In vivo transition in chromatin accessibility during differentiation of deep-layer excitatory neurons in the neocortex

  1. Seishin Sakai
  2. Yurie Maeda
  3. Mai Saeki
  4. Daijiro Konno
  5. Keita Kawaji
  6. Fumio Matsuzaki
  7. Yutaka Suzuki
  8. Yukiko Gotoh
  9. Yusuke Kishi

Reviewed by

Read the full article See related articles

Discuss this preprint

Start a discussion What are Sciety discussions?

Listed in

Log in to save this article

Abstract

During neuronal differentiation, gene transcription patterns change in response to both intrinsic and extrinsic cues. Chromatin regulation at regulatory elements plays a key role in this process. However, how chromatin accessibility evolves in vivo in cortical neurons remains unclear. Here, we established a method for labeling differentiating neurons with specific birthdates. Using this method, we traced the 4-day differentiation process of in vivo deep-layer excitatory neurons in the mouse embryonic cortex and examined changes in the genome-wide transcription pattern and chromatin accessibility using RNA sequencing and DNase sequencing, respectively. We found that genomic regions of genes linked to mature neuronal functions, including deep layer-specific and stimulus-responsive genes, became accessible even at the embryonic stage. Additionally, our results indicated the involvement of bivalent marks in neural precursor/stem cells and Dmrt3 and Dmrta2 in the regulation of chromatin accessibility during neuronal differentiation. These findings highlight the importance of chromatin regulation in embryonic neurons, enabling the timely activation of neuronal genes during maturation.

Article activity feed

  1. Version published to 10.1242/dev.204564
  2. Review Commons

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

    Learn more at Review Commons

    Reply to the reviewers

    1. General Statements [optional]

    We are grateful to the reviewers for their many valuable suggestions for improving this paper. In particular, we fully understand the points raised by Reviewers #1 and #2 regarding the insufficient data analysis and the points raised by Reviewers #2 and #3 regarding the insufficient analysis of the mechanism. In future revisions, we will perform sufficient analysis of our datasets and we will also conduct an analysis focusing on Dmrt3 to investigate the mechanisms for chromatin accessibility and changes in gene expression during neuronal differentiation. We will also make revisions to address other minor points.

    2. Description of the planned revisions

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

    The authors have developed a method for labeling a specific stage of differentiating neurons. Using this approach, they tracked the four-day differentiation process of deep-layer excitatory neurons in the mouse embryonic cortex. They investigated genome-wide changes in transcription patterns and chromatin accessibility using RNA-seq and DNase-seq. Additionally, they provided H3K4me3 and H3K27me3 ChIP-seq data from E12.0 NPCs. This resulting omics data would be a valuable resource for the field. While initial data analyses show potentially interesting findings, only part of the analyses are presented in the figures, lacking sufficient detail. Before publishing the manuscript, the authors should include more comprehensive analyses of their datasets. Specific suggestions are below.

    We appreciate this reviewer's positive comments describing our study as 'a valuable resource for the field.' We plan to revise the paper, as noted below, to address this reviewer's concerns.

    Figure 4 focuses on promoter-specific chromatin accessibility analysis. The author can process the data similarly to the transcription data. They should identify differentially accessible promoter regions across E13.0 to E16.0 and generate a heatmap with clustering. Additionally, the author should provide matched gene expression data, either in the form of a heatmap or box plot, corresponding to those differentially accessible promoter regions. Currently, Figure 4 only presents E16.0 data compared to E12.0, which is not comprehensive.

    We thank the reviewer for the useful suggestions. In the following submission, we will determine gene sets for all chromatin accessibility change patterns, not just open/closed gene sets from E12 to E16. We will then illustrate the changes in gene expression for each gene set.

    Reviewer #1 (Significance (Required)):

    Multi-omics data from the differentiation process of deep-layer excitatory neurons would be a valuable resource for the field.

    Once again, we would like to thank the reviewers for their positive comments.

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

    Summary: The manuscript from Sakai et al. examines changes in chromatin accessibility during the differentiation of deep-layer excitatory neurons in the neocortex. The authors establish a novel genetic labelling method that tracks differentiating neurons based on their birthdates allowing following neuronal differentiation in vivo. By combining RNA-seq and DNase-seq they provide a comprehensive dataset of gene expression and chromatin accessibility changes during neuronal differentiation of deep-layer neurons and reveal that key genes linked to mature neuronal functions and bivalent genes in neural precursor cells become accessible during early differentiation. These findings underscore the crucial role of chromatin regulation in preparing neurons for maturation and unravel novel key insights into the regulatory mechanisms governing deep-layer neuronal differentiation.

