Histone deacetylase 1 maintains lineage integrity through histone acetylome refinement during early embryogenesis

  1. Jeff Jiajing Zhou
  2. Jin Sun Cho
  3. Han Han
  4. Ira L Blitz
  5. Wenqi Wang
  6. Ken WY Cho  Is a corresponding author
  1. Department of Developmental and Cell Biology, University of California, Irvine, United States
  2. Center for Complex Biological Systems, University of California, Irvine, United States

Abstract

Histone acetylation is a pivotal epigenetic modification that controls chromatin structure and regulates gene expression. It plays an essential role in modulating zygotic transcription and cell lineage specification of developing embryos. While the outcomes of many inductive signals have been described to require enzymatic activities of histone acetyltransferases and deacetylases (HDACs), the mechanisms by which HDACs confine the utilization of the zygotic genome remain to be elucidated. Here, we show that histone deacetylase 1 (Hdac1) progressively binds to the zygotic genome from mid-blastula and onward. The recruitment of Hdac1 to the genome at blastula is instructed maternally. Cis-regulatory modules (CRMs) bound by Hdac1 possess epigenetic signatures underlying distinct functions. We highlight a dual function model of Hdac1 where Hdac1 not only represses gene expression by sustaining a histone hypoacetylation state on inactive chromatin, but also maintains gene expression through participating in dynamic histone acetylation–deacetylation cycles on active chromatin. As a result, Hdac1 maintains differential histone acetylation states of bound CRMs between different germ layers and reinforces the transcriptional program underlying cell lineage identities, both in time and space. Taken together, our study reveals a comprehensive role for Hdac1 during early vertebrate embryogenesis.

Editor's evaluation

In this revised and important study, the authors investigate the roles of histone deacetylases in the spatial epigenetic regulation of zygotic gene expression during embryo gastrulation. They provide convincing evidence for HDAC1 binding to genes around the timing of large-scale genome activation, and that inhibition of histone acetylation blocks gastrulation, blurring cell lineage integrity, tied to both positive and negative regulatory effects on transcription in space and time. The research reveals new insight on the role of histone acetylation-deacetylation in dynamics in epigenetic control of gene expression and cell fate determination during early tissue patterning in embryogenesis.

https://doi.org/10.7554/eLife.79380.sa0

Introduction

A fundamental question in early development is the mechanism of zygotic genome activation (ZGA), which requires the degradation of maternal mRNAs and the activation of embryonic transcription (Blitz and Cho, 2021, Tadros and Lipshitz, 2009). During ZGA, the embryonic genome undergoes a dramatic reprogramming of gene expression, which is also accompanied by remodeling of the embryonic epigenome. Post-translational modifications to histones are a major epigenetic regulation influencing chromatin structure and thus play a central role in ZGA. Histone acetylation appears during the onset of both minor and major ZGA waves in many species. In Drosophila, histone acetylation occurs at mitotic cycle 8 on a few early zygotic genes (Li et al., 2014). miR430, the first zygotically active gene, is marked by H3K27ac in 64-cell staged zebrafish embryos (Chan et al., 2019). Genome-wide H3K27ac is detected at mid-blastula shortly after the onset of ZGA in Xenopus (Gupta et al., 2014). In mice, the zygotic genome is increasingly marked by H3K27ac from immature and metaphase II oocytes to 2-cell-stage embryos (Dahl et al., 2016). Despite these findings, there remain several major questions. How is the interplay of enzymes regulating histone acetylation deployed in developing embryos? What is the role of observed histone acetylation on gene expression? How are the spatial and temporal patterns of histone acetylation established during ZGA?

Histone acetylation occurs on the ε-amino group of the lysine residues within N-terminal tails of all four core histones (Inoue and Fujimoto, 1969; Seto and Yoshida, 2014). Acetylation is a reversible process that is directly catalyzed by opposing activities of histone acetyltransferases (HATs) and histone deacetylases (HDACs). In addition, HATs and HDACs can also regulate the acetylation of lysine residues on non-histone proteins (Choudhary et al., 2009). Histone acetylation is often associated with active gene transcription because the acyl groups neutralize the positive charge on the lysine residues, thereby reducing the affinity of histones to DNA (Wang et al., 2000; Anderson et al., 2001); it also serves as a binding platform for bromodomain (BRD) proteins which scaffold and stimulate the transcriptional machinery (Hassan et al., 2007; Filippakopoulos et al., 2012). The balance between HATs and HDACs directly shapes histone acetylation landscapes and subsequently affects transcriptomes.

HDACs are critical epigenetic regulators because they reset chromatin states by returning acetylated lysine residues on histones to the basal state, which can subsequently be subjected to alternative modifications such as methylation. HDACs are grouped into four classes based on phylogenetic conservation. Class I (HDAC1, 2, 3, 8), Class II (HDAC4, 5, 6, 7, 9, 10), and Class IV (HDAC11) HDACs are zinc dependent and are related to yeast Rpd3, Had1, and Hos3, respectively; Class III (SIRT1, 2, 3, 4, 5, 6, 7) HDACs, also known as Sirtuins, are NAD+ dependent and are related to yeast Sir2 (Gregoretti et al., 2004; Milazzo et al., 2020). HDACs are well-characterized negative regulators of gene expression during development. For example, Hdac1 silences homeotic genes in cooperation with Polycomb group repressors in Drosophila (Chang et al., 2001). In zebrafish, Hdac1 represses Notch targets during neurogenesis (Cunliffe, 2004; Yamaguchi et al., 2005). In Xenopus, HDAC activity suppresses Vegt-induced ectopic mesoderm in ectoderm lineages (Gao et al., 2016), represses multi-lineage marker genes at blastula (Rao and LaBonne, 2018), and desensitizes dorsal Wnt signaling at late blastula (Esmaeili et al., 2020). Conversely, HDACs can also positively regulate gene expression. For instance, inhibition of HDAC activities rapidly down-regulates some genes in yeast, suggesting an activator function of HDACs (Bernstein et al., 2000). Genetic deletions or pharmacological application of HDAC inhibitors in cell lines results in both up- and down-regulation of genes (Reid et al., 2005; Zupkovitz et al., 2006; Meganathan et al., 2015). Furthermore, genome-wide studies showed that HDACs occupy genomic loci of active genes, and their binding correlates with gene activities (Kurdistani et al., 2002; Wang et al., 2002; Wang et al., 2009; Kidder and Palmer, 2012). These seemingly opposing functions of HDACs raise an important question as to the exact roles of HDACs on chromatin states and transcriptomes in developing embryos.

In this study, we focus on the role of Hdac1 in regulating the zygotic epigenome and transcriptome during Xenopus germ-layer specification coinciding with ZGA. Current evidence in Xenopus as well as in other non-mammalian systems suggests that the early embryonic genome is rather naive and that major chromatin modifications occur at or after ZGA (Bonn et al., 2012; Vastenhouw et al., 2010; Gupta et al., 2014; van Heeringen et al., 2014; Hontelez et al., 2015). Thus, the system allows us to probe the earliest establishment of histone acetylation and dissect the link between actions of Hdac1, the zygotic histone acetylome, and zygotic transcriptome during the first cell lineage segregation event. Here, we report that the major Hdac1 binding to the embryonic genome occurs at blastula; the binding of Hdac1 during this stage is dependent on maternal factors. We highlight a dual function model for Hdac1. First, Hdac1 keeps inactive chromatin free of histone acetylation, preventing gene misactivation in respective germ layers. Second, Hdac1 participates in dynamic histone acetylation–deacetylation cycles on active chromatin, sustaining the expression of genes that are enriched in respective germ layers. Taken together, our study reveals a coordinated spatial and temporal regulation by Hdac1 during ZGA.

Results

Hdac1 binds to the genome progressively during blastula and onward

To identify the major functional candidates of HDACs during the early Xenopus embryogenesis, we examined the temporal RNA expression profiles (Owens et al., 2016) of HDAC family members (Class I, II, and III HDACs) from the zygote to the beginning of the neurula stage in Xenopus tropicalis. The RNA expression level of hdac1 is the highest among all HDACs examined, followed by hdac2 (Figure 1—figure supplement 1A). Both Hdac1 and Hdac2 proteins are present in the unfertilized egg to the mid gastrula stage, and the overall expression levels of Hdac1 and Hdac2 are relatively constant during this period (Figure 1A). These data reveal that Hdac1/2 are the major maternally endowed HDACs functioning during this window of development.

Figure 1 with 1 supplement see all
Hdac1 binds to the genome gradually during early Xenopus development.

