Author response:

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

Reviewer #1 (Public review):

Summary:

This manuscript by Pournejati et al investigates how BK (big potassium) channels and CaV1.3 (a subtype of voltage-gated calcium channels) become functionally coupled by exploring whether their ensembles form early-during synthesis and intracellular trafficking-rather than only after insertion into the plasma membrane. To this end, the authors use the PLA technique to assess the formation of ion channel associations in the different compartments (ER, Golgi or PM), single-molecule RNA in situ hybridization (RNAscope), and super-resolution microscopy.

Strengths:

The manuscript is well written and addresses an interesting question, combining a range of imaging techniques. The findings are generally well-presented and offer important insights into the spatial organization of ion channel complexes, both in heterologous and endogenous systems.

Weaknesses:

The authors have improved their manuscript after revisions, and some previous concerns have been addressed.

Still, the main concern about this work is that the current experiments do not quantitatively or mechanistically link the ensembles observed intracellularly (in the endoplasmic reticulum (ER) or Golgi) to those found at the plasma membrane (PM). As a result, it is difficult to fully integrate the findings into a coherent model of trafficking. Specifically, the manuscript does not address what proportion of ensembles detected at the PM originated in the ER. Without data on the turnover or halflife of these ensembles at the PM, it remains unclear how many persist through trafficking versus forming de novo at the membrane. The authors report the percentage of PLApositive ensembles localized to various compartments, but this only reflects the distribution of pre-formed ensembles. What remains unknown is the proportion of total BK and Ca_V1.3 channels (not just those in ensembles) that are engaged in these complexes within each compartment. Without this, it is difficult to determine whether ensembles form in the ER and are then trafficked to the PM, or if independent ensemble formation also occurs at the membrane. To support the model of intracellular assembly followed by coordinated trafficking, it would be important to quantify the fraction of the total channel population that exists as ensembles in each compartment. A comparable ensemble-to-total ratio across ER and PM would strengthen the argument for directed trafficking of pre-assembled channel complexes.

We appreciate the reviewer’s thoughtful comment and agree that quantitatively linking intracellular hetero-clusters to those at the plasma membrane is an important and unresolved question. Our current study does not determine what proportion of ensembles at the plasma membrane originated during trafficking. It also does not quantify the fraction of total BK and Ca_V1.3 channels engaged in these complexes within each compartment. Addressing this requires simultaneous measurement of multiple parameters—total BK channels, total Ca_V1.3 channels, hetero-cluster formation (via PLA), and compartment identity—in the same cell. This is technically challenging. The antibodies used for channel detection are also required for the proximity ligation assay, which makes these measurements incompatible within a single experiment.

To overcome these limitations, we are developing new genetically encoded tools to enable real-time tracking of BK and Ca_V1.3 dynamics in live cells. These approaches will enable us to monitor channel trafficking and the formation of hetero-clusters, as detected by colocalization. This kind of experiments will provide insight into their origin and turnover. While these experiments are beyond the scope of the current study, the findings in our current manuscript provide the first direct evidence that BK and CaV channels can form hetero-clusters intracellularly prior to reaching the plasma membrane. This mechanistic insight reveals a previously unrecognized step in channel organization and lays the foundation for future work aimed at quantifying ensemble-to-total ratios and determining whether coordinated trafficking of pre-assembled complexes occurs.

This limitation is acknowledged in the discussion section, page 23. It reads: “Our findings highlight the intracellular assembly of BK-Ca_V1.3 hetero-clusters, though limitations in resolution and organelle-specific analysis prevent precise quantification of the proportion of intracellular complexes that ultimately persist on the cell surface.”

Reviewer #2 (Public review):

Summary:

The co-localization of large conductance calcium- and voltage activated potassium (BK) channels with voltage-gated calcium channels (CaV) at the plasma membrane is important for the functional role of these channels in controlling cell excitability and physiology in a variety of systems.

An important question in the field is where and how do BK and CaV channels assemble as 'ensembles' to allow this coordinated regulation - is this through preassembly early in the biosynthetic pathway, during trafficking to the cell surface or once channels are integrated into the plasma membrane. These questions also have broader implications for assembly of other ion channel complexes

Using an imaging based approach, this paper addresses the spatial distribution of BKCaV ensembles using both overexpression strategies in tsa201 and INS-1 cells and analysis of endogenous channels in INS-1 cells using proximity ligation and superesolution approaches. In addition, the authors analyse the spatial distribution of mRNAs encoding BK and Cav1.3.

The key conclusion of the paper that BK and Ca_V1.3 are co-localised as ensembles intracellularly in the ER and Golgi is well supported by the evidence.However, whether they are preferentially co-translated at the ER, requires further work. Moreover, whether intracellular pre-assembly of BK-Ca_V1.3 complexes is the major mechanism for functional complexes at the plasma membrane in these models requires more definitive evidence including both refinement of analysis of current data as well as potentially additional experiments.

The reviewer raises the question of whether BK and Ca_V1.3 channels are preferentially co-translated. In fact, I would like to propose that co-translation has not yet been clearly defined for this type of interaction between ion channels. In our current work, we 1) observed the colocalization between BK and Ca_V1.3 mRNAs and 2) determined that 70% of BK mRNA in active translation also colocalizes with Ca_V1.3 mRNA. We think these results favor the idea of translational complexes that can underlie the process of co-translation. However, and in total agreement with the Reviewer, the conclusion that the mRNA for the two ion channels is cotranslated would require further experimentation. For instance, mRNA coregulation is one aspect that could help to define co-translation. 

To avoid overinterpretation, we have revised the manuscript to remove references to “co-translation” in the Results section and included the word “potential” when referring to co-translation in the Discussion section. We also clarified the limitations of our evidence in the Discussion that can be found on page 25: “It is important to note that while our data suggest mRNA coordination, additional experiments are required to directly assess co-translation.”

Strengths & Weaknesses

(1) Using proximity ligation assays of overexpressed BK and CaV1.3 in tsa201 and INS1 cells the authors provide strong evidence that BK and CaV can exist as ensembles (ie channels within 40 nm) at both the plasma membrane and intracellular membranes, including ER and Golgi. They also provide evidence for endogenous ensemble assembly at the Golgi in INS-1 cells and it would have been useful to determine if endogenous complexes are also observe in the ER of INS-1 cells. There are some useful controls but the specificity of ensemble formation would be better determined using other transmembrane proteins rather than peripheral proteins (eg Golgi 58K).

We thank the reviewer for their thoughtful feedback and for recognizing the strength of our proximity ligation assay data supporting BK–Ca_V1.3 hetero-clusters formation at both the plasma membrane and intracellular compartments. As for specificity controls, we appreciate the suggestion to use transmembrane markers. To strengthen our conclusion, we have performed an additional experiment comparing the number of PLA puncta formed by the interaction of Ca_V1.3 and BK channels with the number of PLA puncta formed by the interaction of Ca_V1.3 channels and ryanodine receptors in INS-1 cells. As shown in the figure below, the number of interactions between Ca_V1.3 and BK channels is significantly higher than that between Ca_V1.3 and RyR₂. Of note, RyR₂ is a protein resident of the ER. These results provide additional evidence of the existence of endogenous complex formation in INS-1 cells. We have added this figure as a supplement.

(2) Ensemble assembly was also analysed using super-resolution (dSTORM) imaging in INS-1 cells. In these cells only 7.5% of BK and CaV particles (endogenous?) co-localise that was only marginally above chance based on scrambled images. More detailed quantification and validation of potential 'ensembles' needs to be made for example by exploring nearest neighbour characteristics (but see point 4 below) to define proportion of ensembles versus clusters of BK or Cav1.3 channels alone etc. For example, it is mentioned that a distribution of distances between BK and Cav is seen but data are not shown.

We thank the reviewer for this comment. To address the request for more detailed quantification and validation of ensembles, we performed additional analyses:

Proportion of ensembles vs isolated clusters: We quantified clusters within 200 nm and found that 37 ± 3% of BK clusters are near one or more CaV1.3 clusters, whereas 15 ± 2% of CaV1.3 clusters are near BK clusters. Figure 8– Supplementary 1A

Distance distribution: As shown in Figure 8–Supplementary 1B, the nearestneighbor distance distribution for BK-to-CaV1.3 in INS-1 cells (magenta) is shifted toward shorter distances compared to randomized controls (gray), supporting preferential localization of BK–CaV1.3 hetero-clusters.

Together, these analyses confirm that BK–CaV1.3 ensembles occur more frequently than expected by chance and exhibit an asymmetric organization favoring BK proximity to CaV1.3 in INS-1 cells. We have included these data and figures in the revised manuscript, as well as description in the Results section. 

(3) The evidence that the intracellular ensemble formation is in large part driven by cotranslation, based on co-localisation of mRNAs using RNAscope, requires additional critical controls and analysis. The authors now include data of co-localised BK protein that is suggestive but does not show co-translation. Secondly, while they have improved the description of some controls mRNA co-localisation needs to be measured in both directions (eg BK - SCN9A as well as SCN9A to BK) especially if the mRNAs are expressed at very different levels. The relative expression levels need to be clearly defined in the paper. Authors also use a randomized image of BK mRNA to show specificity of co-localisation with Cav1.3 mRNA, however the mRNA distribution would not be expected to be random across the cell but constrained by ER morphology if cotranslated so using ER labelling as a mask would be useful?

We thank the reviewer for these constructive suggestions. We measured mRNA colocalization in both directions as recommended. As shown in the figure below, colocalization between KCNMA1 and SCN9A transcripts was comparable in both directions, with no statistically significant difference, supporting the specificity of the observed associations. We decided not to add this to the original figure to keep the figure simple. 

We agree that co-localization of BK protein with BK mRNA is not conclusive evidence of co-translation, and we do not intend to mislead readers in our conclusion. Consequently, we were careful in avoiding the use of co-translation in the result section and added the word “potential” when referring to co-translation in the Discussion section. We added a sentence in the discussion to caution our interpretation: “It is important to note that while our data suggest mRNA coordination, additional experiments are required to directly assess cotranslation.”

Author response image 1.

(4) The authors attempt to define if plasma membrane assemblies of BK and CaV occur soon after synthesis. However, because the expression of BK and CaV occur at different times after transient transfection of plasmids more definitive experiments are required. For example, using inducible constructs to allow precise and synchronised timing of transcription. This would also provide critical evidence that co-assembly occurs very early in synthesis pathways - ie detecting complexes at ER before any complexes 

We appreciate the reviewer’s insightful suggestion regarding the use of inducible constructs to synchronize transcription timing. This is an excellent approach and would allow direct testing of whether co-assembly occurs early in the synthesis pathway, including detection of complexes at the ER prior to plasma membrane localization. These experiments are beyond the scope of the present work but represent an important direction for future studies.

We have added the following sentence to the Discussion section (page 24) to highlight this idea. “Future experiments using inducible constructs to precisely control transcription timing will enable more precise quantification of heterocluster formation in the ER compartment prior to plasma membrane insertion and reduce the variability introduced by differences in expression timing after plasmid transfection.” 

(5) While the authors have improved the definition of hetero-clusters etc it is still not clear in superesolution analysis, how they separate a BK tetramer from a cluster of BK tetramers with the monoclonal antibody employed ie each BK channel will have 4 binding sites (4 subunits in tetramer) whereas Cav1.3 has one binding site per channel. Thus, how do authors discriminate between a single BK tetramer (molecular cluster) with potential 4 antibodies bound compared to a cluster of 4 independent BK channels.

We appreciate the reviewer’s thoughtful comment regarding the interpretation of super-resolution data. We agree that distinguishing a single BK tetramer from a cluster of multiple BK channels is challenging when using an antibody that can bind up to four sites per channel. To clarify, our analysis does not attempt to resolve individual subunits within a tetramer; rather, it focuses on the nanoscale spatial proximity of BK and Ca_V1.3 signals.

We want to note that this limitation applies only to the super-resolution maps in Figures 8C and 9D and does not affect Airyscan-based analyses or measurements of BK–Ca_V1.3 proximity.

To address how we might distinguish between a single BK tetramer and a cluster of multiple BK channels, we considered two contrasting scenarios. In the first case, we assume that all four α-subunits within a tetramer are labeled. Based on cryoEM structures, a BK tetramer measures approximately 13 nm × 13 nm (≈169 nm²). Adding two antibody layers (primary and secondary) would increase the footprint by ~14 nm in each direction, resulting in an estimated area of ~41 nm × 41 nm (≈1681 nm²). Under this assumption, particles smaller than ~1681 nm² would likely represent individual tetramers, whereas larger particles would correspond to clusters of multiple tetramers. 

In the second scenario, we propose that steric constraints at the S9–S10 segment, where the antibody binds, limit labeling to a single antibody per tetramer. If true, the localization precision would approximate 14 nm × 14 nm—the combined size of the antibody complex and the channel—close to the resolution limit of the microscope. To test this, we performed a control experiment using two antibodies targeting the BK C-terminal domain, raised in different species and labeled with distinct fluorophores. Super-resolution imaging revealed that only ~12% of particles were colocalized, suggesting that most channels bind a single antibody.

If multiple antibodies could bind each tetramer, we would expect much greater colocalization.

Although these data are not included in the manuscript, we have added the following clarification to the Results section (page 19): “It is important to note that this technique does not allow us to distinguish between labeling of four BK αsubunits within a tetramer and labeling of multiple BK channel clusters. Hence, particles smaller than ~1680 nm² may represent either a single tetramer or a cluster. This limitation applies to Figures 8C and 9D and does not affect measurements of BK–Ca_V1.3 proximity.”

Author response image 2.

(6) The post-hoc tests used for one way ANOVA and ANOVA statistics need to be defined throughout

We thank the reviewer for highlighting the need for clarity regarding our statistical analyses. We have now specified the post-hoc tests used for all one-way ANOVA and ANOVA comparisons throughout the manuscript, and updated figure legends.

Reviewer #3 (Public review):

Summary:

The authors present a clearly written and beautifully presented piece of work demonstrating clear evidence to support the idea that BK channels and Cav1.3 channels can co-assemble prior to their assertion in the plasma membrane.

Strengths:

The experimental records shown back up their hypotheses and the authors are to be congratulated for the large number of control experiments shown in the ms.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

The authors have sufficiently addressed the specific points previously raised and the manuscript has improved clarity in those aspects. My main concern, which still remains, is stated in the public review.

Reviewer #3 (Recommendations for the authors):

I am content that the authors have attempted to fully address my previous criticisms.

I have only three suggestions

(1) I think the word Homo-clusters at the bottom right of Figure 1 is erroneously included.

We thank the reviewer for bringing this to our attention. The figure has been corrected accordingly.

(2) The authors should, for completeness, to refer to the beta, gamma and LINGO subunit families in the Introduction and include appropriate references:

Knaus, H. G., Folander, K., Garcia-Calvo, M., Garcia, M. L., Kaczorowski, G. J., Smith, M., & Swanson, R. (1994). Primary sequence and immunological characterization of betasubunit of high conductance Ca2+-activated K+ channel from smooth muscle. The Journal of Biological Chemistry, 269(25), 17274-17278.

Brenner, R., Jegla, T. J., Wickenden, A., Liu, Y., & Aldrich, R. W. (2000a). Cloning and functional characterization of novel large conductance calcium-activated potassium channel beta subunits, hKCNMB3 and hKCNMB4. The Journal of Biological Chemistry, 275(9), 6453-6461.

Yan, J & R.W. Aldrich. (2010) LRRC26 auxiliary protein allows BK channel activation at resting voltage without calcium. Nature. 466(7305):513-516

Yan, J & R.W. Aldrich. (2012) BK potassium channel modulation by leucine-rich repeatcontaining proteins. Proceedings of the National Academy of Sciences 109(20):7917-22

Dudem, S, Large RJ, Kulkarni S, McClafferty H, Tikhonova IG, Sergeant, GP, Thornbury, KD, Shipston, MJ, Perrino BA & Hollywood MA (2020). LINGO1 is a novel regulatory subunit of large conductance, Ca2+-activated potassium channels. Proceedings of the National Academy of Sciences 117 (4) 2194-2200

Dudem, S., Boon, P. X., Mullins, N., McClafferty, H., Shipston, M. J., Wilkinson, R. D. A., Lobb, I., Sergeant, G. P., Thornbury, K. D., Tikhonova, I. G., & Hollywood, M. A. (2023). Oxidation modulates LINGO2-induced inactivation of large conductance, Ca2+-activated potassium channels. The Journal of Biological Chemistry, 299 (3) 102975.

We agree with the reviewer’s suggestion and have revised the Introduction to include references to the beta, gamma, and LINGO subunit families. Appropriate citations have been added to ensure completeness and contextual relevance.

Additionally, BK channels are modulated by auxiliary subunits, which fine-tune BK channel gating properties to adapt to different physiological conditions. The β, γ, and LINGO1 subunits each contribute distinct structural and regulatory features: β-subunits modulate Ca²⁺ sensitivity and can induce inactivation; γ-subunits shift voltage-dependent activation to more negative potentials; and LINGO1 reduces surface expression and promotes rapid inactivation (18-24). These interactions ensure precise control over channel activity, allowing BK channels to integrate voltage and calcium signals dynamically in various cell types.

(3) I think it may be more appropriate to include the sentence "The probes against the mRNAs of interest and tested in this work were designed by Advanced Cell Diagnostics." (P16, right hand column, L12-14) in the appropriate section of the Methods, rather than in Results.

We thank the reviewer for this helpful suggestion. In response, we have relocated the sentence to the appropriate section of the Methods, where it now appears with relevant context.

The possibility of changing becomes increasingly more evident in this part of the song, which gives shape to a conversation between a child and an old man. The resignation seems to belong to the viewpoint of the old man, whereas the child questions the very existence of the "rules" who discriminate them for the color of his skin ("you don't look like they do") and the origin of the authority of those who created them. At this point of the song, resignation seems to resemble a compliant attitude, which may be even benefitting from an injustice society. Notice that the singer evidently sides with the boy, ultimately converging his point of view with his own.

This first stanza describes a specific image: the songs opens with a symbolical "line" where people in distress are waiting to receive some help from the government.

Hornsby's posture seems clear: in these verses, he states that legal measures can only "go so far", that is, they can only operate up to a certain extent. He hints at the fact that another change has to occur: a cultural one. According to the author, indeed, equality can be achieved only in a co-constructive process that implicates the law on one side, and the culture on the other. In other words, what Hornsby is trying to state is that acts can be signed into law and assure people their rights, but it is equally important that people change their mindsets.

Se podría escribir un último párrafo breve respecto a esta hipótesis de interacción, con tal de que se despeje su carácter explotario, y así también dar espacio a justificar porqué se decidió moderar por nivel de gobernanza y no por desigualdad, por ejemplo

creo que esto debería ir antes de definir market justice

esta idea la tiraría más bien al final de esta sección como anuncio de que se viene meritocracia al baile

esto no queda claro

Note: This response was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Reply to the reviewers

Reviewer #1 (Evidence, reproducibility and clarity):

A previous study by Komada et al. demonstrated that MAP7 is expressed in both Sertoli and germ cells, and that Map7 gene-trap mutant mice display disrupted microtubule bundle formation in Sertoli cells, accompanied by defects in spermatid manchettes and germ cell loss. In the current study, Kikuchi et al. investigated the role of MAP7 in the formation of the Sertoli cell apical domain during the first wave of spermatogenesis. They generated a GFP-tagged MAP7 mouse line and demonstrated that the endogenous MAP7 protein localizes to the apical microtubules in Sertoli cells and to the manchette microtubules in step 9-11 spermatids. They also generated a new Map7 knockout (KO) mouse line in a genetic background distinct from the one used in the previous study. Focusing on stages before the emergence of step 9-11 spermatids, the authors aimed to isolate defects caused by the function of MAP7 in Sertoli cells. They report that loss of MAP7 impairs Sertoli cell polarity and apical domain formation, accompanied by the microtubule remodeling defect. Using the GFP-tagged MAP7 line, they performed immunoprecipitation-mass spectrometry and identified several MAP7-interacting proteins in the testis, including MYH9. They further observed that MAP7 deletion alters the distribution of MYH9. Single-cell RNA sequencing revealed that the loss of MAP7 in Sertoli cells resulted in slight transcriptomic shifts but had no significant impact on their functional differentiation. Single-cell RNA sequencing analysis also showed delayed meiotic progression in the MAP7-deficient testis. Overall, while the study provides some interesting discoveries of early Sertoli cell defects in MAP7-deficient testes, some conclusions are premature and not fully supported by the presented data. The mechanistic investigations remain limited in depth.

Response: We thank the reviewer for this insightful summary. We agree that some of our initial interpretations were speculative and have revised the relevant sections to more accurately reflect the limitations of the current data. We also acknowledge that further mechanistic studies will be important to strengthen our conclusions, and we have outlined these plans in the individual responses below.

Major comments:

• Although the infertility phenotype of the Map7 gene-trap mutant mice has been reported previously, it remains essential to assess fertility in this newly generated MAP7 knockout line. While the authors present testis size and histological differences between WT and KO mice (Extended Fig. 2e and 2f), there is no corresponding description or interpretation in the main text regarding fertility outcomes.

Response: We thank the reviewer for raising this point. Although we had presented the differences in testis size and histology between wild-type and Map7-/- mice, we agree that a description of the corresponding fertility outcomes was missing from the main text. We have now revised the relevant part of the Results section as follows: “Consistent with observations in Map7 gene-trap mice, Map7-/- males exhibited reduced testis size and spermatogenic defects (Supplemental Fig. 2E, F). Notably, the cauda epididymis of Map7-/- males contained no mature spermatozoa (Supplemental Fig. 2F), indicating male infertility.” (page 5, line 33–page 6, line 2)

• In Figure 2C, the authors identified Sertoli cells, spermatogonia cells, and spermatocytes using SEM, based on their cell morphology and adhesion to the basement membrane. Given that the loss of MAP7 disrupts the polarity and architecture of Sertoli cells, the position of germ cells will be affected, making this identification criterion less reliable.

Response: We appreciate the reviewer’s comment. While the reviewer notes that cell identification was based on cell morphology and adhesion to the basement membrane, we clarify that nuclear morphology was also considered, as described in the original manuscript. Specifically, germ cells have spherical nuclei, whereas Sertoli cell nuclei are irregularly shaped (representative segmentation results can be provided as an additional Supplemental Figure upon request). Round spermatids at P21 can be distinguished from spermatocytes by their smaller nuclear size. In addition, spermatogonia remain attached to the basement membrane even in Map7-/- testes, as confirmed by GFRα1-positive spermatogonial stem cells (Figure 6A). Together, these features ensure reliable identification of each cell type, independent of the altered polarity observed in Map7-deficient Sertoli cells.

• In Figure 2e, the number of Sox9-positive Sertoli cells in MAP7 knockout mice appears higher than that in the control at P17. Quantification of total Sox9-positive cells should be done to determine whether MAP7 deletion increases Sertoli cell numbers.

Response: As suggested by the reviewer, we will quantify the density of SOX9-positive Sertoli cells per unit area of seminiferous tubule at P10 and P17 in Map7+/- and Map7-/- testes, and include the results in the revised manuscript.

• To determine whether MAP7's role in regulating Sertoli cell polarity relies on germ cells, the authors treated mice with busulfan at P28 to delete germ cells, a stage after Sertoli cell polarity defect has developed in MAP7 knockout mice. This data is insufficient to support the conclusion that MAP7 regulates Sertoli cell polarity independently of the presence of germ cells. Germ cell deletion should be done before the Sertoli cell defect develops to address this question.

Response: We appreciate the reviewer’s thoughtful comment regarding the interpretation of the busulfan experiments. While depletion of germ cells at P28 enabled us to assess Sertoli cell polarity in the absence of postnatal spermatogonia, these experiments do not definitively determine whether MAP7 regulates Sertoli cell polarity independently of germ cells. Neonatal germ-cell depletion would more directly test germ cell–independent effects; however, systemic busulfan administration at early developmental stages is highly toxic, often causing bone marrow failure and multi-organ damage, which precludes survival and confounds analysis of testis-specific effects. Although germ cell ablation could, in principle, be achieved using transgenic approaches that exploit the natural resistance of mice to diphtheria toxin (DTX) (reviewed in Smith et al., Andrology, 2015), these strategies require multiple transgenes and show minor variability in efficiency, making them impractical for our current experiments. Generating the necessary genetic combinations would require considerable time. We therefore plan to pursue alternative genetic approaches in future work.

In the revised manuscript, we have modified the relevant section to more accurately reflect the limitations of the current experiments, as follows: “Busulfan was administered at P28, and testes were analyzed 6 weeks later, after complete elimination of germ cell lineages. Following treatment, Map7+/- mice showed testis-to-body weight ratios comparable to untreated Map7-/- mice (Supplemental Fig. 3D), and hematoxylin-eosin (HE) staining confirmed germ cell depletion (Fig. 2F; Supplemental Fig. 3E). In Map7+/- testes, most Sertoli nuclei remained basally positioned, indicating that once apical–basal polarity is established, it is stably maintained even in the absence of germ cells. In contrast, Map7-/- Sertoli nuclei were frequently misoriented toward the lumen under the same conditions (Fig. 2F; Supplemental Fig. 3E), suggesting that polarity defects in Map7-deficient Sertoli cells occur independently of germ cell presence.” (page 7, lines 20–28)

In addition, we have added the following sentences to the Discussion section to highlight the implication of these findings: “In addition, even after germ cell depletion by busulfan treatment, Map7-deficient Sertoli cells failed to reestablish basal nuclear positioning, indicating that loss of MAP7 causes an intrinsic polarity defect. These findings suggest that MAP7 acts as a cell-autonomous regulator of Sertoli cell polarity, rather than mediating effects indirectly through germ cell–Sertoli cell interactions.” (page 15, lines17–21)

• The resolution of the SEM images in Figure 3c is insufficient to evaluate tight and adherens junctions clearly. As such, these images do not convincingly support the claim that adherens junctions are absent in the KO testes.

Response: We thank the reviewer for this insightful comment. Tight junctions can be reliably identified in SEM images as dense intercellular structures accompanied by endoplasmic reticulum aligned along the cell boundaries. The region immediately apical to the tight junctions likely corresponds to adherens junctions, which are also associated with the endoplasmic reticulum. Unlike tight junctions, these regions exhibit wider intercellular spaces, consistent with the looser membrane apposition characteristic of adherens junctions, although they cannot be unambiguously distinguished from gap junctions or desmosomes based on morphology alone. In the original figure, 2× binning reduced image resolution, which may have contributed to the reviewer’s concern.

In the revised manuscript, we have re-acquired the SEM images in high-resolution mode, focusing on the relevant regions. The new high-resolution images have replaced the original panels in revised Figure 3C, providing clearer visualization of junctional structures at P10 and P21 in Map7+/- and Map7-/- testes. The original Figure 3C images have been moved to Supplemental Figure 4B for reference.

The corresponding section in the Results has been revised as follows in the updated manuscript: “We then performed SEM to examine the effects of Map7 KO. In P21 Map7+/- testes, electron-dense regions along the basal side of Sertoli–Sertoli junctions corresponded to tight junctions closely associated with the endoplasmic reticulum, consistent with previous reports (Luaces et al. 2023) (Fig. 3C; Supplemental Fig. 4B). The region immediately apical to the tight junctions likely represents adherens junctions, which were also associated with the endoplasmic reticulum. Unlike tight junctions, these regions displayed wider intercellular spaces, reflecting the looser membrane apposition typical of adherens junctions, though they could not be definitively distinguished from gap junctions or desmosomes based on morphology alone (Fig. 3C; Supplemental Fig. 4B). At P10, both Map7+/- and Map7-/- testes lacked clearly defined tight junctions and adherens junction–like structures (Fig. 3C; Supplemental Fig. 4B). In P21 Map7-/- mice, Sertoli cells formed expanded basal tight junctions but failed to establish adherens junction–like structures (Fig. 3C; Supplemental Fig. 4B).” (page 8, line 34–page 9, line 12)

• GFP-tagged reporter mice and HeLa cells were used for immunoprecipitation-mass spectrometry to identify proteins that interact with MAP7. Given that the authors aimed to elucidate the mechanism by which MAP7 regulates Sertoli cell cytoskeleton organization, the rationale for including HeLa cells is unclear and should be better justified or reconsidered.

Response: We thank the reviewer for this comment. MAP7-egfpKI HeLa cells were used as a complementary system to identify MAP7-associated proteins, providing sufficient material and a controlled environment for robust detection. By comparing IP-MS results from MAP7-egfpKI HeLa cells and P17–P20 Map7-egfpKI testes, we can distinguish proteins that are specific to polarized Sertoli cells: proteins detected exclusively in P17–P20 testes may be involved in Sertoli cell polarization, whereas proteins detected in both systems likely represent general MAP7-associated factors that are not specific to Sertoli cell polarity.

This rationale has been clarified in the revised manuscript by adding the following sentence to the Results section: “MAP7-egfpKI HeLa cells were used as a complementary system, providing sufficient material and a controlled environment for robust detection of MAP7-associated proteins. Comparison of IP-MS results between MAP7-egfpKI HeLa cells and P17–P20 Map7-egfpKI testes allows identification of MAP7-associated proteins that are specific to polarized Sertoli cells, whereas proteins detected in both systems likely represent general MAP7-associated proteins.” (page 9 lines 27-32)

• The authors observed that MYH9, one of the MAP7-interacting proteins, does not colocalize with ectopic microtubule and F-actin structures in MAP7 KO testes and concluded that MAP7 facilitates the integration of microtubules and F-actin via interaction with NMII heavy chains. This conclusion is speculative and not adequately supported by the presented data.

Response: We thank the reviewer for this insightful comment. We agree that our initial conclusion was speculative and have revised the relevant section to more accurately reflect the limitations of the current data. The revised text now reads as follows: “These findings indicate that MYH9 localization at the luminal interface depends on MAP7, and suggest that MAP7 helps coordinate microtubules and F-actin, potentially via its association with NMII heavy chains.” (page 10, lines 13–15)

To further elucidate this mechanism, we will perform biochemical domain-mapping to define the MAP7 region responsible for MYH9 complex formation. We have already established a series of human MAP7 deletion mutants (as reported previously, EMBO Rep., 2018) and will conduct co-immunoprecipitation assays in HEK293 cells to identify the specific MAP7 domain required for complex formation with MYH9. Based on these results, we plan to use AlphaFold3 to predict the three-dimensional structure of the MAP7–MYH9 complex. These analyses will help clarify how MAP7 associates with the actomyosin network and provide additional mechanistic insights that complement our in vivo observations of MYH9 mislocalization in Map7-/- testes.

• The authors used Spearman correlation coefficients to analyze six Sertoli cell clusters and generated a minimum spanning tree to infer differentiation trajectories. However, details on the method used for constructing the tree are lacking. Moreover, relying solely on Spearman correlation to define differentiation topology is oversimplified.

Response: We appreciate the reviewer’s valuable feedback. We agree that Spearman correlation alone is insufficient to infer differentiation topology. In response, we reanalyzed the data using Monocle3, which implements branch-aware pseudotime inference to capture both cluster continuity and differentiation directionality. This reanalysis provides a more accurate reconstruction of differentiation trajectories among the six Sertoli cell clusters. Although the overall trajectories appeared different and a higher proportion of Map7-/- Sertoli cells exhibited very low pseudotime values, comparison of the control and Map7-/- trajectories revealed that the average node degree was nearly identical, indicating that the local graph structure—reflecting the connectivity among neighboring cells—was largely preserved. The numbers of branch points and the graph diameter differed slightly, likely due to differences in sample size (311 control vs. 434 Map7-/- Sertoli cells) and distribution bias rather than major topological changes. Accordingly, Figures 5C and 5D have been replaced with the updated Monocle3-based trajectory analysis, and the corresponding text in the Results section and figure legend have been revised as follows:

“To reconstruct differentiation trajectories among the six Sertoli cell clusters, we reanalyzed the datasets using Monocle3, which incorporates branch-aware pseudotime inference. Cluster C1 was selected as the root based on shared specificity and entropy scores, consistent with its metabolically active and transcriptionally diverse profile (Fig. 5B, C; Supplemental Fig. 7). While the overall trajectories appeared altered, the proportion of Map7-/- Sertoli cells with very low pseudotime values was only modestly increased (Fig. 5D). Comparison with controls showed that the average node degree was nearly identical (Fig. 5C), indicating that the local graph structure, reflecting connectivity among neighboring cells, remained largely intact. Minor differences in branch points and graph diameter likely reflect inherent variability in the data rather than major topological changes (Supplemental Fig. 6B). Consistent with this, the relative proportions of the six clusters showed only modest shifts, suggesting that the overall architecture of Sertoli cell differentiation is largely preserved in the absence of MAP7.” (page 11, lines 7-18)

“(C) Control and Map7-/- Sertoli cells were visualized separately using UMAPs constructed in Seurat. Using the same datasets, pseudotime trajectories were inferred with Monocle3. For root selection, shared_score (cluster overlap), specificity_score (cluster uniqueness), and entropy_score (transcriptional diversity) were computed, resulting in cluster 1 being selected as the root. The numbers of nodes, edges, branch points, average degree, and diameter of each trajectory are shown below the corresponding UMAPs. (D) Parallel comparison of pseudotime distributions between control and Map7-/- populations.” (page 30, lines 5-12)

Minor comments:

• Several extended data figures are redundant with main figures and do not provide additional value (e.g., Fig. 2d vs. Extended Data Fig. 3a; Fig. 2f vs. Extended Data Fig. 3d; Fig. 2C vs. Extended Data Fig. 4b; Fig. 3d vs. Extended Data Fig. 4c). The authors should consolidate or remove duplicates.

Response: Regarding the concerns about redundancy between main and Supplemental figures, we would like to clarify the rationale for retaining certain Supplemental figures.

Fig. 2D vs. Supplemental Fig. 3A: Due to space limitations in the main figure, only the merged three-color image was shown. We believe that the single-color grayscale images in Supplemental Fig. 3A provide additional clarity, allowing easier visualization of SOX9-positive Sertoli cell distribution and differences in F-actin structure.

Fig. 2F vs. Supplemental Fig. 3E: In the main figure, only the high-magnification image was shown due to space constraints. The lower-magnification image in Supplemental Fig. 3E demonstrates that the selected field was not chosen arbitrarily, providing context for the observed structures. In addition, Supplemental Fig. 3E includes both low- and high-magnification images of age-matched busulfan (-) testes as a control for the busulfan (+) condition, further supporting the validity of the comparison.

For the above-mentioned cases (Fig. 2D vs. Supplemental. 3A; Fig. 2F vs. Supplemental Fig. 3E), as well as other potentially overlapping figures (e.g., Fig. 3D vs. Supplemental Fig. 4C), we believe that the additional single-channel and lower-magnification images provide important context that cannot be fully conveyed in the main figures due to space limitations. Nevertheless, to address the reviewer’s concern, we will (i) clearly state the purpose of each Supplemental figure in the corresponding legends, and (ii) re-evaluate all figures to consolidate or remove any truly redundant panels. Our goal is to ensure that all figures collectively convey the data in the most concise and informative manner.

• Figure citations in the main text do not consistently match figure content. For example, on page 7 (lines 5-6), the text refers to Extended Data Fig. 4a for SOX9 staining. Yet, it is the extended Data Fig. 3a that contains the relevant data. Similarly, the reference to Extended Data Fig. 4b and 4c on page 7 (lines 7-8) for adult defects is inaccurate.

Response: We thank the reviewer for drawing attention to these inconsistencies. We have carefully checked all figure citations throughout the main text and corrected them so that they consistently match the figure content. The revised manuscript reflects these corrections.

• In Figure 2e, percentages of Sertoli cells across three layers are shown. The figure legend should specify which layer(s) show statistically significant differences between WT and KO.

Response: We are grateful to the reviewer for highlighting this point. Statistical comparisons were performed between Map7+/- and Map7-/- mice within each corresponding layer at P17. Statistical significance was assessed using Student’s t-test, and all three layers showed significant differences between Map7+/- and Map7-/- (P < 2.20 × 10⁻⁴). The figure legend has been revised accordingly as follows: “Statistical comparisons between Map7+/- and Map7-/- mice were performed for each corresponding layer at P17 using Student’s t-test. All three layers showed significant differences between Map7+/- and Map7-/- mice (*, P<2.20 × 10⁻⁴).” (page 28, lines 5-8)

• The current color scheme for F-actin and TUBB3 in Figure 3 lacks sufficient contrast. Adjusting to more distinguishable colors would improve readability.

Response: Response: We thank the reviewer for this helpful suggestion. In the original merged images, four channels (DNA, TUBB3, F-actin, and β-catenin) were displayed together, which reduced contrast between cytoskeletal signals. To improve clarity, we generated new merged images showing only TUBB3 and F-actin, allowing better visual distinction between these components. In addition, β-catenin and DNA are now displayed together as a separate merged image (β-catenin in yellow and DNA in blue) in the final column, highlighting the altered localization of β-catenin in Map7-/- testes.

• Since multiple scale bars with different units are present within the same figures, adding units directly above or beside each scale bar would improve readability.

Response: We thank the reviewer for the suggestion. Following this recommendation, we have added units directly above each scale bar in all figures to improve readability.

• It is recommended to directly mark Sertoli cells, spermatogonia, and spermatocytes on the SEM images in Figure 2C for clearer visualization.

Response: We thank the reviewer for the suggestion. We will follow this recommendation by performing segmentation and directly marking Sertoli cells, spermatogonia, and spermatocytes on the SEM images in Figure 2C to improve visualization.

• The quantification of Sertoli cell positioning shown in Fig. 2C is already described in the main text and is unnecessary in the figure.

Response: We appreciate the reviewer’s comment regarding the quantification of Sertoli cell positioning. Although the results are described in the main text, we believe that the visual presentation in Figure 2C is essential for conveying the spatial distribution pattern in an intuitive and comparative manner. To address the concern about redundancy, we have slightly revised the figure legend (page 27, lines 28–29) to clarify that this panel provides a visual summary of the quantitative data described in the text, thereby improving clarity without unnecessary duplication.

_Referee cross-commenting_

I concur with Reviewer 2 that the Map7-eGFP mouse model is a valuable tool for the research community. I also agree that performing MAP7-MYH9 double immunofluorescence staining to demonstrate their colocalization would further strengthen the authors' conclusions regarding their interaction. My overall assessment of the manuscript remains unchanged: the study represents an incremental advance that extends previous findings on MAP7 function but provides limited new mechanistic insight.

Reviewer #1 (Significance):

This study investigates the role of the microtubule-associated protein MAP7 in Sertoli cell polarity and apical domain formation during early stages of spermatogenesis. Using GFP-tagged and MAP7 knockout mouse models, the authors show that MAP7 localizes to apical microtubules and is required for Sertoli cell cytoskeletal organization and germ cell development. While the study identifies early Sertoli cell defects and candidate MAP7-interacting proteins, the mechanistic insights remain limited, and several conclusions require stronger experimental support. Overall, the discovery represents an incremental advance that extends prior findings on MAP7 function, providing additional but modest insights into the role of MAP7 in cytoskeletal regulation in male reproduction.

Response: We thank the reviewer for their constructive comments and thoughtful evaluation of our manuscript. We appreciate the positive feedback regarding the value of the Map7-egfpKI mouse model for the research community. We also thank the reviewer for the suggestion to perform MAP7–MYH9 double immunofluorescence staining to demonstrate colocalization, which we agree will further strengthen the mechanistic support.