    Overall, this manuscript presents a novel technique for tracking neuron development from NPCs with specific birthdates. However, in its current form, it is largely descriptive and relies on correlative observations rather than elucidating a clear mechanism underlying chromatin and transcriptional changes. The provided data could be further leveraged to gain deeper insights into the molecular mechanisms governing deep-layer neuron development.

    We would like to thank the reviewer for recognizing the methods used in this paper as 'a novel technique for tracking neuron development from NPCs with specific birthdates'. As the reviewer commented, this paper was descriptive, and we plan to prepare a revised version that includes results that approach 'the molecular mechanisms governing deep-layer neuron development' by analyzing the role of Dmrt3 in neuronal differentiation, as shown in the response below, especially for point 9.

    Major comments:

    The authors have generated extensive RNA- and DNAse-seq datasets across different developmental time points following birthdate labelling. However, the bioinformatics analyses and interpretations are limited and need further clarification and refinement:

    The violin plots used to demonstrate expression and accessibility changes across developmental time points and the conclusions drawn from them are not convincing. The authors used a rank test to assess significant changes in expression, which only indicates the enrichment of genes with increased or decreased expression in each group. This cannot be directly interpreted as "significant upregulation." For instance, in Figures 4a and 4b, similar violin plots yield different statistical outcomes. The mean values on both graphs are comparable, yet Figure 4a suggests significant changes, while Figure 4b does not conclude significant downregulation of closing DHS genes. This is unconvincing. A more robust approach would be identifying DEGs between time points and analysing functional terms associated with these genes. The current plots do not support interpretations of gene upregulation, as each dot represents a gene, and the violin plot serves more as a population representation. The authors should either revisit their explanations and conclusions or include additional analyses and appropriate plots that support their claims of significant upregulation and downregulation of specific genes during development. We would like to thank the reviewer for their helpful suggestions on presenting the data in Figure 4 more effectively. In future reanalysis, we will add an analysis focusing on DEGs, as suggested by the reviewer. Specifically, we will examine the overlap between DEGs identified by RNA-seq and genes with altered chromatin accessibility and test this using Fisher's exact test and other methods. This will allow us to verify the conclusions of this paper from multiple perspectives.

    Figure 6b lacks clarity regarding the cutoff value used to categorise genes as K4me3 and K27me3 negative or positive from the heatmap. Even the "K4me3 negative" cluster displays a detectable signal of the mark, albeit at lower levels. Since only one plot of the entire gene body is provided, it is unclear what levels of enrichment are present, particularly at the promoter region. The authors are encouraged to provide additional informative plots and analyses of this ChIP-seq experiment, as this is a critical point where they draw conclusions about bivalent genes. This would not only strengthen their claims but could also uncover additional findings with more detailed analyses. A heatmap of clustered ChIP-seq signals of K4me3 and K27me3 alongside expression levels of the same genes (similar to Figure 2c) and differential accessibility (e.g., between NPC and E16) would better visualise and correlate histone modifications with chromatin and gene expression states.

    We would also like to thank this reviewer for their useful suggestions regarding Figure 6. In the next submission, we will try different methods to quantify H3K4me3 and H3K27me3 signals. Specifically, we plan to try methods using peak calling and methods that quantify signals in promoter regions.

    We also plan to show new figures for changes in gene expression and chromatin accessibility in gene sets categorized by H3K4me3 and H3K27me3 signals.

    The DNase-seq dataset can be better utilised to investigate differentially accessible motifs through development. Is this something the authors already looked into? This could strengthen mechanism investigation together with the ChIP-atlas results in Fig.6a

    In the revised version, we will perform motif analysis and ChIP-atlas analysis for all genomic region sets showing differential accessibility. We will then use the results obtained to discuss the mechanisms of chromatin accessibility changes during the neuronal differentiation process in more depth.

    The two distinct modes of H3K4me3 enrichment observed are not addressed and should be explained. Which genes belong to these two clusters? Is there a difference in DHS and gene expression between them?

    In relation to point 2 of this reviewer, we will also re-analyze the differences in H3K4me3 patterns and changes in gene expression and chromatin accessibility. We believe that we can answer this reviewer's questions through the analyses using peak calling and signal quantification, as described in point 2.