(A) Western blot analyses showing Hdac1 and Hdac2 proteins over a time course of early development. β-Tubulin is used as a loading control. (B) Venn diagram comparing Hdac1 irreproducibility discovery rate (IDR) peaks among examined stages. The sums of peaks in st8 and st9 are smaller due to instances where more than one peak from st8 or st9 overlaps the same single st10.5 peak. (C) Clustered heatmap showing Hdac1 ChIP-seq signals at each stage over a window of 5 kb centered on the summit of all Hdac1 IDR peaks in descending order. (D) The expression levels (TPM, transcripts per million) of genes associated with Hdac1 clusters at different developmental periods. (E) Distributions of Hdac1 IDR peaks at each stage across five defined genomic features. The promoter is defined as −1 kb to +100 bp from TSS (transcription start site) while the TTS (transcription termination site) is defined as −100 bp to +1 kb from TTS. (F) Distributions of Hdac1 ChIP-seq signals within the intervals of 5 kb upstream of gene model 5′ ends, gene bodies (normalized for length), and 5 kb downstream of gene model 3′ ends at each stage. The signal of st9 input DNA ChIP-seq is used as a negative control. Y-axis values represent reads quantified by bins per million (BPM) at a bin size of 50 bp. (G) ChIP-qPCR showing Hdac1 enrichment on selected genomic regions (nine positive regions: alkbh2, fgf5, foxi4.2, gdnf, hhex, miR428a, not, snai1, and sp8; two negative regions: hspa4 and klf11) at indicated stages of early embryogenesis.

Figure 1—source data 1

Western blot analyses showing Hdac1 and Hdac2 proteins during Xenopus embryogenesis.

https://cdn.elifesciences.org/articles/79380/elife-79380-fig1-data1-v2.zip

Since Hdac1 modulates various aspects of transcriptional regulation and the chromatin landscape, we examined genome-wide Hdac1-binding patterns during early germ-layer development (Figure 1—figure supplement 1B) using chromatin immunoprecipitation (ChIP) assays followed by sequencing (ChIP-seq). A set of high-confidence peaks at each stage were obtained using irreproducibility discovery rate (IDR) analysis (Li et al., 2011) from two biologically independent samples (Figure 1—figure supplement 1C). Hdac1 binds to 1340 regions at the mid-blastula (st8), 5136 regions at the late blastula (st9), and 22,681 regions at the early gastrula (st10.5) stages (Figure 1B). Overall, a minority of Hdac1 peaks are unique to each of the blastula stages (Clusters a and b) and a majority of Hdac1 peaks are present across multiple stages (Cluster c) and at the early gastrula stage (Cluster d) (Figure 1C, Figure 1—figure supplement 1D). The expression levels of zygotic genes associated with Cluster c Hdac1 peaks are higher than those associated with Clusters a, b, and d, suggesting that the genes associated with sustained Hdac1 binding are more active during gastrulation (Figure 1D). Similarly, Hdac2 gradually accumulates on the embryonic genome (Figure 1—figure supplement 1E). Interestingly, 99% of Hdac1 peaks at stage 9 and 97% of Hdac1 peaks at stage 10.5 overlap with Hdac2 peaks (Figure 1—figure supplement 1F, G). When the levels of Hdac2 peaks associated with Hdac1 were compared to Hdac2 peaks , Hdac1 peaks display a significantly higher Hdac2 peak enrichment (Figure 1—figure supplement 1H). Together, we revealed that Hdac1/2 are progressively directed to the genome during early development. Since a majority of Hdac1-bound peaks are similarly bound by Hdac2, we focus on Hdac1 peaks in subsequent analyses.

To investigate the differences of Hdac1 genomic occupancy across early stages, we examined various genomic features of regions bound by Hdac1 at each stage. The majority of Hdac1 peaks are found in either intergenic or intronic regions; a minor fraction of Hdac1 peaks is present within exons or transcriptional termination sites. A notable observation is the increased Hdac1 binding to more promoter regions over developmental times (Figure 1E). We then analyzed the distribution of Hdac1 across three stages along the gene bodies of genes bound by Hdac1. A higher enrichment of Hdac1 binding is located near the promoters of genes as development proceeds (Figure 1F). Importantly, the timing of Hdac1-binding accumulation coincides with the mid-blastula stage (st8) (Figure 1G), suggesting that Hdac1 is involved in the epigenetic regulation of these genes during ZGA.

Blastula Hdac1 binding is maternally instructed

Since Hdac1 binds progressively to the genome during ZGA, we investigated the contribution of zygotic factors to the recruitment of Hdac1 at the late blastula stage. We injected α-amanitin, which blocks both transcription initiation and elongation (Chafin et al., 1995), to block embryonic transcription (Figure 2—figure supplement 1A). A high Pearson correlation between α-amanitin-injected and uninjected embryos is observed genome-wide or at all Hdac1 IDR peaks (Figure 2A), which is higher than the Pearson correlation representing the batch effect when compared to a different batch of same staged embryos (WT) (Figure 2—figure supplement 1B). There is no significant signal difference between α-amanitin-injected and uninjected embryos at all Hdac1 IDR peaks (Figure 2—figure supplement 1C, H). We conclude that Hdac1 binding is independent of zygotic transcription at the late blastula stage, suggesting the importance of maternal factors in Hdac1 recruitment.

Figure 2 with 1 supplement see all
Maternal factors instruct Hdac1 recruitment during blastula stages.

(A) Pairwise Pearson correlation analyses comparing ChIP-seq signals of st9 Hdac1 irreproducibility discovery rate (IDR) peaks between α-amanitin-injected and uninjected embryos across the genome (left) and among Hdac1 IDR peaks (right). (B) Motif analyses of st9 Hdac1 peaks (500 bp centered on IDR peak summit). Motif sequence to the corresponding factor is retrieved from JASPAR. Hey1 is an example of TF motif with low significance. (C) Clustered heatmap depicting st9 Foxh1 and Sox3 ChIP-seq signals in a window of 10 kb centered on st9 Hdac1 IDR peaks with descending order. (D) ChIP-seq signal enrichment of Hdac1, Foxh1, and Sox3 within the intervals of 5 kb upstream of gene model 5′ ends, gene bodies (normalized for length), and 5 kb downstream of gene model 3′ ends of st9 Hdac1 associated genes. The signal of st9 input DNA ChIP-seq is used as a negative control. Y-axis values represent reads quantified by bins per million (BPM) at a bin size of 50 bp. (E) St9 sequential ChIP-qPCR analyses for Foxh1 and Hdac1 co-bound regions and Sox3 and Hdac1 co-bound regions. anti-HA is used as a negative control. The error bars represent the variation from two technical replicates. (F) ChIP-qPCR analysis of Hdac1 peaks that are also Foxh1/Sox3 co-bound after Foxh1 and/or Sox3 depletion. Genomic loci associated with klf11 (no Hdac1 signals), erfl and insm2 (Hdac1 signals without Foxh1 or Sox3 signals) are negative controls. The error bars represent the variation from two technical replicates. Student’s t-test is used to calculate p-values over Hdac1 enrichment of uninjected embryos. * is for p< 0.05** is for p< 0.01.

To identify maternal factors that facilitate the recruitment of Hdac1 to the genome, we performed a de novo motif search of the DNA sequences of 5136 st9 Hdac1peaks. Sox, Foxh1, and Pou motifs are found to be the most frequent maternal TF motifs (Figure 2B, Supplementary file 1). We thus compared the genomic binding profiles of Hdac1 to two maternal TFs, Foxh1 (Charney et al., 2017) and Sox3. A majority of Hdac1-bound regions (Cluster 1) overlaps with both Foxh1- and Sox3-bound regions while only a small fraction of Hdac1-bound regions (Cluster 2) overlaps the binding of either Foxh1, or Sox3, or neither (Figure 2C). More than 80% of Hdac1 peaks overlap with Foxh1 or Sox3 peaks (Figure 2—figure supplement 1D, E). A positive correlation between Hdac1 binding and Foxh1/Sox3 binding is observed at all Hdac1 IDR peaks (Figure 2—figure supplement 1F, G). We noted frequent overlapping binding of Hdac1 with Foxh1/Sox3 (Figure 2—figure supplement 1I, J), and highly enriched signals of Hdac1, Foxh1, and Sox3 present around promoters of genes (Figure 2D). All these observations suggest a role for Foxh1 and Sox3 in Hdac1 recruitment. We confirmed the co-occupancy of Hdac1 with each of Foxh1 and Sox3 TFs on the same DNA molecules using sequential ChIP-qPCR (Figure 2E, Figure 2—figure supplement 1K, L). Depletion of Foxh1, Sox3, or both TFs by morpholino injections showed a reduced binding of Hdac1 around Foxh1/Sox3 co-occupied genomic regions (Figure 2F). Hence, we propose that Foxh1 and Sox3 maternal TFs facilitate Hdac1 recruitment during ZGA.

