells; SN: NK cells cocultured with supernatant of cancer cells; MN: NK cells cocultured in fresh medium.

Comment #7: The identity of the published datasets used for analysis is not provided, and references are not cited in the results section.

Thank you for your questions. We are sorry for the neglect of our previous work. We have added the information in the revised manuscript (section of Materials and Methods) (Line 123-128).

Comment #8: References are difficult to locate, as the main text follows APA style while the reference section is organized numerically with no clear order.

Thank you for your questions. We have modified the format of the references in the revised manuscript.

Comment #9: Figure 3 shows volcano plots showing DEG genes between tumor and healthy tissue NK cells are not described clearly, and authors did not discuss the significance of these genes, highlighted in the plot.

Thank you for your questions. Volcano plots of Figure 3 showed the DEGs between colon cancer with metastasis and without metastasis in TCGA database. We focused on the genes which were enriched in the pathway of “Natural killer cell mediated cytotoxicity” and found nearly all the genes enriched in the pathway were down-regulated in the colon cancer with metastasis. We have modified the description in the result section and added the description of importance of these genes in the discussion section in the revise manuscript (Line 322-326).

Comment #10: The meaning of "M0" and "M1" in Figures 5A and 5B is unclear and should be defined in the text.

Thank you for your questions. "M0" and "M1" in Figure 5A and 5B means “colon cancer without metastasis” and “colon cancer with metastasis”, respectively. We have modified in the revise manuscript (Line 350-354).

Comment #11: Terms such as "dynamic remodelling of NK cells" and "landscape of NK cells" are used without explanation, necessitating clarification of their meaning.