    The same concern regarding the use of violin plots to correlate gene expression with bivalent genes through development (Figure 6c) as mentioned earlier. It would be better to use DEGs and intersect them. This is particularly important given the wide range of gene expression levels in the already poised state.

    In relation to this reviewer's point 1, we will also perform a reanalysis focusing on DEGs in Figure 6.

    The authors limited their analyses to promoter/gene body regions. A survey of the bivalent marks and accessibility at enhancer regions would be also beneficial for understanding the changes at the chromatin landscape through development.

    The results of Figure 3 showed that chromatin accessibility in the promoter region changes significantly during neuronal differentiation, and this paper has focused on the promoter region. However, as this reviewer has commented, we have realized that analysis of enhancers is also useful. We plan to re-analyze the changes in chromatin accessibility in the enhancer region for the revised version.

    The mechanisms driving the activation and expression of poised neuronal genes through the development of deep-layer neurons is not uncovered. The authors suggest certain histone modifiers and the DNA methyltransferase Dnmt3 as potential drivers of chromatin landscape and transcriptional regulation changes; however, this remains speculative, as there is no direct evidence or validation of these factors binding to the identified target regions or changes in DNA methylation states. The authors should provide validation of their candidate factors' presence at potential targets, as well as changes in DNA methylation if they want to conclude these as the mechanisms driving deep-layer neuron development.

    We thank the reviewer for pointing out the critical issue of the mechanism for the activation of poised genes. We agree that investigating the mechanism in more depth would improve our paper.

    To this end, we will analyze the role of Dmrt3, not Dnmt3, in activating poised genes. Dmrt3 is a transcription factor mainly involved in transcriptional repression, and our RNA-seq results indicate that it is highly expressed in NPCs, and its expression decreases during neuronal differentiation. Therefore, Dmrt3 may suppress poised genes in NPCs. Indeed, our preliminary results using public data have shown that knocking out Dmrt3 increased the expression of poised genes.

    In future analyses, we plan to analyze the role of Dmrt3 using RNA-seq data from Dmrt3 knockout NPCs and Dmrt3 ChIP-seq data from NPCs.

    Minor comments:

    The motif analysis can be included in the main figures.

    We appreciate the reviewer's positive suggestions. Regarding point 9, we will move the results of the motif analysis to the main figure after reanalysis about Dmrt3.

    Reviewer #2 (Significance (Required)):

    By introducing a novel genetic labelling method that tracks neurons based on their birthdates, the study provides a precise way to examine differentiation in vivo, adding valuable insights beyond traditional in vitro approaches. The combination of RNA-seq and DNase-seq analyses reveals how chromatin accessibility changes, particularly in bivalent genes, play a crucial role in neuronal maturation. This work highlights the importance of chromatin dynamics in establishing neuronal identity. The techniques and findings provide a useful framework for future studies, offering a path for deeper exploration of chromatin regulation across different neuronal types, stages of development, or disease contexts, making it a valuable contribution to the field of developmental neurobiology.

    While the manuscript suggests the involvement of chromatin regulators such as Trithorax and Polycomb proteins, as well as Dnmt3 and DNA methylation, it lacks direct mechanistic evidence, such as ChIP-seq, bisulfite-seq, or loss-of-function experiments, to substantiate these claims.

    The bioinformatics analyses and interpretations are limited and require further clarification and refinement.

    The proposed mechanisms are not fully explored, leaving the manuscript largely descriptive rather than providing a detailed mechanistic understanding.

    We would like to thank the reviewer again for their various suggestions for improving our manuscript. By performing the experimental plan described above, we try to resolve the reviewer's concerns and improve this paper.

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

    In this manuscript the authors use in utero electroporation of tamoxifen inducible reporters to permanently mark cortical neurons with a common birthdate. They then FACS harvest these cells for bulk DNAse seq and RNA seq to see changes in chromatin regulation and gene expression as these newborn immature cortical neurons become deep layer neurons. As has been shown in prior studies that have addressed other neuronal types or used different methods to isolate developmental cell stages in the CNS, the authors find correlated changes between the opening or closing of chromatin with changes in gene expression. They use this information to localize chromatin marks that are associated with the differential expression of genes and conclude that many of the differential genes are bivalent for active and repressive chromatin marks. Finally the authors cross this dataset with a microarray they did of BDNF-inducible genes in cortical culture and suggest enrichment of this program in the differentially regulated gene set from in vivo.