Hdac1 binds to genomic regions with distinct epigenetic signatures

To further characterize regions bound by Hdac1 across early germ-layer development, we examined epigenetic signatures (Gupta et al., 2014; Hontelez et al., 2015; Charney et al., 2017) on Hdac1 peaks across various stages. Ep300 binding (a HAT), which catalyzes the acetylation of histone, is observed in Hdac1 peaks (Clusters b–d) from the late blastula and onward where RNA polymerase II signals also emerge (Figure 3—figure supplement 1A). This indicates that Hdac1 and Ep300 share similar binding profiles on many transcriptionally active genes. We next surveyed several histone acetylation modifications in regions bound by Hdac1 (Figure 3A). Consistent with the overlapping binding of Ep300, Hdac1 peaks display signals of H3K9ac (Clusters c and d), H3K18ac (Clusters b–d), H3K27ac (Clusters b–d), and pan-H3 lysine acetylation (pan-H3Kac) (Clusters b–d). We then examined several histone methylation modifications that are associated with gene activation (Figure 3—figure supplement 1B). H3K4me1, a primed enhancer mark (Creyghton et al., 2010), and H3K4me3, an active promoter mark (Heintzman et al., 2007), display signals at Hdac1-bound regions (Clusters c and d); H3K36me3, a transcription elongation mark (Kolasinska-Zwierz et al., 2009), displays minimal signals at any Hdac1-bound regions. These observations reveal that Hdac1 binds to genomic regions with active epigenetic signatures.

Figure 3 with 1 supplement see all
Hdac1 binds to cis-regulatory modules (CRMs) containing functionally distinct histone modifications.

Clustered heatmaps showing signals from several stages of various (A) histone acetylation marks and (B) repressive histone methylation marks on Hdac1 peaks. Each cluster corresponds to the same regions present in Figure 1C. (C) Venn diagram illustrating Hdac1 peaks overlapping with H3K27me3 and H3K27ac peaks from both st9 and st10.5 combined. (D) Clustered heatmaps depicting signals of H3K27me3 and H3K27ac on combined st9 and st10.5 Hdac1 peaks. Clusters denote the same genomic regions in C. Numbers on the right side indicate the total number of regions in each cluster. The signals are shown in a window of 5 kb centered on the summits of Hdac1 peaks presented in descending order of track signal intensities within each cluster.

Hdac1 resets the state of acetylated histones by removing acetyl groups, and thus can facilitate the formation of repressive chromatin. Here, we analyzed several histone methylation modifications associated with gene repression (Figure 3B). Hdac1 binds to genomic regions mostly free of H3K9me2, H3K9me3, or H4K20me3. All three modifications are known to mark constitutive heterochromatin denoting gene-poor areas consisting of tandem repeats (Richards and Elgin, 2002). In addition, a fraction of Hdac1-bound regions (Clusters c and d) display signals of H3K27me3 (Figure 3B), which is known to mark facultative heterochromatin consisting of developmental-cue silenced genes (Trojer and Reinberg, 2007). These observations suggest that Hdac1 binds to facultative heterochromatic regions facilitating the repression of genes.

Since the majority of Hdac1 peaks (Clusters b–d) are marked by both active and repressive epigenetic signatures, we wonder how Hdac1 functions in epigenetically distinct genomic loci. We compared peaks between two functionally opposing histone modifications H3K27ac and H3K27me3 to Hdac1 peaks, then subdivided these Hdac1 peaks into four clusters representing functionally distinct CRM types (Figure 3C, Figure 3—figure supplement 1C). Cluster I denotes 3548 Hdac1 peaks marked by both H3K27ac and H3K27me3 (Figure 3D). Given that both H3K27ac and H3K27me3 are modified on the same lysine residue, we speculate that these regions are differentially marked in space due to heterogeneous cell populations present in the whole embryo. Therefore, Hdac1 Cluster I peaks are referred to as heterogeneous CRMs. Cluster II denotes 1389 Hdac1 peaks marked by only H3K27me3 indicating that these regions are associated with inactive developmental genes (Figure 3D). Hdac1 Cluster II peaks represent repressive CRMs. Cluster III denotes 13,669 Hdac1 peaks marked by only H3K27ac, suggesting that these are active CRMs. Cluster IV denotes 4836 Hdac1 peaks with neither H3K27ac nor H3K27me3 modifications (Figure 3D). At genomic loci marked with two distinct H3K27 modifications (Figure 3—figure supplement 1D), the expression levels of genes bound by Hdac1 (I, II, and III) are generally higher than that of genes unbound (I’, II’, and III’) (Figure 3—figure supplement 1E), suggesting that Hdac1 binding correlates with transcriptional activity of the genes. Together, we show that Hdac1-bound CRMs are subject to distinct epigenetic modifications, which confer differential CRM activities.

HDAC activity is required for differential germ-layer histone acetylomes

A major function of HDACs is to catalyze the removal of acetyl groups from histones. We hypothesize that Hdac1 differentially regulates histone acetylation of four different Hdac1 CRM Clusters (Clusters I–IV in Figure 3C). To test this hypothesis, we treated embryos continuously with a widely used pan-HDAC inhibitor, Trichostatin A (TSA) (Yoshida et al., 1990) beginning at the 4-cell stage and followed the development up to tailbud stages. Embryos treated with TSA are developmentally arrested at gastrula (Figure 4A). We observed the presence of dorsal blastopore lip, suggesting that the progression but not the initiation of gastrulation is defective. A Class I HDAC inhibitor, valproic acid (VPA) (Göttlicher et al., 2001), also produces a similar phenotype (Figure 4—figure supplement 1A). We first showed that protein levels of Hdac1 and Hdac2 are not affected by TSA treatment (Figure 4—figure supplement 1B). Next, Hdac1 ChIP-seq experiments display high Pearson correlations on embryos treated with solvent control or TSA at st9 (Figure 4—figure supplement 1C) and st10.5 (Figure 4—figure supplement 1D), which is further supported by higher Pearson correlations upon TSA treatment than the batch effect (Figure 4—figure supplement 1E, F). Lastly, differential peak analysis on Hdac1 IDR peaks in embryos treated with solvent control or TSA showed barely any differential Hdac1 signals (0.1% peaks or less) at st9 (Figure 4—figure supplement 1G) and st10.5 (Figure 4—figure supplement 1H). These results suggest that TSA treatment of early Xenopus embryos does not alter the recruitment of Hdac1 to the genome. To examine the efficacy of HDAC activity inhibition by TSA, we surveyed six well-known histone acetylation modifications by western blot. Drastically increased levels of H3K9ac, H3K18ac, and H3K27ac are observed in TSA-treated embryos (Figure 4B; Rao and LaBonne, 2018), whereas H3K14ac, H3K56ac, and H4K20ac are not detected during this stage of development. These data indicate that HDAC activity is required to maintain the proper level of histone acetylation during gastrulation.

Figure 4 with 3 supplements see all
Hdac1 maintains differential H3 acetylomes between germ layers.

(A) Embryos treated with 100 nM Trichostatin A (TSA) displaying gastrulation defects 24 hr post-fertilization. Asterisk denotes the dorsal side containing the early blastopore lip. (B) Western blot analyses showing various histone acetylation modifications affected by HDAC inhibition. Anti-H3 is used as a loading control. (C) Clustered heatmap depicting signals of pan-H3Kac at st10.5 in whole embryos (WE), animal cap (AC), and vegetal mass (VG) cells. The signals are shown in a window of 5 kb centered on the summits of combined AC and VG peaks presented in descending order within each cluster. (D) Stacked bar graph representing proportions of localized (LC) versus non-localized (NL) pan-H3Kac signals found at Hdac1 cis-regulatory module (CRM) clusters in Figure 3C. A Hdac1 peak is considered to exhibit localized pan-H3Kac if it overlaps with either AC- or VG-specific pan-H3Kac peaks (Cluster B or C in C); a Hdac1 peak is considered to exhibit non-localized pan-H3Kac if it overlaps with pan-H3Kac peaks shared between AC and VG (Cluster A in C). NP: not overlap with any pan-H3Kac peak. (E) Spike-in normalized pan-H3Kac signals across Hdac1 CRM clusters (clusters in Figure 3C) in Dimethy Sulfoxide (DMSO)- or TSA-treated AC explants. Δ¯ represents the log2 scaled average differences of spike-in normalized pan-H3Kac signals between DMSO- and TSA-treated AC explants. Randomized genomic regions (n = 23,442) are used as the negative control. (F) Fold changes (FC) of pan-H3Kac signals at Hdac1 CRM clusters (clusters in Figure 3C) in DMSO- or TSA-treated AC explants. Red dotted line denotes the level of zero. *** denotes p < 0.001 (Student’s t-test). (G) Fold changes (FC) of pan-H3Kac signals in Cluster I of Hdac1 CRM clusters (clusters in Figure 3C) for each spatial CRM category. d denotes effect size calculated by Cohen’s d.