We would like to clarify that several aspects of our findings represent novel contributions within a field where the mechanisms of microtubule remodeling during apical domain formation have remained largely unresolved. In particular, our study provides evidence that MAP7 is asymmetrically enriched at the apical microtubule network in Sertoli cells and contributes to the directional organization of these microtubules—an aspect of Sertoli cell polarity that has not been previously characterized. Our results further indicate that dynamic microtubule turnover, rather than stabilization alone, is required for proper apical domain formation, addressing a gap in current understanding of how microtubules are reorganized during early polarity establishment. In addition, the data support a role for MAP7 in coordinating microtubule and actomyosin organization, suggesting a scaffolding function that links these cytoskeletal systems. We also observe that Sertoli cell polarity can be functionally separated from cell identity and that disruptions in apical domain architecture precede delays in germ cell developmental progression. Taken together, these observations provide mechanistic insight that expands upon previous studies of MAP7 function at the cellular level.

The conclusions are supported by multiple, complementary lines of evidence, including knockout and Map7-egfpKI mouse models, high-resolution electron microscopy, immunoprecipitation–mass spectrometry, and single-cell RNA sequencing. While we agree that further experiments, such as MAP7–MYH9 double staining, will strengthen the mechanistic framework, we will also perform complementary biochemical analyses to provide additional insight. Specifically, we plan to conduct domain-mapping experiments to identify the MAP7 region required for MYH9 complex formation, coupled with co-immunoprecipitation assays in cultured cells to validate this association.

Although generating new mutant mouse lines is not feasible within the scope of this revision, and no in vitro system fully recapitulates Sertoli cell polarization, these complementary approaches will provide further mechanistic support. We believe that these planned experiments, together with the current dataset, will clarify the underlying mechanisms and reinforce the significance of our findings, while appropriately acknowledging the current limits of experimental evidence.

Reviewer #2 (Evidence, reproducibility and clarity):

In this manuscript the authors evaluate the role of Microtubule Associated Protein 7 (MAP7) in postnatal Sertoli cell development. The authors build two novel transgenic mouse lines (Map7-eGFP, Map7 knockout) which will be useful tools to the community. The transgenic mouse lines are used in paired advanced sequencing experiments and advanced imaging experiments to determine how Sertoli cell MAP7 is involved in the first wave of spermatogenesis. The authors identify MAP7 as an important regulator of Sertoli cell polarity and junction formation with loss of MAP7 disrupting intracellular microtubule and F-actin arrangement and Sertoli cell morphology. These structural issues impact the first wave of spermatogenesis causing a meiotic delay that limits round spermatid numbers. The authors also identify possible binding partners for MAP7, key among those MYH9.

The authors did a great job building a complex multi-modal project that addressed the question of MAP7 function from many angles. The is an excellent balance of using many advanced methods while still keeping the project narrowed, to use only tools to address the real questions. The lack of quality testing on the germ cells outside of TUNEL is disappointing, but the Conclusion section implies that this sort of work is being done currently so the omission in this manuscript is acceptable. However, there is an issue with the imaging portion of the work on MYH9. The conclusions from the MYH9 data is currently overstated, super-resolution imaging of Map7 knockouts with microtubule and F-actin stains, and imaging that uses MYH9 with either Map7-eGFP or anti-MAP7 are also needed to both support the MAP7-MYH9 interaction normally and lack of interaction with failure of MYH9 to localize to microtubules and F-actin in knockouts. Since a Leica SP8 was used for the imaging, using either Leica LIGHTNING or just higher magnification will likely be the easiest solution.

Response: We sincerely appreciate the reviewer’s thorough and positive evaluation of our study. We are encouraged that the reviewer recognized the overall strength of our multi-modal approach and the scientific value of the Map7-egfp knock-in and Map7 knockout genome-edited mouse models that we generated. We also thank the reviewer for highlighting the balance between methodological breadth and focused, hypothesis-driven investigation in our work.

Regarding the reviewer’s valuable comments on the imaging data, we have addressed them as follows. We improved the cytoskeletal imaging data as described in response to the reviewer’s minor comments. Specifically, in the revised Figure 3B, we replaced the original images with higher-resolution confocal images to provide a clearer view of cytoskeletal organization. In addition, following Reviewer #1’s suggestion, we modified the panel layout to enlarge each field and enhance the contrast between TUBB3 and F-actin channels, allowing better visualization of their altered localization in Map7-/- testes.

We agree that super-resolution imaging comparing control and Map7-/- testes stained for TUBB3 and F-actin would further strengthen the analysis. If the current resolution is still considered insufficient, we plan to perform additional imaging using a Carl Zeiss Airyscan or Leica Stellaris 5 system to further improve spatial resolution and confirm the observed cytoskeletal phenotypes. Finally, we will perform co-imaging of MYH9 with MAP7 to validate their spatial relationship under normal conditions, complementing the existing data obtained from Map7-/- testes.

This manuscript is nicely organized with almost all of the results spelled out very clearly and almost always paired with figures that make compelling and convincing support for the conclusions. There are minor revision suggestions for improving the manuscript listed below. These include synching up Figure and Supplemental Figure reference mismatches. There are also many minor, but important, details that need to be added to the Methods section including many catalog numbers and some references.

- Some of the imaging, especially Fig4F could benefit and be more convincing with super-resolution imaging in the 150nm range (SIM, Airyscan, LIGHTNING, SoRa) possibly even just imaging with a higher magnification objective (60x or 100x)

Response: We appreciate the reviewer’s suggestion to improve the resolution of the imaging data. In addition to revising Figure 3B as described above, we have also replaced the images in Figure 4F with higher-resolution confocal images to provide a clearer view of MYH9 localization relative to microtubules and F-actin. These revised images highlight that MYH9 specifically accumulates at apical regions where microtubules and F-actin intersect, forming the apical ES, but is not localized to the basal ES-associated F-actin structures. To retain spatial context and allow readers to appreciate the overall distribution pattern, the original lower-magnification images from Figure 4F have been moved to Supplemental Figure 5.

- SuppFig1D: Please add context in the legend to the meaning of the Yellow Stars and "O->U" labels. The latter would seem to be to indicate the Ovarian and Uterine sides of the image

Response: In response to this comment, we revised the figure legend to clarify the annotations. The legend now states: “O, ovary side; U, uterus side. Asterisks indicate secretory cells that lack planar cell polarity.”

- Pg6Line7: up to P23 or up to P35?

Response: We appreciate the reviewer’s attention to this detail. The text has been revised for clarity as follows: “To examine the temporal dynamics of Sertoli cell polarity establishment, we analyzed seminiferous tubule morphology across the first wave of spermatogenesis, from postnatal day (P)10 to P35. To specifically assess the role of MAP7 in Sertoli cells while minimizing contributions from germ cells, our analysis focused on stages up to P23, before MAP7 expression becomes detectable in step 9–11 spermatids (Fig. 1), to exclude potential secondary effects resulting from MAP7 loss in germ cells.” (page 6, lines 5-10)

- SuppFig4B: Does SuppFig4B reference back to Fig3B or Fig3C? If the latter please update this in the legend.

- Pg7Line21-23: Is SuppFig3D,E meant to be referenced and not SuppFig5A,B?

- Pg8Line22-25: Is SuppFig4A meant to be reference and not SuppFig5?

- Pg8Line34-Pg9Line: Is SuppFig4B meant to be reference and not SuppFig5B?

Response: We appreciate the reviewer’s careful reading. All mismatches in Supplemental figure references have been corrected, ensuring that each reference in the text now accurately corresponds to the appropriate data.

- Pg9Line28-33: Would the authors be willing to rework this figure to include images that more closely match the reported findings? The current version does not strongly support the idea that MYH9 fails to localize to microtubule and F-actin domains in Map7 knockout P17 seminiferous tubules. This could also just be a matter of acquiring these images at a higher magnification or with a lower-end (150nm range) super-resolution system (SIM, Airyscan, LIGHTNING, SoRa etc)

Response: Following the reviewer’s recommendation, we replaced the images in Figure 4F with higher-resolution confocal images to better visualize MYH9 localization relative to microtubules and F-actin in Map7+/- and Map7-/- testes. These revised images demonstrate that MYH9 specifically accumulates at apical regions where microtubules and F-actin intersect, but not at the basal ES-associated F-actin structures. To preserve spatial context, the original low-magnification images have been moved to Supplemental Figure 5. If additional resolution is required, we are prepared to acquire further images using an Airyscan or Stellaris 5 system.

- SuppFig7A: The legend notes these are P23 samples but the image label says 8W. Please update this to whichever is the correct age.

Response: We thank the reviewer for pointing out this discrepancy. The figure legend for Supplemental Figure 7A (now revised as Supplemental Figure 8A) has been corrected to indicate that the samples are from 8-week-old mice, consistent with the image label.

- Pg16Line4-5: Please include in the text the vendor and catalog number for the C57BL/6 mice

Response: The text now specifies: “C57BL/6NJcl mice were purchased from CLEA Japan (Tokyo, Japan)” (page 17, line 4). CLEA Japan does not assign catalog numbers to mouse strains.

- Pg16Line18-19: Please include in the text the catalog number for the DMEM

- Pg16Line19-20: Please include in the text the vendor and catalog number for the FBS

- Pg16Line20: Please include in the text the vendor and catalog number for the Pen-Strep

Response: We have added vendor and catalog information as follows: “Wild-type and MAP7-EGFPKI HeLa cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM, 043-30085; Fujifilm Wako Pure Chemical, Osaka, Japan) supplemented with 10% fetal bovine serum (FBS, 35-015-CV; Corning, Corning, NY, USA) and penicillin–streptomycin (26253-84; Nacalai, Kyoto, Japan) at 37 °C in a humidified atmosphere containing 5% CO₂ 18.” (page 17, lines 18-22)

- Pg17Line6-12: Thank you for including organized and detailed information about the primers, please also define the PCR protocol used including temperatures, timing, and cycles for Map7 knockout genotyping

- Pg17Line20-27: Thank you for including organized and detailed information about the primers, please also define the PCR protocol used including temperatures, timing, and cycles for Map7-eGFP genotyping

Response: The text has been updated to include the PCR conditions used for genotyping as follows: “Genotyping PCR was routinely performed as follows. Genomic DNA was prepared by incubating a small piece of the cut toe in 180 µL of 50 mM NaOH at 95 °C for 15 min, followed by neutralization with 20 µL of 1 M Tris-HCl (pH 8.0). After centrifugation for 20 min, 1 µL of the resulting DNA solution was used as the PCR template. Each reaction (8 µL total volume) contained 4 µL of Quick Taq HS DyeMix (DTM-101; Toyobo, Osaka, Japan) and a primer mix. PCR cycling conditions were as follows: 94 °C for 2 min; 35 cycles of 94 °C for 30 s, 65 °C for 30 s, and 72 °C for 1 min; followed by a final extension at 72 °C for 2 min and a hold at 4 °C. PCR products were analyzed using agarose gel electrophoresis. This protocol was also applied to other mouse lines and alleles generated in this study.” (page 18, lines 17–25)

- Pg17Line30: Please include in the text the vendor and catalog number for the Laemmli sample buffer

Response: We clarified that the buffer was prepared in-house.

- Pg17Line32&SuppTable1: Thank you for including an organized and detailed table for the primary antibodies used, please also make either a similar table or expand the current table to include secondary antibody information

- Pg17Line32: Please note in the text which primary antibodies and secondary antibodies from Supp Table 1

Response: Supplementary Table 1 has been updated to include both primary and HRP-conjugated secondary antibodies. In the Immunoblotting section of the Materials and Methods, we specified the antibodies used: “The following primary antibodies were used: mouse anti-Actin (C4, 0869100-CF; MP Biomedicals, Irvine, CA, USA), mouse anti-Clathrin heavy chain (610500; BD Biosciences, Franklin Lakes, NJ, USA), rat anti-GFP (GF090R; Nacalai, 04404-84), rabbit anti-MAP7 (SAB1408648; Sigma-Aldrich, St. Louis, MO, USA), rabbit anti-MAP7 (C2C3, GTX120907; GeneTex, Irvine, CA, USA), and mouse anti-α-tubulin (DM1A, T6199; Sigma-Aldrich). Corresponding HRP-conjugated secondary antibodies were used for detection: goat anti-mouse IgG (12-349; Sigma-Aldrich), goat anti-rabbit IgG (12-348; Sigma-Aldrich), and goat anti-rat IgG (AP136P; Sigma-Aldrich). Detailed information for all primary and secondary antibodies is provided in Supplementary Table 1.” (page 19, lines 14-22)

- Pg18Line2: Please include in the text the vendor and catalog number for the Bouin's

Response: The text has been updated to indicate that Bouin’s solution was prepared in-house

- Pg18Line3: Please include in the text the catalog number for the CREST-coated glass slides

- Pg18Line7: Please include in the text the catalog number for the OCT compound

- Pg18Line11: Please include in the text the vendor and catalog number for the Donkey Serum

- Pg18Line11: Please include in the text the vendor and catalog number for the Goat Serum

Response: The text now includes vendor and catalog information for all these reagents, including CREST-coated slides (SCRE-01; Matsunami Glass, Osaka, Japan), OCT compound (4583; Sakura Finetechnical, Tokyo, Japan), donkey serum (017-000-121; Jackson ImmunoResearch Laboratories, PA, USA), and goat serum (005-000-121; Jackson ImmunoResearch Laboratories).

- Pg18Line13: Thank you for including an organized and detailed table for the primary antibodies used, please also make either a similar table or expand the current table to include secondary antibody information

Response: We thank the reviewer for the suggestion. Supplementary Table 1 already includes information for the antibodies used for immunoblotting, and we have now added information for the Alexa Fluor-conjugated secondary antibodies used for immunofluorescence in this study.

- Pg18Line18: Please include in the text the vendor and catalog number for the DAPI

Response: The text has been updated to include the vendor and catalog number for DAPI (D9542; Sigma-Aldrich).

- Pg18Line19: Please also include information about the objectives used including catalog numbers, detectors used (PMT vs HyD)

Response: We thank the reviewer for the suggestion. The following information has been added to the Histological analysis section in Materials and Methods: “Objectives used were HC PL APO 40×/1.30 OIL CS2 (11506428; Leica) and HC PL APO 63×/1.40 OIL CS2 (11506350; Leica), with digital zoom applied as needed for high-magnification imaging. DAPI was detected using PMT detectors, while Alexa Fluor 488, 594, and 647 signals were captured using HyD detectors. Images were acquired in sequential mode with detector settings adjusted to prevent signal bleed-through.” (page 20, lines 13-17)

- Pg18Line23: Please cite in the text the reference paper for Fiji (Schindelin et al. 2012 Nature Methods PMID: 22743772) and note the version of Fiji used

- Pg18Line24: Please note the version of Aivia used

Response: We have revised the text accordingly by citing the reference paper for Fiji (Schindelin et al., 2012, Nature Methods, PMID: 22743772) and noting the version used (v.2.16/1.54p). In addition, we have added the version of Aivia used in this study (version 14.1).

- Pg18Line25: If possible, please use a more robust and reliable system than Microsoft Excel to do statistics (Graphpad Prism, Stata, R, etc), if this is not possible please note the version of Microsoft Excel used

Response: We appreciate the reviewer’s suggestion. For basic statistical analyses such as the Student’s t-test, we used Microsoft Excel (Microsoft Office LTSC Professional Plus 2021), which has been sufficient for these standard calculations. For more advanced analyses, including ANOVA and single-cell RNA-seq analyses, we used R. These details have now been added to the text.

- Pg18Line25: Please cite in the text the reference paper for R (R Core Team 2021 R Foundation for Statistical Computing "R: A Language and Environment for Statistical Computing") and note the version of R used

- Pg18Line25: Please note the specific R package with version used to do ANOVA, and cite in the text the reference for this package

Response: We have cited the reference for R (R Core Team, 2021. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria) and noted the version used (version 4.4.0) in the text. In addition, regarding ANOVA, we have added the following description: “For ANOVA analysis, linear models were fitted using the base stats package (lm function), and analysis of variance was conducted with the anova function.” (page 20, lines 23-25)

- Pg18Line25: Please clarify, was a R package called "AVNOVA" used to do ANOVA or is this a typo?

Response: We thank the reviewer for pointing this out. It was a typographical error — the correct term is “ANOVA”. The text has been corrected accordingly.

- Pg18Line32: Please include in the text the catalog number for the EPON 812 Resin

- Pg19Line3: Please include the version number for Stacker Neo

- Pg19Line5: Please include the vendor and version number for Amira 2022

- Pg19Line5: Please include the version number for Microscopy Image Browser

- Pg19Line5: Please include the version number for MATLAB that was used to run Microscopy Image Browser

Response: We added the catalog number for the EPON 812 resin and the vendor and version information for the software used. The following details have been included in the revised text:

EPON 812 resin: TAAB Embedding Resin Kit with DMP-30 (T004; TAAB Laboratory and Microscopy, Berks, UK)

Stacker Neo: version 3.5.3.0; JEOL

Amira 2022: version 2022.1; Thermo Fisher Scientific

Microscopy Image Browser: version 2.91

Note that although Microscopy Image Browser is written in MATLAB, we used the standalone version that does not require a separate MATLAB installation.

- Pg19Line: 9-10: Please include in the text the catalog number for the complete protease inhibitor

- Pg19Line14: Please include in the text the catalog number for the Magnetic Agarose Beads

- Pg19Line16: Please include in the text the catalog number for the GFP-Trap Magnetic Agarose Beads

Response: We have added the catalog numbers for the complete protease inhibitor (4693116001), control magnetic agarose beads (bmab), and GFP-Trap magnetic agarose beads (gtma).

- Pg19Line21: Please note in the text which primary antibodies and secondary antibodies from Supp Table 1

- Pg19Line21-22: Please include in the text the catalog number for the ECL Prime

Response: We thank the reviewer for the helpful suggestions. The description regarding immunoblotting (“Eluted samples were separated by SDS–PAGE, transferred to PVDF membranes…”) was reorganized: overlapping content has been removed, and the necessary information has been integrated into the “Immunoblotting” section, where details of the primary and secondary antibodies (listed in Supplementary Table 1) are already provided. In addition, the information for ECL Prime has been updated to “Amersham ECL Prime (RPN2236; Cytiva, Tokyo, Japan)”.

- Pg20Line2: Please include the version number for Xcalibur

Response: The version of Xcalibur used in this study (version 4.0.27.19) has been added to the text.

- Pg20Line5: Please cite in the text the reference paper for SWISS-PROT (Bairoch and Apweiler 1999 Nucleic Acid Research PMID: 9847139)

Response: The reference paper for SWISS-PROT (Bairoch and Apweiler, 1999, Nucleic Acids Research, PMID: 9847139) has been cited in the text.

- Pg19Line26: Please include in the text the catalog number for the NuPAGE gels

- Pg19Line28: Please include in the text the catalog number for the SimpleBlue SafeStain

Response: Both catalog numbers have been added in the Mass spectrometry section as follows: 4–12% NuPAGE gels (NP0321PK2; Thermo Fisher Scientific) and SimplyBlue SafeStain (LC6060; Thermo Fisher Scientific).

- Pg20Line26: Please include in the text the catalog number for the Chromium Singel Cell 3' Reagent Kits v3

Response: The catalog number for the Chromium Single Cell 3′ Reagent Kits v3 (PN-1000075; 10x Genomics) has been added to the text.

- Pg21Line3: Please cite in the text the reference paper for R (R Core Team 2021 R Foundation for Statistical Computing "R: A Language and Environment for Statistical Computing")

Response: The reference for R (R Core Team, 2021. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria) has already been cited in the “Histological analysis” section, where ANOVA analysis is described.

- Pg21Line3 Please cite in the text the reference for RStudio (Posit team (2025). RStudio: Integrated Development Environment for R. Posit Software, PBC, Boston, MA. URL http://www.posit.co/.)

Response: The reference for RStudio (Posit team, 2025. RStudio: Integrated Development Environment for R. Posit Software, PBC, Boston, MA, USA. URL: http://www.posit.co/) has been added to the text.

- Pg21Line23: Please include the version number for Metascape

Response: The version of Metascape used in this study (v3.5.20250701) has been added to the text.

- SuppFig12: please update the legend to include a description after the title and update the figure labeling to correspond to the legend. Also, this figure is currently not referenced anywhere in the text.

Response: We have updated the legend for Supplemental Figure 12 (Supplemental Figure 13) to include a descriptive sentence after the title and have adjusted the figure labeling to match the legend. The revised legend now reads: “Full-scan images of the agarose gels shown in Supplemental Figs. 1B and 2C are displayed in the upper and lower left panels, respectively, while the corresponding full-scan images of the immunoblots shown in Supplemental Figs. 1C and 2D are presented in the upper and lower right panels, respectively.”

As these images serve as source data, they are not referenced directly in the main text.

_Referee cross-commenting_

I generally agree with Reviewer 1 and specifically concur related to adding details about fertility assessment of the Map7 Knockout line, and enhancing the SEM imaging.

Response: As noted in our response to Reviewer #1, we have re-acquired the SEM images in high-resolution mode, focusing on the relevant regions. The new high-resolution images have replaced the original panels in revised Figure 3C, providing clearer visualization of junctional structures at P10 and P21 in Map7+/- and Map7-/- testes. The original Figure 3C images have been moved to Supplemental Figure 4B for reference.

Reviewer #2 (Significance):

There are mouse lines, and datasets that will be useful resources to the field. This work also advances our understanding of a period in Sertoli cell development that is critical to fertility but very understudied.

Response: We thank the reviewer for the positive comments and for recognizing the potential value of our mouse lines and datasets to the field, as well as the significance of our work in advancing the understanding of this critical but understudied period in Sertoli cell development.

self-knowledge -> rational control, liberate from outside things that are out of control

Learn how to select the best mobile app development company in Riyadh with our expert guide. Discover the 7 essential questions to ask before hiring to ensure quality, security, and ROI in your app project.

Version 4 of this preprint has been peer-reviewed and recommended by Peer Community in Evolutionary Biology.<br /> See the peer reviews and the recommendation.

The Art of Learning by Josh Waitzkin, wrt learning chess.

Bad advice. Safe to fail probes are not 'unserious', bc of aiming to limit the downside should they fail. The point is having the probe to explore 'cheaply' within the actual context of work. First part, the interest should be genuine, is completely true. It does not mean you can't do risk containment though. Probes are real and safe simultaneously. [[Probe proberend handelen 20201111162752]]

How to Start a Zettelkasten When You Are Stuck in Theory • Zettelkasten Method<br /> by [[sascha]]<br /> accessed on 2025-12-11T14:08:33

Meh...

one of the shortest routes, but it also presumes a knowledge of too much other theory and is geared at getting one moving at creating rather than reading the theory

Note: This response was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Reply to the reviewers

Reviewer #1 (Evidence, reproducibility and clarity):

Summary:

The manuscript titled "Unravelling the Progression of the Zebrafish Primary Body Axis with Reconstructed Spatiotemporal Transcriptomics" presents a comprehensive analysis of the development of the primary body axis in zebrafish by integrating bulk RNA-seq, 3D images, and Stereo-Seq. The authors first clearly demonstrate the application of Palette for integrating RNA-seq and Stereo-Seq using published spatial transcriptomics data of Drosophila embryos. Subsequently, they produced serial bulk RNA-seq data for certain developmental stages of Danio rerio embryos and utilized published Stereo-Seq data. Through robust validation, the authors observe the molecular network involved in AP axis formation. While the authors show that integrating bulk RNA-seq data with Stereo-Seq improves spatial resolution, additional proof is required to demonstrate the extent of this improvement.

Response: We thank the reviewer for the positive feedback on our Palette pipeline, zSTEP construction and analysis of primary body axis development. We appreciate the constructive suggestions provided, which we can implement to improve our manuscript. As pointed out by the reviewer, some analysis procedures were not described in sufficient detail. To address this, we have added more explanatory texts and additional schematic diagrams to make the methods clearer and more understandable. We also thank the reviewer for the meticulous reading and for reminding us to include parameters, references and essential texts, which significantly improve the manuscript quality and make the manuscript more rigorous. Furthermore, as suggested by the reviewer, the extent of the improvement on the spatial resolution was not clearly demonstrated in the manuscript. Therefore, we have provided an additional figure to show the original expression on the stacked Stereo-seq slices and 3D live image compared to the expression from zSTEP, and the results indicate that zSTEP provides better, more continuous expression patterns. We still have two remaining tasks that are expected to be completed within the next month. We hope our responses have address the concerns raised by the reviewer, and we are pleased to provide any additional proof as needed.

Major Comments:

1. Lines 66-68: Discuss the limitations of existing tools and explicitly state the advantages of using Palette.

Response: We thank the reviewer for the valuable suggestion. We have added the following new texts after line 68 to emphasize the features and advantages of Palette.

"Newly developed tools are committed to integrating bulk and/or scRNA-seq data with ST data to enhance spatial resolution, focusing on expression at the spot level. However, gene expression patterns are closely correlated to the biological functions and are more critical for understanding biological processes. Therefore, a tool focusing on inferring spatial gene expression patterns would be desirable."

1. Body Pattern Genes Analysis: For both Drosophila and Danio rerio, it would be valuable to examine body pattern genes in Stereo-Seq and apply Palette to determine if the resolution of the segments improves or merges. The resolution of the A-P axis is convincing, but further evidence for other segments would be beneficial.

Response: We thank the reviewer for the suggestions. For the Drosophila data, we only used two adjacent slices for Palette performance assessment, and thus were only able to evaluate the expression patterns within the slice.

For the zebrafish data, although we have construct zSTEP as a 3D transcriptomic atlas, we have to admit that the left-right (LR) and dorsal-ventral (DV) patterning is not satisfactory enough. Here we show a section from the dorsal part of 16 hpf zSTEP that displays a relatively well-defined left-right pattern (Fig. 2). Along the left-right axis, the notochord cells are centrally located, flanked by somite cells on either side, with the outermost cells being pronephros.

One reason for the limited LR and DV patterning is that the original annotation of the ST data does not clearly distinguish all the cell types. Another reason is likely due to the disordered cell positions when stacking ST slices. Thus, our zSTEP is most suitable for investigating the AP patterns, while the performances on LR and DV patterns may not achieve the same level of accuracy.

See response letter for the figure.

1. Figure 2d: Include the A-P line for which the intensity profile was plotted in the main figure, rather than just in the supplementary material. Additionally, consider simplifying the plot by not combining three lines into one, as it complicates the interpretation of observations.

Response: We thank the reviewer for the helpful suggestions. We have updated Figure 2d and Figure S1b by adding a A-P line on each subfigure (Fig. 3). Additionally, as the reviewer suggested, we have separated the intensity plots so that each subfigure now includes a dedicated intensity plot along A-P axis.

See response letter for the figure.

1. Drosophila Data Analysis: While the alignment and validation of Danio rerio sections are clearly explained, the analysis and validation of Drosophila data are insufficiently detailed. Provide a more thorough explanation of how the intensity profiles between BDGP in situ data and Stereo-Seq data are adjusted.

Response: We thank the reviewer for raising this issue. To make the analysis procedure clearer, we have updated Figure 2a (Fig. 4) and added explanatory texts in the figure legends to describe the processing procedure for the Drosophila ST data.

See response letter for the figure.

Additionally, the following sentences have been added into the Methods section to describe the generation of the intensity profiles.

"The intensity plot profiles along AP axis were generated through the following steps: The expression pattern plot images or in situ hybridization images were imported into ImageJ and converted to grayscale. The colour was then inverted, and a line of a certain width (here set as 10) was drawn across from the anterior part to the posterior part (Fig. S1a). The signal intensities along the width of the line were measured and imported into R for generating intensity plots."

1. Figure 3d: Present a plot with the expected expression profiles of the three genes if the embryo is aligned as anticipated.

Response: We thank the reviewer for this helpful suggestion, which improves the clarity of our manuscript. We have added the following subfigure in as Figure 3d (Fig. 5) to show the expected expression profiles of the three midline genes along left-right axis.

See response letter for the figure.

1. Analysis Without Palette: Between lines 277-438, the outcome of using Palette with bulk RNA-seq and Stereo-Seq is convincing. However, consider the following:

o What would be the observations if the analysis were conducted solely with Stereo-Seq data, without incorporating bulk RNA-seq data and employing Palette?

Response: We thank the reviewer for raising this important question. Here we show the comparison of ST expression on stacked Stereo-seq slices, ST expression projected on 3D live images, and the Palette-inferred expression (Fig. 6). The stacked ST slices do not fully reflect the zebrafish morphology, and the gene expression appears sparse, making it look massive (the first row). While after projecting ST expression onto the live image, the expression patterns can be observed on zebrafish morphology, but the expression is still sparsely distributed in spots (the second row). However, the expression patterns captured by Palette in zSTEP show more continuous expression patterns (the third row), which are more similar to the observations in in situ hybridization images (the fourth row). We are considering put these analyses into the supplementary figure.

See response letter for the figure.

o This study uses only Stereo-Seq as the spatial transcriptomics reference. It would strengthen the argument to use at least one other spatial transcriptomics method, such as Visium or MERFISH, in conjunction with bulk RNA-seq and Palette, to demonstrate whether Palette consistently improves gene expression resolution.

Response: We thank the reviewer for raising this professional question. To demonstrate a broad application of Palette, it would be necessary to test Palette performance using different types of ST references. We plan to perform extra analyses to evaluate Palette performance using Visium and MERFISH data as ST references, respectively. Additionally, our Palette pipeline only takes the overlapped genes for inference. As only hundreds of genes can be detected by MERFISH, Palette can only infer the expression patterns of these genes. As mentioned in the work of Liu et al. (2023), MERFISH can independently resolve distinct cell types and spatial structures, and thus we believe Palette will also show great performance when using MERFISH as ST reference. We've already started the analyses and expect to accomplish it within the next month. And we will update the analyses as separated tutorials to the GitHub repository.

Reference:

Liu, J. et al. Concordance of MERFISH spatial transcriptomics with bulk and single-cell RNA sequencing. Life Sci Alliance 6 (2023).

1. PDAC Data Analysis: Provide a more detailed explanation of the PDAC data analysis and use appropriate colors in the tissue images to clearly distinguish cell types.

Response: We thank the reviewer for the suggestions. We have updated the colours used in the tissue images to be consistent to the colours in tissue clustering analysis. Additionally, we have added an additional subfigure in supplementary figure (Fig. 7) with more explanatory texts in the figure legends to provide a more thorough explanation for the analysis.

See response letter for the figure.

1. Comparison with Other Methods: State the limitations of not using STitch3D and Spateo for alignment and explain why these methods were not employed.

Response: We thank the reviewer for raising this constructive comment. We fully agree with you that the introduction of published alignment algorithms would be helpful in our analysis. Currently, the slice alignment is adjusted manually, and thus the main limitation of not using these tools is that manual operation may induce bias compared to the alignment generated by computational algorithm. Unfortunately, STitch3D and Spateo are not included in this study because of two reasons. First, these two newly developed tools have been recently posted, and our analyses were largely completed before that. Therefore, we only mentioned these tools in the Discussion section. Second, we do not want to embed too many external tools into our analysis, which may increase the difficulties for researchers' operation. Specifically, STitch3D and Spateo are configured to run in Python environment, while Palette is based on R packages. Moreover, without these tools, our current manual alignment also achieves desired performance. However, we value this enlightening suggestion by the reviewer and therefore plan to further compare the performance of manual alignment versus the mentioned two alignment tools. At present, we have a preliminary comparison scheme and collected relevant datasets. Hopefully, we will complete this analysis within the next 1 to 2 weeks.

Minor Comments:

1. References: Add references to the statements in lines 51-53.

Response: We thank the reviewer for reminding us of the missing references. We have added the works of Junker et al. (2014), Liu et al. (2022), Chen et al. (2022), Wang et al. (2022), Shi et al. (2023) and Satija et al. (2015) as references in line 53 as follows.

"Thus, great efforts are ongoing to construct gene expression maps of these models with higher resolution, depth, and comprehensiveness1-6."

References:

1. Junker, J.P. et al. Genome-wide RNA Tomography in the zebrafish embryo. Cell 159, 662-675 (2014).
2. Liu, C. et al. Spatiotemporal mapping of gene expression landscapes and developmental trajectories during zebrafish embryogenesis. Dev Cell 57, 1284-1298 e1285 (2022).
3. Chen, A. et al. Spatiotemporal transcriptomic atlas of mouse organogenesis using DNA nanoball-patterned arrays. Cell 185, 1777-1792 e1721 (2022).
4. Wang, M. et al. High-resolution 3D spatiotemporal transcriptomic maps of developing Drosophila embryos and larvae. Dev Cell 57, 1271-1283 e1274 (2022).
5. Shi, H. et al. Spatial atlas of the mouse central nervous system at molecular resolution. Nature 622, 552-561 (2023).
6. Satija, R. et al. Spatial reconstruction of single-cell gene expression data. Nature biotechnology 33, 495-502 (2015)

1. Scientific Name Consistency: Ensure consistency in using either "Danio rerio" or "zebrafish" throughout the manuscript.

Response: We thank the reviewer for this suggestion. We have changed "Danio rerio" to "zebrafish" to make "zebrafish" consistent throughout the manuscript.

1. Related References: Include the following relevant references:

o https://academic.oup.com/bib/article/25/4/bbae316/7705532

o https://www.life-science-alliance.org/content/6/1/e202201701

Response: We thank the reviewer for bringing these two relevant works to us. Baul et al. (2024) presented STGAT leveraging Graph Attention Networks for integrating spatial transcriptomics and bulk RNA-seq, and Liu et al. (2023) demonstrated the concordance of MERFISH ST with bulk and single-cell RNA-seq. Both are excellent works and relevant to our work. We have added these two references in line 61 and line 68, respectively.

References:

Baul, S. et al. Integrating spatial transcriptomics and bulk RNA-seq: predicting gene expression with enhanced resolution through graph attention networks. Brief Bioinform 25 (2024).

Liu, J. et al. Concordance of MERFISH spatial transcriptomics with bulk and single-cell RNA sequencing. Life Sci Alliance 6 (2023).

1. Figure 1a: In the Venn diagram, include the number of genes in the bulk and Stereo-Seq datasets, as well as the number of overlapping genes.

Response: We thank the reviewer reminding us to include these important numbers. And in our current manuscript, we have added the following sentences in the Methods section to provide the gene numbers (Fig. 8). While the Venn diagram in Figure 1a serves as a schematic representation, so we did not include the gene numbers, as these may vary depending on the actual data.

"Palette was performed on the aligned slices using the overlapped genes. For the 10 hpf embryo, there were 24,658 genes in the bulk data, 18,698 genes in the Stereo-seq data, and 16,601 overlapped genes. For the 12 hpf embryo, there were 23,018 genes in the bulk data, 18,948 genes in the Stereo-seq data, and 16,401 overlapped genes. For the 16 hpf embryo, there were 24,357 genes in the bulk data, 23,110 genes in the Stereo-seq data, and 19,539 overlapped genes."

See response letter for the figure.

1. Figure 1 Improvement: Enlarge Figure 1 and reduce repetitive elements, such as parts of the deconvolution and Figure 1b.

Response: We thank the reviewer for the helpful suggestion. We agree with the reviewer that the deconvolution sections appear repetitive. We have updated Figure 1 (Fig. 9) by replacing these repetitive elements with a clearer and simpler diagram.

See response letter for the figure.

1. Figure 3f: Explain the black discontinuous line in the plot.

Response: We thank the reviewer for the reminder. We are sorry about the lack of the explanation. We have added the below explanation for the black discontinuous line in the legend of Figure 3 (Fig. 10) as follows.

See response letter for the figure.

1. Line 610: State the percentage of unpaired imaging spots.

Response: We thank the review for the reminder. We are sorry about not including the paired and unpaired spot number. We have added the number of paired spots with the percentage in the total spots in the Method section as follows.

"The numbers of mapped spots for the 10 hpf, 12 hpf and 16 hpf embryos are 15,379 (69.4% of the total spots), 14,697 (70.5% of the total spots) and 21,605 (77.2% of the total spots), respectively."

1. Lines 616-618: Specify the unit for the spot diameter.

Response: We thank the reviewer for the reminder. Again, we are sorry about not including the spot diameter information in our previous version of manuscript. We have added the spot diameter in Method section as follows.

"In the Stereo-seq data, each spot contained 15 × 15 DNA nanoball (DNB) spots (The diameter of each spot is near 10 μm)."

Reviewer #1 (Significance):

This algorithm will be useful not only for the field of developmental biology but also for wider applications in spatial omics. Although I have expertise in spatial omics technology development, my understanding of computational biology is limited, which restricts my ability to fully evaluate the Palette algorithm presented in this paper.

Response: We thank the reviewer for recognizing our work, and we greatly appreciate the constructive suggestions from the reviewer. Although the reviewer acknowledged limited expertise in computational biology, the comments from the reviewer are highly professional and valuable. Following the suggestions from the reviewer, we have not only included more explanatory texts and figures to make the analysis procedures clearer and more understandable, but also supplemented the important parameters that were missing in our previous manuscript. We also provided extra figure to demonstrate the improvements of zSTEP on gene expression patterns. We believe that our work is now more scientific and more understandable, and we will continue working to solve the remaining issues as planned. We express our thanks for the reviewer again.

Reviewer #2 (Evidence, reproducibility and clarity):

The authors of the study introduce the Palette method, a novel approach designed to infer spatial gene expression patterns from bulk RNA-sequencing (RNA-seq) data. This method is complemented by the development of the DreSTEP 3D spatial gene expression atlas of zebrafish embryos, establishing a comprehensive resource for visualizing gene expression and investigating spatial cell-cell interactions in developmental biology.

Response: We sincerely appreciate the reviewer's positive feedback on our Palette pipeline and the zSTEP 3D spatial expression atlas of zebrafish embryos. We also thank the reviewer for the professional comments and constructive suggestions. The reviewer raised the concerns from the aspect of algorithm design and computational biology, which we did not address well in our previous manuscript. We agree with the reviewer that we did not clarify the selection criteria of the parameters in detail, and we are now working on the additional analyses to address this issue.

We also agree with the reviewer that we did not provide enough discussion of the strategies used in the pipeline, the features of Palette and the application scenarios of Palette and zSTEP. For wide use of our tools, it is significantly important to state these aspects. In this revised version, we have added more paragraphs in the Discussion section to address this issue. Additionally, we acknowledge that we did not adequately demonstrate the computational efficacy and computational requirements, which are important for researchers. We are also working on the additional analyses to address this issue.

Finally, we thank the reviewer again for the professional and constructive suggestions. These suggestions are addressable, and by following them, we believe our manuscript will see a significant improvement, especially in the Palette pipeline part, making the pipeline more rigorous and easier to access. We are confident that we can complete the planned additional tasks within the next 1-2 months.

1. The efficacy of the Palette method may be compromised by its dependency on the quality of the reference spatial transcriptomics data. As highlighted in the study, variations in data quality can lead to significant challenges in reconstructing accurate spatial expression patterns from bulk data. This underscores the necessity of evaluating quality parameters, such as the number of gene detections and spatial resolution, to ensure reliable outcomes. Additional studies should rigorously assess how these quality factors influence the accuracy and efficiency of the algorithm in various data contexts, particularly under diverse conditions of gene detection.

