приложениями?

неразрывный пробел

неразрывный пробел

приложениями?

неразрывный пробел

неразрывный пробел

неразрывный пробел

Я бы написала "Возможности Kaspersky Secure Connection" или "Что может Kaspersky Secure Connection?"

неразрывный пробел

Author Response

Reviewer #2 (Public Review):

Here, a simple model of cerebellar computation is used to study the dependence of task performance on input type: it is demonstrated that task performance and optimal representations are highly dependent on task and stimulus type. This challenges many standard models which use simple random stimuli and concludes that the granular layer is required to provide a sparse representation. This is a useful contribution to our understanding of cerebellar circuits, though, in common with many models of this type, the neural dynamics and circuit architecture are not very specific to the cerebellum, the model includes the feedforward structure and the high dimension of the granule layer, but little else. This paper has the virtue of including tasks that are more realistic, but by the paper’s own admission, the same model can be applied to the electrosensory lateral line lobe and it could, though it is not mentioned in the paper, be applied to the dentate gyrus and large pyramidal cells of CA3. The discussion does not include specific elements related to, for example, the dynamics of the Purkinje cells or the role of Golgi cells, and, in a way, the demonstration that the model can encompass different tasks and stimuli types is an indication of how abstract the model is. Nonetheless, it is useful and interesting to see a generalization of what has become a standard paradigm for discussing cerebellar function.

We appreciate the Reviewer’s positive comments. Regarding the simplifications of our model, we agree that we have taken a modeling approach that abstracts away certain details to permit comparisons across systems. We now include an in-depth discussion of our simplifying assumptions (Assumptions & Extensions section in the Discussion) and have further noted the possibility that other biophysical mechanisms we have not accounted for may also underlie differences across systems.

Our results predict that qualitative differences in the coding levels of cerebellum-like systems, across brain regions or across species, reflect an optimization to distinct tasks (Figure 7). However, it is also possible that differences in coding level arise from other physiological differences between systems.

Reviewer #3 (Public Review):

1) The paper by Xie et al is a modelling study of the mossy fiber-to-granule cell-to-Purkinje cell network, reporting that the optimal type of representations in the cerebellar granule cell layer depends on the type task. The paper stresses that the findings indicate a higher overall bias towards dense representations than stated in the literature, but it appears the authors have missed parts of the literature that already reported on this. While the modelling and analysis appear mathematically solid, the model is lacking many known constraints of the cerebellar circuitry, which makes the applicability of the findings to the biological counterpart somewhat limited.

We thank the Reviewer for suggesting additional references to include in our manuscript, and for encouraging us to extend our model toward greater biological plausibility and more critically discuss simplifying assumptions we have made. We respond to both the comment about previous literature and about applicability to cerebellar circuitry in detail below.

2) I have some concerns with the novelty of the main conclusion, here from the abstract: ’Here, we generalize theories of cerebellar learning to determine the optimal granule cell representation for tasks beyond random stimulus discrimination, including continuous input-output transformations as required for smooth motor control. We show that for such tasks, the optimal granule cell representation is substantially denser than predicted by classic theories.’ Stated like this, this has in principle already been shown, i.e. for example: Spanne and Jo¨rntell (2013) Processing of multi-dimensional sensorimotor information in the spinal and cerebellar neuronal circuitry: a new hypothesis. PLoS Comput Biol. 9(3):e1002979. Indeed, even the 2 DoF arm movement control that is used in the present paper as an application, was used in this previous paper, with similar conclusions with respect to the advantage of continuous input-output transformations and dense coding. Thus, already from the beginning of this paper, the novelty aspect of this paper is questionable. Even the conclusion in the last paragraph of the Introduction: ‘We show that, when learning input-output mappings for motor control tasks, the optimal granule cell representation is much denser than predicted by previous analyses.’ was in principle already shown by this previous paper.

We thank the Reviewer for drawing our attention to Spanne and Jo¨rntell (2013). Our study shares certain similarities with this work, including the consideration of tasks with smooth input-output mappings, such as learning the dynamics of a two-joint arm. However, our study differs substantially, most notably the fact that we focus our study on parametrically varying the degree of sparsity in the granule cell layer to determine the circumstances under which dense versus sparse coding is optimal. To the best of our ability, we can find no result in Spanne and J¨orntell (2013) that indicates the performance of a network as a function of average coding level. Instead, Spanne and Jo¨rntell (2013) propose that inhibition from Golgi cells produces heterogeneity in coding level which can improve performance, which is an interesting but complementary finding to ours. We therefore do not believe that the quantitative computations of optimal coding level that we present are redundant with the results of this previous study. We also note that a key contribution of our study is mathemetical analysis of the inductive bias of networks with different coding levels which supports our conclusions.

We have included a discussion of Spanne and Jo¨rntell (2013) and (2015) in the revised version of our manuscript:

"Other studies have considered tasks with smooth input-output mappings and low-dimensional inputs, finding that heterogeneous Golgi cell inhibition can improve performance by diversifying individual granule cell thresholds (Spanne and J¨orntell, 2013). Extending our model to include heterogeneous thresholds is an interesting direction for future work. Another proposal states that dense coding may improve generalization (Spanne and Jo¨rntell, 2015). Our theory reveals that whether or not dense coding is beneficial depends on the task."

3) However, the present paper does add several more specific investigations/characterizations that were not previously explored. Many of the main figures report interesting new model results. However, the model is implemented in a highly generic fashion. Consequently, the model relates better to general neural network theory than to specific interpretations of the function of the cerebellar neuronal circuitry. One good example is the findings reported in Figure 2. These represent an interesting extension to the main conclusion, but they are also partly based on arbitrariness as the type of mossy fiber input described in the random categorization task has not been observed in the mammalian cerebellum under behavior in vivo, whereas in contrast, the type of input for the motor control task does resemble mossy fiber input recorded under behavior (van Kan et al 1993).

We agree that the tasks we consider in Figure 2 are simplified compared to those that we consider elsewhere in the paper. The choice of random mossy fiber input was made to provide a comparison to previous modeling studies that also use random input as a benchmark (Marr 1969, Albus 1971, Brunel 2004, Babadi and Sompolinsky 2014, Billings 2014, LitwinKumar et al., 2017). This baseline permits us to specifically evaluate the effects of lowdimensional inputs (Figure 2) and richer input-output mappings (Figure 2, Figure 7). We agree with the Reviewer that the random and uncorrelated mossy fiber activity that has been extensively used in previous studies is almost certainly an unrealistic idealization of in vivo neural activity—this is a motivating factor for our study, which relaxes this assumption and examines the consequences. To provide additional context, we have updated the following paragraph in the main text Results section:

"A typical assumption in computational theories of the cerebellar cortex is that inputs are randomly distributed in a high-dimensional space (Marr, 1969; Albus, 1971; Brunel et al., 2004; Babadi and Sompolinsky, 2014; Billings et al., 2014; Litwin-Kumar et al., 2017). While this may be a reasonable simplification in some cases, many tasks, including cerebellumdependent tasks, are likely best-described as being encoded by a low-dimensional set of variables. For example, the cerebellum is often hypothesized to learn a forward model for motor control (Wolpert et al., 1998), which uses sensory input and motor efference to predict an effector’s future state. Mossy fiber activity recorded in monkeys correlates with position and velocity during natural movement (van Kan et al., 1993). Sources of motor efference copies include motor cortex, whose population activity lies on a lowdimensional manifold (Wagner et al., 2019; Huang et al., 2013; Churchland et al., 2010; Yu et al., 2009). We begin by modeling the low dimensionality of inputs and later consider more specific tasks."

4) The overall conclusion states: ‘Our results....suggest that optimal cerebellar representations are task-dependent.’ This is not a particularly strong or specific conclusion. One could interpret this statement as simply saying: ‘if I construct an arbitrary neural network, with arbitrary intrinsic properties in neurons and synapses, I can get outputs that depend on the intensity of the input that I provide to that network.’ Further, the last sentence of the Introduction states: ‘More broadly, we show that the sparsity of a neural code has a task-dependent influence on learning...’ This is very general and unspecific, and would likely not come as a surprise to anyone interested in the analysis of neural networks. It doesn’t pinpoint any specific biological problem but just says that if I change the density of the input to a [generic] network, then the learning will be impacted in one way or another.

We agree with the Reviewer that our conclusions are quite general, and we have removed the final sentence as we agree it was unspecific. However, we disagree with the Reviewer’s paraphrasing of our results.

First, we do not select arbitrary intrinsic properties of neurons and synapses. Rather, we construct a simplified model with a key quantity, the neuronal threshold, that we vary parametrically in order to assess the effect of the resulting changes in the representation on performance. Second, we do not vary the intensity/density of inputs provided to the network – this is fixed throughout our study for all key comparisons we perform. Instead, we vary the density (coding level) of the expansion layer representation and quantify its effect on inductive bias and generalization. Finally, our study’s key contribution is an explanation of the heterogeneity in average coding level observed across behaviors and cerebellum-like systems. We go beyond the empirical statement that there is a dependence of performance on the parameter that we vary by developing an analytical theory. Our theory describes the performance of the class of networks that we study and the properties of learning tasks that determine the optimal expansion layer representation.

To clarify our main contributions, we have updated the final paragraph of the Introduction. We have also removed the sentence that the Reviewer objects to, as it was less specific than the other points we make here.

"We propose that these differences can be explained by the capacity of representations with different levels of sparsity to support learning of different tasks. We show that the optimal level of sparsity depends on the structure of the input-output relationship of a task. When learning input-output mappings for motor control tasks, the optimal granule cell representation is much denser than predicted by previous analyses. To explain this result, we develop an analytic theory that predicts the performance of cerebellum-like circuits for arbitrary learning tasks. The theory describes how properties of cerebellar architecture and activity control these networks’ inductive bias: the tendency of a network toward learning particular types of input-output mappings (Sollich, 1998; Jacot et al., 2018; Bordelon et al., 2020; Canatar et al., 2021; Simon et al., 2021). The theory shows that inductive bias, rather than the dimension of the representation alone, is necessary to explain learning performance across tasks. It also suggests that cerebellar regions specialized for different functions may adjust the sparsity of their granule cell representations depending on the task."

5) The interpretation of the distribution of the mossy fiber inputs to the granule cells, which would have a crucial impact on the results of a study like this, is likely incorrect. First, unlike the papers that the authors cite, there are many studies indicating that there is a topographic organization in the mossy fiber termination, such that mossy fibers from the same inputs, representing similar types of information, are regionally co-localized in the granule cell layer. Hence, there is no support for the model assumption that there is a predominantly random termination of mossy fibers of different origins. This risks invalidating the comparisons that the authors are making, i.e. such as in Figure 3. This is a list of example papers, there are more: van Kan, Gibson and Houk (1993) Movement-related inputs to intermediate cerebellum of the monkey. Journal of Neurophysiology. Garwicz et al (1998) Cutaneous receptive fields and topography of mossy fibres and climbing fibres projecting to cat cerebellar C3 zone. The Journal of Physiology. Brown and Bower (2001) Congruence of mossy fiber and climbing fiber tactile projections in the lateral hemispheres of the rat cerebellum. The Journal of Comparative Neurology. Na, Sugihara, Shinoda (2019) The entire trajectories of single pontocerebellar axons and their lobular and longitudinal terminal distribution patterns in multiple aldolase C-positive compartments of the rat cerebellar cortex. The Journal of Comparative Neurology.

6) The nature of the mossy fiber-granule cell recording is also reviewed here: Gilbert and Miall (2022) How and Why the Cerebellum Recodes Input Signals: An Alternative to Machine Learning. The Neuroscientist. Further, considering the re-coding idea, the following paper shows that detailed information, as it is provided by mossy fibers, is transmitted through the granule cells without any evidence of re-coding: Jo¨rntell and Ekerot (2006) Journal of Neuroscience; and this paper shows that these granule inputs are powerfully transmitted to the molecular layer even in a decerebrated animal (i.e. where only the ascending sensory pathways remains) Jo¨rntell and Ekerot 2002, Neuron.

We agree that there is strong evidence for a topographic organization in mossy fiber to granule cell connectivity at the microzonal level. We thank the Reviewer for pointing us to specific examples. We acknowledge that our simplified model does not capture the structure of connectivity observed in these studies.

However, the focus of our model is on cerebellar neurons presynaptic to a single Purkinje cell. Random or disordered distribution of inputs at this local scale is compatible with topographic organization at the microzonal scale. Furthermore, while there is evidence of structured connections at the local scale, models with random connectivity are able to reproduce the dimensionality of granule cell activity within a small margin of error (Nguyen et al., 2022). Finally, our finding that dense codes are optimal for learning slowly varying tasks is consistent with evidence for the lack of re-coding – for such tasks, re-coding may absent because it is not required.

We have dedicated a section on this issue in the Assumptions and Extensions portion of our Discussion:

"Another key assumption concerning the granule cells is that they sample mossy fiber inputs randomly, as is typically assumed in Marr-Albus models (Marr, 1969; Albus, 1971; LitwinKumar et al., 2017; Cayco-Gajic et al., 2017). Other studies instead argue that granule cells sample from mossy fibers with highly similar receptive fields (Garwicz et al., 1998; Brown and Bower, 2001; J¨orntell and Ekerot, 2006) defined by the tuning of mossy fiber and climbing fiber inputs to cerebellar microzones (Apps et al., 2018). This has led to an alternative hypothesis that granule cells serve to relay similarly tuned mossy fiber inputs and enhance their signal-to-noise ratio (Jo¨rntell and Ekerot, 2006; Gilbert and Chris Miall, 2022) rather than to re-encode inputs. Another hypothesis is that granule cells enable Purkinje cells to learn piece-wise linear approximations of nonlinear functions (Spanne and J¨orntell, 2013). However, several recent studies support the existence of heterogeneous connectivity and selectivity of granule cells to multiple distinct inputs at the local scale (Huang et al., 2013; Ishikawa et al., 2015). Furthermore, the deviation of the predicted dimension in models constrained by electron-microscopy data as compared to randomly wired models is modest (Nguyen et al., 2022). Thus, topographically organized connectivity at the macroscopic scale may coexist with disordered connectivity at the local scale, allowing granule cells presynaptic to an individual Purkinje cell to sample heterogeneous combinations of the subset of sensorimotor signals relevant to the tasks that Purkinje cell participates in. Finally, we note that the optimality of dense codes for learning slowly varying tasks in our theory suggests that observations of a lack of mixing (J¨orntell and Ekerot, 2002) for such tasks are compatible with Marr-Albus models, as in this case nonlinear mixing is not required."

7) I could not find any description of the neuron model used in this paper, so I assume that the neurons are just modelled as linear summators with a threshold (in fact, Figure 5 mentions inhibition, but this appears to be just one big lump inhibition, which basically is an incorrect implementation). In reality, granule cells of course do have specific properties that can impact the input-output transformation, PARTICULARLY with respect to the comparison of sparse versus dense coding, because the low-pass filtering of input that occurs in granule cells (and other neurons) as well as their spike firing stochasticity (Saarinen et al (2008). Stochastic differential equation model for cerebellar granule cell excitability. PLoS Comput. Biol. 4:e1000004) will profoundly complicate these comparisons and make them less straight forward than what is portrayed in this paper. There are also several other factors that would be present in the biological setting but are lacking here, which makes it doubtful how much information in relation to the biological performance that this modelling study provides: What are the types of activity patterns of the inputs? What are the learning rules? What is the topography? What is the impact of Purkinje cell outputs downstream, as the Purkinje cell output does not have any direct action, it acts on the deep cerebellar nuclear neurons, which in turn act on a complex sensorimotor circuitry to exert their effect, hence predictive coding could only become interpretable after the PC output has been added to the activity in those circuits. Where is the differentiated Golgi cell inhibition?

Thank you for these critiques. We have made numerous edits to improve the presentation of the details of our model in the main text of the manuscript. Indeed, granule cells in the main text are modeled as linear sums of mossy fiber inputs with a threshold-linear activation function. A more detailed description of the model for granule cells can now be found in Equation 1 in the Results section:

"The activity of neurons in the expansion layer is given by: h = φ(Jeffx − θ), (1) where φ is a rectified linear activation function φ(u) = max(u,0) applied element-wise. Our results also hold for other threshold-polynomial activation functions. The scalar threshold θ is shared across neurons and controls the coding level, which we denote by f, defined as the average fraction of neurons in the expansion layer that are active."

Most of our analyses use the firing rate model we describe above, but several Supplemental Figures show extensions to this model. As we mention in the Discussion, our results do not depend on the specific choice of nonlinearity (Figure 2-figure supplement 2). We have also considered the possibility that the stochastic nature of granule cell spikes could impact our measures of coding level. In Figure 7-figure supplement 1 we test the robustness of our main conclusion using a spiking model where we model granule cell spikes with Poisson statistics. When measuring coding level in a population of spiking neurons, a key question is at what time window the Purkinje cell integrates spikes. For several choices of integration time windows, we show that dense coding remains optimal for learning smooth tasks. However, we agree with the Reviewer that there are other biological details our model does not address. For example, our spiking model does not capture some of the properties the Saarinen et al. (2008) model captures, including random sub-threshold oscillations and clusters of spikes. Modeling biophysical phenomena at this scale is beyond the scope of our study. We have added this reference to the relevant section of the Discussion:

"We also note that coding level is most easily defined when neurons are modeled as rate, rather than spiking units. To investigate the consistency of our results under a spiking code, we implemented a model in which granule cell spiking exhibits Poisson variability and quantify coding level as the fraction of neurons that have nonzero spike counts (Figure 7-figure supplement 1; Figure 7C). In general, increased spike count leads to improved performance as noise associated with spiking variability is reduced. Granule cells have been shown to exhibit reliable burst responses to mossy fiber stimulation (Chadderton et al., 2004), motivating models using deterministic responses or sub-Poisson spiking variability. However, further work is needed to quantitatively compare variability in model and experiment and to account for more complex biophysical properties of granule cells (Saarinen et al., 2008)."

A second concern the Reviewer raises is our implementation of Golgi cell inhibition as a homogeneous rather than heterogeneous input onto granule cells. In simplified models, adding heterogeneous inhibition does not dramatically change the qualitative properties of the expansion layer representation, in particular the dimensionality of the representation (Billings et al., 2014, Cayco-Gajic et al., 2017, Litwin-Kumar et al., 2017). We have added a section about inhibition to our Discussion:

"We also have not explicitly modeled inhibitory input provided by Golgi cells, instead assuming such input can be modeled as a change in effective threshold, as in previous studies (Billings et al., 2014; Cayco-Gajic et al., 2017; Litwin-Kumar et al., 2017). This is appropriate when considering the dimension of the granule cell representation (Litwin-Kumar et al., 2017), but more work is needed to extend our model to the case of heterogeneous inhibition."

Regarding the mossy fiber inputs, as we state in response to paragraph 3, we agree with the Reviewer that the random and uncorrelated mossy fiber activity that has been used in previous studies is an unrealistic idealization of in vivo neural activity. One of the motivations for our model was to relax this assumption and examine the consequences: we introduce correlations in the mossy fiber activity by projecting low-dimensional patterns into the mossy fiber layer (Figure 1B):

"A typical assumption in computational theories of the cerebellar cortex is that inputs are randomly distributed in a high-dimensional space (Marr, 1969; Albus, 1971; Brunel et al., 2004; Babadi and Sompolinsky, 2014; Billings et al., 2014; Litwin-Kumar et al., 2017). While this may be a reasonable simplification in some cases, many tasks, including cerebellumdependent tasks, are likely best-described as being encoded by a low-dimensional set of variables. For example, the cerebellum is often hypothesized to learn a forward model for motor control (Wolpert et al., 1998), which uses sensory input and motor efference to predict an effector’s future state. Mossy fiber activity recorded in monkeys correlates with position and velocity during natural movement (van Kan et al., 1993). Sources of motor efference copies include motor cortex, whose population activity lies on a low-dimensional manifold (Wagner et al., 2019; Huang et al., 2013; Churchland et al., 2010; Yu et al., 2009). We begin by modeling the low dimensionality of inputs and later consider more specific tasks.

We therefore assume that the inputs to our model lie on a D-dimensional subspace embedded in the N-dimensional input space, where D is typically much smaller than N (Figure 1B). We refer to this subspace as the “task subspace” (Figure 1C)."

The Reviewer also mentions the learning rule at granule cell to Purkinje cell synapses. We agree that considering online, climbing-fiber-dependent learning is an important generalization. We therefore added a new supplemental figure investigating whether we would still see a difference in optimal coding levels across tasks if online learning were used instead of the least squares solution (Figure 7-figure supplement 2). Indeed, we observed a similar task dependence as we saw in Figure 2F. We have added a new paragraph in the Discussion under Assumptions and Extensions describing our rationale and approach in detail:

"For the Purkinje cells, our model assumes that their responses to granule cell input can be modeled as an optimal linear readout. Our model therefore provides an upper bound to linear readout performance, a standard benchmark for the quality of a neural representation that does not require assumptions on the nature of climbing fiber-mediated plasticity, which is still debated. Electrophysiological studies have argued in favor of a linear approximation (Brunel et al., 2004). To improve the biological applicability of our model, we implemented an online climbing fiber-mediated learning rule and found that optimal coding levels are still task-dependent (Figure 7-figure supplement 2). We also note that although we model several timing-dependent tasks (Figure 7), our learning rule does not exploit temporal information, and we assume that temporal dynamics of granule cell responses are largely inherited from mossy fibers. Integrating temporal information into our model is an interesting direction for future investigation."

Finally, regarding the function of the Purkinje cell, our model defines a learning task as a mapping from inputs to target activity in the Purkinje cell and is thus agnostic to the cell’s downstream effects. We clarify this point when introducing the definition of a learning task:

"In our model, a learning task is defined by a mapping from task variables x to an output f(x), representing a target change in activity of a readout neuron, for example a Purkinje cell. The limited scope of this definition implies our results should not strongly depend on the influence of the readout neuron on downstream circuits."

8) The problem of these, in my impression, generic, arbitrary settings of the neurons and the network in the model becomes obvious here: ‘In contrast to the dense activity in cerebellar granule cells, odor responses in Kenyon cells, the analogs of granule cells in the Drosophila mushroom body, are sparse...’ How can this system be interpreted as an analogy to granule cells in the mammalian cerebellum when the model does not address the specifics lined up above? I.e. the ‘inductive bias’ that the authors speak of, defined as ‘the tendency of a network toward learning particular types of input-output mappings’, would be highly dependent on the specifics of the network model.

We agree with the Reviewer that our model makes several simplifying assumptions for mathematical tractability. However, we note that our study is not the first to draw analogies between cerebellum-like systems, including the mushroom body (Bell et al., 2008; Farris, 2011). All the systems we study feature a sparsely connected, expanded granule-like layer that sends parallel fiber axons onto densely connected downstream neurons known to exhibit powerful synaptic plasticity, thus motivating the key architectural assumptions of our model. We have constrained anatomical parameters of the model using data as available (Table 1). However, we agree with the Reviewer that when making comparisons across species there is always a possibility that differences are due to physiological mechanisms we have not fully understood or captured with a model. As such, we can only present a hypothesis for these differences. We have modified our Discussion section on this topic to clearly state this.

"Our results predict that qualitative differences in the coding levels of cerebellum-like systems, across brain regions or across species, reflect an optimization to distinct tasks (Figure 7). However, it is also possible that differences in coding level arise from other physiological differences between systems."

9) More detailed comments: Abstract: ‘In these models [Marr-Albus], granule cells form a sparse, combinatorial encoding of diverse sensorimotor inputs. Such sparse representations are optimal for learning to discriminate random stimuli.’ Yes, I would agree with the first part, but I contest the second part of this statement. I think what is true for sparse coding is that the learning of random stimuli will be faster, as in a perceptron, but not necessarily better. As the sparsification essentially removes information, it could be argued that the quality of the learning is poorer. So from that perspective, it is not optimal. The authors need to specify from what perspective they consider sparse representations optimal for learning.

This is an important point that we would like to clarify. It is not the case that sparse coding simply speeds up learning. In our study and many related works (Barak et al. 2013; Babadi and Sompolinsky 2014; Litwin-Kumar et al. 2017), learning performance is measured based on the generalization ability of the network – the ability to predict correct labels for previously unseen inputs. As our study and previous studies show, sparse codes are optimal in the sense that they minimize generalization error, independent of any effect on learning speed. To communicate this more effectively, we have added the following sentence to the first paragraph of the Introduction:

"Sparsity affects both learning speed (Cayco-Gajic et al., 2017), and generalization, the ability to predict correct labels for previously unseen inputs (Barak et al., 2013; Babadi and Sompolinsky, 2014; Litwin-Kumar et al., 2017)."

10) Introduction: ‘Indeed, several recent studies have reported dense activity in cerebellar granule cells in response to sensory stimulation or during motor control tasks (Knogler et al., 2017; Wagner et al., 2017; Giovannucci et al., 2017; Badura and De Zeeuw, 2017; Wagner et al., 2019), at odds with classic theories (Marr, 1969; Albus, 1971).’ In fact, this was precisely the issue that was addressed already by Jo¨rntell and Ekerot (2006) Journal of Neuroscience. The conclusion was that these actual recordings of granule cells in vivo provided essentially no support for the assumptions in the Marr-Albus theories.

In our reading, the main finding of J¨orntell and Ekerot (2006) is that individual granule cells are activated by mossy fibers with overlapping receptive fields driven by a single type of somatosensory input. However, there is also evidence of nonlinear mixed selectivity in granule cells in support of the re-coding hypothesis (Huang et al., 2013; Ishikawa et al., 2015). Jo¨rntell and Ekerot (2006) also suggest that the granule cell layer shares similar topographic organization as mossy fibers, organized into microzones. The existence of topographic organization does not invalidate Marr-Albus theories. As we have suggested earlier, a local combinatorial expansion can coexist with a global topographic organization.

We have described these considerations in the Assumptions and Extensions portion of the Discussion:

"Another key assumption concerning the granule cells is that they sample mossy fiber inputs randomly, as is typically assumed in Marr-Albus models (Marr, 1969; Albus, 1971; LitwinKumar et al., 2017; Cayco-Gajic et al., 2017). Other studies instead argue that granule cells sample from mossy fibers with highly similar receptive fields (Garwicz et al., 1998; Brown and Bower, 2001; J¨orntell and Ekerot, 2006) defined by the tuning of mossy fiber and climbing fiber inputs to cerebellar microzones (Apps et al., 2018). This has led to an alternative hypothesis that granule cells serve to relay similarly tuned mossy fiber inputs and enhance their signal-to-noise ratio (Jo¨rntell and Ekerot, 2006; Gilbert and Chris Miall, 2022) rather than to re-encode inputs. Another hypothesis is that granule cells enable Purkinje cells to learn piece-wise linear approximations of nonlinear functions (Spanne and J¨orntell, 2013). However, several recent studies support the existence of heterogeneous connectivity and selectivity of granule cells to multiple distinct inputs at the local scale (Huang et al., 2013; Ishikawa et al., 2015). Furthermore, the deviation of the predicted dimension in models constrained by electron-microscopy data as compared to randomly wired models is modest (Nguyen et al., 2022). Thus, topographically organized connectivity at the macroscopic scale may coexist with disordered connectivity at the local scale, allowing granule cells presynaptic to an individual Purkinje cell to sample heterogeneous combinations of the subset of sensorimotor signals relevant to the tasks that Purkinje cell participates in. Finally, we note that the optimality of dense codes for learning slowly varying tasks in our theory suggests that observations of a lack of mixing (J¨orntell and Ekerot, 2002) for such tasks are compatible with Marr-Albus models, as in this case nonlinear mixing is not required."

We have also included the Jo¨rntell and Ekerot (2006) study as a citation in the Introduction:

"Indeed, several recent studies have reported dense activity in cerebellar granule cells in response to sensory stimulation or during motor control tasks (Jo¨rntell and Ekerot, 2006; Knogler et al., 2017; Wagner et al., 2017; Giovannucci et al., 2017; Badura and De Zeeuw, 2017; Wagner et al., 2019), at odds with classic theories (Marr, 1969; Albus, 1971)."

11) Results: 1st para: There is no information about how the granule cells are modelled.

We agree that this should information should have been more readily available. We now more completely describe the model in the main text. Our model for granule cells can be found in Equation 1 in the Results section and also the Methods (Network Model):

"The activity of neurons in the expansion layer is given by: h = φ(Jeffx − θ), (2)

where φ is a rectified linear activation function φ(u) = max(u,0) applied element-wise. Our results also hold for other threshold-polynomial activation functions. The scalar threshold θ is shared across neurons and controls the coding level, which we denote by f, defined as the average fraction of neurons in the expansion layer that are active."

12) 2nd para: ‘A typical assumption in computational theories of the cerebellar cortex is that inputs are randomly distributed in a high-dimensional space.’ Yes, I agree, and this is in fact in conflict with the known topographical organization in the cerebellar cortex (see broader comment above). Mossy fiber inputs coding for closely related inputs are co-localized in the cerebellar cortex. I think for this model to be of interest from the point of view of the mammalian cerebellar cortex, it would need to pay more attention to this organizational feature.