Figure 4—source data 1

Western blot analyses showing various histone acetylation modifications after HDAC inhibition.

https://cdn.elifesciences.org/articles/79380/elife-79380-fig4-data1-v2.zip

Given that Hdac1 CRM Clusters (Clusters I–IV in Figure 3C) are subjected to both active and repressive epigenetic modifications presumably in different germ layers, we compared the general status of the H3 acetylome (pan-H3Kac) between two distinct germ layers, the animal cap (AC, presumptive ectoderm) and the vegetal mass (VG, presumptive endoderm). We performed pan-H3Kac ChIP-seq at early gastrula (st10.5) on dissected AC and VG explants from embryos treated with either TSA or solvent control (Figure 4—figure supplement 2A). A comparison of high-confidence peaks from each explant reveals that a majority (~75%) of pan-H3Kac are specifically marked in ectodermal (~37%) and endodermal (~38%) germ layers (Figure 4C, Figure 4—figure supplement 2B). We observed the signal intensity of pan-H3Kac in AC/VG shared genomic regions (Cluster A in Figure 4C) is significantly higher than those of AC- or VG-specific genomic regions (Clusters B and C in Figure 4C), which also coincides with higher gene expression levels (Figure 4—figure supplement 2C). To correlate the differential pan-H3Kac states and the differential gene expression profiles between the two germ layers, we assigned high enrichment pan-H3Kac peaks (~top 30%) to the nearest genes within 10 kb and compared the expression levels of these genes in each germ layer (Blitz et al., 2017). Indeed, the expression level of genes enriched with AC-specific pan-H3Kac are higher in AC than VG (Figure 4—figure supplement 2D), and the expression levels of genes enriched with VG-specific pan-H3Kac are higher in VG than AC (Figure 4—figure supplement 2E). Well-known genes with germ-layer-specific expression exhibit localized pan-H3Kac signals between germ layers (Figure 4—figure supplement 2F). These results illustrate that the histone acetylation profile generally coincides with the animally and vegetally localized expression of transcripts.

To uncover the role of Hdac1 in regulating histone acetylation states of CRM clusters (Clusters I–IV in Figure 3D), we first examined the general distribution of pan-H3Kac in these Hdac1 CRM Clusters. Consistent with H3K27 modifications, both heterogeneous CRMs (Cluster I, both H3K27ac and H3K27me3) and active CRMs (Cluster III, only H3K27ac), but not repressive CRMs (Cluster II, only H3K27me3), are marked by pan-H3Kac. Nearly half of pan-H3Kac marks on heterogeneous or active CRMs are localized either animally or vegetally (Figure 4C, D), suggesting that these CRMs are regionally active in specific germ layers. We then quantitatively (Egan et al., 2016) compared the levels of pan-H3Kac signals (read density) on each CRM within Hdac1 CRM Clusters I–IV with and without TSA treatment. A global increase of pan-H3Kac signals across all Hdac1 CRM Clusters is observed after TSA treatment (Figure 4E, Figure 4—figure supplement 3A), which is consistent with the western blot data (Figure 4B). TSA-induced HDAC inhibition leads to elevated pan-H3Kac signals on both readily acetylated genomic loci (Cluster α of Figure 4—figure supplement 3B, C) and other genomic regions (Cluster β of Figure 4—figure supplement 3B, C). Interestingly, we found that CRMs of Hdac1 Clusters I–IV respond differently upon HDAC inhibition: repressive CRMs (Cluster II, only H3K27me3) show the highest fold increase of pan-H3Kac signals, while active CRMs (Cluster III, only H3K27ac) show the lowest fold increase of pan-H3Kac signals when compared to other clusters (Figure 4F, Figure 4—figure supplement 3D). We also note that the increased amount of pan-H3Kac is very similar across different Hdac1 clusters, irrespective of CRMs being repressive or active CRMs (Figure 4—figure supplement 3E, F). This suggests that HDACs catalytic activities are similar whether CRMs are repressive or active. In sum, upon HDAC inhibition, Hdac1-bound repressive CRMs are subject to histone hyperacetylation, while Hdac1-bound active CRMs exhibit a further increase of histone acetylation.

Lastly, we explored how Hdac1 CRM Clusters (Clusters I–IV in Figure 3C) differentially respond to HDAC inhibition in specific germ layers. For heterogeneous CRMs (Cluster I, both H3K27ac and H3K27me3), the fold increase of pan-H3Kac signals after TSA treatment is examined (Figure 4—figure supplement 3G). Interestingly, CRMs with low levels of histone acetylation tend to be more responsive to HDAC inhibition, than CRMs with high levels of histone acetylation. Next, heterogeneous CRMs were subdivided into three spatial categories, that are pan-H3Kac enriched animally (AC CRMs), vegetally (VG CRMs), and ubiquitous CRMs. The fold changes of pan-H3Kac signals of these CRMs were examined after TSA treatment. Germ-layer-specific AC and VG CRMs show the greater changes upon HDAC inhibition (Figure 4G). AC CRMs show an increase in pan-H3Kac signals in VG but not in AC after HDAC inhibition. Similarly, VG CRMs acquire an increase of pan-H3Kac signals in AC but not in VG upon HDAC inhibition. A similar trend is also observed on active CRMs (Cluster III, only H3K27ac), emphasizing germ-layer-specific functions of this CRM cluster (Figure 4—figure supplement 3H, I). Taken together, these data demonstrate that Hdac1 activity is spatially regulated in development to maintain proper germ layer specific gene expression patterns.

HDAC activity modulates developmental genes between germ layers

HDACs are considered as transcriptional corepressors because histone deacetylation is generally associated with transcriptional repression. To determine how Hdac1-bound CRM Clusters I–III (Figure 3C) influence the activities of their corresponding genes, each CRM within a cluster was assigned to a nearest gene located within 10 kb and genes were placed into one of three classes (Figure 5A). Class 1 denotes 3104 genes that have CRMs with mixed marks of H3K27ac and/or H3K27me3 suggesting that these genes are differentially expressed in different germ layers. Class 2 denotes 629 genes whose CRMs are marked with only H3K27me3 indicating that these genes may be repressed. Class 3 denotes 5913 genes whose CRMs are marked with only H3K27ac suggesting that these genes are differentially active in various germ layers. We first investigated the temporal expression pattern (Owens et al., 2016) of each gene class. To exclude the interference posed by residual maternal transcripts at these early stages, we examined the expression patterns of exclusively zygotic genes in each class and found that Class 1 and 3 genes are gradually activated after ZGA while Class 2 genes remain mostly silent even at late gastrula (Figure 5B). To assess whether Class 2 genes remain inactive throughout development, we extended our temporal expression analysis until tailbud stage 26 (23 hpf). Class 1 and 3 genes are continuously active after ZGA, whereas Class 2 genes are gradually activated from early neurula and onward (Figure 5C). These data suggest that Hdac1 regulates both transcriptionally active and silent genes at gastrulation.

Figure 5 with 1 supplement see all
Hdac1 regulates germ-layer transcriptomes both in time and space.

Venn diagram comparing genes associated with Hdac1 cis-regulatory module (CRM) clusters (Clusters I–III in Figure 3C). Class 1 genes include genes closest to Cluster I peaks and genes overlapped in Classes 2 and 3; Class 2 are unique genes closest to Cluster II peaks; Class 3 are unique genes closest to Cluster III peaks. (B) Time-course TPM expression of zygotic genes in Classes 1–3 from fertilization to 10 hr post-fertilization (hpf, late st12.5). Red dotted line denotes zero. Bold number ‘4’ in red denotes the onset of zygotic genome activation (ZGA). (C) Time-course TPM expression of zygotic genes in Classes 1–3 up to 23 hpf (tailbud st26). Black vertical dotted lines denote the time window when Hdac1 binding is examined. Trend lines for each class are generated by connecting mean TPM values at each time point. (D) The expression profiles of genes affected by Trichostatin A (TSA) or valproic acid (VPA) in each class. The total number of genes in each heatmap cluster is denoted. (E) Gene ontology enrichment analysis of genes co-regulated by TSA- and VPA-mediated HDAC inhibition. Only genes with matched gene synonym to Homo sapiens are used in the analysis. (F) Spatial expression pattern of Class 1 genes co-regulated by TSA and VPA in proportions. The total number of genes in each category is listed at the top of each bar. Only *** denoting p < 0.001 (Fisher’s exact test) is shown. AC: animal cap, presumptive ectoderm; MZ: marginal zone, presumptive mesoderm; VG: vegetal mass, presumptive endoderm; NL: non-localized genes; NE: non-expressed genes. (G, H) Models of Hdac1 functioning at both inactive and active CRMs: (G) on inactive CRMs, H3K27 residue is maintained as unmodified by Hdac1, which may be subjected to H3K27me3-mediated suppression; (H) on active CRMs, the state of histone acetylation is dynamically modulated by Ep300- and-Hdac1-mediated acetylation–deacetylation cycles.

