Reorganization of postmitotic neuronal chromatin accessibility for maturation of serotonergic identity

Version of Record

Accepted for publication after peer review and revision.

Download
Cite
Share
CommentOpen annotations (there are currently 0 annotations on this page).

Version of Record published: May 12, 2022 (This version)
Accepted Manuscript published: April 26, 2022 (Go to version)
Accepted: April 12, 2022
Received: November 29, 2021
Preprint posted: October 14, 2021 (Go to version)

1. Of interest
Engineering of the endogenous HBD promoter increases HbA2

Mandy Y Boontanrart, Elia Mächler ... Jacob E Corn

Research Article Jun 2, 2023
Further reading

Article
Figures and data
Abstract
Editor's evaluation
Introduction
Results
Discussion
Materials and methods
Data availability
References
Decision letter
Author response
Article and author information
Metrics

Abstract

Assembly of transcriptomes encoding unique neuronal identities requires selective accessibility of transcription factors to cis-regulatory sequences in nucleosome-embedded postmitotic chromatin. Yet, the mechanisms controlling postmitotic neuronal chromatin accessibility are poorly understood. Here, we show that unique distal enhancers define the Pet1 neuron lineage that generates serotonin (5-HT) neurons in mice. Heterogeneous single-cell chromatin landscapes are established early in postmitotic Pet1 neurons and reveal the putative regulatory programs driving Pet1 neuron subtype identities. Distal enhancer accessibility is highly dynamic as Pet1 neurons mature, suggesting the existence of regulatory factors that reorganize postmitotic neuronal chromatin. We find that Pet1 and Lmx1b control chromatin accessibility to select Pet1-lineage-specific enhancers for 5-HT neurotransmission. Additionally, these factors are required to maintain chromatin accessibility during early maturation suggesting that postmitotic neuronal open chromatin is unstable and requires continuous regulatory input. Together, our findings reveal postmitotic transcription factors that reorganize accessible chromatin for neuron specialization.

Editor's evaluation

This study will be of interest to developmental biologists who study the gene regulatory mechanisms necessary for induction and maintenance of post-mitotic neuronal identity. The authors generated a useful resource of genomic data and provided new insights into the dynamic regulation of accessible chromatin regions in post-mitotic serotonin (5-HT) neurons of the mouse hindbrain. This work proposes two transcription factors (Pet1, Lmx1b) as necessary for establishment and maintenance of accessible chromatin regions in serotonin neurons.

https://doi.org/10.7554/eLife.75970.sa0

Introduction

Specialized neuronal identities are acquired through the differential expression of genes that specify transmitter usage, build dendritic and axonal architecture, and control synapse formation (Hobert, 2011). The maturation of neuronal functions is an extended process that plays out over several weeks of mid fetal to early postnatal life in mice and through the first few years of life in humans (Stiles and Jernigan, 2010), as postmitotic neuronal transcriptomes mature to a stable state. Unique combinations of sequence-specific transcription factors (TFs) selectively activate and repress genes to assemble the transcriptomes that encode individual neuron identities (Polleux et al., 2007). Genes with variants associated with autism and other neurodevelopmental disorders exhibit enriched expression in maturing neurons and are often involved in chromatin remodeling and transcriptional regulation; their developmental dysfunction is thought to contribute to disease pathogenesis (De Rubeis et al., 2014; Heavner and Smith, 2020).

Selective expression of neuronal identity genes depends on selective accessibility of promoter and enhancer sequences embedded in postmitotic nucleosomal-organized chromatin to regulatory factor inputs (Gallegos et al., 2018; Perino and Veenstra, 2016). Chromatin accessibility is dynamic as brain regions (Gorkin et al., 2020; de la Torre-Ubieta et al., 2018) and neurons (Frank et al., 2015; Preissl et al., 2018; Stroud et al., 2020; Trevino et al., 2021) mature. Yet, we have a poor understanding of how the dynamics of accessible chromatin in postmitotic neurons prefigures acquisition of neurotransmitter identities and myriad other molecular and morphological features of specialized neurons. Further, the regulatory factors involved in selecting postmitotic accessible chromatin regions to enable sequence-specific activation of unique combinations of identity features have not been identified.

Control of 5-HT neuron development is of particular interest as 5-HT has expansive modulatory effects on central neural circuitry (Celada et al., 2013). Altered serotonergic gene expression, brought about by either genetic or environmental factors, has been implicated in neuropsychiatric disorders including depression, stress-related anxiety disorders, autism, obsessive compulsive disorder, and schizophrenia (Deneris and Wyler, 2012). Postmitotic 5-HT neuron precursors terminally differentiate in the ventral hindbrain to acquire the capacity for 5-HT synthesis, reuptake, vesicular transport, and degradation through the coordinate expression of the 5-HT neurotransmission genes Tph2, Ddc, Gch1, Slc6a4 and Slc22a3, Slc18a2, and Maoa/b (Deneris and Gaspar, 2018). Subsequently, newborn 5-HT neurons enter a maturation stage during which serotonergic transcriptomes are highly dynamic as 5-HT neurons refine their terminal identities, migrate to populate the various raphe nuclei, acquire mature firing properties, and establish expansive connectivity throughout the brain and spinal cord (Deneris and Gaspar, 2018).

The ETS domain factor Pet1 and the LIM homeodomain (HD) factor Lmx1b are terminal selector TFs in 5-HT neurons as they are continuously expressed beginning at the postmitotic precursor stage, control their own expression through direct positive autoregulation, and are required for the acquisition of 5-HT neuron terminal identity through sequence-specific transcriptional activation of 5-HT pathway genes (Deneris and Gaspar, 2018; Hobert, 2008). Pet1 and Lmx1b are also required for the acquisition of serotonergic firing characteristics and formation of long-distance profusely arborized 5-HT axon architectures throughout the brain and spinal cord (Donovan et al., 2019; Liu et al., 2010; Wyler et al., 2016). Pet1’s function in 5-HT neurons is clinically significant as recent studies identified biallelic loss of function of FEV, the human ortholog of Pet1, in autism cases suggesting that defects in transcriptional control of 5-HT neurons are a potential path to human neurodevelopmental disorders (Doan et al., 2019).

The low abundance of many postmitotic neuron types has hindered the assay of their accessible chromatin landscapes and cis-regulatory architectures. Here, we adapted bulk and single-cell ATAC-seq (scATAC-seq), scRNA-seq, and ChIPmentation protocols to map accessible chromatin and cis-regulatory elements (CREs) of developing 5-HT neurons. We then used these combined maps along with Pet1 and Lmx1b knockout mouse models and occupancy mapping to identify a previously unrecognized function of these TFs in regulating the chromatin accessibility of 5-HT neurotransmission and synapse genes. Together, our findings show that Pet1 and Lmx1b control the maturation of 5-HT connectivity not only through sequence-specific activation of select terminal effector genes but also by reorganizing postmitotic accessible chromatin at putative CREs.

Results

The unique accessible chromatin landscape of maturing postmitotic Pet1 neurons

5-HT neurons are generated from Pet1-positive postmitotic precursors located in the embryonic ventral hindbrain from the isthmus to the caudal region of the myelencephalon (Alonso et al., 2013; Hendricks et al., 1999). Therefore, we used the Pet1-EYFP transgenic mouse line to flow sort E14.5 Yfp-positive neurons (Pet1 neurons) from the entire longitudinal extent of the hindbrain for assay of accessible chromatin in maturing 5-HT neurons (Scott et al., 2005a; Scott et al., 2005b; Figure 1—figure supplement 1a-c). We generated ATAC-seq biological replicates with a mean of 29M unique paired-end mapped reads per sample (Figure 1—figure supplement 1d-g) and identified 68,871 peaks of transposase accessible chromatin (TAC) at E14.5, and 59,323 TACs in an equal number of control Yfp-negative, Pet1-negative cells (‘Pet1^neg’) collected from the same hindbrain regions at the same developmental age, which comprise diverse non-serotonergic cell types. Principal component analysis (PCA) showed that cell type identity accounted for 93% of the variance when comparing Pet1 neuron ATAC-seq replicates with Pet1^neg ATAC-seq samples and with previously published ATAC-seq datasets obtained from bulk tissue dissections (Figure 1a; Gorkin et al., 2020). The accessible chromatin landscape of Pet1 neurons is distinct from previously published bulk hindbrain ATAC-seq (Gorkin et al., 2020), which indicates that tissue level analyses did not capture the chromatin accessibility profiles of low abundance Pet1 neurons (Figure 1a). Indeed, of the TACs we identified in Pet1 neurons at E14.5, 31% are unique (Pet1 neuron-specific TACs) and not found in Pet1^neg cells and 22% are not previously identified in any other cell type (Gorkin et al., 2020; Figure 1b).

Figure 1 with 1 supplement see all

Download asset Open asset

Unique distal enhancers and super-enhancers define the *Pet1* lineage.

(a) Principal component analysis (PCA) reveals the unique accessible chromatin landscape of *Pet1* neurons. Dots of the same color represent biological replicates. (b) Bar plot showing percentage of unique ATAC-seq peaks (transposase accessible chromatin [TAC] regions) from E14.5 *Pet1* neurons when compared to cis-regulatory elements (CREs) annotated by ENCODE datasets (Gorkin et al., 2020) (left bar) or with transposase accessible chromatins (TACs) in E14.5 Pet1-negative (*Pet1*^neg) cells from the same hindbrain region (right bar). (c) Emission probabilities for histone modifications in 10 ChromHMM states and the percent genomic coverage of each chromatin state for E14.5 *Pet1* neuron-specific TACs and *Pet1* neuron TACs shared with other tissues in ENCODE ATAC-seq datasets.(d) Distribution of H3K27ac signal across all enhancers identified in E14.5 *Pet1* neurons. Enhancers with exceptionally high read counts (bracket) were designated super-enhancers. (e) Bar plot of *Pet1* neuron super-enhancers that are unique to *Pet1* neurons or overlap (‘shared’) with annotated enhancers from ENCODE datasets. (f) Genome browser view of the *Pet1* neuron-specific super-enhancer across the *Lmx1b* locus defined as an extended stretch of H3K27ac and H3K4me1 enrichment.(g) Genome browser views of ATAC-seq and ChIPmentation signals at the *Pet1/Fev* locus. Red bar denotes the conserved *Pet1* enhancer region that is sufficient to direct 5-HT neuron-specific transgene expression (Scott et al., 2005a; Scott et al., 2005b; Krueger and Deneris, 2008). (h) Gene ontology (GO) of *Pet1* neuron-specific TACs (not found in other tissues in ENCODE ATAC-seq datasets). (i) Diagram of 5-HT neurotransmission. (j) Genome browser views of ATAC-seq and ChIPmentation peaks at 5-HT neurotransmission genes. Light blue shading highlights *Pet1* neuron-specific TACs.

To determine the functions of 5-HT neuron TACs, we next used ChIPmentation (Schmidl et al., 2015) to profile the genome-wide distribution of histone posttranslational modifications associated with transcriptional activation or repression in flow sorted E14.5 Pet1 neurons. With the segmentation algorithm ChromHMM (Ernst and Kellis, 2017), we defined 10 chromatin states by assessing the combinatorial pattern of histone marks at each genomic locus (Figure 1c; Figure 1—figure supplement 1h-i). Active and poised distal enhancer regions are over-represented among the Pet1 neuron-specific TACs (Figure 1c). Our ChIPmentation analysis also identified 1919 genomic regions that fulfill the criteria for super-enhancers based on the presence of unusually expansive H3K27ac signals (Figure 1d; Lovén et al., 2013; Whyte et al., 2013). Comparing these to previously annotated enhancers in other cell types (Gorkin et al., 2020), we found that a significant percentage (42%) of the super-enhancers we detected in Pet1 neurons were not detected in bulk mouse forebrain, hindbrain, or liver tissues (Figure 1e), suggesting that they are highly specific to the Pet1 lineage. As an example, a 82.9 kb super-enhancer detected in Pet1 neurons but not Pet1^neg neurons spans the Lmx1b locus (Figure 1f).

Notably, a distinct TAC was found immediately upstream of the Pet1 promoter in Pet1 neurons but not in Pet1^neg cells (Figure 1g). This TAC shows enrichment of H3K4me1 and H3K27ac marks suggesting the presence of an active enhancer (Figure 1g). Previous transgenic assays, in vivo, demonstrated that this upstream region directs robust reporter expression to developing and adult 5-HT neurons present in each of the midbrain, pontine, and medullary raphe nuclei (Krueger and Deneris, 2008; Scott et al., 2005a; Scott et al., 2005b), hence providing functional validation of serotonergic enhancers we have defined with ATAC-seq and ChIPmentation. As further validation, we intersected all identified Pet1 neuron TACs with the VISTA enhancer database (Visel et al., 2007). Of all experimentally validated VISTA enhancers that show activity in the hindbrain, 38% overlap with Pet1 neuron TACs (Figure 1—figure supplement 1j). However, only 1% of hindbrain-annotated VISTA enhancers overlap with Pet1 neuron-specific TACs, thus demonstrating that analyses of chromatin accessibility in low abundance neuron types can reveal previously unrecognized CREs (Figure 1—figure supplement 1j). Importantly, genes associated with Pet1 neuron-specific TACs are enriched for axon and synapse functions (Figure 1h). Indeed, we found Pet1 neuron-specific TACs for 5-HT neurotransmission genes (Figure 1i–j). Together, our ATAC-seq and ChIPmentation analyses define the open chromatin landscape of maturing 5-HT neurons and reveal that distal TACs putatively assigned to genes encoding axon and synapse-related functions define the Pet1 postmitotic neuron lineage.

Heterogeneous TAC accessibility and gene expression define Pet1 neuron subtypes at embryonic stages

Adult 5-HT neurons possess diverse transcriptomes that encode functional subtypes (Okaty et al., 2015; Okaty et al., 2020; Ren et al., 2019; Wylie et al., 2010), yet the development of 5-HT neuron heterogeneity is poorly understood. To determine whether the 5-HT neuron accessible chromatin is heterogeneous early in development, we carried out scATAC-seq using E14.5 flow sorted Pet1 neurons, which were processed using the 10× Genomics Chromium microfluidic droplet method. The resulting scATAC-seq library was sequenced to a depth sufficient to yield >50,000 paired-end reads per nucleus. After filtering based on standard quality control measures (Figure 2—figure supplement 1a), we retained 1692 nuclei for analysis. This led to the identification of 124,111 TACs, 60,117 of which were not identified in our bulk Pet1 neuron ATAC-seq (Figure 2a). We observed high concordance (Spearman’s correlation = 0.9) between our pseudo-bulk scATAC-seq dataset and our bulk ATAC-seq data as illustrated by the highly similar TAC patterns in the Pet1 upstream enhancer region (Figure 2b). In parallel we also performed scRNA-seq using E14.5 flow sorted Pet1 neurons. We obtained over 80,000 reads per cell for 1612 cells, with a median of 4010 genes per cell and a total of 19,707 genes across all cells (Figure 2—figure supplement 1b); 95% of the flow sorted Pet1 neurons robustly expressed both Pet1 and Yfp, validating that our cell sorting method yields nearly pure samples of Pet1 neurons. The small numbers of Yfp-negative cells were removed from further analysis.

Figure 2 with 2 supplements see all

Download asset Open asset

Heterogeneity in chromatin accessibility and transcriptomes reveals transcriptional programs determining *Pet1* neuron subtype identities.

(a) Venn diagram showing overlap of single-cell *Pet1* neuron TACs vs. bulk ATAC-seq TACs. (b) Genome browser view of *Pet1* locus showing high concordance between the pseudobulk scATAC-seq and bulk FACS-purified *Pet1* neuron ATAC-seq at E14.5. (c) scATAC-seq tracks showing the aggregated chromatin accessibility peaks of *En1* (left) and *Meis2* (right) for each cluster. (d) Uniform manifold approximation and projection (UMAP) visualization of chromatin accessibility of single E14.5 *Pet1*-lineage neurons. scATAC-seq is projected onto scRNA-seq UMAP space. Cells are colored by scATAC-seq cluster assignment. Genes associated with the top two marker TACs for each cluster are labeled. (e) The chromatin accessibility profiles of select genes across UMAP visualized scATAC-seq clusters. (f) UMAP visualization of single E14.5 *Pet1* neurons based on gene expression. Cells are colored by scRNA-seq cluster assignment. (g) The expression profile of example 5-HT neuron identity genes across all scRNA-seq clusters in UMAP. (h) Number of TAC-gene associations for all genes. (i) Representative gene with high number of associated TACs. Loops represent statistically significant TAC-gene association. Loop height represents the p-value of TAC-gene correlation. (j) Number of significantly correlated TACs for each gene. Genes are ranked by the numbers of associated peaks. Select TFs with known or potential functions in the *Pet1* lineage are labeled.

We grouped single Pet1 neurons to identify 11 major subtypes with unique chromatin accessibility and expression patterns (Figure 2d and f; Figure 2—figure supplement 1c-d). Pet1 and Lmx1b loci TACs are robustly accessible in virtually all cells within the 11 identified subtypes (Figure 2e). Like Pet1 and Lmx1b, two-thirds (63%) of TACs are broadly accessible among multiple subtypes, including TACs of 5-HT neurotransmission genes Tph2, Slc22a3, and Slc6a4 (Figure 2e). In contrast, one-third of TACs (36.6%) are unique to a single cluster. Subtype-specific TACs are predominantly novel distal gene enhancers as previously defined by chromatin state (Figure 1c; Figure 2—figure supplement 1e-g) and are often observed near genes that mark known 5-HT neuron subtypes (Figure 2c; Figure 2—figure supplement 1c). For example, TF En1 is enriched in the rostral rhombencephalic 5-HT neurons, whereas Hox gene cofactor Meis2 is selectively expressed in the caudal rhombencephalic 5-HT neurons (Wylie et al., 2010). Six of the scATAC-seq clusters share a unique TAC region immediately upstream of the TSS of the rostral marker gene En1 while the remaining five clusters have more accessible TAC regions near the promoter of the caudal 5-HT neuron marker gene Meis2 (Figure 2c–e). To compare our embryonic scRNA-seq clustering results to published adult scRNA-seq data, we performed a unimodal uniform manifold approximation and projection (UMAP) of the adult 5-HT scRNA-seq data from Okaty et al., 2020, onto our E14.5 dataset (Figure 2—figure supplement 1h). Since the adult scRNA-seq data was obtained from the dorsal raphe nuclei (DRN), which is solely derived from the isthmus and rhombomere 1, we predicted that the adult scRNA-seq data would largely map to our rostral clusters expressing high levels of Tph2 and En1. Indeed, 93% of the adult DRN cells map to the E14.5 Tph2/Slc6a4 and Pax5/En1 clusters that have high expression of rhombomere 1 markers (Figure 2—figure supplement 1h). Thus, our data demonstrate that diverse single-cell 5-HT neuron chromatin landscapes and transcriptomic profiles are established early in postmitotic Pet1 neurons and likely prefigure the heterogeneity of adult 5-HT neurons at least in the DRN.

Mapping accessible chromatin regions to genes

To analyze the association between CRE accessibility and target gene expression, we co-embedded scATAC-seq with scRNA-seq data, then linked TACs to a gene if logistic regression detected a statistical association between gene expression and the binarized chromatin accessibility at a genomic region within 500 kb of a gene (Fang et al., 2021). Applying this method to our single-cell datasets yielded a total of 41,191 TAC-gene pairings (Source data 1). At this developmental time point, each gene is on average linked to only 1.8 TACs and most genes are associated with either none or a single TAC (Figure 2h). However, 10% of genes are associated with >5 TAC regions, and a select group of genes (0.045%) are associated with >20 TACs (Figure 2h–j; Figure 2—figure supplement 2a). Gene ontology (GO) analysis of the top 1500 genes with the highest number of linked CREs revealed that they are highly enriched for genes involved in axon and synapse development (Figure 2—figure supplement 2b), suggesting that a large number of CREs converge to fine-tune or safeguard the transcription of key neuronal regulators (Ma et al., 2020; Trevino et al., 2021). Among these genes are known TFs of the 5-HT neuron gene regulatory network such as Pet1 and Lmx1b (Figure 2j).

We identified many subtype-specific TFs constituting a potential TF code that directs 5-HT neuron heterogeneity (Figure 2—figure supplement 2c-d). As examples, Nkx6-1 gene expression, chromatin accessibility, and ChromVAR-inferred binding at NK motifs are enriched in the Nkx6-1/Ndnf cluster, while expression, chromatin accessibility, and inferred binding of the homeodomain TF Meis2 are enriched in the Meis2/Penk cluster (Figure 2—figure supplement 2c-d; Schep et al., 2017). In contrast, the expression and accessibility of TFs such as Pet1/Fev and Lmx1b are more evenly distributed among all Pet1 neuron subtypes (Figure 2e and g; Figure 2—figure supplement 2d). Overall, most subtype-specific TACs are correlated with the subtype-specific expression pattern of their linked genes; however, we noticed that the expression of 5-HT identity genes Tph2, Slc6a4, and Slc22a3 are low or undetectable in most Pet1 neurons at E14.5 despite their relatively uniform accessible chromatin in Pet1 neurons at this stage (Figure 2e and g). The vast majority of Pet1-lineage neurons robustly express these 5-HT identity genes in the adult mouse brain (Figure 2—figure supplement 2e; Donovan et al., 2019; Okaty et al., 2015; Okaty et al., 2019; Ren et al., 2019). Tph2 immunofluorescence analysis at E14.5 revealed that Tph2-positive cells emerge across all raphe nuclei with particular enrichment in the dorsal raphe (Figure 2—figure supplement 2f). This suggests that chromatin accessibility may be remodeled early during maturation to prime neuron identity before the transcription of terminal effector genes.

A highly dynamic stage of postmitotic accessible chromatin reorganization

We next investigated whether Pet1 neurons are born with fully mature chromatin that is accessible to all regulatory factors needed for the eventual transcription of 5-HT genes. In this case, developmental gene expression trajectory is likely supported by the dynamic interactions of transcriptional regulatory factors with DNA binding sites. Alternatively, Pet1 neurons are perhaps born with immature chromatin that must be remodeled to generate the mature transcriptomes needed to encode adult characteristics. We performed bulk Pet1 neuron ATAC-seq (Figure 3—figure supplement 1a-b) to compare TACs at E11.5 when postmitotic 5-HT neuron precursors are predominant, E14.5 when Pet1 neurons are actively extending dendrites and axons, and E17.5 when Pet1 neurons are coalescing to form the raphe nuclei and innervating distal target brain regions (Donovan et al., 2019). PCA showed high concordance of ATAC-seq replicates within each time point with developmental state accounting for the majority of variance (Figure 3a). We found that significant gains and losses in chromatin accessibility of TSS-distal TACs occurred throughout embryonic maturation (Figure 3b–d; Figure 3—figure supplement 1c-d,i). Of the TACs detected in Pet1 neurons, 16,493 open, that is, gain accessibility (fold change >2 and FDR < 0.01) between E11.5 and E14.5, while 4148 open from E14.5 to E17.5 (Figure 3c–d). Concurrently, 9,402 E11.5 TACs close, that is, lose accessibility (fold change >2 and FDR < 0.01) by E14.5, and 1504 TACs close between E14.5 and E17.5 (Figure 3c–d). This suggests that chromatin remodeling is highly dynamic immediately following cell cycle exit and then gradually stabilizes. Assessment of chromatin states showed that active enhancers are over-represented among TACs opening between E11.5 and E17.5 (Figure 3b; Figure 3—figure supplement 1e-g). Pet1 neuron super-enhancer accessibility is particularly highly dynamic during postmitotic neuron maturation, with the vast majority (93%) of TACs either gaining or losing accessibility between E11.5 and E17.5 (Figure 3—figure supplement 1h).