As we discuss in our response to paragraphs 5 and 6, we see the random distribution assumption at the local scale (inputs presynaptic to a single Purkinje cell) as being compatible with topographic organization occurring at the microzone scale. Furthermore, as discussed earlier, we specifically model low-dimensional input as opposed to the random and high-dimensional inputs typically studied in prior models.

"A typical assumption in computational theories of the cerebellar cortex is that inputs are randomly distributed in a high-dimensional space (Marr, 1969; Albus, 1971; Brunel et al., 2004; Babadi and Sompolinsky, 2014; Billings et al., 2014; Litwin-Kumar et al., 2017). While this may be a reasonable simplification in some cases, many tasks, including cerebellumdependent tasks, are likely best-described as being encoded by a low-dimensional set of variables. For example, the cerebellum is often hypothesized to learn a forward model for motor control (Wolpert et al., 1998), which uses sensory input and motor efference to predict an effector’s future state. Mossy fiber activity recorded in monkeys correlates with position and velocity during natural movement (van Kan et al., 1993). Sources of motor efference copies include motor cortex, whose population activity lies on a low-dimensional manifold (Wagner et al., 2019; Huang et al., 2013; Churchland et al., 2010; Yu et al., 2009). We begin by modeling the low dimensionality of inputs and later consider more specific tasks. We therefore assume that the inputs to our model lie on a D-dimensional subspace embedded in the N-dimensional input space, where D is typically much smaller than N (Figure 1B). We refer to this subspace as the “task subspace” (Figure 1C)."

References

Albus, J.S. (1971). A theory of cerebellar function. Mathematical Biosciences 10, 25–61.

Apps, R., et al. (2018). Cerebellar Modules and Their Role as Operational Cerebellar Processing Units. Cerebellum 17, 654–682.

Babadi, B. and Sompolinsky, H. (2014). Sparseness and expansion in sensory representations. Neuron 83, 1213–1226.

Badura, A. and De Zeeuw, C.I. (2017). Cerebellar granule cells: dense, rich and evolving representations. Current Biology 27, R415–R418.

Barak, O., Rigotti, M., and Fusi, S. (2013). The sparseness of mixed selectivity neurons controls the generalization–discrimination trade-off. Journal of Neuroscience 33, 3844– 3856.

Bell, C.C., Han, V., and Sawtell, N.B. (2008). Cerebellum-like structures and their implications for cerebellar function. Annual Review of Neuroscience 31, 1–24.

Billings, G., Piasini, E., Lo˝rincz, A., Nusser, Z., and Silver, R.A. (2014). Network structure within the cerebellar input layer enables lossless sparse encoding. Neuron 83, 960–974.

Bordelon, B., Canatar, A., and Pehlevan, C. (2020). Spectrum dependent learning curves in kernel regression and wide neural networks. International Conference on Machine Learning 1024–1034.

Brown, I.E. and Bower, J.M. (2001). Congruence of mossy fiber and climbing fiber tactile projections in the lateral hemispheres of the rat cerebellum. Journal of Comparative Neurology 429, 59–70.

Brunel, N., Hakim, V., Isope, P., Nadal, J.P., and Barbour, B. (2004). Optimal information storage and the distribution of synaptic weights: perceptron versus Purkinje cell. Neuron 43, 745–757.

Canatar, A., Bordelon, B., and Pehlevan, C. (2021). Spectral bias and task-model alignment explain generalization in kernel regression and infinitely wide neural networks. Nature Communications 12, 1–12.

Cayco-Gajic, N.A., Clopath, C., and Silver, R.A. (2017). Sparse synaptic connectivity is required for decorrelation and pattern separation in feedforward networks. Nature Communications 8, 1–11.

Chadderton, P., Margrie, T.W., and Ha¨usser, M. (2004). Integration of quanta in cerebellar granule cells during sensory processing. Nature 428, 856–860.

Churchland, M.M., et al. (2010). Stimulus onset quenches neural variability: a widespread cortical phenomenon. Nature Neuroscience 13, 369–378.

Farris, S.M. (2011). Are mushroom bodies cerebellum-like structures? Arthropod structure & development 40, 368–379.

Garwicz, M., Jorntell, H., and Ekerot, C.F. (1998). Cutaneous receptive fields and topography of mossy fibres and climbing fibres projecting to cat cerebellar C3 zone. The Journal of Physiology 512 ( Pt 1), 277–293.

Gilbert, M. and Chris Miall, R. (2022). How and Why the Cerebellum Recodes Input Signals: An Alternative to Machine Learning. The Neuroscientist 28, 206–221.

Giovannucci, A., et al. (2017). Cerebellar granule cells acquire a widespread predictive feedback signal during motor learning. Nature Neuroscience 20, 727–734.

Huang, C.C., et al. (2013). Convergence of pontine and proprioceptive streams onto multimodal cerebellar granule cells. eLife 2, e00400.

Ishikawa, T., Shimuta, M., and Ha¨usser, M. (2015). Multimodal sensory integration in single cerebellar granule cells in vivo. eLife 4, e12916.

Jacot, A., Gabriel, F., and Hongler, C. (2018). Neural tangent kernel: Convergence and generalization in neural networks. Advances in Neural Information Processing Systems 31.

Jo¨rntell, H. and Ekerot, C.F. (2002). Reciprocal Bidirectional Plasticity of Parallel Fiber Receptive Fields in Cerebellar Purkinje Cells and Their Afferent Interneurons. Neuron 34, 797–806.

Jorntell, H. and Ekerot, C.F. (2006). Properties of Somatosensory Synaptic Integration in Cerebellar Granule Cells In Vivo. Journal of Neuroscience 26, 11786–11797.

Knogler, L.D., Markov, D.A., Dragomir, E.I., Stih, V., and Portugues, R. (2017). Senso-ˇ rimotor representations in cerebellar granule cells in larval zebrafish are dense, spatially organized, and non-temporally patterned. Current Biology 27, 1288–1302.

Litwin-Kumar, A., Harris, K.D., Axel, R., Sompolinsky, H., and Abbott, L.F. (2017). Optimal degrees of synaptic connectivity. Neuron 93, 1153–1164. Marr, D. (1969). A theory of cerebellar cortex. Journal of Physiology 202, 437–470.

Nguyen, T.M., et al. (2022). Structured cerebellar connectivity supports resilient pattern separation. Nature 1–7.

Saarinen, A., Linne, M.L., and Yli-Harja, O. (2008). Stochastic Differential Equation Model for Cerebellar Granule Cell Excitability. PLOS Computational Biology 4, e1000004.

Simon, J.B., Dickens, M., and DeWeese, M.R. (2021). A theory of the inductive bias and generalization of kernel regression and wide neural networks. arXiv: 2110.03922.

Sollich, P. (1998). Learning curves for Gaussian processes. Advances in Neural Information Processing Systems 11.

Spanne, A. and Jo¨rntell, H. (2013). Processing of Multi-dimensional Sensorimotor Information in the Spinal and Cerebellar Neuronal Circuitry: A New Hypothesis. PLOS Computational Biology 9, e1002979.

Spanne, A. and Jo¨rntell, H. (2015). Questioning the role of sparse coding in the brain. Trends in Neurosciences 38, 417–427.

van Kan, P.L., Gibson, A.R., and Houk, J.C. (1993). Movement-related inputs to intermediate cerebellum of the monkey. Journal of Neurophysiology 69, 74–94.

Wagner, M.J., Kim, T.H., Savall, J., Schnitzer, M.J., and Luo, L. (2017). Cerebellar granule cells encode the expectation of reward. Nature 544, 96–100.

Wagner, M.J., et al. (2019). Shared cortex-cerebellum dynamics in the execution and learning of a motor task. Cell 177, 669–682.e24.

Wolpert, D.M., Miall, R.C., and Kawato, M. (1998). Internal models in the cerebellum. Trends in Cognitive Sciences 2, 338–347.

Yu, B.M., et al. (2009). Gaussian-process factor analysis for low-dimensional single-trial analysis of neural population activity. Journal of Neurophysiology 102, 614–635.

如果是windows wsl ,命令前需要添加 sudo

This work has been peer reviewed in GigaScience (see https://doi.org/10.1093/gigascience/giad035 ), which carries out open, named peer-review. These reviews are published under a CC-BY 4.0 license and were as follows:

Reviewer Chris Armit

This Technical Note paper describes "FriendlyClearMap: An optimized toolkit for mouse brain mapping and analysis".

Whereas the core concept of a data analysis tool to assist in spatial mapping of cleared mouse tissues is perfectly reasonable, there are multiple issues with the documentation that renders this toolkit very difficult to use. I detail below some of the issues I have encountered.

1. GitHub repositoryThe installation instructions are missing from the following GitHub repository: https://github.com/MoritzNegwer/FriendlyClearMap-scriptsThe closest reference I could find to installation instructions is the following: "Please see the Appendices 1-3 of our <X_upcoming> publication for detailed instructions on how to use the pipelines. <X_protocols.io goes here>"Step-bystep installation instructions should be included in the GitHub repository. In addition, the authors should add the protocols.io links to their GitHub repository.

2. Protocols.ioThe installation instructions are missing from the following protocols.io links:Run Clearmap 1 docker dx.doi.org/10.17504/protocols.io.eq2lynnkrvx9/v1Run Clearmap 2 docker dx.doi.org/10.17504/protocols.io.yxmvmn9pbg3p/v1Both of these protocols include the following instruction:* "Then, download the docker container from our repository: XXX docker container goes here"In the documentation, the authors need to unambiguously refer to the specific Docker container that a user needs to install for each software tool.

3. Test Data I could not find the test data in the form of image stacks that would be needed to test the FriendlyClearMap protocols. Figure 1 refers to 16-bit TIFF image stacks, and I presume these to be the input data that is needed for the image analysis pipelines described in the manuscript. The authors should provide details of the test imaging dataset, including links if necessary to where the image stacks data can be downloaded, in the 'Data Availability' section of the manuscript.

4. Platform / Operating SystemsIn the 'Data Availability' section of the manuscript, the authors specify that the Operating Systems are "platform-independent". However, the protocols.io documents lists a set of requirements for Windows and LINUX, but not for MacOS. The authors should provide installation instructions and system requirements for MacOS.I reject this manuscript on the grounds that, due to lack of appropriate documentation and installation instructions, the software tool is too difficult to use and therefore has extremely low reuse potential.

Yet another example of why it's dumb for Microsoft to lock Community threads. This is in the Bing search results as the top article for my issue with 1,911 views. Since 2011 though, there have been new developments! The new Media Player app in Windows 10 natively supports Zune playlist files! Since the thread is locked, I can't put this news in a place where others following my same search path will find it.

Guess that's why it makes sense to use Hypothes.is 🤷‍♂️

So with the Apple lisa being the first I can see a few things that are simlar to today. Icons, little images that one can select with describing text. Windows with closing buttons and other stuff like that. I think it probably flopped because of the price. Lisa 1983.

CONSIDERATION

ISSUE_REPRODUCTION

PROBLEM_DESCRIPTION

Author Response

Reviewer #2 (Public Review):

1) The authors in reality do not analyze oscillations themselves in this manuscript but only the power of signals filtered at determined frequency bands. This is particularly misleading when the authors talk about "spindles". Spindles are classically defined as a thalamico-cortical phenomenon, not recorded from hippocampus LFPs. Thus, the fact that you filter the signal in the same frequency range matching cortical spindles does not mean you are analyzing spindles. The terminology, therefore, is misleading. I would recommend the authors to change spindles to "beta", which at least has been reported in the hippocampus, although in very particular behavioral circumstances. However, one must note that the presence of power in such bands does not guarantee one is recording from these oscillations. For example, the "fast gamma" band might be related to what is defined as fast gamma nested in theta, but it might also be related to ripples in sleep recordings. The increase of "spindle" power in sleep here is probably related to 1/f components arising from the large irregular activity of slow wave sleep local field potentials. The authors should avoid these conceptual confusions in the manuscript, or show that these band power time courses are in fact matching the oscillations they refer to (for example, their spindle band is in fact reflecting increased spindle occurrence).

We thank the reviewer for allowing us to clarify this subject. We completely agree with concerns raised in the comments. To avoid any confusion, we have replaced throughout the manuscript the word ‘spindle’ with ‘beta’.

2) The shuffling procedure to control for the occupancy difference between awake and sleep does not seem to be sufficient. From what I understand, this shuffling is not controlling for the autocorrelation of each band which would be the main source of bias to be accounted for in this instance. Thus, time shifts for each band would be more appropriate. Further, the controls for trial durations should be created using consecutive windows. If you randomly sample sleep bins from distant time points you are not effectively controlling for the difference in duration between trial types. Finally, it is not clear from the text if the UMAP is recomputed for each duration-matched control. This would be a rigorous control as it would remove the potential bias arising from the unbalance between awake and sleep data points, which could bias the subspace to be more detailed for the LFP sleep features. It is very likely the results will hold after these controls, given it is not surprising that sleep is a more diverse state than awake, but it would be good practice to have more rigorous controls to formalize these conclusions.

We are grateful to the reviewer for suggesting alternative analysis. We have used this direction, to create surrogate datasets obtained by time shifting each band and obtained their respective UMAP projections (see modified Figure 2D). Additionally, as suggested, for duration-matched controls, we have selected consecutive windows, rather than random points (Figure 2 – figure supplement 1C). UMAP projections were obtained for each duration-matched control and occupancy was computed. The text in the method section has been modified to indicate the analysis. As expected, the results were identical.

3) Lots of the observations made from the state space approach presented in this manuscript lack any physiological interpretation. For example, Figure 4F suggests a shift in the state space from Sleep1 to Sleep2. The authors comment there is a change in density but they do not make an effort to explain what the change means in terms of brain dynamics. It seems that the spectral patterns are shifting away from the Delta X Spindle region (concluding this by looking at Fig4B) which could be potentially interesting if analyzed in depth. What is the state space revealing about the brain here? It would be important to interpret the changes revealed by this method otherwise what are we learning about the brain from these analyses? This is similar to the results presented in Figure 5, which are merely descriptions of what is seen in the correlation matrix space. It seems potentially interesting that non-REM seems to be split into two clusters in the UMAP space. What does it mean for REM that delta band power in pyramidal and lm layers is anti-correlated to the power within the mid to fast gamma range? What do the transition probabilities shown in Figures 6B and C suggest about hippocampal functioning? The authors just state there are "changes" but they don't characterize these systematically in terms of biology. Overall, the abstract multivariate representation of the neural data shown here could potentially reveal novel dynamics across the awake-sleep cycle, but in the current form of this manuscript, the observations never leave the abstract level.

We thank the reviewer for allowing us to clarify this aspect of the manuscript. We have now edited the main text to include considerations on the biological relevance of the findings of Figure 4, 5 and 6.

Additions to figure 4: In particular, non-REM states in sleep2 tended to concentrate in a region of increased power in the delta and beta bands, which could be the results of increased interactions with cortical activity modulated in the same range. It is also likely that such effect was induced by the exposure to relevant behavioral experience. In fact, changes in density of individual oscillations after learning have been reported using traditional analytical methods and are thought to support memory consolidation (Bakker et al., 2015; Eschenko et al., 2008, 2006). Nevertheless, while traditional methods provide information about individual components, the novel approach used here provides additional information about the combinatorial shift in the dynamics of network oscillations after learning or exploration. Thus, it provides the basis for identifying how coordinated activity among different oscillations supports memory consolidation processes, as those occurring during non-REM sleep after exploration, which cannot be elucidated using traditional analytical methods.

Additions to figure 5: Gamma segregation and delta decoupling offer a picture of hippocampal REM sleep as being more akin to awake locomotion (with the major difference of a stronger medium gamma presence) while also suggesting a substantial independence from cortical slow oscillations. On the other hand, the across-scale coherence of non-REM sleep is consistent with this sleep stage being dominated by brain-wide collective fluctuations engaging oscillations at every range. Distinct cross frequency coupling among various individual pairs of oscillations such as theta-gamma, delta-gamma etc., have been already reported (Bandarabadi et al., 2019; Clemens et al., 2009; Hammer et al., 2021; Scheffzük et al., 2011). However, computing cross frequency coupling on the state space provides the additional information on how multiple oscillations, obtained from distinct CA1 hippocampal layers (stratum pyramidale, stratum radiatum and stratum lacunosum moleculare), are coupled with each other during distinct states of sleep and wakefulness. Furthermore, projecting the correlation matrices on 2D plane, provides a compact tool that allows to visualize the cross-frequency interactions among various hippocampal oscillations. Altogether, this approach reveals the complex nature of coupling dynamics occurring in hippocampus during distinct behavioral states

Additions to Figure 6: We found that transitions occurring from REM-to-REM sleep and non-REM-to-non-REM sleep (intra-state transitions) are more vulnerable to plasticity after exploration as compared to inter-state transitions (such as non-REM to REM, REM-to-intermediate etc.) (Fig 6E, F). These changes in intra-state transitions were observed to be beyond randomness (Fig S9 E, F) indicating a specificity in plastic changes in state transitions after exploration. In particular, while the average REM period duration is unaltered after exploration (Fig 4G), REM temporal structure is reorganized. In fact, increased probability of REM to REM transitions indicates a significant prolongation of REM bout duration. Similarly, the increase in non-REM to non-REM transition probability reflects an increased duration of non-REM bouts. Therefore, environment exploration was accompanied by an increased separation between REM and non-REM periods, possibly as a response to increased computational demands. More in general, the network state space allows to characterize the state transitions in hippocampus and how they are affected by novel experience or learning. By observing the state transition patterns, this analytical framework allows to detect and identify state-specific changes in the hippocampal oscillatory dynamics, beyond the possibilities offered by more traditional univariate and bivariate methods. We next investigated how fast the network flows on the state space and assessed whether the speed is uniform, or it exhibits specific region-dependent characteristics.

Reviewer #3 (Public Review):

1) My primary concern is to provide clear evidence that this approach will provide key insights of high physiological significance, especially for readers who may think the traditional approaches are advantageous (for example due to their simplicity). I think the authors' findings of distinct sleep state signatures or altered organization of the NLG3-KO mouse could serve this purpose. However, right now the physiological significance of these results is unclear. For example, do these sleep state signatures predict later behavior performance, or is altered organization related to other functional impairments in the disease model? Do neurons with distinct sleep state signatures form distinct ensembles and code for related information?

We are thankful to the reviewer for raising a very interesting line of questioning regarding sleep signatures and distinct ensemble. In this study, we show that sleep state signatures can predict how individual cells may participate in information processing during open field exploration. However, further analysis exploring the recruitment of neuronal ensembles are in preparation for another manuscript and is beyond the scope of this article.

We have further modified the description of the results (as also suggested by other reviewers) to highlight the key advantages of this approach over traditional methods.

Regarding functional impairment: as described in the manuscript, the altered organization in animal model of autism could possibly due to alterations in cellular and synaptic mechanisms as those described in previous reports (Modi et al 2019, Foldy et al 2013)

2) For cells with different mean firing rates during exploration: is that because they are putative fast-spiking interneurons and pyramidal cells? From the reported mean firing rates, I think some of these cells are interneurons. Since mean firing rates are well known to vary with cell type, this should be addressed. For example, the sleep state signatures may be distinct for different putative pyramidal cells and interneurons. This would be somewhat expected considering prior work that has shown different cell types have different oscillatory coupling characteristics. I think it would be more interesting to determine if pyramidal cells had distinct sleep state signatures and, if so, whether pyramidal cells from the same sleep state signature have similar properties like they code for similar things or commonly fire together in an ensemble ms the number of cells in Fig. 8 may be limited for this analysis. The authors could use the hc-11 data in addition, which was also tested in this work.

We thank the reviewer for suggesting this additional analysis to better describe the data. To this end, we have added an additional Figure in supplementary data (analysis of hc11 dataset: Figure Figure 8 – figure supplement 3), to demonstrate that interneurons and pyramidal cells have distinct sleep signatures. These findings are in agreement with dataset presented in Figure 8D, E.

As shown in the manuscript, the spatial firing (sparsity) has large variability for cells having similar network signatures (Fig 8E). Thus, additional parameters beside oscillations may be involved in cells encoding. Different network state spaces are required to be explored in future studies to further understand this phenomenon in detail.

We agree that investigating neuronal ensembles and state space are an interesting direction to follow. In another study (in preparation) which are investigating in detail the recruitment of neuronal ensemble by oscillatory state space. Thus, those findings are beyond the scope of this introductory article.

3) Example traces are needed to show how LFPs change over the state-space. Example traces should be included for key parts of the state-space in Figures 2 and 3.

We thank the reviewer for this key insight on data representation. Example traces of how LFP varies on the state space have been added (see Figure 4 – figure supplement 1).

4) What is the primary rationale for 200ms time bins? Is this time scale sufficient to capture the slow dynamics of delta rhythm (1-5Hz) with a maximum of 1s duration?

Time scale of binning depends on the scale of investigation. We also replicated the results with different time bins (such as 50 ms and 1 seconds) and the results are identical. For delta rhythms, with 200 ms time bins, the dynamics will be captured across multiple bins. Additionally, the binned power time series are also smoothed before obtaining projections.

5) Since oscillatory frequency and power are highly associated with running speed, how does speed vary over the state space. Is the relationship between speed and state-space similar to the results of previous studies for theta (Slawinska and Kasicki, Brain Res 1998; Maurer et al, Hippocampus 2005) and gamma oscillations (Ahmed and Mehta J. Neurosci 2012; Kemere et al PLOS ONE 2013), or does it provide novel insights?

We thank the reviewer for highlighting this crucial link between oscillation and locomotion. While various articles have focused on individual oscillations, the combinatorial effects of multiple oscillations from multiple brain areas in regulating the speed of the animal during exploration is definitely worth exploring with this novel approach. These set of results will be introduced in another study, currently in preparation.

6) The separation of 9 states (Fig. 6ABC) seems arbitrary, where state 1 (bin 1) is never visited. I suggest plotting the density distribution of the data in Fig. 2A or Fig. 6A to better determine how many states are there within the state space. For example, five peaks in such a density plot might suggest five states. Alternately, clustering methods could be useful to determine how the number of states.

We thank the reviewer for this this useful suggestion. We agree that additional clustering methods can be used to identify non-canonical sleep states. These are currently being explored in our lab and will be part of future studies. As for this dataset, the density plots are available in figure 4E, which determines how many states are in each part of the state space.

7) The results in Fig. 4G are very interesting and suggest more variation of sub-states during non REM periods in sleep1 than in sleep2. What might explain this difference? Was it associated with more frequent ripple events occurring in sleep2?

The reviewer is right in looking for the source of the decreased of state variability in sleep2. Considering the distribution of relative frequency power in the state space, the higher concentration in sleep 2 corresponds to higher content in the slower delta and spindle frequency bands, rather than the higher frequencies of SWRs. This result can be interpreted in the light of enhanced cortical activity (which is known to heavily recruit those bands) and possibly of enhanced cortical-hippocampal communication following relevant behavioral experience. In fact, it is also necessary to mention that with our recording setup we cannot rule out the effects of volume conductance completely, and thus we cannot exclude that the increase in the delta and spindle bands in the hippocampus were a spurious effect of purely cortical frequency modulations.

8) The state transition results in Fig. 6 are confusing because they include two fundamentally different timescales: fast transitions between oscillatory states and slow dynamics of sleep states. I recommend clarifying the description in the results and the figure caption. Furthermore, how can an animal transition between the same sleep state (Fig. 6EF)? Would they both be in a single sleep state?

The transitions capture the fast oscillatory scales (as they are investigated over a timeframe of 1 second). The sleep stages (REM, non-REM etc.) are used as labels from which the states originate on the state space. This allows us to characterize fast oscillatory dynamics in various sleep stages.

Regarding same state transition: An increase in same state transition probability corresponds to increase in prolongation of that particular state, thereby altering the temporal structure of a given sleep state.

Reviewer #2 (Public Review):

Modi and colleagues describe a multivariate framework to analyze local field potentials, which is specifically applied to CA1 data in this work. Multivariate approaches are welcome in the field and the effort of the authors should be appreciated. However, I found the analyses presented here are too superficial and do not seem to bring new insights into hippocampal dynamics. Further, some surrogate methods used are not necessarily controlling for confounding variables. These concerns are further detailed below.

1. The authors in reality do not analyze oscillations themselves in this manuscript but only the power of signals filtered at determined frequency bands. This is particularly misleading when the authors talk about "spindles". Spindles are classically defined as a thalamico-cortical phenomenon, not recorded from hippocampus LFPs. Thus, the fact that you filter the signal in the same frequency range matching cortical spindles does not mean you are analyzing spindles. The terminology, therefore, is misleading. I would recommend the authors to change spindles to "beta", which at least has been reported in the hippocampus, although in very particular behavioral circumstances. However, one must note that the presence of power in such bands does not guarantee one is recording from these oscillations. For example, the "fast gamma" band might be related to what is defined as fast gamma nested in theta, but it might also be related to ripples in sleep recordings. The increase of "spindle" power in sleep here is probably related to 1/f components arising from the large irregular activity of slow wave sleep local field potentials. The authors should avoid these conceptual confusions in the manuscript, or show that these band power time courses are in fact matching the oscillations they refer to (for example, their spindle band is in fact reflecting increased spindle occurrence).

2. The shuffling procedure to control for the occupancy difference between awake and sleep does not seem to be sufficient. From what I understand, this shuffling is not controlling for the autocorrelation of each band which would be the main source of bias to be accounted for in this instance. Thus, time shifts for each band would be more appropriate. Further, the controls for trial durations should be created using consecutive windows. If you randomly sample sleep bins from distant time points you are not effectively controlling for the difference in duration between trial types. Finally, it is not clear from the text if the UMAP is recomputed for each duration-matched control. This would be a rigorous control as it would remove the potential bias arising from the unbalance between awake and sleep data points, which could bias the subspace to be more detailed for the LFP sleep features. It is very likely the results will hold after these controls, given it is not surprising that sleep is a more diverse state than awake, but it would be good practice to have more rigorous controls to formalize these conclusions.

3. Lots of the observations made from the state space approach presented in this manuscript lack any physiological interpretation. For example, Figure 4F suggests a shift in the state space from Sleep1 to Sleep2. The authors comment there is a change in density but they do not make an effort to explain what the change means in terms of brain dynamics. It seems that the spectral patterns are shifting away from the Delta X Spindle region (concluding this by looking at Fig4B) which could be potentially interesting if analyzed in depth. What is the state space revealing about the brain here? It would be important to interpret the changes revealed by this method otherwise what are we learning about the brain from these analyses? This is similar to the results presented in Figure 5, which are merely descriptions of what is seen in the correlation matrix space. It seems potentially interesting that non-REM seems to be split into two clusters in the UMAP space. What does it mean for REM that delta band power in pyramidal and lm layers is anti-correlated to the power within the mid to fast gamma range? What do the transition probabilities shown in Figures 6B and C suggest about hippocampal functioning? The authors just state there are "changes" but they don't characterize these systematically in terms of biology. Overall, the abstract multivariate representation of the neural data shown here could potentially reveal novel dynamics across the awake-sleep cycle, but in the current form of this manuscript, the observations never leave the abstract level.

[Window][Tab]

It's not exclaimation, it's default beep.

Future work is needed.

delete

When will it come?

That sounds like a terrible sight to witness.

https://getupnote.com/

Suggested by Cato Minor<br /> Mac, Windows, Linux, iOS, Android

Reviewer #3 (Public Review):

In their manuscript, Schneider et al. aim to develop voyAGEr, a web-based tool that enables the exploration of gene expression changes over age in a tissue- and sex-specific manner. The authors achieved this goal by calculating the significance of gene expression alterations within a sliding window, using their unique algorithm, Shifting Age Range Pipeline for Linear Modelling (ShARP-LM), as well as tissue-level summaries that calculated the significance of the proportion of differentially expressed genes by the windows and calculated enrichments of pathways for showing biological relevance. Furthermore, the authors examined the enrichment of cell types, pathways, and diseases by defining the co-expressed gene modules in four selected tissues. The voyAGEr was developed as a discovery tool, providing researchers with easy access to the vast amount of transcriptome data from the GTEx project. Overall, the research design is unique and well-performed, with interesting results that provide useful resources for the field of human genetics of aging. I have a few questions and comments, which I hope the authors can address.

1. In the gene-centric analyses section of the result, to improve this manuscript and database, linear regression tests accounting for the entire range of age should be added. The authors' algorithm, ShARP-LM, tests locally within a 16-year window which makes it has lower power than the linear regression test with the whole ages. I suspect that the power reduction is strongly affected in the younger age range since a larger number of GTEx donors are enriched in old age. By adding the results from the lm tests, readers would gain more insight and evidence into how significantly their interest genes change with age.<br /> 2. In line with the ShARP-LM test results, it is not clear which criterion was used to define the significant genes and the following enrichment analyses. I assume that the criterion is P < 0.05, but it should be clearly noted. Additionally, the authors should apply adjusted p-values for multiple-test correction. The ideal criterion is an adjusted P < 0.05. However, if none or only a handful of genes were found to be significant, the authors could relax the criteria, such as using a regular P < 0.01 or 0.05.<br /> 3. In the gene-centric analyses section, authors should provide a full list of donor conditions and a summary table of conditions as supplementary.<br /> 4. The tissue-specific assessment section has poor sub-titles. Every title has to contain information.<br /> 5. I have an issue understanding the meaning of NES from GSEA in the tissue-specific assessment section. The authors performed GSEA for the DEGs against the background genes ordered by t-statistics (from positive to negative) calculated from the linear model. I understand the p-value was two-tailed, which means that both positive and negative NES are meaningful as they represent up-regulated expression direction (positive coefficient) and down-regulated expression direction (negative coefficient) with age, respectively, within a window. However, in the GSEA section of Methods, authors were not fully elaborate on this directionality but stated, "The NES for each pathway was used in subsequent analyses as a metric of its over- or down-representation in the Peak". The authors should clearly elaborate on how to interpret the NES from their results.<br /> 6. In the Modules of co-expressed genes section, the authors did not explain how or why they selected the four tissues: brain, skeletal muscle, heart (left ventricle), and whole blood. This should be elaborated on.<br /> 7. In the modules of the co-expressed genes section, the authors did not provide an explanation of the "diseases-manual" sub-tab of the "Pathway" tab of the voyAGEr tool. It would be helpful for readers to understand how the candidate disease list was prepared and what the results represent.