To understand how HDAC activity affects the expression of nearby genes, we performed RNA-seq using early gastrula AC or VG explants treated with or without two different Hdac inhibitors, TSA (Finnin et al., 1999) and VPA (Davie, 2003; Krämer et al., 2003). Differential gene expression analyses identified many genes that are affected after TSA or VPA treatment in both AC and VG explants (Figure 5—figure supplement 1A, C). Many of the genes affected by TSA treatment were similarly affected by VPA treatment, and vice versa (Figure 5—figure supplement 1B, C), suggesting that these are bona-fide HDAC targets. Gene ontology analyses revealed that genes affected by of HDAC inhibition primarily function in early embryonic development such as cell fate commitment, tissue morphogenesis and pattern specification (Figure 5E), consistent with the notion that HDACs are important in regulating the genes involved in early embryonic development.

Integrity of germ-layer-specific transcriptomes is maintained by spatiotemporal HDAC activity

We attempt to correlate how the expression of different classes of Hdac1-bound genes (Classes 1–3 in Figure 5A) is affected by HDAC inhibition. We examined the activation of Class 2 genes (Hadc1-bound and H3K27me3 marked) upon HDAC inhibition, which are usually not transcribed until much after gastrulation (Figure 5B, C). Interestingly, greater than 85% of Class 2 genes are prematurely activated during gastrulation upon HDAC inhibition (Figure 5D), supporting the idea that Hdac1 temporally regulates the expression of Class 2 genes (Figure 5—figure supplement 1H ). This is also consistent with observed histone hyperacetylation at repressive CRMs upon HDAC inhibition (Figure 4E, Figure 4—figure supplement 3A). Furthermore, we observed that differentially regulated genes in Class 1 (CRMs are marked by a mixture of active H3K27ac and repressive H3K27me3) and Class 3 (whose CRMs are only marked by active H3K27ac) can be up- or down-regulated upon HDAC inhibition (Figure 5D). We speculate that HDAC activity Class 1 and 3 genes are spatially regulated.

To gain insights into the spatial gene regulation by Hdac1, we analyzed the localized expression patterns of genes at early gastrula (Blitz et al., 2017). Expression profiles of Class 1 genes revealed that genes normally expressed in the vegetal region (endoderm) are significantly up-regulated in AC upon HDAC inhibition (Figure 5F, second bar). Coincidently, VG CRMs are hyperacetylated in AC upon HDAC inhibition (VG CRMs in Figure 4G), which in turn leads to the misactivation of endodermal genes in AC (Figure 5—figure supplement 1E, Supplementary file 2). Surprisingly, we found that genes normally expressed in ectoderm and mesoderm are significantly down-regulated in AC upon HDAC inhibition (Figure 5F, third bar) and genes normally expressed in mesoderm and endoderm are significantly down-regulated in VG (Figure 5F, fifth bar), upon HDAC inhibition. Contrary to the repressive role of Hdac1, this finding suggests that Hdac1 positively influences the transcription of active genes in each germ layer. We propose that the state of histone acetylation on CRMs associated with active genes is dynamic, and disruption to such equilibrium impairs normal transcription activities. Based on the quantitative analysis of pan-H3Kac profiles upon TSA treatment, we found that Hdac1-bound CRMs gain increased levels of pan-H3Kac globally including already acetylated CRMs (Figure 4E, Figure 4—figure supplement 3A). This TSA-induced excessive histone acetylation on active CRMs may contribute to the attenuated expression of associated active genes in respective germ layers (Figure 5—figure supplement 1E). A similar trend is observed among TSA responsive Class 3 genes (Figure 5—figure supplement 1F). These data demonstrate that Hdac1 not only prevents aberrant activation of silent genes, but also maintains proper gene expression levels in each germ layer (Figure 5—figure supplement 1I,J).

Since genes in Class 1 (CRMs having a mixture of H3K27ac and H3K27me3) and Class 3 (CRMs marked by only H3K27ac) and show localized transcriptomic profiles, we examined the expression differences between these two gene classes. Class 1 genes display a higher variability in expression levels between different germ layers when compared to Class 3 genes (Figure 5—figure supplement 1G). This indicates that the expression of genes undergoing active H3K27me3 suppression is more intimately associated with germ-layer determination. Altogether, these results show that Hdac1 maintains the integrity of germ-layer genes both in time and space during gastrulation.

Discussion

Here, we defined a critical role for Hdac1 during early Xenopus embryogenesis. Our findings demonstrate a close link between homeostasis of histone acetylation and transcriptional activities in developing embryos. Progressive binding of Hdac1 to the genome shapes the zygotic histone acetylome, thereby reinforcing a proper germ-layer-specific transcriptome, both in time and space. Thus, Hdac1 is an essential epigenetic regulator in the control of embryonic cell identity and lineage. We propose that TFs inducing differentiation programs exploit the activity of HDACs to confine the expression of zygotic genes.

Gradual binding of Hdac1 coincides with ZGA

The genome-wide binding of Hdac1 begins at mid-blastula and gradually accumulates at thousands of loci. Such progressive binding of Hdac1 around ZGA raises the question of whether Hdac1 recruitment requires zygotic factors during these stages. Surprisingly, ChIP-seq analyses of Hdac1 from α-amanitin-injected embryos show that zygotic transcription is dispensable for Hdac1 recruitment to the genome, at the very least, up to late blastula (Figure 2A, Figure 2—figure supplement 1B, C). Recent work in yeast shows that active transcription is required to shape histone acetylation patterns largely due to a direct role of RNAPII in recruitment and activation of H4 HATs but not HDACs (Martin et al., 2021). Our result agrees with the notion that Hdac1 recruitment is not directed by on-going transcription. We, therefore, speculate that maternal factors instruct early Hdac1 recruitment. Our results provide evidence that maternal Foxh1/Sox3 plays a role in Hdac1 genomic recruitment (Figure 2E, F), highlighting the importance of these two maternal TFs in early embryonic epigenome establishment. Though maternal TFs such as Foxh1, Vegt, and Otx1 are shown to bind the genome as early as the 32- to 64-cell stage (Charney et al., 2017; Paraiso et al., 2019), the genome-wide binding of Hdac1 likely begins at blastula and is not significantly widespread until early gastrula (Figure 1C, G). It is also known that Hdac1 containing complexes bind to pre-existing epigenetic marks such as DNA methylation and H3K4me3 (Wade et al., 1999; Lee et al., 2018). Therefore, maternally instructed histone modifications may also play a role in Hdac1 recruitment.

Hdac1 functions differently on active versus inactive CRMs

One perplexing finding is that Hdac1 occupies both active and repressive genomic loci (Figure 3A, B). Hdac1 binds to repressive genomic regions that are facultative but not constitutive heterochromatin. Though histone modifications underlying constitutive heterochromatin have been shown to regulate developmental genes (Riddle et al., 2011; Wang et al., 2018; Methot et al., 2021), the profile of H3K27me3 is largely different (~90% non-overlapping peaks) from profiles of H3K9me2, 3, and H4K20me3 in early Xenopus development (data not shown, van Kruijsbergen et al., 2017). These observations suggest that Hdac1-mediated suppression is largely dictated by developmental programs. In contrast to commonly accepted repressor function of Hdac1, the binding of Hdac1 at active genomic regions is surprising. However, our findings are consistent with previous studies in yeast and mammalian cell culture emphasizing a dynamic equilibrium of histone acetylation at active loci (Kurdistani et al., 2002; Wang et al., 2002; Wang et al., 2009; Kidder and Palmer, 2012). Hence, we hypothesize that Hdac1 functions at both active and inactive CRMs in early embryos.