Figure 3 with 1 supplement see all

Download asset Open asset

Gene regulatory dynamics in maturing *Pet1* neurons.

(a) PCA showing the distribution of *Pet1* neuron ATAC-seq data over two principal components. (b) Distribution of chromatin states that exhibit gain (green), loss (red), or no change (blue) in chromatin accessibility from E11.5 to E17.5. (c) Number of TACs detected at each developmental time point (circles) and the number of TACs that are gained, lost, or propagated to the next time point (arrows). In parenthesis are the numbers of TACs that gain or lose >2-fold accessibility with FDR < 0.01. (d) Volcano plots showing the differential chromatin accessibility of TACs between consecutive time points, with fold change cutoff of 2-fold and significance cutoff of FDR < 0.01. Each dot represents one ATAC-seq peak (TAC). (e) Box plots showing the distribution of the log2 expression values of the linked genes for TACs that exhibit loss (blue), no change (black), or gain (red) in accessibility between E11.5 and E17.5. Significance of accessibility to gene expression correlation is calculated by Wilcoxon rank sum test. (f) Genome browser view of ATAC-seq and E14.5 ChIPmentation signals at representative genes that decrease (e.g. *Ascl1*) or increase (e.g. *Ntrk2*) in accessibility between E11.5 and E17.5. Parallel trends are seen in gene expression trajectories (panels on the right of tracks). (g) Top five GO biological process enrichment terms for genes linked to TACs that lose (left) or gain (right) accessibility from E11.5 to E17.5. (h) Bar graphs showing the enrichment ratio of the top gene ontology biological process enrichment terms for genes linked to TACs that lose (blue) or gain (orange) accessibility from E11.5 to E14.5 and from E14.5 to E17.5. (i) Heatmap showing the log2 fold change in chromatin accessibility for TACs of select genes associated with the indicated GO biological process terms for each developmental stage. Representative genes in GO terms from (h) were selected for display in (i).

Next we compared our bulk ATAC-seq datasets at E11.5, E14.5, and E17.5 with our previously published bulk 5-HT neuron RNA-seq data obtained at E11.5, E15.5, and postnatal days 2–3 (Wyler et al., 2016). Using our TAC-to-gene assignments derived from single-cell data (Figure 2i–j), we found that changes in chromatin accessibility at cis-regulatory regions are significantly correlated with the temporal expression trajectories of their target genes (Figure 3e–f). GO analysis revealed that genes associated with closing TACs are important for neural precursor functions, whereas the genes linked to opening TACs are strongly enriched for axon development between E11.5 and E14.5 and synaptic signaling between E14.5 and E17.5 (Figure 3g–i). Dynamic TACs are highly enriched for diverse TF binding motifs, suggesting that combinatorial TF activities contribute to Pet1 neuron maturation (Figure 3—figure supplement 1j). Together these findings support a model in which the dynamic remodeling of CRE accessibility drives gene expression trajectories required for maturation of 5-HT identity while shutting down expression of genes required at earlier stages of development.

Pet1 reorganizes chromatin accessibility of 5-HT neurotransmission genes

The highly dynamic reorganization of TACs in maturing postmitotic Pet1 neurons predicts the existence of regulatory factors that shape postmitotic neuronal chromatin landscapes. However, such factors have not been described. Notably, TACs that are unique to Pet1-lineage neurons are enriched for ETS-factor binding sites (Figure 4—figure supplement 1a). Thus, we hypothesized that Pet1 controls acquisition of 5-HT identity by controlling chromatin accessibility. Pet1^-/- 5-HT neurons are an ideal model with which to examine TF function in shaping the chromatin, as these neurons are retained in the brain in normal numbers and Pet1-EYFP transgene expression in these cells is robustly retained at fetal stages to permit their isolation and analysis (Wyler et al., 2016). PCA of ATAC-seq datasets obtained from Pet1-EYFP-positive, Pet1^-/- hindbrain cells at E11.5, E14.5, and E17.5 (Figure 4—figure supplement 1b-c) showed that Pet1-deficient Pet1-lineage neurons (Pet1^-/- neurons) cluster closely with wildtype Pet1 neurons (Figure 4a; Figure 4—figure supplement 1d), consistent with our observation that Pet1-deficient 5-HT neurons do not adopt an overt alternative identity. However, we found a striking overall reduction of Pet1 footprints and decreased accessibility of many TSS-distal TACs in Pet1^-/- neurons compared to wildtype neurons at all three embryonic stages (Figure 4b–c; Figure 4—figure supplement 1e). To determine whether Pet1 directly controls chromatin accessibility in these chromatin regions, we performed CUT&RUN with E14.5 flow sorted Pet1 neurons to generate a genomic occupancy map for Pet1 (Figure 6—figure supplement 1a-c). The CUT&RUN map significantly overlapped with our previous ChIP-seq of Pet1 (Figure 4—figure supplement 1f) but identified occupancy at several additional 5-HT neurotransmission genes that are controlled by Pet1 (Figure 4i–j; Wyler et al., 2016). CUT&RUN data showed that Pet1 directly occupies at least 53% of Pet1-dependent TACs (Figure 4d–e). Moreover, 73% of all Pet1-regulated TACs and 80% of Pet1-regulated TACs with Pet1 occupancy are TACs unique to Pet1 neurons (Figure 4f).

Figure 4 with 1 supplement see all

Download asset Open asset

Pet1 reorganizes postmitotic chromatin accessibility.

(a) PCA showing the distribution of ATAC-seq data for wildtype and *Pet1^-/- Pet1*-lineage neurons (*Pet1^-/-* neurons) over two principal components. (b) In vivo Pet1 footprints derived from Tn5 insertion frequency (ATAC-seq reads) over representative ETS-factor motifs within *Pet1* neuron accessible chromatin regions. *Pet1^-/-* neurons show reduced Pet1 footprints at E11.5 (left), E14.5 (middle), and E17.5 (right) compared to wildtype neurons for Pet1-regulated TACs. (c) Box plots showing the distribution of the log2 fold change in accessibility of the TACs that show differential accessibility between *Pet1^-/-* vs. wildtype *Pet1* neurons at three developmental time points. *Pet1^-/-* neurons show an overall decrease in chromatin accessibility. (d) The top significantly enriched motif within Pet1 CUT&RUN peaks as determined by de novo MEME motif analysis significantly matches (p = 5.33e-06) the Pet1/FEV high-affinity binding site previously defined in vitro (Wei et al., 2010) based on analysis by the TOMTOM Motif Comparison Tool (top panel). (e) The fraction of Pet1-regulated TACs containing Pet1 CUT&RUN peaks. (f) The fraction of all *Pet1* neuron TACs, *Pet1* neuron TACs that are dependent on Pet1, and Pet1-dependent TACs occupied by Pet1 that are specific to *Pet1*/5-HT lineage or shared with *Pet1*^neg cells. A significant fraction of Pet1-regulated TACs is unique to *Pet1* neurons and absent from *Pet1*^neg cells. (g) Volcano plot showing the differential chromatin accessibility of TACs between *Pet1^-/-* and wildtype *Pet1* neurons at E14.5, with fold change cutoff of 2-fold and significance cutoff of FDR < 0.01 (top). Highest FDR GO terms for Pet1-dependent TACs (in red) that are occupied by Pet1 show strong enrichment for serotonin biosynthesis. Select genes required for 5-HT synthesis are labeled on the volcano plot. (h) The numbers of gain and loss *Pet1^-/-* neuron TACs that map to TACs specific to a subtype of *Pet1* neurons or shared among multiple subtypes in scATAC-seq. (**i–j**) Genome browser tracks showing the gene-enhancer pairings and the ATAC-seq, Pet1 CUT&RUN, and histone modification ChIPmentation signals at 5-HT identity genes. Zoomed-in inserts display the changes in chromatin accessibility between wildtype (red) and *Pet1^-/-* (green) *Pet1* neurons within the highlighted (blue) regions at three developmental time points. Two TACs of the gene *Slc6a4* (red dashed line box) gain accessibility in wildtype 5-HT neurons from E11.5 to E17.5 but fail to open to the same extent in *Pet1^-/-* neurons. Asterisks denote FDR < 0.01 and fold change >2 relative to wildtype at the same time point.

Pet1-regulated TACs are predominantly those shared among multiple subtypes of Pet1-lineage neurons (Figure 4h). GO analysis revealed that the Pet1 occupied TACs that close in Pet1^-/- neurons are associated with genes required for serotonin synthesis (Figure 4g; Figure 4—figure supplement 1g). Indeed, we found a significant loss of accessibility of enhancers either upstream of the TSS or within introns of Tph2, Gch1, and Ddc (Figure 4i). TACs associated with the serotonin reuptake transporter genes Slc22a3 and Slc6a4 are also significantly reduced in accessibility (Figure 4j). Importantly, TACs that require Pet1 for accessibility are mostly unique to Pet1 neurons, while TACs of 5-HT pathway genes that are not Pet1 neuron-specific are not significantly affected by Pet1 deficiency (Figure 4i–j). These findings indicate that Pet1 controls the chromatin accessibility of Pet1 neuron-specific TACs of 5-HT neurotransmission genes.

Interestingly, the enhancers associated with Slc6a4 that are directly occupied by Pet1 show little accessibility to transposase at E11.5 but gradually gain accessibility from E11.5 to E17.5 in wildtype Pet1 neurons (Figure 4j). In contrast, these TAC regions fail to fully open in Pet1^-/- neurons (Figure 4j), suggesting Pet1 is directly required for initial conversion of these regions to a euchromatic state. Among the Pet1-regulated TACs, we identified 47 regions that similarly fail to fully open between E11.5 and E17.5 in Pet1^-/- neurons, 20 of which show direct Pet1 occupancy at E14.5 (Figure 4—figure supplement 1h-i). TACs that require Pet1 for chromatin opening are linked to genes enriched for synaptic products, suggesting that Pet1 opens condensed chromatin at TACs controlling synapse functions (Figure 4—figure supplement 1h).

Continued Pet1 function is required to sustain 5-HT regulatory element accessibility

To determine whether early postmitotic neuronal accessible chromatin landscapes require continuous terminal selector input, we analyzed flow sorted E14.5 Pet1 conditional knockout (Pet1^fl/fl; Pet1-Cre; Rosa26^Ai9) Pet1-lineage neurons (‘Pet1-cKO neurons’) in which Pet1 transcript level is maintained for about 2 days till E12.5 (Liu et al., 2010; Figure 4—figure supplement 1j-l). We found 1834 Pet1-maintained TACs that are mainly distal CREs in wildtype Pet1 neurons lose their accessibility in Pet1-cKO (Figure 5a–d). The accessibility of 936 of these TACs does not change in wildtype neurons from E11.5 to E14.5, indicating that Pet1 is required to sustain TACs after they open (Figure 5a). Many 5-HT neurotransmission gene TACs lose accessibility as in Pet1^-/- neurons including the enhancers of 5-HT neurotransmission genes Tph2, Gch1, Ddc, and Slc22a3 (Figure 5f–g).

Figure 5

Download asset Open asset

Sustained Pet1 input is required for chromatin accessibility.

(a) Top: Volcano plot showing the differential chromatin accessibility of TACs between Pet1-cKO and wildtype *Pet1* neurons, with fold change cutoff of 2-fold and significance cutoff of FDR < 0.01 (in red). 936 TACs (in parenthesis) show no change in accessibility in wildtype *Pet1* neurons between E11.5 and E14.5 but show loss of accessibility in Pet1-cKO. Bottom: Highest FDR GO terms for the Pet1-regulated TACs (in red) that are occupied by Pet1 show enrichment for 5-HT biosynthesis. (b) Overlap between differential TACs in *Pet1^-/-* and Pet1-cKO 5-HT neurons. (c) Comparison of in vivo Pet1 footprints over representative ETS-factor motif between Pet1-cKO and wildtype *Pet1* neurons within all differentially accessible chromatin regions. (d) The ChromHMM chromatin states genomic coverage for TACs that are maintained by Pet1 at E14.5. (e) The numbers of gain and loss *Pet1-cKO Pet1* neuron TACs that map to TACs specific to a subtype of *Pet1* neurons or shared among multiple subtypes in scATAC-seq data. (**f–g**) Genome browser tracks of gene-enhancer pairings, ATAC-seq, Pet1 CUT&RUN, and histone modification ChIPmentation signals at 5-HT identity genes. Zoomed-in inserts display the changes in chromatin accessibility between wildtype (red) and Pet1-cKO (orange) 5-HT precursors within the highlighted (blue) regions at E14.5. Asterisks denote FDR < 0.01 and fold change >2 compared to wildtype. (h) Genome browser tracks showing the ATAC-seq, Pet1 CUT&RUN, and histone modification ChIPmentation signals at the *Pet1* locus.

In addition to selecting 5-HT identity genes for transcriptional activation, Pet1 maintains its expression through positive autoregulation (Scott et al., 2005a). CUT&RUN showed that Pet1 directly occupies the TAC that overlaps with its autoregulatory enhancer (Scott et al., 2005a; Scott et al., 2005b), which is consistent with our previous Pet1 ChIP-seq findings (Wyler et al., 2016; Figure 5h; Figure 4—figure supplement 1f). We found that Pet1 is required for optimal accessibility of this enhancer region and for sustaining its accessibility suggesting that maintenance of chromatin accessibility at autoregulatory enhancers is an important mechanism for terminal selector autoregulation (Figure 5h). Together, these findings suggest the early euchromatin landscape in postmitotic 5-HT neurons is a reversible state that requires ongoing Pet1 input for maintenance of accessibility.

Lmx1b co-regulates reorganization of chromatin accessibility at 5-HT identity genes

To determine whether reorganization of chromatin accessibility in postmitotic neurons is unique to Pet1, we assessed whether Lmx1b also controls chromatin architecture. Since Lmx1b occupancy in the brain has not been defined, we performed Lmx1b CUT&RUN in E14.5 Pet1 neurons, which revealed that a significant portion of Pet1 and Lmx1b binding sites coincide within the same TAC regions (Figure 6—figure supplement 1a-c). In Lmx1b-cKO (Lmx1b^fl/fl; Pet1-Cre;Rosa26^Ai9) Pet1 neurons (Donovan et al., 2019; Figure 6—figure supplement 1d-f), chromatin landscape is altered (Figure 6a), Lmx1b footprints are reduced (Figure 6b), and more TACs are differentially accessible than in Pet1-cKO neurons (Figures 5a and 6c), suggesting that Lmx1b has a broader impact than Pet1 on 5-HT neuron chromatin. Lmx1b directly binds 19% of the differentially accessible TACs in Lmx1b-cKO (comparable to Pet1 which binds 20% of differentially accessible TACs in Pet1-cKO) indicating direct maintenance of accessibility by Lmx1b at those regions (Figure 6d; Figure 4—figure supplement 1m), which are predominantly distal enhancers (Figure 6e). Like Pet1-regulated TACs, Lmx1b-regulated TACs are predominantly shared among multiple Pet1 neuron subtypes (Figure 6f). The Lmx1b-dependent TACs with direct Lmx1b occupancy, of which 65% are specific to Pet1 neurons (Figure 6i), are also most significantly associated with monoamine/serotonin synthesis (Figure 6c). Moreover, the conditional ablation of Lmx1b results in an even greater closing of 5-HT neurotransmission gene TACs than Pet1 ablation (Figure 6j–k). Many TACs whose accessibility was not significantly reduced in Pet1-cKO neurons (fold change >2, FDR < 0.01) significantly lost accessibility in Lmx1b-cKO neurons (Figure 6—figure supplement 1j) including the TACs of Gchfr, Slc18a2, and Maoa (Figure 6j–k). This suggests that although Lmx1b is known to activate Pet1 expression (Donovan et al., 2019), the reduction in Pet1 expression in the Lmx1b-cKO (Figure 6—figure supplement 1h) cannot entirely account for all the changes in chromatin accessibility in the Lmx1b-cKO neurons. Lmx1b directly occupies the enhancers of many 5-HT neurotransmission genes (Figure 6j–k) whose expression is known to be critically dependent on Lmx1b (Ding et al., 2003; Donovan et al., 2019).

Figure 6 with 1 supplement see all

Download asset Open asset

Lmx1b co-regulates chromatin accessibility with Pet1 at 5-HT neurotransmission genes but has broader impact on accessibility landscape.

(a) PCA showing the distribution of wildtype, *Pet1^-/-*, Pet1-cKO, Lmx1b-cKO, and DKO neuron ATAC-seq data over two principal components. (b) Comparison of in vivo Lmx1b footprints over representative LIM-HD motif, between Lmx1b-cKO and wildtype *Pet1* neurons within differentially accessible chromatin regions. (c) Top: Volcano plot showing the differential chromatin accessibility of TACs between Lmx1b-cKO and wildtype *Pet1* neurons, with fold change cutoff of 2-fold and significance cutoff of FDR < 0.01 (in red). 1228 TACs (in parenthesis) show no change in accessibility in wildtype *Pet1* neurons between E11.5 and E14.5 but show loss of accessibility in Lmx1b-cKO. Bottom: Highest FDR GO terms for Lmx1b-regulated TACs (in red) that are occupied by Lmx1b show enrichment of 5-HT biosynthesis. (d) The topmost significantly enriched motif within Pet1 CUT&RUN peaks as determined by de novo MEME motif analysis matches (p = 4.03E-02) the Lmx1b LIM-HD binding site previously defined in vitro (Jolma et al., 2013) based on analysis by the TOMTOM Motif Comparison Tool (left panel). Fraction of Lmx1b-maintained TACs containing Lmx1b CUT&RUN peaks (right panel). (e) The ChromHMM chromatin states genomic coverage for TACs that lose accessibility in Lmx1b-cKO (blue) and DKO (red) in E14.5 *Pet1* neurons. (f) The numbers of gain and loss *Lmx1b-cKO Pet1* neuron TACs that map to TACs specific to a subtype of *Pet1* neurons or shared among multiple subtypes in scATAC-seq data. (g) Comparison of the in vivo Pet1 (top) and Lmx1b (bottom) footprints between DKO and wildtype *Pet1* neurons within differentially accessible chromatin regions. (h) Top: Volcano plot showing the differential chromatin accessibility of TACs between DKO and wildtype *Pet1* neurons, with fold change cutoff of 2-fold and significance cutoff of FDR < 0.01 (in red). 815 TACs (in parenthesis) show no change in accessibility in wildtype *Pet1* neurons between E11.5 and E14.5 but show loss of accessibility in DKO. Bottom: Highest FDR GO terms for Pet1- and Lmx1b-regulated TACs (in red) that are occupied by the two TFs show enrichment for serotonin biosynthesis. (i) The fraction of Pet1-cKO, Lmx1b-cKO, and DKO differentially accessible TACs that is specific to *Pet1* lineage. A significant fraction of the Pet1- and Lmx1b-regulated TACs is unique to *Pet1* neurons and absent from *Pet1*^neg cells. (**j–k**) Genome browser tracks of gene-enhancer pairings, ATAC-seq, Pet1 CUT&RUN, and histone modification ChIPmentation signals at 5-HT identity genes. Display data ranges for the zoomed-out panels are scaled to the peak in the blue highlighted regions. Zoomed-in inserts display the changes in chromatin accessibility between wildtype (red), Lmx1b-cKO (orange), and DKO (orange) 5-HT precursors within the highlighted (blue) regions at E14.5. Asterisks denote FDR < 0.01 and fold change >2 compared to wildtype.

We also found that the loss of chromatin accessibility in Lmx1b-cKO is highly similar to that in Pet1 and Lmx1b double conditional knockout (DKO: Lmx1b^fl/fl; Pet1^fl/fl; Pet1-Cre; Rosa26^Ai9) Pet1 neurons (Figure 6b–c , and g–h), suggesting that the ablation of both TFs does not have greater effect than ablation of Lmx1b alone. PCA also indicates close clustering of Lmx1b-cKO neurons with DKO neurons (Figure 6—figure supplement 1g). These findings are consistent with the model that Lmx1b is required for Pet1-dependent chromatin regulation. At a substantial subset of Lmx1b-dependent TACs such as those of Tph2, Gch1, Ddc, and Slc6a4 (Figures 5f–g ,–6j–k), Pet1 and Lmx1b non-redundantly co-regulate the accessibility of the CREs. In parallel, Lmx1b further sustains, independent of Pet1, an additional subset of TACs that are required for axon and synaptic formation, neurotrophin signaling, and neuronal growth (Figure 6—figure supplement 1i-j).

Motif analysis on the differentially accessible regions in the cKO and DKO neurons showed that the binding sites of several TF families such as the Rfx-related factors are less accessible in Pet1- and Lmx1b-deficient Pet1 neurons, demonstrating that together Pet1 and Lmx1b shape the global gene regulatory landscape of available TF binding sites (Figure 6—figure supplement 1k).

Pet1 and Lmx1b regulate 5-HT synapse formation

In addition to controlling TACs associated with 5-HT transmission genes that define serotonergic identity (Figure 6h), Pet1 and Lmx1b program TAC accessibility of many genes involved in axon growth and synaptic functions (Figure 7a; Figure 7—figure supplement 1a). The majority of regulated TACs identified from Lmx1b-cKO and DKO neurons are those that gain accessibility in wildtype Pet1 neurons from E11.5 to E17.5 (Figure 7b). These include the TACs of Syt1, which encodes a calcium sensor protein that mediates the release of synaptic vesicles in response to calcium (Yoshihara and Littleton, 2002), and Slc17a8, which encodes a vesicular glutamate transporter (vGlut3) that enables glutamatergic co-transmission in a subset of adult 5-HT neurons (Vigneault et al., 2015; Figure 7c). The Pet1- and Lmx1b-dependent TACs of Syt1 and Slc17a8 are Pet1 neuron specific, suggesting that the expression of some widely expressed neuronal genes are regulated in different neurons by neuron-type-specific CREs (Figure 7c). In support of the contribution of these Pet1- and Lmx1b-regulated enhancers to developmental gene expression, Syt1 and Slc17a8 expression are decreased in Pet1-cKO, Lmx1b-cKO, and DKO neurons (Figure 7d). Other axon and synapse-related genes show similar correlation between TAC accessibility and gene expression (Figure 7—figure supplement 1b-d). Further, there is significant correlation between TACs that close in Lmx1b-cKO and DKO and the expression of the linked genes (Figure 7e). GO analysis of down-regulated genes in DKO neurons revealed that Pet1 and Lmx1b collectively maintain the transcriptome necessary for axon, synapse, and dendrite development (Figure 7—figure supplement 1b). Together these data suggest that in addition to controlling accessibility of putative 5-HT transmitter identity enhancers, Pet1 and Lmx1b modulate the accessibility of putative CREs for other 5-HT neuron functions.