    Reviewer #3 (Significance (Required)):

    The idea that chromatin regulation coordinates developmental changes in gene expression in neurons has been addressed with several different strategies over the past decade including prior strategies that allow for isolation of neurons with common birth dates. Many current strategies (well cited by the authors) use single cell sequencing and computational algorithms to deconvolve differentiation state from complex mixtures. This study takes an alternative approach to experimentally label these developmental stages which is nice to see for the validation of ground truth. However the study does not go far beyond current knowledge to use this method to add new concepts to the field. The main point of innovation seems to be the observation that the newborn neurons are primed at the chromatin level to express deep layer markers at the time they are born during embryonic life. This is useful to see but not unexpected on the basis of large scale single cell datasets. They also show that bivalent promoters prime developmental stage specific gene expression (in addition to the well-established function of this form of regulation in fate determination), however this too has been shown already in other neuron types.

    We are very pleased that the reviewer evaluated our method as 'nice to see for the validation of ground truth' and distinguished it from the current mainstream method to trace the differentiation process computationally using single-cell analysis that tracks. On the other hand, we also agree with the reviewer's assessment that our results do not exceed previous knowledge. Therefore, as mentioned in our response to Reviewer #2, we plan to analyze the role of Dmrt3 in gene expression and chromatin structure during the neuronal differentiation process. This will allow us to clarify the novel insight into the neuronal differentiation process.

    In addition to these conceptual limitations, there are some poorly supported comments in the text. For example, the fact that their microarray shows some genes in a category called "apoptosis" that are BDNF-sensitive does not meaningful suggest that BDNF induces excitotoxicity in embryonic cortical culture. BDNF has been well established as a survival factor for many kinds of neurons and is a common additive to serum-free media supplements (like B27). The appearance of "apoptosis" terms in the upregulated genes on the microarray more likely suggests either that the microarray is a poor detector of differential gene expression or that the genes in question are inaccurately categorized as "apoptotic" (GO terms are not terribly specific indicators of gene function). If the authors really wanted to test if BDNF was inducing apoptosis their cultures they could test this. However to use only the GO term data in such a strong statement about the biology of their system caused me to question the rigor of either their data or their analysis.

    We are grateful to the reviewers for their important comments. We also agree that BDNF is an important neurotrophic factor and do not believe that it induces cell death. Therefore, we checked the following 40 genes, which showed chromatin closing from E12 to E16, upregulation upon BDNF stimulation, and the GO term 'programmed cell death'.

    Cdip1, Diablo, Pla2g6, Braf, Tnfrsf25, Pa2g4, Mcl1, Hpn, Cebpb, Epha2, Plk3, Herpud1, Crip1, Dusp1, Sphk1, Irf5, Bag3, Stil, Fosl1, Cadm1, Lhx3, Hip1r, Relt, Irs2, Bmp8a, Ptcra, Mef2d, Prkcz, Rnf41, Pcid2

    As a result, we found that there were no genes involved in the main pathway of apoptosis. From this, we understand that the GO terms related to cell death are listed in Figure 5f because 'the genes in question are inaccurately categorized as "apoptotic" ', as this reviewer pointed out.

    We apologize for the misleading discussion in the previous manuscript and would like to thank the reviewer again for realizing this important point. We have corrected this in the new manuscript (page 9, line 263).

    In addition, we will perform a reanalysis to confirm this conclusion of chromatin opening at neuronal activity-associated gene loci using public gene expression analysis data of neuronal stimulation.

    A second example is the section about promoters being the focus of their discussion for DHS sites. Sure figure 3c shows promoters are more likely to be open compared with their contribution to the genome overall, but this is entirely expected since they are major gene TF binding sites, which is what DNAse detects. However promoters do not look to be more likely to be differentially regulated over time (3c vs 3e), and the statement that promoters are more enriched in opening compared with closing sites would require a statistical statement. Distal DHS sites appear equally more abundant in opening sites too.

    We thank the reviewer for their thoughtful comments on our results. As the reviewer points out, the proportion of promoter regions in the opening DHS in Figure 3e is not so high compared to that in Figure 3c. However, as described in the Abstract and Introduction sections, we are interested in how neurons acquire their function during the differentiation process, and our main focus was on comparing neuron-specific and NPC-specific DHS here. In the comparison within Figure 3e, it is clear that the opening DHS has a higher proportion of promoter regions than the closing DHS. We made the necessary revisions to avoid any misunderstanding on this point (page 7, line 192).