Response: We thank the reviewer for this valuable suggestion. We agree with the reviewer that the quality of the reference ST data may greatly influence the performance and efficacy of the Palette, and we have added paragraphs in the Discussion section to further discuss the impact of ST data quality on Palette performance. As mentioned by the reviewer, gene detections and spatial resolution are two important parameters that can influence the Palette performance. Low gene detection may impact the clustering process, making the cell types of spots not distinguished well. To evaluate the performance of Palette when ST data shows low gene detection, we plan to applied Palette using MERFISH data as the ST reference, which only captures hundreds of genes. Moreover, we will also investigate the impact of spatial resolution on Palette performance by merging ST spots to simulate lower resolution scenarios, as well as the impact of gene detection by randomly reducing detected genes. Through the comparison among the inferred expression patterns with ST data of different spatial resolutions or different numbers of detected genes, we can better access the performance of Palette and provide guidance to researchers on the appropriate ST data requirements for optimal performance. These analyses will take another one month to accomplish after this round of revision due to the limited response time.

1. The methodology raises pertinent questions regarding how the clustering results from different algorithms may affect the reconstructions by the Palette method. The authors would better provide a detailed discussion/comparison of clustering processes that optimize the reconstruction of spatial patterns, ensuring precision in the downstream analyses.

Response: We thank the reviewer for the constructive comments. We agree with the reviewer that the differences in clustering results would impact the inference of the Palette. In our Palette pipeline, rather than develop a new methodology for clustering, we employ the BayesSpace for spot clustering, which considers both spot transcriptional similarity and neighbouring structure for clustering. In this case, researchers may adjust the parameters in the BayesSpace package to achieve optimal clustering results. Actually, in most cases, the spot identities were achieved through UMAP analysis, which only considers the transcriptional differences but does not consider the spatial information. This kind of clustering strategy will potentially lead to an intricate arrangement of spots belonging to different clusters, and may result in sparse gene expression in Palette outcome, which is different from the patterns in bona fide tissues. Therefore, a suitable clustering strategy will definitely help capture the local patterns.

Moreover, our Palette pipeline also can use the clustering results from the tissue histomorphology. Using tissue histomorphology for clustering would be a good choice, as it is closer to the real case. The following Figure (Fig. 11) displays the Palette performance on PDAC datasets using both spatial clustering and histomorphology clustering strategies. The result using histomorphology clustering captures the weak pattern (indicated by the red circle) that were missed when using the spatial clustering (Fig. 11d).

See response letter for the figure.

1. The choice to utilize only highly expressed genes in the initial stages of the Palette algorithm also warrants further exploration. Addressing the criteria for determining which genes qualify as "highly expressed" and outlining robust cutoff will enhance the algorithm's rigor and applicability. Similarly, in the iterative estimation of gene expression across spatial spots, establishing optimal iteration conditions is crucial. Implementing a loss function may offer a systematic method for concluding iterations, thus refining computational efficiency.

Response: We thank the reviewer for the professional suggestions. As pointed out by the reviewer, the selection of highly expressed genes and the iteration times are two important parameters in our pipeline. The definition of highly expressed genes and the number of highly expressed genes are important for achieving a satisfactory clustering performance. We tested the impact of different numbers of highly expressed genes on cluster performance in our preliminary analyses, while we did not summarize these tests and specify the parameters. Therefore, we plan to include a supplementary figure showing the clustering performances under different definitions of highly expressed genes and different numbers of highly expressed genes. Additionally, for the iteration conditions, we have tested different iteration numbers to find out a suitable iteration number to achieve a stable expression in each spot. The following figure (Fig. 1) shows the results after performing Palette with different iteration times. We randomly selected 20 cells and compared their expression across tests with varying iteration times. The results indicate that for a ST dataset with 819 spots, the expression in each spot becomes nearly stable after 5000 iteration times. We previously did not consider the computational efficiency, while here the reviewer raises a valuable and professional suggestion to implement a loss function to determine the optimal number of iterations. We greatly appreciate this suggestion, and plan to apply a loss function to summarize the optimal iteration times for ST datasets of different sizes. This will provide guidance for potential researchers in selecting iteration times and enhance computational efficiency.

See response letter for the figure.

1. Performance metrics relating to processing speed and computational demands remain inadequately addressed in the current framework. Understanding how the Palette method scales across varying gene counts and bulk RNA-seq datasets will be essential for potential applications in larger biological contexts. Notably, the quantitative demands of analyzing 20,000 genes when processing 10, 100, or 1,000 bulk RNA profiles must be articulated to guide researchers in planning accordingly.

Response: We thank the reviewer for this valuable and professional suggestion. In our previous analyses, we did not consider the computation efficiency, processing speed and computational demands, which are important information for potential researchers. To address this issue, we will list our computer configuration first. And under this configuration, we plan to run Palette on datasets with different numbers of overlapped genes or ST references with varying spot numbers, and then summarize the running times into a metrics table. This will help researchers estimate the running time for their datasets and guide them in planning the analyses. We will begin the analyses soon and expect to complete the analysis within the next 1 to 2 months.

Minor opinions:

1. Despite the promising advances offered by the zebrafish 3D reconstruction, there is a lack of details regarding numbers of the spatial transcriptomics (ST) data utilized, and the number of bulk RNA-seq data employed in the analyses. These parameters need to be clarified.

Response: We thank the reviewer for reminding us of these parameters. We are sorry for not including these parameters in our previous manuscript. We have now included the numbers of bulk, ST and overlap genes in the Methods section as follows (Fig. 12).

"Palette was performed on the aligned slices using the overlapped genes. For the 10 hpf embryo, there were 24,658 genes in the bulk data, 18,698 genes in the Stereo-seq data, and 16,601 overlapped genes. For the 12 hpf embryo, there were 23,018 genes in the bulk data, 18,948 genes in the Stereo-seq data, and 16,401 overlapped genes. For the 16 hpf embryo, there were 24,357 genes in the bulk data, 23,110 genes in the Stereo-seq data, and 19,539 overlapped genes."

See response letter for the figure.

1. Issues regarding spatial cell-cell communication, especially concerning interactions over longer distances, necessitate careful consideration. Introducing spatial distance constraints could help formulate more realistic models of cellular interactions, a vital aspect of embryonic development.

Response: We thank the reviewer for this essential comment. We agree with the reviewer that the spatial distance is an essential factor to investigate in vivo cell-cell communication during embryonic development. Therefore, in our analyses, we employed CellChat for spatial cell-cell communication analysis, which can be used to infer and visualize spatial cell-cell communication network for ST datasets, considering the spatial distance as constrains of the computed communication probability. However, during our analyses, we observed that there were interactions between cell types over longer distances, as mentioned by the reviewer. We then investigated how these interactions of longer distances occurred. Here, we show the FGF interaction between tail bud and neural crest cells from our spatial cell-cell analysis as an example, and the distance between these two cell types appears quite significant (Fig. 13). We labelled tail bud cells and neural crest cells on the selected midline section and observed that, although most neural crest cells are distributed anteriorly, a small number of neural crest cells are located at tail, close to the tail bud cells. Therefore, the observed interaction between tail bud and neural crest cells is likely due to their adjacent distribution in the tail region, while the anteriorly distributed of neural crest spot in spatial cell-cell communication analysis reflects the anterior positioning of most neural crest cells. As a result, the distances shown on the spatial cell-cell communication analysis are not the real distance between two cell types.

In most cases in our spatial cell-cell communication analyses, the observed interactions over longer distances are likely influenced by this visualization strategy. Additionally, pre-processing the dataset may enhance the performance of the analyses. Here we performed systematic analyses of the entire embryo, which can make the interactions between cell types appear massive. To investigate specific biological questions, researchers can subset cell types of interest or categorize them into different subtypes based on their positions.

See response letter for the figure.

1. Evaluation metrics such as the Adjusted Rand Index (ARI) and Root Mean Square Error (RMSE) represent critical tools for systematically measuring the similarity of inferred spatial patterns, yet their specific application within this context should be elaborated.

Response: We thank the reviewer for recommending these two tools. We have applied them to evaluate the similarity between the expression patterns (Fig. 14). The inclusion of these statistical values makes our comparisons of expression patterns more scientific and convincing. And we have added the following texts in the Methods section to describe the calculation of these two values.

"The Adjusted Rand Index (ARI) and Root Mean Square Error (RMSE) were used to evaluate the similarity of the expression patterns. The expression patterns of in situ hybridization images were considered as the expected values, and the expression patterns of ST data and inferred expression patterns were compared to the expected values. Common positions along the AP axis within all three expression profiles were used, and the RMSE were calculated based on the scaled intensity of these positions. Values greater than the threshold were set to 1; otherwise, they were set to 0, and the ARI was then calculated based on the intensity category. Higher ARI and lower RMSE indicate greater similarity."

See response letter for the figure.

1. The study's limitations surrounding ST data quality cannot be overstated. Discussing scenarios where only limited or poor-quality ST data are available will be crucial for guiding future studies. Furthermore, a clear explanation of how enhanced specificity and accuracy translate into tangible biological insights is essential for demystifying the underlying mechanisms driving developmental processes.

Response: We thank the reviewer for raising this essential suggestion. We have realized that in our previous manuscript, our discussion on the advantages and limitations of Palette and zSTEP was neither broad nor detailed enough.

Therefore, in our revised manuscript, we have added the following paragraphs to further discuss the advantages and limitations of Palette and zSTEP, as well as the potential application of zSTEP in developmental biology.

In this section, we have emphasized again the impact of ST data quality on the performance of Palette and zSTEP, and then compared Palette with the strategy that uses well-established marker genes to infer spatial information. We demonstrated that although Palette cannot achieve single cell resolution, it captures the major expression patterns, which are closely correlated to biological functions and critical for embryonic development. Furthermore, we further discussed that zSTEP is not only a valuable tool for investigating gene expression patterns, but also has the potential in evaluating the reaction-diffusion model to investigate the complicated and well-choreographed pattern formation during embryonic development.

As here we have provided a more comprehensive discussion about Palette and zSTEP, we think that the potential researchers will better understand the application scenarios of our inference pipeline and our datasets. We hope our study can assist and inspire further research in the field of spatial transcriptomics and developmental biology.

"Thirdly, the performance of Palette and zSTEP heavily relied on the quality of ST data. If the quality of ST data is not of sufficient quality, the low-expression genes may not be detected or only appear in very few scattered spots, and the performance of spot clustering could also be affected. Moreover, in this study, for example, the Stereo-seq data of 12 hpf zebrafish embryo had fewer slices on the right side (Fig. S3b), resulting in more blank spots in the right part of zSTEP for the 12 hpf embryo. However, with the ongoing advancements in spatial resolution and data quality, the performance of Palette is expected to be enhanced and demonstrate even greater potential for analysing spatiotemporal gene expression.

On the other hand, compared to the brilliant strategy that infers spatial information of scRNA-seq data from well-established genes, our Palette pipeline cannot achieve single cell resolution. However, our Palette pipeline is based on the ST reference, and thus preserves the real positional relationships between spots. Furthermore, the focus of our pipeline is to infer the gene expression patterns, which are closely correlated to biological functions and critical for embryonic development, rather than the sparse expression within individual spots. In this regard, our Palette pipeline can be advantageous, as it allows for reconstruction of the major expression profiles, which are often more relevant for understanding developmental processes. Additionally, our Palette can be applied to serial sections, enabling the construction of 3D ST atlas.

Finally, while the current analyses demonstrated that zSTEP can serve as a valuable tool for identifying genes having specific patterns at certain developmental stages, the exploration of zSTEP is still limited. During animal development, pattern formation is always one of the most important developmental issues. As demonstrated by the reaction-diffusion (RD) model, morphogen molecules are produced at specific regions of the embryo, forming morphogen gradients to guide cell specification, while interactions between different morphogens instruct more complicated and well-choreographed pattern formation. Our Palette constructed zSTEP, as a comprehensive transcriptomic expression pattern during development, could be leveraged to evaluate and prove the RD model during development, including AP patterning. Moreover, the investigation of gene expression patterns should not be limited to morphogens and TFs, and further investigation of their roles in AP patterning is desirable. Additionally, here a random forest model may be sufficient for investigating the most essential morphogens and TFs for AP axis refinement, while more sophisticated machine learning models may be required for addressing more specific biological questions."

Reviewer #2 (Significance):

The Palette pipeline demonstrates a marked improvement in specificity and accuracy when predicting spatial gene expression patterns. Evaluative studies on Drosophila and zebrafish datasets affirm its enhanced performance compared to existing methodologies. By effectively reconstructing spatial information from bulk transcriptomic data, the Palette method innovatively merges the philosophy of leveraging single-cell transcriptomic data for deconvolution analyses. This integration is pivotal, advancing traditional bulk RNA-seq approaches while laying the groundwork for future research.

One of the notable achievements in this work is the construction of the DreSTEP atlas, which integrates serial bulk RNA-seq data with advanced 3D imaging techniques. This resource grants researchers unprecedented access to the visualization of gene expression patterns across the zebrafish embryo, facilitating the investigation of spatial relationships and cell-cell interactions critical for developmental processes. Such capabilities are invaluable for understanding the intricate dynamics of embryogenesis and the distinct roles of individual cell types.

Response: We thank the reviewer for the positive evaluation of our work, either the Palette pipeline or zSTEP. The reviewer has strong expertise in algorithm development and computational biology, and the concerns and suggestions from the reviewer are significantly precious and valuable for us. Regarding the bioinformatics tool development, we did not have extensive experiences, and thus we did not thoroughly address the selection criteria or clarify the parameters used in the pipeline, which may influence the application by other researchers. Therefore, we sincerely appreciate the professional suggestions from the reviewer, which we can follow to address these issues, improve our manuscript and make our work more impactful for researchers. Additionally, we did not consider computation efficiency, processing speed and computational demands, which would be important factors for other researchers to use Palette. We would like to add extra analyses to address these aspects.

Currently, based on the suggestions from the reviewer, we have added extra texts discussing the clustering strategy in Palette pipeline, the advantages and limitations of Palette, and the potential application of zSTEP in developmental biology. We believe that readers will now have a clearer understanding of the performance of Palette and the application scenarios of both Palette and zSTEP. We have not fully addressed the comments raised by the reviewer yet, while we are working on the planned additional analyses and expect to complete all these tasks within the next 1-2 months. We sincerely thank the reviewer for the professional and valuable suggestions, which definitely improve our work and will make it accessible for a wide range of researchers.

Finally, through this review process, we have learned a lot about the important considerations and requirements when designing bioinformatics tools, and we benefit a lot from the thoughtful guidance. We express our thanks to the reviewer again for the guidance, and we will try our best to address the remaining issues to further improve our manuscript.

Reviewer #3 (Evidence, reproducibility and clarity):

Evidence, reproducibility and clarity

In this study, Dong and colleagues developed a computational pipeline to use spatial transcriptomics (ST) datasets as a reference to infer the spatial patterns of gene expression from bulk RNA sequencing data. This approach aims to overcome the low read depth and limited gene detection capabilities in current ST datasets, while exploiting its ability to provide highly resolved spatial information. By combining bulk RNA-seq datasets from 3 developmental stages during early zebrafish development with previously available ST and imaging datasets, the authors build DreSTEP (Danio rerio spatiotemporal expression profiles). Using this approach, they go on to identify the morphogens and transcription factors involved in anteroposterior patterning.

The paper is well written, and the pipeline presented in this study is likely to be useful beyond the case studies included in this study. There are a few questions that, in my view, would be important to clarify to increase the impact of this work:

Response: We sincerely appreciate the positive feedback from the reviewer on the Palette pipeline and zebrafish spatiotemporal expression profiles zSTEP. We thank the reviewer for the constructive suggestions, which have inspired us to think deeply about application and advantages of Palette and zSTEP for future studies.

We fully agree with the reviewer that we do not sufficiently clarify the advantages and limitations of our inference pipeline in the original manuscript. The questions raised by the reviewer are very insightful. For example, while the inference expression patterns may closely resemble the in situ hybridization observation, which we consider as good performance, the reviewer pointed out that we should consider whether weak, yet real expression may have been removed. These questions have motivated us to think more deeply about the underlying principles and assumptions of our inference pipeline. Following the reviewer's questions, we have expanded our discussion on the application of zSTEP in developmental biology and the features of Palette compared to the existing strategies.

We believe that after incorporating the revisions, our current manuscript now demonstrates the application scenario of Palette clearer and suggested the application of zSTEP for investigating biological questions in developmental biology. We are grateful for the reviewer's guidance, which helps us increase the impact of our work.

1. The authors mention that they used a variable factor to adjust expression differences between the ST and bulk RNA-seq datasets. It would be important for the authors to comment on how much overlap in gene expression is necessary between the datasets for an accurate calculation of this variable factor? Can this be directly tested, for instance, by testing how their conclusions vary if expression is adjusted by a variable factor calculated from only a smaller set of genes?

Response: We thank the reviewer for the professional questions. We are sorry about not including the gene numbers in our previous manuscript. And now we have provided the numbers of genes in bulk and ST data and the numbers of the overlapped genes (Fig. 15).

"Palette was performed on the aligned slices using the overlapped genes. For the 10 hpf embryo, there were 24,658 genes in the bulk data, 18,698 genes in the Stereo-seq data, and 16,601 overlapped genes. For the 12 hpf embryo, there were 23,018 genes in the bulk data, 18,948 genes in the Stereo-seq data, and 16,401 overlapped genes. For the 16 hpf embryo, there were 24,357 genes in the bulk data, 23,110 genes in the Stereo-seq data, and 19,539 overlapped genes."

See response letter for the figure.

For Palette implementation, we took all the overlapped genes. To calculate the variable factor, we aggregated the expression of each gene in the ST data, and then used the expression of the bulk data to divide the aggregated expression for variable factor calculation. As a result, each overlapped gene was assigned a variable factor to adjust its expression, based on its difference between bulk and ST data. The rationale behind this approach is that by considering the ST data as a whole, we can effectively reduce the variations among individual spots. This allows the variable factors to provide reasonable adjustment to gene expression.

Above all, the variable factors can be directly calculated. Currently Palette only can infer the expression patterns of overlapped genes. It means when the number of overlapped genes is small, such as MERFISH only detecting hundreds of genes, Palette can only infer the expression patterns of these genes. However, if the MERFISH data have good quality, which enable resolving distinct cell types, we believe Palette will also show good performance when using MERFISH as ST reference. Additionally, we plan to perform Palette using MERFISH as ST reference to further demonstrate its broad application when using different ST references.

1. Palette gives rise to highly spatially precise patterns, which closely match those found in ISH. However, the smoothening of the expression can also remove weak, yet real, local expression patterns, as shown for idgf6 in Fig. 2a. Can the authors test this more extensively for other genes?

Response: We thank the reviewer for this essential question. We agree with the reviewer that weak, yet real expression might be removed in our Palette inference pipeline. The weak, sparse expression may be due to the ST technique itself or the variations in samples. However, that sparse gene expression may not have biological meaning, and the focus of our pipeline in to capture the expression patterns, which are closely correlated with functions and crucial for embryonic development. Therefore, our algorithm considers spot characteristics and emphasize cluster-specific expression, resulting in spatial-specific expression patterns. In most cases, the main gene expression patterns can be captured, which can help understand gene functions and roles in embryonic development. We have updated Supplementary Figure S1a (Fig. 16) to include more gene patterns to demonstrate this point.

See response letter for the figure.

1. Using adjacent slices for ST and "bulk RNA-seq" may provide better results than those obtained when comparing two independent datasets. Could the authors also extend the analysis of Palette's functionalities by using separate, previously available but independent datasets, for ST and bulk RNA-seq in Drosophila as well?

Response: We thank the reviewer for the valuable question. We agree with the reviewer that using adjacent slices may provide better results. The idea here is that the inferred spatial expression patterns from pseudo bulk RNA-seq can be used to compare with the real expression of ST to evaluate Palette performance. We have updated our Figure 2a (Fig. 17) to illustrate the analysis clearer.

See response letter for the figure.

To demonstrate the Palette's functionalities, we have used Palette to infer zebrafish bulk RNA-seq slice (Junker et al., 2014) using Stereo-seq slice (Liu et al., 2022) as ST reference, and these two datasets are separate and independent. We agree with the reviewer that it would be good to use separate datasets to test in Drosophila to further demonstrate the Palette's functionalities. However, unfortunately, we did not find the Drosophila serial bulk RNA-seq data along left-right axis of the corresponding stages, and thus we might be unable to perform the extra analyses using independent Drosophila datasets.

References:

Junker, J.P. et al. Genome-wide RNA Tomography in the zebrafish embryo. Cell 159, 662-675 (2014).

Liu, C. et al. Spatiotemporal mapping of gene expression landscapes and developmental trajectories during zebrafish embryogenesis. Dev Cell 57, 1284-1298 e1285 (2022).

1. The DreSTEP analysis in zebrafish embryos is interesting and validates well-established observations in the field. Can the authors also discuss whether and how their dataset allows them to refine our understanding of the spatial or temporal pattern of the morphogens and TFs involved in AP patterning? This would further validate their approach.

Response: We appreciate the reviewer for recognition of our zSTEP and raising this valuable question, which has inspired us to think more deeply about the potential application of zSTEP in developmental biology. As the reviewer noted, our zSTEP analyses have validated well-established observations in the field. Rather than focusing on the sparse expression detected in ST data, zSTEP emphasizes the gene expression patterns that are closely correlated with biological functions and critical for embryonic development. Therefore, zSTEP can serve as a valuable tool for identifying the genes having specific patterns at certain developmental stages.

Pattern formation is one of the most important developmental issues for all animals. The reaction-diffusion (RD) model is a widely recognized theoretical framework used to explain self-regulated pattern formation in developing animal embryos (Kondo & Miura, 2010). Morphogen molecules are produced at specific regions of the embryo, forming morphogen gradients to guide cell specification. Most importantly, interactions between different morphogens instruct more complicated and well-choreographed pattern formation. Our Palette-constructed zSTEP provides a comprehensive transcriptomic expression pattern, including all morphogens and TFs, across the whole embryo during development. These valuable resources, in our opinion, could be leveraged to evaluate and prove the RD model during development, including AP patterning. In our current zSTEP analyses, we have already identified genes that exhibit specific expression patterns along AP axis, some of which have not been fully characterized. These genes could be potential targets for further investigation into their roles in AP patterning, although they are not the primary focus of this study. Additionally, our analyses only focused on morphogens and TFs, but zSTEP can be used to investigate the expression patterns of other genes as well. Moreover, we employed a random forest model to investigate the most essential morphogens and TFs for AP axis refinement, which is one of the basic applications of zSTEP. To investigate specific biological questions of interest, it would be worth exploring the use of more sophisticated machine learning models.

We have added the following paragraph in the Discussion section to discuss the potential application of zSTEP in future studies.

"Finally, while the current analyses demonstrated that zSTEP can serve as a valuable tool for identifying genes having specific patterns at certain developmental stages, the exploration of zSTEP is still limited. During animal development, pattern formation is always one of the most important developmental issues. As demonstrated by the reaction-diffusion (RD) model, morphogen molecules are produced at specific regions of the embryo, forming morphogen gradients to guide cell specification, while interactions between different morphogens instruct more complicated and well-choreographed pattern formation. Our Palette constructed zSTEP, as a comprehensive transcriptomic expression pattern during development, could be leveraged to evaluate and prove the RD model during development, including AP patterning. Moreover, the investigation of gene expression patterns should not be limited to morphogens and TFs, and further investigation of their roles in AP patterning is desirable. Additionally, here a random forest model may be sufficient for investigating the most essential morphogens and TFs for AP axis refinement, while more sophisticated machine learning models may be required for addressing more specific biological questions."

Reference

Kondo, S. & Miura, T. Reaction-Diffusion model as a framework for understanding biological pattern formation. Science 329, 1616-1620 (2010).

1. Can the authors comment on the limits of this inference pipeline? And how it performs as compared to single-cell RNA sequencing datasets where spatial information is inferred from well-established marker genes?

Response: We appreciate the reviewer for this insightful question, which has inspired us to further explore the advantages and limitations of the Palette pipeline in comparison with other inference strategies. As mentioned in the Discussion section, a key limitation of the inference pipeline is its heavy reliance on the quality of ST data. It is obvious that if the quality of ST data is not of sufficient quality, the low-expression genes may not be detected or only appear in very few scattered spots. We think it is a common issue for any inference tools using ST data as the reference. However, with the ongoing advancements in spatial resolution and data quality, the performance of Palette is expected to be improved.

As a comparison, the single-cell RNA sequencing datasets where spatial information is inferred from well-established marker genes do not face this limitation. The ground-breaking work by Satija et al. (2015) used such a strategy that combined scRNA-seq and in situ hybridizations of well-established marker genes to infer spatial location, enabling single cell resolution, as it maintains the high read depth and gene detection. One advantages of this scRNA-seq-based strategy is that it provides the transcriptomics of individual cells, rather than a combination of cell within a ST spot, although the positional relationships between cells are not real.

However, compared to the inference from ST data, the positional relationships between cells are not directly captured. On the other hand, as the embryonic development progresses, more cell types will be specified, and the body patterning becomes more complex. In this scenario, using well-established marker gene to infer spatial information would be much more challenging. Additionally, there are not many scRNA-seq datasets of serial sections, and thus this strategy may not be used to construct 3D ST atlas.

In contrast, our Palette inference pipeline is based on the ST data, which preserves the real positional relationships between spots. Although our inference pipeline cannot achieve single cell resolution, it focuses on the gene expression patterns rather than the sparse expression within individual spots. By applying Palette to paired serial sections, we were able to generated a 3D spatial expression atlas of zebrafish embryos, which has showed promising performance for investigating gene expression patterns and their involvement in AP patterning.

Reference

Satija, R. et al. Spatial reconstruction of single-cell gene expression data. Nature biotechnology 33, 495-502 (2015)

We have updated the following paragraphs to further demonstrating the limitation of the inference pipeline in details in the Discussion section.

"Thirdly, the performance of Palette and zSTEP heavily relied on the quality of ST data. If the quality of ST data is not of sufficient quality, the low-expression genes may not be detected or only appear in very few scattered spots, and the performance of spot clustering could also be affected. Moreover, in this study, for example, the Stereo-seq data of 12 hpf zebrafish embryo had fewer slices on the right side (Fig. S3b), resulting in more blank spots in the right part of zSTEP for the 12 hpf embryo. However, with the ongoing advancements in spatial resolution and data quality, the performance of Palette is expected to be enhanced and demonstrate even greater potential for analysing spatiotemporal gene expression.

On the other hand, compared to the brilliant strategy that infers spatial information of scRNA-seq data from well-established genes, our Palette pipeline cannot achieve single cell resolution. However, our Palette pipeline is based on the ST reference, and thus preserves the real positional relationships between spots. Furthermore, the focus of our pipeline is to infer the gene expression patterns, which are closely correlated to biological functions and critical for embryonic development, rather than the sparse expression within individual spots. In this regard, our Palette pipeline can be advantageous, as it allows for reconstruction of the major expression profiles, which are often more relevant for understanding developmental processes. Additionally, our Palette can be applied to serial sections, enabling the construction of 3D ST atlas."

Reviewer #3 (Significance):

This study tackles an important challenge in biology - the difficult to resolve gene expression patterns with high spatial precision and in a high-throughput manner. By integrating sequencing datasets from previously published studies, as well as newly-generated datasets, the authors provide evidence that their novel inference pipeline enables them to obtain high-quality spatial information simply from bulk RNA-seq datasets, using ST as a reference. The development of this pipeline - Palette - is a major part of this manuscript and its applicability is validated using datasets from Drosophila and zebrafish embryos. This in an important advance for the field, but it would be nice for the authors to further comment on i) the validity of some of their approaches and how they may influence the quality of their inference, as well as, ii) potential pitfalls/limitations of this approach as compared to others available in the field. This would synthetize both previous and current findings into a conceptual and technological framework that would have a strong impact well beyond cell and developmental biology.

Audience: This study would be relevant for a broad audience of biologists, interested in morphogen signaling, gene regulatory networks and cell fate specification.

Expertise in zebrafish development, gastrulation, morphogen signaling and morphogenesis.

Response: We thank the reviewer for providing the positive feedback, arising these valuable questions, which have motivated us to deeply consider the design concept and further application of Palette and zSTEP. Based on the insightful questions from the reviewer, we have added two extra paragraphs in the Discussion section to further discuss the potential application of zSTEP in developmental biology and application scenarios of the Palette pipeline. Specially, we have demonstrated that the performance of the inference pipeline relies on the spatial resolution and data quality of the ST data. We have then compared the advantages and limitations of Palette with the existing brilliant spatial inference strategy, which infers spatial information of scRNA-seq from well-established marker genes. Although our inference pipeline cannot achieve single cell resolution, it can capture the major expression patterns, which are closely correlated to functions and critical for embryonic development. We believe this will help readers gain a clearer understanding of the advantage and limitations of our pipeline compared to other tools, as well as the tasks for which Palette and our constructed zSTEP can be utilized. We express our thanks to the reviewer again for the valuable comments.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Referee #3

Evidence, reproducibility and clarity

In this study, Dong and colleagues developed a computational pipeline to use spatial transcriptomics (ST) datasets as a reference to infer the spatial patterns of gene expression from bulk RNA sequencing data. This approach aims to overcome the low read depth and limited gene detection capabilities in current ST datasets, while exploiting its ability to provide highly resolved spatial information. By combining bulk RNAseq datasets from 3 developmental stages during early zebrafish development with previously available ST and imaging datasets, the authors build DreSTEP (Danio rerio spatiotemporal expression profiles). Using this approach, they go on to identify the morphogens and transcription factors involved in anteroposterior patterning.

The paper is well written, and the pipeline presented in this study is likely to be useful beyond the case studies included in this study. There are a few questions that, in my view, would be important to clarify to increase the impact of this work:

1. The authors mention that they used a variable factor to adjust expression differences between the ST and bulk RNAseq datasets. It would be important for the authors to comment on how much overlap in gene expression is necessary between the datasets for an accurate calculation of this variable factor? Can this be directly tested, for instance, by testing how their conclusions vary if expression is adjusted by a variable factor calculated from only a smaller set of genes?
2. Palette gives rise to highly spatially precise patterns, which closely match those found in ISH. However, the smoothening of the expression can also remove weak, yet real, local expression patterns, as shown for idgf6 in Fig. 2a. Can the authors test this more extensively for other genes?
3. Using adjacent slices for ST and "bulk RNAseq" may provide better results than those obtained when comparing two independent datasets. Could the authors also extend the analysis of Palette's functionalities by using separate, previously available but independent datasets, for ST and bulk RNAseq in Drosophila as well?
4. The DreSTEP analysis in zebrafish embryos is interesting and validates well-established observations in the field. Can the authors also discuss whether and how their dataset allows them to refine our understanding of the spatial or temporal pattern of the morphogens and TFs involved in AP patterning? This would further validate their approach.
5. Can the authors comment on the limits of this inference pipeline? And how it performs as compared to single-cell RNA sequencing datasets where spatial information is inferred from well-established marker genes?

Significance

This study tackles an important challenge in biology - the difficult to resolve gene expression patterns with high spatial precision and in a high-throughput manner. By integrating sequencing datasets from previously published studies, as well as newly-generated datasets, the authors provide evidence that their novel inference pipeline enables them to obtain high-quality spatial information simply from bulk RNAseq datasets, using ST as a reference. The development of this pipeline - Palette - is a major part of this manuscript and its applicability is validated using datasets from Drosophila and zebrafish embryos. This in an important advance for the field, but it would be nice for the authors to further comment on i) the validity of some of their approaches and how they may influence the quality of their inference, as well as, ii) potential pitfalls/limitations of this approach as compared to others available in the field. This would synthetize both previous and current findings into a conceptual and technological framework that would have a strong impact well beyond cell and developmental biology.

Audience: This study would be relevant for a broad audience of biologists, interested in morphogen signaling, gene regulatory networks and cell fate specification.

Expertise in zebrafish development, gastrulation, morphogen signaling and morphogenesis.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Referee #2

Evidence, reproducibility and clarity

The authors of the study introduce the Palette method, a novel approach designed to infer spatial gene expression patterns from bulk RNA-sequencing (RNA-seq) data. This method is complemented by the development of the DreSTEP 3D spatial gene expression atlas of zebrafish embryos, establishing a comprehensive resource for visualizing gene expression and investigating spatial cell-cell interactions in developmental biology.

Major concerns:

1. The efficacy of the Palette method may be compromised by its dependency on the quality of the reference spatial transcriptomics data. As highlighted in the study, variations in data quality can lead to significant challenges in reconstructing accurate spatial expression patterns from bulk data. This underscores the necessity of evaluating quality parameters, such as the number of gene detections and spatial resolution, to ensure reliable outcomes. Additional studies should rigorously assess how these quality factors influence the accuracy and efficiency of the algorithm in various data contexts, particularly under diverse conditions of gene detection.
2. The methodology raises pertinent questions regarding how the clustering results from different algorithms may affect the reconstructions by the Palette method. The authors would better provide a detailed discussion/comparison of clustering processes that optimize the reconstruction of spatial patterns, ensuring precision in the downstream analyses.
3. The choice to utilize only highly expressed genes in the initial stages of the Palette algorithm also warrants further exploration. Addressing the criteria for determining which genes qualify as "highly expressed" and outlining robust cutoff will enhance the algorithm's rigor and applicability. Similarly, in the iterative estimation of gene expression across spatial spots, establishing optimal iteration conditions is crucial. Implementing a loss function may offer a systematic method for concluding iterations, thus refining computational efficiency.
4. Performance metrics relating to processing speed and computational demands remain inadequately addressed in the current framework. Understanding how the Palette method scales across varying gene counts and bulk RNA-seq datasets will be essential for potential applications in larger biological contexts. Notably, the quantitative demands of analyzing 20,000 genes when processing 10, 100, or 1,000 bulk RNA profiles must be articulated to guide researchers in planning accordingly.

Minor opinions:

1. Despite the promising advances offered by the zebrafish 3D reconstruction, there is a lack of details regarding numbers of the spatial transcriptomics (ST) data utilized, and the number of bulk RNA-seq data employed in the analyses. These parameters need to be clarified.
2. Issues regarding spatial cell-cell communication, especially concerning interactions over longer distances, necessitate careful consideration. Introducing spatial distance constraints could help formulate more realistic models of cellular interactions, a vital aspect of embryonic development.
3. Evaluation metrics such as the Adjusted Rand Index (ARI) and Root Mean Square Error (RMSE) represent critical tools for systematically measuring the similarity of inferred spatial patterns, yet their specific application within this context should be elaborated.
4. The study's limitations surrounding ST data quality cannot be overstated. Discussing scenarios where only limited or poor-quality ST data are available will be crucial for guiding future studies. Furthermore, a clear explanation of how enhanced specificity and accuracy translate into tangible biological insights is essential for demystifying the underlying mechanisms driving developmental processes.

Significance

The Palette pipeline demonstrates a marked improvement in specificity and accuracy when predicting spatial gene expression patterns. Evaluative studies on Drosophila and zebrafish datasets affirm its enhanced performance compared to existing methodologies. By effectively reconstructing spatial information from bulk transcriptomic data, the Palette method innovatively merges the philosophy of leveraging single-cell transcriptomic data for deconvolution analyses. This integration is pivotal, advancing traditional bulk RNA-seq approaches while laying the groundwork for future research.

One of the notable achievements in this work is the construction of the DreSTEP atlas, which integrates serial bulk RNA-seq data with advanced 3D imaging techniques. This resource grants researchers unprecedented access to the visualization of gene expression patterns across the zebrafish embryo, facilitating the investigation of spatial relationships and cell-cell interactions critical for developmental processes. Such capabilities are invaluable for understanding the intricate dynamics of embryogenesis and the distinct roles of individual cell types.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Referee #1

Evidence, reproducibility and clarity

Summary:

The manuscript titled "Unravelling the Progression of the Zebrafish Primary Body Axis with Reconstructed Spatiotemporal Transcriptomics" presents a comprehensive analysis of the development of the primary body axis in zebrafish by integrating bulk RNA-seq, 3D images, and Stereo-Seq. The authors first clearly demonstrate the application of Palette for integrating RNA-seq and Stereo-Seq using published spatial transcriptomics data of Drosophila embryos. Subsequently, they produced serial bulk RNA-seq data for certain developmental stages of Danio rerio embryos and utilized published Stereo-Seq data. Through robust validation, the authors observe the molecular network involved in AP axis formation. While the authors show that integrating bulk RNA-seq data with Stereo-Seq improves spatial resolution, additional proof is required to demonstrate the extent of this improvement.

Major Comments:

1. Lines 66-68: Discuss the limitations of existing tools and explicitly state the advantages of using Palette.
2. Body Pattern Genes Analysis: For both Drosophila and Danio rerio, it would be valuable to examine body pattern genes in Stereo-Seq and apply Palette to determine if the resolution of the segments improves or merges. The resolution of the A-P axis is convincing, but further evidence for other segments would be beneficial.
3. Figure 2d: Include the A-P line for which the intensity profile was plotted in the main figure, rather than just in the supplementary material. Additionally, consider simplifying the plot by not combining three lines into one, as it complicates the interpretation of observations.
4. Drosophila Data Analysis: While the alignment and validation of Danio rerio sections are clearly explained, the analysis and validation of Drosophila data are insufficiently detailed. Provide a more thorough explanation of how the intensity profiles between BDGP in situ data and Stereo-Seq data are adjusted.
5. Figure 3d: Present a plot with the expected expression profiles of the three genes if the embryo is aligned as anticipated.
6. Analysis Without Palette: Between lines 277-438, the outcome of using Palette with bulk RNA-seq and Stereo-Seq is convincing. However, consider the following:<br /> o What would be the observations if the analysis were conducted solely with Stereo-Seq data, without incorporating bulk RNA-seq data and employing Palette?<br /> o This study uses only Stereo-Seq as the spatial transcriptomics reference. It would strengthen the argument to use at least one other spatial transcriptomics method, such as Visium or MERFISH, in conjunction with bulk RNA-seq and Palette, to demonstrate whether Palette consistently improves gene expression resolution.
7. PDAC Data Analysis: Provide a more detailed explanation of the PDAC data analysis and use appropriate colors in the tissue images to clearly distinguish cell types.
8. Comparison with Other Methods: State the limitations of not using STitch3D and Spateo for alignment and explain why these methods were not employed.

Minor Comments:

1. References: Add references to the statements in lines 51-53.
2. Scientific Name Consistency: Ensure consistency in using either "Danio rerio" or "zebrafish" throughout the manuscript.
3. Related References: Include the following relevant references:
4. https://academic.oup.com/bib/article/25/4/bbae316/7705532
5. https://www.life-science-alliance.org/content/6/1/e202201701
6. Figure 1a: In the Venn diagram, include the number of genes in the bulk and Stereo-Seq datasets, as well as the number of overlapping genes.
7. Figure 1 Improvement: Enlarge Figure 1 and reduce repetitive elements, such as parts of the deconvolution and Figure 1b.
8. Figure 3f: Explain the black discontinuous line in the plot.
9. Line 610: State the percentage of unpaired imaging spots.
10. Lines 616-618: Specify the unit for the spot diameter.