Figure 7 with 1 supplement see all

Download asset Open asset

Loss of Pet1 and Lmx1b alters chromatin accessibility at synaptic gene enhancers and disrupts 5-HT neuron connectivity.

(a) GO analysis of all Pet1- and/or Lmx1b-maintained TACs identified from E14.5 DKO ATAC-seq datasets, depicted as volcano plot of enrichment (x-axis) vs. FDR statistical significance (y-axis), showing many synapse and axon biological processes among the top enriched annotations. (b) Percent of Pet1- and/or Lmx1b-regulated TACs that gain or lose accessibility (blue) or gain accessibility (red) in wildtype *Pet1* neurons from E11.5 to E17.5. For example, 33% of all TACs are dynamic and 22% gain accessibility. In contrast, 62% of Lmx1b-regulated TACs are dynamic and 50% gain accessibility. (c) Genome browser tracks of gene-enhancer pairings, ATAC-seq, Pet1 and Lmx1b CUT&RUN, and histone modification ChIPmentation signals at synaptic genes *Syt1* and *Slc17a8*. Tracks show the changes in chromatin accessibility between wildtype (red), Pet1-cKO (yellow), Lmx1b-cKO (yellow), and DKO (yellow) Pet1 neurons within the highlighted (blue) regions at E14.5. Display data ranges for the zoomed-out panels are optimized based on the blue highlighted regions. (d) Relative expression (FPKMs) of *Syt1* and *Slc17a8* in rostral Pet1-cKO, Lmx1b-cKO, and DKO neurons at E17.5. Data presented as mean ± SEM. *, FDR<.05. (e) Differential expression (from E11.5 to E17.5) of the genes associated with Pet1-occupied TACs in Pet1-cKO vs. wildtype 5-HT neurons (left), genes associated with Pet1-occupied TACs in Lmx1b-cKO vs. wildtype 5-HT neurons (middle), and genes associated with Pet1 and Lmx1b co-occupied TACs in DKO vs. wildtype 5-HT neurons (right). (f) TdTomato (red) and DAPI (blue) co-staining of pericellular baskets in the medial septum of P15 *Pet1-Cre; Rosa26^Ai9* wildtype, Pet1-cKO, Lmx1b-cKO, and DKO mice. (g) Diagram showing the analyzed brain area and quantification of TdTomato axons (pixels/μm²) (two-way ANOVA; n = 3 animals per genotype and 2 quantified brain sections per animal) (bottom panels). Images were generated through automatic stitching of individual 20× images. (h) Representative immunohistochemistry images of Syt1, TdTomato, and DAPI in the dorsolateral septum of P15 *Pet1-Cre; Rosa26^Ai9* mice showing the decreased overlap of Syt1 (green) and TdTomato (red) signals in KO animals compared to wildtype. Images were generated through automatic stitching of individual 63× images. (i) Representative immunohistochemistry images of vGlut3 (*Slc17a8*), TdTomato, and DAPI in the medial septum of adult *Pet1-Cre; Rosa26^Ai9* mice showing the decreased overlap of vGlut3 (green) and TdTomato (red) signals in KO animals compared to wildtype. Images were generated through automatic stitching of individual 63× images. (j) Quantification of Syt1 (left) and vGlut3 (right) co-localization with TdTomato in the medial septum as a percentage of the total TdTomato signal (two-way ANOVA; n = 2 animals per genotype and 3–4 quantified images per animal). *, P<.05; **, P<.01; ***, P<.001; ****, P<.0001 in panels e, g, j.

To explore the biological significance of Pet1 and Lmx1b-dependent TAC alterations, we analyzed the integrity of 5-HT pericellular baskets that encase postsynaptic neuronal dendrites and cell bodies in the dorsal lateral septum (dLS) (Senft and Dymecki, 2021). Lmx1b-cKO 5-HT neurons display axon defects in which most axonal projections fail to reach the dLS and remain stalled within the medial forebrain bundle during development. However, a small number of Lmx1b-cKO 5-HT axons do extend into the dLS where they come in close proximity to post-synaptic cell bodies (Donovan et al., 2019). We found that by postnatal day 15 (P15), 5-HT pericellular baskets were readily identified in wildtype brains. Further, as expected, some 5-HT axons reach the dLS in Pet1-cKO, Lmx1b-cKO, and DKO mice (Figure 7f). However, pericellular baskets could not be found in the remaining 5-HT axons (Figure 7f–g). Of the few axonal terminals that do reach postsynaptic cell bodies, co-localization of Syt1 and vGlut3 immunostaining with TdTomato-positive axon terminals is significantly reduced compared to wildtype animals (Figure 7h–j). This suggests that Pet1 and Lmx1b control accessibility of putative CREs for genes that are required for 5-HT axon extension and formation of presynaptic structure that enables 5-HT and glutamatergic co-transmission.

Discussion

Specialized molecular, morphological, and functional properties of neurons are largely acquired as postmitotic neurons mature (Polleux et al., 2007). How postmitotic epigenetic programs control gene expression trajectories to encode mature neuronal identities is poorly understood (Bradke et al., 2020; Gallegos et al., 2018). Here, we have defined the accessible chromatin landscape and the cis-regulatory architecture of developing Pet1 neurons that generate mature 5-HT neurons (Hendricks et al., 1999). We show that the cis-regulatory landscape of embryonic Pet1 neurons is highly dynamic as postmitotic Pet1 neurons acquire serotonergic identities. Our scATAC-seq and scRNA-seq analyses reveal that the chromatin landscapes of neurons derived from a single TF lineage are highly heterogeneous and established early to prefigure and likely encode 5-HT neuron subtypes including those expressing the glutamatergic marker, Slc17a8. Previous studies have demonstrated the key role of terminal selectors, Pet1 and Lmx1b, in the differentiation and maturation of 5-HT neurons through sequence-specific transactivation of serotonergic terminal effector genes. We show here that an additional functional attribute of Pet1 and Lmx1b is to reorganize postmitotic chromatin accessibility to enable sequence-specific activation of serotonergic neurotransmission and synapse genes. Thus, our findings identify developmental regulators of postmitotic neuronal chromatin accessibility.

We found that distal transcriptional CRE accessibility is heterogeneous not only between Pet1 neurons and other cell types but also within the Pet1-lineage neurons. Diversity in regulatory sequence accessibility is predictive of the transcriptomic profile of single Pet1 neurons. Furthermore, the maturation of Pet1 neuron putative CRE accessibility is highly correlated with the expression trajectories of their target genes as serotonergic functional properties progressively emerge. TACs linked to genes involved in synapse organization and axon development gain accessibility while those associated with cell cycle control and early embryonic morphogenesis lose accessibility. Importantly, the chromatin landscape is more dynamic between E11.5 and E14.5 than between E14.5 and E17.5, suggesting that the early postmitotic period comprises a highly active stage of chromatin maturation. Since Pet1 neurons have exited the cell cycle, the temporal changes in open chromatin cannot be explained as an altered abundance of Pet1 neuron subtypes. Rather, chromatin landscape within individual Pet1 neurons is dynamic during postmitotic maturation.

The regulatory factors controlling chromatin dynamics in postmitotic neurons are not well defined. TFs with known pioneer activity such as Ascl1 that are expressed in serotonergic progenitors are significantly downregulated or undetectable in postmitotic Pet1 neurons (Aydin et al., 2019; Wyler et al., 2016). Yet, our data suggests that new postmitotic euchromatic regions dynamically appear while other euchromatic regions are lost as 5-HT neuron identity is being acquired, suggesting the existence of regulatory factors that shape postmitotic neuronal chromatin. Additionally, several recent studies computationally mined the genome-wide accessible chromatin sequences and identified TF binding sites that are enriched in regions of dynamic chromatin in neurons or brain regions (Gray et al., 2017; Preissl et al., 2018; de la Torre-Ubieta et al., 2018; Trevino et al., 2021).

Our findings support a founding paradigm in which terminal selector TFs, at least Pet1 and Lmx1b in mouse 5-HT neurons, reorganize the postmitotic neuronal epigenome to select specific gene CREs. We show that targeting of Pet1 and Lmx1b in the early stages of postmitotic maturation significantly alters the Pet1 neuron accessible chromatin landscape. CUT&RUN assays suggest that Pet1 and Lmx1b directly control the accessibility of 5-HT-specific CREs linked to genes encoding terminal effectors of 5-HT identity and neurotransmission. In fact, we provide evidence in support of a role for Pet1 in initiating the conversion to euchromatin of multiple nucleosome-protected cis-regulatory sequences in postmitotic neurons, including enhancers linked to the serotonin transporter gene, Slc6a4. We further show that Pet1 is required for accessibility of its own autoregulated enhancer region that directs 5-HT neuron expression. Autoregulation is thought to boost and maintain terminal selector expression (Leyva-Díaz and Hobert, 2019). We found that in the absence of Pet1 expression, Pet1 enhancer accessibility is significantly reduced but is not fully closed. This suggests that Pet1 boosts its own expression through augmenting auto-enhancer accessibility that was previously initiated by upstream pioneers but was insufficient for proper expression levels of Pet1.

We present evidence suggesting Pet1 and Lmx1b are required to maintain 5-HT-specific open chromatin regions. Whereas pioneer factors in progenitor cells can drive cell lineage commitment and induce, through transient input, stable ‘memories’ of epigenetic states that are transmitted through further cell divisions (D’Urso and Brickner, 2014), our findings suggest that postmitotic accessible chromatin landscapes are unstable and require continuous terminal selector input at least during early postmitotic maturation to sustain the euchromatin state of CREs controlling terminal effector genes. Surprisingly, a larger number of chromatin accessibility changes were found after Pet1 conditional targeting compared to constitutive germline Pet1 targeting. We do not have a straightforward explanation for this difference but it may be due to compensatory effects from other regulators. While Lmx1b maintains accessibility of a distinct set of CREs independent of Pet1, chromatin accessibility of many Pet1-lineage TACs requires direct and non-redundant co-regulation by both TFs, reminiscent of the cooperative activities of TFs and pioneer factors in other cell types (Holmberg and Perlmann, 2012; Mazzoni et al., 2013; Zaret and Carroll, 2011). Because a large number of cognate motif sites can exist in the genome for each individual TF, selective chromatin accessibility controlled by collaborative postmitotic reorganizers may serve to ensure that neuron-type-specific effector genes are selectively activated only in the presence of multiple positive regulatory inputs. Other TFs may also perform similar function in other neuron types (Allaway et al., 2021). The mechanism by which Pet1 and Lmx1b program neuronal chromatin is not understood, but could involve the recruitment of chromatin remodelers or direct cooperative displacement of histones (Spitz and Furlong, 2012).

In addition to neurotransmitter synthesis, neurons must develop axonal and synaptic connectivity in order to transmit signals (Kratsios et al., 2015; Spencer and Deneris, 2017). Pet1 and Lmx1b reorganize the access to CREs for both 5-HT neurotransmission and synaptic connectivity. As a result, the acquisition of distinct functional properties is developmentally coupled. A growing body of evidence supports the idea that alterations in 5-HT neurotransmission cause behavioral changes related to those that characterize neuropsychiatric disorders such as anxiety and autism (Doan et al., 2019; Suri et al., 2015). As many of these disorders are neurodevelopmental in origin, the study of the chromatin regulatory programs that control 5-HT neuron and other identities may shed light on the pathophysiology of brain disorders at critical stages of brain development.

Materials and methods

Animals

Mice were maintained in ventilated cages on a 12 hr light/dark cycle with access to food and water ad libitum with two to five mice per cage. The transgenic mouse lines used in this study have been described previously (Donovan et al., 2019; Wyler et al., 2016). Conditional knockout mice (Lmx1b-cKO, Pet1-cKO, DKO) were generated by breeding animals bearing Rosa26^tdTomato (Ai9; Jackson Labs Stock #: 007909) and either the Lmx1b^fl/fl (Zhao et al., 2006), Pet1^fl/fl (Liu et al., 2010), or both alleles with animals harboring Pet1-Cre (original name ePet-Cre) transgene. Pet1^+/+; Pet1-EYFP and Pet1^-/-; Pet1-EYFP mice were generated by crossing Pet1^+/- bearing the Pet1-EYFP transgene. The Pet1-EYFP transgene expresses Yfp under the same Pet1 enhancer as the Pet1 enhancer region driving the Pet1-Cre. Genotyping was performed using ear tissues with the following primers: Pet-1: 5’-CGGTGGATGTGGAATGTGTGCG-3’, 5’-CGCACTTGGGGGGTCATTATCAC-3’, 5’-GCCTGATGTTCAAGGAAGACCTCGG-3’; floxed Pet-1: 5’-TAGGAGGGTCTGGTGTCTGG-3’, 5’-GCGTCCTTGTGTGTAGCAGA-3’; Pet-EYFP: 5’-TATATCATGGCCGACAAGCAG-3’, 5’-GAACTCCAGCAGGACCATGT-3’; floxed Lmx1b: 5’- AGGCTCCATCCATTCTTCTC-3’, 5’- AGGCTCCATCCATTCTTCTC-3’; Rosa26^tdTomato: 5’-CCACAATAAGCAAGAGGCAC-3’, 5’- CCGAAAATCTGTGGGAAG TC-3’, 5’- CCGAAAATCTGTGGGAAG TC-3’, 5’- CTGTTCCTGTACGGCATG G-3’; Pet1-Cre: 5’- CGGCATCAACGTTTTCTTTT –3’, 5’- AGTCAGGGCAGAGCCATCTA –3’.

Embryonic age was defined as the number of days following the appearance of a vaginal plug, designated as embryonic day (E) 0.5. For all procedures, animals were euthanized by anesthesia with isoflurane inhalation followed by cervical dislocation, and brain tissues were collected equally from male and female mice or embryos. All animal experiments were approved by the Institutional Animal Care and Use Committee of CWRU.

Neuron dissociation and flow cytometry

Request a detailed protocol

Isolation of Pet1 neurons was performed as previously described with a few modifications (Donovan et al., 2019). Briefly, E11.5 or E14.5 embryos from Pet1-Cre⁺; Rosa26^Ai9 or Pet1-EYFP animals were dissected in cold PBS under dissection microscope to collect all fluorescent cells in the fetal mouse hindbrain. Dissected tissues were transferred into cold 1× TrypLE Express (Gibco) and incubated at 37°C for 15 min with gentle agitation. Following enzymatic trypsin digestion, samples were centrifuged for 1 min at 1500 rpm, TrypLE Express was removed, and samples were washed and resuspended in 500 μL cold Leibovitz’s L-15 media (Thermo Fisher) with 4 μL DNAse I (Invitrogen) and 10% FBS (Gibco). Cells were then slowly triturated with fire-polished glass Pasteur pipettes of decreasing bore size until a single-cell suspension was achieved.

E17.5 embryos were dissociated using a modified adult neuron dissociation method. Pet1-Cre⁺; Rosa26^Ai9 or Pet1-EYFP embryos were dissected in cold aCSF solution (3.5 mM KCl, 126 mM NaCl, 20 mM NaHCO₃, 20 μM dextrose, 1.25 mM NaH₂PO₄, 2 mM CaCl₂, 2 mM MgCl₂, 50 μM AP-V (Tocris), 29 μM DNQX (Sigma), and 100 nM TTX (Abcam)) to collect all fluorescent cells in the hindbrain. Dissected tissues were then incubated in bubbling (95% O₂, 5% CO₂) aCSF containing 1 mg/mL protease from Streptomyces griseus (Sigma) for 15 min at room temperature. Following enzymatic digestion, tissues were transferred to 500 μL cold aCSF/10% FBS solution containing 4 μL DNase I (Invitrogen) and gently triturated with fire-polished glass pipettes to generate single-cell suspension.

Dissociated cells were filtered through a 40 μm filter (BD Biosciences) and sorted on a Becton Dickinson FACS Aria or FACS Aria-SORP digital cell sorter with 85 μm nozzle. Viability of flow sorted cells were routinely checked with Trypan blue staining followed by manual counting of intact cells on small portions of the FACS-purified neurons.

Bulk Pet1 neuron ATAC-seq library preparation

Request a detailed protocol

ATAC-seq was carried out as previously described (Buenrostro et al., 2015) with a few modifications using rostral and caudal Pet1 neurons flow sorted with either Pet1-EYFP (wildtype and Pet1^-/- samples) or Pet1-Cre; Rosa26^Ai9 (Pet1-cKO, Lmx1b-cKO, and DKO samples). Bulk ATAC-seq for each time point/condition was performed with three to five biological replicates. Each biological replicate consisted of embryos flow sorted on the same day from either one or multiple dams that were then pooled to collect a total of 10,000–50,000 Pet1 neurons. Neurons were sorted into L15/10% FBS solution, washed with ice-cold PBS, and resuspended in 25 μL ice-cold ATAC-RSB buffer (10 mM Tris-HCl, pH 7.4, 10 mM NaCl, 3 mM MgCl₂, 0.1% NP40 [Roche cat#11332473001]). Cells were incubated in ATAC-RSB on ice for 2 min, then centrifuged at 800× g for 10 min at 4°C. Pelleted nuclei were resuspended in 25 μL Transposition mix (12.5 μL 2× TD buffer [Illumina], 1.25 μL transposase [100 nM final] [Illumina], 8.25 μL PBS, 3.25 μL H₂O) for E11.5 and E14.5 samples and 15 μL of Transposition mix for E17.5 samples. The transposition reactions were incubated at 37°C for 25 min on a Thermomixer at 1000 rpm. The resulting DNA fragments were purified using the DNA Clean and Concentrator-5 Kit (Zymo) and PCR amplified for 13 cycles with Illumina Nextera adapter primers using the NEBNext High Fidelity 2× Master Mix (NEB) with the following PCR program: (1) 5 min at 72°C, (2) 30 s at 98°C, (3) 10 s at 98°C, (4) 30 s at 63°C, (5) 1 min at 72°C, and (6) repeat steps 3–5, 12×. Final PCR products were cleaned using PCRClean Dx beads (Aline Biosciences), assessed for quality on a Bioanalyzer, and sequenced for 75 cycles to an average depth of 29 M paired-end reads per sample on an Illumina NextSeq 550.

ChIP-seq using Tagmentation (ChIPmentation)

Request a detailed protocol

Pet1 neurons (caudal and rostral) were flow sorted from Pet1-EYFP embryos at E14.5 into PBS/10% FBS. Neurons were frozen in PBS as cell pellets in –80°C until use; 50,000–300,000 frozen cells from multiple days were then thawed on ice, fixed with 1% formaldehyde in PBS for 10 min at room temperature followed by quenching with 0.125 M glycine, and pooled to generate one biological replicate. Two biological replicates were collected for each antibody. ChIP-seq using Tagmentation (ChIPmentation) was then performed as previously described (Schmidl et al., 2015). Cells were washed once with ice-cold PBS and lysed in 130 μL Sonication Buffer (10 mM Tris-HCl pH 8, 0.25% SDS, 2 mM EDTA, 1× protease inhibitor cocktail [Active Motif # 37490], 1 mM PMSF). Sonication was conducted in MicroTUBEs (130 μL) on a Covaris S2 ultrasonicator until DNA fragments were in the range of 200–700 bp. Lysates were diluted 1:1.5 with Equilibrium Buffer (10 mM Tris-HCl pH 8, 233 mM NaCl, 1.66% Triton X-100, 0.166% sodium deoxycholate, 1 mM EDTA, 1× protease inhibitor cocktail [Active Motif]) and incubated with 3 μg of antibody on a rotator overnight at 4°C. The following antibodies were used: H3K4me3 (Diagenode C15410003-50), H3K4me1 (Abcam ab8895), H3K27ac (Abcam ab4729), H3K27me3 (Active Motif 39055). The next day, 10 μL of Protein A/G Dynabeads blocked overnight with 0.1% BSA in RIPA-LS Buffer (10 mM Tris-HCl pH 8, 140 mM NaCl, 1 mM EDTA, 0.1% SDS, 0.1% sodium deoxycholate, 1% Triton X-100) were transferred to each immunoprecipitated lysate and incubated for 2 hr at 4°C on a rotator. Following incubation, beads were washed twice with ice-cold RIPA-LS Buffer, twice with cold RIPA-HS Buffer (10 mM Tris-HCl pH 8, 500 mM NaCl, 1 mM EDTA, 0.1% SDS, 0.1% sodium deoxycholate, 1% Triton X-100), twice with RIPA-LiCl Buffer (10 mM Tris-HCl pH 8, 1 mM EDTA, 250 mM LiCl, 0.5% NP-40, 0.5% sodium deoxycholate), and twice with 10 mM Tris-HCl pH 8. Beads were then resuspended in 25 μL of tagmentation reaction mix containing 12.5 μL of 2× Tagmentation DNA Buffer (Illumina #15027866) and 1 μL of Tagment DNA Enzyme (Illumina #15027865). Tagmentation reaction was incubated at 37°C for 10 min with gentle vortex every 5 min. Tn5 reaction was inactivated by adding 500 μL of ice-cold RIPA-LS Buffer. Beads with washed twice with RIPA-LS Buffer and twice with TE Buffer (10 mM Tris-HCl pH 8, 1 mM EDTA) and resuspended in 48 μL of ChIP Elution Buffer (10 mM Tris-HCl pH 8, 5 mM EDTA, 300 mM NaCl, 0.4% SDS) and 2 μL of Proteinase K solution (Thermo Fisher). Digestion occurred at 55°C for 1 hr followed by sample de-crosslinking at 65°C overnight. Supernatant was transferred the next day to a new tube, and Proteinase K digestion was repeated on remaining bead-bound chromatin by resuspending the beads in 1 μL Proteinase K in 19 μL ChIP elution buffer and eluting for 1 hr at 55°C. Supernatants from both elutions were combined and purified using DNA Clean and Concentrator-5 Kit (Zymo). Library preparation was performed as described for ATAC-seq. ChIPmentation libraries were analyzed on a Bioanalyzer for quality control and then sequenced with 2 × 75 paired-end reads per sample on an Illumina NextSeq 550.

scATAC-seq library construction

Request a detailed protocol

Pet1 neurons from Pet1-EYFP embryos were flow sorted at E14.5 into PBS/10% FBS, washed twice with PBS/0.04% BSA, and resuspended in 100 μL freshly prepared Lysis Buffer (10 mM Tris-HCl pH 7.4, 10 mM NaCl, 3 mM MgCl₂, 0.1% Tween-20, 0.1% IGEPAL). After 3 min of cell lysis on ice, 1 mL of chilled Wash Buffer (10 mM Tris-HCl pH 7.4, 10 mM NaCl, 3 mM MgCl₂, 1% BSA, 0.1% Tween-20) was added to the reaction and nuclei were pelleted and resuspended in 1× Nuclei Buffer (10× Genomics). Nuclei concentration and quality were determined using Trypan blue staining followed by cell counting and inspection on a hemocytometer to ensure that nuclei were free of nuclear membrane blebbing and that there was complete lysis. Nuclei were then added to a transposition mix to perform the Tn5 transposase reaction. The reaction was incubated at 37°C for 60 min. Transposed nuclei were then loaded onto the 10× Genomics Chromium controller to perform nuclei encapsulation with barcoded gel beads, followed by library amplification for 13 cycles. After purification, the library was analyzed on an Agilent Fragment Analyzer to determine fragment size. Library was sequenced on Illumina NextSeq 550 Sequencing System to a depth sufficient to yield >50,000 paired-end reads per nuclei.

scRNA-seq library construction

Request a detailed protocol

Pet1 neurons from Pet1-EYFP embryos were flow sorted at E14.5 into PBS/10% FBS. FACS-purified neuron viability was confirmed to be >80% by staining a portion of the cells with Trypan blue and counting the number of live cells on a hemocytometer. Single cells were loaded in the 10× Single Cell 3’ Chip at a concentration of 1000 cells/μL. Library was prepared according to 10× Genomics Single Cell Protocols Cell Preparation Guide on the 10× Genomics Chromium controller and sequenced to a depth of over 80,000 reads per cell for ~2700 cells on Illumina NextSeq 550 Sequencing System.