This is Dickinson's way of expressing her belief that there is no bright light once you die.

When she mentioned “the windows” she was referring to her eyes(The eyes are commonly known as “the windows of the soul”) as they were shutting down and she lost contact with the outside world.

In these lines, the buzzing sound of the fly is described as a "Blue—uncertain stumbling Buzz." Here, Dickinson compares the sound of the fly to a stumbling movement, emphasizing its presence and intrusion within the solemn atmosphere of death. The use of the word "uncertain" further adds to the unsettling nature of the scene, suggesting a lack of clarity and a disturbance in the narrator's perception.

Ruskin: Praeterita III. 1. "But now that I had a missal of my own and could touch its leaves and tum, and even here and there understand the Latin of it, no girl of seven years old with a new doll is prouder or happier; but the feeling was something between the girl's with the doll and Aladdin's fn a new spirit-slave to build palaces for him with jewel windows. For truly a well illuminated missal is a fairy cathedral full of painted windows, bound together to carry in one's pocket, with the music and the blessing of all its prayers besides. And then followed, of course, the discovery that all beautiful prayers were Catholic, all wise interpretations of the Bible, Catholic; and every manner of Protestant written service whatsoever, either insolently altered corruptions, or washed-out and ground-down rags and debris of the great Catholic collects, litanies and songs of praise." A comparison of the old Hebrew prayer book with its modem substitute, tempts me to cite this opinion of Ruskin with much approval.

Praeterita : outlines of scenes and thoughts, perhaps worthy of memory in my past life / by John Ruskin.

Are you using the same aggregation windows and same number of repeats for the gait score and the individual features?

This is an ordinary bug and not worthy of note in a document about the relative strengths and weaknesses of WebDAV (or any other protocol); it's not an inherent consequence of WebDAV as designed...

Reviewer #1 (Public Review):

We conclude that this descriptive study has some strengths but additionally, we propose several ways in which to increase its potential impact and to strengthen some of the claims. This study describes the remodeling of Merkel cells and their innervating sensory axons in the skin. By using transgenic mouse lines in which these cells were genetically fluorescently labeled, the authors performed a series of analyses mostly focusing on the number and location of Merkel cells and the sensory axons that innervate them.

One of the major strengths of the study is the establishment of intravital imaging techniques to investigate the dynamic simultaneous behavior of Merkel cells and their innervation during homeostasis and hair regeneration. However, how the findings integrate into the existing knowledge of skin development is unfortunately only partially addressed.

To the best of our understanding, a few technical limitations of the study define its major weaknesses: First, Merkel cell loss is dramatic and it's unclear whether this reduction is part of a developmentally controlled reduction in cell number, or whether additional cells are expected to be integrated into the system. Longer windows of imaging might help here. Second, the depilation agent might be too aggressive and lead to cell death and thus better controls might be suggested. Similarly, ablating Merkal cells throughout development might cause developmental issues that might mask the proposed homeostasis analyses. A controlled adult specific ablation might be suggested. Finally, the TrkC based transgenic mouse is expected to be heterozygous - could that be an issue? Either better controls, or a textual addressing of this topic are advised.

All in all, we think this study has the potential to establish a high resolution description of Merkel cells - sensory axon dynamic interactions. We hope that the authors will be encouraged to improve the paper based on our comments, something that will likely improve its potential significance and impact.

Reviewer #1 (Public Review):

This manuscript reports the results of a clever chemogenetic manipulation study designed to probe how stimulation (excitation and inhibition) of D1-expressing spiny projection neurons in the striatum (direct pathway) influences hemodynamic characteristics of local and global regions in the brain in mice. Stimulating in the dorsal caudate, the authors found alteration of the local hemodynamic BOLD responses in adjacent areas of the striatum, regions of the thalamus known to project back to the striatum, and more unimodal cortical regions. The authors also observed a decrease in functional connectivity between several cortical regions and the striatum with direct pathway excitation, and the opposite effect with direct pathway inhibition. Put together the results begin to paint a picture of the macroscopic signatures of direct pathway stimulation (and inhibition) that could be used to help infer global BOLD patterns in task-related experiments.

Overall, this appears to be a timely study. The rise of papers using chemo/optogenetic methods to probe the underlying mechanisms of the BOLD response is crucial for the neuroimaging field to continue moving forward. The methods are, for the most part, rigorous and clear. There are, however, a few open issues that require addressing in order for readers to reach the same conclusion as the authors.

MAJOR CONCERNS

1. Classifier as an evaluation method.

The authors largely rely on a support vector machine (SVM) classifier to predict whether BOLD dynamics within atlas-defined regions reflect stimulation-on or stimulation-off windows. While in one way this is a conservative method for evaluating stimulation effects in the resting BOLD fluctuations, the authors largely report their findings as accuracies of the classifier. Figures 3-5 largely only report model accuracy effects, but we get no sense as to what exactly is happening to the BOLD dynamics in each region. The autocorrelation analysis (Fig. 6) somewhat tries to get at this, but only for a subset of regions and the results are largely unclear (see comment below). As a result, a key goal of the study is left largely unaddressed for the reader: i.e. how do intra-region BOLD dynamics change with direct pathway stimulation? The study needs more effort put into this descriptive level of analysis to complement the rigorous classifier analysis.

Also, the classifier method itself seems highly parameterized. The hctsa method returns 7702 features for each time series. It is unclear exactly how many were in the final set used to run the classification, but even if half of the features were removed, it would still make the classification problem highly overparameterized (e.g., 23 and 25 observations against thousands of features for the excitation and inhibition classifiers respectively). Assuming the authors used cross-validation correctly (which we need more information to support), the risk of inflated classification performance is mitigated. However, we need the details to be able to vet that the bias-variance tradeoff was resolved effectively in this model. In addition, it would be nice to know the features that loaded highly on the final model to resolve the questions about what changes in the local BOLD dynamics from excitation and inhibition of the direct pathway.

2. Local stimulation effects in the striatum

Figure 3 is quite confusing. The classifier is supposed to predict stimulation (excitation or inhibition) on vs. stimulation off (control) periods. This would predict a single number (balanced prediction accuracy) per striatal nuclei. Yet the heat maps shown in this figure show classification accuracies for both stimulation and control conditions. Where do the two numbers come from? Also, given the extremely limited short-range lateral connectivity in the striatum, why are the only stimulation effects observed not in the subnucleus being stimulated (Cpl,dm,cd), but for adjacent subnuclei (CPre, CPivmv) and *only* for excitation conditions? This lack of direct change in BOLD dynamics at the stimulated site seems important and largely ignored.

3. Autocorrelation findings.

The one attempt to characterize what happens in the intra-region BOLD dynamics is the autocorrelation analysis reported on page 11 and in Figure 6. However this analysis a) only focuses on the thalamic nuclei (not also the cortical and the single striatal site shown to exhibit stimulation effects) and b) only focuses on a few time series measures. Why this limited focus of both target (thalamic nuclei) and measure (first lag of the autocorrelation)? There are many measures for characterizing the temporal characteristics of autocorrelated series from the hctsa analysis. This selective focus seems both narrow and incomplete.

4. Connectivity results

The changes in functional connectivity as a result of direct pathway stimulation (excitation and inhibition) are both fascinating and limited. There is a clear excitation/inhibition difference in effects, as shown in Figure 7 B-C. However, Figure 7B suggests something different than the change results shown in Figure 7C. It appears that the application of clozapine increases functional connectivity in the control mice (black line Fig. 7B). This effect is exaggerated in the inhibition condition, but (most importantly) direct pathway excitation does not really reveal a significant change in the BOLD connectivity patterns. Now this does not change the authors' overall conclusions (connectivity is suppressed with direct pathway excitation relative to control mice), but the nuance of what is happening in the control mice is important for interpretation purposes: direct pathway excitation does not necessarily decrease functional connectivity but does not express the increase in connectivity observed from the application of clozapine. This needs to be elaborated more.

Along the same lines, there is an interesting disconnect between the intra-region results and the inter-region (connectivity) results. It is clear that resting BOLD dynamics in thalamic nuclei that project back to the striatum, as well as more unimodal cortical areas, change from direct pathway stimulation in the dorsal caudate. Yet, only one cortical region (MOp) with significant functional connectivity changes overlaps with the set of nuclei that exhibit intra-region BOLD changes. This suggests that local BOLD dynamics and global connectivity are largely disconnected effects. Yet this seems to be largely ignored in the current work. It would be nice to see more analysis, and discussion, of the intra-region and inter-region stimulation effects.

any version level dependencies ? for any of the mentioned OS?

Note: This rebuttal 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

1. General Statements [optional]

We are grateful for the thoughtful comments and suggestions from the reviewers which we feel have resulted in a manuscript that is both clearer to the reader and more rigorous. We have addressed the suggested revisions point by point below:

2. Point-by-point description of the revisions

Reviewer #1 Major comments:

Figure 2E. This shows the residence time of Cse4 without Ndc10 association. How does this compare to the residence time on mutant CEN3 (Supplemental Figure 1). It looks like Cse4 still binds to CEN3 with some specificity even in the absence of Ndc10. Does this suggest that Cse4 has some intrinsic ability to recognize CEN3? Alternatively, Ndc10 is still required for Cse4 binding but is below detection in the Cse4-alone residence lifetimes. Ideally, the authors would compare this with Cse4 binding in an Ndc10 mutant.

We thank the reviewer for this interesting question. As suggested, we analyzed Cse4 behavior on the mutant CDEIIImut CEN3 DNA, which does not stably recruit Ndc10 (Figure 1C), using the real-time colocalization assay. Although overall Cse4 associations were reduced, we still observed transient interactions, consistent with the possibility that Cse4 has some intrinsic ability to recognize CEN3. Kaplan-Meier analysis of the lifetimes of Cse4 colocalizations on CDEIIImut CEN3 DNA were significantly reduced when compared to CEN DNA (Figure EV1G, H). We have added these points to the text (p. 10, lines 29-31 and p. 11, lines 1-8).

Figure 3 explores the very interesting relationship between Scm3 dynamics and Cse4 binding but I feel that the data is not fully flushed out. A key finding is that Cse4 can potentially bind to CEN DNA prior to engaging with Scm3 to be incorporated. This predicts that Cse4 should bind first and then colocalize with Scm3. Can this be detected in the timing of the traces? How often does Scm3 bind to already prebound Cse4 and does this correlate with Cse4 residing longer?

We performed new and more rigorous analyses of the data to address these questions in the revised manuscript. After our analysis to calculate ternaryScm3 off-rates, we analyzed the relative timing of these ternary residences and found that indeed Cse4 can bind to CEN DNA prior to Scm3 and these events do correlate with Cse4 residing longer. Complete analysis of the binding order of Cse4 and Scm3 ternary residences revealed that Scm3 bound to CEN3 DNA prior to Cse4 more often than Cse4 preceding Scm3 (46% vs 34% of ternary residences) with the remaining 20% arriving simultaneously (Figure EV2E). Despite a difference in prevalence, the median lifetimes of both Scm3-Firstand Scm3-Last contexts were similar to each other (Figure EV2F) and both were significantly stabilized when compared to non-ternary residences. These results highlight a potential mechanism where Scm3 catalyzes stable Cse4 incorporation at centromeric DNA instead of delivering it to the centromere regardless of the order of arrival. These data are now reported and discussed in the revision (p. 12, line 11-p.13, line 10).

A related and perhaps even more interesting point is whether Scm3 is involved in "loading" of Cse4. If so, then one would expect that once Cse4 is assembled into nucleosomes it should be stable, even if Scm3 leaves. Can the authors extract this from the data? Alternatively, it is possible that Scm3 remains associated to Cse4 to maintain the nucleosome which would imply a more extended role for Scm3 apart from assembly alone. It would be interesting if information on this can be extracted from the data.

Using our updated analysis of ternaryScm3 Cse4 residences, feel we have addressed this possibility indirectly in a couple of ways. First, we found that in instances where ternary Scm3/Cse4 complexes are formed, Scm3 co-occupied the CEN DNA an average of 56.0% of the total Cse4 residence time, which is distinctly shorter than the 84% of the total Cse4 residence that was occupied by Ndc10 in Cse4/Ndc10 ternary residences. These findings are consistent with a Scm3-turnover mechanism that occurs post ternary complex formation with Cse4 as we found that Cse4 off-rates were significantly reduced after Scm3 association (Figure 3D).

Second, further analysis of Scm3 residence behavior revealed that there was no significant stabilization of Scm3 after ternary Cse4/Scm3 complex formation vs non-ternary Scm3 residences found in either off-rates (33 hr-1 vs 32 hr-1, Figure EV2C) or median lifetimes (45 s vs 52 s, Figure EV2D). These results indicate that Scm3 association is not stabilized like Cse4 after ternary complex formation and points to a potential catalytic role in Cse4 nucleosome formation, leaving a stable Cse4 nucleosome after turnover. We reported these findings in the revision results section (p. 12, lines 11-16) as well as briefly within the discussion.

Even in the presence of Scm3 and CCAN components, Cse4 appears to have a limited lifetime in the in vitro assay compared to in vivo stability. The authors should speculate on whether activities exist in their extract that actively disassembles nucleosomes. Perhaps ATP could be depleted to inactivate remodellers?

This is an excellent suggestion that we addressed with a new experiment. We performed an endpoint localization experiment with lysate containing fluorescently labeled Ndc10 and Cse4 and then removed the lysate and incubated the slide for 24hr in imaging buffer at RT. Strikingly, the proteins were maintained at the CEN DNA with a high protein total colocalization (~75% retention) (shown in Figure EV1B, C). These data suggest that the lysates may contain negative regulatory factors and we have added this point in the revised text (p. 9 lines 6-25).

We were not able to address whether the removal of ATP stabilized the proteins because we previously found that ATP depletion of the lysates completely abrogates kinetochore assembly in extracts. We will need to eventually dissect the role of remodelers in future work using a different approach.

For Figure 6, it is not clear why AT-track mutants of CDEII are labeled as genetically stable and genetically unstable. This is confusing as the "genetically stable" show a more than 10-fold increase in chromosome loss rates. Perhaps these can be renamed into "unstable" and "very unstable" or "weak" and "strong" mutants, just to make clear that these classes are both poorer than wild type.

We had deferred to the nomenclature used in the previous study (Baker and Rogers, 2005) which initially characterized these mutants. To avoid this confusion, we have renamed these mutants “unstable” and “very unstable” as suggested to make it clearer that none of these synthetic mutants fully recapitulate WT CEN3 behavior.

Finally, it would be wonderful to include data to assess whether a full Cse4 nucleosome is assembled or a partial nucleosome e.g. just Cse4/H4 tetrasomes. This could be done by tracking the accumulation of H2A or H2B at the CEN3. This would give further insight into what step Scm3 catalyses.

This is a very interesting suggestion that we were not able to directly address. Epitope tagging of these histone proteins in Saccharomyces cerevisiae with endogenous fluorophores has proved challenging due to gene duplication, overall sensitivity to histone levels within the cells and disruption of histone function by epitope tagging. We hope to find a workable method to address this in the future to address this question directly.

Minor comments:

Typo on page 5, line 1 "nucleosom" missing an e.

We have corrected this in the revised text.

Kaplan-Meyer should be spelled Kaplan-Meier

We have corrected this in the revised text.

The term "censored" is mentioned across many figures but comes up just ones in the methods where it is not clearly explained. Perhaps this could be clarified in the legend.

We have now provided a clear explanation of the term “censored” in the text on p. 28, lines 25-27. It has also been added to the figure legends and reported in the Statistical tests section of the methods section to address this point.

The abstract states that Cse4 can arrive at the centromere without its chaperone. More likely, Cse4 is in complex with other chaperones that may allow it to bind. Perhaps the abstract can be modified to read "Cse4 can arrive at the centromere without its dedicated centromere-specific chaperone Scm3..."

We updated the abstract to reflect this point in the revised text.

Related to this point, the discussion states the possibility that Cse4 can initially bind to CEN3 via other more general chaperones. However, it should be acknowledged that transient Cse4 binding in their assay may simply occur through mass action due to high concentrations of CEN3 DNA. In vivo, this transient binding may not be that relevant.

We acknowledge this potential caveat in the discussion section (p. 20 lines 15-18), although we feel this is somewhat unlikely due to our observation of significantly reduced Cse4 binding on CDEIIImut DNA despite DNA concentrations being similar to previous assays (Figure EV3A). We speculate that some of this transient behavior is at least in part driven by two major factors: negative regulatory factors within our cellular extracts that counter nucleosome formation (as explored in Figure EV1B-C) and photostability of the endogenous fluorophores used within the study (Figure EV1D-E). These points were highlighted within the second paragraph of page 9.

Reviewer #2 Major comments:

1. Figure S1A-D seem like some of the most compelling data in the paper to bolster the rigor of their experimental setup. There appears to be plenty of space to include these data in the main figure set in Figure 1 after panel D. The authors would be well served to consider moving S1A-D somewhere in the main figure set.

We appreciate the helpful feedback on the importance of the date found in Supplemental Figure 1 and have now incorporated it into Figure 1 within the main text as suggested.

The authors conclude that Ndc10 recruits HJURP(Smc3) to the yeast point centromeres. If this is the case, can the TIRFM assay measure ternary residence lifetimes complexes between Ndc10/HJURP(Smc3)/CenDNA?

We made the conclusion that Ndc10 recruits Scm3 based on previous publications showing this requirement in vivo. We have now attempted to address this in our assay indirectly by monitoring Scm3 behavior on the CDEIIImut CEN3 DNA that lacks Ndc10. Surprisingly, we found that Scm3 interacted similarly with CDEIIImut CEN3 DNA and actually showed an increase in median lifetimes vs. CEN3 DNA (Figure EV2B), suggesting its intrinsic DNA-binding activity may be the primary driver of its CEN DNA binding and that stable Cse4 association is required for its turnover. These data suggest that Ndc10 is not driving Scm3 interaction (or targeting) to CEN3. We are grateful to the reviewer for pointing this out and have adjusted our conclusions in the revised manuscript (p. 12, line 26-p.13 line 10).

Throughout the manuscript short- and long- term residence lifespans are mentioned, referencing the figures containing lines with lengths depicting residence times. This is a qualitative reference to short and long residences. Can the authors provide a graph for short-term ( 300 s) residence life-spans for, CENPA alone, CENPA/Ndc10, and CENPA/HJURP on CEN3 DNA? Or some figure similar to Figure 3C, but reporting the proportion of short-term vs long-term residence?

We typically used Kaplan-Meier survival function estimates to compare binding behavior but agree that quantification of residences within these contexts may be easier for the reader to follow. We have therefore quantified short-term ( 300 s) as suggested and added them as a panel (F) to Figure EV1 and as panel (E) in Figure 3.

The choice of CCAN components for analysis in Fig. 5 is interesting, but many readers may be curious why Mif2 wasn't selected for disruption, since it has such a cozy placement with CENP-A and CEN DNA. Can choice to not include Mif2 mutants/degrons be mentioned/justified in the text (unless, even better yet, they choose to address Mif2 role directly in new experimentation)?

We relied on structural models to choose CCAN proteins that are in close proximity to the DNA. Because Mif2 is not in these structures, we did not include it in our studies. We have explained this in the revised text (p. 16 lines 14-17) and agree it is an interesting future area of study.

Minor comments:

1. Are these whole cell extracts (WCE) DNA-free? I'm curious if there is any competition from endogenous DNA from the yeast cellular extract.

The extracts are not DNA-free so it is likely there is some competition from endogenous DNA. We have avoided enzymatically removing the DNA since the TIRF assay depends on the integrity of DNA.

In relation to Mif2 and comment #4 above, do the authors make any connection to their results with synthetic nucleosome sequences not being conducive to yeast centromere formation with the prior observation (Allu et al 2019) using recombinant components that the human version of Mif2 more easily saturates its binding sites on CENP-A nucleosomes when they are assembled with natural centromere DNA rather than the Widom 601 sequence?

We did not speculate on the role of Mif2 and stability of synthetic nucleosome sequences. This is an interesting point but the differences between the yeast and human systems combined with the fact we have not yet started to study Mif2 made it seem too premature to include in this manuscript.

Providing a gel (or other measure) of the DNA templates (750, 250, and 80 bp) used in TIRFM assay would be nice to show to confirm the designed size of the pre-tethered DNA.

We agree this is a helpful control and we have now included it as a panel in Figure EV5B.

1. Some of the references to figures/figure panels in the main text do not match the figures. (discussion, pg 16, paragraph 1 & 2;pg 18, paragraph 1).

We have updated references mentioned to reflect to the correct figure in both sections of the discussion.

Reviewer #3 Major comments: -->Statistics are highly recommended for all the data in the ms.

We have included log-rank analysis to instances where two survival functions were plotted together where appropriate. P-values for all these analyses were reported in the appropriate figure legends.

• At what rate is data collected in the TIRFM setup. For clarity for the reader, it is important to provide imaging details for time-lapse. What is the impact of photobleaching on the stability of the fluorophore signal? Please clarify.

This is a helpful suggestion and we have now included imaging details (like time intervals for each channel) when the real-time assay is introduced in the results section (p. 8, lines 10-12). We have also provided additional details for the photobleaching estimates and how these might censor data in turn (p. 9, lines 14-25). Although photobleaching is a primary limitation of the time-lapse assays, we point out that it is appropriate to compare protein behavior under identical imaging parameters within differing contexts. We also noted that we typically compared time-lapse behavior (which is affected by photobleaching) with endpoint assays to ensure consistent behavior.

• The power of single-molecule technique is precisely that such data can be made quantitative. Indeed, the Kaplan-Meyer graphs do show nice quantitative results. Unfortunately, in the text few quantitative measurements are reported. In fact, the Kaplan-Meyer graphs can be interpreted in a quantitative manner such as probability of residency time. Although in most cases the statistical significance between two conditions can be expected, this is not formally calculated and shown. What is the 50% survival time of Cse4 alone or with Ndc10, for instance? This manuscript would greatly benefit from a quantitative approach to the data, including a summary table of the results of the various conditions tested. Please add this important table.

We initially put the quantitative data in the figure legends but omitted it from the main text for simplicity but appreciate the Reviewer’s point. We note that we performed log-rank tests on all Kaplan-Meier analyses that are plotted on the same graph to provide statistical differences where applicable and have included all P-values in the figure legends. In response to the suggestion, we have now also included a table (Table 1) that contains the median survival time for all proteins tested as well as the median survival times for the differing contexts tested for quick reference and easier comparison for the reader.

• This reviewer would expect that the endpoint (90 min) would roughly reflect the occupancy results from time-lapse (45 min) experiments. Based on the data presented in Figures 1, 2, S1-3, this does not appear to be the case. 50% of Cse4-GFP and 70% Ndc10-mCherry colocalized with CEN3 DNA in the endpoint experiment, whereas in Fig 2C, ~30 and ~52 traces were positive for Cse4-GFP and Ndc10-mCherry, resp. with the former having drastically lower residency survival compared to Ndc10-mCherry. If indeed, 50% of Cse4-GFP makes it to the endpoint, about 50% of all traces should reach the end of the 45 minutes time-lapse window. Only about 1/3 of all positive Cse4-GFP traces can be seen at the end of the 45 min window. Could this be due to technical issues of photostability of GFP? Why does the colocalization of Cse4 signal with the DNA signal disappear so readily? Are Cse4 so unstable? Is the same true for canonical H3 nucleosomes? This unlikely true for nucleosomes in cells.

This is a valid concern, and we appreciate the thoughtful critique. The inconsistency noted between the initial endpoint colocalization and those reported later in the paper is mainly due to the difference between yeast strains carrying Cse4 tagged alone in comparison to multiple kinetochore proteins with tags that exhibit mild genetic interactions. This point is now explained in the revised text (p. 8, line 29-p. 9. line 3).

Photostability is also a factor in the live imaging experiments compared to the endpoint localization assays. However, our photobleaching estimates suggest that the Cse4 lifetimes are not limited by photobleaching (Figure EV1D, E) so we do not believe that accounts for the differences between experiments and it is mainly the presence of multiple epitope tags.

In regard to why Cse4 is not more stable, Reviewer 1 had the same question so we performed an experiment to address whether the lysate contains negative regulatory factors. We found that Cse4 is stable once the lysate is removed (Figure EV1B, C), consistent with the idea that there are factors that disrupt it in the lysate. We discuss these potential reasons for transient association in the revised text (p. 9, lines 4-25).

It should also be noted that there are clear differences in nucleosome formation in reconstitutions and within our extracts, as evident by the Widom-601 DNA data (Figure 6D). This was not necessarily unexpected, as extracts are a much more complex medium, but we are encouraged by the fact that at least once formed, these Cse4-containing particles on CEN DNA are perhaps more stable than their reconstituted counterparts that seem to be so far unsuitable for structural studies.

Along the same lines, in Suppl Fig 3 there is a disconnect between residency survival and endpoint colocalization on either CEN3, CEN7, or CEN9. What could be the underlying mechanism between the discordance of endpoint results and time-lapse results? Could this be the result of experimental differences?

We are grateful that this discrepancy was highlighted to us, as upon closer examination we discovered that endpoint colocalization analysis had not been correctly updated in the figure to include data from equivalent genetic backgrounds as the CEN3 and CEN9 assays. Updating the figure in Appendix Figure S2 to include this data remedied this discrepancy.

• What fraction of particles show colocalization between Cse4-GFP and Ndc10-mCherry? What fraction of occupancy time show colocalization between Cse4-GFP and Ndc10-mCherry? Altogether, understanding the limitation and benefits of endpoint analysis and time-lapse analysis in this particular experimental set-up is important to be able to interpret the results. Please clarify these points.

We have now added particle numbers to all survival estimate plots which makes it much easier for the reader to interpret the proportion of Cse4 residences that are ternary vs. non-ternary in instances where off-rates were quantified and Kaplan-Meier analysis was performed on the resulting lifetimes. We determined that for ternary Cse4-Ndc10 residences, Cse4 and Ndc10 co-occupied the CEN DNA an average of ~84% of the total Cse4 residence.

• Page 9, third sentence of third paragraph it is stated that the "results suggests that Scm3 helps promote more stable binding of Cse4 ...". This is indeed one possible explanation of the results, and this possibility is tested by overexpressing Psh1 or Scm3 by endpoint colocalization analysis. 1) Taking the concerns regarding the endpoint vs time-lapse results into account, wouldn't it be more accurate to compare either time-lapse results against each other or endpoint results? 2) Alternatively, more stable Cse4 particles are able to recruit Scm3, simply because of the increased binding opportunity of a more stable particle. In this scenario, just the residency time of Cse4 alone is the predicting factor in likelihood to associate with Scm3. To test the latter possibility, Cse4 stability would need to be altered. Please consider this experiment- should be relatively easy with the right mutant of either CSE4 or CDEII (see Luger or Wu papers).

We appreciate the points raised here and addressed both as follows. For point (1) we altered the text in the third paragraph of the section, The conserved chaperone Scm3HJURP is a limiting cofactor that promotes stable association Cse4CENP-A with the centromere, to make it more clear to the reader that in the experiments presented in Figure EV4, endpoint analysis results were only compared to each other, and likewise time-lapse experiments were only compared to each other for each genetic background. While the results were consistent between experiments, we did not directly compare results from one to the results of another, but instead we used both assays to characterize Cse4CENP-A behavior more completely in differing contexts.

To test the alternative hypothesis proposed in point (2), we sought to avoid potential selection bias by utilizing off-rate analysis, which enabled us to separate the portions of Cse4 residences that occurred prior to ternary association with either Ndc10 or Scm3. This unbiased approach allowed us to compare Cse4 residence lifetimes pre and post ternary association and we found that there were still significant differences in off-rates and median lifetimes of the associated ternary and non-ternary residences using this updated analysis. We thank the reviewer for helping to guide us towards this more robust analysis.