Significance

This algorithm will be useful not only for the field of developmental biology but also for wider applications in spatial omics. Although I have expertise in spatial omics technology development, my understanding of computational biology is limited, which restricts my ability to fully evaluate the Palette algorithm presented in this paper.

Links to data spaces on Wikipedia: note that that link talks about data spaces as tech only, not in the sense of the EU single market for data, which is a much more meaningful context when it comes to (design)values.

eLife Assessment

The authors studied cognitive control signals in the anterior cingulate cortex (ACC) while rats selected between small immediate and larger delayed rewards. The description of behavioral strategies related to value-tracking signals in ACC is potentially useful. The evidence in support of this finding is incomplete due to issues with the task design, analyses, and modeling.

Reviewer #1 (Public review):

Summary:

Adult (4mo) rats were tasked to either press one lever for an immediate reward or another for a delayed reward. The task had an adjusting amount structure in which (1) the number of pellets provided on the immediate reward lever changed as a function of the decisions made, (2) rats were prevented from pressing the same lever three times in a row.

While the authors have been very responsive to the reviews, and I appreciate that, unfortunately, the new analyses reported in this revision actually lead me to deeper concerns about the adequacy of the data to support the conclusions. In this revision, it has become clear that the conclusions are forced and not supported by the data. Alternative theories are not considered or presented. This revision has revealed deep problems with the task, the analyses, and the modeling.

Data Weaknesses

Most importantly, the inclusion of the task behavior data has revealed a deep problem with the entire structure of the data. As is obvious in Figure 1D, there is a slow learning effect that is changing over the sessions as the animals learn to stop taking the delayed outcome. Unfortunately, the 8s delays came *after* the 4s. The first 20 sessions contain 19 4s delays and 1 8s delay, while the last 20 sessions contain 14 8s delays and 6 4s delays. Given the changes across sessions, it is likely that a large part of the difference is due to across-session learning (which is never addressed or considered).

These data are not shown by subject and I suspect that individual subjects did all 4s then all 8s and some subjects switched tasks at different times. If my suspicion is true, then any comparisons between the 4s and 8s conditions (which are a major part of the author's claims) may have nothing to do with the delays, but rather with increased experience on the task.

Furthermore, the four "groups", which are still poorly defined, seem to have been assessed at a session-by-session level. So when did each animal fall into a given group? Why is Figure 1D not showing which session fell into which group and why are we not seeing each animal's progression? They also admit that animals used a mixture of strategies, which implies that the "group" assignment is an invalid analysis, as the groups do not accommodate strategy mixing.

Figure 2 shows that none of the differences of the group behavior against random choice with a basic p(delay) are significant. The use a KS test to measure these differences. KS tests are notoriously sensitive as KS tests simply measure whether there are any statistical differences between two distributions. They do not report the full statistics for Figure 2, but only say that the 4HI group was not significant (KS p-value = 0.72) and the 8LO showed a p-value of 0.1 (which they interpret as significant). p=0.1 is not significant. They don't report the value of the 4LO or 8HI groups (why not?), but say they are in-between these two extremes. That means *none* of the differences are significant.

They then test a model with additional parameters, and say that the model includes more than the minimal p_D parameter, but never report BIC or AIC model comparisons. In order to claim that the model is better than the bare p_D assumption, they should be reporting model-comparison statistics. But given that the p_D parameters are enough (q.v. Figure 2), this entire model seems unnecessary

It took me a while to determine what was being shown in Figure 3, but I was eventually able to determine that 0 was the time after the animal made the choice to wait out the delay side, so the 4s in Figure 3A1 with high power in the low-frequency (<5 Hz) range is the waiting time. They don't show the full 8s time. Nor do they show the spectrograms separated by group (assuming that group is the analytical tool they are using). In B they show only show theta power, but it is unclear how to interpret these changes over time.

In Figure 4, panel A is mostly useless because it is just five sample sessions showing firing rate plotted on the same panels as the immediate reward amount. If they want to claim correlation, they should show and test it. But moreover, this is not how neural data should be presented - we need to know what the cells are doing, population-wise. We need to have an understanding of the neural ensemble. These data are clearly being picked and chosen, which is not OK.

Figure 4, panels B and C show that the activity trivially reflects the reward that has been delivered to the animal, if I am understanding the graphs correctly. (The authors do not interpret it this way, but the data is, to my eyes, clear.) The "immediate" signal shows up immediately at choice and reflects the size of the immediate reward (which is varying). The "delay" signal shows up after the delay and does not, which makes sense as the animals get 6 pellets on the delayed side no matter what. In fact, the max delayed side activity = the max immediate side activity, which is 6 pellets. This is just reward-related firing.

Figure 5 is poorly laid out, switching the order in 5C to be 2 1 3 in E and F. (Why?!) The statistics for Figure 5 on page 17 should be asking whether there are differences between neuron types, not whether there is a choice x time interaction in a given neuron type. When I look at Figure 5F1-3, all three types look effectively similar with different levels of noise. It is unclear why they are doing this complicated PC analysis or what we should be drawing from it.

Figure 6 mis-states pie charts as "total number" rather than proportions.

Interpretation Weaknesses

The separation of cognitive effort into "resource-based" and "resistance-based" seems artificial to me. I still do not understand why the ability to resist a choice does not also depend on resource or why using resources are not a form of resistance. Doesn't every action in the end depend on the resources one has available? And doesn't every use of a resource resist one option by taking another? Even if one buys these two separate cognitive control processes (which at this point in reading the revision, I do not), the paper starts from the assumption that a baseline probability of waiting out the delays is a "resistance-based cognitive control" (why?) and a probability of choice that takes into account the size of the immediate value (confusingly abbreviated as ival) is a "resource-based cognitive control" (again, why?)

Reviewer #2 (Public review):

Summary:

I appreciate the considerable work the authors have done on the revision. The manuscript is markedly improved.

Strengths still include the strong theoretical basis, well-done experiments, and clear links to LFP / spectral analyses that have links to human data. The task is now more clearly explained, and the neural correlates better articulated.

Weaknesses:

I had remaining questions, many related to my previous questions.<br /> (1) The results have some complexity, but I still had questions about which is resource and which is resistance based. The authors say in the last sentence of the discussion: "Prominent pre-choice theta power was associated with a behavioral strategy characterized by a strong bias towards a resistance-based strategy, whereas the neural signature of ival-tracking was associated with a strong bias towards a resource-based strategy.".<br /> I might suggest making this simpler and clear in the abstract and the first paragraph of the discussion. A simple statement like 'pre-choice theta was biased towards resistance whereas single neurons were biased towards resources" might make this idea come across?

(2) I think most readers would like to see raw single trial LFP traces in Figure 3, single unit rasters in Figure 4, and spike-field records in Figure 5.

(3) What limitations are there to this work? I wonder if readers might benefit from some contextualization - the sample size, heterogenous behavior - lack of cell-type specificity - using PC3 to define spectral relationships - I might suggest pointing these out.

(4) I still wasn't sure what 4 Hz vs. theta 6-12 Hz meant - is it all based on PC3's pos/neg correlation? I wonder if showing a scatter plot with the y-axis being PC3 and the x-axis being theta 4 Hz power would help distinguish these? Is this the first time this sort of analysis has been done? If so, it requires clearer definitions.

Reviewer #3 (Public review):

Summary:

The study investigated decision making in rats choosing between small immediate rewards and larger delayed rewards, in a task design where the size of the immediate rewards decreased when this option was chosen and increased when it was not chosen. The authors conceptualise this task as involving two different types of cognitive effort; 'resistance-based' effort putatively needed to resist the smaller immediate reward, and 'resource-based' effort needed to track the changing value of the immediate reward option. They argue based on analyses of the behaviour, and computational modelling, that rats use different strategies in different sessions, with one strategy in which they preferentially choose the delayed reward option irrespective of the current immediate reward size, and another strategy in which they preferentially choose the immediate reward option when the immediate reward size is large, and the delayed reward option when the immediate reward size is small. The authors recorded neural activity in anterior cingulate cortex. They propose that oscillatory activity in the 6-12Hz theta band occurs when subjects use a 'resistance-based' strategy of choosing the delayed option irrespective of the current value of the immediate reward option. They also examine neural representation of the current value of the immediate reward option, and suggest that this value is more strongly represented when subjects are using this value information to guide choice. They further argue that neurons whose activity is modulated by theta oscillations are less involved in tracking the value of the immediate reward option than neurons whose activity is not theta modulated. If solid, these findings will be of interest to researchers working on cognitive control and ACCs involvement in decision making. However, there are some issues with the modelling and analysis which preclude high confidence in the validity of the conclusions.

Strengths:

The behavioural task used is interesting and the recording methods used (64 channel silicon probes) should enable the collection of good quality single unit and LFP electrophysiology data. The authors recorded from a sizable sample of subjects for this type of study. The approach of splitting the data into sessions where subjects used different strategies and then examining the neural correlates of each is in principle interesting, though I have some reservations about the strength of evidence for the existence of multiple strategies.

Limitations:

The dataset is unbalanced in terms of both the number of sessions contributed by each subject, and their distribution across the different putative behavioural strategies (see Table 1), with some subjects contributing 7 sessions to a given strategy and others 0. Further, only 2 of 10 subjects contribute any sessions to one of the behavioural strategies (8LO), and a single subject contributes >50% of the sessions (7 of 13) sessions to another strategy (8HI). Apparent differences in brain activity between the strategies could therefore in fact reflect differences between subjects, which could arise due to e.g. differences in electrode placement. To make firm conclusions that neural activity is different in sessions where different strategies are thought to be employed, it would be necessary to account for potential cross-subject variation in the data. The current statistical methods don't appear to do this as they use within subject measures (e.g. trials or neurons) as the experimental unit and ignore which subject the neuron/trial came from.

The starting point for the analysis was the splitting of sessions into 4 groups based on the duration of the delay (4 vs 8 seconds) and then clustering within each delay category into two sub-groups. It was not clear why 2 clusters per delay category were used, nor whether the data did in fact have a clear split into two distinct clusters or continuous variation across the population of sessions. The simplified RL model used in the revised manuscript (which is an improvement from that used in the previous version) could in principle help to quantify variation across the populations of sessions, by using model fitting and comparison methods to evaluate variation in strategy across subjects. However, as far as I could tell no model-fitting or comparison was performed, and the only attempt to link the model to data was by simulating data using a fixed probability of choosing the delayed lever (i.e. with no learning across trials) and comparing the distribution of total rewards obtained per session with that of the subjects in each group (Figure 2). Total reward per session is a very coarse behavioural metric and using likelihood-based methods to fit model parameters to subjects trial-by-trial choice data would provide a more sensitive way of using the modelling to assess behavioural strategy across sessions.

Conceptually, it is not obvious that choices towards the delayed vs immediate lever reflect use of different strategies employing different types of cognitive effort. Rather these could reflect a single strategy which compares the estimated value of the two levers, with differences in behaviour between sessions accounted for either by differences in the task itself (between the 8s and 4s delay condition) or differences in the parameters of the strategy, such as the strength of temporal discounting.

Even if one accepts the claim that the task recruits two distinct types of cognitive control, the argument that theta oscillations, which occur on delay choice trials in the 4s delay condition, are a correlate of a 'resistance-based' strategy (resisting the immediate reward), is hard to reconcile with the fact that theta oscillations do not occur on delay choice trials in the 8s delay condition (Figure 3). The authors note this discrepancy, but state that 'The reason was because these groups largely avoided the delayed lever (Figure 1) and thereby abandoned the need to implement resistance-based control altogether.' However, the data in Figure 1D show that even in the 8s condition the subjects choose the delayed lever on around 50% of trials. It is not obvious why choosing the delayed lever on 50% of trials in the 8s condition does not require 'resistance-based' cognitive effort, while choosing it in the 4s delay condition does.

The other main claims regarding the neural data are that the neuronal representation of the value of the immediate reward lever (ival) is stronger in sessions where subjects are choosing that lever more often, particularly the 8LO group, and that neurons whose activity tracks ival are a different population from neurons whose activity is theta modulated. However, the analysis methods used to make these claims are rather convoluted and make it hard to assess the strength of the evidence for them.

To evaluate the strength of ival representation in neural activity, the authors first fit a regression model predicting each neuron's activity at different timepoints as a function of behavioural variables including ival, which is a sensible first step. However, they then perform clustering on the regression coefficients and then plot neural activity only for the cluster which they state 'provided the clearest example of value tracking'. It is not clear how the clustering was done, whether there were in fact well defined clusters in the neural activity, how the clusters whose activity is plotted were chosen, nor the proportion of neurons in this cluster for each group of sessions. The analysis therefore provides only limited information about the strength of ival representation in different session groups. It would be useful to quantify the variance explained by ival in neural activity for each group of sessions using a simpler quantification of the regression analysis, such as cross-validated coefficient of partial determination.

The analysis of how theta modulation related to representation of ival across neurons was also complicated and non-standard. To determine whether individual neurons were theta modulated, the authors did PCA on a matrix comprised of spike train autocorrelations for individual neurons, and then grouped neurons according to the projection of their autocorrelation function onto the 3rd Principal Component, on the basis that neurons with negative projection onto this component showed a peak roughly at theta frequency in the power spectrum of their autocorrelation. Even ignoring the fact that the peak in the power spectrum is broad and centred above the standard theta frequency (see figure 5B3), this is an arbitrary and unnecessarily complex way to determine if neurons are theta modulated. It would be much simpler and greatly preferable to either directly assess the modulation depth of individual neurons spike train autocorrelation in the theta band, or to use a metric of spike-LFP coupling in the theta band instead. The authors do include some analysis of spike field coherence in Figure 6 and this is a much more sensible approach. However, it is worth noting that the only session group which shows a difference in coherence at theta frequency relative to the other groups is 8LO, to which only 2 of 8 animals contribute any data and 70% of sessions come from one animal. It is therefore unclear whether differences in this group are due to differences in behavioural strategy, or reflect other sources of cross-animal variation.

Author response:

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

We would like to thank the reviewers for their efforts and feedback on our preprint. We have elected to rework the manuscript for publication in a different journal. In this process we will alter many of the approaches and re-evaluate the conclusions. With this, many of the points raised by the reviewers will be no longer relevant and therefore do not require a response. Again, we thank the reviewers for their time and helpful feedback.

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

eLife Assessment:

The authors present a potentially useful approach of broad interest arguing that anterior cingulate cortex (ACC) tracks option values in decisions involving delayed rewards. The authors introduce the idea of a resource-based cognitive effort signal in ACC ensembles and link ACC theta oscillations to a resistance-based strategy. The evidence supporting these new ideas is incomplete and would benefit from additional detail and more rigorous analyses and computational methods.

We are extremely grateful for the several excellent and comments of the reviewers. To address these concerns, we have completely reworked the manuscript adding more rigorous approaches in each phase of the analysis and computational model. We realize that this has taken some time to prepare the revision. However, given the comments of the reviewers, we felt it necessary to thoroughly rework the paper based on their input. Here is a (nonexhaustive) overview of the major changes we made:

We have developed a way to more adequately capture the heterogeneity in the behavior

We have completely reworked the RL model

We have added additional approaches and rigor to the analysis of the value-tracking signal. 

Reviewer #1 (Public Review):

Summary:

Young (2.5 mo [adolescent]) rats were tasked to either press one lever for immediate reward or another for delayed reward. 

Please note that at the time of testing and training that the rats were > 4 months old. 

The task had a complex structure in which (1) the number of pellets provided on the immediate reward lever changed as a function of the decisions made, (2) rats were prevented from pressing the same lever three times in a row. Importantly, this task is very different from most intertemporal choice tasks which adjust delay (to the delayed lever), whereas this task held the delay constant and adjusted the number of 20 mg sucrose pellets provided on the immediate value lever.

Several studies parametrically vary the immediate lever (PMID: 39119916, 31654652, 28000083, 26779747, 12270518, 19389183). While most versions of the task will yield qualitatively similar estimates of discounting, the adjusting amount is preferred as it provides the most consistent estimates (PMID: 22445576). More specifically this version of the task avoids contrast effects of that result from changing the delay during the session (PMID: 23963529, 24780379, 19730365, 35661751) which complicates value estimates. 

Analyses are based on separating sessions into groups, but group membership includes arbitrary requirements and many sessions have been dropped from the analyses. 

We have updated this approach and now provide a more comprehensive assessment of the behavior. The updated approach applies a hierarchical clustering model to the behavior in each session. This was applied at each delay to separate animals that prefer the immediate option more/less. This results in 4 statistically dissociable groups (4LO, 4HI, 8LO, 8HI) and includes all sessions. Please see Figure 1. 

Computational modeling is based on an overly simple reinforcement learning model, as evidenced by fit parameters pegging to the extremes. 

We have completely reworked the simulations in the revision. In the updated RL model we carefully add parameters to determine which are necessary to explain the experimental data. We feel that it is simplified yet more descriptive. Please see Figure 2 and associated text. 

The neural analysis is overly complex and does not contain the necessary statistics to assess the validity of their claims.

We have dramatically streamlined the spike train analysis approach and added several statistical tests to ensure the rigor of our results. Please see Figures 4,5,6 and associated text. 

Strengths:

The task is interesting.

Thank you for the positive comment

Weaknesses:

Behavior:

The basic behavioral results from this task are not presented. For example, "each recording session consisted of 40 choice trials or 45 minutes". What was the distribution of choices over sessions? Did that change between rats? Did that change between delays? Were there any sequence effects? (I recommend looking at reaction times.) Were there any effects of pressing a lever twice vs after a forced trial? 

Please see the updated statistics and panels in Figures 1 and 2. We believe these address this valid concern.  

This task has a very complicated sequential structure that I think I would be hard pressed to follow if I were performing this task. 

Human tasks implement a similar task structure (PMID: 26779747). Please note the response above that outlines the benefits of using of this task.   

Before diving into the complex analyses assuming reinforcement learning paradigms or cognitive control, I would have liked to have understood the basic behaviors the rats were taking. For example, what was the typical rate of lever pressing? If the rats are pressing 40 times in 45 minutes, does waiting 8s make a large difference?

Thank you for this suggestion. Our additions to Figure 1 are intended to better explain and quantify the behavior of the animals. Note that this task is designed to hold the rate of reinforcement constant no matter the choices of the animals. Our analysis supports the long-held view in the literature that rats do not like waiting for rewards, even at small delays. Going from the 4 à 8 sec delay results in significantly more immediate choices, indicating that the rats will forgo waiting 8 sec for a larger reinforcer and take a smaller reinforcer at 4 sec.  

For that matter, the reaction time from lever appearance to lever pressing would be very interesting (and important). Are they making a choice as soon as the levers appear? Are they leaning towards the delay side, but then give in and choose the immediate lever? What are the reaction time hazard distributions?

This is an excellent suggestion, we have added a brief analysis of reaction times (Please see the section entitled “4 behavioral groups are observed across all sessions” in the Results). Please note that an analysis of the reaction times has been presented in a prior analysis of this data set (White et al., 2024). In addition, an analysis of reaction times in this task was performed in Linsenbardt et al. (2017). In short, animals tend to choose within 1 second of the lever appearing. In addition, our prior work shows that responses on the immediate lever tend to be slower, which we viewed as evidence of increased deliberation requirements (possibly required to integrate value signals).   

It is not clear that the animals on this task were actually using cognitive control strategies on this task. One cannot assume from the task that cognitive control is key. The authors only consider a very limited number of potential behaviors (an overly simple RL model). On this task, there are a lot of potential behavioral strategies: "win-stay/lose-shift", "perseveration", "alternation", even "random choices" should be considered.

The strategies the Reviewer mentioned are descriptors of the actual choices the rats made. For example, perseveration means the rat is choosing one of the levers at an excessively high rate whereas alternation means it is choosing the two levers more or less equally, independent of payouts. But the question we are interested in is why? We are arguing that the type of cognitive control determines the choice behavior, but cognitive control is an internal variable that guides behavior, rather than simply a descriptor of the behavior. For example, the animal opts to perseverate on the delayed lever because the cognitive control required to track ival is too high. We then searched the neural data for signatures of the two types of cognitive control.

The delay lever was assigned to the "non-preferred side". How did side bias affect the decisions made?

The side bias clearly does not impact performance as the animals prefer the delay lever at shorter delays, which works against this bias.  

The analyses based on "group" are unjustified. The authors compare the proportion of delayed to immediate lever press choices on the non-forced trials and then did k-means clustering on this distribution. But the distribution itself was not shown, so it is unclear whether the "groups" were actually different. They used k=3, but do not describe how this arbitrary number was chosen. (Is 3 the optimal number of clusters to describe this distribution?) Moreover, they removed three group 1 sessions with an 8s delay and two group 2 sessions with a 4s delay, making all the group 1 sessions 4s delay sessions and all group 2 sessions 8s delay sessions. They then ignore group 3 completely. These analyses seem arbitrary and unnecessarily complex. I think they need to analyze the data by delay. (How do rats handle 4s delay sessions? How do rats handle 6s delay sessions? How do rats handle 8s delay sessions?). If they decide to analyze the data by strategy, then they should identify specific strategies, model those strategies, and do model comparison to identify the best explanatory strategy. Importantly, the groups were session-based, not rat based, suggesting that rats used different strategies based on the delay to the delayed lever.

We have completely reworked our approach for capturing the heterogeneity in behavior. We have taken care to show more of the behavioral statistics that have gone into identifying each of the groups. All sessions are included in this analysis. As the reviewer suggests, we used the statistics from each of the behavioral groups to inform the RL model that explores neural signals that underly decisions in this task. We strongly disagree that groups should be rat and not session based as the behavior of the animal can, and does, change from day to day. This is important to consider when analyzing the neural data as rat-based groupings would ignore this potential source of variance. 

The reinforcement learning model used was overly simple. In particular, the RL model assumes that the subjects understand the task structure, but we know that even humans have trouble following complex task structures. Moreover, we know that rodent decision-making depends on much more complex strategies (model-based decisions, multi-state decisions, rate-based decisions, etc). There are lots of other ways to encode these decision variables, such as softmax with an inverse temperature rather than epsilon-greedy. The RL model was stated as a given and not justified. As one critical example, the RL model fit to the data assumed a constant exponential discounting function, but it is well-established that all animals, including rodents, use hyperbolic discounting in intertemporal choice tasks. Presumably this changes dramatically the effect of 4s and 8s. As evidence that the RL model is incomplete, the parameters found for the two groups were extreme. (Alpha=1 implies no history and only reacting to the most recent event. Epsilon=0.4 in an epsilongreedy algorithm is a 40% chance of responding randomly.)

While we agree that the approach was not fully justified, we do not agree that it was invalid. Simply stated, a softmax approach gives the best fit to the choice behavior, whereas our epsilon-greedy approach attempted to reproduce the choice behavior using a naïve agent that progressively learns the values of the two levers on a choice-by-choice basis. Nevertheless, we certainly appreciate that important insights can be gained by fitting a model to the data as suggested. We feel that the new modeling approach we have now implemented is optimal for the present purposes and it replaces the one used in the original manuscript.

The authors do add a "dbias" (which is a preference for the delayed lever) term to the RL model, but note that it has to be maximal in the 4s condition to reproduce group 2 behavior, which means they are not doing reinforcement learning anymore, just choosing the delayed lever.

The dbias term was dropped in the new model implementation

Neurophysiology:

The neurophysiology figures are unclear and mostly uninterpretable; they do not show variability, statistics or conclusive results.

While the reviewer is justified in criticizing the clarity of the figures, the statement that “they do not show variability, statistics or conclusive results” is not correct. Each of the figures presented in the first draft of the manuscript, except Figure 3, are accompanied by statistics and measures of variability. Nonetheless we have updated each of the neurophysiology analyses. We hope that the reviewer will find our updates more rigorous and thorough.   

As with the behavior, I would have liked to have seen more traditional neurophysiological analyses first. What do the cells respond to? How do the manifolds change aligned to the lever presses? Are those different between lever presses?

We have added several figures that plot the mean +/- SEM of the neural activity (see Figures 4 and 5). Hopefully this provides a more intuitive picture of the changes in neural activity throughout the task.  

Are there changes in cellular information (both at the individual and ensemble level) over time in the session? 

We provide several analyses of how firing rate changes over trials in relation to ival over time and trials in the session. In addition, we describe how these signals change in each of the behavioral groups. 

How do cellular responses differ during that delay while both levers are out, but the rats are not choosing the immediate lever?

We were somewhat unclear about this suggestion as the delay follows the lever press. In addition, there is no delay after immediate presses 

Figure 3, for example, claims that some of the principal components tracked the number of pellets on the immediate lever ("ival"), but they are just two curves. No statistics, controls, or justification for this is shown. BTW, on Figure 3, what is the event at 200s?

This comment is no longer relevant based on the changes we’ve made to the manuscript. 

I'm confused. On Figure 4, the number of trials seems to go up to 50, but in the methods, they say that rats received 40 trials or 45 minutes of experience.

This comment is no longer relevant based on the changes we’ve made to the manuscript. 

At the end of page 14, the authors state that the strength of the correlation did not differ by group and that this was "predicted" by the RL modeling, but this statement is nonsensical, given that the RL modeling did not fit the data well, depended on extreme values. Moreover, this claim is dependent on "not statistically detectable", which is, of course, not interpretable as "not different".

This comment is no longer relevant based on the changes we’ve made to the manuscript. 

There is an interesting result on page 16 that the increases in theta power were observed before a delayed lever press but not an immediate lever press, and then that the theta power declined after an immediate lever press. 

Thank you for the positive comment. 

These data are separated by session group (again group 1 is a subset of the 4s sessions, group 2 is a subset of the 8s sessions, and group 3 is ignored). I would much rather see these data analyzed by delay itself or by some sort of strategy fit across delays.

Thank you for the excellent suggestion. Our new group assignments take delay into account. 

That being said, I don't see how this description shows up in Figure 6. What does Figure 6 look like if you just separate the sessions by delay?

We are unclear what the reviewer means by “this description”.  

Discussion:

Finally, it is unclear to what extent this task actually gets at the questions originally laid out in the goals and returned to in the discussion. The idea of cognitive effort is interesting, but there is no data presented that this task is cognitive at all. The idea of a resourced cognitive effort and a resistance cognitive effort is interesting, but presumably the way one overcomes resistance is through resourcelimited components, so it is unclear that these two cognitive effort strategies are different.

The basis for the reviewers assertation that “the way one overcomes resistance is through resourcelimited components” is not clear. In the revised version, we have taken greater care to outline how each type of effort signal facilitates performance of the task and articulate these possibilities in our stochastic and RL models. We view the strong evidence for ival tracking presented herein as a critical component of resource based cognitive effort. 

The authors state that "ival-tracking" (neurons and ensembles that presumably track the number of pellets being delivered on the immediate lever - a fancy name for "expectations") "taps into a resourced-based form of cognitive effort", but no evidence is actually provided that keeping track of the expectation of reward on the immediate lever depends on attention or mnemonic resources. They also state that a "dLP-biased strategy" (waiting out the delay) is a "resistance-based form of cognitive effort" but no evidence is made that going to the delayed side takes effort.

We challenge the reviewers that assertation ival tracking is a “fancy name for expectations”. We make no claim about the prospective or retrospective nature of the signal. Clearly, expectations should be prospective and therefore different from ival tracking. Regarding the resistance signal: First, animals avoid the delay lever more often at the 8 sec delay (Figure 1). We have shown that increasing the delay systematically biases responses AWAY from the delay (Linsenbardt et al., 2017). This is consistent with a well-developed literature that rats and mice do not like waiting for delayed reinforcers. We contend that enduring something you don’t like takes effort. 

The authors talk about theta synchrony, but never actually measure theta synchrony, particularly across structures such as amygdala or ventral hippocampus. The authors try to connect this to "the unpleasantness of the delay", but provide no measures of pleasantness or unpleasantness. They have no evidence that waiting out an 8s delay is unpleasant.

We have added spike-field coherence to better contact the literature on synchrony. Note that we never refer to our results as “synchrony”. However, we would be remiss to not address the growing literature on theta synchrony in effort allocation. There is a well-developed literature that rats and mice do not like waiting for delayed reinforcers. If waiting out the delay was not pleasant then why do the animals forgo larger rewards to avoid it? 

The authors hypothesize that the "ival-tracking signal" (the expectation of number of pellets on the immediate lever) "could simply reflect the emotional or autonomic response". Aside from the fact that no evidence for this is provided, if this were to be true, then, in what sense would any of these signals be related to cognitive control?

This is proposed as an alternative explanation to the ival signal in the discussion. It was added as our due diligence. Emotional state could provide feedback to the currently implemented control mechanism. If waiting for reinforcement is too unpleasant this could drive them to ival tracking and choosing the immediate option more frequently. We provide this option only as a possibility, not a conclusion. We have clarified this in the revised text. Nevertheless, based on our review of the literature, autonomic tracking in some form, seems to be the most likely function of ACC (Seamans & Floresco 2022). While the reviewer may disagree with this, we feel it is at least as valid as all the complex, cognitively-based interpretations that commonly appear in the literature.

Reviewer #2 (Public Review):

Summary:

This manuscript explores the neuronal signals that underlie resistance vs resource-based models of cognitive effort. The authors use a delayed discounting task and computational models to explore these ideas. The authors find that the ACC strongly tracks value and time, which is consistent with prior work. Novel contributions include quantification of a resource-based control signal among ACC ensembles, and linking ACC theta oscillations to a resistance-based strategy.

Strengths:

The experiments and analyses are well done and have the potential to generate an elegant explanatory framework for ACC neuronal activity. The inclusion of local-field potential / spike-field analyses is particularly important because these can be measured in humans.

Thank you for the endorsement of our work.

Weaknesses:

I had questions that might help me understand the task and details of neuronal analyses.

(1) The abstract, discussion, and introduction set up an opposition between resource and resistancebased forms of cognitive effort. It's clear that the authors find evidence for each (ACC ensembles = resource, theta=resistance?) but I'm not sure where the data fall on this dichotomy.

(a) An overall very simple schematic early in the paper (prior to the MCML model? or even the behavior) may help illustrate the main point.

(b) In the intro, results, and discussion, it may help to relate each point to this dichotomy.

(c) What would resource-based signals look like? What would resistance based signals look like? Is the main point that resistance-based strategies dominate when delays are short, but resource-based strategies dominate when delays are long?

(d) I wonder if these strategies can be illustrated? Could these two measures (dLP vs ival tracking) be plotted on separate axes or extremes, and behavior, neuronal data, LFP, and spectral relationships be shown on these axes? I think Figure 2 is working towards this. Could these be shown for each delay length? This way, as the evidence from behavior, model, single neurons, ensembles, and theta is presented, it can be related to this framework, and the reader can organize the findings.

These are excellent suggestions, and we have implemented them, where possible. 

(2) The task is not clear to me.

(a) I wonder if a task schematic and a flow chart of training would help readers.

Yes, excellent idea, we have now included this in Figure 1. 

(b) This task appears to be relatively new. Has it been used before in rats (Oberlin and Grahame is a mouse study)? Some history / context might help orient readers.

Indeed, this task has been used in rats in several prior studies in rats. Please see the following references (PMID: 39119916, 31654652, 28000083, 26779747, 12270518, 19389183).

(c) How many total sessions were completed with ascending delays? Was there criteria for surgeries? How many total recording sessions per animal (of the 54?)

Please note that the delay does not change within a session. There were no criteria for surgery. 

(d) How many trials completed per session (40 trials OR 45 minutes)? Where are there errors? These details are important for interpreting Figure 1.

Every animal in this data set completed 40 trials and we have updated the task description to clarify this issue. There are no errors in this task, but rather the task is designed to the tendency to make an impulsive choice (smaller reward now). 

(3) Figure 1 is unclear to me.

(a) Delayed vs immediate lever presses are being plotted - but I am not sure what is red, and what is blue. I might suggest plotting each animal.

We have updated Figure 1 considerably for clarity. 

(b) How many animals and sessions go into each data point?

We hope this is clarified now with our new group assignments as all sessions were included in the analysis. 

(c) Table 1 (which might be better referenced in the paper) refers to rats by session. Is it true that some rats (2 and 8) were not analyzed for the bulk of the paper? Some rats appear to switch strategies, and some stay in one strategy. How many neurons come from each rat?

We have updated Table 1 based on our new groupings. The rats that contribute the most sessions also tend to be represented across the behavioral groups therefore it is unlikely that effort allocation strategies across groupings are an esoteric feature of an animal. 

(d) Task basics - RT, choice, accuracy, video stills - might help readers understand what is going into these plots

(e) Does the animal move differently (i.e., RTs) in G1 vs. G2?

Excellent suggestion. We have added more analysis of the task variables in the revision (e.g. RT, choice comparisons across delays, etc…)

(4) I wasn't sure how clustered G1 vs. G2 vs G3 are. To make this argument, the raw data (or some axis of it) might help.

(a) This is particularly important because G3 appears to be a mix of G1 and G2, although upon inspection, I'm not sure how different they really are

(b) Was there some objective clustering criteria that defined the clusters?

(c) Why discuss G3 at all? Can these sessions be removed from analysis?

Based on our updates to the behavioral analysis these comments are no longer relevant. 

(5) The same applies to neuronal analyses in Fig 3 and 4

(a) What does a single neuron peri-event raster look like? I would include several of these.

(b) What does PC1, 2 and 3 look like for G1, G2, and G3?

(c) Certain PCs are selected, but I'm not sure how they were selected - was there a criteria used? How was the correlation between PCA and ival selected? What about PCs that don't correlate with ival?

(d) If the authors are using PCA, then scree plots and PETHs might be useful, as well as comparisons to PCs from time-shuffled / randomized data.

We hope that our reworking of the neural data analysis has clarified these issues. We now include several firing rate examples and aggregate data.   

(6) I had questions about the spectral analysis

(a) Theta has many definitions - why did the authors use 6-12 Hz? Does it come from the hippocampal literature, and is this the best definition of theta? What about other bands (delta - 1-4 Hz), theta (4-7 Hz); and beta - 13- 30 Hz? These bands are of particular importance because they have been associated with errors, dopamine, and are abnormal in schizophrenia and Parkinson's disease.

This designation comes mainly from the hippocampal and ACC literature in rodents. In addition, this range best captured the peak in the power spectrum in our data. Note that we focus our analysis on theta give the literature regarding theta in the ACC as a correlate of cognitive controls (references in manuscript). We did interrogate other bands as a sanity check and the results were mostly limited to theta. Given the scope of our manuscript and the concerns raised regarding complexity we are concerned that adding frequency analyses beyond theta obfuscates the take home message.

However, the spectrograms in Figure 3 show a range of frequencies and highlight the ones in the theta band as the most dynamic prior to the choice. 

(b) Power spectra and time-frequency analyses may justify the authors focus. I would show these (yaxis - frequency, x-axis - time, z-axis, power).

Thank you for the suggestion. We have added this to Figure 3.    

(7) PC3 as an autocorrelation doesn't seem the to be right way to infer theta entrainment or spikefield relationships, as PCA can be vulnerable to phantom oscillations, and coherence can be transient. It is also difficult to compare to traditional measures of phase-locking. Why not simply use spike-field coherence? This is particularly important with reference to the human literature, which the authors invoke.

Excellent suggestion. Note that PCA provided a way to classify neurons that exhibited peaks in the autocorrelation at theta frequencies. We have added spike-field coherence, and this analysis confirms the differences in theta entrainment of the spike trains across the behavioral groups. Please see Figure 6D.   

Reviewer #3 (Public Review):

Summary:

The study investigated decision making in rats choosing between small immediate rewards and larger delayed rewards, in a task design where the size of the immediate rewards decreased when this option was chosen and increased when it was not chosen. The authors conceptualise this task as involving two different types of cognitive effort; 'resistance-based' effort putatively needed to resist the smaller immediate reward, and 'resource-based' effort needed to track the changing value of the immediate reward option. They argue based on analyses of the behaviour, and computational modelling, that rats use different strategies in different sessions, with one strategy in which they consistently choose the delayed reward option irrespective of the current immediate reward size, and another strategy in which they preferentially choose the immediate reward option when the immediate reward size is large, and the delayed reward option when the immediate reward size is small. The authors recorded neural activity in anterior cingulate cortex (ACC) and argue that ACC neurons track the value of the immediate reward option irrespective of the strategy the rats are using. They further argue that the strategy the rats are using modulates their estimated value of the immediate reward option, and that oscillatory activity in the 6-12Hz theta band occurs when subjects use the 'resistancebased' strategy of choosing the delayed option irrespective of the current value of the immediate reward option. If solid, these findings will be of interest to researchers working on cognitive control and ACCs involvement in decision making. However, there are some issues with the experiment design, reporting, modelling and analysis which currently preclude high confidence in the validity of the conclusions.

Strengths:

The behavioural task used is interesting and the recording methods should enable the collection of good quality single unit and LFP electrophysiology data. The authors recorded from a sizable sample of subjects for this type of study. The approach of splitting the data into sessions where subjects used different strategies and then examining the neural correlates of each is in principle interesting, though I have some reservations about the strength of evidence for the existence of multiple strategies.

Thank you for the positive comments. 

Weaknesses:

The dataset is very unbalanced in terms of both the number of sessions contributed by each subject, and their distribution across the different putative behavioural strategies (see table 1), with some subjects contributing 9 or 10 sessions and others only one session, and it is not clear from the text why this is the case. Further, only 3 subjects contribute any sessions to one of the behavioural strategies, while 7 contribute data to the other such that apparent differences in brain activity between the two strategies could in fact reflect differences between subjects, which could arise due to e.g. differences in electrode placement. To firm up the conclusion that neural activity is different in sessions where different strategies are thought to be employed, it would be important to account for potential cross-subject variation in the data. The current statistical methods don't do this as they all assume fixed effects (e.g. using trials or neurons as the experimental unit and ignoring which subject the neuron/trial came from).

In the revised manuscript we have updated the group assignments. We have improved our description of the logic and methods for employing these groupings as well. With this new approach, all sessions are now included in the analysis. The group assignments are made purely on the behavioral statistics of an animal in each session. We feel this approach is preferable to eliminating neurons or session with the goal of balancing them, which may introduce bias. Further, the rats that contribute the most sessions also tend to be represented across the behavioral groups therefore it is unlikely that effort allocation strategies across groupings are an esoteric feature of an animal. As neurons are randomly sampled from each animal on a given session, we feel that we’re justified in treating these as fixed effects.   

It is not obvious that the differences in behaviour between the sessions characterised as using the 'G1' and 'G2' strategies actually imply the use of different strategies, because the behavioural task was different in these sessions, with a shorter wait (4 seconds vs 8 seconds) for the delayed reward in the G1 strategy sessions where the subjects consistently preferred the delayed reward irrespective of the current immediate reward size. Therefore the differences in behaviour could be driven by difference in the task (i.e. external world) rather than a difference in strategy (internal to the subject). It seems plausible that the higher value of the delayed reward option when the delay is shorter could account for the high probability of choosing this option irrespective of the current value of the immediate reward option, without appealing to the subjects using a different strategy.

Further, even if the differences in behaviour do reflect different behavioural strategies, it is not obvious that these correspond to allocation of different types of cognitive effort. For example, subjects' failure to modify their choice probabilities to track the changing value of the immediate reward option might be due simply to valuing the delayed reward option higher, rather than not allocating cognitive effort to tracking immediate option value (indeed this is suggested by the neural data). Conversely, if the rats assign higher value to the delayed reward option in the G1 sessions, it is not obvious that choosing it requires overcoming 'resistance' through cognitive effort.