To test the in vivo function of HDACs, we blocked their endogenous activity using an inhibitor and quantitatively examined the changes of general H3 acetylation (pan-H3Kac) upon HDAC inhibition. First, an increase of pan-H3Kac (detecting acetylated forms of H3K9, K14, K18, and K27) is observed across Hdac1-bound CRMs, including active CRMs, consistent with the canonical enzymatic activity of Hdac1 (Figure 4E, Figure 4—figure supplement 3A). Second, repressive CRMs marked by H3K27me3 undergo drastic H3 hyperacetylation compared to active CRMs marked by H3K27ac suggesting a HDAC-activity-dependent suppression of repressive CRMs (Figure 4F, Figure 4—figure supplement 3D). Third, germ-layer-specific pan-H3Kac profiles are disrupted indicating the importance of Hdac1 in defining spatial patterns of histone acetylation (Figure 4G, Figure 4—figure supplement 3I). Based on these results, we propose a dual function model for Hdac1. On the one hand, Hdac1 prevents histone acetylation at inactive CRMs, thereby preserving H3K27 as unacetylated (Figure 5G). Interestingly, H3K27me3 is not always imposed on inactive CRMs. For instance, active CRMs (Cluster III, only H3K27ac) are spatially modified with pan-H3Kac (Figure 4D) but are not subjected to H3K27me3 (Figure 4D, Figure 4—figure supplement 3I). This suggests that HDAC-mediated histone deacetylation and Polycomb-mediated histone methylation are not coupled at inactive CRMs. On the other hand, Hdac1 participates in dynamic histone acetylation–deacetylation cycles at active CRMs (Figure 5H). Although we did not directly test the co-binding of HATs and HDACs, CRMs may be simultaneously bound since (1) the binding profiles of Ep300 and Hdac1 mostly overlap (Figure 3—figure supplement 1A), and (2) pan-H3Kac signals increase at all Hdac1 peaks including active CRMs, upon HDAC inhibition (Figure 4E, Figure 4—figure supplement 3A). Presumably, active CRMs are maintained in a state of dynamic equilibrium. This model is in accordance with a previous study demonstrating that HATs and HDACs simultaneously participate in histone acetylation cycles, which initiate and reset chromatin between rounds of transcription (Wang et al., 2009).

Hdac1 safeguards misactivation of genes both in time and space

We attempted to correlate the activity of CRMs with the transcriptional activity of potential target genes. Genes associated with repressive CRMs (H3K27me3 only) are mainly inactive until neurula (Figure 5C). More than 85% of these genes are prematurely activated upon HDAC inhibition (Figure 5D), suggesting that Hdac1 maintains the state of histone hypoacetylation on repressive CRMs, thereby preventing premature expression of genes (Figure 5—figure supplement 1H). Moreover, genes associated with heterogeneous (both H3K27ac and H3K27me3) and active (H3K27ac only) CRMs are misactivated in different germ layers when HDAC activity is blocked (Figure 5F, Figure 5—figure supplement 1F). This indicates that Hdac1 safeguards differential histone acetylation states in each germ layer (Figure 4G, Figure 4—figure supplement 3I), restricting proper spatial transcription (Figure 5—figure supplement 1I, J). We did not directly address whether hypoacetylated heterogeneous CRMs are subjected to H3K27me3. However, a previous study showed that H3K27me3 is spatially deposited at late gastrula (Akkers et al., 2009). We predict that heterogenous CRMs are differentially marked by opposing H3K27me3 or acetylation in different germ layers. In summary, Hdac1 preserves the histone hypoacetylation state of inactive CRMs resulting in gene suppression both in time and space, thus supporting the transcriptional corepressor role for Hdac1.

Cyclical histone acetylation sustains germ-layer gene transcription

Our study reveals an unexpected role for Hdac1 in sustaining active gene expression during germ-layer formation. Within both ectoderm and endoderm, the expression of HDAC inhibitor down-regulated genes associated with either active (H3K27ac only) or heterogeneous (H3K27ac and H3K27me3) CRMs are enriched in their respective germ layers (Figure 5F, Figure 5—figure supplement 1F). This suggests a paradoxical activator role for Hdac1, which is also reported in previous studies (Vidal and Gaber, 1991; Zupkovitz et al., 2006; Baltus et al., 2009; Hughes et al., 2014; Rao and LaBonne, 2018). We speculate that utilization of HDAC activities at active genomic loci is a general mechanism, as seen in examples of Xenopus ectoderm and endoderm lineages, which deploy distinct gene regulatory networks. Based on our findings, we propose that a dynamic equilibrium between acetylation and deacetylation is essential to sustain gene transcription. The function of HDACs on active genomic regions has been elucidated in several contexts. In yeast, cotranscriptional methylation (H3K36me3 and H3K4me2) recruits HDAC containing complexes (Rpd3S and Set3C) to suppress intragenic transcription and delay induction of genes that overlap non-coding RNAs (Carrozza et al., 2005; Keogh et al., 2005; Li et al., 2007; Kim and Buratowski, 2009; Kim et al., 2012; Heo et al., 2021). Genetic deletion of Set3C affects transcript levels only in altered growth conditions (Lenstra et al., 2011), consistent with the notion that cyclical histone acetylation acts as a mechanism to regulate dynamics and fidelity of transcription. In metazoans HDAC1 can be targeted by Ep300 to transcribing genes through a direct interaction (Simone et al., 2004). Simultaneous binding of both HATs and HDACs at active genomic regions is shown in T cells (Wang et al., 2009). Inhibition of both DNA methyltransferases and HDACs induces cryptic transcription in lung cancer cells (Brocks et al., 2017). Down-regulated genes upon HDAC inhibition exhibit high levels of cryptic transcripts during mouse cardiogenesis (Milstone et al., 2020). These findings suggest a role for HDAC activity in transcriptional fidelity.

Why does excessive histone acetylation due to HDAC inhibition lead to reduced transcription instead of elevated transcription? We observed that active (H3K27ac only) and heterogeneous (both H3K27ac and H3K27me3) CRMs are excessively acetylated (Figure 4E, Figure 4—figure supplement 3A) following HDAC inhibition, which results in the reduced expression of these CRM-associated genes within their respective germ layers (Figure 5F, Figure 5—figure supplement 1F). This leads us to speculate that excessive histone acetylation interferes the activity of the transcriptional machinery. A recent study shows that excessive histone acetylation on chromatin induced by inhibition of HDAC1, 2, and 3 leads to increased aberrant contacts and reduced native contacts between super-enhancer loops (Gryder et al., 2019). This suggests that the excessive histone acetylation impairs active transcription by altering chromatin interactions. Alternatively, excessive histone acetylation can alter the binding of acetyl-histone readers. H4 polyacetylation induced by HDAC inhibition is shown to be preferentially bound by BRD proteins (such as BRD4), thereby sequestering these factors away from active genes (Slaughter et al., 2021). Therefore, HDACs safeguard the function of normal acetyl-histone readers. Further investigation of cyclical histone acetylation regulating developmental programs is needed in the context of germ-layer specification.

Materials and methods

Animal model and subject details

Request a detailed protocol

Xenopus tropicalis embryos were obtained by in vitro fertilization according to Ogino et al., 2006 and staged according to Nieuwkoop and Faber, 1994. All embryos were cultured in 1/9× Marc’s modified Ringers (MMR) at 25°C. For HDAC inhibition, 4-cell stage embryos were immersed in 1/9× MMR containing (1) 100 nM TSA (Esmaeili et al., 2020) or DMSO; or (2) 10 mM VPA (Rao and LaBonne, 2018) or H2O. For α-amanitin injection, each 1-cell stage embryo was injected with 6 pg of α-amanitin (Hontelez et al., 2015). For morpholino injection, 10 ng of morpholino (Foxh1 MO: 5′-TCATCCTGAGGCTCCGCCCTCTCTA-3′, Chiu et al., 2014; Sox3 MO: 5′-GTCTGTGTCCAACATGCTATACATC-3′, Gentsch et al., 2019) was injected into 1-cell staged embryos. For spatial analyses, embryos were dissected at the late blastula stage (6 hpf), and explants were cultured to the early gastrula (7 hpf). Animals were raised and maintained following the University of California, Irvine Institutional Animal Care Use Committee (IACUC). Animals used were raised in the laboratory and/or purchased from the National Xenopus Resource (RRID: SCR_013731).

Western blotting

Request a detailed protocol

Embryos were homogenized in 1× RIPA (50 mM Tris–HCl pH7.6, 1% NP40, 0.25% Na-deoxy-cholate, 150 mM NaCl, 1 mM etheylenediaminetetraacetic acid [EDTA]), 0.1% sodium dodecyl sulfate [SDS], 0.5 mM dithiothreitol (DTT) with protease inhibitors (Roche cOmplete) and centrifuged twice at 14,000 rpm. The supernatant was then subjected to western blotting using anti-HDAC1 (Cell Signaling, 34589S), anti-HDAC2 (Genetex, GTX109642), and anti-Tubulin (Sigma, T5168). For histone modifications, acid-extracted histone lysate was prepared accordingly (Shechter et al., 2007) and subjected to western blotting using anti-H3K9ac (Cell Signaling, 9649), H3K14ac (Cell Signaling, 7627), H3K18ac (Cell Signaling, 13998), H3K27ac (Cell Signaling, 8173), H3K56ac (Cell Signaling, 4243), and H4K20ac (Active Motif, 61531).