CUT&RUN library preparation

Request a detailed protocol

CUT&RUN was performed based on the Epicypher protocol v1.6, which is derived from the method developed by Skene et al., 2018. Briefly, Concanavalin A-coated (ConA) beads were washed twice with CUT&RUN Bead Activation Buffer (20 mM HEPES pH 7.9, 10 mM KCl, 1 mM CaCl₂, 1 mM MnCl₂). FACS-purified Pet1 neurons were washed twice with Wash Buffer (20 mM HEPES pH 7.5, 150 mM NaCl, 0.5 mM spermidine, protease inhibitors [Roche]) and then incubated with activated ConA beads for 15 min at room temperature to allow the cells to absorb to the beads. ConA bead-bound cells were incubated overnight in 50 μL Antibody Buffer (Wash Buffer, 0.01% digitonin, 2 mM EDTA) with 1:100 rabbit anti-FEV (Sigma-Aldrich #HPA067679), guinea pig anti-Lmx1b (Carmen Birchmeier, Max Delbruck Center Berlin), or rabbit monoclonal IgG antibody (Jackson ImmunoResearch Lab 115-007-003). One replicate was collected for each antibody. After overnight incubation, the ConA-bound cells were washed twice with Digitonin Wash Buffer (Wash Buffer, 0.01% digitonin), resuspended in 50 μL Digitonin Buffer containing 700 ng/mL CUTANA pAG-MNase (Epicypher), and incubated for 10 min at room temperature. After binding of pAG-MNase, ConA bead-bound cells were washed twice with Digitonin Buffer, then resuspended in 50 μL Digitonin Buffer containing 2 mM CaCl₂ and incubated at 4°C for 2 hr on a rotator to induce digestion of the chromatin. Reaction was stopped by the addition of 33 μL STOP Buffer to each sample (340 mM NaCl, 20 mM EDTA, 4 mM EGTA, 50 μg/mL RNase A [Thermo Fisher], 50 μg/mL GlycoBlue [Thermo Fisher]) and incubated at 37°C for 15 min to release cleaved chromatin into the supernatant and degrade RNA. DNA was purified from the supernatant using the DNA Clean and Concentrator-5 kit (Zymo). Sequencing library preparation was performed using the NEBNext Ultra II Library Prep Kit for Illumina (NEB) according to the manufacturer’s instructions, with the following modification: following adapter ligation DNA was purified using 1.75× PCRClean Dx beads (Aline Biosciences) and libraries were amplified using the following CUT&RUN-specific PCR cycling conditions to enrich for 100–700 bp fragments: (1) 45 s at 98°C, (2) 15 s at 98°C, (3) 10 s at 60°C, (4) repeat steps 2–3 for a total of 14×, and (5) 1 min at 72°C. PCRClean Dx beads were used to purify the final libraries, which were then sequenced at 75bp x2 on an Illumina NextSeq 550.

RNA-seq library preparation

Request a detailed protocol

Rostral Pet1 neurons were dissected from E17.5 DKO (Pet1-Cre; Pet1^fl/fl; Lmx1b^fl/fl; Rosa26^Ai9) or control Pet1-Cre; Pet1^+/+; Lmx1b^+/+; Rosa26^Ai9 embryos and flow sorted into Trizol LS (Invitrogen) on a Becton Dickonson FACS Aria-SORP sorter. Experiment was performed with three control and two DKO biological replicates. Each biological replicate represented 1000–3000 pool flow sorted neurons from multiple embryos of one or more litters. Total RNA was isolated using chloroform extraction and purified using RNA Clean & Concentrator-5 Kit (Zymo). RNA concentration and quality were measured using Quantifluor RNA system (Promega) and Agilent Fragment Analyzer (Advanced Analytics). DNase I treatment and library construction was performed using the Ovation SoLo RNA-Seq Systems V2 according to manufacturer’s instructions. Final libraries were sequenced on a NextSeq 550 (Illumina) with single-end sequencing for 75 cycles.

Immunohistochemistry, image acquisition, and processing

Request a detailed protocol

Mice were deeply anesthetized by isoflurane inhalation (chamber atmosphere containing 4% isoflurane) and perfused for 20 min with cold 4% paraformaldehyde (PFA) in PBS. Brains were post-fixed in 4% PFA overnight and transferred to 30% sucrose/PBS overnight. Tissues were then frozen in optimal cutting temperature solution and sectioned on a cryostat at 25 μm thickness. Tissue sections were mounted on SuperFrost Plus slides (Thermo Fisher Scientific) and dried. For antigen retrieval, sections were heated up in sodium citrate buffer for 5 min in the microwave at low power. Sections were blocked with 10% NGS in PBS-T for 1 hr, followed by blocking of endogenous biotin with Avidin/Biotin Blocking kit (Vector) only for Syt1 and vGlut3 stained samples. Sections were incubated with primary antibody (1:1000) at room temperature overnight, then treated the next day with secondary antibodies for 1 hr, followed by visualization (Tph2, GFP, and RFP) or further processing with the VECTASTAIN ABC kit (VECTOR) according to manufacturer’s instructions before visualization. Primary antibodies used were: mouse anti-vGlut3 antibody (Abcam ab134310), mouse anti-Synaptotagmin1 (Sigma, SAB1404433), rabbit anti-Tph2 (Millipore Sigma, ABN60), rabbit anti-RFP (Rockland 600-401-379), and chicken anti-GFP (Abcam, ab13970). Secondary antibodies used were: goat anti-mouse, Alexa Fluor 488 or 647 (Invitrogen, A10029 or A28181; 1:500), goat anti-chicken, Alexa Fluor 488 (Invitrogen, A11039, 1:500), goat anti-rabbit, Alexa Fluor 594 (Invitrogen, A11037, 1:500), and biotinylated goat anti-guinea pig IgG antibody (Vector, BA7000; 1:200).

Stained slides were imaged on an LSM800 confocal microscope (Carl Zeiss). In each experiment two animals are analyzed per genotype. Images were acquired per animal using 20× and 63× objective lens and processed in ImageJ. Signal brightness and contrast were edited across whole images equally among genotypes for all pericellular basket synapse and axon images. Signal intensity calculations were normalized for each section. Two-way ANOVA with Welch’s correction analysis was performed.

ATAC-seq data processing and analysis

Request a detailed protocol

ATAC-seq data were processed using the standardized software pipeline from the ENCODE consortium to perform quality-control, read alignment, and peak calling (Landt et al., 2012). FASTQ files from ATAC-seq reads were mapped to UCSC mm10 with Bowtie v2.3.4.3. Samtools v1.9 was used to remove PCR duplicates and chrM reads. Peak calling was performed using MACS2 with the following parameters: --nomodel --shift 37 --ext 73 --pval 1e-2 -B --SPMR --callsummits. Then by intersecting called peaks in replicates using BEDTools v2.29.0, we defined replicated peaks as the set of peaks present independently in each replicated peak call set and also called in the pooled set.

Differential accessibility analysis

Request a detailed protocol

A comprehensive set of ATAC-seq peaks (TACs) from control datasets (E11.5, E14.5, and E17.5) was generated to be used as the reference for assessing differential accessibility between control time points and for control to TF knockout comparisons. All control peak sets were merged using 1 bp overlap to create the union of all control peaks (E11.5, E14.5, and E17.5). Any potential batch effects were corrected using surrogate variable analysis (Leek, 2014). The number of significant surrogate variables was determined using an iteratively re-weighted least squares approach. One significant surrogate variable was found and added to the model design in DESeq2 (Love et al., 2014). Differentially accessible peaks were determined using DESeq2 v1.32.0 with a fold change cutoff of 2-fold and FDR adjusted p-value ≤ 0.01 for all wildtype and conditional knockout comparisons and ≤0.05 for Pet1^-/- comparisons. BEDTools v2.30.0 was used to find shared intersections between identified peaks (1 bp minimum overlap). Peak annotations were performed using ChIPseeker v1.28.3. PCA was performed using the prcomp function within the ‘plotPCA’ wrapper in DESeq2 on variance stabilizing transformed count data to correct for the underlying mean-variance relationship and plotted with ggplot2 (Love et al., 2014).

Volcano plots were generated using the EnhancedVolcano package in R (v1.10.0, https://github.com/kevinblighe/EnhancedVolcano; Blighe, 2021).

Motif enrichment

Request a detailed protocol

TF motifs were identified using monaLisa v0.1.50 (https://fmicompbio.github.io/monaLisa/index; Turaga, 2022). Differentially accessible peak regions were binned into ‘loss’, ‘no change’, and ‘gain’ categories and motif enrichment for each bin was compared against an equal number of sequences randomly sampled from the genome matched by GC composition. Statistically significant enriched motifs were determined using a one-tailed Fisher’s exact test with Benjamini-Hochberg adjusted p-values ≤ 1e-8.

GO enrichment

Request a detailed protocol

GO was performed using Webgestalt requiring >5 and <2000 genes per category and FDR ≤ 0.05 (Liao et al., 2019).

ChIPmentation data processing and analysis

Request a detailed protocol

ChIPmentation data were processed using the standard ENCODE ChIP-seq software pipeline from the ENCODE Consortium. Sequencing reads were mapped to the UCSC mm10 by Bowtie v2.3.4.3. PCR duplicates were removed using samtools v1.9. The unique mapped reads were used to call peaks comparing immunoprecipitated chromatin with input chromatin with MACS2 v2.2.4 with FDR adjusted p-value cutoff of 0.01. Peak lists were filtered to remove peaks overlapping ENCODE blacklisted regions. BEDTools was used to find shared intersections between identified peaks (1 bp minimum overlap). Peak annotations were performed using ChIPseeker v1.28.3.

Chromatin state analysis

Request a detailed protocol

Chromatin state analysis was performed using ChromHMM v1.22 (Ernst and Kellis, 2017). Four histone mark ChIPmentation datasets were binarized at a resolution of 200 bp with a signal threshold of 4. We trained a 10-state model as it characterized the key interactions between the histone marks. The posterior probability state of each genomic bin was calculated using the learned model.

Super-enhancer analysis

Request a detailed protocol

The rank ordering of super-enhancers (ROSE) package was used to identify super-enhancers based on the ranking of H3K27ac signal intensities from E14.5 Pet1 neurons (Lovén et al., 2013; Whyte et al., 2013). Clusters of E14.5 ATAC-seq peaks within 12.5 kb were grouped and ranked for their total input background subtracted H3K27ac signal with super-enhancers called above the inflection point of 6391.9.

scATAC-seq analysis

Request a detailed protocol

Raw scATAC-seq data was processed using the CellRanger ATAC v1.2.0 pipeline (10× Genomics). Reads were aligned to the GRCm38 (mm10) mouse genome. Peaks were called using MACS2 v2.2.4. The peak/cell matrix and all unique single-cell fragments were imported into R using the Signac package (Stuart et al., 2021). Several QC metrics were calculated to assess the quality of the data. Cells with >3000 to <100,000 peak region fragments, >40% reads in peaks, <0.025 blacklist ratio, <4 nucleosome signal, >2 TSS enrichment were retained for further analysis. The data was normalized and dimensionality reduced using latent semantic indexing (Cusanovich et al., 2015). First, term frequency-inverse document frequency (TF-IDF) normalizes across cells for sequencing depth and across peaks to give higher values to rare peaks. Next, we ran singular value decomposition on the TF-IDF matrix using all variable features. A gene activity matrix was constructed by counting the number of fragments that map within the 2 kb upstream region of each gene. Motif enrichment and TF activity for scATAC-seq clusters were quantified using the ‘RunChromVAR’ wrapper function within the Signac package (v1.14.0) (Schep et al., 2017; Stuart et al., 2021). The filtered cell (n = 1692) by peak (n = 124,111) matrix was used as input with the curated database of mm10 cisBP motifs (n = 884). Default settings were used to compute motif accessibility deviation Z-scores.

scRNA-seq analysis

Request a detailed protocol

Raw scRNA-seq data was processed using the CellRanger v3.1.0 pipeline (10× Genomics). Reads were aligned to the GRCm38 (mm10) mouse genome. The unique molecular index count matrix was imported into R using Seurat (Butler et al., 2018; Hao et al., 2021; Satija et al., 2015; Stuart et al., 2019). Cells with <200 or >7500 expressed genes or >10% mitochondrial genes were excluded from further analysis. Gene counts were log normalized, then the dataset was scaled and centered and mitochondrial gene levels were regressed out. Dimensionality of the data was determined using an elbow plot and the first 20 principal components were used to reduce dimensionality using PCA.

Integration of scATAC-seq and scRNA-seq data

Request a detailed protocol

scATAC-seq and scRNA-seq datasets were integrated using SCTransform from Seurat using 3000 variable features. To perform clustering, we first computed a PCA on the normalized count matrix and used the first 20 PCs to generate a shared nearest neighbor graph with the k parameter set to 20. Then, we clustered the data with a resolution of 0.8 using the Louvain algorithm. Next, we performed UMAP to project the clustering onto low dimensional space and plot the first two dimensions for visualizations (30 nearest neighbors, minimum distance = 0.3) (McInnes et al., 2018). Cluster markers were determined using the Wilcoxon rank sum test with a minimum fraction of cells expressing the gene at 0.25 and a log fold change threshold of 0.25 for each cluster vs. all other clusters. The top two marker genes for each cluster are used to identify each cluster.

Adult scRNA-seq projection onto embryonic scRNA-seq reference

Request a detailed protocol

Data from Okaty et al., 2020, were downloaded from the NCBI GEO database (GSE144980). Raw sequencing reads were processed as above for the E14.5 scRNA-seq data using CellRanger. Cells were filtered using criteria as described in the original publication resulting in 2350 cells. The ‘MapQuery’ function in Seurat was used to project the adult scRNA-seq cells onto the E14.5 Pet1 neuron scRNA-seq UMAP reference.

Enhancer-gene (TAC-gene) association

Request a detailed protocol

scATAC-seq and scRNA-seq seurat objects were imported into SnapATAC (Fang et al., 2021). A gene expression profile for each cell based on the weighted average of the nearest neighboring cells (k = 15) was imputed based on the scRNA-seq dataset. From the integrated scATAC-seq and scRNA-seq datasets, a ‘pseudo’ multiomics cell is made that contains both gene expression and chromatin accessibility information. Then, logistic regression was performed to quantify the association between gene expression and the binarized accessibility state at putative enhancer elements in a 500 kb window centered at the TSS of each gene. The gene to putative enhancer associations were filtered with p-value ≤ 0.05.

RNA-seq analysis

Request a detailed protocol

RNA-seq reads were aligned to UCSC mm10 using Hisat2 (v2.1.0) with a penalty of 12 for a non-canonical splice site, maximum and minimum penalty for mismatch of 1,0, and maximum and minimum penalty for soft-clipping of 3,0 parameters (Kim et al., 2019). Gene expression quantification and differential expression were calculated using Cufflinks v2.2.2 with a fold change cutoff of 1.5-fold and FDR cutoff of 0.05 (Roberts et al., 2011; Trapnell et al., 2010; Trapnell et al., 2012).

CUT&RUN data processing and analysis

Request a detailed protocol

CUT&RUN data were processed using the standard ENCODE ChIP-seq software pipeline from the ENCODE Consortium. Sequencing reads were mapped to the UCSC mm10 by Bowtie v2.3.4.3. PCR duplicates were removed using samtools v1.9. The unique mapped reads were used to call peaks comparing TF-bound chromatin with an IgG control with SEACR using both stringent and relaxed threshold settings (Meers et al., 2019). The stringent threshold uses the peak of the total signal curve and the relaxed threshold uses the total signal of the ‘knee’ to the peak of the total signal curve. Both threshold settings have been shown to have high sensitivity and specificity (Meers et al., 2019). BEDTools was used to find shared intersections between identified peaks (1 bp minimum overlap). Peak annotations were performed using ChIPseeker v1.28.3. CUT&RUN data for Pet1 and Lmx1b were normalized together for visualization using the ChIPseqSpikeInFree method to determine the scaling factor for adjusting signal coverage across the genome (Jin et al., 2020).

Footprinting analysis

Request a detailed protocol

TF footprints were identified using RGT-HINT (Gusmao et al., 2016; Li et al., 2019). First, we searched for motifs from HOCOMOCO v11 within the set of differentially accessible ATAC-seq peaks for each TF knockout condition (Kulakovskiy et al., 2018). Then, differential footprints were called on uniquely mapped paired-end ATAC-seq reads with bias correction for the Tn5 transposase within the set of differentially accessible ATAC-seq peaks comparing control to each TF knockout condition.

Genomic visualization

Request a detailed protocol

ATAC-seq, ChIPmentation, and CUT&RUN data were visualized in Integrative Genomic Viewer (v2.9.3). Bam files for all ATAC-seq and ChIPmentation datasets were normalized to 1× genome coverage using deeptools (v3.5.0) bamCoverage.

Statistical analysis

Request a detailed protocol

No statistical methods were used to predetermine sample size. Measurements for all experiments were taken from distinct samples. No methods were used to determine whether the data conformed to the assumptions of the statistical method. Experimenters were blinded to genotype during immunofluorescence analysis of 5-HT pericellular baskets. No other blinding of experiments was performed. Clustering of scATAC-seq and scRNA-seq was performed in an unbiased manner. Cell subtypes were assigned after unbiased clustering. Nuclei that did not meet quality thresholds were removed from further analyses as described in the Results and Materials and methods.

Data visualization

Request a detailed protocol

Data plots were generated using ggplot2 (v3.3.5) in R (v4.0.4) (Wickham et al., 2016) and Microsoft Excel. 5-HT neurotransmission diagram was created using Biorender.com.

Code availability

Request a detailed protocol

No custom code was used in this study. Open-source algorithms were used as described in the analysis methods. Details on how these algorithms were used to analyze data are available from the corresponding author upon request.

Data availability

Sequencing data have been deposited in GEO under accession code GSE185737.

The following data sets were generated

(2022) NCBI Gene Expression Omnibus
ID GSE185737. Reorganization of postmitotic neuronal chromatin accessibility for maturation of serotonergic identity.

http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE185737

The following previously published data sets were used

1. Donovan LJ
2. Spencer WC
3. Kitt MM
4. Eastman BA
5. Lobur KJ
6. Jiao K
7. Silver J
8. Deneris ES
(2019) NCBI Gene Expression Omnibus
ID GSE130514. Lmx1b is required at multiple stages to build expansive serotonergic axon architectures.

https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE130514
1. Wyler SC
2. Spencer WC
3. Green NH
4. Rood BD
5. Crawford L
6. Craige C
7. Gresch P
8. McMahon DG
9. Beck SG
10. Deneris ES
(2016) NCBI Gene Expression Omnibus
ID GSE74315. Pet-1 Switches Transcriptional Targets Postnatally to Regulate Maturation of Serotonin Neuron Excitability.

https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE74315

References

1. Allaway KC
2. Gabitto MI
3. Wapinski O
4. Saldi G
5. Wang CY
6. Bandler RC
7. Wu SJ
8. Bonneau R
9. Fishell G
(2021) Genetic and epigenetic coordination of cortical interneuron development
Nature 597:693–697.
https://doi.org/10.1038/s41586-021-03933-1
- PubMed
- Google Scholar
(2013) Development of the serotonergic cells in murine raphe nuclei and their relations with rhombomeric domains
Brain Structure & Function 218:1229–1277.
https://doi.org/10.1007/s00429-012-0456-8
- PubMed
- Google Scholar
(2019) Proneural factors Ascl1 and Neurog2 contribute to neuronal subtype identities by establishing distinct chromatin landscapes
Nature Neuroscience 22:897–908.
https://doi.org/10.1038/s41593-019-0399-y
- PubMed
- Google Scholar
Software
1. Blighe K
(2021) EnhancedVolcano, version swh:1:rev:7abca284367c4146524397e0bb056fc231a095df
GitHub.