    On the other hand, as noted in the discussion, we are also interested in the role of the alteration in distal DHS. As in our response to Reviewer #2, we also plan to analyze changes in DHS in enhancer regions.

    3. Description of the revisions that have already been incorporated in the transferred manuscript

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

    In Figure 1c, the actual values of the differentially expressed genes are unclear. Is this a Z-score? Please provide the log2 expression values and specify the scale used for the heatmap and clustering.

    We apologize for the unclear expression value of Figure 2c. As this reviewer pointed out, the heatmap shows the Z-score, and we provided the actual scale in the new figure.

    Figure 5: It is somewhat unusual that the authors used microarray instead of RNA-seq for the BDNA stimulation of in vitro cortical neurons. Please provide a justification for this choice.

    Gene expression analysis using microarrays is a well-established technique, though it is currently unfamiliar. Compared to RNA-seq, microarrays have the disadvantage that they can analyze only RNAs with probes and have a lower dynamic range. However, on the other hand, they have the advantages of reasonable cost and a simpler analysis method. In this paper, we performed microarray analysis for BDNF experiment, considering these advantages.

    Figure 6: again, the data analyses are not comprehensively presented. What are the gene expression profiles of the other clusters (H3K27me3+, H3K4me3-/H3K27me3-, H3K4me3+)? Additionally, the sequencing data is inaccessible, and it is unclear how many samples (e.g., replicates) were used in this study for RNA-seq, DNase-seq, and ChIP-seq.

    We apologize for the lack of gene expression patterns of other clusters in Figure 6c. We provided them in the new figure and confirmed that only bivalent genes (H3K4me3+, H3K27me3+) showed increased gene expression levels during neuronal differentiation and other clusters slight reduction (new Figure 6c). This result again suggests that the bivalent state in NPCs contributes to their activation during neuronal differentiation.

              We described these data in the revised manuscript (page 10, line 296).

    Raw sequence datasets (fastq files) and processed data were deposited in the DNA Data Bank of Japan (DDBJ) Sequence Read Archive, a partner of International Nucleotide Sequence Database Collaboration (INSDC), as already described in the Data Availability section. Although DDBJ does not provide a reviewer access system for raw sequence datasets,

    the reviewer's access to the processed data is as follows.

    To review GEA accession E-GEAD-803, E-GEAD-859, E-GEAD-860:

    Please see the instructions below.

    https://www.ddbj.nig.ac.jp/gea/reviewer-access-e.html

    We will provide the access tokens in the final revised manuscript.

    For replicate numbers, we apologize for forgetting to describe them for the BDNF microarray experiment, though those for RNA-seq, DNase-seq, and ChIP-seq were already described in the Methods section. The replicates numbers are as follows:

    RNA-seq: two replicates

    DNase-seq: two replicates

    Microarray: three replicates

    ChIP-seq: two replicates

    We provided the replicate number of the microarray experiment in the revised manuscript (page 17, line 543).

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

    Major comments:

    The authors begin by examining TFs enriched at E16 DHS regions and suggest that TrxG and PcG factors are highly enriched in neurons, initiating their investigation of bivalent marks. However, they later conclude that bivalent marks are present in the NPC state and later become accessible. It is unclear why PRC factors would be enriched at the neuronal stage when the authors conclude that the chromatin becomes more open (potentially by removal of K27me3). The authors should refine this section of the manuscript to better rationalise their methodology and results.

    We are grateful to the reviewers for pointing out our poor explanation in Figure 6.

    This section aimed to investigate the mechanism by which open genomic regions in E16 were established. We used ChIP-atlas to investigate the transcription factors enriched in the E16 DHS and found many of the components of TrxG and PcG in the previous experiments using ES cells, which are the stem cells as NPCs. Therefore, we hypothesized that binding both TrxG and PcG, meaning a bivalent state, in NPCs may be important for chromatin opening until E16.Therefore, we analyzed bivalent genes in NPCs rather than E16 neurons in Figure 6b-d.

    We explained the rationale in detail in the revised version (page 9-10, line 269-288).

    Do the authors find any expressional changes of the suggested candidate proteins at the RNA or protein levels through development?