ChIP and ChIP-seq analysis

Request a detailed protocol

ChIP protocol was performed as described (Chiu et al., 2014). Antibodies used for ChIP were anti-HDAC1 (Cell Signaling, 34589S, 1:100), anti-HDAC2 (Cell Signaling, 57156S, 1:100), anti-H3K18ac (Cell Signaling, 13998, 1:100), and anti-Sox3 (Zhang et al., 2003). ChIP-seq libraries were constructed using the NEBNext Ultra II DNA Kit (NEB, E7645).

For sequential ChIP, the first round of ChIP was performed as described and eluted in 1× Tris-EDTA (TE containing 1% SDS at 37°C for 30 min). The eluate was diluted ten-fold with 1× RIPA (without SDS) and subjected to the second round of ChIP as described (Desvoyes et al., 2018). Real-time quantitative PCR (RT-qPCR) was performed using Power SYBR Green PCR master mix (Roche) to quantify the DNA recovery compared to ChIP input DNA at one embryo equivalency (percent input). The error among technical replicates was calculated using the rule of error propagation. ChIP qPCR primer sequence information is provided in Supplementary file 3.

For dissected pan-H3Kac ChIP, spike-in chromatin (Active Motif, 53083) was added to the chromatin of dissected tissues at a ratio of 1:35. Mixed chromatin was then subjected to ChIP with 5 µg anti-panH3Kac (Active Motif, 39139) and 1 µg of anti-H2Av (Active Motif, 61686) and followed as described.

All experiments were performed in two independent biological replicates unless noted. Sequencing was performed using the Illumina NovaSeq 6000 and 100 bp single-end reads or 100 bp paired-end reads were obtained.

All sequencing data were aligned to Xenopus tropicalis v10.0 genome (http://www.xenbase.org/, RRID:SCR_003280) using Bowtie2 v2.4.4 (Langmead and Salzberg, 2012). PCR duplicates were removed using Samtools v1.10 (Li et al., 2009). ChIP-seq signals were visualized using IGV v2.11.3 (Robinson et al., 2011) after concatenating two biological replicates when available. IDR analysis (Li et al., 2011) was used to identify high-confidence peaks called by Macs2 v2.7.1 (Zhang et al., 2008) against the stage-matched input (Charney et al., 2017) between two biological replicates according to ENCODE3 ChIP-seq pipelines (IDR threshold of 0.05) (https://docs.google.com/document/d/1lG_Rd7fnYgRpSIqrIfuVlAz2dW1VaSQThzk836Db99c/edit). Differential ChIP peak analysis was performed using Homer v4.11 (Heinz et al., 2010).

For dissected pan-H3Kac ChIP, all second replicates were downsampled to 25% to compare equivalent sequencing depth. Drosophila H2Av peaks are generated from published S2 cell samples (Tettey et al., 2019). Normalization factors were then calculated based on reads that mapped to Drosophila H2Av peaks for each ChIP-seq sample (Egan et al., 2016). Detailed normalization factors used are listed Supplementary file 4.

RNA-seq and analysis

Request a detailed protocol

Total RNA from dissected tissues was extracted using Trizol as described (Amin et al., 2014). mRNA was then isolated using NEBNext PolyA mRNA Magnetic Isolation Module (NEB E7490S). Sequencing libraries were prepared using NEBNext Ultra II RNA library prep kit (NEB E7770S) and sequenced by the Illumina NovaSeq 6000 with 100 bp paired-end reads. All experiments were performed in two independent biological replicates.

All sequencing samples were aligned using STAR v2.7.3a (Dobin et al., 2013) to Xenopus tropicalis genome v10.0 (http://www.xenbase.org/, RRID:SCR_003280) to obtain raw read counts. RSEM v1.3.3 (Li and Dewey, 2011) was used to calculate expression values in transcripts per million (TPM) which are used to construct heatmaps depicting gene expression levels. Differentially expressed genes were identified using edgeR v3.36.0 (Robinson et al., 2010) with the following parameters: greater than twofold change and less than 0.05 false discovery rate (also known as the adjusted p-value), in R v4.1.2 (R Development Core Team, 2021). Metascape (Zhou et al., 2019) was used to perform gene ontology enrichment analyses with default parameters (min overlap = 3, p-value cutoff = 0.01, and min enrichment = 1.5).

Additional bioinformatics and statistical analyses

Request a detailed protocol

Samtools v1.10 (Li et al., 2009) was used to convert between SAM and BAM files. DeepTools v3.5.0 (Ramirez et al., 2014) was used to generate: (1) ChIP-seq signal track (bigwig files) normalized by reads per genomic content (-RPGC) at the bin size of 1 bp; (2) heatmaps around peak summits normalized by Bins Per Million mapped reads (-BPM) at the bin size of 50 bps; (3) signal profile along the gene bodies normalized by -BPM at the bin size of 50 bps; (4) Pearson correlation between ChIP-seq samples at peaks. Homer v4.10 (Heinz et al., 2010) was used to annotate genomic features of ChIP peaks. Bedtools v2.29.2 (Quinlan and Hall, 2010) was used to determine peak overlaps among ChIP-seq peaks and obtain counts of reads at ChIP-seq peaks. CentriMo (Bailey and Machanick, 2012) was used to perform local motif enrichment analysis. Welch Two-sample t-test in R v4.1.2 was used to determine the statistical significance between groups. Cohen’s d (effect size) was calculated using lsr v0.5.2 package in R v4.1.2 (R Development Core Team, 2021). Time-course gene expression was obtained from ribosomal RNA-depleted RNA-seq data (Owens et al., 2016). The expression in TPM was calculated as outlined above. Spatial gene expression at early gastrula was obtained from RNA-seq of five dissected tissues (Blitz et al., 2017) consisting of the animal cap (ectoderm), the dorsal marginal zone (dorsal mesoderm), the lateral marginal zone (lateral mesoderm), the ventral marginal zone (ventral mesoderm), and the vegetal mass (endoderm). The expression in TPM was obtained as outlined above. Fisher’s exact test (alternative = "greater") in R v4.1.2 (R Development Core Team, 2021) was used to determine the significance of the proportional enrichment between groups. p-Values from Fisher’s exact test are summarized in Supplementary file 5.

Categorical analyses

Request a detailed protocol

Spatial categorization of CRMs is defined as below: AC CRMs represent CRMs whose pan-H3Kac signals in AC is twofold higher than in VG; VG CRMs represent CRMs whose pan-H3Kac signals in VG is twofold higher than in AC; ubiquitous CRMs represent the remaining CRMs whose pan-H3Kac signals do not exceed twofold enrichment in either of the two examined germ layers. For temporal gene expression analysis, (strictly) zygotic genes are determined by removing (1) genes whose expression levels are greater than 1 TPM during the first 2 hr post-fertilization and (2) genes whose expression levels are less than 1 TPM from 0 to 23 hpf. Spatial categorization of genes at early gastrula stage: the average TPM between three dissected mesoderm tissues (dorsal, marginal, and lateral marginal zones) was used to represent the expression of mesoderm. Genes with the expression in any dissected tissue less than 1 TPM were considered not expressed. Genes with the coefficient of variance of TPM less than 0.1 (10%) were considered evenly expressed. The remaining genes with localized expression were assigned to a germ layer based on the maximum TPM.

Data availability

Raw and processed RNA-seq and ChIP-seq datasets generated from this study are available at NCBI Gene Expression Omnibus using the accession GSE198378. Publicly available datasets used in this study are available at NCBI Gene Expression Omnibus using the accession GSE56000 (Gupta et al., 2014; H3K27ac ChIP-seq), GSE67974 (Hontelez et al., 2015; Ep300, H3K9ac, H3K4me1, H3K4me3, H3K36me3, H3K9me2, H3K9me3, H3K27me3, H4K20me3 ChIP-seq, and st12 RNA Pol2 ChIP-seq), GSE65785 (Owens et al., 2016; temporal profiling of RNA-seq), GSE85273 (Charney et al., 2017; st9 Foxh1 ChIP-seq, st7, 8, 9, and 10.5 RNA Pol2 ChIP-seq), GSE81458 (Blitz et al., 2017; st10.5 dissected germ-layer RNA-seq), and GSE129236 (Tettey et al., 2019; H2Av ChIP-seq in S2 cells). Relevant bioinformatic analysis scripts are accessible at https://github.com/jiajinglz/bioRxiv_05052022_Hdac_dual_roles (copy archived at Zhou, 2023).