https://archive.softwareheritage.org/swh:1:dir:7715031df604f96a2f80c5a2672eceed48f026b9;origin=https://github.com/kevinblighe/EnhancedVolcano;visit=swh:1:snp:1f4a816b4bc938ff3ab838ee7ae3722e98275d9d;anchor=swh:1:rev:7abca284367c4146524397e0bb056fc231a095df
(2020) Neuronal Maturation: Challenges and Opportunities in a Nascent Field
Trends in Neurosciences 43:360–362.
https://doi.org/10.1016/j.tins.2020.02.005
- PubMed
- Google Scholar
(2015) ATAC-seq: A Method for Assaying Chromatin Accessibility Genome-Wide
Current Protocols in Molecular Biology 109:21.
https://doi.org/10.1002/0471142727.mb2129s109
- PubMed
- Google Scholar
1. Butler A
2. Hoffman P
3. Smibert P
4. Papalexi E
5. Satija R
(2018) Integrating single-cell transcriptomic data across different conditions, technologies, and species
Nature Biotechnology 36:411–420.
https://doi.org/10.1038/nbt.4096
- PubMed
- Google Scholar
(2013) Serotonin modulation of cortical neurons and networks
Frontiers in Integrative Neuroscience 7:25.
https://doi.org/10.3389/fnint.2013.00025
- PubMed
- Google Scholar
(2015) Multiplex single cell profiling of chromatin accessibility by combinatorial cellular indexing
Science (New York, N.Y.) 348:910–914.
https://doi.org/10.1126/science.aab1601
- PubMed
- Google Scholar
1. de la Torre-Ubieta L
2. Stein JL
3. Won H
4. Opland CK
5. Liang D
6. Lu D
7. Geschwind DH
(2018) The Dynamic Landscape of Open Chromatin during Human Cortical Neurogenesis
Cell 172:289–304.
https://doi.org/10.1016/j.cell.2017.12.014
- PubMed
- Google Scholar
1. De Rubeis S
2. He X
3. Goldberg AP
4. Poultney CS
5. Samocha K
6. Cicek AE
7. Kou Y
8. Liu L
9. Fromer M
10. Walker S
11. Singh T
12. Klei L
13. Kosmicki J
14. Shih-Chen F
15. Aleksic B
16. Biscaldi M
17. Bolton PF
18. Brownfeld JM
19. Cai J
20. Campbell NG
21. Carracedo A
22. Chahrour MH
23. Chiocchetti AG
24. Coon H
25. Crawford EL
26. Curran SR
27. Dawson G
28. Duketis E
29. Fernandez BA
30. Gallagher L
31. Geller E
32. Guter SJ
33. Hill RS
34. Ionita-Laza J
35. Jimenz Gonzalez P
36. Kilpinen H
37. Klauck SM
38. Kolevzon A
39. Lee I
40. Lei I
41. Lei J
42. Lehtimäki T
43. Lin C-F
44. Ma’ayan A
45. Marshall CR
46. McInnes AL
47. Neale B
48. Owen MJ
49. Ozaki N
50. Parellada M
51. Parr JR
52. Purcell S
53. Puura K
54. Rajagopalan D
55. Rehnström K
56. Reichenberg A
57. Sabo A
58. Sachse M
59. Sanders SJ
60. Schafer C
61. Schulte-Rüther M
62. Skuse D
63. Stevens C
64. Szatmari P
65. Tammimies K
66. Valladares O
67. Voran A
68. Li-San W
69. Weiss LA
70. Willsey AJ
71. Yu TW
72. Yuen RKC
73. DDD Study
74. Homozygosity Mapping Collaborative for Autism
75. UK10K Consortium
76. Cook EH
77. Freitag CM
78. Gill M
79. Hultman CM
80. Lehner T
81. Palotie A
82. Schellenberg GD
83. Sklar P
84. State MW
85. Sutcliffe JS
86. Walsh CA
87. Scherer SW
88. Zwick ME
89. Barett JC
90. Cutler DJ
91. Roeder K
92. Devlin B
93. Daly MJ
94. Buxbaum JD
(2014) Synaptic, transcriptional and chromatin genes disrupted in autism
Nature 515:209–215.
https://doi.org/10.1038/nature13772
- PubMed
- Google Scholar
1. Deneris ES
2. Wyler SC
(2012) Serotonergic transcriptional networks and potential importance to mental health
Nature Neuroscience 15:519–527.
https://doi.org/10.1038/nn.3039
- PubMed
- Google Scholar
1. Deneris E
2. Gaspar P
(2018) Serotonin neuron development: shaping molecular and structural identities
Wiley Interdisciplinary Reviews. Developmental Biology 7:e301.
https://doi.org/10.1002/wdev.301
- PubMed
- Google Scholar
1. Ding YQ
2. Marklund U
3. Yuan W
4. Yin J
5. Wegman L
6. Ericson J
7. Deneris E
8. Johnson RL
9. Chen ZF
(2003) Lmx1b is essential for the development of serotonergic neurons
Nature Neuroscience 6:933–938.
https://doi.org/10.1038/nn1104
- PubMed
- Google Scholar
1. Doan RN
2. Lim ET
3. De Rubeis S
4. Betancur C
5. Cutler DJ
6. Chiocchetti AG
7. Overman LM
8. Soucy A
9. Goetze S
10. Autism Sequencing Consortium
11. Freitag CM
12. Daly MJ
13. Walsh CA
14. Buxbaum JD
15. Yu TW
(2019) Recessive gene disruptions in autism spectrum disorder
Nature Genetics 51:1092–1098.
https://doi.org/10.1038/s41588-019-0433-8
- PubMed
- Google Scholar
1. Donovan LJ
2. Spencer WC
3. Kitt MM
4. Eastman BA
5. Lobur KJ
6. Jiao K
7. Silver J
8. Deneris ES
(2019) Lmx1b is required at multiple stages to build expansive serotonergic axon architectures
eLife 8:e48788.
https://doi.org/10.7554/eLife.48788
- PubMed
- Google Scholar
1. D’Urso A
2. Brickner JH
(2014) Mechanisms of epigenetic memory
Trends in Genetics 30:230–236.
https://doi.org/10.1016/j.tig.2014.04.004
- PubMed
- Google Scholar
1. Ernst J
2. Kellis M
(2017) Chromatin-state discovery and genome annotation with ChromHMM
Nature Protocols 12:2478–2492.
https://doi.org/10.1038/nprot.2017.124
- PubMed
- Google Scholar
1. Fang R
2. Preissl S
3. Li Y
4. Hou X
5. Lucero J
6. Wang X
7. Motamedi A
8. Shiau AK
9. Zhou X
10. Xie F
11. Mukamel EA
12. Zhang K
13. Zhang Y
14. Behrens MM
15. Ecker JR
16. Ren B
(2021) Comprehensive analysis of single cell ATAC-seq data with SnapATAC
Nature Communications 12:1337.
https://doi.org/10.1038/s41467-021-21583-9
- PubMed
- Google Scholar
1. Frank CL
2. Liu F
3. Wijayatunge R
4. Song L
5. Biegler MT
6. Yang MG
7. Vockley CM
8. Safi A
9. Gersbach CA
10. Crawford GE
11. West AE
(2015) Regulation of chromatin accessibility and Zic binding at enhancers in the developing cerebellum
Nature Neuroscience 18:647–656.
https://doi.org/10.1038/nn.3995
- PubMed
- Google Scholar
1. Gallegos DA
2. Chan U
3. Chen LF
4. West AE
(2018) Chromatin Regulation of Neuronal Maturation and Plasticity
Trends in Neurosciences 41:311–324.
https://doi.org/10.1016/j.tins.2018.02.009
- PubMed
- Google Scholar
1. Gorkin DU
2. Barozzi I
3. Zhao Y
4. Zhang Y
5. Huang H
6. Lee AY
7. Li B
8. Chiou J
9. Wildberg A
10. Ding B
11. Zhang B
12. Wang M
13. Strattan JS
14. Davidson JM
15. Qiu Y
16. Afzal V
17. Akiyama JA
18. Plajzer-Frick I
19. Novak CS
20. Kato M
21. Garvin TH
22. Pham QT
23. Harrington AN
24. Mannion BJ
25. Lee EA
26. Fukuda-Yuzawa Y
27. He Y
28. Preissl S
29. Chee S
30. Han JY
31. Williams BA
32. Trout D
33. Amrhein H
34. Yang H
35. Cherry JM
36. Wang W
37. Gaulton K
38. Ecker JR
39. Shen Y
40. Dickel DE
41. Visel A
42. Pennacchio LA
43. Ren B
(2020) An atlas of dynamic chromatin landscapes in mouse fetal development
Nature 583:744–751.
https://doi.org/10.1038/s41586-020-2093-3
- PubMed
- Google Scholar
1. Gray LT
2. Yao Z
3. Nguyen TN
4. Kim TK
5. Zeng H
6. Tasic B
(2017) Layer-specific chromatin accessibility landscapes reveal regulatory networks in adult mouse visual cortex
eLife 6:e21883.
https://doi.org/10.7554/eLife.21883
- PubMed
- Google Scholar
1. Gusmao EG
2. Allhoff M
3. Zenke M
4. Costa IG
(2016) Analysis of computational footprinting methods for DNase sequencing experiments
Nature Methods 13:303–309.
https://doi.org/10.1038/nmeth.3772
- PubMed
- Google Scholar
1. Hao Y
2. Hao S
3. Andersen-Nissen E
4. Mauck WM
5. Zheng S
6. Butler A
7. Lee MJ
8. Wilk AJ
9. Darby C
10. Zager M
11. Hoffman P
12. Stoeckius M
13. Papalexi E
14. Mimitou EP
15. Jain J
16. Srivastava A
17. Stuart T
18. Fleming LM
19. Yeung B
20. Rogers AJ
21. McElrath JM
22. Blish CA
23. Gottardo R
24. Smibert P
25. Satija R
(2021) Integrated analysis of multimodal single-cell data
Cell 184:3573–3587.
https://doi.org/10.1016/j.cell.2021.04.048
- PubMed
- Google Scholar
1. Heavner WE
2. Smith SEP
(2020) Resolving the Synaptic versus Developmental Dichotomy of Autism Risk Genes
Trends in Neurosciences 43:227–241.
https://doi.org/10.1016/j.tins.2020.01.009
- PubMed
- Google Scholar
(1999) The ETS domain factor Pet-1 is an early and precise marker of central serotonin neurons and interacts with a conserved element in serotonergic genes
The Journal of Neuroscience 19:10348–10356.
https://doi.org/10.1523/JNEUROSCI.19-23-10348.1999
- PubMed
- Google Scholar
1. Hobert O
(2008) Regulatory logic of neuronal diversity: terminal selector genes and selector motifs
PNAS 105:20067–20071.
https://doi.org/10.1073/pnas.0806070105
- PubMed
- Google Scholar
1. Hobert O
(2011) Regulation of terminal differentiation programs in the nervous system
Annual Review of Cell and Developmental Biology 27:681–696.
https://doi.org/10.1146/annurev-cellbio-092910-154226
- PubMed
- Google Scholar
1. Holmberg J
2. Perlmann T
(2012) Maintaining differentiated cellular identity
Nature Reviews. Genetics 13:429–439.
https://doi.org/10.1038/nrg3209
- PubMed
- Google Scholar
1. Jin H
2. Kasper LH
3. Larson JD
4. Wu G
5. Baker SJ
6. Zhang J
7. Fan Y
(2020) ChIPseqSpikeInFree: a ChIP-seq normalization approach to reveal global changes in histone modifications without spike-in
Bioinformatics (Oxford, England) 36:1270–1272.
https://doi.org/10.1093/bioinformatics/btz720
- PubMed
- Google Scholar
1. Jolma A
2. Yan J
3. Whitington T
4. Toivonen J
5. Nitta KR
6. Rastas P
7. Morgunova E
8. Enge M
9. Taipale M
10. Wei G
11. Palin K
12. Vaquerizas JM
13. Vincentelli R
14. Luscombe NM
15. Hughes TR
16. Lemaire P
17. Ukkonen E
18. Kivioja T
19. Taipale J
(2013) DNA-binding specificities of human transcription factors
Cell 152:327–339.
https://doi.org/10.1016/j.cell.2012.12.009
- PubMed
- Google Scholar
1. Kim D
2. Kim D
3. Lee G
4. Cho S-I
5. Kim J-S
(2019) Genome-wide target specificity of CRISPR RNA-guided adenine base editors
Nature Biotechnology 37:430–435.
https://doi.org/10.1038/s41587-019-0050-1
- PubMed
- Google Scholar
(2015) Transcriptional coordination of synaptogenesis and neurotransmitter signaling
Current Biology 25:1282–1295.
https://doi.org/10.1016/j.cub.2015.03.028
- PubMed
- Google Scholar
1. Krueger KC
2. Deneris ES
(2008) Serotonergic Transcription of Human FEV Reveals Direct GATA Factor Interactions and Fate of Pet-1-Deficient Serotonin Neuron Precursors
Journal of Neuroscience 28:12748–12758.
https://doi.org/10.1523/JNEUROSCI.4349-08.2008
- PubMed
- Google Scholar
(2018) HOCOMOCO: towards a complete collection of transcription factor binding models for human and mouse via large-scale ChIP-Seq analysis
Nucleic Acids Research 46:D252–D259.
https://doi.org/10.1093/nar/gkx1106
- PubMed
- Google Scholar
1. Landt SG
2. Marinov GK
3. Kundaje A
4. Kheradpour P
5. Pauli F
6. Batzoglou S
7. Bernstein BE
8. Bickel P
9. Brown JB
10. Cayting P
11. Chen Y
12. DeSalvo G
13. Epstein C
14. Fisher-Aylor KI
15. Euskirchen G
16. Gerstein M
17. Gertz J
18. Hartemink AJ
19. Hoffman MM
20. Iyer VR
21. Jung YL
22. Karmakar S
23. Kellis M
24. Kharchenko PV
25. Li Q
26. Liu T
27. Liu XS
28. Ma L
29. Milosavljevic A
30. Myers RM
31. Park PJ
32. Pazin MJ
33. Perry MD
34. Raha D
35. Reddy TE
36. Rozowsky J
37. Shoresh N
38. Sidow A
39. Slattery M
40. Stamatoyannopoulos JA
41. Tolstorukov MY
42. White KP
43. Xi S
44. Farnham PJ
45. Lieb JD
46. Wold BJ
47. Snyder M
(2012) ChIP-seq guidelines and practices of the ENCODE and modENCODE consortia
Genome Research 22:1813–1831.
https://doi.org/10.1101/gr.136184.111
- PubMed
- Google Scholar
1. Leek JT
(2014) svaseq: removing batch effects and other unwanted noise from sequencing data
Nucleic Acids Research 42:e161.
https://doi.org/10.1093/nar/gku864
- PubMed
- Google Scholar
1. Leyva-Díaz E
2. Hobert O
(2019) Transcription factor autoregulation is required for acquisition and maintenance of neuronal identity
Development (Cambridge, England) 146:dev177378.
https://doi.org/10.1242/dev.177378
- PubMed
- Google Scholar
1. Li Z
2. Wang M
3. Lin K
4. Xie Y
5. Guo J
6. Ye L
7. Zhuang Y
8. Teng W
9. Ran X
10. Tong Y
11. Xue Y
12. Zhang W
13. Zhang Y
(2019) The bread wheat epigenomic map reveals distinct chromatin architectural and evolutionary features of functional genetic elements
Genome Biology 20:139.
https://doi.org/10.1186/s13059-019-1746-8
- PubMed
- Google Scholar
1. Liao Y
2. Wang J
3. Jaehnig EJ
4. Shi Z
5. Zhang B
(2019) WebGestalt 2019: gene set analysis toolkit with revamped UIs and APIs
Nucleic Acids Research 47:W199–W205.
https://doi.org/10.1093/nar/gkz401
- PubMed
- Google Scholar
1. Liu C
2. Maejima T
3. Wyler SC
4. Casadesus G
5. Herlitze S
6. Deneris ES
(2010) Pet-1 is required across different stages of life to regulate serotonergic function
Nature Neuroscience 13:1190–1198.
https://doi.org/10.1038/nn.2623
- PubMed
- Google Scholar
1. Love MI
2. Huber W
3. Anders S
(2014) Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2
Genome Biology 15:550.
https://doi.org/10.1186/s13059-014-0550-8
- PubMed
- Google Scholar
1. Lovén J
2. Hoke HA
3. Lin CY
4. Lau A
5. Orlando DA
6. Vakoc CR
7. Bradner JE
8. Lee TI
9. Young RA
(2013) Selective inhibition of tumor oncogenes by disruption of super-enhancers
Cell 153:320–334.
https://doi.org/10.1016/j.cell.2013.03.036
- PubMed
- Google Scholar
1. Ma S
2. Zhang B
3. LaFave LM
4. Earl AS
5. Chiang Z
6. Hu Y
7. Ding J
8. Brack A
9. Kartha VK
10. Tay T
11. Law T
12. Lareau C
13. Hsu Y-C
14. Regev A
15. Buenrostro JD
(2020) Chromatin Potential Identified by Shared Single-Cell Profiling of RNA and Chromatin
Cell 183:1103–1116.
https://doi.org/10.1016/j.cell.2020.09.056
- PubMed
- Google Scholar
1. Mazzoni EO
2. Mahony S
3. Closser M
4. Morrison CA
5. Nedelec S
6. Williams DJ
7. An D
8. Gifford DK
9. Wichterle H
(2013) Synergistic binding of transcription factors to cell-specific enhancers programs motor neuron identity
Nature Neuroscience 16:1219–1227.
https://doi.org/10.1038/nn.3467
- PubMed
- Google Scholar
(2018) UMAP: Uniform Manifold Approximation and Projection
Journal of Open Source Software 3:861.
https://doi.org/10.21105/joss.00861
- Google Scholar
(2019) Peak calling by Sparse Enrichment Analysis for CUT&RUN chromatin profiling
Epigenetics & Chromatin 12:42.
https://doi.org/10.1186/s13072-019-0287-4
- PubMed
- Google Scholar
1. Okaty BW
2. Freret ME
3. Rood BD
4. Brust RD
5. Hennessy ML
6. deBairos D
7. Kim JC
8. Cook MN
9. Dymecki SM
(2015) Multi-Scale Molecular Deconstruction of the Serotonin Neuron System
Neuron 88:774–791.
https://doi.org/10.1016/j.neuron.2015.10.007
- PubMed
- Google Scholar
(2019) Embracing diversity in the 5-HT neuronal system
Nature Reviews. Neuroscience 20:397–424.
https://doi.org/10.1038/s41583-019-0151-3
- PubMed
- Google Scholar
(2020) A single-cell transcriptomic and anatomic atlas of mouse dorsal raphe Pet1 neurons
eLife 9:e55523.
https://doi.org/10.7554/eLife.55523
- PubMed
- Google Scholar
1. Perino M
2. Veenstra GJC
(2016) Chromatin Control of Developmental Dynamics and Plasticity
Developmental Cell 38:610–620.
https://doi.org/10.1016/j.devcel.2016.08.004
- PubMed
- Google Scholar
(2007) Transcriptional regulation of vertebrate axon guidance and synapse formation
Nature Reviews. Neuroscience 8:331–340.
https://doi.org/10.1038/nrn2118
- PubMed
- Google Scholar
1. Preissl S
2. Fang R
3. Huang H
4. Zhao Y
5. Raviram R
6. Gorkin DU
7. Zhang Y
8. Sos BC
9. Afzal V
10. Dickel DE
11. Kuan S
12. Visel A
13. Pennacchio LA
14. Zhang K
15. Ren B
(2018) Single-nucleus analysis of accessible chromatin in developing mouse forebrain reveals cell-type-specific transcriptional regulation
Nature Neuroscience 21:432–439.
https://doi.org/10.1038/s41593-018-0079-3
- PubMed
- Google Scholar
1. Ren J
2. Isakova A
3. Friedmann D
4. Zeng J
5. Grutzner SM
6. Pun A
7. Zhao GQ
8. Kolluru SS
9. Wang R
10. Lin R
11. Li P
12. Li A
13. Raymond JL
14. Luo Q
15. Luo M
16. Quake SR
17. Luo L
(2019) Single-cell transcriptomes and whole-brain projections of serotonin neurons in the mouse dorsal and median raphe nuclei
eLife 8:e49424.
https://doi.org/10.7554/eLife.49424
- PubMed
- Google Scholar
1. Roberts HC
2. Denison HJ
3. Martin HJ
4. Patel HP
5. Syddall H
6. Cooper C
7. Sayer AA
(2011) A review of the measurement of grip strength in clinical and epidemiological studies: towards A standardised approach
Age and Ageing 40:423–429.
https://doi.org/10.1093/ageing/afr051
- PubMed
- Google Scholar
1. Satija R
2. Farrell JA
3. Gennert D
4. Schier AF
5. Regev A
(2015) Spatial reconstruction of single-cell gene expression data
Nature Biotechnology 33:495–502.
https://doi.org/10.1038/nbt.3192
- PubMed
- Google Scholar
(2017) chromVAR: inferring transcription-factor-associated accessibility from single-cell epigenomic data
Nature Methods 14:975–978.
https://doi.org/10.1038/nmeth.4401
- PubMed
- Google Scholar
(2015) ChIPmentation: fast, robust, low-input ChIP-seq for histones and transcription factors
Nature Methods 12:963–965.
https://doi.org/10.1038/nmeth.3542
- PubMed
- Google Scholar
(2005a) A differentially autoregulated Pet-1 enhancer region is A critical target of the transcriptional cascade that governs serotonin neuron development
The Journal of Neuroscience 25:2628–2636.
https://doi.org/10.1523/JNEUROSCI.4979-04.2005
- PubMed
- Google Scholar
1. Scott MM
2. Wylie CJ
3. Lerch JK
4. Murphy R
5. Lobur K
6. Herlitze S
7. Jiang W
8. Conlon RA
9. Strowbridge BW
10. Deneris ES
(2005b) A genetic approach to access serotonin neurons for in vivo and in vitro studies
PNAS 102:16472–16477.
https://doi.org/10.1073/pnas.0504510102
- PubMed
- Google Scholar
1. Senft RA
2. Dymecki SM
(2021) Neuronal pericellular baskets: neurotransmitter convergence and regulation of network excitability
Trends in Neurosciences 44:915–924.
https://doi.org/10.1016/j.tins.2021.08.006
- PubMed
- Google Scholar
(2018) Targeted in situ genome-wide profiling with high efficiency for low cell numbers
Nature Protocols 13:1006–1019.
https://doi.org/10.1038/nprot.2018.015
- PubMed
- Google Scholar
1. Spencer WC
2. Deneris ES
(2017) Regulatory Mechanisms Controlling Maturation of Serotonin Neuron Identity and Function
Frontiers in Cellular Neuroscience 11:215.
https://doi.org/10.3389/fncel.2017.00215
- PubMed
- Google Scholar
1. Spitz F
2. Furlong EEM
(2012) Transcription factors: from enhancer binding to developmental control
Nature Reviews Genetics 13:613–626.
https://doi.org/10.1038/nrg3207
- PubMed
- Google Scholar
1. Stiles J
2. Jernigan TL
(2010) The Basics of Brain Development
Neuropsychology Review 20:327–348.
https://doi.org/10.1007/s11065-010-9148-4
- PubMed
- Google Scholar
1. Stroud H
2. Yang MG
3. Tsitohay YN
4. Davis CP
5. Sherman MA
6. Hrvatin S
7. Ling E
8. Greenberg ME
(2020) An Activity-Mediated Transition in Transcription in Early Postnatal Neurons
Neuron 107:874–890.
https://doi.org/10.1016/j.neuron.2020.06.008
- PubMed
- Google Scholar
1. Stuart T
2. Butler A
3. Hoffman P
4. Hafemeister C
5. Papalexi E
6. Mauck WM
7. Hao Y
8. Stoeckius M
9. Smibert P
10. Satija R
(2019) Comprehensive Integration of Single-Cell Data
Cell 177:1888–1902.
https://doi.org/10.1016/j.cell.2019.05.031
- PubMed
- Google Scholar
1. Stuart T
2. Srivastava A
3. Madad S
4. Lareau CA
5. Satija R
(2021) Single-cell chromatin state analysis with Signac
Nature Methods 18:1333–1341.
https://doi.org/10.1038/s41592-021-01282-5
- PubMed
- Google Scholar
(2015) Monoamine-sensitive developmental periods impacting adult emotional and cognitive behaviors
Neuropsychopharmacology 40:88–112.
https://doi.org/10.1038/npp.2014.231
- PubMed
- Google Scholar
1. Trapnell C
2. Williams BA
3. Pertea G
4. Mortazavi A
5. Kwan G
6. van Baren MJ
7. Salzberg SL
8. Wold BJ
9. Pachter L
(2010) Transcript assembly and quantification by RNA-Seq reveals unannotated transcripts and isoform switching during cell differentiation
Nature Biotechnology 28:511–515.
https://doi.org/10.1038/nbt.1621
- PubMed
- Google Scholar
1. Trapnell C
2. Roberts A
3. Goff L
4. Pertea G
5. Kim D
6. Kelley DR
7. Pimentel H
8. Salzberg SL
9. Rinn JL
10. Pachter L
(2012) Differential gene and transcript expression analysis of RNA-seq experiments with TopHat and Cufflinks
Nature Protocols 7:562–578.
https://doi.org/10.1038/nprot.2012.016
- PubMed
- Google Scholar
1. Trevino AE
2. Müller F
3. Andersen J
4. Sundaram L
5. Kathiria A
6. Shcherbina A
7. Farh K
8. Chang HY
9. Pașca AM
10. Kundaje A
11. Pașca SP
12. Greenleaf WJ
(2021) Chromatin and gene-regulatory dynamics of the developing human cerebral cortex at single-cell resolution
Cell 184:5053–5069.
https://doi.org/10.1016/j.cell.2021.07.039
- PubMed
- Google Scholar
Software
1. Turaga N
(2022) fmicompbio/monaLisa, version swh:1:rev:228cb9532d820f4b12d1f7303ec512cfc832e3c1
GitHub.