    We thank this reviewer for the useful suggestions. We agree that changes in the expression of TrxG and PcG components during neuronal differentiation are important information for considering the mechanism of chromatin structural changes in bivalent genes. Therefore, we checked the expression levels of genes encoding components of PcG or TrxG, determined by Schuettengruber et al., Cell, 2017, in our RNA-seq dataset (new Supplementary Data 5). More than half of them showed significant alteration, suggesting the possible contribution of alteration in the activity of PcG or TrxG or both on chromatin opening.

              We described this point in the revised manuscript (page 12, line 370).

    Minor comments:

    1. The manuscript would improve with proofreading by a native English speaker.

    We have already had proofreading by a native English speaker performed. We will also do it when submitting the revised version.

    4. Description of analyses that authors prefer not to carry out

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

    One additional point, which may be beyond the scope of this paper, is that to demonstrate the temporal resolution of this birthdate tracking method robustly, the authors should also apply the technique to upper-layer neuron development and compare developmental differences that were previously challenging to capture due to lower resolution.

    Reviewer #2 (Significance (Required)):

    The study focuses exclusively on deep-layer excitatory neurons, without comparisons to other neuronal subtypes or non-neuronal cells. Including such comparisons would help determine whether the observed chromatin changes are unique to this specific population or part of a broader developmental process.

    We are grateful for the reviewer's meaningful suggestions. We also think that by comparing with upper-layer neurons and non-neuronal cells, we can more comprehensively understand the development of the cerebral cortex . However, this paper primarily focuses on deep-layer neurons, and analysis of upper-layer neurons and non-neuronal cells will be future work.

    We described this point in the revised manuscript (page 13, line 384).

  3. Review Commons

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

    Learn more at Review Commons

    Referee #3

    Evidence, reproducibility and clarity

    In this manuscript the authors use in utero electroporation of tamoxifen inducible reporters to permanently mark cortical neurons with a common birthdate. They then FACS harvest these cells for bulk DNAse seq and RNA seq to see changes in chromatin regulation and gene expression as these newborn immature cortical neurons become deep layer neurons. As has been shown in prior studies that have addressed other neuronal types or used different methods to isolate developmental cell stages in the CNS, the authors find correlated changes between the opening or closing of chromatin with changes in gene expression. They use this information to localize chromatin marks that are associated with the differential expression of genes and conclude that many of the differential genes are bivalent for active and repressive chromatin marks. Finally the authors cross this dataset with a microarray they did of BDNF-inducible genes in cortical culture and suggest enrichment of this program in the differentially regulated gene set from in vivo.

    Significance

    The idea that chromatin regulation coordinates developmental changes in gene expression in neurons has been addressed with several different strategies over the past decade including prior strategies that allow for isolation of neurons with common birth dates. Many current strategies (well cited by the authors) use single cell sequencing and computational algorithms to deconvolve differentiation state from complex mixtures. This study takes an alternative approach to experimentally label these developmental stages which is nice to see for the validation of ground truth. However the study does not go far beyond current knowledge to use this method to add new concepts to the field. The main point of innovation seems to be the observation that the newborn neurons are primed at the chromatin level to express deep layer markers at the time they are born during embryonic life. This is useful to see but not unexpected on the basis of large scale single cell datasets. They also show that bivalent promoters prime developmental stage specific gene expression (in addition to the well-established function of this form of regulation in fate determination), however this too has been shown already in other neuron types.

    In addition to these conceptual limitations, there are some poorly supported comments in the text. For example, the fact that their microarray shows some genes in a category called "apoptosis" that are BDNF-sensitive does not meaningful suggest that BDNF induces excitotoxicity in embryonic cortical culture. BDNF has been well established as a survival factor for many kinds of neurons and is a common additive to serum-free media supplements (like B27). The appearance of "apoptosis" terms in the upregulated genes on the microarray more likely suggests either that the microarray is a poor detector of differential gene expression or that the genes in question are inaccurately categorized as "apoptotic" (GO terms are not terribly specific indicators of gene function). If the authors really wanted to test if BDNF was inducing apoptosis their cultures they could test this. However to use only the GO term data in such a strong statement about the biology of their system caused me to question the rigor of either their data or their analysis.

    A second example is the section about promoters being the focus of their discussion for DHS sites. Sure figure 3c shows promoters are more likely to be open compared with their contribution to the genome overall, but this is entirely expected since they are major gene TF binding sites, which is what DNAse detects. However promoters do not look to be more likely to be differentially regulated over time (3c vs 3e), and the statement that promoters are more enriched in opening compared with closing sites would require a statistical statement. Distal DHS sites appear equally more abundant in opening sites too.