Based on the recommendation in point (2), we also sought to directly compare the behavior of Cse4 and Scm3 on the “Very Unstable” CDEII mutants described in the section, DNA-composition of centromeres contributes to genetic stability through Cse4CENP-A recruitment. In this case, equivalent extracts were used and Cse4 stability was altered directly via the DNA template. When the off rates of ternaryScm3 Cse4 residences were compared, we found a significant increase in off-rates of Cse4 on the “Very Unstable” CDEII mutant CEN DNA (Appendix Figure S3B) compared to WT CEN DNA. If the alternative hypothesis proposed in point (2) were true, we would expect this reduction in median lifetime to correlate with a lower proportion of Cse4-Scm3 ternary association but quantification yielded proportions that, while varied, were not on average lower than the proportion of Cse4-Scm4 association on CEN3 DNA (.23 vs .31, Appendix Figure S3A). This finding, taken together with the fact that it would be difficult for us to propose an alternative hypothesis that explains the results outlined in Figure EV4, supports our hypothesis that Scm3 helps promote more stable binding of Cse4 and that this stabilization is directly influenced by DNA sequence composition.

• In Figure 1C and Supplemental Figure 5B, there appears to be foci that CEN3-ATTO-647 positive, but Cse4-GFP negative and visa verse. It seems logical that there are DNA molecules that didn't reconstitute Cse4 nucleosomes. But how can there be Cse4-GFP positive foci without a positive DNA signal? Is it possible that unlabeled DNA make it onto the flow chamber? If so, can these unlabeled DNA be visualized by Sytox Orange for instance to confirm that no spurious Cse4 deposition occurred? Please clarify.

Because it is unlikely that random associations will colocalize with the labeled DNA based on control assays (Supp. Figure 1C) that show this occurs rarely (

• On page 10, at the end of the first paragraph, growth phenotype of pGAL-SCM3 and pGAL-PSH1 mutants were tested. On GAL plates, pGAL-PSH1 shows reduced growth, but not pGAL-SCM3. Wouldn't a more accurate conclusion be that Psh1 is limiting stable centromeric nucleosome formation, instead of Scm3?

The growth defects on galactose don’t necessarily mean that a factor is limiting in cells. Instead, they report on whether changing the relative amounts of the complex lead to phenotypes in cells that could be the result of many causes that would require characterization of the phenotypes to understand. In this case, we presume that Psh1 titrates Scm3 away from Cse4 to prevent nucleosome formation in vivo. However, we have not directly tested this so we just concluded that the relative levels of the complexes are important for cell growth.

• In the section where DNA was tethered at either one or both ends, an important control is missing. How does such a set-up impact nucleosome formation in general. Can H3 nucleosomes form on random DNA that is either tethered at one or both ends? Does this potentially affect the unwrapping potential/topology of AT-tract DNA? Please comment.

This is an interesting point and one that we hope to explore further in the future but was beyond the scope of this paper. We suspect that restrictions via tethering would also limit canonical nucleosome formation on random DNA. We envision that unwrapping may be affected as well and hope to explore this via other, potentially better suited techniques like optical tweezers.

Minor comments + Censored data points are not explained in the text.

A brief explanation of censorship was added to the figure legends and we have now provided a clear explanation of the term “censored” in the text on p. 28, lines 25-27. It has also been reported in the Statistical tests section of the methods section to address this point.

• Number of particles tested should be reported in the main and supplemental figures, not just the legends for those readers who first skim the manuscript before deciding to read it.

We add these values to all Kaplan-Meier plots in all figures (Main, Expanded View, and Appendix)

• Typo on page 5, first line: "nucleosom" should be "nucleosome".

We fixed this in the text.

• Typo on page 9, second line: sentence is missing something "... is required for Scm3-dependent ..."

We fixed this in the text.

• It is unclear how the difference in Supplemental figure 5D was calculated.

We included log-rank test generated P-values as well as description in the figure legend of EV4.

• Figure 4C: why are there more Ndc10-mCherry foci observed in double tethered constructs vs single tethered constructs?

There can be variances in DNA density between slides, particularly with non-dye labeled DNA template. We updated figure panel C to include a representative image with similar Ndc10 density.

• For the display of individual traces as shown in Fig 2B, 3A, 4E, and 5E, it might be more informative if the z-normalized signal and the binary read-out are shown in separate windows to better appreciate how the z-normalized signal was interpretated.

Due to spacing limits within figures we attempted to accommodate this by reducing the thickness of the binary read-out and ensured that the raw data traces were overlaid for easier interpretation by the reader.

• Page 17, fifth line of the second paragraph, it is stated that a conserved feature of centromeres is their AT-richness. This is most certainly true for the majority of species studied thus far, but bovine centromeres for instance are about 54% GC rich. Indeed, Melters et al 2013 Genome Biol showed that in certain clades centromeres can be comprised of GC-rich sequences. It might be worthwhile to nuance this statement.

We have updated the text to reflect that AT-rich DNA is widely conserved but not a universal feature of centromeres.

• Page 17, last paragraph. Work by Karolin Luger and Carl Wu is cited in relationship to AT-rich DNA being unfavorable for canonical nucleosome deposition. A citation is missing here: Stormberg & Lyubchenko 2022 IJMS 23(19): 11385. Also, the first person to show that AT-tracts affect nucleosome positioning are Andrew Travers and Drew. This landmark work should be cited.

We thank the Reviewer for noticing this and have added the appropriate citations.

• Page 18, 9th line from the top, it is stated that yeast centromeres are sensitive to negative genetic drift. This reviewer is of the understanding that genetic drift is a statistical fluctuation of allele frequency, which can result in either gain or loss of specific alleles. Population size is a major factor in the potential power of genetic drift. The smaller a population, the greater the effect. Budding yeast is found large numbers, which would mean that drift only has limited predicted impact. Maybe the authors meant to use the term purifying selection?

We appreciate this clarification and agree with the reviewer, we have updated the manuscript to cite purifying selection and not genetic drift as at centromeres.

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

The manuscript "Single molecule visualization of native centromeric nucleosome formation reveals coordinated deposition by kinetochore proteins and centromere DNA sequence" by Popchock and colleagues describes a new high-throughput single-molecule technique that combines both in vitro and in vivo sample sources. Budding yeast centromeres are genetically defined centromeres, which makes them ideal for studying short DNA segments at the single-molecule level. By flowing in whole cell lysates, Cse4 nucleosomes can form under near physiological conditions. Two analytical experiments were performed: endpoint and time-lapse. In the former case, nucleosomes were allowed to form within 90 minutes and the latter case, nucleosome formation was tracked for up to 45 minutes. In addition, well described genetic mutants were used to assess the stability of Cse4 nucleosomes, as well as different DNA sequences (this we particularly liked- well done). Overall, this is a very interesting technique with potential to be useful for studying any DNA-based effect, ranging from DNA repair to kinetochore assembly. This is strong and impactful work, and the potential this kind of microscopy has for solving kinetic problems in the field. We think it's worthy of publication after revising technical and experimental concerns that would elevate the ms significantly.

Major comments:

Statistics are highly recommended for all the data in the ms.

• At what rate is data collected in the TIRFM setup. For clarity for the reader, it is important to provide imaging details for time-lapse. What is the impact of photobleaching on the stability of the fluorophore signal? Please clarify.
• The power of single-molecule technique is precisely that such data can be made quantitative. Indeed, the Kaplan-Meyer graphs do show nice quantitative results. Unfortunately, in the text few quantitative measurements are reported. In fact, the Kaplan-Meyer graphs can be interpreted in a quantitative manner such as probability of residency time. Although in most cases the statistical significance between two conditions can be expected, this is not formally calculated and shown. What is the 50% survival time of Cse4 alone or with Ndc10, for instance? This manuscript would greatly benefit from a quantitative approach to the data, including a summary table of the results of the various conditions tested. Please add this important table.
• This reviewer would expect that the endpoint (90 min) would roughly reflect the occupancy results from time-lapse (45 min) experiments. Based on the data presented in Figures 1, 2, S1-3, this does not appear to be the case. 50% of Cse4-GFP and 70% Ndc10-mCherry colocalized with CEN3 DNA in the endpoint experiment, whereas in Fig 2C, ~30 and ~52 traces were positive for Cse4-GFP and Ndc10-mCherry, resp. with the former having drastically lower residency survival compared to Ndc10-mCherry. If indeed, 50% of Cse4-GFP makes it to the endpoint, about 50% of all traces should reach the end of the 45 minutes time-lapse window. Only about 1/3 of all positive Cse4-GFP traces can be seen at the end of the 45 min window. Could this be due to technical issues of photostability of GFP? Why does the colocalization of Cse4 signal with the DNA signal disappear so readily? Are Cse4 so unstable? Is the same true for canonical H3 nucleosomes? This unlikely true for nucleosomes in cells. Along the same lines, in Suppl Fig 3 there is a disconnect between residency survival and endpoint colocalization on either CEN3, CEN7, or CEN9. What could be the underlying mechanism between the discordance of endpoint results and time-lapse results? Could this be the result of experimental differences?
• What fraction of particles show colocalization between Cse4-GFP and Ndc10-mCherry? What fraction of occupancy time show colocalization between Cse4-GFP and Ndc10-mCherry? Altogether, understanding the limitation and benefits of endpoint analysis and time-lapse analysis in this particular experimental set-up is important to be able to interpret the results. Please clarify these points.
• Page 9, third sentence of third paragraph it is stated that the "results suggests that Scm3 helps promote more stable binding of Cse4 ...". This is indeed one possible explanation of the results, and this possibility is tested by overexpressing Psh1 or Scm3 by endpoint colocalization analysis. 1) Taking the concerns regarding the endpoint vs time-lapse results into account, wouldn't it be more accurate to compare either time-lapse results against each other or endpoint results? 2) Alternatively, more stable Cse4 particles are able to recruit Scm3, simply because of the increased binding opportunity of a more stable particle. In this scenario, just the residency time of Cse4 alone is the predicting factor in likelihood to associate with Scm3. To test the latter possibility, Cse4 stability would need to be altered. Please consider this experiment- should be relatively easy with the right mutant of either CSE4 or CDEII (see Luger or Wu papers).
• In Figure 1C and Supplemental Figure 5B, there appears to be foci that CEN3-ATTO-647 positive, but Cse4-GFP negative and visa verse. It seems logical that there are DNA molecules that didn't reconstitute Cse4 nucleosomes. But how can there be Cse4-GFP positive foci without a positive DNA signal? Is it possible that unlabeled DNA make it onto the flow chamber? If so, can these unlabeled DNA be visualized by Sytox Orange for instance to confirm that no spurious Cse4 deposition occurred? Please clarify.
• On page 10, at the end of the first paragraph, growth phenotype of pGAL-SCM3 and pGAL-PSH1 mutants were tested. On GAL plates, pGAL-PSH1 shows reduced growth, but not pGAL-SCM3. Wouldn't a more accurate conclusion be that Psh1 is limiting stable centromeric nucleosome formation, instead of Scm3?
• In the section where DNA was tethered at either one or both ends, an important control is missing. How does such a set-up impact nucleosome formation in general. Can H3 nucleosomes form on random DNA that is either tethered at one or both ends? Does this potentially affect the unwrapping potential/topology of AT-tract DNA? Please comment.

Minor comments

• Censored data points are not explained in the text.
• Number of particles tested should be reported in the main and supplemental figures, not just the legends for those readers who first skim the manuscript before deciding to read it.
• Typo on page 5, first line: "nucleosom" should be "nucleosome".
• Typo on page 9, second line: sentence is missing something "... is required for Scm3-dependent ..."
• It is unclear how the difference in Supplemental figure 5D was calculated.
• Figure 4C: why are there more Ndc10-mCherry foci observed in double tethered constructs vs single tethered constructs?
• For the display of individual traces as shown in Fig 2B, 3A, 4E, and 5E, it might be more informative if the z-normalized signal and the binary read-out are shown in separate windows to better appreciate how the z-normalized signal was interpretated.
• Page 17, fifth line of the second paragraph, it is stated that a conserved feature of centromeres is their AT-richness. This is most certainly true for the majority of species studied thus far, but bovine centromeres for instance are about 54% GC rich. Indeed, Melters et al 2013 Genome Biol showed that in certain clades centromeres can be comprised of GC-rich sequences. It might be worthwhile to nuance this statement.
• Page 17, last paragraph. Work by Karolin Luger and Carl Wu is cited in relationship to AT-rich DNA being unfavorable for canonical nucleosome deposition. A citation is missing here: Stormberg & Lyubchenko 2022 IJMS 23(19): 11385. Also, the first person to show that AT-tracts affect nucleosome positioning are Andrew Travers and Drew. This landmark work should be cited.
• Page 18, 9th line from the top, it is stated that yeast centromeres are sensitive to negative genetic drift. This reviewer is of the understanding that genetic drift is a statistical fluctuation of allele frequency, which can result in either gain or loss of specific alleles. Population size is a major factor in the potential power of genetic drift. The smaller a population, the greater the effect. Budding yeast is found large numbers, which would mean that drift only has limited predicted impact. Maybe the authors meant to use the term purifying selection?

Significance

This study developed an in vitro imaging technique that allows native proteins from whole cell lysates to associate with a specific DNA sequence that is fixed to a surface. By labeling proteins with specific fluorophore-tags colocalization provides insightful proximity data. By creating mutants, the assembly or disassembly of protein complexes on native or mutated DNAs can therefore be tracked in real time. In a way, this is a huge leap forward from gel shift EMSA assays that have traditionally been used to do comparative kinetics in biochemistry.

This makes this technique ideal for studying DNA binding complexes, and potentially, even RNA-binding complexes. This study shows both the importance of using genetic mutants, as well as testing the effects of the fixed DNA sequence. As many individual fixed DNA molecules can be tracked at one, it allows for high-throughput analysis, similar to powerful DNA curtain work from Eric Greene's lab. Overall, this is a promising new single-molecule technique that combines in vitro and ex vivo sample sources, and will likely appeal to a broad range of molecular and biophysics scientists.

Scrollable windows can cause some confusion, leading to difficulties in understanding and errors in sentence breaks when using TTS.

The room is pierced by sunlight coming from two back windows, highlighting a dramatic contrast between the natural joy and promise of spring outside and the stark, gloomy oppressive and intimidating interior ("solemn, even forbidding").

Joint Public Review:

In the current paper, Jones et al. describe a new framework, named coccinella, for real-time high-throughput behavioral analysis aimed at reducing the cost of analyzing behavior. In the setup used here each fly is confined to a small circular arena and able to walk around on an agar bed spiked with nutrients or pharmacological agent. The new framework, built on the researchers' previously developed platform Ethoscope, relies on relatively low-cost Raspberry Pi video cameras to acquire images at ~0.5 Hz and pull out, in real time, the maximal velocity (parameter extraction) during 10 second windows from each video. Thus, the program produces a text file, and not voluminous videos requiring storage facilities for large amounts of video data, a prohibitive step for many behavioral analyses. The maximal velocity time-series is then fed to an algorithm called Highly Comparative Time-Series Classification (HCTSA)(which itself is based on a large number of feature extraction algorithms) developed by other researchers. HCTSA identifies statistically salient features in the time-series which are then passed on to a type of linear classifier algorithm called support vector machines (SVM). In cases where such analyses are sufficient for characterizing the behaviors of interest this system performs as well as other state-of-the-art systems used in behavioral analysis (e.g., DeepLabCut).

In a pharmacobehavior paradigm testing different chemicals, the authors show that coccinella can identify specific compounds as effectively as other more time-consuming and resource-consuming systems.<br /> The new paradigm should be of interest to researchers involved in drug screens, and more generally, in high-throughput analysis focused on gross locomotor defects in fruit flies such as identification of sleep phenotypes. By extracting/saving only the maximal velocity from video clips, the method is fast. However, the rapidity of the platform comes at a cost--loss of information on subtle but important behavioral alterations. When seeking subtle modifications in animal behavior, solutions like DeepLabCut, which are admittedly slower but far superior in terms of the level of details they yield, would be more appropriate.

The manuscript reads well, and it is scientifically solid.

1- The fact that Coccinella runs on Ethoscopes, an open source hardware platform described by the same group, is very useful because the relevant publication describes Ethoscope in detail. However, the current version of the paper does not offer details or alternatives for users that would like to test the framework, but do not have an Ethoscope. Would it be possible to overcome this barrier and have coccinella run with any video data (and, thus, potentially be used to analyze data obtained from other animal models)?

2- Readers who want background on the analytical approaches that the platform relies on following maximal velocity extraction, will have to consult the original publications. In particular, the current manuscript does not provide much information on Highly Comparative Time-Series Classification (HCTSA) or SVM; this may be reasonable because the methods were developed earlier by others. While some readers may find that the lack of details increases the manuscript's readability, others may be left wanting to see more discussion on these not-so-trivial approaches. In addition, it is worth noting that the same authors who published the HCTSA method also described a shorter version named catch22, that runs faster with a similar output. Thus, explaining in more detail how HCTSA operates, considering that it is a relatively new method, will make the method more convincing.

rime 输入法教程 #todo

Note that some conventional secure deletion methods do not translate well from HDDs to SSDs and data on SSDs deleted with software designed for HDDs may not always be physically overwritten as SSDs often 'pretend' to behave like HDDs to operating systems for legacy compatibility and actually do different things from what they report.

Quote of a digital worker's words. Source (following note 12): "Die Zeit, Feuilleton Ausgabe no. 14, 30 March 2023—a joint cultural supplement project on artificial intelligence by Marie Serah Ebcinoglu, Eike Kühn, Jan Lichte, Peter Neumann, Hanno Rauterberg, Malin Schulz, Hito Steyerl and Tobias Timm. This interview was conducted by Tobias Timm."

Author Response

Reviewer #1 (Public Review):

The authors present a study of visuo-motor coupling primarily using wide-field calcium imaging to measure activity across the dorsal visual cortex. They used different mouse lines or systemically injected viral vectors to allow imaging of calcium activity from specific cell-types with a particular focus on a mouse-line that expresses GCaMP in layer 5 IT (intratelencephalic) neurons. They examined the question of how the neural response to predictable visual input, as a consequence of self-motion, differed from responses to unpredictable input. They identify layer 5 IT cells as having a different response pattern to other cell-types/layers in that they show differences in their response to closed-loop (i.e. predictable) vs open-loop (i.e. unpredictable) stimulation whereas other cell-types showed similar activity patterns between these two conditions. They analyze the latencies of responses to visuomotor prediction errors obtained by briefly pausing the display while the mouse is running, causing a negative prediction error, or by presenting an unpredicted visual input causing a positive prediction error. They suggest that neural responses related to these prediction errors originate in V1, however, I would caution against over-interpretation of this finding as judging the latency of slow calcium responses in wide-field signals is very challenging and this result was not statistically compared between areas.

Surprisingly, they find that presentation of a visual grating actually decreases the responses of L5 IT cells in V1. They interpret their results within a predictive coding framework that the last author has previously proposed. The response pattern of the L5 IT cells leads them to propose that these cells may act as 'internal representation' neurons that carry a representation of the brain's model of its environment. Though this is rather speculative. They subsequently examine the responses of these cells to anti-psychotic drugs (e.g. clozapine) with the reasoning that a leading theory of schizophrenia is a disturbance of the brain's internal model and/or a failure to correctly predict the sensory consequences of self-movement. They find that anti-psychotic drugs strongly enhance responses of L5 IT cells to locomotion while having little effect on other cell-types. Finally, they suggest that anti-psychotics reduce long-range correlations between (predominantly) L5 cells and reduce the propagation of prediction errors to higher visual areas and suggest this may be a mechanism by which these drugs reduce hallucinations/psychosis.

This is a large study containing a screening of many mouse-lines/expression profiles using wide-field calcium imaging. Wide-field imaging has its caveats, including a broad point-spread function of the signal and susceptibility to hemodynamic artifacts, which can make interpretation of results difficult. The authors acknowledge these problems and directly address the hemodynamic occlusion problem. It was reassuring to see supplementary 2-photon imaging of soma to complement this data-set, even though this is rather briefly described in the paper.

We will expand on the discussion of caveats as suggested.

Overall the paper's strengths are its identification of a very different response profile in the L5 IT cells compared other layers/cell-types which suggests an important role for these cells in handling integration of self-motion generated sensory predictions with sensory input. The interpretation of the responses to anti-psychotic drugs is more speculative but the result appears robust and provides an interesting basis for further studies of this effect with more specific recording techniques and possibly behavioral measures.

Reviewer #2 (Public Review):

Summary:

This work investigates the effects of various antipsychotic drugs on cortical responses during visuomotor integration. Using wide-field calcium imaging in a virtual reality setup, the researchers compare neuronal responses to self-generated movement during locomotion-congruent (closed loop) or locomotion-incongruent (open loop) visual stimulation. Moreover, they probe responses to unexpected visual events (halt of visual flow, sudden-onset drifting grating). The researchers find that, in contrast to a variety of excitatory and inhibitory cell types, genetically defined layer 5 excitatory neurons distinguish between the closed and the open loop condition and exhibit activity patterns in visual cortex in response to unexpected events, consistent with unsigned prediction error coding. Motivated by the idea that prediction error coding is aberrant in psychosis, the authors then inject the antipsychotic drug clozapine, and observe that this intervention specifically affects closed loop responses of layer 5 excitatory neurons, blunting the distinction between the open and closed loop conditions. Clozapine also leads to a decrease in long-range correlations between L5 activity in different brain regions, and similar effects are observed for two other antipsychotics, aripripazole and haloperidol, but not for the stimulant amphetamine. The authors suggest that altered prediction error coding in layer 5 excitatory neurons due to reduced long-range correlations in L5 neurons might be a major effect of antipsychotic drugs and speculate that this might serve as a new biomarker for drug development.

Strengths:

• Relevant and interesting research question:

The distinction between expected and unexpected stimuli is blunted in psychosis but the neural mechanisms remain unclear. Therefore, it is critical to understand whether and how antipsychotic drugs used to treat psychosis affect cortical responses to expected and unexpected stimuli. This study provides important insights into this question by identifying a specific cortical cell type and long-range interactions as potential targets. The authors identify layer 5 excitatory neurons as a site where functional effects of antipsychotic drugs manifest. This is particularly interesting as these deep layer neurons have been proposed to play a crucial role in computing the integration of predictions, which is thought to be disrupted in psychosis. This work therefore has the potential to guide future investigations on psychosis and predictive coding towards these layer 5 neurons, and ultimately improve our understanding of the neural basis of psychotic symptoms.

• Broad investigation of different cell types and cortical regions:

One of the major strengths of this study is quasi-systematic approach towards cell types and cortical regions. By analysing a wide range of genetically defined excitatory and inhibitory cell types, the authors were able to identify layer 5 excitatory neurons as exhibiting the strongest responses to unexpected vs. expected stimuli and being the most affected by antipsychotic drugs. Hence, this quasi-systematic approach provides valuable insights into the functional effects of antipsychotic drugs on the brain, and can guide future investigations towards the mechanisms by which these medications affect cortical neurons.

• Bridging theory with experiments

Another strength of this study is its theoretical framework, which is grounded in the predictive coding theory. The authors use this theory as a guiding principle to motivate their experimental approach connecting visual responses in different layers with psychosis and antipsychotic drugs. This integration of theory and experimentation is a powerful approach to tie together the various findings the authors present and to contribute to the development of a coherent model of how the brain processes visual information both in health and in disease.

Weaknesses:

• Unclear relevance for psychosis research

From the study, it remains unclear whether the findings might indeed be able to normalise altered predictive coding in psychosis. Psychosis is characterised by a blunted distinction between predicted and unpredicted stimuli. The results of this study indicate that antipsychotic drugs further blunt the distinction between predicted and unpredicted stimuli, which would suggest that antipsychotic drugs would deteriorate rather than ameliorate the predictive coding deficit found in psychosis. However, these findings were based on observations in wild-type mice at baseline. Given that antipsychotics are thought to have little effects in health but potent antipsychotic effects in psychosis, it seems possible that the presented results might be different in a condition modelling a psychotic state, for example after a dopamine-agonistic or a NMDA-antagonistic challenge. Therefore, future work in models of psychotic states is needed to further investigate the translational relevance of these findings.

We fully agree that it is unclear how the effects of antipsychotics in mice relate to the drug effects that would be observed in schizophrenic patients. It is also correct that the reduction of the difference between closed and open loop locomotion onset response in L5 IT neurons (Figure 4) is not what we would have expected to find under the assumption that psychosis is characterized by a blunted distinction between predicted and unpredicted stimuli. We are not sure how to interpret this finding. However, it is probably important to note that the difference is only reduced when using a normalized comparison. Looking just at the subtraction of the two curves, the difference between closed and open loop locomotion onset responses remains unchanged before and after antipsychotic drug injection. The finding of a decorrelation of layer 5 activity, however, is easier to interpret under the assumption that layer 5 functions as an internal representation. If speech hallucinations, for example, are the consequence of a spurious activation of internal representations in speech processing areas of cortex, then antipsychotics might reduce the probability of these spurious activation events by reducing the lateral influence between layer 5 neurons in different cortical areas.

We do indeed plan to address the question of how antipsychotics influence cortical processing in mouse models of schizophrenia in the future.

• Incomplete testing of predictive coding interpretation

While the investigation of neuronal responses to different visual flow stimuli Is interesting, it remains open whether these responses indeed reflect internal representations in the framework of predictive coding. While the responses are consistent with internal representation as defined by the researchers, i.e., unsigned prediction error signals, an alternative interpretation might be that responses simply reflect sensory bottom-up signals that are more related to some low-level stimulus characteristics than to prediction errors.

This is correct – we will expand on the discussion of this point in the manuscript.

Moreover, This interpretational uncertainty is compounded by the fact that the used experimental paradigms were not suited to test whether behaviour is impacted as a function of the visual stimulation which makes it difficult to assess what the internal representation of the animal actual was. For these reasons, the observed effects might reflect simple bottom-up sensory processing alterations and not necessarily have any functional consequences. While this potential alternative explanation does not detract from the value of the study, future work would be needed to explain the effect of antipsychotic drugs on responses to visual flow. For example, experimental designs that systematically vary the predictive strength of coupled events or that include a behavioural readout might be more suited to draw from conclusions about whether antipsychotic drugs indeed alter internal representations.

We agree that much additional work will be necessary to identify internal representation neurons. However, it is difficult to envision how behavioral output could be used to make inferences about internal representations in sensory areas of cortex. In humans, for example, there is evidence that internal representations in visual cortex and behavioral output are not always directly related: binocular rivalry activates representations of both stimuli shown in visual cortex, while the conscious experience that drives behavioral output is only of one of the two stimuli. Hence, we would assume that the internal representation in visual cortex does not necessarily relate to behavioral output.

• Methodological constraints of experimental design

While the study findings provide valuable insights into the potential effects of antipsychotic drugs, it is important to acknowledge that there may be some methodological constraints that could impact the interpretation of the results. More specifically, the experimental design does not include a negative control condition or different doses. These conditions would help to ensure that the observed effects are not due to unspecific effects related to injection-induced stress or time, and not confined to a narrow dose range that might or might not reflect therapeutic doses used in humans. Hence, future work is needed to confirm that the observed effects indeed represent specific drug effects that are relevant to antipsychotic action.

We agree that both dosages and a broader spectrum of non-antipsychotic compounds will need to be investigated. We are in the process of building a screening pipeline to perform exactly these types of experiments. We would however argue that the paper already includes a control condition in the form of the amphetamine data (Figure 7). While it is possible that amphetamine might have an effect that exactly cancels out potential i.p. injection- or stress-induced changes, we would argue it is more probable that these changes had no measurable effect on Tlx3 positive L5 IT neuron calcium activity per se. We will provide additional evidence that time or injection stress alone do not result in the observed effects.

Conclusion:

Overall, the results support the idea that antipsychotic drugs affect neural responses to predicted and unpredicted stimuli in deep layers of cortex. Although some future work is required to establish whether this observation can indeed be explained by a drug-specific effect on predictive coding, the study provides important insights into the neural underpinnings of visual processing and antipsychotic drugs, which is expected to guide future investigations on the predictive coding hypothesis of psychosis. This will be of broad interest to neuroscientists working on predictive coding in health and in disease.

Reviewer #3 (Public Review):

The study examines how different cell types in various regions of the mouse dorsal cortex respond to visuomotor integration and how antipsychotic drugs impacts these responses. Specifically, in contrast to most cell types, the authors found that activity in Layer 5 intratelencephalic neurons (Tlx3+) and Layer 6 neurons (Ntsr1+) differentiated between open loop and closed loop visuomotor conditions. Focussing on Layer 5 neurons, they found that the activity of these neurons also differentiated between negative and positive prediction errors during visuomotor integration. The authors further demonstrated that the antipsychotic drugs reduced the correlation of Layer 5 neuronal activity across regions of the cortex, and impaired the propagation of visuomotor mismatch responses (specifically, negative prediction errors) across Layer 5 neurons of the cortex, suggesting a decoupling of long-range cortical interactions.

The data when taken as a whole demonstrate that visuomotor integration in deeper cortical layers is different than in superficial layers and is more susceptible to disruption by antipsychotics. Whilst it is already known that deep layers integrate information differently from superficial layers, this study provides more specific insight into these differences. Moreover, this study provides a first step into understanding the potential mechanism by which antipsychotics may exert their effect.

Whilst the paper has several strengths, the robustness of its conclusions is limited by its questionable statistical analyses. A summary of the paper's strengths and weaknesses follow.