https://archive.softwareheritage.org/swh:1:dir:6a40673b98abbb4698e05744f42eb1a3e6866b73;origin=https://github.com/fmicompbio/monaLisa;visit=swh:1:snp:ed69206ebf604e2820184ead3fe42b9fe5b023cb;anchor=swh:1:rev:228cb9532d820f4b12d1f7303ec512cfc832e3c1
1. Vigneault É
2. Poirel O
3. Riad M
4. Prud’homme J
5. Dumas S
6. Turecki G
7. Fasano C
8. Mechawar N
9. El Mestikawy S
(2015) Distribution of vesicular glutamate transporters in the human brain
Frontiers in Neuroanatomy 9:23.
https://doi.org/10.3389/fnana.2015.00023
- PubMed
- Google Scholar
(2007) VISTA Enhancer Browser--a database of tissue-specific human enhancers
Nucleic Acids Research 35:D88–D92.
https://doi.org/10.1093/nar/gkl822
- PubMed
- Google Scholar
1. Wei G-H
2. Badis G
3. Berger MF
4. Kivioja T
5. Palin K
6. Enge M
7. Bonke M
8. Jolma A
9. Varjosalo M
10. Gehrke AR
11. Yan J
12. Talukder S
13. Turunen M
14. Taipale M
15. Stunnenberg HG
16. Ukkonen E
17. Hughes TR
18. Bulyk ML
19. Taipale J
(2010) Genome-wide analysis of ETS-family DNA-binding in vitro and in vivo
The EMBO Journal 29:2147–2160.
https://doi.org/10.1038/emboj.2010.106
- PubMed
- Google Scholar
1. Whyte WA
2. Orlando DA
3. Hnisz D
4. Abraham BJ
5. Lin CY
6. Kagey MH
7. Rahl PB
8. Lee TI
9. Young RA
(2013) Master Transcription Factors and Mediator Establish Super-Enhancers at Key Cell Identity Genes
Cell 153:307–319.
https://doi.org/10.1016/j.cell.2013.03.035
- PubMed
- Google Scholar
(2016) Package ‘ggplot2.’ Create Elegant Data Vis
Using Gramm. Graph. Version 2:1–189.
https://doi.org/10.1007/978-3-319-24277-4
- Google Scholar
1. Wyler SC
2. Spencer WC
3. Green NH
4. Rood BD
5. Crawford L
6. Craige C
7. Gresch P
8. McMahon DG
9. Beck SG
10. Deneris E
(2016) Pet-1 Switches Transcriptional Targets Postnatally to Regulate Maturation of Serotonin Neuron Excitability
The Journal of Neuroscience 36:1758–1774.
https://doi.org/10.1523/JNEUROSCI.3798-15.2016
- PubMed
- Google Scholar
1. Wylie CJ
2. Hendricks TJ
3. Zhang B
4. Wang L
5. Lu P
6. Leahy P
7. Fox S
8. Maeno H
9. Deneris ES
(2010) Distinct Transcriptomes Define Rostral and Caudal Serotonin Neurons
Journal of Neuroscience 30:670–684.
https://doi.org/10.1523/JNEUROSCI.4656-09.2010
- PubMed
- Google Scholar
1. Yoshihara M
2. Littleton JT
(2002) Synaptotagmin I Functions as a Calcium Sensor to Synchronize Neurotransmitter Release
Neuron 36:897–908.
https://doi.org/10.1016/S0896-6273(02)01065-6
- PubMed
- Google Scholar
1. Zaret KS
2. Carroll JS
(2011) Pioneer transcription factors: establishing competence for gene expression
Genes & Development 25:2227–2241.
https://doi.org/10.1101/gad.176826.111
- PubMed
- Google Scholar
1. Zhao ZQ
2. Scott M
3. Chiechio S
4. Wang JS
5. Renner KJ
6. Gereau RW
7. Johnson RL
8. Deneris ES
9. Chen ZF
(2006) Lmx1b is required for maintenance of central serotonergic neurons and mice lacking central serotonergic system exhibit normal locomotor activity
The Journal of Neuroscience 26:12781–12788.
https://doi.org/10.1523/JNEUROSCI.4143-06.2006
- PubMed
- Google Scholar

Decision letter

Paschalis Kratsios

Reviewing Editor; University of Chicago, United States
Catherine Dulac

Senior Editor; Harvard University, United States
Paschalis Kratsios

Reviewer; University of Chicago, United States

Our editorial process produces two outputs: i) public reviews designed to be posted alongside the preprint for the benefit of readers; ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "Reorganization of postmitotic neuronal chromatin accessibility for maturation of serotonergic identity" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, including Paschalis Kratsios as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Catherine Dulac as the Senior Editor.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

This work mapped accessible chromatin regions and gene expression using bulk and single cell analysis, performed chromatin immunoprecipitation studies of Pet-1 and Lmx1b transcription factors and investigated effects of Pet1 and Lmx1b loss of function on chromatin accessibility. The study generated a useful resource of genomic data and provided new insights into the dynamic regulation of accessible chromatin regions in postmitotic neurons. Technically, this is a tour-de-force effort.

Essential revisions:

The reviewers think the manuscript does not require additional experimental data, but it does require additional analysis of the existing data (as listed in the detailed reviewer comments below) and major revisions of the text.

1. While the manuscript provides important new data, the model how serotonergic neuron specification is controlled by Pet1 and Lmx1b remains unclear. Most importantly, the authors need to integrate better the single cell analysis with the global analysis of chromatin accessibility and TF binding data. Specifically, it would be quite interesting to address whether the pan-serotonergic selector factors also contribute to subtype diversification.

2. Another important point that deserves more attention is the analysis of differentially enriched motifs in neuronal subtypes and along the temporal trajectory. We also agree that the discrepancy between proximal vs distal TF binding and changes in chromatin accessibility should be better analyzed and highlighted in the manuscript. On the other hand, we are not sure whether correlative analysis between Pet1/Lmx1b binding sites and gene expression will be that informative, as control of gene expression is typically a result of complex integrated regulatory inputs from multiple distal enhancers.

3. We would like to see a better description of the 11 single cell clusters, and it would be great if the authors attempted to correlate the cluster identity with published adult serotonergic single cell data. However, we do understand that such analysis might be challenging and hard to interpret, so we do not think that the publication should be made conditional on this analysis.

4. Despite the wealth of new genomic data, the focus of the manuscript is somewhat elusive. The goal seems to be to study the mechanisms that control chromatin accessibility in postmitotic serotonergic neurons during their subtype diversification. The manuscript will benefit from clearer writing, explicit stating what hypotheses are being tested and better explanation of the rationale for the performed experiments.

Reviewer #1 (Recommendations for the authors):

1. Can the "pioneer" activity of Pet1 (or Lmx1b) be separated from its transcriptional activity? For example, ATAC-Seq and CUT&RUN data show that open chromatin regions are bound by Pet1. So, if Pet1 binds at a single genomic location upstream of gene x, does this binding event triggers the chromatin to open AND transcription to start? And what happens in cases where multiple binding peaks for Pet1 are found upstream of a gene? Some peaks are responsible for opening chromatin, while others for transcriptional activation?

2. Addressing the following questions by editing the text and/or re-analyzing existing data (RNA-Seq, ATAC-Seq, and Cut&RUN) could help the reader: Do all known Pet1 targets genes from previous RNA-Seq in serotonin neurons also require Pet1 to "open" their cis-regulatory regions? Are there any Pet1 targets whose expression is affected by Pet1 inactivation, but Pet1 does not open their cis-reg region? Lastly, are there any genes for which Pet1 only exerts a "pioneer function"? That is, it only opens their chromatin, but does not control directly the expression of these genes?

The same questions apply to Lmx1b.

3. It is very intriguing that the expression of 5-HT genes (Tph2, Slc6a4, Slc22a3) is low or undetectable at e14.5 despite their "open" chromatin status at this stage. Another intriguing observation comes from a previous study by the authors reporting that Pet1 switches its targets postnatally (Wyler et al., J Neurosci. 2016). One of the genes that is initially responsive to Pet1 is Tph2, but then, at postnatal stages, Tph2 expression does not depend on Pet1. Do the current data (e.g. CUT&RUN, ATAC-Seq, ChIP-seq) on WT and Pet1 cKO neurons help explain how Pet1 switches its targets in 5-HT neurons? For example, is Pet1 required to open the chromatin surrounding Tph2 and also activate Tph2 transcription early, but at later postnatal stages Pet1 only opens the chromatin for other TFs to activate Tph2 transcription?

4. Are ATAC-seq experiments quantitative enough to allow for comparisons between genotypes? For example, the height of each peak is compared between WT and cKO conditions in Figures 5e and 6i,j.

5. The authors discovered a new function for Pet1 and Lmx1b and therefore propose that these TFs define a previously unrecognized category of TFs termed "postmitotic neuron chromatin reorganizers" (PnCRs) that act specifically during postmitotic neuron maturation to shape the accessible chromatin landscapes and thereby select enhancers required for acquisition of specialized neuronal identities. While it is tempting to "name" the TFs, I would suggest to simply "name/state" their new function, especially in light of the fact that Pet1 and Lmx1b have already been termed "terminal selectors". To avoid confusion and help the reader, the authors should consider something along the following lines: Pet1 and Lmx1b exert a dual function: they not only act as terminal selectors, but also as chromatin reorganizers in postmitotic 5-HT neurons.

Reviewer #2 (Recommendations for the authors):

The authors should discuss the identity of the 11 major subtypes with unique chromatin accessibility patterns (Figure 2d; Supplemental Figure 2b) identified by scATAC and RNAseq analysis. How many of the clusters can be mapped on known serotonergic subtypes? How well do the subtypes map on published scRNAseq serotonergic neuron data?

Figure 3c: numbers do not match – gained and lost TACs do not add up to the numbers of peaks detected.

Figure3e: It is unclear how the enhancer-gene association was performed for this data. Were enhancer-gene associations used from the sc-seq data? Those were performed at e14.5 and would not have some of the peaks at e11.5 and e17.5 used in this plot. Were new enhancer-gene associations created from the temporal data- in this case this plot is making a circular argument (if the enhancer-gene associations are made based on co-regulation between expression and accessibility).

Figure 3i: needs more clear labeling of each of the gene sets in the four heatmaps

Figure 4a: pink dots in PCA seem extra?

Figure 4a: the PCA analysis shows that TACs are dominantly influenced by neuronal maturation, but such analysis does not reveal whether the Pet1-/- neurons changed identity. Instead, the authors should perform PCA analysis at one timepoint and include Pet1 negative wt neurons in the PCA analysis. This will address whether Pet1-/- neurons are more similar to Pet1 positive or Pet1 negative wt neurons.

Figure 4c-e: The analysis and interpretation of the correlation of chromatin accessibility with Pet1 binding in wt and mutant cells is extremely confusing. It is not clear what the main conclusion of this section is and what data support the conclusions. Figure 4c shows all TACs at different ages and plots their change in accessibility – the overall decrease of accessibility in Pet1-/- neurons is striking and surprising. Is this perhaps a technical issue? It is hard to believe that the loss of Pet1 TF that binds only to a subset of accessible sites results in a global ~4 fold reduction in chromatin accessibility. What is the number of TACs that gain, lose or do not change in Pet1-/- neurons? Is it correct that only 200 TACs are Pet1 regulated (i.e. bind Pet1 and change in KO)? How many Pet1 binding sites there are? How many of these binding sites are in TACs and how many are outside TACs? How are "Pet1-dependent" TACs defined?

The authors conclude that Pet1 is involved both in opening chromatin and in maintaining open chromatin. What differentiates Pet1 binding sites that gain accessibility compared from the sites that have high accessibility already at e11 (before Pet1 binds) or from the sites that loose accessibility in the conditional Pet1 knockout? Is it possible that sites that exhibit lower accessibility in conditional mice had not acquired full accessibility at e12.5 (i.e. the data cannot be interpreted as evidence for the role of TFs in the maintenance of TACs)

Figure 5: Pet1 cKO has a stronger effect on chromatin accessibility than the Pet1 full knockout. Is this a technical issue? If not, what are the chromatin regions affected in the cKO but not in full KO- are these enhancers based on chromatin profile (generated in Figure 1 at e14.5)? How do you explain these results?

The authors should integrate better the single cell analysis with analysis of Pet1 and Lmx1b binding and mutants. Do Pet1 and Lmx1b regulate TACs and expression levels of subtype specific genes in addition to pan-Pet1 lineage genes? Are these motifs co-enriched with other, more cell type specific motifs at subtype-specific peaks?

While majority of Lmx1b co-binds with Pet1, majority of Pet1 binds independently from Lmx1b. The data on Lmx1b and Pet1 binding overlap should be presented as a scatter plot rather than Venn diagrams. Are co-bound TACs regulated in a different way compared to TACs bound by Pet1 or Lmx1b only?

Considering that Pet1 expression depends on Lmx1b, the authors should address whether Pet1 expression is lost in Lmx1b CKO or whether it is maintained independently of Lmx1b.

Figure 7e: Legend says "Differential expression (from E11.5 to E17.5) of the genes associated with Pet1- occupied TACs in Pet1-cKO vs. wildtype 5-HT neurons (left)…", however the y-axis in plot is labeled "e17.5 KO/wt". If the plot is correctly labeled in the figure, it seems that genes associated with TACs are not changing expression in expected patterns. However, it does look like some synaptic genes do show reduced expression (Figure 7 and supp Figure 7). It would be interesting to know how gene expression is affected globally in cKO and DKO conditions, and if the genes affected are mainly pan-pet1 lineage genes (like 5-HT, synaptic, etc) or also subtype specific genes.

Reviewer #3 (Recommendations for the authors):

1. In general, I felt the work, as it is presently written, does not flow very well, going back and forward in stages and techniques. I am aware it is a lot of data and different methods but maybe it would be better organized as: (1) longitudinal data e11, e14, e17; (2) mutants longitudinal 3) e14.5 single cell data

2. I find authors often focus on specific examples from already well-characterized targets of Pet1 and Lmx1b, such as 5HT pathway genes but I miss a more unbiased analysis. Also, it would be nice to include a more detailed and integrated analysis of the embryonic scATACseq and scRNAseq with the adult scRNAseq (Okaty et al. and/or Ren et al. reports) as well as a more profound study of differential TF expression and TF motif enrichment in different TACs of different 5HT neuron subclasses, that could maybe help identify new TFs involved in subtype specification. Similarly, those analysis could also be done comparing differential TF expression and motif enrichment in differentially accessible TACs along time (e11, e14 and e17 data) to try to unravel TFs involved in maturation or in specific transient events.

3. The abstract mentions several conclusions, I find some of them are not really novel or lack enough data to support them:

a) "heterogeneous chromatin landscapes are established early": this has been already shown for other neuronal types.

b) "reveal the regulatory programs driving subtype identities of Pet1-lineage

neurons that generate serotonin (5-HT) neurons": I don´t think present data identify regulatory programs (that is networks of TFs and their targets).

c) Dynamic enhancers in postmitotic neurons have been already shown.

d) "We find that Pet1 and Lmx1b control chromatin accessibility to select Pet1-lineage specific enhancers for 5-HT neurotransmission and synaptogenesis" The description of changes in chromatin accessibility is interesting, however, the role of Pet1 and Lmx1b controlling gene expression of 5HT neurotransmission and synaptogenesis is not new.

e) "Required to maintain accessible chromatin in maturing neurons", also interesting however both TFs are already described to be required for gene expression maintenance

4. Regarding the mouse model: what is the % of 5HT neurons that are Pet1:YFP negative in each analysed time point? and vice versa, are all Pet1:YFP cells 5HT positive?

5. Why scATAC approach identifies 60,117 accessible sites not identified in the bulk Pet1? Are those uniquely identified TACs present only in subpopulations of Pet1 positive cells?

6. How do the 11 chromatin clusters of Pet1 scATAC correlate with adult scRNAseq data? It would be nice to have an integrated analysis of the embryonic and adult scRNAseq and the scATACseq data (including Okaty et al. and/or Ren et al. reports) to find the embryonic chromatin correlates of the adult 5HT subtypes.

7. Motif enrichment in scATAC clusters (Sup figure 3c). (1) Can the authors speculate on why ETS motif is only enriched in Tph2 cluster and LIM motif in Lrfn5 cluster? (2) could this motif enrichment analysis be combined with cluster specific or enriched TF expression for each family to identify TF candidates for subtype specification? For example, authors mention enriched expression of nkx6.1 for one cluster, which also shows enrichment in NK motifs (this double match TF expression with TF motif is very interesting and should be more clearly mentioned). Similar correlation analysis should be done for other TFs, TF motifs and clusters.

8. Sup figure 4j: motif enrichment of gained and lost TACs. Are the same motifs present in gained from e11 to e14 than those gained from e14 to e17? and the same question applies for the lost motifs. And as mentioned already, it would be advisable to merge TF expression dynamics with TF motif enrichment dynamics.

9. How does gene expression, gains and losses between wt and Pet1 mutants at different stages correlate with TACs: (1) What is the % of genes downregulated in Pet1 mutant compared to control that shows at least one assigned TAC in wt scATACseq or in bulk atacseq (2) from those, what is the % of downregulated genes in Pet1 mutant with TAC that lose accessibility in at least one assigned TAC? Similar analysis will be interesting for the >1000 genes with gained expression in e11.5 Pet1 mutants reported in Wyler et al. 2016.

10. Are genes associated with TACs lost in Pet1 mutant expressed in specific clusters of 5HT or broadly expressed in 5ht neurons? are they 5HT specific?

11. Could authors speculate why Pet1-/- phenotype is milder than Pet1cKO? (372 TACs less accessible and 3 more accessible compared to 1834 and 391 in cKO), is there any compensatory effect in the regular mutant? How do these 372 TACs and 3 TACs and their respective assigned genes match with DEX genes in Pet1 mutants at e14?

12. What genes are assigned to the 391 TACs gained in Pet1cKO? do they correlate with gained gene expression in this mutant background? what motifs are enriched in these TACs? what GO terms?

13. Most differential TACs in Pet1 lineage are found in distal intergenic or introns, and show enhancer chromatin marks, however CUT and RUN binding for both Pet1 and Lmx1b is found preferentially at proximal regions (promoters). Could authors specify de % of proximal TACs with Pet1 or Lmx1b binding compared to % of distal TACs with binding? Is motif enrichment in TACs different in proximal and distal TACs? What is the mechanistic explanation of this discrepancy between CUT and RUN binding and differential TACs and lost TACs in the mutants?

14. Donovan et al. showed Lmx1b is upstream Pet1 expression, at least in some 5HT neurons, Is the effect on chromatin accessibility of Lmx1b mutant mediated, at least in part, by the loss of Pet1? How many lost TACs in Lmx1b are bound by Pet1 but not by Lmx1b?

15. I would remove the last section of the manuscript, as I don´t think the data support a role in synaptogenesis: Authors mention vglut3 and syt1 as panneuronal genes, however they are neuronal effector genes not expressed in all neurons but in specific neurons subtypes. Moreover, they are not involved in synaptogenesis but in synaptic function and neurotransmission. Finally, the analysis of pericellular baskets does not make much sense as these mutants show defects in axon projections to start with.

https://doi.org/10.7554/eLife.75970.sa1

Author response

Essential revisions:

The reviewers think the manuscript does not require additional experimental data, but it does require additional analysis of the existing data (as listed in the detailed reviewer comments below) and major revisions of the text.

1. While the manuscript provides important new data, the model how serotonergic neuron specification is controlled by Pet1 and Lmx1b remains unclear. Most importantly, the authors need to integrate better the single cell analysis with the global analysis of chromatin accessibility and TF binding data. Specifically, it would be quite interesting to address whether the pan-serotonergic selector factors also contribute to subtype diversification.

In response to this suggestion, we have added into the manuscript figure panels to describe the percentage of Pet1- and/or Lmx1b-regulated TACs that are subtype specific vs. common to multiple subtypes. These results show that Pet1 and Lmx1b both regulate some subtype-specific TACs and common TACs. We agree that it would be very interesting to test whether either TF contributes to subtype diversification. However, the focus of this work is on TF regulation of chromatin accessibility in postmitotic neurons, and thoroughly investigating TF regulation of subtype diversification is beyond the scope of this manuscript.

2. Another important point that deserves more attention is the analysis of differentially enriched motifs in neuronal subtypes and along the temporal trajectory. We also agree that the discrepancy between proximal vs distal TF binding and changes in chromatin accessibility should be better analyzed and highlighted in the manuscript. On the other hand, we are not sure whether correlative analysis between Pet1/Lmx1b binding sites and gene expression will be that informative, as control of gene expression is typically a result of complex integrated regulatory inputs from multiple distal enhancers.

We agree that these phenomena are very intriguing while also sharing the view that we are not sure whether correlation between TF binding sites and gene expression is all that informative. We have made an effort to give this more attention. Please see our detailed responses below in our responses to Reviewer #3 major points 8 and 13.

3. We would like to see a better description of the 11 single cell clusters, and it would be great if the authors attempted to correlate the cluster identity with published adult serotonergic single cell data. However, we do understand that such analysis might be challenging and hard to interpret, so we do not think that the publication should be made conditional on this analysis.

This is a good suggestion as it would be interesting to relate our developmental subtype results to published adult subtypes. As the reviewers note, this is a difficult analysis due in particular to the significant time gap between the datasets. To incorporate this suggestion, we focused on the adult serotonergic scRNA-seq data from Okaty et al. (2020) since they profiled the highest number of cells. Cell number is critical to identify any patterns that could emerge from projecting data from one dataset to another. Please see specific responses to reviewer #2 comment 1 below for further details on our approach and results.

4. Despite the wealth of new genomic data, the focus of the manuscript is somewhat elusive. The goal seems to be to study the mechanisms that control chromatin accessibility in postmitotic serotonergic neurons during their subtype diversification. The manuscript will benefit from clearer writing, explicit stating what hypotheses are being tested and better explanation of the rationale for the performed experiments.

The goal of the manuscript is to study the mechanisms that control chromatin accessibility in serotonergic neurons during their early postmitotic maturation, which is a period predominated by axon and synapse formation. We believe the focus became confusing because we introduced some single cell datasets early in the manuscript. The main purpose of the single cell analyses were to (1) demonstrate that chromatin accessibility is significantly heterogeneous not only among different neuron-types in the brain but also within Pet1 lineage neurons that will mature to become 5-HT neuron subtypes, (2) show that there is a strong correlation between chromatin accessibility and gene expression, and (3) associate specific TACs to specific genes that can then be subsequently be used during bulk ATAC-seq and RNA-seq analyses to understand the temporal maturation of chromatin.