The RL modelling used to characterise the subject's behavioural strategies made some unusual and arguably implausible assumptions:

Thank you for the feedback, based on these comments (and those above) we have completely reworked the RL model. In addition, we’ve taken care to separate out the variables that correspond to a resistance- versus a resource-based signal. 

There were also some issues with the analyses of neural data which preclude strong confidence in their conclusions:

Figure 4I makes the striking claim that ACC neurons track the value of the immediately rewarding option equally accurately in sessions where two putative behavioural strategies were used, despite the behaviour being insensitive to this variable in the G1 strategy sessions. The analysis quantifies the strength of correlation between a component of the activity extracted using a decoding analysis and the value of the immediate reward option. However, as far as I could see this analysis was not done in a cross-validated manner (i.e. evaluating the correlation strength on test data that was not used for either training the MCML model or selecting which component to use for the correlation). As such, the chance level correlation will certainly be greater than 0, and it is not clear whether the observed correlations are greater than expected by chance.

We have added more rigorous methods to assess the ival tracking signal (Figure 4 and 5). In addition, we’ve dropped the claim that ival tracking is the same across the behavioral groups. We suspect that this was an artifact of a suboptimal group assignment approach in the previous version. 

An additional caveat with the claim that ACC is tracking the value of the immediate reward option is that this value likely correlates with other behavioural variables, notably the current choice and recent choice history, that may be encoded in ACC. Encoding analyses (e.g. using linear regression to predict neural activity from behavioural variables) could allow quantification of the variance in ACC activity uniquely explained by option values after controlling for possible influence of other variables such as choice history (e.g. using a coefficient of partial determination).

We agree that the ival tracking signal may be influenced by other variables – especially ones that are not cognitive but rather more generated by the autonomic system. We have included a discussion of this possibility in the Discussion section. Our previous work has explored the role of choice history on neural activity, please see White et al., (2024). 

Figure 5 argues that there are systematic differences in how ACC neurons represent the value of the immediate option (ival) in the G1 and G2 strategy sessions. This is interesting if true, but it appears possible that the effect is an artefact of the different distribution of option values between the two session types. Specifically, due to the way that ival is updated based on the subjects' choices, in G1 sessions where the subjects are mostly choosing the delayed option, ival will on average be higher than in G2 sessions where they are choosing the immediate option more often. The relative number of high, medium and low ival trials in the G1 and G2 sessions will therefore be different, which could drive systematic differences in the regression fit in the absence of real differences in the activity-value relationship. I have created an ipython notebook illustrating this, available at: https://notebooksharing.space/view/a3c4504aebe7ad3f075aafaabaf93102f2a28f8c189ab9176d48 07cf1565f4e3. To verify that this is not driving the effect it would be important to balance the number of trials at each ival level across sessions (e.g. by subsampling trials) before running the regression.

This is an excellent point and lead us to abandon the linear regression-based approach to quantify differences in ival coding across behavioral groups.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

This paper was extremely hard to read. In addition to the issues raised in the public review (overly complex and incomplete analyses), one of the hardest things to deal with was the writing.

Thank you for the feedback. Hopefully we have addressed this with our thorough rewrite. 

The presentation was extremely hard to follow. I had to read through it several times to figure out what the task was. It wasn't until I got to the RL model Figure 2A that I realized what was really going on with the task. I strongly recommend having an initial figure that lays out the actual task (without any RL or modeling assumptions) and identifies the multiple different kinds of sessions. What is the actual data you have to start with? That was very unclear.

Excellent idea. We have implemented this in Figure 1.  

Labeling session by "group" is very confusing. I think most readers take "group" as the group of subjects, but that's not what you mean at all. You mean some sessions were one way and some were another. (And, as I noted in the public review, you ignore many of the sessions, which I think is not OK.) I think a major rewrite would help a lot. Also, I don't think the group analysis is necessary at all. In the public review, I recommend doing the analyses very differently and more classically.

We have updated the group assignments in a manner that is more intuitive, reflects the delays, and includes all sessions.  

The paper is full of arbitrary abbreviations that are completely unnecessary. Every time I came to "ival", I had to translate that into "number of pellets delivered on the immediate lever" and every time I came to dLP, I had to translate that into "delayed lever press". Making the text shorter does not make the text easier to read. In general, I was taught that unless the abbreviation is the common term (such as "DNA" not "deoxyribonucleic acid"), you should never use an abbreviation. While there are some edge cases (ACC probably over "anterior cingulate cortex"), dLP, iLP, dLPs, iLPs, ival, are definitely way over the "don't do that" line.

We completely agree here and apologize for the excessive use of abbreviations. We have removed nearly all of them

The figures were incomplete, poorly labeled, and hard to read. A lot of figures were missing, for example

Basic task structure

Basic behavior on the task

Scatter plot of the measures that you are clustering (lever press choice X number of pellets on the immediate lever, you can use color or multiple panels to indicate the delay to the delayed lever) Figure 3 is just a couple of examples. That isn't convincing at all.

Figure 4 is missing labels. In Figure 4, I don't understand what you are trying to say.

I don't see how the results on page 16 arise from Figure 6. I strongly recommend starting from the actual data and working your way to what it means rather than forcing this into this unreasonable "session group" analysis.

We have completely reworked the Figures for clarity and content. 

The statement that "no prior study has explored the cellular correlates of cognitive effort" is ludicrous and insulting. There are dozens of experiments looking at ACC in cognitive effort tasks, in humans, other primates, and rodents. There are many dozens of experiments looking at cellular correlates in intertemporal choice tasks, some with neural manipulations, some with ensemble recordings. There are many dozens of experiments looking at cellular relationships to waiting out a delay.

We agree that our statement was extremely imprecise. We have updated this to say:  “Further, a role for theta oscillations in allocating physical effort has been identified. However, the cellular

mechanisms within the ACC that control and deploy types of cognitive effort have not been identified.”

Reviewer #2 (Recommendations For The Authors):

In Figure 2, the panels below E and F are referred to as 'right' - but they are below? I would give them letters.

I would make sure that animal #s, neuron #s, and LFP#s are clearly presented in the results and in each figure legend. This is important to follow the results throughout the manuscript.

Some additional proofreading ('Fronotmedial') might help with clarity.

Based on our updates, this is no longer relevant.  

Reviewer #3 (Recommendations For The Authors):

In addition to the suggestions above to address specific issues, it would be useful to report some additional information about aspects of the experiments and analyses:

Specify how spike sorting was performed and what metrics were used to select well isolated single units.

Done.

Provide histology showing the recording locations for each subject.

Histological assessments of electrodes placements are provided in White et al. 2024, but we provide an example placement. This has been added to the text. 

Indicate the sequence of recording sessions that occurred for each subject, including for each session what delay duration was used and which dataset the session contributed to, and indicate when the neural probes were advanced between sessions.

We feel that this adds complexity unnecessarily as we make no claims about holding units across sessions for differences in coding in the dorsoventral gradient of ACC. 

Indicate the experimental unit when reporting uncertainty measures in figure legends (e.g. mean +/- SEM across sessions).

Done.

The explanations in the feedback don't correspond well to the titles of the type of response. I suggest changing the titles to a dialogue/action that reflects the type of response

This seems illogical. I think they are referring to collusion, so perhaps another statement about ensuring that collaboration doesn't mean collusion/copying.

eLife Assessment

This study investigates how the HIV inhibitor lenacapavir influences capsid mechanics and interactions with the nuclear pore complex. It provides important insights into how drug-induced hyperstabilization of the viral shell can compromise its structural integrity during nuclear entry. While the modeling is technically sophisticated and the results are promising, some mechanistic interpretations rely on assumptions embedded in the simulations, leaving parts of the evidence incomplete.

Reviewer #1 (Public review):

The paper from Hudait and Voth details a number of coarse-grained simulations as well as some experiments focused on the stability of HIV capsids in the presence of the drug lenacapavir. The authors find that LEN hyperstabilizes the capsid, making it fragile and prone to breaking inside the nuclear pore complex.

I found the paper interesting. I have a few suggestions for clarification and/or improvement.

(1) How directly comparable are the NPC-capsid and capsid-only simulations? A major result rests on the conclusion that the kinetics of rupture are faster inside the NPC, but are the numbers of LENs bound identical? Is the time really comparable, given that the simulations have different starting points? I'm not really doubting the result, but I think it could be made more rigorous/quantitative.

(2) Related to the above, it is stated on page 12 that, based on the estimated free-energy barrier, pentamer dissociation should occur in ~10 us of CG time. But certainly, the simulations cover at least this length of time?

(3) At first, I was surprised that even in a CG simulation, LEN would spontaneously bind to the correct site. But if I read the SI correctly, LEN was parameterized specifically to bind to hexamers and not pentamers. This is fine, but I think it's worth describing in the main text.

Reviewer #2 (Public review):

Here, Hudait et al. use CG modeling to investigate the mechanism by which lenacapavir (LEN) treats HIV capsids that dock to the nuclear pore complex (NPC). However, the manuscript fails to present meaningful findings that were previously unreported in the literature, and is thus of low impact. Many claims made in the manuscript are not substantiated by the presented data. Key mechanistic details that the work purports to reveal are artifacts of the parameterization choices or simulation/analysis design, with the simulations said to reveal details that they were specifically biased to reproduce. This makes the manuscript highly problematic, as its contributions to the literature would represent misconceptions based on oversights in modeling, and thus mislead future readers.

(1) Considering the literature, it is unclear that the manuscript presents new scientific discoveries. The following are results from this paper that have been previously reported:

(a) LEN-bound capsid can dock to the nuclear pore (Figure 2; see e.g. 10.1016/j.cell.2024.12.008 or 10.1128/mbio.03613-24).

(b) NUP98 interacts with the docked capsid (Figure 2; see e.g. 10.1016/j.virol.2013.02.008 or 10.1038/s41586-023-06969-7 or 10.1016/j.cell.2024.12.008).

(c) LEN and NUP98 compete for a binding interface (Figure 2; see e.g. 10.1126/science.abb4808 or 10.1371/journal.ppat.1004459).

(d) LEN creates capsid defects (Figure 3 and 5, see e.g. 10.1073/pnas.2420497122).

(e) RNP can emerge from a damaged capsid (Figure 3 and 5; see e.g. 10.1073/pnas.2117781119 or 10.7554/eLife.64776).

(f) LEN hyperstabilizes/reduces the elasticity of the capsid lattice (Figure 6; see e.g. 10.1371/journal.ppat.1012537).

(2) The mechanistic findings related to how these processes occur are problematic, either based on circular reasoning or unsubstantiated, based on the presented data. In some cases, features of parameterization and simulation/analysis design are erroneously interpreted as predictions by the CG models.

(a) Claim: LEN-bound capsids remain associated with the NPC after rupture. CG simulations did not reach the timescale needed to demonstrate continued association or failure to translocate, leaving the claim unsubstantiated.

(b) Claim: LEN contributes to loss of capsid elasticity. The authors do not measure elasticity here, only force constants of fluctuations between capsomers in freely diffusing capsids. Elasticity is defined as the ability of a material to undergo reversible deformation when subjected to stress. Other computational works that actually measure elasticity (e.g., 0.1371/journal.ppat.1012537) could represent a point of comparison, but are not cited. The changes in force constants in the presence of LEN are shown in Figure 6C, but the text of the scale bar legend and units of k are not legible, so one cannot discern the magnitude or significance of the change.

(c) Claim: Capsid defects are formed along striated patterns of capsid disorder. Data is not presented that correlates defects/cracks with striations.

(d) Claim: Typically 1-2 LEN, but rarely 3 bind per capsid hexamer. The authors state: "The magnitude of the attractive interactions was adjusted to capture the substoichiometric binding of LEN to CA hexamers (Faysal et al., 2024). ... We simulated LEN binding to the capsid cone (in the absence of NPC), which resulted in a substoichiometric binding (~1.5 LEN per CA hexamer), consistent with experimental data (Singh et al., 2024)." This means LEN was specifically parameterized to reproduce the 1-2 binding ratio per hexamer apparent from experiments, so this was a parameterization choice, not a prediction by CG simulations as the authors erroneously claim: "This indicates that the probability of binding a third LEN molecule to a CA hexamer is impeded, likely due to steric effects that prevent the approach of an incoming molecule to a CA hexamer where 2 LEN molecules are already associated. ... Approximately 20% of CA hexamers remain unoccupied despite the availability of a large excess of unbound LEN molecules. This suggests a heterogeneity in the molecular environment of the capsid lattice for LEN binding." These statements represent gross over-interpretation of a bias deliberately introduced during parameterization, and the "finding" represents circular reasoning. Also, if "steric effects" play any role, the authors could analyze the model to characterize and report them rather than simply speculate.

(e) Claim: Competition between NUP98 and LEN regulates capsid docking. The authors state: "A fraction of LEN molecules bound at the narrow end dissociate to allow NUP98 binding to the capsid ... Therefore, LEN can inhibit the efficient binding of the viral cores to the NPC, resulting in an increased number of cores in the cytoplasm." Capsid docking occurs regardless of the presence of LEN, and appears to occur at the same rate as the LEN-free capsid presented in the authors' previous work (Hudait &Voth, 2024). The presented data simply show that there is a fluctuation of bound LEN, with about 10 fewer (<5%) bound at the end of the simulation than at the beginning, and the curve (Figure 2A) does not clearly correlate with increased NUP98 contact. In that case, no data is shown that connects LEN binding with the regulation of the docking process. Further, the two quoted statements contradict each other. The presented data appear to show that NUP outcompetes LEN binding, rather than LEN inhibiting NUP binding. The "Therefore" statement is an attempt to reconcile with experimental studies, but is not substantiated by the presented data.

(f) Claim: LEN binding leads to spontaneous dissociation of pentamers. The CG simulation trajectories show pentamer dissociation. However, it is quite difficult to believe that a pentamer in the wide end of the capsid would dissociate and diffuse 100 nm away before a hexamer in the narrow end (previously between two pentamers and now only partially coordinated, also in a highly curved environment, and further under the force of the extruding RNA) would dissociate, as in Figure 2B. A more plausible explanation could be force balance between pent-hex versus hex-hex contacts, an aspect of CG parameterization. No further modeling is presented to explain the release of pentamers, and changes in pent-hex stiffness are not apparent in the force constant fluctuation analysis in Figure 6C.

(g) Claim: WTMetaD simulations predict capsid rupture. The authors state: "In WTMetaD simulations, we used the mean coordination number (Figure S6) between CA proteins in pentamers and in hexamers as the reaction coordinate." This means that the coordination number, the number of pent-hex contacts, is the bias used to accelerate simulation sampling. Yet the authors then interpret a change in coordination number leading to capsid rupture as a discovery, representing a fundamental misuse of the WTMetaD method. Changes in coordination number cannot be claimed as an emergent property when they are in fact the applied bias, when the simulation forced them to sample such states. The bias must be orthogonal to the feature of interest for that feature to be discoverable. While the reported free energies are orthogonal to the reaction coordinate, the structural and stepwise-mechanism "findings" here represent circular reasoning.

(3) Another major concern with this work is the excessive self-citation, and the conspicuous lack of engagement with similar computational modeling studies that investigate the HIV capsid and its interactions with LEN, capsid mechanical properties relevant to nuclear entry, and other capsid-NPC simulations (e.g., 10.1016/j.cell.2024.12.008 and 10.1371/journal.ppat.1012537). Other such studies available in the literature include examination of varying aspects of the system at both CG and all-atom levels of resolution, which could be highly complementary to the present work and, in many cases, lend support to the authors' claims rather than detract from them. The choice to omit relevant literature implies either a lack of perspective or a lack of collegiality, which the presentation of the work suffers from. Overall, it is essential to discuss findings in the context of competing studies to give readers an accurate view of the state of the field and how the present work fits into it. It is appropriate in a CG modeling study to discuss the potential weaknesses of the methodology, points of disagreement with alternative modeling studies, and any lack of correlation with a broader range of experimental work. Qualitative agreement with select experiments does not constitute model validation.

(4) Other critiques, questions, concerns:

(a) The first Results sub-heading presents "results", complete with several supplementary figures and a movie that are from a previous publication about the development of the HIV capsid-NPC model in the absence of LEN (Hudait &Voth, 2024). This information should be included as part of the introduction or an abbreviated main-text methods section rather than being included within Results as if it represents a newly reported advancement, as this could be misleading.

(b) The authors say the unbiased simulations of capsid-NPC docking were run as two independent replicates, but results from only one trajectory are ever shown plotted over time. It is not mentioned if the time series data are averaged or smoothed, so what is the shadow in these plots (e.g., Figures 1,2, and Supplementary Figure 5)?

(c) Why do the insets showing LEN binding in Figure 2A look so different from the models they are apparently zoomed in on? Both instances really look like they are taken from different simulation frames, rather than being a zoomed-in view.

(d) What are the sudden jerks apparent in the SI movies? Perhaps this is related to the rate at which trajectory frames are saved, but occasionally, during the relatively smooth motion of the capsid-NPC complex, something dramatic happens all of a sudden in a frame. For example, significant and apparently instantaneous reorientation of the cone far beyond what preceding motions suggest is possible (SI movie 2, at timestamp 0.22), RNP extrusion suddenly in a single frame (SI movie 2, at timestamp 0.27), and simultaneous opening of all pentamers all at once starting in a single frame (SI movie 2, at timestamp 0.33). This almost makes the movie look generated from separate trajectories or discontinuous portions of the same trajectory. If movies have been edited for visual clarity (e.g., to skip over time when "nothing" is happening and focus on the exciting aspects), then the authors should state so in the captions.

(e) Figure 3c presents a time series of the degree of defects at pent-hex and hex-hex interfaces, but I do not understand the normalization. The authors state, "we represented the defects as the number of under-coordinated CA monomers of the hexamers at the pentamer-hexamer-pentamer and hexamer-hexamer interface as N_Pen-Hex and N_Hex-Hex ... Note that in N_Pen-Hex and N_Hex-Hex are calculated by normalizing by the total number of CA pentamer (12) and hexamer rings (209) respectively." Shouldn't the number of uncoordinated monomers be normalized by the number of that type of monomer, rather than the number of capsomers/rings? E.g., 12*5 and 209*6, rather than 12 and 209?

(f) The authors state that "Although high computational cost precluded us from continuing these CG MD simulations, we expect these defects at the hexamer-hexamer interface to propagate towards the high curvature ends of the capsid." The defects being reported are apparently propagating from (not towards) the high curvature ends of the capsid.

(g) The first half of the paper uses the color orange in figures to indicate LEN, but the second half uses orange to indicate defects, and this could be confusing for some readers. Both LEN and "defects" are simply a cluster of spheres, so highlighted defects appear to represent LEN without careful reading of captions.

(h) SI Figure S3 captions says "The CA monomers to which at least one LEN molecule is bound are shown in orange spheres. The CA monomers to which no LEN molecule is bound are shown in white spheres. " While in contradiction, the main-text Fig 2 says "The CA monomers to which at least one LEN molecule is bound are shown in white spheres. The CA monomers to which no LEN molecule is bound are shown in orange spheres. " One of these must be a typo.

(i) The authors state that: "CG MD simulations and live-cell imaging demonstrate that LEN-treated capsids dock at the NPC and rupture at the narrow end when bound to the central channel and then remain associated to the NPC after rupture." However, the live cell imaging data do not show where rupture occurs, such that this statement is at least partially false. It is also unclear that CG simulations show that cores remain bound following rupture, given that simulations were not extended to the timescale needed to observe this, again rendering the statement partially false.

(j) The authors state: "We previously demonstrated that the RNP complex inside the capsid contributes to internal mechanical strain on the lattice driven by CACTD-RNP interactions and condensation state of RNP complex (Hudait &Voth, 2024). " In that case, why do the present CG models detect no difference in results for condensed versus uncondensed RNP?

(k) The authors state: "The distribution demonstrates that the binding of LEN to the distorted lattice sites is energetically favorable. Since LEN localizes at the hydrophobic pocket between two adjoining CA monomers, it is sterically favorable to accommodate the incoming molecule at a distorted lattice site. This can be attributed to the higher available void volume at the distorted lattice relative to an ordered lattice, the latter being tightly packed. This also allows the drug molecule to avoid the multitude of unfavorable CA-LEN interactions and establish the energetically favorable interactions leading to a successful binding event. " What multitude of unfavorable interactions are the authors referring to? Data is not presented to substantiate the claim of increased void volume between hexamers in the distorted lattice. Capsomer distortion is shown as a schematic in Figure 6A rather than in the context of the actual model.

(l) The authors state that "These striated patterns also demonstrate deviations from ideal lattice packing. " What does ideal lattice packing mean in this context, where hexamers are in numerous unique environments in terms of curvature? What is the structural reference point?

(m) If pentamer-hexamer interactions are weakened in the presence of LEN, why are differences at these interfaces not apparent in the Figure 6C data that shows stiffening of the interactions between capsomer subunits?

(n) The authors state: "Lattice defects arising from the loss of pentamers and cracks along the weak points of the hexameric lattice drive the uncoating of the capsid." The word rupture or failure should be used here rather than uncoating; it is unclear that the authors are studying the true process of uncoating and whether the defects induced by LEN binding relate in any way to uncoating.

(o) The authors state: "LEN-treated broken cores are stabilized by the interaction with the disordered FG-NUP98 mesh at the NPC." But no data is presented to demonstrate that capsid stability is increased by NUP98 interaction. In fact, the presented data could suggest the opposite since capsids in contact with NUP98 in the NPC appeared to rupture faster than freely diffusing capsids.

(p) The authors state: "LEN binding stimulates similar changes in free capsids, but they occur with lower frequency on similar time scales, suggesting that the cores docked at the NPC are under increased stress, resulting in more frequent weakening of the hexamer-pentamer and hexamer-hexamer interactions, as well as more nucleation of defects at the hexamer-hexamer<br /> Interface. ... Our results suggest that in the presence of the LEN, capsid docking into the NPC central channel will increase stress, resulting in more frequent breaks in the capsid lattice compared to free capsids." The first is a run-on sentence. The results shown support that LEN stimulates changes in free capsids to happen faster, but not more frequently. The frequency with which an event occurs is separate from the speed with which the event occurs.

(q) The authors state: "A possible mechanistic pathway of capsid disassembly can be that multiple pentamers are dissociated from the capsid sequentially, and the remaining hexameric lattice remains stabilized by bound LEN molecules for a time, before the structural integrity of the remaining lattice is compromised." This statement is inconsistent with experimental studies that say LEN does not lead to capsid disassembly, and may even prevent disassembly as part of its disruption of proper uncoating (e.g., 10.1073/pnas.2420497122 previously published by the authors).

(r) Finally, it remains a concern with the authors' work that the bottom-up solvent-free CG modeling software used in this and supporting works is not open source or even available to other researchers like other commonly used molecular dynamics software packages, raising significant questions about transparency and reproducibility.

Author response:

Before providing a brief provisional response to the two reviews, it is important to reiterate a few key points about our work. First, our paper is largely a computational biophysics paper, augmented by experimental results. Generally speaking, computational biophysics work intends to achieve one of two things (or both). One is to provide more molecular level insight into various behaviors of biomolecular systems that have not been (or cannot be) provided by qualitative experimental results alone. The second general goal of computational biophysics it to formulate new hypotheses to be tested subsequently by experiment. In our paper, we have achieved both of these goals and then confirmed the key computational results by experiment..

The first reviewer has some valuable points, which can be addressed as follows (and will be emphasized in the revised version of the paper): (1) Yes the simulations of capsid rupture in the NPC and capsid-only are directly comparable as both have approximately the same number of bound LEN, as determined by following the LEN-capsid interaction protocol described in the main text (around Fig 6) and in the SI section S3; (2) While we have stressed this point in several places in the manuscript, here again we stress that coarse-grained (CG) MD time is not the same as real time. The point of CG simulations is to accelerate the timescale of the MD and the associated sampling, so the CG “time” from the MD integrator needs to be rescaled to associate a real time to it. As such, our CG simulation is not representing a microsecond of real time but rather something much longer. We will emphasize this again in the revised text. (3) Actually, we think that the parameterization of the LEN model and the LEN-capsid interactions is well described in the text associated with Fig 6 and in SI section S3. It is true that this one part of the CG model was parameterized “top-down” given the good experimental structures of bound LEN to capsid and other data, but the rest of the CG model is “bottom-up” (meaning developed from well-defined coarse-graining statistical mechanics as applied to molecular level structures and interactions, see also below). 

As for the second reviewer, this review is quite problematic in our view as the reviewer seems to think that quoting a number of qualitative experimental results is sufficient to undermine the impact of our paper (they are not) and, furthermore, the reviewer appears to have a very minimal understanding of “bottom-up” CG modeling, which we have utilized. This modeling does not in fact rely on the “assumptions” this reviewer alleges we have relied on. (As an aside, it could be helpful for this reviewer to study the review by Jin et al, https://doi.org/10.1021/acs.jctc.2c00643) in order to become more familiar with the field and our approach before criticizing it.) We also note that our main HIV capsid-NPC docking model is already published in PNAS (https://doi.org/10.1073/pnas.2313737121), where it underwent rigorous peer review. In our forthcoming full response to the reviews and in the revised paper we will attempt to address a number of this reviewers comments, but the number, extent, and tone of this collection of criticisms, for us, calls into question the objectivity of this reviewer, not to mention the reviewer’s rather weak understanding of what we have done and how we have done it.

Finally, while we certainly appreciate the overall positive eLife assessment, we are disappointed by the statement “some mechanistic interpretations rely on assumptions embedded in the simulations, leaving parts of the evidence incomplete”. Of course, all simulations (and experiments) rely on certain assumptions, but we have gone to great length to provide a “bottomup” approach to our modeling, based on underlying molecular level structures and interactions, and we have provided experimental validation of the main simulation predictions. It seems that the comments of the second reviewer may have influenced this point of view, but we do not feel it is justified.

Remove 'and' or join the two sentences

eLife Assessment

This study offers valuable insights into the anatomical and physiological features of cold-selective lamina I spinal projection neurons. The evidence supporting the authors' claims is compelling, although including a larger sample size and more quantification would have strengthened the study further, and the claims of monosynaptic connectivity would benefit from being stated more cautiously. The work will interest those in the field of somatosensory biology, especially researchers studying spinal cord dorsal horn circuits and projection neuron cell types.

Reviewer #1 (Public review):

Summary:

Spinal projection neurons in the anterolateral tract transmit diverse somatosensory signals to the brain, including touch, temperature, itch, and pain. This group of spinal projection neurons is heterogeneous in their molecular identities, projection targets in the brain, and response properties. While most anterolateral tract projection neurons are multimodal (responding to more than one somatosensory modality), it has been shown that cold-selective projection neurons exist in lamina I of the spinal cord dorsal horn. Using a combination of anatomical and physiological approaches, the authors discovered that the cold-selective lamina I projection neurons are heavily innervated by Trpm8+ sensory neuron axons, with calb1+ spinal projection neurons primarily capturing these cold-selective lamina I projection neurons. These neurons project to specific brain targets, including the PBNrel and cPAG. This study adds to the ongoing effort in the field to identify and characterize spinal projection neuron subtypes, their physiology, and functions.

Strengths:

(1) The combination of anatomical and physiological analyses is powerful and offers a comprehensive understanding of the cold-selective lamina I projection neurons in the spinal cord dorsal horn. For example, the authors used detailed anatomical methods, including EM imaging of Trpm8+ axon terminals contacting the Phox2a+ lamina I projection neurons. Additionally, they recorded stimulus-evoked activity in Trpm8-recipient neurons, carefully selected by visual confirmation of tdTomato and GFP juxtaposition, which is technically challenging.

(2) This study identifies, for the first time, a molecular marker (calb1) that labels cold-selective lamina I projection neurons. Although calb1+ projection neurons are not entirely specific to cold-selective neurons, using an intersectional strategy combined with other genes enriched in this ALS group or cold-induced FosTRAP may further enhance specificity in the future.

(3) This study shows that cold-selective lamina I projection neurons specifically innervate certain brain targets of the anterolateral tract, including the NTS, PBNrel, and cPAG. This connectivity provides insights into the role of these neurons in cold sensation, which will be an exciting area for future research.

Weaknesses:

(1) The sample size for the ex vivo electrophysiology is small. Given the difficulty and complexity of the preparation, this is understandable. However, a larger sample size would have strengthened the authors' conclusions.

(2) The authors used tdTomato expression to identify brain targets innervated by these cold-selective lamina I projection neurons. Since tdTomato is a soluble fluorescent protein that fills the entire cell, using synaptophysin reporters (e.g., synaptophysin-GFP) would have been more convincing in revealing the synaptic targets of these projection neurons.

(3) The summary cartoon shown in Figure 7 can be misleading because this study did not determine whether these cold-selective lamina I projection neurons have collateral branches to multiple brain targets or if there are anatomical subtypes that may project exclusively to specific targets. For example, a recent study (Ding et al., Neuron, 2025) demonstrated that there are PBN-projecting spinal neurons that do not project to other rostral brain areas. Furthermore, based on the authors' bulk labeling experiments, the three main brain targets are NTS, PBNrel, and cPAG. The VPL projection is very sparse and almost negligible.

Reviewer #2 (Public review):

Summary:

In this study, the authors took advantage of a semi-intact ex vivo somatosensory preparation that includes hindlimb skin to characterize the response of projection neurons in the dorsal horn of the spinal cord to peripheral stimulation, including cold thermal stimuli. The main aim was to characterize the connectivity between peripheral afferents expressing the cold-sensing receptor TRPM8 and a set of genetically tagged neurons of the anterolateral system (ALS). These ALS neurons expressed high levels of the calcium-binding protein calbindin 1.

In addition, combining different viral tracing methods, the authors could identify the anatomical targets of this specific subset of projection neurons within the brainstem and diencephalon.

Strengths:

The use of a relatively new (seldom used previously) transgenic line to label TRPM8-expressing afferents, combined with the genetic characterization of a previously identified subset of projection neurons, adds a specificity to the characterization. The transgenic line appears to capture well the subpopulation of Trpm8-expressing neurons

In addition, the use of electron microscopy techniques makes the interpretation of the structural contacts more compelling.

The writing is clear, and the presentation of findings follows a logical flow.

Overall, this study provides solid, novel information about the brain circuits involved in cold thermosensation.

Weaknesses:

In the characterization of recorded neurons in close contact or in the absence of this contact with TRPM8 afferents, the number of recorded neurons is relatively low. In addition, the strength of thermal stimuli is not very well controlled, preventing a more precise characterization of the connectivity.

The authors could provide some sense of the effort needed to record from the 6 cold-activated neurons described. How many preparations were needed, etc?

Will students know about this framework? I don't think we can assume that they have done this. Can we remove this?

I like the idea of the video and the text is fine. She needs to pronounce 'graduate' as a verb, not a noun. She seems very robotic throughout most of the video, especially in the middle.

But we must have all, available local-first yet omni present ready to share and collaborate autonomously. Tha tis the promise of IPFS

the internet today has been described as this big nervous system for humanity, right?

eLife Assessment

This important work advances our understanding of the single neuron coding types in the mouse gustatory cortex and the functional roles of these neurons for perceptual decision-making. The conclusions are based on compelling evidence from rigorous behavioral experiments, high-density electrophysiology, sophisticated data analysis, and neural network modeling with in silico perturbations of functionally-identified units. This work will be of broad interest to systems neuroscientists.

Reviewer #1 (Public review):

This manuscript provides several important findings that advance our current knowledge about the function of the gustatory cortex (GC). The authors used high-density electrophysiology to record neural activity during a sucrose/NaCl mixture discrimination task. They observed population-based activity capable of representing different mixtures in a linear fashion during the initial stimulus sampling period, as well as representing the behavioral decision (i.e., lick left or right) at a later time point. Analyzing this data at the single neuron level, they observed functional subpopulations capable of encoding the specific mixture (e.g., 45/55), tastant (e.g., sucrose), and behavioral choice (e.g., lick left). To test the functional consequences of these subpopulations, they built a recurrent neural network model in order to "silence" specific functional subpopulations of GC neurons. The virtual ablation of these functional subpopulations altered virtual behavioral performance in a manner predicted by the subpopulation's presumed contribution.

Strengths:

Building a recurrent neural network model of the gustatory cortex allows the impact of the temporal sequence of functionally identifiable populations of neurons to be tested in a manner not otherwise possible. Specifically, the author's model links neural activity at the single neuron and population level with perceptual ability. The electrophysiology methods and analyses used to shape the network model are appropriate. Overall, the conclusions of the manuscript are well supported.

Weaknesses:

One potential concern is the apparent mismatch between the neural and behavioral data. Neural analyses indicate a clear separation of the activity associated with each mixture that is independent of the animal's ultimate choice. This would seemingly indicate that the animals are making errors despite correctly encoding the stimulus. Based solely on the neural data, one would expect the psychometric curve to be more "step-like" with a significantly steeper slope. One potential explanation for this observation is the concentration of the stimuli utilized in the mixture discrimination task. The authors utilize equivalent concentrations, rather than intensity-matched concentrations. In this case, a single stimulus can (theoretically) dominate the perception of a mixture, resulting in a biased behavioral response despite accurate concentration coding at the single neuron level. Given the difficulty of isointensity matching concentrations, this concern is not paramount. However, the apparent mismatch between the neural and behavioral data should be acknowledged/addressed in the text.

Reviewer #2 (Public review):

Lang et al. investigate the contribution of individual neuronal encoding of specific task features to population dynamics and behavior. Using a taste-based decision-making behavioral task with electrophysiology from the mouse gustatory cortex and computational modeling, the authors reveal that neurons encoding sensory, perceptual, and decision-related information with linear and categorical patterns are essential for driving neural population dynamics and behavioral performance. Their findings suggest that individual linear and categorical coding units have a significant role in cortical dynamics and perceptual decision-making behavior.

Overall, the experimental and analytical work is of very high quality, and the findings are of great interest to the taste coding field, as well as to the broader systems neuroscience field.

I have a couple of suggestions to further enhance the authors' important conclusions:

My main comment is the distinction between constrained and unconstrained units. The authors train a small percentage of units to match the real neural data (constrained units), and then find some unconstrained units that are similar to the real neural data and some that are not. As far as I could tell, the relative fraction of constrained and unconstrained units in the trained RNN is not reported; I assume the constrained ones are a much smaller population, but this is unclear. The selection of different groups of neurons for the RNN ablation experiments appears to be based on their response profiles only. Therefore, if I understood correctly, both constrained and unconstrained units and ablated together for a given response category (e.g., linear or step-perception). It would be useful, therefore, to separately compare the effects of constrained vs. unconstrained RNN units.

Specifically:

(1) For the analyses in the initial version of the manuscript, the authors should specify how many units in each ablation category are constrained and unconstrained.

(2) The authors should repeat Figure 6, but only for unconstrained units to test how much of the effects in the initial version of Figure 6 are driven by constrained vs. unconstrained RNN units.

(3) The authors should repeat Figure 7, but performing ablations separately on the constrained and unconstrained units to examine how the network behaves in each case and the resulting "behavioral" effect.

Reviewer #3 (Public review):

Primary taste cortex neurons show a variety of dynamic response profiles during taste decision-making tasks, reflecting both sensory and decision variables. In the present study, Lang et al. set out to determine how neurons with distinct response profiles contribute to perceptual decisions about taste stimuli.

The methods, with reference to the behavioral task and electrophysiological recordings/data analysis, are straightforward, solid, and appropriate. The computational model is presented in a clear and conceptually intuitive manner, although the details are outside of my area of expertise.

The experimental design features a simple 2-alternative forced-choice design that yielded clear psychometric curves across a range of stimuli. In vivo recordings were performed using Neuropixels and yielded an appropriate sample of single neuron responses. The strength of the model lies in the fact that it consists of single neurons whose response profiles mimic those recorded in vivo, and allows neuron-selective manipulation.

By virtually lesioning specific subsets of neurons in the network, the authors demonstrate that a relatively small population of neurons with specific tuning profiles was sufficient to produce the observed neural dynamics and behavioral responses. This effect was selective as lesioning other responsive neurons did not affect overall response dynamics or performance.

These findings provide new insight into the relation between the response profiles of single neurons in sensory cortex, their population-level activity dynamics, and the perceptual decisions they inform.

The approach is particularly innovative as it uses computational modeling to target functionally-defined "cell types", which cannot necessarily be targeted by more conventional genetic approaches.

Author response:

Reviewer #1 (Public review):

This manuscript provides several important findings that advance our current knowledge about the function of the gustatory cortex (GC). The authors used high-density electrophysiology to record neural activity during a sucrose/NaCl mixture discrimination task. They observed population-based activity capable of representing different mixtures in a linear fashion during the initial stimulus sampling period, as well as representing the behavioral decision (i.e., lick left or right) at a later time point. Analyzing this data at the single neuron level, they observed functional subpopulations capable of encoding the specific mixture (e.g., 45/55), tastant (e.g., sucrose), and behavioral choice (e.g., lick left). To test the functional consequences of these subpopulations, they built a recurrent neural network model in order to "silence" specific functional subpopulations of GC neurons. The virtual ablation of these functional subpopulations altered virtual behavioral performance in a manner predicted by the subpopulation's presumed contribution.

Strengths:

Building a recurrent neural network model of the gustatory cortex allows the impact of the temporal sequence of functionally identifiable populations of neurons to be tested in a manner not otherwise possible. Specifically, the author's model links neural activity at the single neuron and population level with perceptual ability. The electrophysiology methods and analyses used to shape the network model are appropriate. Overall, the conclusions of the manuscript are well supported.

Weaknesses:

One potential concern is the apparent mismatch between the neural and behavioral data. Neural analyses indicate a clear separation of the activity associated with each mixture that is independent of the animal's ultimate choice. This would seemingly indicate that the animals are making errors despite correctly encoding the stimulus. Based solely on the neural data, one would expect the psychometric curve to be more "step-like" with a significantly steeper slope. One potential explanation for this observation is the concentration of the stimuli utilized in the mixture discrimination task. The authors utilize equivalent concentrations, rather than intensity-matched concentrations. In this case, a single stimulus can (theoretically) dominate the perception of a mixture, resulting in a biased behavioral response despite accurate concentration coding at the single neuron level. Given the difficulty of isointensity matching concentrations, this concern is not paramount. However, the apparent mismatch between the neural and behavioral data should be acknowledged/addressed in the text.

We thank the Reviewer for the insightful comments and thoughtful suggestions. Our electrophysiological recordings show that GC dynamically encodes stimulus concentration of mixture elements, dominant perceptual quality, and decisions of directional lick. With regard to the encoding of mixtures, the clear separation of activity associated with each mixture (Figure 3) is present at a trial-averaged pseudo-population level, and average activities associated with more similar, intermediate mixtures are closer to each other in this space. In fact, at a single trial level activity evoked by similar, intermediate mixtures can be hard to separate. This increased similarity can lead to behavioral errors resulting from either incorrect encoding of the stimulus or from the inability to interpret the stimuli to guide the correct decision.

The psychometric function, which shows that more distinct stimuli (100/0 vs 0/100) lead to fewer mistakes than more ambiguous, intermediate mixtures (55/45 vs 55/45), is consistent with the increased ambiguity of responses to intermediate mixtures and with the possibility that, compared to pure stimuli, intermediate mixtures lead to more trials in which the binary choice component of neural activity is inverted, resulting in more directional errors.