Strengths:

The authors perform an extensive investigation of how different cortical cell types (including Layer 2/3, 4 , 5, and 6 excitatory neurons, as well as PV, VIP, and SST inhibitory interneurons) in different cortical areas (including primary and secondary visual areas as well as motor and premotor areas), respond to visuomotor integration. This investigation provides strong support to the idea that deep layer neurons are indeed unique in their computational properties. This large data set will be of considerable interest to neuroscientists interested in cortical processing.

The authors also provide several lines of evidence that visuomotor information is differentially integrated in deep vs. superficial layers. They show that this is true across experimental paradigms of visuomotor processing (open loop, closed loop, mismatch, drifting grating conditions) and experimental manipulations, with the demonstration that Layer 5 visuomotor integration is more sensitive to disruption by the antipsychotic drug clozapine, compared with cortex as a whole.

The study further uses multiple drugs (clozapine, aripiprazole and haloperidol) to bolster its conclusion that antipsychotic drugs disrupt correlated cortical activity in Layer 5 neurons, and further demonstrates that this disruption is specific to antipsychotics, as the psychostimulant amphetamine shows no such effect.

In widefield calcium imaging experiments, the authors effectively control for the impact of hemodynamic occlusions in their results, and try to minimize this impact using a crystal skull preparation, which performs better than traditional glass windows. Moreover, they examine key findings in widefield calcium imaging experiments with two-photon imaging.

Weaknesses:

A critical weakness of the paper is its statistical analysis. The study does not use mice as its independent unit for statistical comparisons but rather relies on other definitions, without appropriate justification, which results in an inflation of sample sizes.

We will expand on both analyses and justifications throughout.

For example, in Figure 1, independent samples are defined as locomotion onsets, leading to sample sizes of approx. 400-2000 despite only using 6 mice for the experiment. This is only justified if the data from locomotion onsets within a mouse is actually statistically independent, which the authors do not test for, and which seems unlikely. With such inflated sample sizes, it becomes more likely to find spurious differences between groups as significant. It also remains unclear how many locomotion onsets come from each mouse; the results could be dominated by a small subset of mice with the most locomotion onsets. The more disciplined approach to statistical analysis of the dataset is to average the data associated with locomotion onsets within a mouse, and then use the mouse as an independent unit for statistical comparison. A second example, for instance, is in Figure 2L, where the independent statistical unit is defined as cortical regions instead of mice, with the left and right hemispheres counting as independent samples; again this is not justified. Is the activity of cortical regions within a mouse and across cortical hemispheres really statistically independent? The problem is apparent throughout the manuscript and for each data set collected.

This may partially be a misunderstanding. Figures 1F-1K indeed use locomotion onsets as a unit, but there were no statistical comparisons. In these Figures we were addressing the question of whether locomotion onsets in closed loop differ from those in open loop. Thus, we quantify variability as a unit of locomotion onsets. The question of mouse-to-mouse variability of this analysis is a slightly different one. We did include the same analysis (for visual cortex) with the variability calculated across mice as Figure S2. We will expand this supplementary figure with the equivalent data of Figure 3 to further address this concern.

For Figure 1L (we assume the reviewer means Figure 1L, not Figure 2L), the unit we used for analysis was cortical area. We will update and improve the analysis. This was indeed not optimal, and we will replace the statistical testing with hierarchical bootstrap (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7906290/) to account for nested data.

An additional statistical issue is that it is unclear if the authors are correcting for the use of multiple statistical tests (as in for example Figure 1L and Figure 2B,D). In general, the use of statistics by the authors is not justified in the text.

We will update and improve the analysis shown in Figure 1L.

In Figures 2B and 2D, we think adding family-wise error correction would be slightly misleading. We could add a correction – our conclusions would remain unchanged almost independent of the choice of correction (most of the significant p values are infinitesimally small, see Table S1). However, our interpretation is not focusing on one particular comparison (of many possible comparisons) that is significant - all comparisons between closed and open loop data points were significant in the L5 IT recordings and none of them were significant in the recordings in C57BL/6 mice that expressed GCaMP brain-wide.

Finally, it is important to note that whilst the study demonstrates that antipsychotics may selectively impact visuomotor integration in L5 neurons, it does not show that this effect is necessary or sufficient for the action of antipsychotics; though this is likely beyond the scope of the study it is something for readers to keep in mind.

We fully agree, it is still unclear how the effects we observe in our work relate to the treatment relevant effects in patients. We will expand on this point in the discussion.

Reviewer #3 (Public Review):

The study examines how different cell types in various regions of the mouse dorsal cortex respond to visuomotor integration and how antipsychotic drugs impacts these responses. Specifically, in contrast to most cell types, the authors found that activity in Layer 5 intratelencephalic neurons (Tlx3+) and Layer 6 neurons (Ntsr1+) differentiated between open loop and closed loop visuomotor conditions. Focussing on Layer 5 neurons, they found that the activity of these neurons also differentiated between negative and positive prediction errors during visuomotor integration. The authors further demonstrated that the antipsychotic drugs reduced the correlation of Layer 5 neuronal activity across regions of the cortex, and impaired the propagation of visuomotor mismatch responses (specifically, negative prediction errors) across Layer 5 neurons of the cortex, suggesting a decoupling of long-range cortical interactions.

The data when taken as a whole demonstrate that visuomotor integration in deeper cortical layers is different than in superficial layers and is more susceptible to disruption by antipsychotics. Whilst it is already known that deep layers integrate information differently from superficial layers, this study provides more specific insight into these differences. Moreover, this study provides a first step into understanding the potential mechanism by which antipsychotics may exert their effect.

Whilst the paper has several strengths, the robustness of its conclusions is limited by its questionable statistical analyses. A summary of the paper's strengths and weaknesses follow.

Strengths:

The authors perform an extensive investigation of how different cortical cell types (including Layer 2/3, 4 , 5, and 6 excitatory neurons, as well as PV, VIP, and SST inhibitory interneurons) in different cortical areas (including primary and secondary visual areas as well as motor and premotor areas), respond to visuomotor integration. This investigation provides strong support to the idea that deep layer neurons are indeed unique in their computational properties. This large data set will be of considerable interest to neuroscientists interested in cortical processing.

The authors also provide several lines of evidence that visuomotor information is differentially integrated in deep vs. superficial layers. They show that this is true across experimental paradigms of visuomotor processing (open loop, closed loop, mismatch, drifting grating conditions) and experimental manipulations, with the demonstration that Layer 5 visuomotor integration is more sensitive to disruption by the antipsychotic drug clozapine, compared with cortex as a whole.

The study further uses multiple drugs (clozapine, aripiprazole and haloperidol) to bolster its conclusion that antipsychotic drugs disrupt correlated cortical activity in Layer 5 neurons, and further demonstrates that this disruption is specific to antipsychotics, as the psychostimulant amphetamine shows no such effect.

In widefield calcium imaging experiments, the authors effectively control for the impact of hemodynamic occlusions in their results, and try to minimize this impact using a crystal skull preparation, which performs better than traditional glass windows. Moreover, they examine key findings in widefield calcium imaging experiments with two-photon imaging.

Weaknesses:

A critical weakness of the paper is its statistical analysis. The study does not use mice as its independent unit for statistical comparisons but rather relies on other definitions, without appropriate justification, which results in an inflation of sample sizes. For example, in Figure 1, independent samples are defined as locomotion onsets, leading to sample sizes of approx. 400-2000 despite only using 6 mice for the experiment. This is only justified if the data from locomotion onsets within a mouse is actually statistically independent, which the authors do not test for, and which seems unlikely. With such inflated sample sizes, it becomes more likely to find spurious differences between groups as significant. It also remains unclear how many locomotion onsets come from each mouse; the results could be dominated by a small subset of mice with the most locomotion onsets. The more disciplined approach to statistical analysis of the dataset is to average the data associated with locomotion onsets within a mouse, and then use the mouse as an independent unit for statistical comparison. A second example, for instance, is in Figure 2L, where the independent statistical unit is defined as cortical regions instead of mice, with the left and right hemispheres counting as independent samples; again this is not justified. Is the activity of cortical regions within a mouse and across cortical hemispheres really statistically independent? The problem is apparent throughout the manuscript and for each data set collected.

An additional statistical issue is that it is unclear if the authors are correcting for the use of multiple statistical tests (as in for example Figure 1L and Figure 2B,D). In general, the use of statistics by the authors is not justified in the text.

Finally, it is important to note that whilst the study demonstrates that antipsychotics may selectively impact visuomotor integration in L5 neurons, it does not show that this effect is necessary or sufficient for the action of antipsychotics; though this is likely beyond the scope of the study it is something for readers to keep in mind.

Behavior factors seems like an artificial distinction, at least based on these examples. These would be better classified as Knowledge factors. Drawing a pattern that you've memorized is conceptually no different than typing a code. Or should I point out that typing a code is also a behavior? You have to press your fingers in a certain location on your keyboard and in a certain order.

It is important that it work on Windows, but three-quarters of all PCs do not run Windows. Three-quarters of all traditional laptops and desktops? Sure, but most personal computers these days are mobile phones.

巨帧设置：

帧的大小主要是会影响相机的Packet Size的大小，MVS会跟据PC的巨帧大小自动调整Packet Size的大小。我们一般设置巨型帧为9K，即9014字节。巨型帧设置路径：网络和共享中心>本地连接>属性配置>高级>巨型帧（巨帧数据包、大型数据包、Jumbo Frames、Jumbo Mtu、MTU等），若不设置，则相机很容易丢包。

环境设置：

III. 关闭Windows防火墙 操作方法：控制面板>Windows防火墙>打开或关闭Windows 防火墙 （1）需要确认每个网络对应的防火墙都关闭 （2）有些PC装完系统后windows防火墙是未启用的，需要启用任务管 理器-服务中防火墙后，再通过以上方法关闭，这样防火墙才算是关闭掉 IV. 确认千兆的网络运行环境 使用相机时，需要确保网络链接速度为千兆网络： （1）PC网卡需要是千兆网卡，且确保是以千兆的速度运行； （2）网线需要使用超五类或六类以上的千兆网线。 VI. 卸载杀毒、防护软件 VII. 安装相机的金属结构接地，或者PC工控机电源/外 壳接地：避免静电干扰

光电触发线： （从左到右，为 3 和 6口） 默认的蓝、紫线序为 line0. 默认光电为 NPN 还是 PNP 型，对应不同接线方式。

读码调节：

光圈值调节 • 相机位置不同，景深要求不同对应的光圈值也相应不同；大景深对应小 光圈值，小景深对应大光圈值。。 建议值：F8与F16中间，螺丝锁紧，如下图所示： 三、读码调节流程 TIPS： 光圈位置调节要求比较精确，请务 必按照图示中的位置调节，偏差一 点就会影响到后续的调试

海康的 ID6000系列读码相机。

关闭Windows防火墙 操作方法：控制面板>Windows防火墙>打开或关闭Windows 防火墙 （1）需要确认每个网络对应的防火墙都关闭 （2）有些PC装完系统后windows防火墙是未启用的，需要启用任务管 理器-服务中防火墙后，再通过以上方法关闭，这样防火墙才算是关闭掉 IV. 确认千兆的网络运行环境 使用相机时，需要确保网络链接速度为千兆网络： （1）PC网卡需要是千兆网卡，且确保是以千兆的速度运行； （2）网线需要使用超五类或六类以上的千兆网线。 VI. 卸载杀毒、防护软件 VII. 安装相机的金属结构接地，或者PC工控机电源/外 壳接地：避免静电干扰

Author Response

Reviewer #1 (Public Review):

In this interesting manuscript, Nasser et al explore long-term patterns of behavior and individuality in C. elegans following early-life nutritional stress. Using a rigorous, highly quantitative, high-throughput approach, they track patterns of motor behavior in many individual nematodes from L1 to young adulthood. Interestingly, they find that early-life food deprivation leads to decreased activity in young larvae and adults, but that activity between these times, during L2-L4, is largely unaffected. Further, they show that this "buffering" of stress requires dopamine signaling, as L2-L4 activity is significantly reduced by early-life starvation in cat-2 mutants. The paper also provides evidence that serotonin signaling has a role in modulating sensitivity to stress in L1 larvae and adults, but the size of these effects is modest. To evaluate patterns of individuality, the authors use principal components analysis to find that three temporal patterns of activity account for much of the variation in the data. While the paper refers to these as "individuality types," it may be more reasonable to think of these as "dimensions of individuality." Further, they provide evidence that stress may alter the strength and/or features of these dimensions. Though the circuit mechanisms underlying individuality and stress-induced changes in behavior remain unknown, this paper lays an important foundation for evaluating these questions. As the authors note, the behaviors studied here represent only a small fraction of the behavioral repertoire of this system. As such, the findings here are an interesting and very promising entry point for a deeper understanding of behavioral individuality, particularly because of the cellular/synaptic-level analysis that is possible in this system. This paper should be of interest to those studying C. elegans behavior and also more generally to those interested in behavioral plasticity and individuality.

We thank the reviewer for finding our results interesting.

Reviewer #2 (Public Review):

This paper set out to understand the impact of early life stress on the behavior and individuality of animals, and how that impact might be amplified or masked by neuromodulation. To do so, the authors built on a previously established assay (Stern et al 2017) to measure the roaming fraction and speed of individuals. This technique allowed the authors to assess the effects of early life starvation on behavior across the entire developmental trajectory of the individual. By combining this with strains with mutant neuromodulatory systems, this enabled the authors to produce a rich dataset ripe for analysis to analyze the complicated interactions between behavior, starvation intensity, developmental time, individuality, and neuromodulatory systems.

The richness of this dataset - 2 behavioral measures continuous across 5 developmental stages, 3 different neuromodulatory conditions (with the dopamine system subject to decomposition by receptor types) and 4 different levels of starvation, with ~50-500 individuals in each condition-underlies the strength of this paper. This dataset enabled the authors to convincingly demonstrate that starvation triggers a behavioral effect in L1 and adult animals that is largely masked in intermediate stages, and that this effect becomes larger with increased severity of starvation. Furthermore, they convincingly show that the masking of the effect of starvation in L2-L4 animals depends on dopaminergic systems. The richness of the dataset also allowed a careful analysis of individuality, though only neuromodulatory mutants convincingly manipulated individuality, recapitulating earlier research. Nonetheless, a few caveats exist on some of their findings and conclusions:

We thank the reviewer for the constructive comments. In the revised manuscript we include additional analyses and textual changes as detailed below, to address the points raised.

1) Lack of quantitative analysis for effects within developmental stages. In making the argument for buffered effects of starvation on behavior during periods of larval development, the authors make claims regarding the temporal structure of behavior within specific stages. However, no formal analysis is performed and and the traces are provided without confidence intervals, making it difficult to judge the significance of potential deviations between starvation conditions.

In the revised manuscript, we include additional analyses of roaming fraction effects across shorter developmental-windows, showing within-stage differences in behavioral patterns following starvation (Figure 1 - figure supplement 1E; Figure 3 - figure supplement 1C). In addition, we further temper and rewrite our conclusions to clearly describe these effects (now- “…while 1 day of early starvation modified within-stage temporal behavioral structures by shifting roaming activity peaks to later time-windows during the L2 and L3 stages…” in p. 4 and “Interestingly, during the L2 intermediate stage the effects on roaming activity patterns were more pronounced during earlier time-windows of the stage…” in p. 8).

2) Incorrect inferences from differences in significance demonstrating significant differences. The authors claim that there is an increase in PC1 inter-individual variation in tph-1 individuals, however the difference in significance is not evidence of a significant difference between conditions (see Nieuwenhuis et al. 2011). This undermines claims about an interaction of starvation, neuromodulators, and individuality.

In the revised manuscript we provide now a direct comparison of PCs inter-individual variances between starved and unstarved populations, demonstrating significant differences in inter-individual variation in specific PC individuality dimensions following stress (Figure 6 and Figure 6 - figure supplement 1). These results include the increase in PC1 inter-individual variation in tph-1 mutants following 3 and 4 days of starvation (Figure 6A,E).

3) Sensitivity of analysis to baseline effects and assumptions of additive/proportional effects. The neuromodulatory and stress conditions in this paper have a mixture of effects on baseline activity and differences from baseline. The authors normalize to the roaming fraction without starvation, making the reasonable assumption that the effect due to starvation is proportional to baseline, rather than an additive effect. This confound is most visible in the adult subpanel of figure 5d, where an ~2-3 fold difference in relative roaming due to starvation is clearly noted, however, this is from a baseline roaming fraction in tph-1 animals that are ~2 fold higher, suggesting that the effect could plausibly be comparable in absolute terms.

Unavoidably, any such assumptions on the expected interaction between multiple effects will be a gross simplification in complicated nonlinear systems, and the data are largely shown with sufficient clarity to allow the reader to make their own conclusions. However, some of the interpretations in the paper lean heavily on an assumption that the data support a direct interpretation (e.g. "neuronal mechanisms actively buffer behavioral alterations at specific development times") rather than an indirect interpretation (e.g. that serotonin reduces baseline roaming fraction which makes a fixed sized effect more noticeable). Parsing the differences requires either more detailed mechanistic study or careful characterization of the effect of different baselines on the sensitivity of behavior to perturbation-barring that it's worth noting that many of these interactions may be due to differences in biological and experimental sensitivity to change under different conditions, rather than a direct interaction of stress and neuromodulatory processes or evidence of differing neuromodulatory activity at different stages of development.

In the revised manuscript we added a discussion of the potential complicated interactions between neuromodulation and stress, altering baseline levels and deviations from baseline. We also discuss the interpretation of the results in the context of non-linear systems in which sensitivity of the behavioral response to underlying variations may be modified by specific neuromodulatory and environmental perturbations, without assuming direct differences in neuromodulatory states over development or across individuals (p. 16).

Reviewer #3 (Public Review):

In this study, Nasser et al. aim to understand how early-life experience affects 1) developmental behavior trajectory and 2) individuality. They use early life starvation and longitudinal recording of C. elegans locomotion across development as a model to address these questions. They focus on one specific behavioral response (roaming vs. dwelling) and demonstrate that early life (right after embryo hatching) starvation reduces roaming in the first larval (L1) and adult stages. However, roaming/dwelling behavior during mid-larval stages (L2 through L4) is buffered from early life starvation. Using dopamine and serotonin biosynthesis null mutant animals, they demonstrated that dopamine is important for the buffering/protection of behavioral responses to starvation in mid-larval stages, while in contrast, serotonin contributes to early-life starvation's effects on reduced roaming in the L1 and adult stages. While the technique and analysis approaches used are mostly solid and support many of the conclusions made in the manuscript for part 1), there are some technical limitations (e.g., whether the method has sufficient resolution to analyze the behaviors of younger animals) and confounding factors (e.g., size of the animal) that the authors do not yet sufficient address, and can affect interpretation of the results. Additionally, much of the study is descriptive and lacks deep mechanistic insight. Furthermore, the focus on a single behavioral parameter (dwelling vs. roaming) limits the broad applicability of the study's conclusions. Lastly, the manuscript does not provide clear presentation or analysis to address part 2), the question of how early life experience affect individuality.

We thank the reviewer for these important comments. As described below, in the revised manuscript we include new analyses (following extraction of size data), showing behavioral modifications across different conditions/genotypes also in size-matched individuals (within the same size range) (Figure 1 - figure supplement 1F; Figure 3 - figure supplement 1D,E; Figure 5 - figure supplement 1B,D). We also made edits to the text to describe these results (Methods p. 21 and Results section). In addition, while we can detect behavioral changes using our imaging method even in young L1 worms across conditions and genotypes (described in Stern et al. 2017 and this manuscript), as the reviewer correctly pointed out, we may miss some milder behavioral effects due to lower spatial imaging resolution in younger worms. We are now referring to this spatial resolution limitation in the revised manuscript (discussion part). Lastly, in the revised manuscript we added clearer and more direct analyses of changes in inter-individual variation in multiple PC dimensions following early stress, by directly comparing variation between starved and unstarved individuals within the mutant and wild-type populations (Figure 6; Figure 6 - figure supplement 1). These analyses show significant changes in inter-individual variation within specific PC individuality dimensions following early stress. Also, we made textual changes along the manuscript to increase the clarity of presentation of these results.

This work has been peer reviewed in GigaScience (see paper https://doi.org/10.1093/gigascience/giad025), which carries out open, named peer-review. These reviews are published under a CC-BY 4.0 license and were as follows:

Reviewer 1: Weilong Guo, PhD

Patrick König and colleagues have built a web application for the interactive query, visualization and analysis of genomic diversity data, supportting population structure analysis on specific genetic elements, and data export. The application can also be easily used as a plugin for existing web application. According to its documentation, this application can be easily installed form pip, Docker and conda, which would be useful for population genomic studies. There are still several concerns about this manuscript.

Major concerns:

1. As for the SNP visualization function, there are only very limited numbers of SNPs can be read on the webpage, without function such as "zoom in" or "zoom out"(it is suggested to add such functions or similar functions). Although the application can export almost all the SNP sites of a whole VCF file, it is far from user-friendly.It is suggested to add a track of chromosomes showing the genomic windows under querying, allowing the cursor to select or adjust the genomic regions (UCSC-browser style), which is necessary for an intuitive user experience.

2. The BLAST function could serve as a useful entry point. But what is the starting position of the query sequence when mapped on minus strand? The authors should make it more clearly explained on the website.

3. TThe authors mentioned that their application would convert the inputted VCF file into Zarr format. Thus, more performance evaluation should be declared to show the advantages of this strategy (rather than using the VCF file directly).

4. The authors should also compared the their applications with other similar existing web applications, such as CanvasDB, Gigwa, SNiPlay and SnpHub, to highlight their advantages and improvemences.

Minor concerns:

1. The analysis functions are still insufficient. Commonly used analysis tools or methods, such as haplotype analysis, STRUCTURE analysis, distribution of nucleotide diversity and selection sweep analysis, are also suggested to be supported.

2. Ref. 22 is not completed.

From LAWLER 216: Added in the typescript.

ZABROUSKI: After these sentences, Wilde revised the rest of the paragraph in 1891: "Others crowded round the swinging doors of the coffee-house in the piazza. The heavy cart-horses slipped and stamped upon the rough stones, shaking their bells and trappings. Some of the drivers were lying asleep on a pile of sacks. Iris-necked and pink-footed, the pigeons ran about picking up seeds. After a little while, he hailed a hansom and drove home. For a few moments he loitered upon the doorstep, looking round at the silent square, with its blank, close-shuttered windows and its staring blinds. The sky was pure opal now, and the roofs of the houses glistened like silver against it. From some chimney opposite a thin wreath of smoke was rising. It curled, a violet riband, through the nacre-coloured air."

See below the number of ways to define the same path on Windows

UNC paths on Windows

Paths on Windows are much more complex

ok

The quote raises questions about the nature of freedom and when it will truly arrive for the woman, as she wonders if she will ever be able to answer the question, "Can I live?" This uncertainty and the longing for a more genuine sense of freedom underscore the complex nature of American womanhood, as women throughout history have grappled with questions of freedom, rights, and equality.

eventlog chain

This is important information for everybody to be aware of. Even if you think something you're doing online is a secret, it isn't, and you need to be careful.

Moin, Herr Baumgartner,

ich persönlich halte dies nicht mehr für unbedingt nötig.

Jedenfalls solange bei DeppL (KI-basiert unter https://www.deepl.com/translator) in der freien Version 3000 Zeichen für eine einzelne Übersetzung weiterhin möglich bleiben. Bei einer Lesegeschwindigkeit von rd. 1200 Zeichen/min. sind das etwa 2,5 Minuten. Habe es mit lorem ipsum, auch mit einfügen in eine Word-Datei getestet, es bringt schon was. Für die Hilfe-Dateien von Obsidian auf alle Fälle. Und vielfach braucht man die Übersetzung ja eigentlich nur als Ergänzung.

Neben der Browser-Variante gibt es auch iOS und Android Apps, als Windows- und Mac-App und Erweiterung für ein paar Browser.

Ich habe die Android ab, da lassen sich die Übersetzungen auch speichern.

B. Frenzel

If I were required to live in such way, I would feel so sad, then angry, maybe try to find way to show my angry

that is wondful!!!

Language itself the technology that was meant for humans to talk to humans. People complain about social media sites' bot populaces. If sex spam bots can degrade the Tumblr experience and crypto spam bots can degrade the Twitter experience will these new bots degrade the language experience?

I never thought of talk shows as windows into imagined Americas before but it is very interesting to think about

I'm super intrigued by the use/utility of simulation here. I know that these simulations were conducted using msprime, but I can't help but wonder about the inclusion of varying intensities of positive selection on loci in these simulations. This of course would increase the complexity of parameter space significantly with regards to potential simulations, but seems explicitly tied to the hypotheses being tested here.

More generally, it seems to me that this framework could lend itself nicely to the application of machine or deep learning approaches for assignment of genomic windows to alternative evolutionary histories (e.g. migration, selection), as well as potentially even parameter inference, such as through the use of a CNN. Obviously this is outside of the scope of the current paper, but I'm just curious as to whether this is a thought/space you all have explored?

The mention of the Menagerie in Bentham's work may be intended as a metaphor for the way in which power operates through the control and display of bodies, whether they be animal or human.

The idea behind the Panopticon was to create a prison design that would allow a single watchman to observe all of the prisoners without the prisoners knowing when they were being watched.

He’s like a teacher taking roll during a fire drill lol

Another just like when Covid was here, many people were spraying down houses when someone in their house was sick and when they did they did not leave a mark uncleaned or untouched.

This shows that with tesla coming out with all this new tech that some people are still not familiar with it.

To capture a Youtube video screenshot without the player controls, use the Hyde browser extension and the keystroke Ctrl+M to hide/unhide the controls.

Further, one can use the Windows Snipping Tool via Win-Shift-S to effect a screenshot to the clipboard for saving and editing.

https://youtubedownload.minitool.com/youtube/how-to-hide-youtube-bar.html

Note: This rebuttal 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

General comments:

We thank the reviewers for recognizing the importance of our work and for their supportive and insightful comments.

Our planned revisions focus on addressing all the comments and especially in further elucidating the molecular mechanism underpinning our observations, their consequences for cell phenotypes and reproducing our observations in an additional cell line. Our revision plan is backed up in many cases by preliminary data.

Our submitted manuscript demonstrated that DNMT3B’s recruitment to H3K9me3-marked heterochromatin was mediated by the N-terminal region of DNMT3B. Data generated since submission suggest that DNMT3B binds indirectly to H3K9me3 nucleosomes through an interaction mediated by a putative HP1 motif in its N-terminal region.

Specifically, we have found that DNMT3B can pull down HP1a and H3K9me3 from cell extracts and that this interaction is abrogated when we remove the N-terminal region of DNMT3B (revision plan, figure 1a). Using purified proteins in vitro, we have shown binding of DNMT3B to HP1a that is dependent on the presence of DNMT3B’s N-terminus suggesting that the interaction with HP1a is direct and that this mediates DNMT3B’s recruitment to H3K9me3 (revision plan, figure 1b). Alphafold multimer modelling identified that DNMT3B's N-terminus binds the interface of a HP1 dimeric chromoshadow domain through a putative HP1 motif. Two point mutations in this motif ablate DNMT3B’s interaction with HP1a in vitro (revision plan, figure 1b - DNMT3B L166S I168N).

We propose to further characterize DNMT3B’s interaction with HP1a in vitro and determine the significance of these observations in cells by microscopy in a revised manuscript. Together with the other proposed experiments and analyses, we believe the extra detail regarding the molecular mechanisms through which DNMT3B is recruited to H3K9me3 heterochromatin will help address the reviewer’s comments.

Point by point response:

We have reproduced the reviewer’s comments in their entirety and highlighted them in blue italics.

Reviewer #1 (Evidence, reproducibility and clarity (Required)): Summary:

This paper by Francesca Taglini, Duncan Sproul, and their coworkers, examines the mechanisms of DNA methylation in a human cancer cell line. They use the human colorectal cancer line HCT116, which has been very widely used to look at epigenetics in cancer, and to dissect the contribution of different proteins and chromatin marks to DNA methylation.

The authors focus on the role of the de novo methyltransferase DNMT3B. It has been shown in ES cells in 2015 that its PWWP domain directs it to H3K36me3, typically found in gene bodies. More recently, the authors showed similar conclusions in colorectal cancer (Masalmeh Nat Comm 2021). Here they examine, more specifically, the role of the PWWP. The conclusions are described below.

Major comments:

• *1-I feel that this paper has several messages that are somewhat muddled. The main message, as expressed in the title and in the model, is that the PWWP domain of DNMT3B actively drags the protein to H3K36me3-marked regions. Inactivation of this domain by a point mutation, or removal of the Nter altogether, causes DNMT3B to relocate to other genomic regions that are H3K9me3-rich, and that see their DNA methylation increase in the mutant conditions. This first message is clear.

We thank the reviewer for their positive comments on our observations. However, we note that our results suggest that removal of the N-terminal region has a different effect to point mutations in the PWWP domain. The data we present suggest that the N-terminus facilitates recruitment to H3K9me3 regions.

The second message has to do with ICF. A mutant form of DNMT3B bearing a mutation found in ICF, S270P, is actually unstable and, therefore, does not go to H3K9me3 regions. I feel that here the authors go on a tangent that distracts from message #1. This could be moved to the supp data. At any rate, HCT116 are not a good model for ICF. In addition, a previous paper has looked at the S270P mutant, and it did not seem unstable in their hands (Park J Mol Med 2008, PMID: 18762900). So I really feel the authors do not do themselves a favor with this ICF angle.