In response to the reviewers’ feedback we clarified our writing in the summary to better explain the rationale for the experiments we performed.

Reviewer #1 (Recommendations for the authors):

1. Can the "pioneer" activity of Pet1 (or Lmx1b) be separated from its transcriptional activity? For example, ATAC-Seq and CUT&RUN data show that open chromatin regions are bound by Pet1. So, if Pet1 binds at a single genomic location upstream of gene x, does this binding event triggers the chromatin to open AND transcription to start? And what happens in cases where multiple binding peaks for Pet1 are found upstream of a gene? Some peaks are responsible for opening chromatin, while others for transcriptional activation?

We believe the simplest interpretation of our findings is that Pet1 is able to open partially or completely inaccessible chromatin and subsequently engage in transcriptional activation. However, the details of Pet1 action are likely more complex and will require future study to elucidate.

It is difficult to definitively separate the chromatin “pioneering” and transcriptional activities of TFs, but we do observe instances suggesting the two processes may be decoupled. For example, there is significant Pet1 and Lmx1b occupancy at gene promoters that correlates strongly with Pet1/Lmx1b dependent gene expression, yet ablating Pet1 and/or Lmx1b did not result in a significant loss of chromatin accessibility at those promoters, suggesting that factors other than Pet1 and Lmx1b could be responsible for maintaining accessibility in those TSS proximal regions. As another example, at E14.5 Pet1 and Lmx1b are detected in all Pet1 neurons by scRNA-seq and the TACs of 5-HT neurotransmission genes Tph2, Slc6a4, and Slc22a3 are in an accessible state, yet the expression of these genes is low or undetected in many individual Pet1 neurons, suggesting that there is a delay in Pet1/Lmx1b transactivation and chromatin opening preceded gene transcription.

The reviewer raises a number of other interesting questions, some of which cannot be answered definitively with our present findings. For example, we do not believe our findings shed light on whether Pet1 occupancy at some of an array of binding sites strictly controls opening while others strictly control transcriptional activation. It seems addressing this question would require systematic editing of individual binding sites, in vivo, followed by chromatin mapping and expression analyses.

Single cell studies of dividing cells (Ma et al., 2020; Trevino et al., 2021) drew a similar conclusion that chromatin accessibility often occurs prior to mRNA expression.

The reviewer’s proposal that Pet1 binding at different sites could have different functions (some for opening chromatin, while others for transcriptional activation) is a very intriguing idea. However, it is difficult to test this idea because TAC regions usually contain the binding sites of multiple TFs, so ablating the TF or mutating a binding site of the TF that is maintaining chromatin access also prevents the regional binding of other TFs that could be responsible for gene transcription.

2. Addressing the following questions by editing the text and/or re-analyzing existing data (RNA-Seq, ATAC-Seq, and Cut&RUN) could help the reader: Do all known Pet1 targets genes from previous RNA-Seq in serotonin neurons also require Pet1 to "open" their cis-regulatory regions? Are there any Pet1 targets whose expression is affected by Pet1 inactivation, but Pet1 does not open their cis-reg region? Lastly, are there any genes for which Pet1 only exerts a "pioneer function"? That is, it only opens their chromatin, but does not control directly the expression of these genes?

The same questions apply to Lmx1b.

Our present findings suggest that not all known Pet1 target genes in serotonin neurons require Pet1 to open their cis-regulatory regions. There are Pet1 targets whose expression is affected by Pet1 inactivation, but Pet1 does not appear to open their associated CREs. Conversely, there are also examples for which Pet1 only exerts a “pioneer function” and does not control gene expression.

There are several possible explanations for these phenomena. First, not all differentially regulated genes in the Pet1 KO 5-HT neurons have direct nearby Pet1 occupancy, which suggests that some differential gene expression is due to either indirect or compensatory effects through the actions of other TFs, and therefore we do not expect to find Pet1 mediated chromatin accessibility changes near those genes. Secondly, Pet1 likely occupies CREs in combination with many other TFs similar to what we have shown between Pet1 and Lmx1b. In some instances, the removal of Pet1 is sufficient to destabilize the protein complex at a genomic region and result in the collapse of chromatin accessibility, whereas in other instances, chromatin accessibility in a genomic region is maintained in the absence of Pet1. We hypothesize that this difference can be attributed to the activities of the other regulatory factors that occupy CREs in combination with Pet1.

Lastly, as the reviewer proposed, it is very possible that the “pioneering” and transcriptional activities of TFs can be decoupled, though our data lack the resolution to definitively conclude this with certainty. To prioritize accuracy, we limited our TAC-to-gene association parameter so that TACs were only assigned to genes that were within 500kb, whereas enhancers are known to act across greater distances than 500kb, so we do not claim to have captured all TAC-to-gene interactions. In addition, CRE association with genes can be very fluid. For example, studies have shown that deletion of one TSS proximal enhancer can lead to compensatory activities from a more distal enhancer (e.g. Hong et al., 2008; Frankel et al., 2010; Perry et al., 2010). Therefore, it is possible that TACs dependent on Pet1 are not paired to all actual target genes.

Same for Lmx1b.

3. It is very intriguing that the expression of 5-HT genes (Tph2, Slc6a4, Slc22a3) is low or undetectable at e14.5 despite their "open" chromatin status at this stage. Another intriguing observation comes from a previous study by the authors reporting that Pet1 switches its targets postnatally (Wyler et al., J Neurosci. 2016). One of the genes that is initially responsive to Pet1 is Tph2, but then, at postnatal stages, Tph2 expression does not depend on Pet1. Do the current data (e.g. CUT&RUN, ATAC-Seq, ChIP-seq) on WT and Pet1 cKO neurons help explain how Pet1 switches its targets in 5-HT neurons? For example, is Pet1 required to open the chromatin surrounding Tph2 and also activate Tph2 transcription early, but at later postnatal stages Pet1 only opens the chromatin for other TFs to activate Tph2 transcription?

We agree that the dynamic changes in gene target sensitivity to Pet1 are interesting; this observation is what initially motivated our interest in probing the dynamics of postmitotic chromatin accessibility. Unfortunately, we do not believe our current findings offer any insight into the switching mechanism. We envision many potential mechanisms: (1) At around birth, Pet1 is no longer needed to remodel chromatin around some serotonergic genes such as Tph2 and other TFs such Lmx1b are sufficient for this job. To begin to investigate this idea we would need to target Pet1 at around P0 to see if loss of Pet1 at that time no longer had an effect on chromatin.

4. Are ATAC-seq experiments quantitative enough to allow for comparisons between genotypes? For example, the height of each peak is compared between WT and cKO conditions in Figures 5e and 6i,j.

ATAC-seq experiments are quantitative enough to allow for comparisons between genotypes. The height of each peak reflects the number of sequencing reads that map to that genomic region. Datasets are normalized to adjust for signal coverage across the genome. The height of visualized peaks can be compared between WT and cKO conditions.

5. The authors discovered a new function for Pet1 and Lmx1b and therefore propose that these TFs define a previously unrecognized category of TFs termed "postmitotic neuron chromatin reorganizers" (PnCRs) that act specifically during postmitotic neuron maturation to shape the accessible chromatin landscapes and thereby select enhancers required for acquisition of specialized neuronal identities. While it is tempting to "name" the TFs, I would suggest to simply "name/state" their new function, especially in light of the fact that Pet1 and Lmx1b have already been termed "terminal selectors". To avoid confusion and help the reader, the authors should consider something along the following lines: Pet1 and Lmx1b exert a dual function: they not only act as terminal selectors, but also as chromatin reorganizers in postmitotic 5-HT neurons.

We do appreciate the reviewer’s acknowledgment that we have discovered a chromatin reorganizing function for terminal selectors in maturing postmitotic neurons. We agree that our proposal to designate these as PnCRs, may be confusing since they already go by the descriptor, terminal selector. Yet, we maintain our findings reveal reorganization of postmitotic chromatin accessibility, which is a previously unrecognized functional attribute of how terminal selectors work. Thus, our findings advance the understanding of how terminal selectors determine specialized neuronal identities: terminal selectors, at least Pet1 and Lmx1b in mouse 5-HT neurons, remodel postmitotic accessible chromatin for selection of enhancers. Subsequently they transactivate via those selected enhancers, genes whose expression contributes to specialized neuronal identities. In this interpretation, Pet1 and Lmx1b do not possess dual and distinct terminal selector and chromatin reorganizing functions but rather the chromatin reorganizing function of these TFs is an inherent feature of their terminal selector action that operates prior to or together with transactivation. We believe it is reasonable and appropriate to discuss this duality of terminal selector action that has been inferred from the study of Pet1 and Lmx1b. How do these TFs reorganize postmitotic chromatin? One possibility is that they directly alter nucleosome positioning upon binding to ETS/HD motifs. Alternatively, we hypothesize that through their initial binding at ETS/HD motifs they recruit chromatin remodeling complexes that further open chromatin at these motifs for optimal sequence specific gene transactivation.

Reviewer #2 (Recommendations for the authors):

The authors should discuss the identity of the 11 major subtypes with unique chromatin accessibility patterns (Figure 2d; Supplemental Figure 2b) identified by scATAC and RNAseq analysis. How many of the clusters can be mapped on known serotonergic subtypes? How well do the subtypes map on published scRNAseq serotonergic neuron data?

Okaty et al. performed 10X scRNA-seq at 1.5 to 2.5 mo on only DRN cells. The Ren et al. (2019) study used Sert positive neurons in adult dorsal and median raphe. Our analysis included cells from all raphe nuclei at E14.5. Since there is a large time gap between our datasets, it can be difficult for the integration algorithm to identify anchors between the two datasets. Nevertheless, we performed a reciprocal unimodal UMAP projection of our E14.5 scRNA with the Okaty et al. ~2mo scRNA data. First, we projected our scRNA clustering and UMAP structure onto the adult scRNA data. 93% of Okaty 10X DRN cells map to our Tph2/Slc6a4 (1099) and Pax5/En1 (1082) clusters. Since Tph2, Slc6a4, and the rhombomere 1 markers (e.g. En1) are highest in these two clusters, it is not surprising that the adult DRN neurons, which are only derived from isthmus rhombomere 1, are the most highly correlated with these clusters. Next, we projected the Okaty DRN scRNA clustering/UMAP onto our E14.5 scRNA. Since we did not have the exact clustering from Okaty et al., we analyzed their data according to their methods as precisely as possible, although it was not possible to reproduce the clustering exactly. Despite this issue, our clustering (14 clusters) and UMAP are strikingly similar to the results in Okaty et al. Then, we used the top cluster marker genes from Okaty et al., to determine which of our clusters are most similar to the Okaty clusters. By projecting the adult UMAP structure onto our E14.5 scRNA, most of our cells mapped to adult cluster 13, which are high Slc17a8 and low Tph2+ cells. This is reasonable since most of our cells have high Slc17a8 expression. The next largest number of cells mapped to adult cluster 1, which are high Tph2 expressing cells. 138 cells were mapped to cluster 7, which co-express Tph2 and Slc17a8. Lastly, 35 cells map to cluster 12, which are the Met expressing cells. For clarity, we added the projection of the Okaty data onto the embryonic clusters the manuscript in Figure 2 —figure supplement 1, since we do not have exactly the same clustering as published for the adult data.

Figure 3c: numbers do not match – gained and lost TACs do not add up to the numbers of peaks detected.

In our original figure panel, the gain/loss TAC numbers referred to the TACs with greater or equal than 2 fold change (increase = gain; decrease = loss) in accessibility and FDR <0.01, not the number of TACs that entirely appear/disappear between time points, so it did not add up to the total number of peaks called at each developmental time point. We have modified the figure to include both sets of numbers and included an explanation in the figure legends.

Figure3e: It is unclear how the enhancer-gene association was performed for this data. Were enhancer-gene associations used from the sc-seq data? Those were performed at e14.5 and would not have some of the peaks at e11.5 and e17.5 used in this plot. Were new enhancer-gene associations created from the temporal data- in this case this plot is making a circular argument (if the enhancer-gene associations are made based on co-regulation between expression and accessibility).

The enhancer-gene association was calculated using the E14.5 scRNA-seq and scATAC-seq datasets and tested on a separate set of bulk time course RNA-seq and ATAC-seq datasets to avoid making a circular argument. The single cell SnapATAC algorithm only paired a TAC with one or more genes if there was correlation between gene expression and accessibility within a single cell, so TACs with no differential accessibility between subtypes were not paired with a gene. Furthermore, to ensure the best accuracy we used the stringent parameter that a pairing is made only if the detected TAC was within 500kb of the gene. As a result, we only paired a small fraction of TACs with candidate genes. We then asked whether the accessibility of this specific subset of TACs is temporally correlated with gene expression in E11.5-E17.5 bulk datasets generated at another time and found a strong correlation.

Figure 3i: needs more clear labeling of each of the gene sets in the four heatmaps

We edited the figure to label the gene sets next to the heatmaps.

Figure 4a: pink dots in PCA seem extra?

They were indeed extra because of misalignment of the colored dots during figure compilation and we have corrected the figure.

Figure 4a: the PCA analysis shows that TACs are dominantly influenced by neuronal maturation, but such analysis does not reveal whether the Pet1-/- neurons changed identity. Instead, the authors should perform PCA analysis at one timepoint and include Pet1 negative wt neurons in the PCA analysis. This will address whether Pet1-/- neurons are more similar to Pet1 positive or Pet1 negative wt neurons.

We added PCA analysis at E14.5 with Pet1^neg neurons, wildtype Pet1 neurons, and Pet1-/- Pet1 neurons to Figure 4 —figure supplement 1.

Figure 4c-e: The analysis and interpretation of the correlation of chromatin accessibility with Pet1 binding in wt and mutant cells is extremely confusing. It is not clear what the main conclusion of this section is and what data support the conclusions. Figure 4c shows all TACs at different ages and plots their change in accessibility – the overall decrease of accessibility in Pet1-/- neurons is striking and surprising. Is this perhaps a technical issue? It is hard to believe that the loss of Pet1 TF that binds only to a subset of accessible sites results in a global ~4 fold reduction in chromatin accessibility. What is the number of TACs that gain, lose or do not change in Pet1-/- neurons? Is it correct that only 200 TACs are Pet1 regulated (i.e. bind Pet1 and change in KO)? How many Pet1 binding sites there are? How many of these binding sites are in TACs and how many are outside TACs? How are "Pet1-dependent" TACs defined?

We thank the reviewer for catching an error in Figure 4c and its legend that we presume has led to confusion. Figure 4c box plots display the distribution of the log2 fold change in accessibility of TACs showing differential accessibility between Pet1-/- vs. wildtype Pet1 neurons (not all TACs that were identified) at the three embryonic time points. We present the findings this way to make the point that there is an overall decrease in accessibility of TACs in Pet1-/- neurons compared to wildtype. This supports our interesting and novel conclusion that Pet1 is mainly required for opening or maintaining chromatin accessibility in postmitotic neurons. 372 TACs lost accessibility and 3 TACs gained accessibility (fold change ≥2, FDR<0.05) at E14.5 while the remaining E14.5 Pet1 neuron TACs did not change accessibility in a statistically significant manner. The reviewer is correct that only 200 of these TACs are Pet1 regulated (i.e. bind Pet1 and change in Pet1-/-). As shown in Figure 6 —figure supplement 1, Pet1 CUT&RUN peak calling identified 23,594 Pet1 binding sites, 22,285 of which are in TACs and 3280 are outside of TACs. We defined “Pet1-dependent” and “Pet1-regulated” TACs as TACs that show differences in accessibility (gain or loss) in the absence of Pet1.

The authors conclude that Pet1 is involved both in opening chromatin and in maintaining open chromatin. What differentiates Pet1 binding sites that gain accessibility compared from the sites that have high accessibility already at e11 (before Pet1 binds) or from the sites that loose accessibility in the conditional Pet1 knockout? Is it possible that sites that exhibit lower accessibility in conditional mice had not acquired full accessibility at e12.5 (i.e. the data cannot be interpreted as evidence for the role of TFs in the maintenance of TACs)

In this study we define maintenance as ongoing or continued requirement for Pet1 in sustaining wildtype chromatin landscape, in contrast to pioneer factors that induce “epigenetic memory” without ongoing expression of those factors. Therefore, if accessibility is lost in the cKO, we interpret that to mean that ongoing Pet1 activity is needed to maintain/sustain/preserve accessibility at the normal level.

About half of Pet1 or Lmx1b regulated TACs do not significantly change in accessibility between E11.5 and E14.5 in wildtype Pet1 neurons but lose accessibility in the cKOs, including the majority of 5-HT neurotransmission gene TACs. We added these numbers to the volcano graphs in Figures5 and 6.

Figure 5: Pet1 cKO has a stronger effect on chromatin accessibility than the Pet1 full knockout. Is this a technical issue? If not, what are the chromatin regions affected in the cKO but not in full KO- are these enhancers based on chromatin profile (generated in Figure 1 at e14.5)? How do you explain these results?

Our best hypothesis is that there is a compensatory effect to losing Pet1 after it has already been expressed and has activated other genes, which include other transcription factors. Only ~20% of Pet1 maintained TACs are directly occupied by Pet1 in wildtype neurons at E14.5 (and very few of the TACs that gain accessibility in Pet1-cKO have direct Pet1 occupancy), whereas 53% of Pet1-dependent TACs in KO are directly occupied by Pet1, which suggests that the larger number of TAC changes in the cKO are attributable to indirect effects.

Similar to ATAC-seq, we have observed in our lab differences between Pet1-cKO and Pet1-/- 5-HT neuron transcriptomic changes in RNA-seq experiments, which suggest that the difference between the two knockout models is unlikely to be a simple technical issue. We have also shown previously by knocking down Pet1 at different developmental time points that Pet1 switches transcriptional targets during different stages of 5-HT neuron development (Wyler et al., 2016), so it is possible that Pet1 regulates more TACs at later time points.

The chromatin regions affected in the cKO but not in germ line KO are predominantly (59%) active and poised enhancers distal to the TSS.

The authors should integrate better the single cell analysis with analysis of Pet1 and Lmx1b binding and mutants. Do Pet1 and Lmx1b regulate TACs and expression levels of subtype specific genes in addition to pan-Pet1 lineage genes? Are these motifs co-enriched with other, more cell type specific motifs at subtype-specific peaks?

The majority of the Pet1 or Lmx1b regulated TACs belong to pan-Pet1 lineage genes shared among multiple subtypes, but there are also subtype-specific genes that are regulated by the two TFs. 21% of Pet1 regulated TACs and 15% of the Lmx1b regulated TACs are subtype-specific.

We show motif activity enrichment for each scATAC cluster in Figure 2—figure supplement 2. We do not expect Pet1 (ETS) and Lmx1b (LIM-HD) motifs to be enriched in every cluster since they are equally expressed in all clusters.

While majority of Lmx1b co-binds with Pet1, majority of Pet1 binds independently from Lmx1b. The data on Lmx1b and Pet1 binding overlap should be presented as a scatter plot rather than Venn diagrams. Are co-bound TACs regulated in a different way compared to TACs bound by Pet1 or Lmx1b only?

We added to the manuscript a scatter plot showing the binding overlap between Pet1 and Lmx1b as suggested.

On average, in wildtype Pet1-lineage neurons co-bound TACs show greater ATAC-seq signal than TACs bound by only Pet1 or only Lmx1b. Loss of chromatin accessibility is seen in cKO conditions both for TACs that are co-bound and for those bound by only Pet1 or only Lmx1b.

Author response image 1

Download asset Open asset

Considering that Pet1 expression depends on Lmx1b, the authors should address whether Pet1 expression is lost in Lmx1b CKO or whether it is maintained independently of Lmx1b.

Pet1 expression is reduced to 52% of wildtype level in the Lmx1b-cKO Pet1 neurons based on our RNA-seq analyses at E17.5. However, the reduction in Pet1 expression cannot entirely account for the dramatic changes in chromatin accessibility seen in the Lmx1b-cKO neurons because 1) significant residual Pet1 expression remains 2) Lmx1b-cKO neurons show 2823 differentially accessible TACs relative to wildtype neurons that were not observed in the Pet1-cKO (Figure 6 —figure supplement 1), including TACs associated with the 5-HT neurotransmission genes Gchfr, Slc18a2, and Maoa. We added a graph in Figure 6 —figure supplement 1 to show the expression of Pet1 in Lmx1b-cKO and the expression of Lmx1b in Pet1-cKO.

Figure 7e: Legend says "Differential expression (from E11.5 to E17.5) of the genes associated with Pet1- occupied TACs in Pet1-cKO vs. wildtype 5-HT neurons (left)…", however the y-axis in plot is labeled "e17.5 KO/wt". If the plot is correctly labeled in the figure, it seems that genes associated with TACs are not changing expression in expected patterns. However, it does look like some synaptic genes do show reduced expression (Figure 7 and supp Figure 7). It would be interesting to know how gene expression is affected globally in cKO and DKO conditions, and if the genes affected are mainly pan-pet1 lineage genes (like 5-HT, synaptic, etc) or also subtype specific genes.

The graphs show that the genes linked to TACs that lose accessibility in the Lmx1b-cKO (“Loss”) are downregulated in expression relative to genes linked to TACs that show no change in accessibility in the Lmx1b-cKO (“No Change”) (middle panel). Similarly, genes linked to TACs that lose accessibility in the DKO are downregulated in expression relative to genes linked to TACs with no change in accessibility in the DKO (right panel). These two associations were the only ones that were statistically significant and are consistent with the genes Syt1 and Slc17a8 that we used as examples.

Other correlations did not reach statistical significance, including Pet1-cKO Loss TACs and the expression of their associated genes. One possible explanation for this is that while the majority of Pet1- or Lmx1b-regulated TACs are likely enhancers, a significant percentage did not have common enhancer or promoter associated histone modifications (e.g. Figures 5d and 6e) and could therefore be silencers that negatively influence target gene expression, or insulators that do not directly influence expression. For example, we have previously found that the loss of Pet1 can lead to the activation of repressed genes (Wyler et al., 2016), suggesting that Pet1 could have repressor activities. Another possibility is that enhancer usage is dynamic in neurons. When one enhancer becomes decommissioned in the KO neuron, shadow enhancers that act redundantly on the same gene may compensate to maintain gene expression.

As suggested by the reviewer we added a panel to show how gene expression is affected globally in DKO condition in the figure supplement for Figure 7. The top biological function gene ontology terms related to the downregulated genes in DKO Pet1 neurons are indole-containing compound (e.g. serotonin) metabolic process and neuron projection guidance. The top biological function gene ontology terms related to the up-regulated genes are response to lipoprotein particle and membrane invagination. The most significant (lowest FDR) GO term is serotonin biosynthetic process.

Pet1-cKO and Lmx1b-cKO RNA-seq datasets were previously generated and analyzed in Donovan et al., 2019, so we did not include detailed analyses of these datasets in this paper. The genes affected in cKO and DKO are mostly genes shared among multiple subtypes of Pet1 lineage neurons but some Pet1 and Lmx1b regulated genes are subtype specific.