  4. Review Commons

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

    Learn more at Review Commons

    Referee #2

    Evidence, reproducibility and clarity

    Summary:

    The manuscript from Sakai et al. examines changes in chromatin accessibility during the differentiation of deep-layer excitatory neurons in the neocortex. The authors establish a novel genetic labelling method that tracks differentiating neurons based on their birthdates allowing following neuronal differentiation in vivo. By combining RNA-seq and DNase-seq they provide a comprehensive dataset of gene expression and chromatin accessibility changes during neuronal differentiation of deep-layer neurons and reveal that key genes linked to mature neuronal functions and bivalent genes in neural precursor cells become accessible during early differentiation. These findings underscore the crucial role of chromatin regulation in preparing neurons for maturation and unravel novel key insights into the regulatory mechanisms governing deep-layer neuronal differentiation.

    Overall, this manuscript presents a novel technique for tracking neuron development from NPCs with specific birthdates. However, in its current form, it is largely descriptive and relies on correlative observations rather than elucidating a clear mechanism underlying chromatin and transcriptional changes. The provided data could be further leveraged to gain deeper insights into the molecular mechanisms governing deep-layer neuron development.

    One additional point, which may be beyond the scope of this paper, is that to demonstrate the temporal resolution of this birthdate tracking method robustly, the authors should also apply the technique to upper-layer neuron development and compare developmental differences that were previously challenging to capture due to lower resolution.

    Major comments:

    The authors have generated extensive RNA- and DNAse-seq datasets across different developmental time points following birthdate labelling. However, the bioinformatics analyses and interpretations are limited and need further clarification and refinement:

    1. The violin plots used to demonstrate expression and accessibility changes across developmental time points and the conclusions drawn from them are not convincing. The authors used a rank test to assess significant changes in expression, which only indicates the enrichment of genes with increased or decreased expression in each group. This cannot be directly interpreted as "significant upregulation." For instance, in Figures 4a and 4b, similar violin plots yield different statistical outcomes. The mean values on both graphs are comparable, yet Figure 4a suggests significant changes, while Figure 4b does not conclude significant downregulation of closing DHS genes. This is unconvincing. A more robust approach would be identifying DEGs between time points and analysing functional terms associated with these genes. The current plots do not support interpretations of gene upregulation, as each dot represents a gene, and the violin plot serves more as a population representation. The authors should either revisit their explanations and conclusions or include additional analyses and appropriate plots that support their claims of significant upregulation and downregulation of specific genes during development.
    2. Figure 6b lacks clarity regarding the cutoff value used to categorise genes as K4me3 and K27me3 negative or positive from the heatmap. Even the "K4me3 negative" cluster displays a detectable signal of the mark, albeit at lower levels. Since only one plot of the entire gene body is provided, it is unclear what levels of enrichment are present, particularly at the promoter region. The authors are encouraged to provide additional informative plots and analyses of this ChIP-seq experiment, as this is a critical point where they draw conclusions about bivalent genes. This would not only strengthen their claims but could also uncover additional findings with more detailed analyses. A heatmap of clustered ChIP-seq signals of K4me3 and K27me3 alongside expression levels of the same genes (similar to Figure 2c) and differential accessibility (e.g., between NPC and E16) would better visualise and correlate histone modifications with chromatin and gene expression states.
    3. The DNase-seq dataset can be better utilised to investigate differentially accessible motifs through development. Is this something the authors already looked into? This could strengthen mechanism investigation together with the ChIP-atlas results in Fig.6a
    4. The two distinct modes of H3K4me3 enrichment observed are not addressed and should be explained. Which genes belong to these two clusters? Is there a difference in DHS and gene expression between them?
    5. The same concern regarding the use of violin plots to correlate gene expression with bivalent genes through development (Figure 6c) as mentioned earlier. It would be better to use DEGs and intersect them. This is particularly important given the wide range of gene expression levels in the already poised state.
    6. The authors limited their analyses to promoter/gene body regions. A survey of the bivalent marks and accessibility at enhancer regions would be also beneficial for understanding the changes at the chromatin landscape through development.
    7. The authors begin by examining TFs enriched at E16 DHS regions and suggest that TrxG and PcG factors are highly enriched in neurons, initiating their investigation of bivalent marks. However, they later conclude that bivalent marks are present in the NPC state and later become accessible. It is unclear why PRC factors would be enriched at the neuronal stage when the authors conclude that the chromatin becomes more open (potentially by removal of K27me3). The authors should refine this section of the manuscript to better rationalise their methodology and results.
    8. Do the authors find any expressional changes of the suggested candidate proteins at the RNA or protein levels through development?
    9. The mechanisms driving the activation and expression of poised neuronal genes through the development of deep-layer neurons is not uncovered. The authors suggest certain histone modifiers and the DNA methyltransferase Dnmt3 as potential drivers of chromatin landscape and transcriptional regulation changes; however, this remains speculative, as there is no direct evidence or validation of these factors binding to the identified target regions or changes in DNA methylation states. The authors should provide validation of their candidate factors' presence at potential targets, as well as changes in DNA methylation if they want to conclude these as the mechanisms driving deep-layer neuron development.