While we agree with the reviewer that HCT116 cells as a cancer cell line are not a good model for ICF1 syndrome, our observation that S270P destabilizes DNMT3B is important to consider in the context of this disease. In addition, the S270P mutant was reported to abrogate the interaction between DNMT3B and H3K36me3 (Baubec et al 2015 Nature PMID: 25607372) making it important to compare it to the other mutations we examine. In our revised version of the manuscript, we propose to move these data to the supplementary materials and add a statement to the discussion noting the caveat that HCT116 cells are likely not to model many aspects of ICF1.

With regard to the differences between our results and that of Park et al, we note that stability of the S270P mutant was not assessed in that study whereas we directly assess stability in vitro and in cells. We propose to add discussion of this previous study to the revised manuscript.

2-I feel that some major confounders exist that endanger the conclusions of the work. The most worrisome one, in my opinion, is the amount of WT or mutant DNMT3B in the cells. It is clear in figure 4C that the WT rescue construct is expressed much more than the W263A mutant (around 3 or 4 times more). Unless I am mistaken, we are never shown how the level of exogenous rescue protein compares to the level of DNMT3B in WT cells. This bothers me a lot. If the level is too low, we may have partial rescue. If it is too high, we might have artifactual effects of all that DNMT3B. I would also like to see the absolute DNA methylation values (determined by WGBS) compared to the value found in WT. From figure S1A, it looks like WT is aroun 80% methylation, and 3BKO is around 77% or so. I wonder if the rescue lines may actually have more methylation than WT?

The rescue cell lines do express DNMT3B to a greater level than observed endogenously. In our manuscript we controlled for this effect by generating the knock-in W263A cells and, as reported in the manuscript, we observe similar effects to the rescue cells (manuscript, figure 2d) suggesting that our observations are not driven by the overexpression.

We also expressed ectopic DNMT3B from a weaker promoter (EF1a) in DNMT3B KO cells but did not include these data in the submitted manuscript. We have previously shown that this promoter expresses DNMT3B at lower levels than the CAG promoter used in the submitted manuscript (Masalmeh et al 2021 Nature Communications PMID: 33514701). Bisulfite PCR of representative non-repetitive loci within heterochromatic H3K9me3 domains show that we observe similar gains of methylation with DNMT3BW263A (revision plan, figure 2).

Revision plan figure 2. Expression of DNMT3BW236A from a weaker promoter leads to increased DNA methylation at selected H3K9me3 loci. Barplot of mean methylation by BS-PCR at H3K9me3 loci alongside the H3K4me3-marked BRCA2 promoter in DNMT3B mutant cells were DNMT3B is expressed from the EF1 promoter. P-values are from two-sided Wilcoxon rank sum tests.

To reinforce that our conclusions are not solely a result of the level of DNMT3B expression, we propose to include these data in the revised manuscript.

The reviewer is also correct that by WGBS, the rescue cell lines have higher levels of overall DNA methylation than HCT116 cells. We will note this in revised manuscript and include HCT116 cells in a revised version of Figure S1e.

3-I guess the unarticulated assumption is that the gain of DNA methylation seen at H3K9me3 region upon expression of a mutant DNMT3B is due to DNMT3B itself. But we do not know this for sure, unless the authors test a double mutant (PWWP inactive, no catalytic activity). I am not necessarily asking that they do it, but minimally they should mention this caveat.

The hypothesis that the gains in DNA methylation at H3K9me3 loci result from the direct catalytic activity of DNMT3B is supported by our observation that a catalytic dead DNMT3B does not remethylate heterochromatin (manuscript, figures 1d and e). However, we acknowledge that we have not formally shown that the additional DNA methylation seen with DNMT3BW263A are a direct result of its catalytic activity. We will conduct an analysis of the effect of catalytically dead DNMT3BW263A on DNA methylation at Satellite II and selected H3K9me3 loci and include this in the revised manuscript.

4-I am confused as to why the authors look at different genomic regions in different figures. In figure 1 we are looking at a portion of the "left" arm of chr 16. But in figure 2B, we now look at a portion of the "right" arm of the same chromosome, which has a large 8-Mb block of H3K9me3, and is surprisingly lowly methylated in the 3BKO. This seems quite odd, and I wonder if there is a possible artifact, for instance mapping bias, deletion, or amplification in HCT116. Showing the coverage along with the methylation values would eliminate some of these concerns.

By choosing different regions of the genome for different figures, we intended to reassure the reader that our results were not specific to any one region of the genome. In the revised manuscript, we propose to display a consistent genomic region between these figures.

With regard to the low levels of DNA methylation in H3K9me3 domains in DNMT3B KO cells, H3K9me3 domains are partially methylated domains which have reduced methylation in HCT116 cells (see page 5 of the manuscript):

… we found that hidden Markov model defined H3K9me3 domains significantly overlapped with extended domains of overall reduced methylation termed partially methylated domains (PMDs) defined in our HCT116 WGBS (Jaccard=0.575,p=1.07x10-6, Fisher’s test).

These domains lose further DNA methylation in DNMT3B KO cells leading to the low methylation level noted by the reviewer. The methylation percentages calculated from WGBS are based on the ratio of methylated to total reads. Thus, a lack of coverage generates errors from division by zero rather than the low values observed in this domain in DNMT3B KO cells.

We include a modified version of figure 2b from the manuscript below. This includes coverage for the 3 cell lines (revision plan, figure 3). Although WGBS coverage is slightly reduced in H3K9me3 domains, reads are still present and overall coverage equal between different cell lines.

While we could potentially include the coverage tracks in revised versions of figures, we note that doing so for multiple cell lines would make these figures extensively cluttered and it would likely be difficult to observe the differences in DNA methylation in these figure panels due to shrinkage of the other tracks.

Minor comments:

1-The WGBS coverage is not very high, around 2.5X on average, occasionally 2X. I don't believe this affects the findings, as the authors look at large H3K9me3 regions. But the info in table S2 was hard to find and it is important. I would make it more accessible.

In the revised manuscript we will specify the mean coverage in the text to ensure this is clearer.

2-It would be nice to have a drawing showing exactly what part of the Nter was removed.

We will add this in the figure in the revised manuscript.

3-some figures could be clearer. I was not always sure when we were looking at a CRISPR mutant clone (W263A) versus a piggyBac rescue.

In the revised manuscript we will clarify in the figure labels to ensure it is clear which data were generated using CRISPR clones.

4-unless I am mistaken, all the ChIP-seq data (H3K9me3, H3K36me3 etc) come from WT cells. It is not 100% certain that they remain the same in the 3BKO, is it? This should be discussed.

We performed ChIP-seq on both HCT116 and 3BKO cell lines and used ChIP-seq data from the 3BKO cell line for the rescue experiments where DNMT3Bs were expressed in 3BKO cells. We will ensure this is clearer in the revised version.

Reviewer #1 (Significance (Required)):

Strengths:

The experiments are for the most part well done and well interpreted (save for the limitations mentioned above). The techniques are appropriate and well mastered by the team. The paper is well written, the figures are nice. The authors know the field well, which translates into a good intro and a good discussion. The bioinformatics are convincing.

Limitations:

All the work is done in a single cancer cell line. One might assume the conclusions will hold in other systems, but there is no certainty at this point.

We acknowledge this limitation. To demonstrate that our results are applicable beyond HCT116 cells, we will include analysis of experiments on an independent cell line in the revised manuscript.

HCT116 are not the best model system to study ICF, which mostly affects lymphocytes

At present, I feel that the biological relevance of the findings is fairly unclear. The authors report what happens when DNMT3B has no functional PWWP domain. I am convinced by their conclusions, but what do they tell us, biologically? Are there, for instance, mutant forms of DNMT3B expressed in disease that have a mutant PWWP? Are there isoforms expressed during development or in certain cell types that do not have a PWWP? In these cell types, does the distribution of DNA methylation agree with what the authors predict?

As stated in response to point 1, although we acknowledge the limitations of HCT116 cells as a model of ICF, we believe are finding that the S270P mutation results in unstable DNMT3B are still important to consider for ICF syndrome.

We are not aware of reports of mutations affecting the residues of DNMT3B’s PWWP domain we have studied. Our preliminary analysis suggests that although mutations in DNMT3B’s PWWP domain are frequent, residues in the aromatic cage such as W263 and D266 are absent from the gnomAD catalogue (Karczewski et al 2020 Nature, PMID: 32461654). This suggests that they are incompatible with healthy human development.

A number of different DNMT3B splice isoforms have been reported. These include DDNMT3B4 which lacks the PWWP domain and a portion of the N-terminal region (Wang et al 2006 International Journal of Oncology, PMID: 16773201). DDNMT3B4 is proposed to be expressed in non-small cell lung cancer (Wang et al 2006 Cancer Research, PMID: 16951144).

We will include analysis of gnomAD and discussion of these points in the revised manuscript.

In its present state, I feel the appeal of the findings is towards a semi-specialized audience, that is interested in aberrant DNA methylation in cancer and other diseases. This is not a small audience, by the way.

We thank the reviewer for their comments and the suggestion that our findings are of interest to a cross-section of researchers.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

Note, we have added numbers to the comments made by reviewer 2 to aid cross-referencing.

In this manuscript, Taglini et al., describe an increased activity of DNMT3B at H3K9me3-marked regions in HCT cells. They first identify that DNA methylation at K9me3-marked regions is strongly reduced in absence of DNMT3B. Next, the authors re-express DNMT3B and DNMT3B mutant variants in the DNMT3B-KO HCT cells and assess DNA methylation by WGBS where they identify a strong preference for re-methylation of K9me3 sites. Based on genome-wide binding maps for DNMT3B, including the mutant variants, they address how the localization of DNMT3B relates to the observed changes in methylation.

Major points:

• The authors show increased reduction of mCG at H3K9me3 (and K27me3) sites in absence of DNMT3B. This is based on correlating delta %mCG with histone modifications in 2kb bins. I find this approach to not fully support the major claim. First, the correlation coefficients are very small -0.124 for K9me3 and -0.175 for K27me3, and just marginally better compared to, for example, K36me3 that does not seem to have any influence on mCG according to Sup Fig S1b. While I agree that mCG seems more reduced at K9me3 in absence of DNMT3B (e.g. in Fig 1a), is there a better way to visualize the global effect? The delta mCG Boxplots based on bins are not ideal (this applies to many figures using the same approach in the current manuscript).*

Our choice to examine the global effects using correlations in windows across the genome was motivated by similar previous analyses in other studies (for example: Baubec et al. 2015 Nature PMID 25607372, Weinberg at al 2021 Nature PMID:33986537, Neri et al 2017 Nature PMID: 28225755). These global analyses result in modest correlation coefficients because the vast majority of genomic windows are negative for a given mark. For this reason, we included specific analyses of H3K36me3, H3K9me3 and H3K27me3 domains in the manuscript (eg manuscript figure 1b, c and d) which reinforce the conclusions drawn from our global analyses.

However, we acknowledge that while our data support a specific activity at H3K9me3 marked heterochromatin, these are not the only changes in DNMT3B KO cells as DNMTs are promiscuous enzymes that are localized to multiple genomic regions. We will add discussion of this point to the revised manuscript.

2. Second, the calculation based on delta mCpG does not allow to see how much methylation was initially there. For example, S1b shows a median decrease of ~ 10% in K9me3 and ~7-8% in H3K4me3. What does this mean given that the starting methylation for both marks is completely different?

Following this point, the authors mention that mCG is already low at K9me3 domains in HCT cells (compared to other sites in the genome). I am curious if this may influence the accelerated loss of methylation in absence of DNMT3B? Any comments on this?

The observation that there is a greater loss at H3K9me3 domains than H3K27me3-only domains which also have low DNA methylation levels in HCT116 argue that the losses are not solely driven by the lower initial level of methylation in H3K9me3 domains. Our analyses later in the manuscript also support a specific activity at H3K9me3. In addition, we propose to reinforce this point through further data on exploring how DNMT3B interacts with HP1a (see general comments, revision plan figure 1).

However, we acknowledge the possibility that part of the loss seen at H3K9me3 domains in DNMT3B KO cells could be in part a result of their low initial level of methylation. In the revised manuscript we propose to include discussion of this possibility.

3. One issue is the lack of correlation in DNMT3B binding to H3K9me3 sites in WT cells (Fig 3). How does this explain the requirement for DNMT3B for maintenance of methylation at H3K9me3? While some of the tested mutants show some weak increase at K9me3 sites, these are not comparable to the strong binding preferences observed at K36me3 for the wt or delta N- term version.

Using ChIP-seq we cannot say that DNMT3BWT does not bind at H3K9me3, only that it binds here to a lower level than at K36me3-marked loci. The normalized DNMT3BWT signal at H3K9me3 domains is higher than the background signal from DNMT3B KO cells (manuscript figure 3d) supporting the hypothesis that DNMT3BWT localizes to H3K9me3. This hypothesis is also supported by the observation that the correlation between DNMT3BDN and H3K9me3 is reduced compared to that of DNMT3BWT (manuscript figure 6c compared to figure 3c).

There are several reasons why the apparent enrichment of DNMT3B at H3K9me3 may appear weaker than at H3K36me3 by ChIP-seq. Previous work has also suggested that formaldehyde crosslinking fails to capture transient interactions in cells (Schmiedeberg et al. 2009 PLoS One PMID: 19247482). H3K9me3-marked heterochromatin is also resistant to sonication (Becker et al. 2017 Molecular Cell PMID: 29272703) and this could further affect our ability to detect DNMT3B in these regions using ChIP-seq. Our new data also suggest that DNMT3B binds to H3K9me3 indirectly through HP1a (see general comments, revision plan figure 1) and this may also lead to weaker ChIP-seq enrichment at H3K9me3 compared to the direct interaction with H3K36me3 through DNMT3B’s PWWP domain.

We propose to add discussion of these issues to the revised manuscript.

4. Following the above comment, what about other methyltransferases in HCT cells? Could DNMT1 or DNMT3A function be altered in absence of DNMT3B, and the observed methylation changes could be indirectly influenced by DNMT3B? The authors could create a DNMT-TKO HCT cell line and re-introduce DNMT3B in this background and measure methylation to exclude that DNMT1 or DNT3A could have an influence. In this case, only H3K9me3 should gain DNA methylation.

As discussed in response to reviewer 1 (point 3), we propose to examine the changes in DNA methylation upon expression of catalytically dead DNMT3BW263A to further strengthen the evidence that DNMT3BW263A is directly responsible for the increased DNA methylation at H3K9me3-marked loci.

5. DNMT3B lacking N-terminal shows reduced K9me3 methylation & some localization by imaging. While the presented experiments show some support for this conclusion, I suggest to re-introduce a W263A mutant lacking the N-terminal part and measure changes in DNA methylation at H3K9. This should help to test the requirement for the N-terminal regions and further indicate which protein part (PWWP or N-term) is more important in regulating the balance between K9me3 and K36me3.

We have performed this experiment and the data are shown in manuscript figure S6c and d. The results of these experiments show that DNMT3BΔN+W263A cells showed less methylation at H3K9me3 loci than DNMT3BW263A cells, supporting a role for the N-terminus in recruiting DNMT3B to H3K9me3-marked heterochromatin. In the revised version, we will ensure that these data are more clearly indicated.

In the first paragraph of the discussion, the authors state: "Our results demonstrate that DNMT3B is recruited to and methylates heterochromatin in a PWWP- independent manner that is facilitated by its N-terminal region." Same statement is found in the abstract. This contradicts the ChIP-seq results that do not indicate a recruitment of DNMT3B to heterochromatin, and the N- terminal deletions are not fully supporting a role in this targeting since there is no localization to K9me3 to begin with. While changes in methylation are observed, it remains to be determined if this is indeed through direct DNMT3B delocalization or indirectly through influencing the remaining DNMTs.

As discussed above, there are several potential reasons why DNMT3B ChIP-seq signal at H3K9me3 is weak (reviewer 2, point 3). The additional experiments we propose to include in the revised manuscript could reinforce this statement by clarifying whether DNMT3B is directly responsible for methylating H3K9me3-marked regions (reviewer 1, point 3) and by delineating the role of the putative HP1a motif in DNMT3B’s N-terminal region (general comments, revision plan figure 1).

Reviewer #2 (Significance (Required)):

Advance: detailed analysis of DNMT3B mutants in relation to K9me3. Builds up on previous studies. Audience: specialised audience

We thank the reviewer for their insights.

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

• *In this work, Taglini et al. examine how the de novo DNA methyltransferase DNMT3B localizes to constitutive heterochromatin marked by the repressive histone modification H3K9me3. The authors utilize a previously generated DNMT3B KO colorectal carcinoma cell line, HCT116 to study recruitment and activity of DNMT3B at constitutive H3K9me3 heterochromatin. The authors noted preferential decrease of DNA methylation (DNAme) at regions of the genome marked with H3K9me3 in DNMT3B KOs. The authors then rescued the deficiency through overexpression of WT and catalytic dead DNMT3A/B and confirmed that DNA methylation increase at H3K9me3+ region in the WT DNMT3B, but not catalytically inactive mutant nor DNMT3A. To examine which protein domains may be mediating DNMT3B's recruitment to H3K9me3 regions, the authors designed a series of mutants, primarily focusing on the PWWP domain which normally recognizes H3K36me3. In the PWWP mutants, DNMT3B binding to the genome is altered, showing depletion at some H3K36me3-marked regions and gain at H3K9me3 heterochromatin, which coincides with DNAme increase at satellites. In contrast, the clinically relevant ICF1 mutation S270P, shows DNMT3B protein destabilization and no such loss of DNAme at heterochromatin. Finally, the authors truncate the N-terminal portion of DNMT3B, and saw that this region of the protein is necessary for heterochromatin localization and subsequent DNAme of H3K9me3+ regions.

The experiments are well done with extensive controls, and the results are interesting and convincing. The structure of the manuscript could be improved for clarity and flow - for example, the PWWP mutations and truncations should be mentioned and compared together. I also found the section on ICF1 mutant to be out-of-place.

As described above (reviewer 1, point 1), we propose to move these data to the supplementary materials in the revised manuscript.

• *

More emphasis should be placed on the N- terminal mutant as this region seems to be critical to heterochromatin recruitment, and this may address whether the interaction to H3K9me3 is direct or indirect.

As described above (general comments), the revised manuscript will include experiments clarifying the nature of DNMT3B’s interaction with H3K9me3. Our preliminary data support that it is an indirect interaction mediated through HP1a (revision plan figure 1).

Finally, while the epigenetic crosstalk is well-examined in this work, I would strongly urge the authors to add RNA-seq data to determine the transcriptional consequence of such chromatin disruptions (e.g. are repetitive sequences up-regulated in DNMT3B KOs?).

As suggested by the reviewer, we propose to generate and analyse RNA-seq data in the revised manuscript to understand the impact of DNMT3B on transcriptional programs.

Comments

1. A potential caveat to the study is the use of a single cell line - colorectal cancer cell HCT116 - to draw major conclusions on the function of DNMT3B. It is worth noting the Baubec et al. study examining DNMT3B recruitment to H3K36me3 was mainly performed in murine embryonic stem cells (mESCs). It would greatly strengthen the study if the authors could perform similar type of data analysis on an independent DNMT3B KO cell line. For example, does DNMT3B localize to H3K9me3 regions in WT mESCs?

As described above in response to reviewer 1, we will include analysis in an additional cell line in the revised manuscript to demonstrate that our results are generalizable beyond HCT116 cells.

2. Did the PWWP mutant W263A show the expected loss of DNAme at H3K36me3-marked regions? In other words, was there evidence of DNAme redistribution in loss at H3K36me3+ regions and inappropriate gain at H3K9me3+ regions? Please perform intersection analysis of DMRs with other epigenomic marks (e.g. H3K27me3, H3K36me3, CpG shores) in the PWWP mutants.

Our analysis of DNMT3B KO cells (manuscript figure s1d) show that losses of DNA methylation in these cells are not correlated with H3K36me3 in gene bodies suggesting that DNMT3A and DNMT1 are sufficient to compensate in maintaining their methylation in DNMT3B KO cells. To clarify this point for the DNMT3BW263A knock in clones, in the revised manuscript we will directly examine whether these cells show loss of methylation at H3K36me3 marked gene bodies in a similar analysis and add discussion of these results.

The study would also be strengthen greatly with the in addition of biochemical studies to confirm direct loss of binding, and possibly gain of H3K9me3 binding, in the DNMT3B PWWP mutants.

As detailed above (general comments, revision plan figure 1), our data suggest that DNMT3B interacts indirectly with H3K9me3 through an HP1 motif in its N-terminal region. We will undertake further biochemical studies on this interaction which will be included in the revised manuscript. Specifically we will focus on using EMSAs with synthetic nucleosomes to clarify the degree to which the HP1a interaction is responsible for binding of DNMT3B to H3K9me3 modified nucleosomes.

We also propose to undertake in vitro biochemical characterization of the effect of DNMT3B PWWP mutations on interaction with H3K36me3 using synthetic nucleosomes. However, we note that in the manuscript we have shown similar effects using two independent point mutations that are predicted to affect H3K36me3 binding (W263A and D266A) and deletion of the entire PWWP domain.

3. Examining the tracks in Figure 3A,B, the PWWP mutants showed almost indiscriminate increase across the genome, and not specifically to H3K9me3-marked regions. Would ask the authors to speculate as to why the ChIP-seq of DNMT3B mutants do not recapitulate the heterochromatin co-localization shown by immunofluorescence.

As discussed in response to reviewer 2 (point 3) we believe that the weak DNMT3B ChIP-seq signal at H3K9me3 loci is likely due to the nature of the interaction that DNMT3B has with chromatin in these regions. We will add discussion of these points to the revised manuscript.

4. It's a shame that the ICF1 mutation S270P was not characterized to the same extent as the PWWP mutants. Would consider adding WGBS for this clinically relevant mutation.

We have shown that this mutant does not produce stable protein in vitro or in our cells and we observe little difference in DNA methylation at selected loci. As WGBS is expensive, we believe that carrying out this experiment is not an efficient use of limited research resources.

5. Figure 7 - please draw in the ICF1 and the N-terminal mutations in the model figure. Also provide legends.

We will modify the manuscript to include these details in the revised manuscript.

Reviewer #3 (Significance (Required)):

This is an intersting study on a timely subject. It will be of interest to multiple fields from epigenetics to development and cancer. My expertise is in cancer, epigenetics, development.

We thank the reviewer for highlighting the broad interest of our study.

Dock terminal then use this

Author Response

Reviewer #1 (Public Review):

During the height of the Covid19-pandemic, there was great and widely spread concern about the lowered protection the screening programs within the cancer area could offer. Not only were programs halted for some periods because of a lack of staff or concern about the spreading of SARS CoV2. When screening activities were upheld, participation decreased, and follow-up of positive test results was delayed. Mariam El-Zein and coworkers have addressed this concern in the context of cervical screening in Canada, one of the rather few countries in the world with well organized, population-based, although regionalized, cervical screening program.

Comment 1: Despite the existence of screening registries, they choose to do this in form of a survey on the internet, to different professional groups within the chain of care in cervical screening and colposcopy. The reason for taking this "soft data" approach is somewhat diffuse.

We are happy to provide a counterargument to the reviewer’s concern about the “soft data” approach. Our unit – McGill’s Division of Cancer Epidemiology – is a major stakeholder in policymaking and cervical screening guideline development in Canada. It is one of the components in a McGill Task Force on COVID-19 and Cancer that has been widely engaged in assessing the pandemic’s impact on the entire spectrum of cancer control and care (examples: PMID: 33669102, PMID: 34843106). Canada is a country of continental size, and during the pandemic even travel between provinces was interrupted. It is only via a web-based survey that one could have captured the required information. We took advantage of our unit’s credibility and stature to secure a substantial response to the survey, which elicited a high level of detail.

The survey questionnaire instrument was thoughtfully developed with input from Canadian experts who are active in the field of cervical cancer prevention and involved in clinical care to comprehensively formulate informative questions (and practical, reasonable responses) underpinning each of the themes covered. Of note, some of these coinvestigators, having executive roles in relevant clinical professional bodies, advised our team on the logistics of circulating the survey to members. The administration of the survey was coordinated with the pertinent societies. Our aim was to provide an overall portrait across Canada of the extent of the harms to cervical cancer screening and treatment processes at the beginning of the COVID-19 pandemic (specifically a snapshot from mid-March to mid-August 2020), as perceived by professional groups in multiple health disciplines.

Indeed, as the reviewer mentioned, there are fully (i.e., for Saskatchewan) and partially (i.e., for British Columbia, Alberta, Manitoba, Ontario) organized cervical cancer screening programs in Canada in addition to opportunistic programs (i.e., for North West Territories, Yukon, Nunavut, Quebec). The Canadian Partnership Against Cancer also collects information on cervical cancer screening programs and/or strategies across Canada. Using data from these different sources enables a quantitative assessment of the impact of the pandemic on cervical cancer screening, but this was not the research methodology used; the survey approach was our research strategy as we attempted to collect responses from all provinces and territories, regardless of the different screening programs and modalities implemented across the country, and including regions that do not have an official screening program.

Since the effects of the COVID-19 pandemic will stay with us for years to come, our research team is also examining – using a “hard data” approach via administrative healthcare datasets – the long-term effects that will accrue on cervical cancer morbidity and mortality from the interruptions and delays in screening processes and other activities in the process of care. A discussion of this is, however, beyond the scope and objectives of our manuscript.

No modifications were made in the manuscript to address this comment.

Comment 2: The authors claim they want to "capture modifications". However, the suggestions that come from this study are limited and are submitted for publication 2 years after the survey when the height of the pandemic has passed long since, and its burden on the screening program has largely disappeared. The value of the study had been larger if either the conclusions had been communicated almost directly, or if the survey had been done later, to sum up the total effect of the pandemic on the Canadian cervical screening program.

We appreciate this comment. As part of our commitment to transparency, we now plainly acknowledge that considerable time (1.5 years) has elapsed between the time the survey data were available (March 2021) and manuscript submission (September 2022) for publication in the special issue, curated by eLife, on the impact of the COVID-19 pandemic on cancer prevention, control, care and survivorship. However, we also argue that this lag time is reasonable given the undertaking of data management, analysis, and reporting of a large amount of data, including the synthesis of replies to open-ended questions. We also took this opportunity to expose two graduate students to the research process.

Changes made: Page 15, Lines 437-440.

In terms of assessing the total effect of the pandemic on the Canadian cervical screening program, this work is in progress, but not within the current manuscript. The PubMed references mentioned above show examples of directions we are taking. Also, as mentioned in our response 1 to comment 1, we will use data from administrative healthcare datasets (medical and drug claims, hospitalization data, death registry data) and hospital cancer registries (clinical characteristics such as cancer stage, grade, and biomarkers) on cancer patients diagnosed in Quebec between 2010 and 2026. Using these datasets, we intend to compare the pre- and post-pandemic eras in order to analyze changes in patterns of cancer care, cancer prognosis, and survival, including shifts at stage at diagnosis.

Comment 3: Another major problem with this study is the coverage. The results of persistent activities to get a large uptake is somewhat depressing although this is not expressed by the authors. 510 professionals filled out the survey partially or in total. 10 professions were targeted. The authors make no attempt to assess the coverage or the validity of the sample. They state the method used does not make that possible. But the number of family practicians, colposcopists, cytotechnicians, etc. involved in the program should roughly be known and the proportion of those who answered the survey could have been calculated. My guess is that it is far below 10%.

There were no extensive additional efforts to increase participation rate, apart from follow-up reminder emails to complete the survey, which is standard practice followed by the societies that administered the survey to their constituents. We respectfully disagree with the reviewer concerning coverage being a major limitation, particularly in view of the difficulty in general to secure a high response rate in a survey such as ours, at a time like the middle of the pandemic. Although it appears to be a seemingly easy to compute classic non-response rate, information on the “population of interest” (i.e., number of professionals approached in addition to the advertisement of the survey on social media platform”) is not available to estimate the extent of non-response. Even if the response rate is below 10% as suggested by the Reviewer, our survey and findings should be considered on their merits; the target population was involved in the survey design to ensure the validity of coverage of the questions along the continuum of care in cervical cancer screening and treatment. In addition, we followed the Checklist for Reporting Results of Internet E-surveys to inform the design, conduct, and reporting of our survey research.

Changes made: Page 14, Lines 421-425.

Comment 4: The national distribution seems shewed despite the authors boosting its pan-Canadian character. I am just faintly familiar with the Canadian regions, but, as an example, only 2 replies from Quebec must question the national validity of this survey.

We apologize for this typo error in Table 1; many cells were accidently shifted down (the last couple of provinces had the wrong numbers). There were actually 21 survey respondents from the province of Quebec. This has now been corrected.

Changes made: Page 19.

Comment 5: The result section is dominated by quantitative data from the responses to the 61 questions. All questions and their answers are tabulated. As there is no way to assess the selection bias of the answers these quantitative results have no real value from an epidemiological standpoint.