Reviewer #3 (Recommendations for the authors):

1. In general, I felt the work, as it is presently written, does not flow very well, going back and forward in stages and techniques. I am aware it is a lot of data and different methods but maybe it would be better organized as: (1) longitudinal data e11, e14, e17; (2) mutants longitudinal 3) e14.5 single cell data

We have discussed this comment at length to consider the reviewer’s proposed reorganization. After careful consideration our view is that although there may be issues with flow in our chosen organization it appears to work best given the varied datasets and methods used. The reason that the single cell datasets were presented towards the beginning was that we used those data to link TACs to their putative regulated genes, and later used these TAC-to-gene pairings on bulk sequencing data to probe the importance of chromatin regulation for developmental/temporal gene expression trajectories. Single cell data gave us much higher resolution for enhancer-gene mapping than mapping using bulk ATAC-seq and RNA-seq data.

2. I find authors often focus on specific examples from already well-characterized targets of Pet1 and Lmx1b, such as 5HT pathway genes but I miss a more unbiased analysis. Also, it would be nice to include a more detailed and integrated analysis of the embryonic scATACseq and scRNAseq with the adult scRNAseq (Okaty et al. and/or Ren et al. reports) as well as a more profound study of differential TF expression and TF motif enrichment in different TACs of different 5HT neuron subclasses, that could maybe help identify new TFs involved in subtype specification. Similarly, those analysis could also be done comparing differential TF expression and motif enrichment in differentially accessible TACs along time (e11, e14 and e17 data) to try to unravel TFs involved in maturation or in specific transient events.

We maintain that Pet1 lineage neuron specific TACs are of interest because these determine overt serotonergic identity. As suggested by the reviewer, we added more detailed and integrated correlation of our embryonic scATAC-seq and scRNA-seq with the Okaty adult scRNA-seq (more discussion on this in response to reviewer 2 comment #1). We have included in Figure 2 —figure supplement 2 some data on differential TF expression and TF motif enrichment in TACs of different 5-HT neuron subclasses; while this is an interesting topic, the focus of this manuscript is on how Pet1 and Lmx1b regulate shared 5-HT neuron TACs that are important for 5-HT connectivity and neurotransmission.

We included in Figure 3 —figure supplement 1j the motifs that are differentially accessible along time to analyze the TFs involved in maturation. See response to comment #8 for more discussion on this.

3. The abstract mentions several conclusions, I find some of them are not really novel or lack enough data to support them:

a) "heterogeneous chromatin landscapes are established early": this has been already shown for other neuronal types.

Heterogeneity in chromatin landscapes has not been shown for 5-HT/Pet1 neurons. With the exception of the very recent publication from Allaway et al. (2021) and Closser et al. (2022), studies on brain neuron chromatin diversity relied on single cell analyses of dissect bulk brain dissections, largely focusing on the differences between different neuronal and non-neuronal cell types instead of profiling the interesting heterogeneity that exists even within a single neurotransmitter subtype or single transcriptional factor lineage subpopulation. Thus, we maintain our presentation of heterogeneous chromatin landscapes within the single population of Pet1 lineage neurons distinguishes our work.

b) "reveal the regulatory programs driving subtype identities of Pet1-lineage

neurons that generate serotonin (5-HT) neurons": I don´t think present data identify regulatory programs (that is networks of TFs and their targets).

We maintain that our analyses do in fact shed light on the Pet1 neuron gene regulatory network. First, our study identifies over 15,000 putative cis-regulatory regions that uniquely define the Pet1 neurons and are not found in other hindbrain neurons or annotated ENCODE ATAC-seq datasets from many tissues and cell types. We then used motif analysis to identify candidate TFs that drive Pet1 neuron identity. This analysis revealed that ETS motifs and LIM-homeodomain motifs, among many TF families, are highly enriched in Pet1 neuron specific TACs. We then associated Pet1 neuron subtype specific cis-regulatory elements with subtype specific transcription factors and downstream target genes using motif discovery and network inference. We tested our enhancer-to-gene associations using independently generated bulk RNA-seq time course datasets and found a strong correlation between TAC accessibility and target gene expression. And lastly, as proof of principle we ablated two of these candidate TFs, Pet1 and Lmx1b, and show that they directly regulate neuronal chromatin accessibility. Lastly, we show that the expression of genes associated with Lmx1b and Pet1 maintained TACs are overall significantly downregulated in expression in the absence of Lmx1b and Pet1.

c) Dynamic enhancers in postmitotic neurons have been already shown.

We agree that dynamic enhancer accessibility in postmitotic neurons has been reported in postnatal cerebellar development (Frank et al. 2015), Sst and Vip neurons (Stroud et al., 2020), and in developing forebrain tissue (Nord et al. 2013). Although we realize we were not the first to report dynamic enhancers in postmitotic neurons, enhancers have not been mapped, or shown to be dynamic, for 5-HT or Pet1-lineage neurons. Our study reports dynamic enhancer accessibility in embryonic Pet1 lineage neurons, which we presented as a basis to predict (or raise the possibility) that certain factors exist to remodel postmitotic chromatin and thereby support postmitotic neuron maturation.

d) "We find that Pet1 and Lmx1b control chromatin accessibility to select Pet1-lineage specific enhancers for 5-HT neurotransmission and synaptogenesis" The description of changes in chromatin accessibility is interesting, however, the role of Pet1 and Lmx1b controlling gene expression of 5HT neurotransmission and synaptogenesis is not new.

We certainly agree that we have previously shown using RNAseq and RT-qPCR that Pet1 and Lmx1b control expression of 5-HT neurotransmission genes such as Tph2 and Slc6a4. So yes, we have already shown that these TFs control 5HT neurotransmission gene expression. We note however, that none of our previous papers have focused on gene expression underlying synaptogenesis or synapse functions. What is new in our present work is a deeper mechanistic understanding of how these TFs control gene expression: we show that Pet1 and Lmx1b remodel postmitotic chromatin accessibility to enable the expression of neurotransmission genes and genes needed for synaptogenesis in maturing 5-HT neurons. The identification of TFs that remodeling postmitotic neuronal chromatin is new and this discovery was made possible by our earlier published work at the level of 5HT neurotransmission gene expression.

e) "Required to maintain accessible chromatin in maturing neurons", also interesting however both TFs are already described to be required for gene expression maintenance

Yes, it is true that we have previously reported findings indicating that Pet1 and Lmx1b maintain the expression of 5-HT neurotransmission genes such as Tph2. That is not the point of our current paper. The current study builds on this knowledge to put forth a new and surprising finding: these TFs maintain gene expression by maintaining accessible CREs in postmitotic neurons. The remodeling of accessible chromatin by Pet1and Lmx1b is the most novel and interesting of our findings in our present manuscript and was made possible by our previous studies of these TFs.

Our findings suggest that by controlling chromatin accessibility, these TFs not only directly modulate transcription activation as previously shown, but also impact the global landscape of active CREs and available TF binding sites which, in turn, can alter the global network of regulatory factor inputs. It broadens our understanding of how neuronal identity is maintained in postmitotic neurons (e.g. not just maintenance of transcription activation but also maintenance of CRE landscape).

4. Regarding the mouse model: what is the % of 5HT neurons that are Pet1:YFP negative in each analysed time point? and vice versa, are all Pet1:YFP cells 5HT positive?

The number of 5-HT neurons that are Pet1YFP positive is nearly 100% at all three time point. Therefore the % that are Pet1YFP negative is close to zero or is zero. At 11.5, <5% of Pet1YFP+ cells are 5-HT+, at E14.5 roughly 22% of Pet1YFP+ cells are 5-HT+, and at E17.5, roughly 42% of Pet1YFP+ cells are 5-HT+. In adult 5-HT neurons nearly all Pet1YFP+ cells are 5-HT+. This is consistent with our scRNA-seq findings showing that not all Pet1-lineage neurons are Tph2 positive at embryonic time points.

Our findings are consistent with Alonso et al. (2013) and the Allen Brain Atlas: Tph2 expression is very sparse and limited to a few rostral neurons at E11.5 and then gradually expands throughout Pet1-lineage neurons at E14.5 and E17.5 and into adulthood.

5. Why scATAC approach identifies 60,117 accessible sites not identified in the bulk Pet1? Are those uniquely identified TACs present only in subpopulations of Pet1 positive cells?

Yes, 63% (38043/60117) of the scATAC peaks not identified in the bulk E14.5 WT ATAC are present only in subpopulations of Pet1+ cells. The remainder may not be called peaks in the bulk ATAC due to more stringent peak calling parameters compared to standard scATAC-seq methods.

6. How do the 11 chromatin clusters of Pet1 scATAC correlate with adult scRNAseq data? It would be nice to have an integrated analysis of the embryonic and adult scRNAseq and the scATACseq data (including Okaty et al. and/or Ren et al. reports) to find the embryonic chromatin correlates of the adult 5HT subtypes.

See response to Reviewer 2 first point.

7. Motif enrichment in scATAC clusters (Sup figure 3c). (1) Can the authors speculate on why ETS motif is only enriched in Tph2 cluster and LIM motif in Lrfn5 cluster? (2) could this motif enrichment analysis be combined with cluster specific or enriched TF expression for each family to identify TF candidates for subtype specification? For example, authors mention enriched expression of nkx6.1 for one cluster, which also shows enrichment in NK motifs (this double match TF expression with TF motif is very interesting and should be more clearly mentioned). Similar correlation analysis should be done for other TFs, TF motifs and clusters.

Bulk ATAC-seq results suggest that ETS motifs are highly enriched in general in Pet1-lineage neurons compared to other cell types. Within Pet1 neurons, the scATAC-seq suggests that one subpopulation (“Tph2/Slc6a4” cluster) has slightly more accessible (<2 fold) ETS motifs than the other clusters, but that does not mean that other clusters have no accessible ETS motifs. Similar scenario applies to the LIM homeodomain motifs in the Lrfn5/Nr2f2 cluster. The Tph2/Slc6a4 cluster maps to rostral Pet1 neurons with higher Tph2 and Slc6a4 expression. Rostral 5-HT neurons are known to mature and synthesize 5-HT a day earlier than caudal 5-HT neurons. Pet1::EYFP is also activated earlier in the rostral 5-HT neurons. We would therefore speculate that this cluster is capturing that subpopulation that is advancing the fastest in development, perhaps due to earlier Pet1 activity.

We present Figure 2 —figure supplement 2d as a list of TF candidates for 5-HT neuron subtype specification. In response to the reviewer’s suggestion we mentioned in the text a second TF, Meis2, from the single cell data in addition to Nkx6.1, which the reviewer found to be interesting. As the reviewer mentioned, there is a lot of data in this study, so for space saving reasons we chose to focus mostly on our bulk ATAC-seq analyses, as the TFs we tested, Pet1 and Lmx1b, are mainly responsible for the maturation of shared Pet1 neuron TACs. We show the candidate subtype specific TFs to emphasize that there is further heterogeneity even within Pet1-lineage neurons, which is an interesting topic for future studies.

8. Sup figure 4j: motif enrichment of gained and lost TACs. Are the same motifs present in gained from e11 to e14 than those gained from e14 to e17? and the same question applies for the lost motifs. And as mentioned already, it would be advisable to merge TF expression dynamics with TF motif enrichment dynamics.

Motifs present in gained from E11 to E14 are roughly similar to those gained from E14 to E17, the same for loss TACs. Bioinformatics correlation of motifs to gene expression changes is challenging because many TFs that are known to be active in developing 5-HT neurons belong to the same TF families. For example, Meis1, Meis2, Lmx1b, En1, En2, Pbx2, Hox factors, and POU domain factors are all homeodomain proteins that are expressed in Pet1 neurons. As another example, multiple Rfx-related factors are expressed in developing Pet1 neurons. However, these TFs with very similar binding motifs have different spatiotemporal expression patterns. For instance, Rfx2 and Rfx4 are downregulated from E11.5-E17.5, but Rfx1 is significantly upregulated, whereas Rfx7 and Rfx3 show little change in expression level. Because these TFs within the same or related families bind very similar motif sequences, it is difficult to assign specific genomic sites containing their cognate motif to the activity of a specific TF. Overall, we could detect a significant correlation between TF expression and motif accessibility if all members of the TF family show the same general trend in gene trajectories (e.g. correlation in downregulated Sox2 expression and reduction in Sox family motifs within TACs from E11.5 to E17.5), but when multiple TFs have similar motifs, the results are difficult to interpret.

Additionally, it is possible for TF levels to not be tightly correlated with binding site accessibility level because some TFs may lack chromatin remodeling activity of their own such that its genomic binding is dependent on the priming of the chromatin or recruitment by other TFs and regulatory factors.

9. How does gene expression, gains and losses between wt and Pet1 mutants at different stages correlate with TACs: (1) What is the % of genes downregulated in Pet1 mutant compared to control that shows at least one assigned TAC in wt scATACseq or in bulk atacseq (2) from those, what is the % of downregulated genes in Pet1 mutant with TAC that lose accessibility in at least one assigned TAC? Similar analysis will be interesting for the >1000 genes with gained expression in e11.5 Pet1 mutants reported in Wyler et al. 2016.

65% (499) of Pet1 downregulated at E15.5 in PKO have at least one assigned TAC.

20% (98) of those genes have a TAC that loses accessibility in PKO/PCKO.

73% (734) of Pet1 upregulated at E15.5 in PKO have at least one assigned TAC.

5% (38) of those have a TAC that gains accessibility in PKO/PCKO.

10. Are genes associated with TACs lost in Pet1 mutant expressed in specific clusters of 5HT or broadly expressed in 5ht neurons? are they 5HT specific?

Of the TACs that are differentially accessible in the Pet1-cKO Pet1 neurons, 1432 are shared/broadly expressed among Pet1 neurons and 391 are specific to one cluster. Of the TACs that are differentially accessible in the Pet1-/- Pet1 neurons, 368 are shared and 20 are specific to one cluster.

Of the TACs that are differentially accessible in the Pet1-cKO Pet1 neurons, 33% are “5-HT specific”. Of the TACs that are both differentially accessible in the Pet1-cKO Pet1 neurons and contain Pet1 occupancy, 50% are “5-HT specific”. Of the TACs that are differentially accessible in the Pet1-/- Pet1 neurons, 73% are 5-HT specific. Of the TACs that are both differentially accessible in the Pet1-/- Pet1 neurons and contain Pet1 occupancy, 83% are 5-HT specific (Figure 4f).

11. Could authors speculate why Pet1-/- phenotype is milder than Pet1cKO? (372 TACs less accessible and 3 more accessible compared to 1834 and 391 in cKO), is there any compensatory effect in the regular mutant? How do these 372 TACs and 3 TACs and their respective assigned genes match with DEX genes in Pet1 mutants at e14?

Our hypothesis is described above in response to Rev 2 comment #9.

289 genes are linked to the 372 TACs that show loss in accessibility in Pet1-/- 5-HT neurons. Of these 289 genes, 35 are down-regulated in expression in Pet1-/- 5-HT neurons compared to wildtype neurons at E15.5. One of the 3 TACs that show gain in accessibility in Pet1-/- is a distal bivalent enhancer (based on ChromHMM) linked to the gene Rnf2. Rnf2 is down-regulated in expression in Pet1-/- 5-HT neurons at E15.5.

12. What genes are assigned to the 391 TACs gained in Pet1cKO? do they correlate with gained gene expression in this mutant background? what motifs are enriched in these TACs? what GO terms?

The 391 TACs gained in Pet1-cKO are mapped to 135 genes that are not significantly correlated with gained gene expression in the Pet1-cKO mutant background. These genes are most closely associated with the GO terms “negative regulation of cell volume”, “regulation of axon diameter”, “long-term synaptic potentiation”, and “potassium ion transport”, but none of these GO terms reached statistical significance. The 391 TACs are enriched for Rfx, Gata, HD, and ETS motifs.

13. Most differential TACs in Pet1 lineage are found in distal intergenic or introns, and show enhancer chromatin marks, however CUT and RUN binding for both Pet1 and Lmx1b is found preferentially at proximal regions (promoters). Could authors specify de % of proximal TACs with Pet1 or Lmx1b binding compared to % of distal TACs with binding? Is motif enrichment in TACs different in proximal and distal TACs? What is the mechanistic explanation of this discrepancy between CUT and RUN binding and differential TACs and lost TACs in the mutants?

Pet1 directly binds 59% of proximal promoter Pet1-dependent TACs and only 21% of the distal Pet1-dependent TACs. Lmx1b directly binds 32% of the proximal promoter Lmx1b-dependent TACs and only 9% of distal Lmx1b-dependent TACs.

Proximal and distal Pet1 and Lmx1b genome occupancy sites do show different nearby motif enrichments. Promoter proximal Pet1 occupied TACs are most highly enriched for ETS motifs and C2H2 zinc finger factor motifs, whereas distal Pet1 bound TACs are highly enriched for Rfx and homeodomain motifs. Similarly, promoter proximal Lmx1b bound TACs are enriched for C2H2 zinc finger factor motifs, CREB-related factor motifs, and ETS motifs, whereas distal Lmx1b bound TACs are also enriched for Rfx motifs and homeodomain motifs.

It’s been documented since the initial mapping of DNase-hypersensitivity sites that distal CREs are highly cell type selective compared to TSS-proximal CREs (e.g. Thurman et al., 2012), but the precise reason for this remains poorly understood. One possible mechanistic explanation is that promoter proximal TACs are occupied and more stably maintained by trans factors that sustain chromatin accessibility even in the absence of cell type specific pioneer TFs. We do not currently have enough evidence to draw more specific conclusions on what these factors are based on our data, as TF motif sites can be opened by chromatin remodeling factors other than the cognate TF itself.

14. Donovan et al. showed Lmx1b is upstream Pet1 expression, at least in some 5HT neurons, Is the effect on chromatin accessibility of Lmx1b mutant mediated, at least in part, by the loss of Pet1? How many lost TACs in Lmx1b are bound by Pet1 but not by Lmx1b?

Pet1 expression is reduced to 52% of wildtype level in the Lmx1b-cKO Pet1 neurons based on our RNA-seq analyses at E17.5. We added a panel into Figure 6 —figure supplement 1 to clarify this. However, the reduction in Pet1 expression cannot completely account for the dramatic changes in chromatin accessibility seen in the Lmx1b-cKO neurons because 1) significant residual Pet1 expression remains 2) Lmx1b-cKO neurons show 2823 differentially accessible TACs relative to wildtype neurons that were not observed in the Pet1-cKO (Figure 6 – supplemental Figure 1), including TACs associated with the 5-HT neurotransmission genes Gchfr, Slc18a2, and Maoa.

Pet1 binds to 23% (699/3102) of LKO down TACs that are not bound by Lmx1b.

15. I would remove the last section of the manuscript, as I don´t think the data support a role in synaptogenesis: Authors mention vglut3 and syt1 as panneuronal genes, however they are neuronal effector genes not expressed in all neurons but in specific neurons subtypes. Moreover, they are not involved in synaptogenesis but in synaptic function and neurotransmission. Finally, the analysis of pericellular baskets does not make much sense as these mutants show defects in axon projections to start with.

Our reasoning is that for synaptogenesis to occur many effector genes must be activated to express presynaptic machinery e.g. Syt1. If this is prevented functional synapse formation is impaired. We showed that loss of Pet1 or Lmx1b impairs Syt1 chromatin accessibility and gene expression. Our interpretation is that these terminal selectors are needed to reorganize postmitotic chromatin to enable expression of genes needed for the formation of presynaptic machinery, hence synaptogenesis, but we agree with the reviewer that this wording choice can be confusing and have revised it to “synapse function” in the text. We would like to point out that while most axon projections to the dorsolateral septum fail to form some make it there. These successfully projections are the ones we investigated to determine whether they form pericellular baskets.

https://doi.org/10.7554/eLife.75970.sa2

Article and author information

Author details

Xinrui L Zhang

Department of Neurosciences, Case Western Reserve University, Cleveland, United States

Contribution
Conceptualization, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing - original draft, Writing - review and editing

Contributed equally with
William C Spencer

Competing interests
No competing interests declared

"This ORCID iD identifies the author of this article:" 0000-0002-7728-6970
William C Spencer

Department of Neurosciences, Case Western Reserve University, Cleveland, United States

Contribution
Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing - original draft, Writing - review and editing

Contributed equally with
Xinrui L Zhang

Competing interests
No competing interests declared

"This ORCID iD identifies the author of this article:" 0000-0002-9700-8011
Nobuko Tabuchi

Department of Neurosciences, Case Western Reserve University, Cleveland, United States

Contribution
Investigation, Methodology, Validation, Visualization, Writing - review and editing

Competing interests
No competing interests declared
Meagan M Kitt

Department of Neurosciences, Case Western Reserve University, Cleveland, United States

Contribution
Conceptualization, Writing - review and editing

Competing interests
No competing interests declared
Evan S Deneris

Department of Neurosciences, Case Western Reserve University, Cleveland, United States

Contribution
Conceptualization, Funding acquisition, Supervision, Writing - original draft, Writing - review and editing

For correspondence
esd@case.edu

Competing interests
No competing interests declared

"This ORCID iD identifies the author of this article:" 0000-0003-4211-9934

Funding

National Institute of Mental Health (R01 MH117643)

Evan S Deneris

National Institute of Mental Health (R01 MH062723)

Evan S Deneris

National Institute of Mental Health (F30 MH122173)

Xinrui L Zhang

National Institute of General Medical Sciences (T32 GM007250)

Xinrui L Zhang

Uehara Memorial Foundation (201940009)

Nobuko Tabuchi

Japan Society for the Promotion of Science (Overseas Research Fellowship)

Nobuko Tabuchi

National Institute of General Medical Sciences (T32 GM008056)

Meagan M Kitt

National Institute of Mental Health (R01 MH125918)

Evan S Deneris
William C Spencer

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

This research was supported by NIH grants R01 MH117643 and R01 MH062723 to ESD, and R01 MH125918 to ESD and WCS (co-I), F30 MH122173, and T32 GM007250 to XLZ, and the Uehara Memorial Foundation Fellowship (201940009) and Japan Society for the Promotion of Science Overseas Research Fellowship to NT, and T32GM008056 to MMK. We thank Katherine Lobur for assistance with mouse breeding and Laura Marsland for assisting with immunohistochemistry analysis of Tph2 in embryonic mouse hindbrain. We thank Dr Polyxeni Philippidou for helpful comments on the manuscript. We thank Dr Carmen Birchmeier for the anti-Lmx1b antibody. This work was supported by the Cytometry and Microscopy Shared Resource at Case Comprehensive Cancer Center, the CWRU Genomics Core, and the CWRU Applied Functional Genomics Core.

Ethics

All animal procedures used in this study were in strict accordance with the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was approved by the Case Western Reserve University School of Medicine Institutional Animal Care and Use Committee (Animal Welfare Assurance Number A3145-01, protocol #: 2014-0044).

Senior Editor

Catherine Dulac, Harvard University, United States

Reviewing Editor

Paschalis Kratsios, University of Chicago, United States

Reviewer

Paschalis Kratsios, University of Chicago, United States

Publication history

Preprint posted: October 14, 2021 (view preprint)
Received: November 29, 2021
Accepted: April 12, 2022
Accepted Manuscript published: April 26, 2022 (version 1)
Version of Record published: May 12, 2022 (version 2)

Copyright

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.