    Minor comments:

    1. The manuscript would improve with proofreading by a native English speaker.
    2. The motif analysis can be included in the main figures.

    Significance

    By introducing a novel genetic labelling method that tracks neurons based on their birthdates, the study provides a precise way to examine differentiation in vivo, adding valuable insights beyond traditional in vitro approaches. The combination of RNA-seq and DNase-seq analyses reveals how chromatin accessibility changes, particularly in bivalent genes, play a crucial role in neuronal maturation. This work highlights the importance of chromatin dynamics in establishing neuronal identity. The techniques and findings provide a useful framework for future studies, offering a path for deeper exploration of chromatin regulation across different neuronal types, stages of development, or disease contexts, making it a valuable contribution to the field of developmental neurobiology.

    While the manuscript suggests the involvement of chromatin regulators such as Trithorax and Polycomb proteins, as well as Dnmt3 and DNA methylation, it lacks direct mechanistic evidence, such as ChIP-seq, bisulfite-seq, or loss-of-function experiments, to substantiate these claims.

    The study focuses exclusively on deep-layer excitatory neurons, without comparisons to other neuronal subtypes or non-neuronal cells. Including such comparisons would help determine whether the observed chromatin changes are unique to this specific population or part of a broader developmental process. The bioinformatics analyses and interpretations are limited and require further clarification and refinement. The proposed mechanisms are not fully explored, leaving the manuscript largely descriptive rather than providing a detailed mechanistic understanding.

  5. Review Commons

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

    Learn more at Review Commons

    Referee #1

    Evidence, reproducibility and clarity

    The authors have developed a method for labeling a specific stage of differentiating neurons. Using this approach, they tracked the four-day differentiation process of deep-layer excitatory neurons in the mouse embryonic cortex. They investigated genome-wide changes in transcription patterns and chromatin accessibility using RNA-seq and DNase-seq. Additionally, they provided H3K4me3 and H3K27me3 ChIP-seq data from E12.0 NPCs. This resulting omics data would be a valuable resource for the field. While initial data analyses show potentially interesting findings, only part of the analyses are presented in the figures, lacking sufficient detail. Before publishing the manuscript, the authors should include more comprehensive analyses of their datasets. Specific suggestions are below.

    In Figure 1c, the actual values of the differentially expressed genes are unclear. Is this a Z-score? Please provide the log2 expression values and specify the scale used for the heatmap and clustering.

    Figure 4 focuses on promoter-specific chromatin accessibility analysis. The author can process the data similarly to the transcription data. They should identify differentially accessible promoter regions across E13.0 to E16.0 and generate a heatmap with clustering. Additionally, the author should provide matched gene expression data, either in the form of a heatmap or box plot, corresponding to those differentially accessible promoter regions. Currently, Figure 4 only presents E16.0 data compared to E12.0, which is not comprehensive.

    Figure 5: It is somewhat unusual that the authors used microarray instead of RNA-seq for the BDNA stimulation of in vitro cortical neurons. Please provide a justification for this choice.

    Figure 6: again, the data analyses are not comprehensively presented. What are the gene expression profiles of the other clusters (H3K27me3+, H3K4me3-/H3K27me3-, H3K4me3+)? Additionally, the sequencing data is inaccessible, and it is unclear how many samples (e.g., replicates) were used in this study for RNA-seq, DNase-seq, and ChIP-seq.

    Significance

    Multi-omics data from the differentiation process of deep-layer excitatory neurons would be a valuable resource for the field.

  6. Version published to 10.1101/2024.08.26.609258 on bioRxiv