Indeed, we opted to provide the reader with descriptive results on all the questions and sub-questions that were asked, with explicit annotation to each question number and clear reference to the formulated question by appending the full survey instrument to the manuscript. We designed the survey as a descriptive and not an analytical study, contrary to traditional epidemiology studies that investigate a specific exposure-outcome relationship.

Changes made: Page 12, Lines 366-368.

In the spirit of other papers in the special issue on COVID-19 and cancer, curated by eLife, we measured the impact of the pandemic on the process of care like many other eLife articles did. The eLife collection is a snapshot of a period when not only was cancer control disrupted, but the ability to conduct valid research was also severely curtailed. The reviewer will likely agree that our paper is not the only one to suffer from these methodological shortcomings. Yet, taken together, the gestalt value of the eLife collection will inform epidemiologic modellers for the next long while on how this period affected cancer control. We are happy to contribute with this paper a few more pieces of the puzzle, adding to that which eLife published for many other jurisdictions.

Comment 6: The replies to the open-ended questions are summarized in a table and in the text. The main conclusion of the content analysis of the answers to the direct questions, and one of the main conclusions of the study, is that the majority favors HPV self-sampling in light of the pandemic. However, this not-surprising view is taken by only 80 responders while almost as many (n=60) had no knowledge about HPV self-sampling.

Another aim of our survey was to identify the windows of opportunity that were created by the pandemic and pinpoint positive aspects that could enable the transformation of cervical cancer screening (i.e., HPV primary based screening and HPV self-sampling). We found that 33% of respondents were of the opinion that the pandemic context could facilitate the implementation of self-sampling and that 50.1% were in favor of the implementation of this new screening practice (described in Results Theme 1: Screening Practice and Stable 5).

Changes made: Page 4, Lines 93-97.

The reviewer is correct that in the open-ended sub-question of Question 23 “Are you in favor of the implementation of HPV self-sampling as an alternative screening method in your clinical practice?”, 60 respondents justified their answer to the nominal question by their lack of familiarity with HPV self-sampling, compared to 80 who shared positive comments. However, we would like to draw the reviewer’s attention to the responses to the nominal part of the question in Stable 5. Of those who answered “Maybe”, 47.1% said that they were not familiar enough to express a favorable or unfavorable opinion. We would also like to draw the reviewer’s attention to the results of our cross-tabulation of profession and the question of relevance (described in Results Theme 1: Screening Practice). The lack of familiarity with novel screening practices such as self-sampling can be explained by the fact that most (75.0%) of those who expressed these views were primary healthcare professionals, and not secondary and tertiary specialists.

Changes made: Page 12, Lines 344-346

Comment 7: The authors conclude that their study identified the need for recommendations and strategies and building resilience in the screening system. No one would dispute the need, but the additional weight this study adds, unfortunately, is low, from a scientific standpoint.

Although no one would dispute the need as the reviewer is suggesting, but as epidemiologists we needed to collect this empirical evidence. We urge the reviewer to consider that this article is to contribute to a more complete picture of the collective process of discovery of the impact of the pandemic initiated by eLife’s special issue.

No modifications were made in the manuscript to address this comment.

Comment 8: The conclusion I draw from this study is that the authors have done a good job in identifying some possible areas within the Canadian screening programs where the SARS-Cov2 pandemic had negative effects and received some support for that in a survey. Furthermore, they listed a few actions that could be taken to alleviate the vulnerability of the program in a future similar situation, and received limited support for that. No more, no less.

We thank the Reviewer for the positive feedback provided in the first part of the comment. As for the rest, we believe we have addressed above the reviewer’s concerns.

Reviewer #2 (Public Review):

The study aimed to provide information on the extent to which the COVID-19 pandemic impacted cervical cancer (CC) screening and treatment in 3 Canadian provinces. The survey methodology is appropriate, and the results provide detailed descriptive statistics by province and type of practice. The results support the authors' conclusions. This evidence together with data gathered from other national surveys may provide baseline data on the impact of the pandemic on CC outcomes such as late-stage diagnoses and CC treatment outcomes due to these delays.

We are flattered by the Reviewer’s overall assessment of our manuscript.

Comment: This study relies mostly on descriptive statistics and open-ended questions that provide details about what CC screening and treatment procedures were delayed. It is unclear how the reader would use the results to affect current or future practice.

As mentioned in our reply above to a similar comment raised by reviewer 1, our overarching aim was to portray in a purely descriptive manner the negative and positive impacts of the COVID-19 pandemic on cervical cancer screening-related activities, as perceived by healthcare professionals. Please refer to arguments above.

Changes made: Page 12, Lines 366-368; Page 15, Lines 437-440.

evidence.

from https://news.ycombinator.com/item?id=8439611

This part of the article is difficult to understand, I think I would need more context about what relevance the type of software used has to see why this detail was included.

Clarification is needed:

For environments with 3 configservers 1 primary and only1 backup host must be specified

I never thought about this before. I did not try this before. I have not thought about that it there have a restriction for some of the edivices and browsers.

I think these are very important and I want to highlight them for others as well as keep them around for myself for when I need to come back to look at them!

This sounds little bit like Vivaldi :).

Author Response

Reviewer #1 (Public Review):

This is thorough, quantitative microbial ecology research on one of the most important problems of species coexistence in infection biology. The intermediate disturbance hypothesis is supported once again, and they show unsurprisingly that nutrition matters for their ratio of coexistence, but more specifically as a novel function of the ratio of metabolic fueling to reproductive rate, which the authors term absolute growth. I like this study for its care and completeness even though the results are fairly intuitive to those in the field of cystic fibrosis microbial ecology.

We would like to thank the reviewer for acknowledging the importance, care, and completeness of our original manuscript. We have continued to employ our standards of rigor for this revision.

Reviewer #2 (Public Review):

The authors present a manuscript that addresses an important topic of bacterial co-existence. Specifically modeling infection-relevant scenarios to determine how two highly antibiotic-resistant pathogens will develop over time. Understanding how such organisms can persist and tolerate therapeutic interventions has important consequences for the design of future treatment strategies.

We would like to thank the reviewer for acknowledging the importance of our work.

A major strength of this paper is the methodical approach taken to assess the dynamics between the two bacterial species. Using carbon sources to regulate growth to test different community structures provides a level of control to be able to directly assess the impact of one dominant pathogen over another.

The modeling aspect of this manuscript provides a basis for testing other disturbances and/or the impact of additional incoming pathogens. This could easily be applied to other infection settings where multiple microbes are observed ( for example viral/bacterial interactions in the lung).

Thank you for acknowledging the rigor in our experimental and modeling approaches.

The authors clearly show that by altering the growth rate and metabolism of various carbon sources, population structure can be modified, with one out-competing the other. Both modeling and experimental approaches support this.

The exploration of the role of virulence factors is less clear, for example how strains unable to produce virulence factors are impacted in regard to their overall growth and whether S. aureus is able to sense virulence factors without transcriptional assays here. Although the hypothesis is strong, the experimental data does not fully support this conclusion.

In addressing your comments below, we hope that we have increased your confidence in our hypotheses presented in our manuscript as it pertains to the involvement of virulence factors.

Spatial disturbance has a significant impact on community structure. Although using one approach to assess this, it is not clear if the spatial structure is impacted without the comparable microscopy evaluation.

We have indeed acknowledged this short coming in our revised manuscript. In the discussion, we write:

“While we did not explicitly quantify spatial organization experimentally owing to technical limitations of our microplate reader and microscope setups, in theory, co-culture in an undisturbed condition should facilitate the creation of spatial organization.”

In fact, we would really like to be able to track the position of each bacterium during shaking events. However, the plate reader cannot accommodate a microscope setup. While we could remove the plate from the plate reader and transport it to the microscope (two floors down), we cannot be certain that the position of the bacterium would not be altered during transport. We have thought about fixing the bacterium in place prior to transport. However, the injection of liquid for the purposes of fixation would likely alter the positioning of bacteria. Thus, we chose a modeling approach using an agent based model that is parametrized based on our experimental approach. Accordingly, we agree that this is a limitation of our current study. We hope that acknowledging this limitation in the discussion sits well with the reviewer.

Overall this paper highlights the use of modeling approaches in combination with wet lab experiments to predict microbial interactions in changing environments.

Reviewer #3 (Public Review):

This is an intriguing manuscript with a rigorous experimental and computational methodology looking at the interaction of Pseudomonas aeruginosa (Pa) and Staphylococcus aureus (Sa). These two pathogens frequently co-habit infections but in standard liquid media often show a winner-take-all outcome. This study seeks to be mechanistically predictive as to the outcome of the co-culture based on the addition of specific carbon sources as filtered through the lens of metabolic efficiency or, as the authors term - absolute growth. Overall, the study is sound, but there are some specific caveats that I would like to present:

We would like to thank the reviewer for acknowledging the rigor of our work.

1) The study undersells the knowledge in the literature of what allows or prohibits the stability of the Pa and Sa co-cultures. While most of the correct papers are cited, the outcomes of those studies are downplayed in favor of the current predictive study. While the current study is indeed more "predictive", it strays exceedingly far from an infection-relevant media, whereas other studies show reasonable co-existence in host-relevant media.

We have addressed this comment two different ways. First, we have included an entire paragraph in the discussion that acknowledges previous work and how our results fit into previous findings. We write:

“Given the clinical importance of co-infection with both P. aeruginosa and S. aureus, multiple previous studies have identified mechanisms of co-existence. Indeed, long term co-existence of both species can result in physiological changes that reduce their competitive interactions. Strains of P. aeruginosa isolated from patients that enter into a mucoid state show reduced production of siderophores, pyocyanin, rhamnolipids and HQNO, which facilitates the survival of S. aureus [23, 24]. These strains can also overproduce the polysaccharide alginate, which in itself is sufficient to decrease the production of these virulence factors. Moreover, exogenously supplied alginate can reduce the production of pyoverdine and expression from the PQS quorum sensing system, which is responsible for the production of HQNO [25]. Changes in the physiology of S. aureus can also facilitate co-existence. Strains of S. aureus isolated from patients with cystic fibrosis show multiple changes in the abundance of proteins including super oxide dismutase, the GroEL chaperone protein, and multiple surface associated proteins [26]. Interestingly, the majority of proteins that show changes in abundance in S. aureus are related to central metabolism, which is consistent with our findings demonstrating that metabolism can influence the co-existence of both species. While it is unclear as to how long-term co-culture would affect the ratio of absolute growth, our findings provide an additional mechanism that can determine the co-existence of these bacterial species.”

Second, as noted in our response in the ‘essential revisions’ section, we have tested the relationship between the final density ratio and the absolute growth ratio in SCFM medium, which we believe is host relevant. Our findings were fully consistent with the trends that we saw in our original submission. This data is presented in Fig. 3 and Figure 5 – figure supplement 3.

2) The major weakness in the ability of this study to be extrapolatable to infection conditions is the basal media selected for this analysis. The authors choose TSB, which is an incredibly rich media from the start, and proceed to alter only 11% of the available carbon (per mass) with their carbon source manipulations. This suggests an underappreciation for the amino acid metabolism routes of these two pathogens that are taking advantage of the roughly 89% of carbon as amino acid content in the TSB components of tryptone and soytone (17g and 3g, respectively vs the 2.5g carbon source). There are a few major issues with this basal formulation:

a) Comparison to all extant literature on Pa - The media historically used to assess Pa include (rich) LB, BHI, MH; (minimal) MOPS, M63, M9; (host-associated) ASM, SCFM, SCFM2, Serum, and DMEM. TSB is not a historically evaluated formulation for Pa (though it is often for non-mammalian pathogenic Pseudomonads and environmental species). Thus, this study is not inherently integrated into the Pa literature and presents an offshoot study for which a direct connection to extant literature is difficult. Explicitly testing these predictions in the most minimal media possible and then in a host-relevant model would be optimal.

We would truly like to thank the reviewer for their rigor in reviewing our manuscript. We, admittedly, overlooked how amino acids might be influencing the growth of P. aeruginosa in TSB medium. We originally chose TSB medium as previous studies that have examined the co-culture of S. aureus and P. aeruginosa, or their mechanisms of interaction, have used this medium (e.g., [29-34]).

To address this comment directly, we grew co-cultures in AMM minimal medium. This medium, to our knowledge, is the only minimal medium that allows growth of S. aureus. We, and others, have not reported growth of S. aureus in M9 or MOPS minimal medium despite the addition of components such as casamino acids and increases in the concentration of thiamine.

While AMM as reported is quite complex relative to media such as MOPS and M9, we removed several vitamins (nicotinic acid, thiamine, calcium pantothenate, biotin), decreased the concentration of some salts, used a low concentration of casamino acids (0.01%), and used a higher concentration of carbon source (0.04%). In doing so, we hoped to reduce any ‘background effect’ of media components and thus absolute growth could be driven more by carbon source.

Importantly, in using AMM medium, we continue to find a strong and significant relationship between the final density ratio and the absolute growth ratio. This data is presented in the Figure 3 and is described in a standalone paragraph in the results, along with our findings using SCFM.

b) TSB is not remotely host-relevant. The Whiteley lab has done monumental work evaluating in vitro models that mimic human infection (scrupulously matching transcriptomes) and TSB is about as far as you can get. Thus, the ability to extrapolate from the current work to infection without testing in host-relevant media is limited.

As noted above, we repeated our core experimental analysis in SCFM. The results are fully consistent with our original submission. This data is presented Figure 3 and in Figure 5- figure supplement 3.

c) The experimental situation has a component that is both good and bad- O2 tension. By overlaying with mineral oil, the authors immediately bias Staph (a more versatile fermenter) to success, whereas Pa deals with most of these carbon sources better at body level or higher O2 levels. The benefit of this is that many of the infection sites in which these two species co-occur are low in O2.

This was an interesting observation that we have partially addressed experimentally and acknowledged in the discussion.

First, we acknowledged the limitations of our experimental approach as it pertains to O2 levels in the discussion as follows:

“We note that our findings may be relevant to infections occurring in both high and low O2 environments. While P. aeruginosa is limited in its ability to perform fermentation [35], we have provided evidence that the absolute growth ratio can affect community composition in both aerobic (Figures 2-5) and more anaerobic environments (Figure 2 - figure supplement 1, panel H). The limited ability of P. aeruginosa to grow in anaerobic environments was apparent in SCFM as we could not obtain reliable or robustly quantifiable growth of this bacteria when succinate or -ketoglutarate was provided as a carbon source.”

Second, we tested the effect of placing mineral oil over top of the co-culture experiments, thus increasing the anaerobic nature of the environment. We found that, in general, as the ratio of absolute growth increased, so did the dominance of P. aeruginosa in the growth medium. This new data is presented in Figure 2 - figure supplement 1, panel H.

Taken together, we hope that these two modifications meet the Reviewer’s expectations.

d) Some of the tested metabolites are osmotically active (sucrose), while others are not (acetate), confounding the interpretation of what absolute metabolism means in the context of this study since the concentrations of all tested metabolites vary from above to below physiologic-dependent on the metabolite. A much better approach would have been to vary a single metabolite or combination to alter 'absolute metabolism' and test whether the stability of the co-culture held.

e) The manuscript never goes into the fact that for some of these "the carbon source" sources, they are catabolite repressed compared to the basal TSB amino acids (or not). Both organisms show exquisite catabolite repression control, yet this is not addressed at all within the text of the manuscript. Since this response in both organisms is sensitive to relative proportions of the various C-sources, failure to vary C-sources or compare utilization compared to the massive excess tryptone and soytone in the media makes the 'absolute metabolism' difficult to interpret.

To address comments d and e, and to acknowledge the potential limitations of our findings, we have included the following in the discussion. In this paragraph, we acknowledge the osmotic activity of the different carbon sources and preferential consumption of amino acids in TSB medium.

“One drawback of our approach in using different carbon sources to manipulate absolute growth is that some carbon sources are osmotically active, whereas others are not, which could have additional physiological effects on the bacteria outside of changing growth and metabolism. Moreover, both species of bacteria have different carbon source preferences; as above S. aureus tends to prefer carbon sources such as glucose [36] whereas P. aeruginosa prefers organic and amino acids [37]. Given the carbon source preferences of each species, in complex medium such as TSB, there is the potential that P. aeruginosa consumes amino acids prior to consuming the supplied carbon source. This is perhaps less of a concern in AMM medium or SCFM where the concentration of amino acids and additional nutrient components is reduced as compared to TSB medium. Along this line, it is certainly worth investigating how each nutrient component and its ordered utilization by both species contributes to changes in absolute growth. Minor or transient changes in absolute growth owing to preferential nutrient consumption may provide windows of opportunity for one species to increase its relative density to the other.”

f) The authors left out the 'favorite' sources of Pa that are known to be relevant in vivo - the TCA intermediates: citrate, succinate, fumarate (and directly relevant to host-pathogen interactions, itaconate)

We have included the analysis of succinate as a carbon source in both TSB medium (Figs. 1 and 2) and AMM medium (Fig. 3). However, we could not achieve reliable or a quantifiable growth rate of P. aeruginosa in SCFM medium supplemented with succinate in our experimental setup. Accordingly, this carbon source was not used in SCFM.

3) Statistics: Most of the experiments presented are comparisons in which there are more than two experimental groups and the t-tests employed therefore need to be corrected for multiple comparisons. The standard way to do this is to employ an ANOVA with the appropriate multiple-comparison-corrected post-test. These appear to be appropriate for Dunnett's post-testing but the comparator group is not directly defined within the figure legends. Multiple comparison testing is critical for this analysis, as the H0 is that all are the same - the more samples potentially pulled from the same distribution will result in a higher likelihood that one or more will appear as from a distinct population (i.e. H0 rejected). Multiple comparisons correct for this and are absolutely critical for the evaluation of the data presented in this manuscript.

We have addressed this comment two different ways.

First, where there was a clear control group, we performed either a Dunnett’s (for normally distributed data) or a Dunn’s (for non-parametric data sets) following either an ANOVA or Kruskal-Wallis, respectively. These tests were applied to the data presented in Figure 2B, 5H (top and bottom panels) and in Figure 2 - figure supplement 1, panels K-L.

Second, we did not broadly perform multiple comparisons across all data sets. The reason is that this approach would test the significance of relationships that are not relevant to the central premise of the manuscript. For example, a multiple comparison for figure 1B would test the growth rate of all carbon sources against all carbon sources. However, we are only interested if S. aureus or P. aeruginosa grows faster than one another. However, we do understand the need for a corrected P value to reduce the occurrence of Type 1 errors. To accomplish this, we applied a Benjamini-Hochberg Procedure [38] with a 8.5% discovery rate to all P values in the manuscript, including those that tested the distribution of data. This reduced the P value to indicate significance at < 0.0472. We have updated all claims and indications of significance in the figures based on this adjusted P value.

4) The authors missed including Alves et Maddocks 2018 in relation to priority effects and other contributing factors to stable Pa/Sa co-culture.

We have indeed included this manuscript and its findings in the introduction where we write:

“While S. aureus can initially aid in the establishment of the P. aeruginosa population [8], production of N-acetylglucosamine from S. aureus augments…..”

在引入时用于目录链接，后来扩展到所有文件

This is beautiful imagery.

Author Response:

Reviewer #1 (Public Review):<br /> <br /> Roberts et al have developed a tool called "XTABLE" for the analysis of publicly available transcriptomic datasets of premalignant lesions (PML) of lung squamous cell carcinoma (LUSC). Detection of PMLs has clinical implications and can aid in the prevention of deaths by LUSC. Hence efforts such as this will be of benefit to the scientific community in better understanding the biology of PMLs.

The authors have curated four studies that have profiled the transcriptomes of PMLs at different stages. While three of them are microarray-based studies, one study has profiled the transcriptome with RNA-seq. XTABLE fetches these datasets and performs analysis in an R shiny app (a graphical user interface). The tool has multiple functionalities to cover a wide range of transcriptomic analyses, including differential expression, signature identification, and immune cell type deconvolution.

The authors have also included three chromosomal instability (CIN) signatures from literature based on gene expression profiles. They showed one of the CIN signatures as a good predictor of progression. However, this signature performed well only in one study. The authors have further utilised the tool XTABLE to identify the signalling pathways in LUSC important for its developmental stages. They found the activation of squamous differentiation and PI3K/Akt pathways to play a role in the transition from low to high-grade PMLs

The authors have developed user-friendly software to analyse publicly available gene expression data from premalignant lesions of lung cancer. This would help researchers to quickly analyse the data and improve our understanding of such lesions. This would pave the way to improve early detection of PMLs to prevent lung cancer.

Strengths:

1. XTABLE is a nicely packaged application that can be used by researchers with very little computational knowledge.<br /> 2. The tool is easy to download and execute. The documentation is extensive both in the article and on the GitLab page.<br /> 3. The tool is user-friendly, and the tabs are intuitively designed for successive steps of analysis of the transcriptome data.<br /> 4. The authors have properly elaborated on the biological interest in investigating PMLs and their clinical significance.

Weaknesses:

The article is focused on the development and the utility of the tool XTABLE. While the tool is nicely developed, the need for a tool focussing only on the investigation of PMLs is not justified. Several shiny apps and online tools exist to perform transcriptomic analysis of published datasets. To list a few examples - i) http://ge-lab.org/idep/ ; ii) http://www.uusmb.unam.mx/ideamex/ ; iii) RNfuzzyApp (Haering et al., 2021); iv) DEGenR (https://doi.org/10.5281/zenodo.4815134); v) TCC-GUI (Su et al., 2019). While some of these are specific to RNA-seq, there are plenty of such shiny apps to perform both RNA-seq and microarray data analysis. Any of these tools could also be used easily for the analysis of the four curated datasets presented in this article. The authors could have elaborated on the availability of other tools for such analysis and provided an explanation of the necessity of XTABLE. Since 3 of the 4 datasets they curated are from microarray technology, another good example of a user-friendly tool is NCBI GEO2R. This is integrated with the NCBI GEO database, and the user doesn't need to download the data or run any tools. iDEP-READS (http://bioinformatics.sdstate.edu/reads/) provide an online user-friendly tool to download and analyse data from publicly available datasets. Another such example is GEO2Enrichr (https://maayanlab.cloud/g2e/). These tools have been designed for non-bioinformatic researchers that don't involve downloading datasets or installing/running other tools.

Two of these tools (IDEP and TCC-GUI) were reviewed in a literature review covering 20 Shiny apps performed two years ago prior to work on XTABLE starting. Three of the suggested tools (IDEP, RNFuzzyApp, TCC-GUI) are for processing only RNA-seq datasets. IDEAMEX appears to be for RNA-seq data only and is severely limited in its downstream analysis capabilities. DEGenR appears to handle microarray datasets and features an option to retrieve data directly from GEO. However, it appears to be based on GEO2R (with additional downstream analyses) where it automatically logtransforms already log-transformed data and unlike GEO2R, you do not have the option to not apply a log-transformation. A refreshed literature search focusing on microarray datasets highlighted three additional tools. iGEAK which hasn’t been updated in three years and seems to have compatibility issues running on new Windows and Mac machines. sMAP, an upcoming Shiny app for microarray data published in bioRxiv on 29 May 2022. MAAP which has the same issue of log-transforming already log-transformed data. iDEP-READS does not list the datasets used in XTABLE. GEO2Enrichr appears to require the counts table and experimental design in one file, performs a “characteristic direction” DEG test and outputs enriched pathways. These apps require not just downloading of datasets but reformatting and renaming of expression data files and creation of additional files for setting up the DEG analysis which is not practical for the number of samples we have (122, 63, 33, 448) even if these apps handled microarray data. XTABLE also incorporates AUC metrics, which is appropriate given the number of samples in each dataset and tool known for adequately controlling FDR, which is not seen in other apps as well as emphasis on individual gene results and interrogation.

A new paragraph on the discussion section (lines 361-370) of the discussion addresses the potential use of existing applications instead of XTABLE

Secondly, XTABLE doesn't provide a solution to integrate the four datasets incorporated in the tool. One can only analyse one dataset at a time with XTABLE. The differences in terms of methodology and study design within these four datasets have been elaborated on in the article. However, attempts to integrate them were lacking.

We repeatedly considered different strategies of integrating the analysis of the four datasets and we always reached the conclusion that it was hardly going to offer any advantage, or that it might be counterproductive.

Integration can occur at multiple levels. One possibility is to carry out the same analysis (e.g. expression of a given gene in two groups of samples) in all datasets. Since the design and methodologies of the four studies differ substantially (different stages, different definitions of progression status, etc), a unique stratification for all datasets is not possible. Moreover, interrogating the four datasets simultaneously would slow the analysis, with no significant advantage in terms of speed. Another possibility is the integration of results in the same output. For instance, obtain a single chart with the expression of a given gene in multiple subgroups of the four datasets. We think that the results from each cohort should be kept separately and then compared with a similar analysis from other datasets due to differences in design. Scientifically, this is the best way to proceed as it avoids confusions.

Nevertheless, XTABLE allows the export of data for further analysis. The user can use this option to integrate data using other applications or statistical packages.

We do understand the attractiveness of integration between the four datasets is and we seriously considered it. But there is a fine balance between user-friendliness, flexibility, and scientific rigour. We think that XTABLE achieves this balance. Increasing integration of datasets might lead to error and wrong conclusions due to biological and methodological differences between studies. We believe that comparing analyses obtained independently from the four cohorts is the most sensible way to proceed.

We propose to discuss these aspects accordingly.

The integrative analysis of two or more datasets has been discussed in a new paragraph (382-391)

The tool also lacks the flexibility for users to add more datasets. This would be helpful when there are more datasets of PMLs available publicly.

This was also a permanent topic for discussion while designing XTABLE. Creating a tool that could be used to analyse other cohorts of precancerous lesions, while maintaining the ease of use was certainly a challenge. We had to adapt XTABLE to the characteristics of each one of the four databases: specific stratification criteria, different nomenclatures for the different sample types, etc. Designing a shiny app that can be adapted to other present or future datasets without the need of changing the code is simply not practical.

The flexibility that these other Shiny apps incorporate to analyse any RNA-seq dataset requires the contrasts used for the differentially expressed gene analysis be manually defined. IDEP requires an experimental design file where sample names in the counts file must match exactly the sample names in this experimental design file and pre-processing visualisation is limited to the first 100 samples. RNFuzzyApp is similar but we could not format the experimental design file in a way that did not result in the app crashing upon upload. TCC-GUI requires all the sample names to be renamed to the contrast group with the addition of the replicate number. Apps that allow datasets to be uploaded do not have a practical or easy way to set up the DEG analysis of more than a couple dozen samples.

Future versions of XTABLE can be updated to include additional curated PML datasets that would enhance hypothesis generation upon request. Importantly, the code is freely available and can be modified by other scientists to add their cohorts of interest, although we agree that a high level of expertise in coding will be needed. We propose to add these considerations to the text.

The possibilities of expansion of XTABLE to new databases are discussed in lines 392-398

Understanding the biology of PML progression would require a multi-omics approach. XTABLE analyses transcriptome data and lacks integration of other omics data. The authors mention the availability of data from whole exome, methylation, etc from the four studies they have selected. However, apart from the CIN scores, they haven't integrated any of the other layers of omics data available.

Only one dataset (GSE108104) contains whole-exome sequencing and methylation data. We considered that a multi-omics approach in XTABLE would result in an overcomplicated application. As far as early detection and biomarker discovery is concerned, transcriptomic data is the most interesting parameter.

Also discussed in lines 382-391

Lastly, the authors could have elaborated on the limitations of the tool and their analysis in the discussion.

We propose to raise these limitations accordingly in the discussion.

See above.

Reviewer #2 (Public Review):

In this manuscript, Roberts et al. present XTABLE, a tool to integrate, visualise and extract new insights from published datasets in the field of preinvasive lung cancer lesions. This approach is critical and to be highly commended; whilst the Cancer Genome Atlas provided many insights into cancer biology it was the development of accessible visualisation tools such as cbioportal that democratised this knowledge and allowed researchers around the world to interrogate their genes and pathways of interest. XTABLE is trying to do this in the preinvasive space and should certainly be commended as such. We are also very impressed by the transparency of the approach; it is quite simple to download and run XTABLE from their Gitlab account, in which all data acquisition and analysis code can be easily interrogated.

We would however strongly advocate deploying XTABLE to a web-accessible server so that researchers without experience in R and git can utilise it. We found it a little buggy running locally and cannot be sure whether this is due to my setup or the code itself. Some issues clearly need development; Progeny analysis brings up a warning "Not working for GSE109743 on the server and not sure why". GSEA analysis does not seem to work at all, raising an error "Length information for genome hg38 and gene ID ensGene is not available". In such relatively complex software, some such errors can be overlooked, as long as the authors have a clear process for responding to them, for example using Gitlab issue reporting. Some acknowledgement that this is an ongoing development would be helpful.

We thank the reviewer for these comments. We will inspect the code to address those warnings, implement a system for issue reporting, and add the acknowledgements suggested by the reviewer. Regarding the deployment of XTABLE to a web-accessible server, this could present a challenge in the long term as computing resources need to be allocated for years and the economic cost involved.

The code has been inspected to remove the warning and errors pointed out by the reviewer.