The Reviewer is correct that there could be a slight mismatch in the perceived intensity of the mixture components. This mismatch could be the reason for the slight asymmetry in our psychometric function (Figure 1B). However, it is not uncommon for mice in these 2AC tasks to also have a motor laterality bias in their responses that manifests itself for the more ambiguous stimuli. We chose not to model this bias given its subtlety and its unknown origin. Rather, we chose to model an ideal scenario in which stimuli have matched intensity and no motor bias exists. In the revised version we will discuss this issue.

Reviewer #2 (Public review):

Lang et al. investigate the contribution of individual neuronal encoding of specific task features to population dynamics and behavior. Using a taste-based decision-making behavioral task with electrophysiology from the mouse gustatory cortex and computational modeling, the authors reveal that neurons encoding sensory, perceptual, and decision-related information with linear and categorical patterns are essential for driving neural population dynamics and behavioral performance. Their findings suggest that individual linear and categorical coding units have a significant role in cortical dynamics and perceptual decision-making behavior.

Overall, the experimental and analytical work is of very high quality, and the findings are of great interest to the taste coding field, as well as to the broader systems neuroscience field.

I have a couple of suggestions to further enhance the authors' important conclusions:

My main comment is the distinction between constrained and unconstrained units. The authors train a small percentage of units to match the real neural data (constrained units), and then find some unconstrained units that are similar to the real neural data and some that are not. As far as I could tell, the relative fraction of constrained and unconstrained units in the trained RNN is not reported; I assume the constrained ones are a much smaller population, but this is unclear. The selection of different groups of neurons for the RNN ablation experiments appears to be based on their response profiles only. Therefore, if I understood correctly, both constrained and unconstrained units and ablated together for a given response category (e.g., linear or step-perception). It would be useful, therefore, to separately compare the effects of constrained vs. unconstrained RNN units.

We thank the Reviewer for the constructive feedback and are pleased that the work is considered of broad interest. The Reviewer is correct that ablations were carried out with respect to response categories only and included both constrained and unconstrained units.

The ratio of total units to constrained units is fixed at 5.88, thus constrained units are ~17% of the network and unconstrained units are ~83%. This value is specified in the Methods (RNN: Components and dynamics), but we will report it in the Results of the revised manuscript as well for clarity.

Specifically:

(1) For the analyses in the initial version of the manuscript, the authors should specify how many units in each ablation category are constrained and unconstrained.

In the revised manuscript, we will specify the fractions of constrained and unconstrained units within each response category. For convenience, they are reported here: Linear = 194 constrained and 691 unconstrained units; Step-perception = 147 constrained and 840 unconstrained units; Step-choice = 129 constrained and 814 unconstrained units; Other = 353 constrained and 1739 unconstrained units.

(2) The authors should repeat Figure 6, but only for unconstrained units to test how much of the effects in the initial version of Figure 6 are driven by constrained vs. unconstrained RNN units.

In the revised version we will add a Supplemental Figure in which the contribution of constrained vs unconstrained units is addressed.

(3) The authors should repeat Figure 7, but performing ablations separately on the constrained and unconstrained units to examine how the network behaves in each case and the resulting "behavioral" effect.

The revised version will include a Supplemental Figure with these simulations.

Reviewer #3 (Public review):

Primary taste cortex neurons show a variety of dynamic response profiles during taste decision-making tasks, reflecting both sensory and decision variables. In the present study, Lang et al. set out to determine how neurons with distinct response profiles contribute to perceptual decisions about taste stimuli.

The methods,with reference to the behavioral task and electrophysiological recordings/data analysis, are straightforward, solid, and appropriate. The computational model is presented in a clear and conceptually intuitive manner, although the details are outside of my area of expertise.

The experimental design features a simple 2-alternative forced-choice design that yielded clear psychometric curves across a range of stimuli. In vivo recordings were performed using Neuropixels and yielded an appropriate sample of single neuron responses. The strength of the model lies in the fact that it consists of single neurons whose response profiles mimic those recorded in vivo, and allows neuron-selective manipulation.By virtually lesioning specific subsets of neurons in the network, the authors demonstrate that a relatively small population of neurons with specific tuning profiles was sufficient to produce the observed neural dynamics and behavioral responses. This effect was selective as lesioning other responsive neurons did not affect overall response dynamics or performance.These findings provide new insight into the relation between the response profiles of single neurons in sensory cortex, their population-level activity dynamics, and the perceptual decisions they inform.

The approach is particularly innovative as it uses computational modeling to target functionally-defined "cell types", which cannot necessarily be targeted by more conventional genetic approaches.

We thank the Reviewer for the positive assessment of our study.

eLife Assessment

This valuable study leverages a large global dataset of tens of thousands of tuberculosis samples to place recurrent protein-coding mutations into their three-dimensional structural context, offering an expanded view of how antibiotic resistance emerges compared to traditional genetic analyses alone. The strength of evidence is convincing, supported by the scale and breadth of the dataset and the systematic structural analysis, although some of the assumptions made in the the modeling approach are only partially supported. Overall, the work will be of broad interest to researchers studying microbial evolution, antibiotic resistance, and structure-function relationships in pathogens.

Reviewer #1 (Public review):

Summary:

In this manuscript, Green et al. attempt to use large-scale protein structure analysis to find signals of selection and clustering related to antibiotic resistance. This was applied to the whole proteome of Mycobacterium tuberculosis, with a specific focus on the smaller set of known antibiotic-resistance-related proteins.

Strengths:

The use of geospatial analysis to detect signals of selection and clustering on the structural level is really intriguing. This could have a wider use beyond the AMR-focussed work here and could be applied to a more general evolutionary analysis context. Much of the strength of this work lies in breaking ground into this structural evolution space, something rarely seen in such pathogen data. Additional further research can be done to build on this foundation, and the work presented here will be important for the field.

The size of the dataset and use of protein structure prediction via AlphaFold, giving such a consistent signal within the dataset, is also of great interest and shows the power of these approaches to allow us to integrate protein structure more confidently into evolution and selection analyses.

Weaknesses:

There are several issues with the evolutionary analysis and assumptions made in the paper, which perhaps overstate the findings, or require refining to take into account other factors that may be at play.

(1) The focus on antimicrobial resistance (AMR) throughout the paper contains the findings within that lens. This results in a few different weaknesses:

(a) While the large size of the analysis is highlighted in the abstract and elsewhere, in reality, only a few proteins are studied in depth. These are proteins already associated with AMR by many other studies, somewhat retreading old ground and reducing the novelty.

(b) Beyond the AMR-associated proteins, the proteome work is of great interest, but only casually interrogated and only in the context of AMR. There appears to be an assumption that all signals of positive selection detected are related to AMR, whereas something like cas10 is part of the CRISPR machinery, a set of proteins often under positive selection, and thus unlikely to be AMR-related.

(2) The strength of the signal from the structural information and the novelty of the structural incorporation into prediction are perhaps overstated.

(a) A drop of 13% in F1 for a gain of 2% in PPV is quite the trade-off. This is not as indicative of a strong predictor that could be used as the abstract claims. While the approach is novel and this is a good finding for a first attempt at such complex analysis, this is perhaps not as significant as the authors claim

(b) In relation to this, there is a lack of situating these findings within the wider research landscape. For instance, the use of structure for predicting resistance has been done, for example, in PncA (https://academic.oup.com/jacamr/article/6/2/dlae037/7630603, https://www.sciencedirect.com/science/article/pii/S1476927125003664, https://www.nature.com/articles/s41598-020-58635-x) and in RpoB (https://www.nature.com/articles/s41598-020-74648-y). These, and other such works, should be acknowledged as the novelty of this work is perhaps not as stark as the authors present it to be.

(3) The authors postulate that neutral AA substitutions would be randomly distributed in the protein structure and thus use random mutations as a negative control to simulate this neutral evolution. However, I am unsure if this is a true negative control for neutral evolution. The vast majority of residues would be under purifying selection, not neutral selection, especially in core proteins like rpoB and gyrA. Therefore, most of these residues would never be mutated in a real-world dataset. Therefore, you are not testing positive selection against neutral selection; you are testing positive against purifying, which will have a much stronger signal. This is likely to, in turn, overestimate the signal of positive selection. This would be better accounted for using a model of neutral evolution, although this is complex and perhaps outside the scope. Still, it needs to be made clear that these negative controls are not representative of neutral evolution.

(4) In a similar vein, the use of 15 Å as a cut-off for stating co-localisation feels quite arbitrary. The average radius of a globular protein is about 20 Å, so this could be quite a large patch of a protein. I think it may be good to situate the cut-off for a 'single location' within a size estimator of the entire protein, as 15 Å could be a neighbourhood in a large protein, but be the whole protein for smaller ones.

Reviewer #2 (Public review):

Summary:

This is an important study that, for the first time, systematically places the homoplastic genetic variation observed in the coding regions in a large collection of >31,000 M. tuberculosis samples into the protein structural context. This should be much more informative when, e.g. predicting antimicrobial resistance. The authors imaginatively apply the Getis-Ord score, which originated in geographical spatial analysis but has also been used in human disease to demonstrate that missense mutations in M. tuberculosis known to be associated with antimicrobial resistance are clustered in space. That they are able to consider almost all of the proteome using a large dataset of 31,000 M. tuberculosis complex clinical samples, which makes the evidence convincing.

Strengths:

To my knowledge, this is the first study to place the homoplastic missense mutations from a large clinical dataset into their protein structural context and attempt to look for clustering in space, which could be indicative of a recent evolutionary pressure, such as the use of antibiotics. The field usually only views resistance through the genetic paradigm, so it is delightful to see a structural paradigm being brought to bear, as this should, in theory, be much more informative, as protein structure is much closer to function. In addition, the dataset used is large (>31,000 clinical M. tuberculosis samples), and the authors are able to consider almost all of the ORFs (3,687/3,996) in the M. tuberculosis reference, and hence the analysis is comprehensive.

Weaknesses:

It is not apparent at the time of this review if the study could be reproduced by other researchers as e.g. whilst the authors state that the raw sequencing files (FASTQ) underpinning the dataset of 31,428 M. tuberculosis isolates can be downloaded the table in the Supplement containing the sample and accession identifiers contains rows that do not contain NCBI accessions e.g. '01R0685' or 'IDR 1600023875' or '1479144813357T181715lib5022nextseqn0035151bp' instead of the expected form e.g. 'SAMEA1016138'. I have searched the NCBI SRA using these terms and got no results, so they cannot be used to download any FASTQ files. There is also no information in the preprint on how the reads were processed (which is a complex process) and the dataset of SNPs subsequently built. One can trace back through the references, but I cannot find anywhere where one can download the SNP dataset, which would permit researchers to reproduce at least the latter stages of the work -- one obvious option would be to make the SNP dataset available. Likewise, the authors have constructed a "M. tuberculosis structureome", which would be very useful for the community but does not appear to be publicly available. At the time of the review, not all the GitHub repositories were public, so these points may have been rectified when that was corrected.

The authors correctly point out in the Introduction that supervised methods like GWAS or ML need datasets with matching genetic and phenotypic drug susceptibility data, which are much difficult/expensive to obtain, but don't then close the loop by comparing their results back to such supervised methods. They pick out RnJ as having previously been identified by a GWAS, but it would have provided a useful validation of their method to e.g. demonstrating that X% of the genes they identify were also identified by GWAS/ML studies, and therefore their method can achieve similar results but without having to collect pDST data.

Whilst the authors acknowledge that assuming all sites are equally likely to mutate in their random shuffling procedure is a shortcoming, a bigger weakness is, I suspect, that one should also only consider which amino acids could arise at each codon due to a SNP. Shuffling assumes any amino acid can arise at any codon which is only possible with multiple nucleotide changes, which is possible but highly unlikely.

Finally, the authors implicitly assume that the mutations do not perturb the structure of the proteins, which is likely to be generally true for essential genes but less likely to be true for non-essential genes. This assumption underpins their entire approach and should be borne in mind when evaluating the results.

eLife Assessment

This valuable study shows that combining reactivation-based training with anodal tDCS yields an unusually broad generalization of visual perceptual learning, while preserving robust learning gains and markedly reducing total training time. Although the empirical evidence is solid, the proposed mechanistic account, i.e., the GABA modulation, disrupted offline consolidation and reduced perceptual overfitting, remains insufficiently substantiated, as these assumptions lack direct neurochemical support, and several alternative behavioral explanations and necessary control comparisons have not been fully addressed. The work will be of broad interest to researchers investigating brain plasticity, perceptual learning, and rehabilitation training.

Reviewer #1 (Public review):

Summary:

This manuscript by Xie and colleagues presents an intriguing behavioral finding for the field of perceptual learning (PL): combining the reactivation-based training paradigm with anodal tDCS induces complete generalization of the learning effect. Notably, this generalization is achieved without compromising the magnitude of learning effects and with an 80% reduction in total training time. The experimental design is well-structured, and the observed complete generalization is robustly replicated across two stimulus dimensions (orientation and motion direction).

However, while the empirical results are methodologically valid and scientifically surprising, the theoretical framework proposed to explain them appears underdeveloped and, in some cases, difficult to reconcile with the existing literature. Several arguments are insufficiently justified. In addition, the introduction of a non-standard metric (NGI: normalized learning gain index) raises concerns about the interpretability and comparability with existing PL literature.

Strengths:

(1) Rigorous experimental design

In this study, Xie and colleagues employed a 2×2 factorial design (Training paradigm: Reactivation vs. Full-Practice × tDCS protocols: Anodal vs. Sham), which allowed clear dissociation of the main and interaction effects.

(2) High statistical credibility

Sample sizes were predetermined using G*Power, non-significant effects were evaluated using the Bayes factor, and the core behavioral findings were replicated in a second stimulus dimension. These strengthen the credibility of the findings.

(3) Strong translational potential

The observed complete generalization could have useful implications for sensory rehabilitation. The large reduction (80%) in total training time is particularly compelling.

Weaknesses:

(1) NGI (Normalized learning gain index) is a non-standard behavioral metric and may distort interpretability.

NGI (pre - post / ((pre + post) / 2)) is rarely used in PL studies to measure learning effects. Almost all PL studies rely on raw thresholds and percent improvements (pre - post / pre), making it difficult to contextualize the current NGI-based results within the broader field. The current manuscript provides no justification for adopting NGI.

A more critical issue is the NGI's nonlinearity: by normalizing to the mean of pre- and post-test thresholds, it disproportionately inflates learning effects for participants with lower post-test thresholds. Notably, the "complete generalization" claims are illustrated mainly with NGI plots. Although the authors also analyze thresholds directly and the results also support the core claim, the interpretation in the text relies heavily on NGI.

The authors may consider rerunning key analyses using the standard percent improvement metric. If retaining NGI, the authors should provide explicit justification for why NGI is superior to standard measures.

(2) The proposed theoretical framework is sometimes unclear and insufficiently supported.

The authors propose the following mechanistic chain:

(a) reactivation-based learning depends on offline consolidation mediated by GABA (page 4 line 73);

(b) online a-tDCS reduces GABA (page 4, line 76), thereby disrupting offline consolidation (page 11, line 225);

(c) disrupted offline consolidation reduces perceptual overfitting (page 4, line 77; page 11, line 225), thereby enabling generalization;

(d) under full-practice training, a-tDCS increases specificity via a different mechanism (page 11 line 235).

While this framework is plausible in broad terms, several components are speculative at best in the absence of neurochemical or neural measurements.

(3) Several reasoning steps require further clarification.

(a) Mechanisms of Reactivation-based Learning.

The manuscript focuses on the neurochemical basis of reactivation-based learning. However, reactivation-induced neurochemical changes differ across brain regions. In the motor cortex, Eisenstein et al. (2023) reported that after reactivation, increased GABA and decreased E/I ratio were associated with offline gains. In contrast, Bang et al. (2018) demonstrated that, in the visual cortex, reactivation decreased GABA and increased E/I ratio. While both studies are consistent with GABA involvement, the direction of GABA modulation differs. The authors should clarify this discrepancy.<br /> More importantly, Bang et al. (2018) demonstrated that reactivation-based (3 blocks) and full-practice (16 blocks) training produced similar time courses of E/I ratio changes in V1: an initial increase followed by a decrease. Given this similarity, the manuscript would benefit from a more thorough discussion of how the two paradigms diverge mechanistically. For example, behaviorally, Song et al. (2021) reported greater generalization with reactivation-based training than with full-practice training, aligning with Kondat et al. (2025). Neurally, Kondat et al. (2024) showed that reactivation-based training increased activity in higher-order brain regions (e.g., IPS), whereas full practice training reduced connectivity between temporal and parietal regions.

(b) tDCS Mechanisms and Protocols.

The effect of a-tDCS on GABA is not consistent across brain regions. While a-tDCS reliably reduces GABA in the motor cortex, recently, a more related work (Abuleli et al., 2025) reports no significant modulation of GABA or Glx in V1, challenging the authors' assumption of tDCS-induced GABA reduction in the visual cortex.

The manuscript proposes that online a-tDCS disrupts offline consolidation is somewhat difficult to interpret conceptually. Online tDCS typically modulates processes occurring during stimulation (e.g., encoding process, attentional state), whereas consolidation occurs afterward. Thus, stating that online tDCS protocols only disrupt offline consolidation without considering the possibility that they first modulate the encoding process is difficult to interpret. Even if tDCS has prolonged effects, the link between online stimulation and disruption of offline consolidation remains unelucidated.

(c) Missing links between GABA modulation and perceptual overfitting.

The proposed chain ("tDCS disrupts consolidation → reduced overfitting → improved generalization") skips a critical step: how GABA modulation translates to changes in neural representational properties (e.g., tuning width, representational overlap between trained/untrained stimuli) that define "perceptual overfitting." The PL literature has not established a link between GABA levels and these representational changes, leaving a key component of the mechanistic explanation underspecified.

(d) Insufficient explanation of the opposite effects.

The manuscript does not fully explain why the same a-tDCS promotes generalization in reactivation-based training but increases specificity in full-practice training. Both paradigms engage offline consolidations, and, as mentioned above, the time courses of E/I ratio changes are similar for 3-block reactivation-based or 16-block training. Thus, if offline consolidation mechanisms (and their associated E/I changes) are comparable across paradigms, it is unclear why identical a-tDCS would produce opposite outcomes in the two paradigms.

Reviewer #2 (Public review):

Xie et al., combined transcranial direct current brain stimulation (tDCS) and a reactivation-based training protocol to investigate the generalization of learning. Using visual perceptual learning as a model, they found that a reactivation-based training protocol, when combined with anodal tDCS over the visual cortex, can induce learning transfer to untrained visual orientations and motion directions. Interestingly, extending reactivation-based training to a full-training protocol with more training trials did not induce generalization of learning. Furthermore, even when paired with tDCS, extending the training protocol did not provide benefits for generalization of learning. This study provides interesting insights into the mechanisms of brain plasticity and how future training protocols could be designed to achieve robust and generalizable learning outcomes.

The authors supported their arguments with a series of well-constructed experiments. The conclusions are largely supported by the data, although some clarifications about their hypotheses and control analyses could strengthen the work:

(1) The authors hypothesize that tDCS can reduce perceptual overfitting through reduced GABA concentrations in the visual cortex, which leads to learning transfer. However, without a clear description of the role of GABA in perceptual learning and perceptual overfitting, it is difficult for the reader to understand why reduced GABA concentrations would contribute to generalization. Do the authors imply that increased GABA can lead to specificity? Are there studies that can support this argument? The authors also did not describe clearly how reactivation-based visual perceptual learning can modify GABA levels in the visual cortex differently (compared to full-practice) during training and during the offline consolidation phase. In order for the reader to better understand their hypotheses and the motivation of the current study, it is beneficial for the authors to provide a concise but clearer description of the roles of GABA in perceptual learning with a focus on the roles of GABA in generalization and during off-line consolidation for different types of training protocols (see for instance Bang et al., 2018; Frangou et al., 2019; Frank et al., 2022; Jia et al., 2024; Shibata et al., 2011; Tamaki et al., 2020; Yamada et al., 2024).

(2) Based on the results, an alternative explanation is that the amount of transfer to the untrained visual feature might be related to the amount of learning for the trained visual feature, which might be different depending on the training protocol and brain stimulation combination. Is it beneficial to compare the amount of learning gains across different training and stimulation protocols to rule out this possibility? Would more learning gains for the trained visual feature predict less transfer for the untrained visual feature? Are there correlations between learning gains and learning transfer?

(3) The authors argued that a reactivation-based training protocol, rather than the amount of training, was critical for the generalization of learning. The control experiment in the study showed that full-practice training combined with tDCS did not lead to transfer, as in reactivation-based training. However, in order to rule out the confounding effects from the amount of training, it is crucial to examine whether a training protocol in which a similar number of trials as in the reactivation-based training but not separated across training sessions would lead to similar generalization of learning.

Reviewer #3 (Public review):

Summary:

This research focuses on a long-lasting and interesting phenomenon in human plasticity. When humans learn basic perceptual skills such as judging the orientation of a simple line, the learned abilities are often limited to the trained condition but not generalizable to untrained conditions. The authors hypothesized that this learning specificity was related to GABA, an inhibitory neurotransmitter in the brain. Using a novel training method that combines reactivation and a brain stimulation method (tDCS) that hypothetically inactivates GABA, the authors hypothesized that learned visual perceptual skills would show greater transfer.

Strengths:

The authors conducted a list of well-conceived behavior studies to demonstrate the effectiveness of their proposed method in enabling learning transfer in two different visual tasks, and carefully conducted comparison studies to elucidate other possible explanations. The sample size was adequate to convey convincing results, and the analyses were thorough.

Weaknesses:

While the authors built their training paradigm on

(1) the hypothetical role GABA plays in inhibiting learning transfer, and

(2) the hypothetical impact tDCS may have on GABA, there was no direct evidence supporting these hypotheses in the current study.

Further, learning specificity takes many formats from features to locations to tasks; it is not yet clear the scope of the observed transfer with the proposed method.

Dit is belangrijk, dat degene margin die groter is, wordt toegepast.

eLife Assessment

This important study establishes the first vertebrate models of DeSanto-Shinawi Syndrome, revealing conserved craniofacial and social and behavioral phenotypes across mouse and zebrafish that mirror key clinical features. The solid evidence is supported by behavioral, anatomical, and molecular analyses of Wac animal mutants that broadly support the authors' claims, though additional mechanistic investigation would strengthen the conclusions. This study sets a baseline for future mechanistic studies and reports a platform to test approaches to reverse phenotypes.

Reviewer #1 (Public review):

Summary:

The authors generated mouse and zebrafish models for DeSanto-Shinawi Syndrome, caused by loss-of-function variants in the WAC gene. Using these vertebrate systems, they demonstrate conserved craniofacial and social-behavioral phenotypes that parallel human clinical features, along with deficits in GABAergic markers. They observe increased seizure susceptibility and male-biased brain volumetric changes in Wac mutant mice. Together, these findings begin to define the biological consequences of Wac haploinsufficiency and provide valuable resources for future mechanistic studies.

Strengths:

WAC is a high-confidence neurodevelopmental disorder gene and one of the genes identified by large-scale exome sequencing efforts, including the Satterstrom et al. (2020) autism spectrum disorder cohort. This study establishes the first vertebrate Wac models, addressing a major gap in the understanding of DeSanto-Shinawi Syndrome, and provides a framework for studying other syndromic forms of autism. The models generated will be impactful and useful to the community to study and understand DeSanto-Shinawi Syndrome.

The cross-species analysis is important and well executed, and reveals both conserved and divergent phenotypes. The behavioral and anatomical assays are rigorously executed and well-controlled, and the inclusion of RNA-sequencing analyses adds valuable insights into the mechanisms underlying brain function in Wac mutants. Notably, the RNA-seq data reveal upregulation of several clustered protocadherins, genes central to neuronal identity and cell-cell interactions, which are known to be regulated by dynamic developmental regulation of chromatin architecture. This observation provides an intriguing hint that could link Wac function to higher-order chromatin organization and neuronal connectivity.

Weaknesses:

The evidence is solid, but the study remains incomplete in its mechanistic depth and molecular interpretation. The authors compellingly describe behavioral, anatomical, and transcriptomic phenotypes associated with WAC loss, yet do not explore how WAC mechanistically regulates chromatin or transcription. Given prior evidence that WAC interacts with the RNF20/40 ubiquitin ligase complex and promotes histone H2B ubiquitination and transcriptional elongation, the paper would benefit from a discussion of these functions as a potential link between Wac haploinsufficiency and the observed changes in neuronal gene expression. Similarly, the authors mention WAC's WW and coiled-coil domains but do not consider how these domains could mediate nuclear interactions or recruitment of transcriptional cofactors that shape gene regulation and chromatin organization in neurons.

The transcriptomic analysis is rich but largely descriptive. Although the upregulation of clustered protocadherins is particularly intriguing, these findings are not validated or localized to specific neuronal populations. The study would be strengthened by independently validating the most significant RNA-seq changes, such as protocadherin gamma genes, using in situ hybridization methods to confirm the spatial and cellular specificity of expression changes.

Finally, while the behavioral and MRI results add valuable breadth, their interpretation would be improved by clearer reporting of sample sizes, statistical corrections, and effect sizes to support claims of sex-specific and regional brain volume differences.

Reviewer #2 (Public review):

The authors describe the first deep neurological characterization of WAC mutation in two vertebrate species (zebrafish and mouse). They examine these at various levels, guided by the work in humans that has associated a heterozygous WAC mutation with DeSantos Shinawi Syndrome (DESSH). Therefore, they investigate the animals for a variety of phenotypes, following a template for what is seen when characterizing a new mouse/fish model of a developmental disability gene. Investigations include analysis of skull and jaw for abnormalities(both species), MRI of brain structure(in mice), electrophysiology(mice), assessment of signaling pathways (by Western blot, in mice), cell counts (both, more in mice), transcriptomics (mice), and behavior (both).

Generally, this describes an important first characterization of the consequences of the mutation. Most of the studies appear well-conducted and reasonably powered, thus solid or convincing. However, there are a few places where the data presentation could be improved for clarity, and a few concerns about some choices in analytical approach for a couple of the experiments, where improved statistical approaches could improve their sensitivity and/or better rule out false positives, and thus the support of some of these claims is currently incomplete. There is also some lack of clarity about the rationale for some decisions regarding the fish genetics. Nonetheless, this is an important and useful first characterization of many phenotypes of these lines. Such experiments form a baseline for future mechanistic studies in the same lines and a platform to test approaches to reverse phenotypes.

Individual claims and their strength & weaknesses:

(1) The authors developed mouse and zebrafish models of WAC deletion

They used the existing KOMP floxed WAC line to generate a null allele. For the mouse, there is a Western showing that it is indeed null for the protein. The fish data is less robustly validated - they don't confirm the allele in null at the protein or RNA level, and fish have two paralogs (waca and wacb), and this paper only characterizes one of these. So this evidence is less clear. The evaluated mice are heterozygous (Het), similar to patients, while the fish appear to be evaluated as homozygous mutants.

(2) The authors show that both species show altered craniofacial features

These data appear well powered, and the findings are robust.

(3) Each model altered GABAergic neurons

In mice, the authors stained with PV antibodies and saw a decrease in cells positive for this staining. A second marker, Lhx6, does not show a difference, suggesting this might be a change in PV expression rather than cell number. They could maybe look into the literature to see if this loss of just the protein also occurs in other models. Overall, the sample size here is a bit smaller than other parts of the paper (n=3), and the methods on the cell counts were less clear, so it is not as clear that this finding is as robust. The authors counted several other broad classes of cells, and those appear normal. Interestingly, there might also be some TBR1 mislocalization in layer 6 that might be significant with added power.

The fish data is based on an in situ hybridization for GAD. The measure shown is the width of the positive area in the forebrain. This measure is not one I have seen much before, and has potential to be driven by something unrelated to GABA (e.g., if the whole forebrain were simply a bit smaller). So this analysis could use a couple of other approaches (density of signal?) and/or a control probe for some other brain gene showing the measure is normal, and thus it is not just a size issue.

(4) Mice were more susceptible to the seizure-inducing agent PTZ

These data appear well powered, and the findings are robust. The authors also did a fair amount of useful electrophysiology that was all normal, but appeared to be well executed.

(5) Mice had changes in brain volume that interact with sex

The authors conducted an MRI on a good number of mice and reported a slight increase in global volume just in males. Sample size is fair, but the statistical approach here may be better if it puts males and females in the same model (to boost power and explicitly test for sex by genotype interaction that they report), and there is some chance that the brain region level differences that they report could include some false positives. They tested many regions, and it is not clear whether or not they corrected for the number of tests. Often, an FDR correction would be used in such imaging studies. It may be that only the most robust regional findings will survive those corrections. It is interesting data either way, but the analysis could be improved.

(6) Several behaviors are altered in the mice as well

These studies were fairly well-powered (n=15,16), and they found several positive and negative results, including alterations in memory and sociability in both species. There is a minor statistical flaw in the three-chamber analysis (they don't actually compare the Hets directly to the wildtypes in their statistical testing - a common mistake in neuroscience that should be addressed. But the data look like they will probably still be significant when correctly analyzed. In the supplement, the authors could do a bit more with the data they have to look at hyperactivity (i.e., show total motion in open field, not just time in center vs. periphery), and adding sex to their model might improve sensitivity for genotype effects.

(7) Some biochemical signaling pathways are altered in the brain

These are n=4 immunoblots, and show altered phospho ERK, but no changes in other signaling events predicted from prior WAC literature like H2B ubiquitination. They appear well done, and the authors share the full blots in the supplement.

(8) WAC deletion also alters gene expression in the brain

These studies were well-powered for RNAseq, with 10 and 14 samples, using neonates (P2), just the forebrain. The sequencing quality metrics all looked good, and the approach to analysis was okay. It would be stronger to again include sex in the model, rather than separate by sex. There were some typos in this part of the paper that made part of the conclusions unclear, but the RNAseq nicely confirmed the mutation of the mice, and discovered many differentially expressed genes, consistent with the role of this gene as a regulator of transcription. The presentation could be expanded to make more use of the data. Overall, though, this is a useful first characterization of the transcriptome in the line.

eLife Assessment

This fundamental study reports solid evidence for early verbal episodic memory formation. The findings demonstrate that speaker identity is a crucial feature, enabling episodic-like memories from birth, and will be of interest to cognitive neuroscientists working on brain development, memory, language learning and social cognition.

Reviewer #1 (Public review):

Summary:

This manuscript investigates whether newborns can use speaker identity to separate verbal memories, aiming to shed light on the earliest mechanisms of language learning and memory formation. The authors employ a well-designed experimental paradigm using functional near-infrared spectroscopy (fNIRS) to measure neural responses in newborns exposed to familiar and novel words, with careful counterbalancing and acoustic controls. Their main finding is that newborns show differential neural activation to novel versus familiar words, particularly when speaker identity changes, suggesting that even at birth, infants can use indexical cues to support memory.

Strengths:

Major strengths of the work include its innovative approach to a longstanding question in developmental science, the use of appropriate and state-of-the-art neuroimaging methods for this age group, and a thoughtful experimental design that attempts to control for order and acoustic confounds. The study addresses a significant gap in our understanding of how infants process and remember speech, and the data are presented transparently, with clear reporting of both significant and non-significant results.

Weaknesses:

However, there are notable weaknesses that limit the strength of the conclusions. The main recognition effect is restricted to a specific subgroup of participants and emerges only during a particular testing window, raising questions about the robustness and generalizability of the findings. The sample size, while typical for infant neuroimaging, is modest, and the statistical power is further reduced by missing data and group-dependent effects. Additionally, the claims regarding episodic memory and evolutionary implications are somewhat overstated, as the paradigm primarily demonstrates memory retention over a few minutes without evidence of the rich, contextually bound recall characteristic of fully developed episodic memory.

Overall, the authors have achieved their primary aim of demonstrating that speaker identity can facilitate memory separation in newborns, providing valuable preliminary evidence for early indexical processing in language learning. The results are intriguing and likely to stimulate further research, but the limitations in effect robustness and theoretical interpretation mean that the findings should be viewed as an important step forward rather than a definitive answer. The methods and data will be of interest to researchers studying infant cognition, memory, and language, and the study highlights both the promise and the challenges of probing complex cognitive processes in the earliest stages of life.

Reviewer #2 (Public review):

Summary:

Previous studies by some of the same authors of the actual manuscript showed that healthy human newborns memorize recently learned nonsense words. They exposed neonates to a familiarization period (several minutes) when multiple repetitions of a bisyllabic word were presented, uttered by the same speaker. Then they exposed neonates to an "interference period" when newborns listened to music or the same speaker uttering a different pseudoword. Finally, neonates were exposed to a test period when infants hear the familiarized word again. Interestingly, when the interference was music, the recognition of the word remained. The word recognition of the word was measured by using the NIRS technique, which estimates the regional brain oxygenation at the scalp level. Specifically, the brain response to the word in the test was reduced, unveiling a familiarity effect, while an increase in regional brain oxygenation corresponds to the detection of a "new word" due to a novelty effect. In previous studies, music does not erase the memory traces for a word (familiarity effect), while a different word uttered by the same speaker does.

The current study aims at exploring whether and how word memory is interfered with by other speech properties, specifically the changes in the speaker, while young children can distinguish speakers by processing the speech. The author's main hypothesis anticipates that new speaker recognition would produce less interference in the familiarized word because somehow neonates "separate" the processing of both words (familiarized uttered by one speaker, and interfering word, uttered by a different speaker), memorizing both words as different auditory events.

From my point of view, this hypothesis is interesting, since the results would contribute to estimating the role of the speaker in word learning and speech processing early in life.

Strengths:

(1) New data from neonates. Exploring neonates' cognitive abilities is a big challenge, and we need more data to enrich the knowledge of the early steps of language acquisition.

(2) The study contributes new data showing the role of speaker (recognition) on word learning (word memory), a quite unexplored factor. The idea that neonates include speakers in speech processing is not new, but its role in word memory has not been evaluated before. The possible interpretation is that neonates integrate the process of the linguistic and communicative aspects of speech at this early age.

(3) The study proposes a quite novel analytic approach. The new mixed models allow exploring the brain response considering an unbalanced design. More than the loss of data, which is frequent in infants' studies, the familiarization, interference and learning processes may take place at different moments of the experiment (e.g. related to changes in behavioural states along the experiment) or expressed in different regions (e.g. related to individual variations in optodes' locations and brain anatomy).

Weaknesses:

I did not find major weaknesses. However, I would like to have more discussion or explanation on the following points.

(1) It would be fine to report the contribution of each infant to the analysis, i.e. how many good blocks, 1 to 5 in sequence 1 and 2, were provided by each infant.

(2) Why did the factor "blocknumber" range from 0 to 4? The authors should explain what block zero means and why not 1 to 5.

(3) I may suggest intending to integrate the changes in brain activity across the 3 phases. That is, whether changes in familiarization relate to changes in the test and interference phases. For instance, in Figure 2, the brain response distinguishes between same and novel words that occurred over IFG and STG in both hemispheres. However, in the right STG there was no initial increase in the brain response, and the response for the same was higher than the one for novels in the 5th block.

(4) Similarly, it is quite amazing that the brain did not increase the activity with respect to the familiarization during the interference phase, mainly over the left hemisphere, even if both the word and speaker changed. Although the discussion considers these findings, an integrated discussion of the detection of novel words and the detection of a novel speaker over time may benefit from a greater integration of the results.

Appraisal:

The authors achieved their aims because the design and analytic approaches showed significant differences. The conclusions are based on these results. Specifically, the hypothesis that neonates would memorize words after interference, when interfered speech is pronounced by a different speaker, was supported by the data in blocks 2 and 5, and the potential mechanisms underlying these findings were discussed, such as separate processing for different speakers, likely related to the recognition of speaker identity.

I think the discussion is well-structured, although I may suggest integrating the changes into the three phases of the study. Maybe comparing with other regions, not related to speech processing.

Evaluating neonates is a challenge. Because physiology is constantly changing. For instance, in 9 minutes, newborns may transit from different behavioral states and experience different physiological needs.

This study offers the opportunity to inspire looking for commonalities and individual differences when investigating early memory capacities of newborns.

eLife Assessment

This study offers a valuable contribution to understanding how working memory (WM) shapes neural processing in extrastriate cortex. By applying spectral decomposition to LFP recordings from primate middle temporal area (MT) during a spatial WM task, the authors show that lower-frequency components (theta, alpha, and beta, but not gamma or high-gamma) correlate with trial-by-trial gain modulation of visually evoked responses. However, certain aspects of the gain-modulation and statistical analyses are incomplete. A clearer and more comprehensive description of these components would substantially strengthen the manuscript.

Reviewer #1 (Public review):

Working memory affects sensory processing. Observers make faster and more accurate perceptual decisions at remembered locations, and corresponding regions of retinotopic visual cortex display enhanced response gain and modulations in oscillatory activity and spike-phase coupling.

Roshanaei et al investigate the relationship between working memory, oscillatory activity, and response gain by reanalyzing extracellular laminar probe recordings from area MT of rhesus monkeys performing a spatial working memory task. During the memory period, visual probes were flashed in the receptive field of the recorded neurons, allowing a comparison of visual responses when memory overlapped with this receptive field (IN) or a location in the opposite hemifield (OUT). They first replicate a range of findings, including increased power in lower frequency bands (theta and alpha/beta) and increased visually-evoked responses in the IN condition. The authors next deployed a spectral technique (MODWT) to decompose the local field potential on single trials into 6 non-arbitrary component frequency bands. This approach allows the authors to observe shifts in peak spectral frequencies across IN and OUT trials. Finally, these single-trial spectral decompositions allowed the authors to relate frequency band power and response gain. This analysis revealed that response gain tended to increase with power in lower (alpha, beta, and theta) frequency bands, and this effect minimally interacted with the remembered location.

Together, these interesting results provide correlational evidence that the effect of working memory on response gain may be mediated by oscillatory power. As the authors note, these results are also consistent with theories positing that lower frequency oscillatory activity primarily reflects working-memory related feedback signals from prefrontal and parietal cortex.

These findings also suggest opportunities for further exploration. From a methodological perspective, it's not clear if the particular spectral decomposition highlighted here is necessary for obtaining these results, or if applying more standard approaches to single trials (as in Lundqvist et al., 2016) would have provided similar sensitivity. Additionally, although the relationship among working memory, oscillatory power, and response gain explored here is necessarily correlational, it could be of interest to subject these factors to a mediation analysis in this or future studies. Finally, the careful analysis of oscillatory phenomena reported here can ideally be used to inform large-scale circuit models and constrain the underlying mechanism.

Reviewer #2 (Public review):

Summary:

Roshanaei et al investigate how working memory (WM) modulates neural activity in the primate visual system by examining local field potentials (LFPs) and spiking activity recorded in area MT. This work is an extension and the reuse of the dataset of the group's prior manuscript, Bahmani et al, Neuron 2018. The animals perform a spatial working memory task where they need to remember the location of a probe stimulus presented within (IN condition) or outside (OUT condition) the neuron's mapped receptive field (RF).

As the first step, the authors replicate the findings in their Neuron 2018 paper by showing:<br /> (1) Significant modulation of the LFP power in αβ band during the working memory period in IN vs OUT conditions. This effect was absent in the gamma band.<br /> (2) A significant increase in phase-coded mutual information for probe location for the IN condition compared to the OUT condition.