The authors discuss some very important differences between the datasets in the text. Most notably they differ in endpoints and in the presence of laser capture. We would advocate including some warning text within the XTABLE application to explain these. For example, the "persistent/progressive" endpoint used in Beane et al (next biopsy is the same or higher grade) is not the same as the "progressive" endpoint in Teixeira et al (next biopsy is cancer); samples defined as "persistent/progressive" may never progress to cancer. This may not be immediately obvious to a user of XTABLE who wishes to compare progressive and regressive lesions. Similarly, the use of laser capture is important; the authors state that not using laser capture has the advantage of capturing microenvironment signals, but differentiating between intra-lesional and stromal signals is important, as shown in the Mascaux and Pennycuick papers. The authors cannot do much about the different study designs, but as the goal is to make these data more accessible We think some brief description of these issues within the app would help to prevent non-expert users from drawing incorrect conclusions.

The authors themselves illustrate this clearly in their analysis of CIN signatures in progression potential. They observe that there is a much clearer progressive/regressive signal in GSE108124 compared to GSE114489 and GSE109743. This does not seem at all surprising, since the first study used a much stricter definition of progression - these samples are all about to become cancer whereas "progressive" samples in GSE109743 may never become cancer - and are much enriched for CIN signals due to laser capture. Their discussion states "CIN scores as a predictor of progression might be limited to microdissected samples and CIS lesions"; you cannot really claim this when "progression" in the two cohorts has such a different meaning. To their credit, the authors do explain these issues but they really should be clearly spelled out within the app.

This is a very good point. We will add the warning text about the differences between studies regarding the definition of progression potential and the differences and sample processing (LCM or o not) so that the user is permanently aware of the differences between cohorts.

A new tab (Dataset) has been added table with the methodologies used in each of each study, and the differences in progression status definitions. Additionally, we emphasized these differences in the main text of the manuscript (lines 296-300 and 403-409).

We are not sure we agree with their analysis of CDK4/Cyclin-D1 and E2F expression in early lesions. The authors claim these are inhibited by CDKN2A and therefore are markers of CDKN2A loss of function. But these genes are markers of proliferation and can be driven by a range of proliferative processes. Histologically, low-grade metaplasias and dysplasias all represent proliferative epithelium when compared to normal control, but most never become cancer. It is too much of a leap to say that these are influenced by CDKN2A because that gene is inactivated in LUSC; do the authors have any evidence that this gene is altered at the genomic level in low-grade lesions?

We are grateful for this comment. There is currently not evidence that CDKN2A mutations occur in low-grade lesions and therefore, we cannot argue that the of CDK4/Cyclin-D1 and E2F expression signature are the result of CDKN2A inactivation in low-grade lesions. We propose to modify the text to introduce these caveats to our conclusion an make our interpretations more accurate.

We have modified the discussion (lines 443-454) to address the interpretation of our results regarding the connection between CDKN2A inactivation and the CDK4/cyclin-D1 and E2F signatures. We now focus our conclusions on the pathway itself and we mention Cyclin-D1 and CDKN2A alterations as a potential modulator of the changes in the pathway, but leaving the discussion open to other drivers.

Overall this tool is an important step forwards in the field. Whilst we are a little unconvinced by some of their biological interpretations, and the tool itself has a few bugs, this effort to make complex data more accessible will be greatly enabling for researchers and so should be commended. In the future, we would like to see additional molecular data integrated into this app, for example, the whole genome and methylation data mentioned in line 153. However, we think this is an excellent start to combining these datasets.

1-click Beautiful Screenshots on Windows

Interesting workflow here for taking a screenshot quickly, saving it as a file, saving it to clipboard, and sharing it to various services.

ShareX - The best free and open source screenshot tool for Windows

via https://www.youtube.com/watch?v=6q-GQ4KiJEk

Important study and one that I had never heard of before.

Unfortunately a common sight in these small mining towns after their boom in the 20th century. They are often left mostly abandoned and deserted.

Check VC++ compiler version

Author Response

Reviewer #1 (Public Review):

Demographic inference is a notoriously difficult problem in population genetics, especially for non-model systems in which key population genetic parameters are often unknown and where the reality is always a lot more complex than the model. In this study, Rose et al. provided an elegant solution to these challenges in their analysis of the evolutionary history of human specialization in Ae. aegypti mosquitoes. They first applied state-of-the-art statistical phasing methods to obtain haplotype information in previously published mosquito sequences. Using this phased data, they conducted cross-coalescent and isolation-with-migration analyses, and they innovatively took advantage of a known historical event, i.e., the spread of Ae. aegypti to South America, to infer the key model parameters of generation time and mutation rate. With these parameters, they were able to confirm a previous hypothesis, which suggests that human specialists evolved at the end of the African Humid Period around 5,000 years ago when Ae. aegypti mosquitoes in the Sahel region had to adapt to human-derived water storage as their breeding sites during intense dry seasons. The authors further carried out an ancestry tract length analysis, showing that human specialists have recently introgressed into Ae. aegypti population in West African cities in the past 20-40 years, likely driven by rapid urbanization in these cities.

Given all the complexities and uncertainties in the system, the authors have done outstanding jobs coming up with well-informed research questions and hypotheses, carrying out analyses that are most appropriate to their questions, and presenting their findings in a clear and compelling fashion. Their results reveal the deep connections between mosquito evolution and past climate change as well as human history and demonstrate that future mosquito control strategies should take these important interactions into account, especially in the face of ongoing climate change and urbanization. Methodologically, the analytical approach presented in this paper will be of broad interest to population geneticists working on demographic inference in a diversity of non-model organisms.

In my opinion, the only major aspect that this paper can still benefit from is more explicit and in-depth communication and discussion about the assumptions made in the analyses and the uncertainties of the results. There is currently one short paragraph on this in the discussion section, but I think several other assumptions and sources of uncertainties could be included, and a few of them may benefit from some quantitative sensitivity analyses. To be clear, I don't think that most of these will have a huge impact on the main results, but some explicit clarification from the authors would be useful.

Below are some examples:

Thank you very much for your kind words and your feedback! We have expanded our discussion of assumptions and uncertainties – we have responded to each point below:

1) Phasing accuracy: statistical phasing is a relatively new tool for non-model species, and it is unclear from the manuscript how accurate it is given the sample size, sequencing depth, population structure, genetic diversity, and levels of linkage disequilibrium in the study system. If authors would like to inspire broader adoption of this workflow, it would be very helpful if they could also briefly discuss the key characteristics of a study system that could make phasing successful/difficult, and how sensitive cross-coalescent analyses are to phasing accuracy.

We agree that this is an important topic to expand on. We have clarified as follows:

Results, Page 4, last paragraph: “Over 95% of prephase calls had maximal HAPCUT2 phred-scaled quality scores of 100 and prephase blocks (i.e. local haplotypes) were 728bp long on average (interquartile range 199-1009bp). We then used SHAPEIT4.2 to assemble the prephase blocks into chromosome-level haplotypes, using statistical linkage patterns present across our panel of 389 individuals (25).”

Discussion, Page 8, last paragraph: “Overall linkage disequilibrium is relatively low in Ae. aegypti, dropping off quickly over a few kilobases and reaching half its maximum value within about 50kb (37); this is likely sufficient for assembling shorter, high-confidence prephase blocks into longer haplotypes in many cases. However, phase-switch errors may be common across longer distances – potentially affecting inferences in the most recent time windows. Nevertheless, the similar results we obtain using different proxy populations (and thus different input haplotype structures) for human-specialist and generalist lineages (see Figure S1) suggest that our results are robust to potential mistakes in long-range haplotype phasing.”

Discussion, Page 9, paragraph 2: “Here, we take advantage of a continent-wide set of genomes, combined with read-based prephasing and population-wide statistical phasing to develop a phasing panel that should enable future studies in Ae. aegypti with a lower barrier to entry. The same approach may work for other study organisms with similar population genomic properties; high levels of diversity are helpful for prephasing and at least moderate levels of linkage disequilibrium are important for the assembly of prephase blocks.”

2) Estimation of mutation rate and generation time: the estimation of these importantparameters is made based on the assumption that they should maximize the overlap between the distribution of estimated migration rate and the number of enslaved people crossing the Atlantic, but how reasonable is this assumption, and how much would the violation of this assumption affect the main result? Particularly, in the MSMC-IM paper (Wang et al. 2020, Fig 2A), even with a simulated clean split scenario, the estimated migration rate would have a wide distribution with a lot of uncertainty on both sides, so I believe that the exact meaning and limitations of such estimated migration rate over time should be clarified. This discussion would also be very helpful to readers who are thinking about using similar methods in their studies. Furthermore, the authors have taken 15 generations per year as their chosen generation time and based their mutation rate estimates on this assumption, but how much will the violation of this assumption affect the result?

This is a great point. We have expanded our discussion of how this assumption affects our conclusions (see Discussion page 9, first paragraph): “Furthermore, we chose a scaling factor that maximized overlap between the peak of estimated Ae. aegypti migration and the peak of the Atlantic Slave Trade (Fig. 2B). If we instead consider alternative scenarios where peak migration occurred at the very beginning of the slave trade era, around 1500, then our inferred mutation rate would be lower (about 2.4e-9, assuming 15 generations per year), pushing back the split of human-specialist lineages to about 10,000 years before present. This scenario seems less plausible, in part because our isolation-with-migration analyses suggest a gradual onset of migration between continents rather than a single, early-pulse model. It would also make it harder to explain the timing of the bottleneck we see in invasive populations; the first signs of this bottleneck occur at the beginning of the slave trade (~500 years ago) with our current calibration (Fig. S1A), but would be pushed to a pre-trade date in this alternative scenario. We can also consider a scenario in which peak Ae. aegypti migration occurred more recently, perhaps around 1850, corresponding to increased global shipping traffic outside the slave trade alone. In this case, our inferred mutation rate would be higher (or generation time lower), and the split of human-specialist lineages would be placed at about 3,000 years ago. Overall, the best match between the existing literature and our data corresponds to our main estimates, but alternative scenarios could gain support if future research finds evidence for a different time course of invasion than is suggested by the epidemiological literature.”

We have slightly expanded our description of calibration in Results, page 5, last paragraph: “The fact that we see good overlap between the two distributions (yellow–white color) across a wide range of reasonable mutation rates and generation times for Ae. aegypti is consistent with our understanding of the species’ recent history and supports our approach. For example, if we take the common literature value of 15 generations per year (0.067 years per generation) (17, 20), the de novo mutation rate that maximizes correspondence between the two datasets is 4.85x10-9 (black dot in Figure 2A, used in Figure 2B), which is on the order of values documented in other insects. We chose to carry forward this calibrated scaling factor (corresponding to any combination of mutation rate and generation time found along the line in Figure 2A) into subsequent analyses.”

We have also expanded on the uncertainty of our analyses (see Discussion page 8, last paragraph): “First, the temporal resolution of our inferences is relatively low, and both previously published simulations (39) and our own bootstrap replicates (Figure 2B–D, grey lines) suggest relatively wide bounds for the precise timing of events.”

3) The effect of selection: all analyses in this paper assume that no selection is at play,and the authors have excluded loci previously found to be under selection from these analyses, but how effective is this? In the ancestry tract length analysis, in particular, the authors have found that the human-specialist ancestry tends to concentrate in key genomic regions and suggested that selection could explain this, but doesn't this mean that excluding known loci under selection was insufficient? If the selection has indeed played an important role at a genome-wide level, how would it affect the main results (qualitatively)?

We have clarified that we excluded those loci from our timing estimates for both MSMC and ancestry tract analyses, but then re-ran the ancestry tract analysis with all regions included to visualize and assess how tracts were distributed along chromosomes. See Methods, page 12, paragraph 2: “Since selection associated with adaptation to urban habitats could shape lengths of admixture tracts, we masked regions previously identified as under selection between human-specialists and generalists when estimating admixture timing—namely, the outlier regions in (2). However, we used an unmasked analysis to determine and visualize the genome-wide distribution of ancestries (Fig. 3).”

We have also added additional discussion of the expected effects of selection on our analyses (see Discussion, page 9, last paragraph): “Positive selection during adaptive introgression can increase tract lengths and make admixture appear to be more recent than it actually is. For this reason, we masked regions of the genome thought to underlie adaptation to human habitats before running our analysis. Nevertheless, if selection has acted outside these regions, admixture may be somewhat older than we estimate.”

Poor Larkin, his pessimistic outlook on the universe doesn't take him much farther than the sun's comprehension of glass, which he sees as both nothing and perpetually nowhere.

Those individuals' imaginations that are enthralled by emotions become nothing more than flabbergasted bananas. The evidence is when they become entranced when they have no proof to support the things that they believe in. This demonstrates that Larkin has animosity against Christians since, in his view, the cross represents nothing more than a pointless symbol.

The nihilistic perspective on the world that Larkin has is enough to make me throw up into his mouth. When Larking says "or sweating in the dark," I find it quite prothetic of him since that is precisely what he is going to be doing for the rest of his precious everlasting life. I dont find that amusing.

Larking believes that dreams are what keep the creative imagination prisoner, or in bondage he adds the gestures of people who wanted to bring damage or offer compassion. This vision depicts the harvest season with field workers cutting down the crop.

There is an image of harvest season and field workers mowing down the crop, which Larking suggests are dreams that hold the creative imagination hostage, or in bondage he includes the gestures of those who intended to cause harm or show mercy.

Larkin has a very archaic view of the world. His paradise has become a couple kids fucking taking pills and birth control. What a looser.

High windows is written in 5 quatrains. Lines 6,8 begin a water-downed rhyme scheme with the words side and slide. lines 10 and 12 have a rhyme scheme with the words back and dark. 13, 15 have a rhyme scheme with the words hide and slide. 17, 19 words windows and shows both have the "ows." 18, 20 scheme with words glass and endless.

但windows自己出的也够用了

The narrator describes the lake area as a "defeated jumble of sheds and small junkyards, the sidewalk gives up and we are walking on a sandy path with burdocks, plantains, humble nameless weeds all around." it sounds like its almost a deserted area as he tells us they walk onto a vacant lot. Although his description is vivid Lake Huron sounds to be less desired when he describes "The Pavilion"... "full of farmers in stiff good clothes on Sundays."

Blessed once again Jonathan is given back his home nearly all intact after the war. For survivors like the Iwegbu family the small things like the bike and the house begin to add up and it is this humble house they find standing a msiracleious.

Author Response

Reviewer #1 (Public Review):

• The statistical procedures used are not completely described and may not be appropriate.

We revised the text in Methods and Results sections to give more details about the methods used.

-As only two levels of delay were tested, it is not possible to directly test whether the subjective discounting function is hyperbolic or exponential and hence whether the delay is encoded subjectively or objectively.

We agree with the reviewer. A higher number of task parameters may offer a better resolution to evaluate the discounting functions. Fortunately, this does not affect our main results.

• The task has several variable interval lengths (hold in: 1.2-2.8 s, short delay: 1.8-2.3 s, long delay: 3.5-4s) that frustrate interpretation. The distribution of these delays is not described, for example as it reads it seems possible that some long delay rewards are delivered with shorter latency between cue and reward than some short delay rewards (1.2 + 3.5 = 4.7s vs. 2.8+2.3 = 5.1 s).

We revised the text to address that ambiguity. In the new version of the manuscript, we describe short versus long delays considering the total delay intervals between instruction cue onset and reward delivery [short delay (3.5-5.6s) and long delay (5.2-7.3s)]. Within each delay category, individual delays were distributed in a gaussian fashion such that the two delay ranges overlapped for 9% of trials. These details are now described in the revised Methods section (pg. 22).

-The authors have not considered that if the delay value is encoding, then the value, both objectively and subjectively, may be changing as the delay elapses. The variation of these task intervals may have an effect on the value of delay.

In the present study, we report a dynamic integration between the desirability of the expected reward and the imposed delay to reward delivery across the waiting period. Our results (e.g. see Fig. 6) do not fit with simple linear (or logarithmic) effects corresponding to continuous regular changes as the delay elapses. We found different types of interactions (Discounting± and Compounding±) at different periods of the hold period and in different single units. We did not find a way to model all these types of interactions with this type of approach.

Reviewer #2 (Public Review):

• Plots of "rejection rate" (trials where the monkeys failed to wait until the rewards) as a function of delay and reward size seem to indicate that the monkeys understood the visual cue. The rejection rates were very low (less than 4% for almost all conditions) which indicates that the monkeys did not have a hard time inhibiting their behavior. It also meant that the authors could not compare trials where the monkeys successfully waited with trials where they failed to wait. This missing comparison weakens the link between the neurophysiological observations and the conclusions the authors made about the signals they observed.

Here, our main goal was to describe the dynamic STN signals engaged during the waiting period without studying action-related activities. In the discussion (pg. 20), we clearly wrote ‘Further research is needed to determine whether the neural signals identified here causally drive animals’ behavior or rather just participate to reflect or evaluate the current situation.’ Consequently, our conclusions were already tempered by that point.

In addition, we address the same limitation by writing (pg. 20): “An important avenue for future research will be to determine how STN signals, such as those described here, change when animals run out of patience and finally decide to stop waiting. To do this, however, smaller reward sizes and longer delays might be used to promote more escape behaviors during the delay interval.”

• The authors examined the STN activity aligned to the start of the delay and also aligned to the reward. Most of the "delay encoding" in the STN activity was observed near the end of the waiting period. The trouble with the analysis is that a neuron that responded with exactly the same response on short and long trials could appear to be modulated by delay. This is easiest to see with a diagram, but it should be easy to imagine a neural response that quickly rose at the time of instruction and then decayed slowly over the course of 2 seconds. For long trials, the neuron's activity would have returned to baseline, but for short trials, the activity would still be above baseline. As such, it is not clear how much the STN neurons were truly modulated by delay.

We agree with the reviewers. Our original analyses using two-time windows had the potential to introduce biases in the detection of neuronal activities modulated by the delay. To overcome this issue, we modified the time frame of all of our analyses (neuronal activity, eye position, EMG). Now, the revised version of the manuscript only reports activities across one-time window aligned to the time of instruction cue delivery (i.e., -1 to 3.5s relative to instruction cue onset). This time frame corresponds to the minimum possible interval between instruction cues and reward delivery. We have revised all of the figures and we re-calculated all of the statistics using that one analysis window. Despite these major modifications, our key findings were not changed substantially. We found the same pattern in STN activities, with a strong encoding of reward (48% of neurons) preceding a late encoding of delay (39% of neurons). We also updated the text in Methods and Results sections to reflect the revised analyses.

• Another concern is the presence of eye movement variables in the regressions that determine whether a neuron is reward or delay encoding. If the task variables modulated eye movements (which would not be surprising) and if the STN activity also modulated eye movements, then, even if task variables did not directly modulate STN activity, the regression would indicate that it did. This is commonly known as "collider bias". This is, unfortunately, a common flaw in neuroscience papers.

Because the presence of eye variables did not influence how neurons were selected by the GLM, we do not think it likely that our analysis was susceptible to “collider bias”. Nonetheless, to control for that possibility directly, we have now repeated the GLM analyses with eye movement variables excluded. Results are shown in a new figure (Fig.4 – supplementary 1). Exclusion of eye parameters produced results that are very similar to those from the GLM that included eye parameters (differences <3 degrees). We have added text to the manuscript describing this added control analysis.

but it also relies on the "deserving" and "undeserving" distinction of graffiti. Some graffiti artists are making art that, if it were not graffiti, would be acceptable to the wider public. But not every graffiti artist is. So is only "good" graffiti allowed?

I do love that this shows how humans are the same--we create wall art then, we create wall art now.

Reviewer #2 (Public Review):

The present study aimed to demonstrate the utility of brain signal decoding for the differentiation of asynchronous motor signs in Parkinson's disease (PD). To this end, thirty-one PD patients undergoing deep brain stimulation electrode implantation were recruited to participate in an intraoperative motor task. Task performance was compared to extra-operative experiments in healthy subjects. Neural activity and movement traces were segmented into 7-second windows and attributed tremor and slowness measures. To integrate the two symptom domains an additional decoding state termed effective motor control was introduced, which represented the absence of symptoms. Support vector machine regression was used as the model of choice that was trained on individual recording sessions within subjects. All decoding targets from each neurophysiological modality reached significant prediction performances. This represents an important milestone in the current state of research towards machine learning-based intelligent adaptive deep brain stimulation.

Strengths

1. The present analysis is among the first to demonstrate the potential utility of brain signal decoding for the differentiation of asynchronous motor symptoms in Parkinson's disease. In the future, such approaches may be adopted in clinical brain-computer interfaces that can adapt stimulation in real time to concurrent therapeutic demand.

2. The effort from the research team and patients to acquire this important dataset is commendable. The time pressure in the operation room combined with the current trend of asleep surgery for deep brain stimulation makes such data very rare.

3. No relevant difference in decoding performance was found for subthalamic micro vs. macroelectrode recordings. This has practical significance because current sensing-enabled deep brain stimulation implants only allow for macro-recordings, which according to this study has no severe disadvantage over microelectrode recordings for movement decoding. Note that this question could only be answered in the intraoperative setting, which on the other hand can have disadvantages further described below.

4. Beyond the subthalamic nucleus, the authors corroborate the superiority of electrocorticography over subthalamic activity for movement and symptom decoding in Parkinson's disease. This provides further evidence that additional sensing electrodes may complement the subthalamic signals for adaptive deep brain stimulation.

5. Finally, the idea of decoding the presence of an effective motor state is creative and may inspire future developments in adaptive stimulation control algorithms.

Weaknesses

(Note that I take more words for weaknesses, not because they outweigh the strengths, but because I want to justify my criticism in more detail)

1. One inherent limitation of this study is the intraoperative setting, which demands the patients' skull be fixed to the stereotactic frame. This setting is not naturalistic per se and likely comes with additional perturbations in the brain states that are recorded. Thus, the generalization to real-world scenarios is limited. Given the unique opportunity to record invasive brain signals in humans, this limitation has to be accepted and should be taken into account for the interpretation of the results. As mentioned in the strengths, this is currently the only setting that allows for a comparison of micro- and macroelectrode recordings for brain signal decoding.

2. Similarly, the medication state is defined by the intraoperative scenario, as deep brain stimulation implantations are performed after the withdrawal of dopaminergic medication in the so-called dopaminergic OFF state. In this state, PD symptoms are aggravated, which is used clinically to provide a more reliable assessment of deep brain stimulation-induced symptom alleviation. This may also lead to an overestimation of decoding performances as the difference between the absence and presence of PD motor signs in the dopaminergic medication ON state during activities of daily living could be more nuanced.

3. The task design is very interesting as it allows for a continuous definition of symptom severity and motor performance. The comparison to healthy subjects demonstrated clearly higher tremor scores in PD but no significant differences in movement velocity (depicted as trending but p>0.2). This is somewhat unexpected as slowness of movement, also called bradykinesia, is a defining symptom of Parkinson's disease (PD). By definition, this symptom is present in all PD patients, also indicated in the clinical scores shown in the present study. Action tremor, i.e. the presence of tremulous muscle activity during motor performance, is comparatively rare. To support the clinical relevance of the movement tremor observed during the task, the authors show a correlation with the "resting tremor" score from the clinical assessment. It is unclear to me why resting, instead of action tremor scores are shown, as both are part of the clinical assessment (Unified Parkinson's disease rating scale - UPDRS part III). Ultimately, even though resting tremor is significantly more common in Parkinson's disease, not all patients of the current cohort had resting tremor (as indicated in the clinical score correlation). Thus, it remains somewhat puzzling how precise the 3-8 Hz activity actually captures tremor vs. motor noise or inaccuracy. A more fine-grained analysis comparing patients with clinically diagnosed action tremor (as defined by preoperative UPDRS assessment) and without tremor could have helped to support the clinical claims on symptom-specific decoding. On the other hand, the lack of a significant difference in the slowness of movement in the patient cohort relative to healthy controls questions the ability of the task to capture this symptom. Here, I am not sure whether the normalization procedure may have an influence on the comparability. Finally, movement velocity is an easy target that is distributed across a spectrum, so despite the lack of a significant difference in the healthy cohort, I am relatively confident that the decoding of movement slowness in the present cohort is clinically meaningful.

4. Overall, the pathophysiological framework is well placed in the current state of literature, while almost the entire field of brain signal decoding for adaptive deep brain stimulation was neglected. Successful decoding to address Parkinson's and essential tremor (another disorder with more common action tremor) was achieved by multiple groups in impactful studies representing more naturalistic extraoperative or fully embedded settings (Hirschmann et al., 2017, He et al., 2021, Opri et al., 2021). Additionally, other symptoms, like gait disturbances have been the target of machine learning analyses more recently (Louie et al., 2022 and Thenaisie et al., 2022). Here, the manuscript appears to avoid a discussion of the present endeavour in comparison to the current state of the field. One of our own studies has provided the first demonstration of the superiority of electrocorticography over subthalamic LFP for movement decoding, which I am happy to see replicated for the first time in the present manuscript. Importantly, the referenced study showed modality-dependent model performances, with gradient-boosted decision trees performing significantly better than linear models for electrocorticography, while Wiener filters have been repeatedly shown to perform well for subthalamic local field potentials (e.g. see Shah et al., 2018 IEEE Trans Neural Syst Rehabil Eng). The present study does not compare different machine learning architectures. Thus, decoding performances could potentially be further improved with more refined computational approaches. A more thorough overview of the literature from the many laboratories that are invested in this research across the globe would have improved the interpretation with respect to the broader impacts of the present manuscript.

5. The authors also present analyses of the spatial localization of relative decoding performances. They demonstrate higher tremor decoding performance in the dorsolateral subthalamic nucleus and higher decoding performance for the slowness of movement in the more central and ventral subthalamic regions. The authors interpret this as potential evidence to support clinical decision-making for optimized stimulation control of these symptoms at the respective locations. This is overly speculative and currently not backed by the data. First of all, the results only show the contrast of tremor vs. slowness of movement and not each individually. Thus, the spatial peak with each symptom domain could be very similar, e.g. in the dorsolateral STN, but a reversal of the difference only occurs at relatively low performances, e.g. in the ventral STN. Thus, showing both spatial distributions individually could be more informative. However, the claim that this could also be used to adjust stimulation location to alleviate the respective target symptoms is by no means backed by the data and remains an interesting speculation.

6. Finally, as in many brain signal decoding studies, the presented decoding performances are relatively low. The authors decided to present linear correlation metrics as Pearson's r values. These values are by definition higher than the commonly chosen Coefficient of determination or R² that provides a more interpretable performance metric. The amount of variance in the symptom scores that could be explained by the models ranged between 10% and 30% at a temporal resolution of 7 seconds. Moreover, the validity of the linear score is not entirely clear as Pearson's r can be heavily biased by non-normal distributions which were not assessed or at least not reported for the performance evaluation. These considerations do not severely limit the validity of the results themselves, as the authors have convincingly shown that significant decoding performances are possible and other studies in this field range in similar performance ranks. However, this point should remind us that a short-term clinical adoption of such methods is not yet in sight and further research is warranted. Before machine learning-based clinical computer interfaces can reach the clinical routine, the field has to work on more refined methods. In my opinion, the field will have to provide robust decoding performances with R² > 0.8 without patient-specific training to get into the realm of widespread clinical adoption.

Reviewer #1 (Public Review):

Temporal patterning allows a neural stem cell to generate different neural identities through the course of development. While this relationship has been demonstrated in many instances of stem cells and/or neurons, it is unclear how birth order translates to target specification. In this manuscript, the authors use live imaging and new tools generated from single-cell RNA sequencing data to address this issue.

They find that neurons born from a given time window (at the resolution of early>middle>late) innervate together - and distinctly from - those born at different temporal windows, though the specifics of the innervations differ between neural stem cell lineages. They also find that neurons achieve this by extending their dendrites in exploratory directions and selectively stabilising the ones in the appropriate direction. This process likely occurs at the sub-second timescales. Finally, they also demonstrate that embryonic-born (larval-specific) neurons that remodel to integrate into adulty-specific circuitry simultaneously perform pruning and dendrite extension to integrate into the circuitry at the appropriate time.

This is a valuable description of how developmental programmes imparted to neurons at the time of their birth might translate to their targetting and connectivity. It lays down a framework for understanding the defects in these processes.

It's crazy how easy this all seemed. Like how others mentioned, it does seem like it was premeditated, inside help. If it was always this easy, people would have struck before. But why now?

Connection: I agree with my classmate who mentioned this is more than just a citizen attack. I believe the government, the staff from the White House I would say, had something to do with this and were, in fact, involve because I don't think there is just a way citizens would know exactly what window, how were they just not reinforced? The White House just needs maximum MAXIMUM security. It shouldn't be this easy for citizens to break in, especially during elections.

This fact really surprised me because when I was watching this on the news I was in shock when I heard they broke through a window to get into the capital building. As previously stated there was a website called TheDonald.win and it probably included the information on what window to go through. I'm assuming this information was from someone with power in the government so it was definitely more than just citizen involvement.

There doesn't seem to be a web version. Only mobile, and windows app. Personally I use Linux or web on the desktop

Greetings Colgate Faculty. Welcome to AnnotatED!

You are seeing this annotation because you have successfully activated Hypothes.is in your browser.

To add your annotations to the private AnnotatED - Colgate space, select the group in the upper left hand corner of the Hypothes.is sidebar.

Emacs on windows<br /> by John D. Cook