The authors then apply the Maximal Overlap Discrete Wavelet Transform (MODWT) to decompose LFP signals at the single-trial level, an approach that allows them to identify oscillatory components without imposing pre-defined frequency bands. They find that the precise frequencies of low-frequency oscillations (theta, alpha, and beta) correlate with the visually evoked firing rates of MT neurons.

Strengths:

The work addresses an important question: how cognitive states such as working memory modulate sensory processing in the visual cortex. More specifically, as we are expanding our understanding of the role of feedback in the brain, a me role of oscillations.

The application of MODWT to single-trial LFPs represents a methodological advance over traditional bandpass filtering, which typically relies on trial-averaged power and may miss fine-grained frequency variability.

The work aligns with ongoing efforts to understand how feedback and oscillatory dynamics contribute to top-down modulation in the brain.

Weaknesses:

(1) Several early results (e.g., increases in alpha/beta power and phase coding) closely replicate previous work from the same group and may be better placed in the Supplementary Information or omitted entirely. The novelty of the current paper lies mainly in the single-trial decomposition and frequency-rate relationship. However, the manuscript fails to expand the prior findings using the traditional methods, or at least offer a more mechanistic insight into the role of top-down modulation of the MT area during working memory tasks. Single-trial analysis can offer new avenues for mechanistic insight. For example, authors could have investigated the relationship of Cross-frequency coupling (CFC) with trial-by-trial behavior of the animal (Voytek et al., 2010) or transient synchronous oscillations for memory maintenance (Buschman et al, 2012).

(2) The statistical methods require greater transparency. Details such as whether tests were one- or two-sided, how multiple comparisons were controlled, and how correlations among nearby electrodes were handled are not fully reported.

Worth mentioning that someone says the GPU definition here is a bit outdated. https://x.com/140ismymax/status/1999170916371300493

do.how

A program that integrates rigorous computing, applied mathematics, and specialized domain knowledge is critical for developing students' complete skills. These programs not only provide a solid technical basis, but also make it easier to apply these talents to real-world problems in specialized industries.

giải thích rõ hơn cho tôi

Even though their political power didn’t last long, the cultural influence of Charlemagne’s family spread far and remained important for centuries.

Both sides got something: the Byzantines protected their pride, and Charlemagne kept his huge territory.

Two brothers first shared the kingdom, but when Carloman died young, Charles became the only ruler.

When the Western Jin fell apart, a related dynasty called the Eastern Jin took over parts of China and ruled from Jiankang.

Believing in heaven or eternal life helped people cope with hardship and suffering in their everyday lives.

Many of Paul’s letters encourage non-jews to join the Christian movement, making Christianity open to everyone, not just people of jewish background.

The Triumvirate created “proscription lists,” which named people they wanted arrested or killed. Around 2,300 people were targeted, including the famous orator Cicero.

The senators who killed Caesar hoped Rome would return to its old system with no single ruler, but things didn’t go the way they planned.

After winning against the Xiongnu, Emperor Wu sent Zhang Qian on another mission with 300 people. They brought valuable gifts like silk, gold, and lacquerware to give to the leaders of the Wusun, another nomadic group.

Emperor Wu sent Zhang qian and a group of 100 people to make contact with the yuezhi, a nomadic group that had been pushed out of their land by the xiongnu

Emperor Wu made Confucianism the official belief system of the government and used it to guide how officials should behave.

As other cultures were spreading around the Mediterranean sea, rome located in central Italy was slowly becoming more important.

Zeno believed the whole universe was like a god, and that people had a small piece of that divine reason inside them.

The Ramayana started as an old story passed down orally, and over time it grew into a long poem in the Sanskrit language.

Thuydides is famous because he tried to study history in a careful, logical way almost like a scientist.

The author says that details about the Hebrewd religion will come later, this section is mainly about historical events

Most people today know baal because the Hebrew Bible speaks negatively about him. The writers saw him as a false or rival god.

This means about 1,000 years passed between the Bronze Age Collapse and the beginning of year 1 CE.

I’ve learned how much students light up when I point out something they did well, even if it seems small. When I make an effort to notice the good choices such as when a student stays focused, helps a friend, or gives their best effort the entire class environment changes. It reminds me that praise builds trust, confidence, and a stronger classroom environment.

I’ve found that a quiet reminder can go a long way. It keeps the student from feeling called out in front of their peers, and it lets them know I’m on their side. When that isn’t enough, creating a simple behavior plan will build more trust with the student. It shows them that I’m trying to support them and not punish them. This usually leads to better behavior and a more positive relationship.

In my own observations, I’ve seen how much smoother class runs when students help make the rules and know exactly what will happen if they break them. When the classroom guidelines are consistent and fair, the students trust the process more. The students know I’m not out to get them and I’m simply keeping the classroom safe and respectful so we can make music together.

I’ve learned in my own teaching that having a student sit closer to me can make a huge difference. Sometimes they just need that extra sense of security or a quicker reminder to stay on track. I’ve also realized how important it is to speak up when the class size makes it hard to meet everyone’s needs. Asking for extra support is a way to make sure every student, especially those who need more guidance, can be successful in my music class.

Even though students can have trouble in the classroom, I’ve learned that when I meet students where they are and show that I truly see them, they give that trust right back. This shows progress from Carson and the small agreements we make together feel big for him personally and watching him succeed on his own terms has taught me what it means to have patience, flexibility, and how you can support students.

Around the early 300s, Rome was very unstable. Different leaders were fighting each other for power, causing the empire to break apart politically.

This made me think about how in my own household, which I manage mostly on my own, the things I rely on, such as housing and education systems and how they exist because of the colonial systems, even though they may appear to me as just "regular" life.

The author argued that daily lives are where decolonization begins.

This presents the main idea, which is that colonialism is not a historical act, it is ongoing, happens in daily life and progresses even in this day. Even just waking up is connected to colonization.

Standpoint theory?

These 2 paragraphs introduced and explained their background and identities, one is indigenous queer and one white queer, this helped audiences to understand the influence the author's background has in the way they are shaped today.

gem

to IPFS Forum Kubo v-.28 annuncemnt

caching peers

to IPNS Record Protocol

converging supporting personal firs

interpersonal networks

Provider records may be sent to peers that have gone offline or are no longer closest to the CID, since the snapshot is only refreshed hourly

or more, you should increase the number of provide workers - to handle the higher throughput.

Reprovide Sweep

Default

tradeoff

continuous crawling vs zero-latency lookup

serves from map lookup no network request

100k CIDs within a 3-hour window single thread

this burst approach creates temporary peaks in resource usage—high CPU load, network bandwidth consumption, and numerous concurrent connections

it pops all CIDs from the same region as the head of the queue and provides them together in a single batch

Wow caps at 3k lookups

independently of the number of CIDs being provided

since every content is contextualized =

can just lookup adjacent peers by shift of focus

keyspace prefix

20 peers

replication factor

into a smooth, continuous background process that distributes the load over time.

Reprovide Sweep

lots of records to the same server

if you have far more CIDs to provide than there are DHT servers in the network, then each server must be allocated multiple CIDs. Since each CID must be stored on 20 servers (the replication factor), and there are approximately 10,000 active DHT servers, a node providing hundreds of thousands of CIDs will send dozens or even hundreds of provider records to the same servers.

CIDs that are close together

inpractical for smaller scale IPFS users

expensive

consider timebound focus

peer ids of near contacts two hops!

lookup

complete map of the network

Interpersonal connectivity may help to scale this

scaling salience

conrifmed

new content arrive faster

can't keep up

bottleneck lookup

indeed key

to validity 48 hours

I'm curious about a few things: - who specificially comes to mind as exemplars of high development in this phase space? - feels attractive to me to train up our volumes in this phase space in different stages; from games to practices touching the basics all the way to finding identity and core lifeworld framing / ways of being grounded in mixtures of these; know anyone else doing this?

Hit me up at yuli.jadov@gmail.com if you have some thoughts - I'd love to hear from others. I imagine I will live & run training at various levels over time.

Crenshaw's humility notwithstanding, I'm frequently caught off guard by frequent interactions with people whose conversational posture reveals that they have a similar conception of software like, say, Windows—their comments a manifestation of a seemingly total unwillingness to confront the fact that, no, some observable behavior in software isn't just how computers work, but that someone—a human programmer—sat down and decided to make it work that way—that it isn't just some natural property of computers that someone has coaxed out of one, and that other instances of software creation are not mere parlor tricks. They're procedures. They have to be conceived of and then worked out and (ideally) made airtight against a whole range of conditions.

Try to keep it short and to the point so it isn't drawn out and wordy.

Keep an open mind

Take some time to think it over before crafting the response

Reflection is important in all aspects of life as it gives you time to think over choices and abilities and allows you to correct mistakes and build on successes

Reflection is a great way, personally, to help the information that I've learned really stick.

María Corina Machado revealed in Oslo that she escaped Venezuela with undisclosed assistance from the U.S. government after more than a year in hiding, a detail she kept vague to protect collaborators as she arrived to celebrate her Nobel Peace Prize.

How to Be a Stoic: Using Ancient Philosophy to Live a Modern Life by Massimo Pigliucci | Goodreads

That's a powerful scene

There is a thesis in smut author Ayn Rand's novel, "Atlas Shrugged" that goes something like,

The Woman a Man Marries is a reflection of what he truly seems himself as and values.

Look up the character, James_Taggart

everything is mate suppression - YouTube

I feel like I know the type

I read this article the other day I wonder if Aria Grande became sickly skinny because she felt like she was getting undressed by every mans eyes in every room she walked into

What about the self esteem of fat women, the modern man has no capacity to empathize with that. You can point it out to them and they will acknowledge that but they can be fucking brutal.

Can we acknowledge the cast system when it comes to dating,

Men can lack enough theory of mind to say to their romantic partner that "Women hit the wall at 30". Shake my head, how does a guy like that get a date?

Someone needs to generate an infographic

The thesis that women select for a good fuck tonight or a good provider for her and her children in the future,

Strategic pluralism - Wikipedia

I wonder if there is any subculture associated with Woman that men participate in, the Woke Mind Virus might be one.

By showing women and men differently from older crime shows did, CSI tells viewers that authority isn't only about brute strength of traditional gender roles. It shows us that intelligence and competence matter more than old stereotypes.

Crime dramas definitely shape the way we think about justice and authority. By constantly portraying police officers as the heroes and lawyers as the people who get in the way, these shows create a very simplified version of the legal system. They make it seem like the police are always right and always moral, while lawyers are the ones causing problems. This is a clear example of how the media doesn't just show reality, it also influences the way that we understand it and what we believe about the world around us. The depiction can affect how viewers see real-life institutions, making them more likely to trust police without questioning them and less likely to understand the important role lawyers play in protecting rights.

Version 3 of this preprint has been peer-reviewed and recommended by Peer Community in Evolutionary Biology.<br /> See the peer reviews and the recommendation.

Fascinating screen showing how ineffective many common actin inhibitors can be in a non-traditional system, and how essential it is to have a diverse toolkit of compounds with different chemistries, binding affinities, and environmental stabilities. It really highlights that inhibitors developed for eukaryotic actin behave very differently in Lokiactin-based cells, and that relying on a single ‘standard’ drug would have missed most of the biology.

Related to the poly-L-lysine experiments in Fig. S2, did you ever see protrusions actually breaking off or being left behind on the surface at higher poly-L-lysine concentrations, or was the main effect just a slowdown in protrusion dynamics and cell migration? And did stronger adhesion noticeably change protrusion length distributions—for example, by increasing retraction/catastrophe events or biasing protrusions toward shorter lengths?

Does the "Round" phenotype include Bulbs and Complex Arms? or just Bulbs?

Carthaginians followed Phoenician religion. Their main gods were Tanit and Baal, but early on they didn’t build big temples like other civilizations did.

Heavy exposure to media causes people to develop a distorted view of reality. Watching a lot of TV shows, news, or scrolling through social media can make the world feel much more dangerous than it actually is. Constant reports of crime, disasters, or dramatic events can create a false sense of threat, even when our daily lives might be generally safe. Media is powerful in shaping our perception of reality. Therefore, it's important to step back and think critically to separate what we are seeing on screen from what's actually happening in the world around us.

This is an exceptionally cool and ambitious piece of work. It’s impressive to see such a deep level of structural and dynamic characterization developed from an organism that was essentially uncharacterized and self-isolated by the authors. The combination of live imaging, high-resolution cytoskeletal staining, and careful morphometric analysis provides a remarkably detailed view of Filoreta’s network architecture across different states. This represents a substantial contribution both technically and conceptually, and it’s exciting to see such a comprehensive foundation being built for a new model system. Amazing work!

Incredibly cool analysis of this amazing phenotype!

Do you see evidence of these actin-dense patches staying at the base of newly formed branchlets? I was curious whether you have observations that help distinguish patches functioning as nucleation sites for new protrusions versus representing increased endocytic activity in the nutrient-enriched conditions. Also, do you think these puncta might correspond to Arp2/3-mediated branched actin networks, or is that still unclear?

I was curious whether you ever see a change in feeding behavior when the medium is enriched with YET. Since the network responds so quickly to added nutrients, do think cells would reduce their phagocytosis of bacteria under these conditions, or do you think feeding branchlets would stay just as active?

Just curious if you ever see the "loops" close overtime or if "feeding" branchlets might extend into the loop incase a food source is captured within, or if you believe these are fully functioning as support systems for the reticulated network?

What outdated notions am I carrying?

What's an example of retrieval-augmented generation?

The format of a movie is always appealing to me. When Sinners came out, the 13 IMAX screen was a huge deal for everyone to see for a better experience, as that is what it was filmed for. Those that seen it in standard theatre’s, hurried to rewatch it and enjoyed it in IMAX greatly. I have never seen the difference between the standard and other forms, however after Sinners, it put that into a different perspective for me.

Movies that are made now have long credits that people do not sit for to read, however in older films, it would generally consist of a small preview of those who worked on the film. This reminds me of how Marvel would include a sneak peak of a film at the end of their movies, showcasing the credits. I think that that is good because it truly honors everyone’s hard work and you can see most names which is good.

It has always been very interesting to me that today’s films can not only cost much money, but also accumulate billions of dollars as well. The more that I read about the technology used and also the amount actors make, it continues to make sense. I also understand that makeup and hair artists, and so many other jobs are required to be on set and work on films which increases the amount tremendously.

It's rather more a testament to the durability of W3C standards generally—something that is now under attack by WHATWG member orgs.

Should we add the Accessibility Hub to the top of this list? PSU's Accessible Teaching Resource Hub

This entire section is plagiarized from Eliason. Bizarre.

institution

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

Reply to the reviewers

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Reply to the Reviewers

I thank the Referees for their...

Referee #1

1. The authors should provide more information when...

Responses + The typical domed appearance of a hydrocephalus-harboring skull is apparent as early as P4, as shown in a new side-by-side comparison of pups at that age (Fig. 1A). + Though this is not stated in the MS 2. Figure 6: Why has only...

Response: We expanded the comparison

Minor comments:

1. The text contains several...

Response: We added...

Referee #2

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Referee #3

Evidence, reproducibility and clarity

This is a good manuscript, well performed and well presented. I have several suggestions/questions to enhance the clarity of the concept, as technically the work is rather well performed.

1. I suggest that the authors explain better the mesenchymal-to-epithelial (MET) transition in reprogramming. Perhaps, explaining that epithelial gene acquisition (e.g., CDH1) and epidermal cell fate are not exactly the same. This approach could also be used to divide the genes they study further in their analyses.
2. KLF4 is both a repressor and an activator in different cell contexts including reprogramming. Does HIC2 act only as repressor? Is it possible that HIC2 is repressing KLF4-activated genes bad for reprogramming (including epidermal genes) and activating KLF4-suppressed genes ncessary for reprogramming? This should not be too difficult to explore with their current dataset and they also could look at available datasets for histone modifications in reprogramming.
3. Does HIC2 bind to genes related to somatic cell identify that need to be suppressed in reprogramming before the MET phase takes place?
4. Does HIC2 influence proliferation during reprogramming?

Referee cross-commenting

Comments by the other reviewers are sound and will help improve the manuscript.

Significance

In this manuscript, Kaji and colleagues perform a CRISPR/Cas9 screen to identify genes involved in mouse somatic cell reprogramming, identifying HIC2 as a target that they further validate. They conclude that HIC2 acts by repressing the epidermal/epithelial program induced by KLF4 during reprogramming. Studying the complex role of transcription factor interactions in the context of cell fate conversions (of any kind and not just somatic cell reprogramming) is highly relevant. This work helps clarify such complexity in a specific context but the work has wider conceptual implications.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Referee #2

Evidence, reproducibility and clarity

The study by Beniazza et al. aims to address the inefficiencies associated with OSKM-mediated reprogramming. Through a genome-wide CRISPR/Cas9 knockout screen, the authors identified 14 genes essential for iPSC reprogramming but dispensable for ESC self-renewal. Among these, HIC2 significantly enhanced reprogramming efficiency, yielding approximately a tenfold increase compared to standard conditions. scRNA-seq analyses revealed that HIC2-overexpressing cells follow a more direct trajectory toward pluripotency, bypassing the KLF4-dependent activation of keratinocyte and epidermal gene programs. ChIP-seq profiling further demonstrated that HIC2 and KLF4 co-occupy approximately 60% of their genomic targets, indicating substantial regulatory overlap. Notably, this co-binding and its functional effects are dose-dependent on KLF4, as shown by experiments comparing high KLF4 expression systems (standard OSKM and STEMCCA+9 constructs) with low KLF4 conditions (STEMCCA cassette lacking additional KLF4). The authors conclude that HIC2's modulatory effect occurs specifically under high KLF4 levels.

Major Comments

Figure 1D: What is the efficiency of gRNA library transduction into MEFs? What percentage of MEF cells were successfully knocked out? Figures 2B/C: To rule out the possibility that the observed variability in reprogramming efficiency among the tested factor combinations stems from differences in MKOS expression levels, the authors should provide evidence showing that the expression levels of all MKOS factors are comparable across samples. Figures 2D/E: To rule out a fibroblast-specific effect, can the authors show whether the epidermal gene signature is also upregulated during NSC reprogramming and whether Hic2 overexpression suppresses this signature? Figure 2H: Are the 13 signature genes that distinguish MKOS-Hic2-iPSCs from MKOS-iPSCs consistently identified across independent Hic2-iPSC lines, or does each reprogramming event produce a distinct gene set? If the signature is consistent, this is an important observation and should be further addressed and discussed. Figure 3K: Can the authors show the expression levels of MKOS and Hic2 transgenes in all samples? The same concern applies to Figure 4I. The reviewer wishes to be confident that the reduction in epidermal gene expression observed in MEFs is not due to variable transgene expression caused by multiple vector introductions (e.g., KLF4 alone versus KLF4 + Hic2), which could potentially lead to lower KLF4 expression through co-transfection competition. Does KLF4 overexpression in Hic2-knockdown MEFs lead to greater upregulation of the epidermal gene signature compared to the wild-type control? Figure 4C: It appears that only about half of the Hic2 binding sites overlap with KLF4 sites. What are the characteristics of the other Hic2-specific sites, and how might they contribute to reprogramming, if at all? Can the authors perform a reprogramming experiment using a combination that lacks KLF4 (e.g., replacing KLF4 with Esrrb or BMP4, as shown in PMID: 19136965 and PMID: 21135873) and test the effect of Hic2 under these conditions? Do KLF4 and HIC2 physically interact? The authors should perform a co-immunoprecipitation assay to address this question. What is the effect of Hic2 during human reprogramming? Does it play a similar regulatory role?

Minor Comments

• Typographical errors should be checked and avoided; for example, on page 10, the word 'colonies' was misspelled.
• Some blank squares appear in the Methods section; please correct these formatting errors.

Referee cross-commenting

All suggestions are feasible within a relatively short time frame and will improve the manuscript.

Significance

Overall, this study is of significant interest to the stem cell community and presents a well-designed and carefully executed experimental framework. However, several concerns remain that should be addressed prior to publication.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Referee #1

Evidence, reproducibility and clarity

Summary: This work identified 14 genes essential for iPSC reprogramming but not essential for ESC maintenance and MEF proliferation by analyzing three CRISPR/Cas9-mediated genome-wide KO screens. Among them, they found that overexpression of the Hic2 gene can greatly promote OSKM-driven reprogramming. By using scRNA-seq in time points of the reprogramming process, they found that Hic2 can bypass the epidermal gene expressing state during reprogramming. Then, using ChIP-seq, they found that HIC2 and KLF4 have common binding sites on epidermal genes. Finally, by expressing KLF4 alone or KLF4 and HIC2 together, they demonstrated that HIC2 can inhibit KLF4-driven epidermal gene expression.

Major comments: The claims and conclusions are well-supported by the data and do not require additional experiments or analysis. The data and methods are presented in a reproducible way.

Minor comments: There seem to be some typos. For example, "we selected 30 genes with low FDRs in ESC maintenance" may be "high depletion FDR," since you want nonessential genes for ESC maintenance. The "log10(-FDR)" may be "-log10(FDR)." Some figures lack P values. Perhaps it would be useful to analyze whether Hic2 reduces reprogramming heterogeneity. Validation experiments, such as trilineage differentiation, could be considered to demonstrate that Hic2 does not affect the pluripotency and differentiation capacity of iPSCs.

Significance

General Assessment: This work is based on three CRISPR/Cas9-mediated genome-wide KO screens, which makes it comprehensive and reliable. They discovered that HIC2 and OSKM can drive reprogramming without an epidermal gene expression intermediate. They also found extensive common binding sites of HIC2 and KLF4 at target genes. This work not only enables more efficient reprogramming but also expands our understanding of the reprogramming process. Among HIC2 and KLF4 common target genes, some are repressed while others are activated, and it will be very interesting to study the mechanism of this selective function.

Advance: Compared to natural `embryonic development, OSKM-driven reprogramming is very inefficient, and our understanding of the mechanisms of efficient reprogramming remains poor. The specific role of the epidermal gene expression state in the reprogramming process remains unclear. This work strongly supports the idea that repression of the epidermal gene expression state can promote iPSC generation. Moreover, previous studies on Hic2 are limited, and this work enriches our understanding of its mechanisms and functions.

Audience: This study may be of interest to those interested in basic research on reprogramming mechanisms or Hic2, as well as those developing efficient reprogramming technologies.

My field: Reprogramming, stem cells, aging, transcription factors.

A question that you as well as other people want to answer. Make sure to include evidence that is reliable as well as accessible.

This one was phenomenal. Generally on the nose, I have to squint a bit to get the fruity notes, but I really was getting the mango/papaya!

eLife Assessment

This fundamental work substantially advances our understanding of episodic memory by proposing a biologically plausible mechanism through which hippocampal barcode activity enables efficient memory binding and flexible recall. The evidence supporting the conclusions is convincing, with rigorously validated computational models and alignment with experimental findings. The work will be of broad interest to neuroscientists and computational modelers studying memory and hippocampal function.

Reviewer #1 (Public review):

Summary:

In this paper, the authors develop a biologically plausible recurrent neural network model to explain how the hippocampus generates and uses barcode-like activity to support episodic memory. They address key questions raised by recent experimental findings: how barcodes are generated, how they interact with memory content (such as place and seed-related activity), and how the hippocampus balances memory specificity with flexible recall. The authors demonstrate that chaotic dynamics in a recurrent neural network can produce barcodes that reduce memory interference, complement place tuning, and enable context-dependent memory retrieval, while aligning their model with observed hippocampal activity during caching and retrieval in chickadees.

Strengths:

(1) The manuscript is well-written and structured.

(2) The paper provides a detailed and biologically plausible mechanism for generating and utilizing barcode activity through chaotic dynamics in a recurrent neural network. This mechanism effectively explains how barcodes reduce memory interference, complement place tuning, and enable flexible, context-dependent recall.

(3) The authors successfully reproduce key experimental findings on hippocampal barcode activity from chickadee studies, including the distinct correlations observed during caching, retrieval, and visits.

(4) Overall, the study addresses a somewhat puzzling question about how memory indices and content signals coexist and interact in the same hippocampal population. By proposing a unified model, it provides significant conceptual clarity.

Weaknesses:

The recurrent neural network model incorporates assumptions and mechanisms, such as the modulation of recurrent input strength, whose biological underpinnings remain unclear. The authors acknowledge some of these limitations thoughtfully, offering plausible mechanisms and discussing their implications in depth. It may be worth exploring the robustness of the results to certain modeling assumptions. For instance, the choice to run the network for a fixed amount of time and then use the activity at the end for plasticity could be relaxed.

Reviewer #2 (Public review):

Summary:

Striking experimental results by Chettih et al 2024 have identified high-dimensional, sparse patterns of activity in the chickadee hippocampus when birds store or retrieve food at a given site. These barcode-like patterns were interpreted as "indexes" allowing the birds to retrieve from memory the locations of stored food.

The present manuscript proposes a recurrent network model that generates such barcode activity and uses it to form attractor-like memories that bind information about location and food. The manuscript then examines the computational role of barcode activity in the model by simulating two behavioral tasks, and by comparing the model with an alternate model in which barcode activity is ablated.

Strengths of the study:

proposes a potential neural implementation for the indexing theory of episodic memory\

Provides a mechanistic model of striking experimental findings: barcode-like, sparse patterns of activity when birds store a grain at a specific location

A particularly interesting aspect of the model is that it proposes a mechanism for binding discrete events to a continuous spatial map, and demonstrates the computational advantages of this mechanism

Weaknesses:

The importance of different modeling ingredients and dynamical mechanisms could be made more clear.

Author response:

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

Reviewer #1 (Public review):

Summary:

In this paper, the authors develop a biologically plausible recurrent neural network model to explain how the hippocampus generates and uses barcode-like activity to support episodic memory. They address key questions raised by recent experimental findings: how barcodes are generated, how they interact with memory content (such as place and seed-related activity), and how the hippocampus balances memory specificity with flexible recall. The authors demonstrate that chaotic dynamics in a recurrent neural network can produce barcodes that reduce memory interference, complement place tuning, and enable context-dependent memory retrieval, while aligning their model with observed hippocampal activity during caching and retrieval in chickadees.

Strengths:

(1) The manuscript is well-written and structured.

(2) The paper provides a detailed and biologically plausible mechanism for generating and utilizing barcode activity through chaotic dynamics in a recurrent neural network. This mechanism effectively explains how barcodes reduce memory interference, complement place tuning, and enable flexible, context-dependent recall.

(3) The authors successfully reproduce key experimental findings on hippocampal barcode activity from chickadee studies, including the distinct correlations observed during caching, retrieval, and visits.

(4) Overall, the study addresses a somewhat puzzling question about how memory indices and content signals coexist and interact in the same hippocampal population. By proposing a unified model, it provides significant conceptual clarity.

Weaknesses:

The recurrent neural network model incorporates assumptions and mechanisms, such as the modulation of recurrent input strength, whose biological underpinnings remain unclear. The authors acknowledge some of these limitations thoughtfully, offering plausible mechanisms and discussing their implications in depth.

One thread of questions that authors may want to further explore is related to the chaotic nature of activity that generates barcodes when recurrence is strong. Chaos inherently implies sensitivity to initial conditions and noise, which raises questions about its reliability as a mechanism for producing robust and repeatable barcode signals. How sensitive are the results to noise in both the dynamics and the input signals? Does this sensitivity affect the stability of the generated barcodes and place fields, potentially disrupting their functional roles? Moreover, does the implemented plasticity mitigate some of this chaos, or might it amplify it under certain conditions? Clarifying these aspects could strengthen the argument for the robustness of the proposed mechanism.

In our model, chaos is used to produce a random barcode when forming memories, but memory retrieval depends on attractor dynamics. Specifically, the plasticity update at the end of the cache creates an attractor state, and then afterwards for successful memory retrieval the network activity must settle into this attractor rather than remaining chaotic. This attractor state is a conjunction of memory content (place and seed activity) and memory index (barcode activity). Thus a barcode is ‘reactivated’ when network dynamics during retrieval settle into this cache attractor, or in other words chaotic dynamics do not need to generate the same barcode twice.

The reviewer raises an important point, which is how sensitivity to initial conditions and noise would affect the reliability of our proposed mechanism. The key question here is how noise will affect the network’s dynamics during retrieval. Would adding noise to the dynamics make memory retrieval more difficult? We thank the reviewer for suggesting we investigate this further, and below describe our experiments and changes to the manuscript to better address this topic.

We first experimented with adding independent gaussian distributed noise into each unit, drawn independently at each timestep. We analyzed recall accuracy using the same task and methods as Fig. 4F while varying the magnitude of noise. Memory recall was quite robust to this form of noise, even as the magnitude of noise approached half of the signal amplitude. This first experiment added noise into the temporal dynamics of the network. We subsequently examined adding static noise into the network inputs, which can also be thought of as introducing noise into initial conditions. Specifically, we added independent gaussian distributed noise into each unit, with the random value held constant for the extent of temporal dynamics. This perturbation decreased the likelihood of memory recall in a graded manner with noise magnitude, without dramatically changing the spatial profile. Examination of dynamics on individual trials revealed that the network failed to converge onto a cache attractor on some random fraction of trials, with other trials appearing nearly identical to noiseless results. We now include these results in the text and as a new supplementary figure, Figure S4AB.

To clarify the network dynamics and the purpose of chaos in our model, we make the following modifications in text:

Section 2.3, paragraph 2 (starting at “To store memories…”):

“…place inputs arrive into the RNN, recurrent dynamics generate an essentially random barcode, seed inputs are activated, and then Hebbian learning binds a particular pattern of barcode activity to place- and seed-related activity.”

Section 2.3, paragraph 3 (starting at “Memory recall in our network…”): As an example, consider a scenario in which an animal has already formed a memory at some location l, resulting in the storage of an attractor \vec{a} into the RNN. The attractor \vec{a} can be thought of as a linear combination of place input-driven activity $p(l)$, seed input-driven activity $s$, and a recurrent-driven barcode component $b$. Later, the animal returns to the same location and attempts recall (i.e. sets r \= 1, Figure 3B). Place inputs for location l drive RNN activity towards $p(l)$, which is partially correlated with attractor \vec{a}, and the recurrent dynamics cause network activity to converge onto attractor \vec{a}. In this way, barcode activity $b$ is reactivated, along with the place and seed components stored in the attractor state, $p(l)$ and $s$. The seed input can also affect recall, as discussed in the following section.

Section 2.4, final paragraph (starting “We further examined how model hyperparameters affected performance on these tasks”), added the following describing new results on adding noise: We found that adding noise to the network's temporal dynamics had little effect on memory recall performance (Figure S4A). However, large static noise vectors added to the network's input and initial state decreased the overall probability of memory recall, but not its spatial profile (Figure S4B).

It may also be worth exploring the robustness of the results to certain modeling assumptions.  For instance, the choice to run the network for a fixed amount of time and then use the activity  at the end for plasticity could be relaxed.

As described above, chaotic dynamics are necessary to generate a barcode during a cache, but not to reactivate that barcode during retrieval. During a successful memory retrieval, network activity settles into an attractor state and thus does not depend on the duration of simulated dynamics. The choice of duration to run dynamics during caching is important, but only insofar as activity significantly decorrelates from the initial state. We show in Figure S1B that decorrelation saturates ~t=25, and thus any random time point t > 25 would be similarly effective. We used a fixed duration runtime for caches only to avoid introducing unnecessary complication into our model.

Reviewer #2 (Public review):

Summary:

Striking experimental results by Chettih et al 2024 have identified high-dimensional, sparse patterns of activity in the chickadee hippocampus when birds store or retrieve food at a given site. These barcode-like patterns were interpreted as "indexes" allowing the birds to retrieve from memory the locations of stored food.

The present manuscript proposes a recurrent network model that generates such barcode activity and uses it to form attractor-like memories that bind information about location and food. The manuscript then examines the computational role of barcode activity in the model by simulating two behavioral tasks, and by comparing the model with an alternate model in which barcode activity is ablated.

Strengths of the study:

Proposes a potential neural implementation for the indexing theory of episodic memory - Provides a mechanistic model of striking experimental findings: barcode-like, sparse patterns of activity when birds store a grain at a specific location

A particularly interesting aspect of the model is that it proposes a mechanism for binding discrete events to a continuous spatial map, and demonstrates the computational advantages of this mechanism.

Weaknesses:

The relation between the model and experimentally recorded activity needs some clarification

The relation with indexing theory could be made more clear

The importance of different modeling ingredients and dynamical mechanisms could be made more clear

The paper would be strengthened by focusing on the most essential aspects

Comments:

The model distinguishes between "barcode activity" and "attractors". Which of the two corresponds to experimentally-recorded barcodes? I would presume the attractors. A potential issue is that the attractors are, as explained in the text (l.137), conjunctions of place activity, barcode activity and "seed" inputs. The fact that the seed activity is shared across attractors seems to imply that they have a non-zero correlation independent of distance. Is that the case in the model? If I understand correctly, Fig 3D shows correlations between an attractor and barcodes at different locations, but correlations between attractors at different locations are not shown. Fig 1 F instead shows that correlations between recorded retrieval activities decay to zero with distance.

More generally, the fact that the expression "barcode" is apparently used with different meanings in the model and in the experiments is potentially confusing (in the model they correspond to activity generating during caching, and this activity is distinct from the memories; my understanding is that in the experiments barcodes correspond to both caching and retrieval, but perhaps I am mistaken?).

Our intent is to use the expression “barcode” as similarly as possible between model and experimental work. The reviewer points out that the connection between barcodes in experimental and modeling work is unclear, as well as the relation of “attractors” in our model to previous experimental results. The meaning of ‘barcode’ is absolutely critical—we clarify below our intended meaning, and then describe changes to the manuscript to highlight this.

In experiments, we observed that activity during caching looked different than ordinary hippocampal activity (i.e. typical “place activity” observed during visits). Empirically there were two major differences. First, there was a pattern of neural activity which was present during every cache . This pattern was also present when birds visually inspected sites containing a cached seed, but not when visually inspecting an empty site. This is what we refer to as “seed activity”. Second, there was a pattern of neural activity which was unique to each cache. This pattern re-occurred during retrieval, and was orthogonal to place activity (see Fig. 1E-F). This is what we refer to as “barcode activity”. In summary, activity during a cache (or retrieval) contains a combination of three components: place activity, seed activity, and barcode activity.

These experimental findings are recapitulated in our model, as activity during a cache contains a combination of three components: place activity driven by place inputs, seed activity driven by seed inputs, and barcode activity generated by recurrent dynamics. Cache activity in the model corresponds to cache activity in experiments, and barcodes in the model correspond to barcodes in experiments. Our model additionally has “attractors”, meaning that network connectivity changes so that the activity generated during a simulated cache becomes an attractor state of network dynamics. “Attractors” refers to a feature of network dynamics, not a distinct activity state, and we do not yet know if these attractors exist in experimental data.

Figure 3D, as described in the figure legend, is a correlation of activity during cache and retrieval (in purple), for cache-retrieval pairs at the same or at different sites. We believe this is what the reviewer asks to see: the correlation between attractor states for different cache locations. The reviewer makes an important point: seed activity is shared across all attractors, so then why are correlations not high for all locations? This is because attractors also have a place component, which is anti-correlated for distant locations. This is evident in Fig. 3D by noticing that visit-visit correlations (black line, corresponding to place activity only) are negative for distant locations, and the correlation between attractors (purple line, cache-retrieval pairs) is subtly shifted up relative to the black line (place code only) for these distant locations. The size of this shift is due to the relative magnitude of place and seed inputs. For example, if we increase the strength of the seed input during caching (blue line), we can further increase the correlation between attractors even for quite distant sites:

Author response image 1.

To clarify the manuscript, we made the following modifications:

Section 2.2, first paragraph: We model the hippocampus as a recurrent neural network (RNN) (Alvarez and Squire, 1994; Tsodyks, 1999; Hopfield, 1982) and propose that recurrent dynamics can generate barcodes from place inputs. As in experiments, the model’s population activity during a cache should exhibit both place and barcode activity components.

Section 2.3, paragraph 3 (starting at “Memory recall in our network…”): As an example, consider a scenario in which an animal has already formed a memory at some location l , resulting in the storage of an attractor \vec{a} into the RNN . The attractor \vec{a} can be thought of as a linear combination of place input-driven activity $p(l)$, seed input-driven activity $s$, and a recurrent-driven barcode component $b$. Later, the animal returns to the same location and attempts recall (i.e. sets r \= 1, Figure 3B). Place inputs for l drive RNN activity towards $p(l)$, which is partially correlated with attractor \vec{a}, and the recurrent dynamics cause network activity to converge onto attractor \vec{a}. In this way, barcode activity $b$ is reactivated as part of attractor \vec{a}, along with the place and seed components stored in the attractor state, $p(l)$ and $s$. The seed input can also affect recall, as discussed in the following section.

The insights obtained from the network model for the computational role of barcode activity could be explained more clearly. The introduction starts by laying out the indexing theory, which proposes that the hippocampus links an index with each memory so that the memory is reactivated when the index is presented. The experimental paper suggests that the barcode activations play the role of indexes. Yet, in the model reactivations of memories are driven not by presenting bar-code activity, but by presenting place activity (Cache Presence task) or seed activity (Cache Location task). So it seems that either place activity and seed activity play the role of indexes. Section 2.5 nicely shows that ultimately the role of barcode activity is to decorrelate attractors, which seems different from playing the role of indexes. I feel it would be useful that the Discussion reassess more critically the relationship between barcodes, indexing theory, and key-value architectures.

The reviewer highlights a failure on our part to clearly identify the connection between our findings on barcodes, indexing theory, and key-value architectures. This is another major component of the paper, and below we propose changes to the manuscript to clarify these concepts and their relationships. First, we will summarize the key points that were unclear in our original manuscript.

The reviewer equates the concept of an ‘index’ with that of a ‘query’: the signal that drives memory reactivation. This may be intuitive, but it is not how a memory index was defined in indexing theory (e.g. Teyler & DiScenna 1986). In indexing theory, the index is a pattern of hippocampal activity that is (a) generated during memory formation, (b) separate from the activity encoding memory content, and (c) linked to memory content via associative plasticity. After memory formation, a memory might be queried by activating a partial set of the memory contents, which would then drive reactivation of the hippocampal index, leading to pattern completion of memory contents. See, for example, figure 1 of Teyler and DiScenna 1986. The ‘index’ is thus not the same as the ‘query’ that drives recall.

We propose in this work that barcode activity is such an index. Indexing theory originally posited that memory content was encoded by neocortex, and memory index was encoded by hippocampus. However the experiments of Chettih et al. 2024 revealed that the hippocampus contained both memory content and memory index signals, and furthermore there was no division of cells into ‘content’ and ‘index’ subtypes. Thus our model drops the assumption of earlier work that index and content signals correspond to different neurons in different brain areas—a significant advance of our work. Otherwise, the experimentally observed barcodes and the barcodes generated by our computational model play the role of indices as originally defined.

Our original manuscript was unclear on the relationship of indexing theory and key-value systems. Our work connects diverse areas of memory models, including attractor dynamics, key-value memory systems, and memory indexing. A full account of these literatures and their relationships may be beyond the scope of this manuscript, and we note that a recent review article (Gershman, Fiete, and Irie, 2025) further clarifies the relationship between key-value memory, indexing theory, and the hippocampus. We will cite this work in our discussion as a source for the interested reader.

Briefly, a key-value memory system distinguishes between the address where a memory is stored, the ‘key’, and the content of that memory, the ‘value’. An advantage of such systems is that keys can be optimized for purposes independent of the value of each memory. The use of barcodes in our model to decorrelate memories is related to this optimization of keys in key-value memory systems. By generating barcodes and adding this to the attractor state corresponding to a cache memory, the ‘address’ of the memory in population activity is differentiated from other memories. Our work is thus consistent with the idea that hippocampus generates keys and implements a key storage system. However it is not so straightforward to equate barcodes with keys, as they are defined in key-value memory. As the reviewer points out, memory recall can be driven by location and seed inputs, i.e. it is content-addressable. We think of the barcode as modifying the memory address to better separate similar memories, without changing memory content, and the resulting memory can be recalled by querying with either content or barcode. Given the complex and speculative nature of these relationships, we prefer to note the salient connection of our work with ongoing efforts applying the key-value framework to biological memory, and leave the precise details of this connection to future work.

We make the following changes in the manuscript to clarify these ideas:

Introduction, first paragraph: In this scheme, during memory formation the hippocampus generates an index of population activity, and the neurons representing this index are linked with the neurons representing memory content by associative plasticity . Later, re-experience of partial memory contents may reactivate the index, and reactivation of the index drives complete recall of the memory contents.

Discussion, 4th paragraph on key-value: Interestingly, prior theoretical work has suggested neural implementations for both key-value memory and attention mechanisms, arguing for their usefulness in neural systems such as long term memory (Kanerva, 1988; Tyulmankov et al., 2021; Bricken and Pehlevan, 2021; Whittington et al., 2021; Kozachkov et al., 2023; Krotov and Hopfield, 2020; Gershman 2025 ). In this framework, the address where a memory is stored (the key) may be optimized independently of the value or content of the memory. In our model, barcodes improve memory performance by providing a content-independent scaffold that binds to memory content, preventing memories with overlapping content from blurring together. Thus barcodes can be considered as a change in memory address, and our model suggests important connections between recurrent neural activity and key generation mechanisms. However we note that barcodes should not be literally equated with keys in key-value systems as our model’s memory is ‘content-addresable’—it can be queried by place and seed inputs.

The model includes a number of non-standard ingredients. It would be useful to explain which of these ingredients and which of the described mechanisms are essential for the studied phenomenon. In particular:

- the dynamics in Eq.2 include a shunting inhibition term. Is it essential and why?

The shunting inhibition is important as it acts to normalize the network activity to prevent runaway excitation. We hope to clarify this further by amending the following sentence in section 2.2: “g (·) is a leak rate that depends on the average activity of the full network, representing a form of global shunting inhibition that normalizes network activity to prevent runaway excitation from recurrent dynamics.”

- same question for the global inhibition included in the random connectivity;

The distribution from which connectivity strengths are drawn has a negative mean (global inhibition). This causes activity during caching (i.e. r = 1) to be sparser than activity during visits (i.e. r = 0), and was chosen to match experimental findings. In figures 2B and S2B we show that our model can transition between a mode with place code only, barcode only, or a mode containing both, by changing the variance of the weight distribution while holding the mean constant. We suggest clarifying this by editing the following in section 2.2, paragraph 2: “We initialize the recurrent weights from a random Gaussian distribution, . where 𝑁_𝑋 is the number of RNN neurons and μ < 0, reflecting global subtractive inhibition that encourages sparse network activity to match experimental findings (Chettih et al. 2024).”

- the model is fully rate-based, but for certain figures, spikes are randomly generated. This seems superfluous.

Spikes are simulated for one analysis and one visualization, where it is important to consider noise or variability in neural responses across trials. First, for Fig. 2H,J, we generated spikes to allow a visual comparison to figures that can be easily generated from experimental data. Second, and more significantly, for the analysis underlying Fig. 3D, it is essential to simulate variability in neural responses. Because our rate-based models are noiseless, the RNN’s rate vector at site distance = 0 will always be the same and result in a correlation of 1 for both visit-visit and cache-retrieval. However, we show that, if one interprets the rate as a noisy Poisson spiking process, the correlation at site distance = 0 between a cache-retrieval pair is higher than that of two visits. This is because under a Poisson spiking model, the signal-to-noise ratio is higher for cache-retrieval activity, where rates are higher in magnitude. The greater correlation for a cache-retrieval pair at the same site, relative to visits at the same site, is an experimental finding that was critical for our model to reproduce. We detail clarifications to the manuscript below in response to the reviewer’s following and related question.

How are the correlations determined in the model (e.g., Fig 2 B)? The methods explain that they are computed from Poisson-generated spikes, but over which time period? Presumably during steady-state responses, but are these responses time-averaged?

The reviewer points out a lack of clarity in our original manuscript. Correlations for events (caches, retrievals and visits) at different sites are calculated in two sections of the paper (2B, 3D), for different purposes and with slight differences in methods:

- For figure 2B, no spikes are simulated. Note that the methods mentioning poisson spike generation specify only Fig. 2H,J and Fig. 3D. We simply take the network’s rate vector at timestep t=100 (when the decorrelating effect of chaotic dynamics has saturated, S1A-B) and correlate this vector when generated at different locations. We now clarify this in the legend for Figure 2B: “We show correlation of place inputs (gray) and correlation of the RNN's rate vector at t = 100 (black).”

- For Figure 3D, we want to compare the model to empirical results from Chettih et al. 2024, and reproduced in this paper in Fig. 1E-F. These empirical results are derived from correlating vectors of spiking activity on pairs of single trials, and are thus affected by noise or variability in neural responses as described in our response to the reviewer’s previous question. We thus took the RNN’s rate vector at t=100 and simulated spiking data by drawing samples from a poisson distribution to get spike counts. Our original manuscript was unclear about this, and we suggest the following changes:

- Legend for Figure 3D: D. Correlation of Poisson-generated spikes simulated from RNN rate vectors at two sites, plotted as a function of the distance between the two sites.

- Section 2.3, last paragraph: Population activity during retrieval closely matches activity during caching, and is substantially decorrelated from activity during visits (Figure 3C). To compare our model with the empirical results reproduced in Figure 1E,F, we ran in silico experiments with caches and retrievals at varying sites in the circular arena. We simulated Poisson-generated spikes drawn from our network's underlying rates to match the intrinsic variability in empirical data (see Methods).

- Methods, subsection Spatial correlation of RNN activity for cache-retrieval pairs at different sites: To calculate correlation values as in Figure \ref{fig3}D, we simulated experiments where 5 sites were randomly chosen for caching and retrieval. To compare model results to the empirical data in Fig. 1E,F, which includes intrinsic neural variability, we sampled Poisson-generated spike counts from the rates output by our model. Specifically, for RNN activity \vec{r_i} at location i, using the rates at t=100 as elsewhere, we first generate a sample vector of spikes…

I was confused by early and late responses in Fig 2 C. The text says that the activity is initialized at zero, so the response at t=0 should be flat (and zero). More generally, I am not sure I understand why the dynamics matter for the phenomenon at all, presumably the decorrelation shown in Fig 2B depends only on steady state activity (cf previous question).

Thanks for catching this mistake. The legend has been updated to indicate that the ‘early’ response is actually at t=1, when network activity reflects place inputs without the effects of dynamics. The reviewer is correct that we are primarily interested in the ‘late’ response of the network. All other results in the paper use this late response at t=100. As shown in Fig. S2A,B, this timepoint is not truly a steady state, as activity in the network continues to change, but the decorrelation of network activity with place-driven activity has saturated.

We include the early response in Fig. 2C for visual comparison of the purely place-driven early activity with the eventual network response. It is also relevant since, as the reviewer points out above, there is a shunting inhibition term in the dynamics that is present during both low and high recurrent strength simulations.

Related to the previous point, the discussion of decorrelation (l.79 - 97) is somewhat confusing. That paragraph focuses on chaotic activity, but chaos decorrelates responses across different time points. Here the main phenomenon is the decorrelation of responses across different spatial inputs (Fig 2B). This decorrelation is presumably due to the fact that different inputs lead to different non-trivial steady-state responses, but this requires some clarification. If that is correct, the temporal chaos adds fluctuations around these non-trivial steady-state responses, but that alone would not lead to the decorrelation shown in Fig 2B.

We agree with the reviewer that chaotic activity produces a decorrelation across time points. Because of chaotic dynamics, network activity does not settle into a trivial steady-state, and instead evolves from the initial state in an unpredictable way. The network does not settle into a steady-state pattern, but both the decorrelation of network state with initial state and the rate of change in the network state saturate after ~t=25 timesteps, as shown in Fig. S2A-B.

The initial activity for nearby states is similar, due to them receiving similar place inputs.

Because network activity is chaotically decorrelated from this initial state by temporal dynamics, ‘late stage’ network activity between nearby spatial states is less correlated than ‘early stage’ activity. Thus the temporal decorrelation produces a spatial decorrelation. We believe that the changes we have introduced to the manuscript in revision will make this point clearer in our resubmission.

A key ingredient of the model is that the recurrent interactions are switched on and off between "caching" and "visits". The discussion argues that a possible mechanism for this is recurrent inhibition (l.320), which would need to be added. However two forms of inhibition are already included in the model. The text also says that it is unclear how units in the model should be mapped onto E and I neurons. However the model makes explicit assumptions about this, in particular by generating spikes from individual neurons. Altogether, I did not find that part of the Discussion convincing.

We agree with the reviewer that this section is a limitation of our current work, and in fact it is an ongoing area of future research. However we think the advances in this current work warrant publication despite this topic requiring further research. We attempted to discuss this limitation explicitly, and note that the other reviewer pointed this section out as particularly helpful. We do not think it is problematic for a realistic model of the brain to ultimately include 3, or even more forms of inhibition. We do not think that poisson-generated spikes commit us to interpreting network units as single neurons. Spikes are not a core part of our model’s mechanism, and were used only as a mechanism of introducing variability on top of deterministic rates for specific analyses. Furthermore one could still view network units as pools of both E and I spiking neurons. We would welcome further recommendations the reviewer believes are important to note in this section on our model’s limitations.

On lines 117-120 the text briefly mentions an alternate feed-forward model and promptly discards it. The discussion instead says that a "separate possibility is that barcodes are generated in a circuit upstream of where memories are stored, and supplied as inputs to the hippocampal population", and that this possibility would lead to identical conclusions. The two statements seem a bit contradictory. It seems that the alternative possibility would replace the need for switching on and off recurrent interactions, with a mechanism where barcode inputs are switched on and off. This alternate scenario is perhaps more plausible, so it would be useful to discuss it more explicitly.

We apologize for the confusion here, which seems to be due to our phrasing in the discussion section. We do reject the idea that a simple feed-forward model could generate the spatial correlation profile observed in data, as mentioned in the text and included as Fig. S2. Our statement in the discussion may have seemed contradictory because here we intended to discuss the possibility that an upstream area generates barcodes, for example by the chaotic recurrent dynamics proposed in our work, while a downstream network receives these barcodes as inputs and undergoes plasticity to store memories as attractors. We did not intend to suggest any connection to the feedforward model of barcode generation, and apologize for the confusion. Our claim that this ‘2 network’ solution would lead to similar conclusions is because the upstream network would need an efficient means of barcode generation, and the downstream network would need an efficient means of storing memory attractors, and separating these functions into different networks is not likely to affect for example the advantage of partially decorrelating memory attractors. Moreover, the downstream network would still require some form of recurrent gating, so that during visits it exhibits place activity without activating stored memory attractors!

We thus chose a 1 network instead of a 2 network solution because it was simpler and, we believe, more interesting. It is challenging in the absence of more data to say which is more plausible, thus we wanted to mention the possibility of a 2 network solution. We suggest the following changes to the manuscript:

- Discussion, 3rd paragraph: “Alternatively, other mechanisms may be involved in generating barcodes. We demonstrated that conventional feed-forward sparsification (Babadi and Sompolinsky, 2014; Xie et al., 2023) was highly inefficient, but more specialized computations may improve this (Földiak, 1990; Olshausen and Field, 1996; Sacouto and Wichert, 2023; Muscinelli et al., 2023). Another possibility is that barcodes are generated in a separate recurrent network upstream of the recurrent network where memories are stored. In this 2-network scenario, the downstream network receives both spatial tuning and barcodes as inputs. This would not obviate the need for modulating recurrent strength in the downstream network to switch between input-driven modes and attractor dynamics. We suspect separating barcode generation and memory storage in separate networks would not fundamentally affect our conclusions.”

As a minor note, the beginning of the discussion states that the presented model is similar to previous recurrent network models of the hippocampus. It would be worth noting that several of the cited works assign a very different role to recurrent interactions: they generate place cell activity, while the present model assumes it is inherited from upstream inputs.

We are not sure how best to modify the paper to address this suggestion. As far as we know, all of the cited models which deal with spatial encoding do assume that the hippocampus receives a spatially-modulated or spatially-tuned input. For example, the Tsodyks 1999 paper cited in this paragraph uses exponentially-decaying place inputs to each neuron highly similar to our model. Furthermore we explore how our model would perform if we change the format of spatial inputs in Fig. S4, and find key results are unchanged. It is unclear how hippocampal place fields could emerge without inputs that differentiate between spatial locations. We think it is appropriate to highlight the similarity of our model to well known hopfield-type recurrent models, where memories are stored as attractor states of the network dynamics.

On the other hand, we agree that a common line of hippocampal modeling proposes that recurrent interactions reshape spatial inputs to produce place fields. This often arises in the context of hippocampus generating a predictive map, where inputs may be one-hot for a single spatial state, in a grid cell-like format, or a random projection of sensory features. We attempted to address this in section 2.6, using a model which superimposes the random connectivity needed for barcode generation with the structured connectivity needed for predictive map formation. We found that such a model was able to perform both predictive and barcode functions, suggesting a path forward to connecting different lines of hippocampal modeling in future work.

Not needd anymore just go InterPlanetary with IPFS and interpersonal Weaving the IndyWeb

10 years later

going Inter - - Personal and - Planetary

It is not semantics

we need Mutual learning Trails

TrailBlazers for Human InterIntellect Aumentation

Indy Learning commons

Symmathesy

Symmatheticist in the medium of Softwrae

Let People do it All For Themeselves

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

Reply to the reviewers

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

In this manuscript, Xiong and colleagues investigate the mechanisms operating downstream to TRIM32 and controlling myogenic progression from proliferation to differentiation. Overall, the bulk of the data presented is robust. Although further investigation of specific aspects would make the conclusions more definitive (see below), it is an interesting contribution to the field of scientists studying the molecular basis of muscle diseases.

We thank the Reviewer for appreciating our work and for their valuable suggestions to improve our manuscript. We have carefully addressed some of the concerns raised, as detailed here, while others, which require more experimental efforts, will be addressed as detailed in the Revision Plan.

In my opinion, a few aspects would improve the manuscript. Firstly, the conclusion that Trim32 regulates c-Myc mRNA stability could be expanded and corroborated by further mechanistic studies:

1. Studies investigating whether Tim32 binds directly to c-Myc RNA. Moreover, although possibly beyond the scope of this study, an unbiased screening of RNA species binding to Trim32 would be informative. Authors’ response. This point will be addressed as detailed in the Revision Plan

If possible, studies in which the overexpression of different mutants presenting specific altered functional domains (NHL domain known to bind RNAs and Ring domain reportedly involved in protein ubiquitination) would be used to test if they are capable or incapable of rescuing the reported alteration of Trim32 KO cell lines in c-Myc expression and muscle maturation.

Authors’ response. This point will be addressed as detailed in the Revision Plan

An optional aspect that might be interesting to explore is whether the alterations in c-Myc expression observed in C2C12 might be replicated with primary myoblasts or satellite cells devoid of Trim32.

Authors’ response. This point will be addressed as detailed in the Revision Plan

I also have a few minor points to highlight:

• • It is unclear if the differences highlighted in graphs 5G, EV5D, and EV5E are statistically significant.*

Authors’ response. We thank the Reviewer for raising this point. We now indicated the statistical analyses performed on the data presented in the mentioned figures (according also to a point of Reviewer #3). According to the conclusion that Trim32 is necessary for proper regulation of c-Myc transcript stability, using 2-way-ANOVA, the data now reported as Figure 5G show the statistically significant effect of the genotype at 6h (right-hand graph) but not at D0 (left-hand graph). In the graphs of Fig. EV5 D and E at D0 no significant changes are observed whereas at 6h the data show significant difference at the 40 min time point. We included this info in the graphs and in the corresponding legends.

- On page 10, it is stated that c-Myc down-regulation cannot rescue KO myotube morphology fully nor increase the differentiation index significantly, but the corresponding data is not shown. Could the authors include those quantifications in the manuscript?

Authors’ response. As suggested, we included the graph showing the differentiation index upon c-Myc silencing in the Trim32 KO clones and in the WT clones, as a novel panel in Figure 6 (Fig. 6D). As already reported in the text, a partial recovery of differentiation index is observed but the increase is not statistically significant. In contrast, no changes are observed applying the same silencing in the WT cells. Legend and text were modified accordingly.

Reviewer #1 (Significance (Required)):

The manuscript offers several strengths. It provides novel mechanistic insight by identifying a previously unrecognized role for Trim32 in regulating c-Myc mRNA stability during the onset of myogenic differentiation. The study is supported by a robust methodology that integrates CRISPR/Cas9 gene editing, transcriptomic profiling, flow cytometry, biochemical assays, and rescue experiments using siRNA knockdown. Furthermore, the work has a disease relevance, as it uncovers a mechanistic link between Trim32 deficiency and impaired myogenesis, with implications for the pathogenesis of LGMDR8. * * At the same time, the study has some limitations. The findings rely exclusively on the C2C12 myoblast cell line, which may not fully represent primary satellite cell or in vivo biology. The functional rescue achieved through c-Myc knockdown is only partial, restoring Myogenin expression but not the full differentiation index or morphology, indicating that additional mechanisms are likely involved. Although evidence supports a role for Trim32 in mRNA destabilization, the precise molecular partners-such as RNA-binding activity, microRNA involvement, or ligase function-remain undefined. Some discrepancies with previous studies, including Trim32-mediated protein degradation of c-Myc, are acknowledged but not experimentally resolved. Moreover, functional validation in animal models or patient-derived cells is currently lacking. Despite these limitations, the study represents an advancement for the field. It shifts the conceptual framework from Trim32's canonical role in protein ubiquitination to a novel function in RNA regulation during myogenesis. It also raises potential clinical implications by suggesting that targeting the Trim32-c-Myc axis, or modulating c-Myc stability, may represent a therapeutic strategy for LGMDR8. This work will be of particular interest to muscle biology researchers studying myogenesis and the molecular basis of muscle disease, RNA biology specialists investigating post-transcriptional regulation and mRNA stability, and neuromuscular disease researchers and clinicians seeking to identify new molecular targets for therapeutic intervention in LGMDR8. * * The Reviewer expressing this opinion is an expert in muscle stem cells, muscle regeneration, and muscle development.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

Summary: * * In this study, the authors sought to investigate the molecular role of Trim32, a tripartite motif-containing E3 ubiquitin ligase often associated with its dysregulation in Limb-Girdle Muscular Dystrophy Recessive 8 (LGMDR8), and its role in the dynamics of skeletal muscle differentiation. Using a CRISPR-Cas9 model of Trim32 knockout in C2C12 murine myoblasts, the authors demonstrate that loss of Trim32 alters the myogenic process, particularly by impairing the transition from proliferation to differentiation. The authors provide evidence in the way of transcriptomic profiling that displays an alteration of myogenic signaling in the Trim32 KO cells, leading to a disruption of myotube formation in-vitro. Interestingly, while previous studies have focused on Trim32's role in protein ubiquitination and degradation of c-Myc, the authors provide evidence that Trim32-regulation of c-Myc occurs at the level of mRNA stability. The authors show that the sustained c-Myc expression in Trim32 knockout cells disrupts the timely expression of key myogenic factors and interferes with critical withdrawal of myoblasts from the cell cycle required for myotube formation. Overall, the study offers a new insight into how Trim32 regulates early myogenic progression and highlights a potential therapeutic target for addressing the defects in muscular regeneration observed in LGMDR8.

We thank the Reviewer for valuing our work and for their appreciated suggestions to improve our manuscript. We have carefully addressed some of the concerns raised as detailed here, while others, which require more laborious experimental efforts, will be addressed as reported in the Revision Plan.

Major Comments:

The work is a bit incremental based on this:

https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0030445 * * And this:

https://www.nature.com/articles/s41418-018-0129-0 * * To their credit, the authors do cite the above papers.

Authors’ response. We thank the Reviewer for this careful evaluation of our work against the current literature and for recognising the contribution of our findings to the understanding of myogenesis complex picture in which the involvement of Trim32 and c-Myc, and of the Trim32-c-Myc axis, can occur at several stages and likely in narrow time windows along the process, thus possibly explaining some reports inconsistencies.

The authors do provide compelling evidence that Trim32 deficiency disrupts C2C12 myogenic differentiation and sustained c-Myc expression contributes to this defective process. However, while knockdown of c-Myc does restore Myogenin levels, it was not sufficient to normalize myotube morphology or differentiation index, suggesting an incomplete picture of the Trim32-dependent pathways involved. The authors should qualify their claim by emphasizing that c-Myc regulation is a major, but not exclusive, mechanism underlying the observed defects. This will prevent an overgeneralization and better align the conclusions with the author's data.

Authors’ response. We agree with the Reviewer and we modified our phrasing that implied Trim32-c-Myc axis as the exclusive mechanism by explicitly indicated that other pathways contribute to guarantee proper myogenesis, in the Abstract and in Discussion.

The Abstract now reads: “… suggesting that the Trim32–c-Myc axis may represent an essential hub, although likely not the exclusive molecular mechanism, in muscle regeneration within LGMDR8 pathogenesis.”

The Discussion now reads: “Functionally, we demonstrated that c-Myc contributes to the impaired myogenesis observed in Trim32 KO clones, although this is clearly not the only factor involved in the Trim32-mediated myogenic network; realistically other molecular mechanisms can participate in this process as also suggested by our transcriptomic results.”

The authors provide a thorough and well-executed interrogation of cell cycle dynamics in Trim32 KO clones, combining phosphor-histone H3 flow cytometry of DNA content, and CFSE proliferation assays. These complementary approaches convincingly show that, while proliferation states remain similar in WT and KO cells, Trim32-deficient myoblasts fail in their normal withdraw from the cell cycle during exposure to differentiation-inducing conditions. This work adds clarity to a previously inconsistent literature and greatly strengthens the study.

Authors’ response. We thank the Reviewer for appreciating our thorough analyses on cell cycle dynamics in proliferation conditions and at the onset of the differentiation process.

The transcriptomic analysis (detailed In the "Transcriptomic analysis of Trim32 WT and KO clones along early differentiation" section of Results) is central to the manuscript and provides strong evidence that Trim32 deficiency disrupts normal differentiation processes. However, the description of the pathway enrichment results is highly detailed and somewhat compressed, which may make it challenging for readers to following the key biological 'take-homes'. The narrative quickly moves across their multiple analyses like MDS, clustering, heatmaps, and bubble plots without pausing to guide the reader through what each analysis contributes to the overall biological interpretation. As a result, the key findings (reduced muscle development pathways in KO cells and enrichment of cell cycle-related pathways) can feel somewhat muted. The authors may consider reorganizing this section, so the primary biological insights are highlighted and supported by each of their analyses. This would allow the biological implications to be more accessible to a broader readership.

Authors’ response. We thank the Reviewer for raising this point and apologise for being too brief in describing the data, leaving indeed some points excessively implicit. As suggested, we now reorganised this session and added the lists of enriched canonical pathways relative to WT vs KO comparisons at D0 and D3 (Fig. EV3B) as well as those relative to the comparison between D0 and D3 for both WT and Trim32 KO samples (Fig. EV3C), with their relative scores. We changed the Results section “Transcriptomic analysis of Trim32 WT and Trim32 KO clones along early differentiation” as reported here below and modified the legends accordingly.

The paragraph now reads: “Based on our initial observations, the absence of Trim32 already exerts a significant impact by day 3 (D3) of C2C12 myogenic differentiation. To investigate how Trim32 influences early global transcriptional changes during the proliferative phase (D0) and early differentiation (D3), we performed an unbiased transcriptomic profiling of WT and Trim32 KO clones (Fig. 2A). Multidimensional Scaling (MDS) analysis revealed clear segregation of gene expression profiles based on both time of differentiation (Dim1, 44% variance) and Trim32 genotype (Dim2, 16% variance) (Fig. 2A). Likewise, hierarchical clustering grouped WT and Trim32 KO clones into distinct clusters at both timepoints, indicating consistent genotype-specific transcriptional differences (Fig. EV3A). Differentially Expressed Genes (DEGs) were detected in the Trim32 KO transcriptome relative to WT, at both D0 and D3. In proliferating conditions, 72 genes were upregulated and 189 were downregulated whereas at D3 of differentiation, 72 genes were upregulated and 212 were downregulated. Ingenuity Pathway Analysis of the DEGs revealed the top 10 Canonical Pathways displayed in Fig. EV3B as enriched at either D0 or D3 (Fig. EV3B). Several of these pathways can underscore relevant Trim32-mediated functions though most of them represent generic functions not immediately attributable to the observed myogenesis defects.

Notably, the transcriptional divergence between WT and Trim32 KO cells is more pronounced at D3, as evidenced by a greater separation along the MSD Dim2 axis, suggesting that Trim32-dependent transcriptional regulation intensifies during early differentiation (Fig. 2A). Given our interest in the differentiation process, we therefore focused our analyses comparing the changes occurring from D0 to D3 in WT (WT D3 vs. D0) and in Trim32 KO (KO D3 vs. D0) RNAseq data.

Pathway enrichment analysis of D3 vs. D0 DEGs allowed the selection of the top-scored pathways for both WT and Trim32 KO data. We obtained 18 top-scored pathways enriched in each genotype (-log(p-value) ³ 9 cut-off): 14 are shared while 4 are top-ranked only in WT and 4 only in Trim32 KO (Fig. EV3C). For the following analyses, we employed thus a total of 22 distinct pathways and to better mine those relevant in the passage from the proliferation stage to the early differentiation one and that are affected by the lack of Trim32, we built a bubble plot comparing side-by-side the scores and enrichment of the 22 selected top-scored pathways above in WT and Trim32 KO (Fig. 2B). A heatmap of DEGs included within these selected pathways confirms the clustering of the samples considering both the genotypes and the timepoints highlighting gene expression differences (Fig. 2C). These pathways are mainly related to muscle development, cell cycle regulation, genome stability maintenance and few other metabolic cascades.

As expected given the results related to Figure 1, moving from D0 to D3 WT clones showed robust upregulation of key transcripts associated with the Inactive Sarcomere Protein Complex, a category encompassing most genes in the “Striated Muscle Contraction” pathway, while in Trim32 KO clones this pathway was not among those enriched in the transition from D0 to D3 (Fig. EV3C). Detailed analyses of transcripts enclosed within this pathway revealed that on the transition from proliferation to differentiation, WT clones show upregulation of several Myosin Heavy Chain isoforms (e.g., MYH3, MYH6, MYH8), α-Actin 1 (ACTA1), α-Actinin 2 (ACTN2), Desmin (DES), Tropomodulin 1 (TMOD1), and Titin (TTN), a pattern consistent with previous reports, while these same transcripts were either non-detected or only modestly upregulated in Trim32 KO clones at D3 (Fig. 2D). This genotype-specific disparity was further confirmed by gene set enrichment barcode plots, which demonstrated significant enrichment of these muscle-related transcripts in WT cells (FDR_UP = 0.0062), but not in Trim32 KO cells (FDR_UP = 0.24) (Fig. EV3D). These findings support an early transcriptional basis for the impaired myogenesis previously observed in Trim32 KO cells.

In addition to differences in muscle-specific gene expression, we observed that also several pathways related to cell proliferation and cell cycle regulation were more enriched in Trim32 KO cells compared to WT. This suggests that altered cell proliferation may contribute to the distinct differentiation behavior observed in Trim32 KO versus WT (Fig. 2B). Given that cell cycle exit is a critical prerequisite for the onset of myogenic differentiation and considering that previous studies on Trim32 role in cell cycle regulation have reported inconsistent findings, we further examined cell cycle dynamics under our experimental conditions to clarify Trim32 contribution to this process”

The work would be greatly strengthened by the conclusion of LGMDR8 primary cells, and rescue experiments of TRIM32 to explore myogenesis.

Authors’ response. This point will be addressed as detailed in the Revision Plan

Also, EU (5-ethynyl uridine) pulse-chase experiments to label nascent and stable RNA coupled with MYC pulldowns and qPCR (or RNA-sequencing of both pools) would further enhance the claim that MYC stability is being affected.

Authors’ response. This point will be addressed as detailed in the Revision Plan

"On one side, c-Myc may influence early stages of myogenesis, such as myoblast proliferation and initial myotube formation, but it may not contribute significantly to later events such as myotube hypertrophy or fusion between existing myotubes and myocytes. This hypothesis is supported by recent work showing that c-Myc is dispensable for muscle fiber hypertrophy but essential for normal MuSC function (Ham et al, 2025)." Also address and discuss the following, as what is currently written is not entirely accurate: https://www.embopress.org/doi/full/10.1038/s44319-024-00299-z and https://journals.physiology.org/doi/prev/20250724-aop/abs/10.1152/ajpcell.00528.2025

Authors’ response. We thank the Reviewer for bringing to our attention these two publications, that indeed, add important piece of data to recapitulate the in vivo complexity of c-Myc role in myogenesis. We included this point in our Discussion.

The Discussion now reads: “On one side, c-Myc may influence early stages of myogenesis, such as myoblast proliferation and initial myotube formation, but it may not contribute significantly to later events such as myotube hypertrophy or fusion between existing myotubes and myocytes. This hypothesis is supported by recent work showing that c-Myc is dispensable for muscle fiber hypertrophy but essential for normal MuSC function (Ham et al, 2025). Other reports, instead, demonstrated the implication of c-Myc periodic pulses, mimicking resistance-exercise, in muscle growth, a role that cannot though be observed in our experimental model (Edman et al., 2024; Jones et al., 2025).”

Minor Comments:

Z-score scale used in the pathway bubble plot (Figure 2C) could benefit from alternative color choices. Current gradient is a bit muddy and clarity for the reader could be improved by more distinct color options, particularly in the transition from positive to negative Z-score.

Authors’ response. As suggested, we modified the z-score-representing colors using a more distinct gradient especially in the positive to negative transition in Figure 2B.

Clarification on the rationale for selecting the "top 18" pathways would be helpful, as it is not clear if this cutoff was chosen arbitrarily or reflects a specific statistical or biological threshold.

Authors’ response. As now better explained (see comment regarding Major point: Transcriptomics), we used a cut-off of -log(p-value) above or equal to 9 for pathways enriched in DEGs of the D0 vs D3 comparison for both WT and Trim32 KO. The threshold is now included in the Results section and the pathways (shared between WT and Trim32 KO and unique) are listed as Fig. EV3C.

The authors alternates between using "Trim 32 KO clones" and "KO clones" throughout the manuscript. Consistent terminology across figures and text would improve readability.

Authors’ response. We thank the Reviewer for this remark, and we apologise for having overlooked it. We amended this throughout the manuscript by always using for clarity “Trim32 KO clones/cells”.

Cell culture methodology does not specify passage number or culture duration (only "At confluence") before differentiation. This is important, as C2C12 differentiation potential can drift with extended passaging.

Authors’ response. We agree with the Reviewer that C2C12 passaging can reduce the differentiation potential of this myoblast cell lines; this is indeed the main reason why we decided to employ WT clones, which underwent the same editing process as those that resulted mutated in the Trim32 gene, as reference controls throughout our study. We apologise for not indicating the passages in the first version of the manuscript that now is amended as per here below in the Methods section:

“The C2C12 parental cells used in this study were maintained within passages 3–8. All clonal cell lines (see below) were utilized within 10 passages following gene editing. In all experiments, WT and Trim32 KO clones of comparable passage numbers were used to ensure consistency and minimize passage-related variability.”

Reviewer #2 (Significance (Required)):

General Assessment:

This study provides a thorough investigation of Trim32's role the processes related to skeletal muscle differentiation using a CRISPR-Cas9 knockout C2C12 model. The strengths of this study lie in the multi-layered experimental approach as the authors incorporated transcriptomics, cell cycle profiling, and stability assays which collectively build a strong case for their hypothesis that Trim32 is a key factor in the normal regulation of myogenesis. The work is also strengthened by the use of multiple biological and technical replicates, particularly the independent KO clones which helps address potential clonal variation issues that could occur. The largest limitation to this study is that, while the c-Myc mechanism is well explored, the other Trim32-dependent pathways associated with the disruption (implicated by the incomplete rescue by c-Myc knockdown) are not as well addressed. Overall however, the study convincingly identifies a critical function for Trim32 during skeletal muscle differentiation. * * Advance: * * To my knowledge, this is the first study to demonstrate the mRNA stability level of c-Myc regulation by Trim32, rather than through the ubiquitin-mediated protein degradation. This work will advance the current understanding and provide a more complete understanding of Trim32's role in c-Myc regulation. Beyond c-Myc, this work highlights the idea that TRIM family proteins can influence RNA stability which could implicate a broader role in RNA biology and has potential for future therapeutic targeting. * * Audience: * * This research will be of interest to an audience that focuses on broad skeletal muscle biology but primarily to readers with more focused research such as myogenesis and neuromuscular disease (LGMDR8 in particular) where the defined Trim32 governance over early differentiation checkpoints will be of interest. It will also provide mechanistic insights to those outside of skeletal muscle that study TRIM family proteins, ubiquitin biology, and RNA regulation. For translational/clinical researchers, it identifies the Trim32/c-Myc axis as a potential therapeutic target for LGMDR8 and related muscular dystrophies.

Expertise: * * My expertise lies in skeletal muscle biology, gene editing, transgenic mouse models, and bioinformatics. I feel confident evaluating the data and conclusions as presented.

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

• In this paper, the authors examine the role of TRIM32, implicated in limb girdle muscular dystrophy recessive 8 (LGMDR8), in the differentiation of C2C12 mouse myoblasts. Using CRISPR, they generate mutant and wild-type clones and compare their differentiation capacity in vitro. They report that Trim32-deficient clones exhibit delayed and defective myogenic differentiation. RNA-seq analysis reveals widespread changes in gene expression, although few are validated by independent methods. Notably, Trim32 mutant cells maintain residual proliferation under differentiation conditions, apparently due to a failure to downregulate c-Myc. Translation inhibition experiments suggest that TRIM32 promotes c-Myc mRNA destabilization, but this conclusion is insufficiently substantiated. The authors also perform rescue experiments, showing that c-Myc knockdown in Trim32-deficient cells alleviates some differentiation defects. However, this rescue is not quantified, was conducted in only two of the three knockout lines, and is supported by inappropriate statistical analysis of gene expression. Overall, the manuscript in its current form has substantial weaknesses that preclude publication. Beyond statistical issues, the major concerns are: (1) exclusive reliance on the immortalized C2C12 line, with no validation in primary/satellite cells or in vivo, (2) insufficient mechanistic evidence that TRIM32 acts directly on c-Myc mRNA, and (3) overinterpretation of disease relevance in the absence of supporting patient or in vivo data. Please find more details below:*

We thank the Reviewer for the in-depth assessment of our work and precious suggestions to improve the manuscript. We have carefully addressed some of the concerns raised, as detailed here, while others, which require more experimental efforts, will be addressed as detailed in the Revision Plan.

- TRIM32 complementation / rescue experiments to exclude clonal or off-target CRISPR effects and show specificity are lacking.

Authors’ response. This point will be addressed as detailed in the Revision Plan

- The authors link their in vitro findings to LGMDR8 pathogenesis and propose that the Trim32-c-Myc axis may serve as a central regulator of muscle regeneration in the disease. However, LGMDR8 is a complex disorder, and connecting muscle wasting in patients to differentiation assays in C2C12 cells is difficult to justify. No direct evidence is provided that the proposed mRNA mechanism operates in patient-derived samples or in mouse satellite cells. Moreover, the partial rescue achieved by c-Myc knockdown (which does not fully restore myotube morphology or differentiation index) further suggests that the disease connection is not straightforward. Validation of the TRIM32-c-Myc axis in a physiologically relevant system, such as LGMD patient myoblasts or Trim32 mutant mouse cells, would greatly strengthen the claim.

Authors’ response. This point will be addressed as detailed in the Revision Plan

-Some gene expression changes from the RNA-seq study in Figure 2 should be validated by qPCR

Authors’ response. We thank the reviewer for this suggestion. This point will be addressed as detailed in the Revision Plan. We have selected several transcripts that will be evaluated in independent samples in order to validate the RNAseq results.

- The paper shows siRNA knockdown of c-Myc in KO restores Myogenin RNA/protein but does not fully rescue myotube morphology or differentiation index. This suggests that Trim32 controls additional effectors beyond c-Myc; yet the authors do not pursue other candidate mediators identified in the RNA-seq. The manuscript would be strengthened by systematically testing whether other deregulated transcripts contribute to the phenotype.

Authors’ response. This point will be addressed as detailed in the Revision Plan

- There are concerns with experimental/statistical issues and insufficient replicate reporting. The authors use unpaired two-tailed Student's t-test across many comparisons; multiple testing corrections or ANOVA where appropriate should be used. In Figure EV5B and Figure 6B, the authors perform statistical analyses with control values set to 1. This method masks the inherent variability between experiments and artificially augments p values. Control sample values need to be normalized to one another to have reliable statistical analysis. Myotube morphology and differentiation index quantifications need clear description of fields counted, blind analysis, and number of biological replicates.

Authors’ response. We thank the Reviewer for raising this point.

Regarding the replicates, we clarified in the Methods and Legends that the Trim32 KO experiments have been performed on 3 biological replicates (independent clones) and the same for the reference control (3 independent WT clones), except for the Fig. 6 experiments that were performed on 2 Trim32 KO and 2 WT clones. All the Western Blots, immunofluorescence, qPCR data are representative of the results of at least 3 independent experiments unless otherwise stated. We reported the number and type of replicates as well as the microscope fields analyzed.

We repeated the statistical analyses of the data in Figure 5G, EV5D, EV5E, employing more appropriately the 2-way-ANOVA test, as suggested, and we now reported this info in the graphs and legends.

We thank the Reviewer for raising this point, we agree and substituted the graphs in Fig. EV5B and 6B showing the control values normalised as suggested. The statistical analyses now reflect this change.

-Some English mistakes require additional read-throughs. For example: "Indeed, Trim32 has no effect on the stability of c-Myc mRNA in proliferating conditions, but upon induction of differentiation the stability of c-Myc mRNA resulted enhanced in Trim32 KO clones (Fig. 5G, Fig. EV5D and 5E)."

Authors’ response. We re-edited this revised version of the manuscript as suggested.

-Results in Figure 5A should be quantified

Authors’ response. We amended this point by quantifying the results shown in Fig. 5A, we added the graph of the quantification of 3 experimental replicates to the Figure. Quantification confirms that no statistically significant difference is observed. The Figure and the relative legend are modified accordingly.