The following data sets were generated
    1. Zhou JJ
    (2022) NCBI Gene Expression Omnibus
    ID GSE198378. Histone deacetylase 1 maintains lineage integrity through histone acetylome refinement during early embryogenesis.
The following previously published data sets were used
    1. Gupta R
    2. Baker JC
    (2014) NCBI Gene Expression Omnibus
    ID GSE56000. Enhancer chromatin signatures predict Smad2/3 binding in Xenopus.
    1. Hontelez S
    2. Veenstra GC
    (2015) NCBI Gene Expression Omnibus
    ID GSE67974. Embryonic transcription is controlled by maternally defined chromatin state.
    1. Owens ND
    2. Blitz IL
    3. Lane MA
    4. Patrushev I
    5. Overton JD
    6. Gilchrist MJ
    7. Cho KW
    8. Khokha MK
    (2016) NCBI Gene Expression Omnibus
    ID GSE65785. Measuring Absolute RNA Copy Numbers at High Temporal Resolution Reveals Transcriptome Kinetics in Development.
    1. Charney RM
    2. Cho KW
    (2017) NCBI Gene Expression Omnibus
    ID GSE85273. Foxh1 marks the embryonic genome prior to the activation of the mesendoderm gene regulatory program.
    1. Blitz IL
    2. Paraiso KD
    (2017) NCBI Gene Expression Omnibus
    ID GSE81458. Regional expression of X. tropicalis transcription factors in early gastrula embryos.
    1. Tettey TT
    2. Gao X
    3. Shao W
    4. Li H
    5. Story BA
    6. Chitsazan AD
    7. Glaser RL
    8. Goode ZH
    9. Seidel CW
    10. Conaway RC
    11. Zeitlinger J
    12. Blanchette M
    13. Conaway JW
    (2019) NCBI Gene Expression Omnibus
    ID GSE129236. Expression profiling by high throughput sequencing Genome binding/occupancy profiling by high throughput sequencing.

References

    1. Inoue A
    2. Fujimoto D
    (1969) Enzymatic deacetylation of histone
    Biochemical and Biophysical Research Communications 36:146–150.
    https://doi.org/10.1016/0006-291x(69)90661-5
  1. Book
    1. Nieuwkoop PD
    2. Faber J
    (1994)
    Normal Table of Xenopus laevis (Daudin)
    Garland Publishing Inc.
  2. Software
    1. R Development Core Team
    (2021) R: A language and environment for statistical computing
    R Foundation for Statistical Computing, Vienna, Austria.
    1. Yoshida M
    2. Kijima M
    3. Akita M
    4. Beppu T
    (1990)
    Potent and specific inhibition of mammalian histone deacetylase both in vivo and in vitro by trichostatin a
    The Journal of Biological Chemistry 265:17174–17179.

Decision letter

  1. Matthew C Good
    Reviewing Editor; University of Pennsylvania Perelman School of Medicine, United States
  2. Marianne E Bronner
    Senior Editor; California Institute of Technology, 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 "Histone deacetylase 1 maintains lineage integrity through histone acetylome refinement during early embryogenesis" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Jessica Tyler as the Senior Editor. The reviewers have opted to remain anonymous.

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

Essential revisions:

The reviewers highlighted a number of novel features of your study. This includes uncovering dual positive and negative roles for HDACs in cell fate specification in the early embryo and identifying a key contribution of HDACs to germ layer cell lineage integrity. Inhibition of deacetylation causes dysregulation of specificity of spatial boundaries in the gastrula, with mesendoderm misexpressed in animal cap cells. Additionally, you demonstrate progressive binding of HDAC1 to the genome with ChIP-seq and find correlative patterns of promoter occupancy relative to other repressive and activating histone epigenetic marks and ChIP-seq peaks for maternal transcription factors, FoxH1 and Sox3. On the whole, this work advances our understanding of HDACs and control of chromatin acetylation and gene expression tied to early cell fate specificity in a vertebrate embryo.

However, the manuscript is not acceptable in its current format and will require major experimental revisions for consideration. Referees brought up a number of concerns.

1) Direct versus indirect effects of deacetylation inhibition. Specifically, whether H3-acetylation peaks and gene expression changes from deacetylase inhibition can be directly assigned to HDAC1 activity. One reviewer suggested a knockdown of HDAC1 would be useful, as performed in a previous study by Rao and Labonne, Development 2018. Additionally, because the authors primarily performed their experiments with TSA, rather than HDAC1-specific VPA, it is suggested that they expand the comparison, at least for RNA-seq, and also consider directly comparing the effects on histone to separately assign chromatin binding and gene regulation individually to HDAC1 versus HDAC2.

2) Lack of evidence that HDAC1 recruitment is likely dependent on maternal factors. The authors relied on the correlation of HDAC1 ChIP-seq peaks from this manuscript to FoxH1 and Sox3 peaks in published ChIP-seq datasets. Although the correlation is supportive, reviewers felt that the conclusion lacked experimental support. They proffer two suggestions: (1) to temporally analyze when HDAC1 binding begins in the egg or early gastrula, or (2) to knockdown maternal FoxH1 and Sox3 and characterize the loss of HDAC1 peaks from ChIP-seq.

3) There are a number of comments from reviewer 2 about improving manuscript clarity, and specific notes from reviewer 3 about analyses and figures. These should be addressed as well. Reviewer 1 suggested highlighting further that TSA causes more extensive gene expression dysregulation in AC versus VG, and to specifically plot mis-expression of endoderm VG genes in AC.

Reviewer #1 (Recommendations for the authors):

This study provides a number of new insights on the role of histone acetylation-deacetylation in spatial regulation of gene expression that contributes to cell lineage regulation as germ layers begin to form in the gastrula embryos. The work is important and interesting to the field. However, my enthusiasm is partly limited by the correlative nature of the experiments and some missing analyses. My comments to the author are below, formatted as ways to improve the manuscript.

1. Extent of mis-regulation of germ layer-specific genes upon TSA and VPA expression. The authors provide volcano plots in Figure S5B, and differential gene expression analysis spatially within AC and VG in Figure 4F. They find that up-regulation of H3Kac peaks tends to occur for opposite lineage, ie. VG genes are misexpressed in AC. Further, that downregulated genes tend to be more lineage constrained. It would be helpful if they could also plot a smaller subset of the data for ectoderm and endoderm-specific gene lists. Based on Figure 4F, one would expect that a large fraction of endoderm-specific genes would be inappropriately upregulated in the AC, and a large fraction of ectoderm-specific genes upregulated in the VG. To what extent is this true.

I also think the authors should highlight and discuss that TSA treatment has a much more profound effect on gene expression dysregulation in the AC compared to the VG (Figure S5B). There are many more genes up and downregulated in the AC. This is quite a novel and interesting result. Can the authors comment on why?

2. Timing of HDAC1 binding to the genome. It is interesting that HDAC1 peaks are widespread in the genome by Stage 8, presumably during zygotic genome activation, and that these peaks largely strengthen over time into the early gastrula. This however begs the question of when HDAC1 begins to bind to the genome. Is it already bound in the egg, or upon fertilization, or are peaks only detectable starting around the time of genome activation? Given that the protein is present in the egg and throughout cleavage stages (Figure 1A), and that the authors propose FoxH1 and Sox3 may be contributing to recruitment, the manuscript would be strengthened by characterization of the time of recruitment.

3. Differential role for HDAC1 and HDAC2 in lineage maintenance? The TSA treatment makes it difficult to disentangle specific subsets of genes modulated by promoter-proximal deacetylation. Can the authors contrast TSA vs. VPA inhibition experiments to identify peaks that could likely be attributed to HDAC2 rather than HDAC1, and perform GO analyses? Alternatively, can the authors provide direct binding data with ChIP-seq of HDAC2 in blastula and gastrula.

4. Direct evidence that HDAC1/2 binding contributes to gene expression regulation. The authors utilize inhibition of deacetylation (TSA) or HDAC1 specifically (VPA) to analyze changes to pan-H3K-acetylation and gene expression, regionally within the AC and VG. In general, I like this line of experiments, however, I wonder whether the TSA inhibition might be affecting other deacetylase enzymes, and what the effect would be of MO knockdown of the HDACs (or their recruitment via FoxH1, Sox3). Other groups have found that HAT-HDAC binding to promoters and enhancers does not always correlate to gene expression regulation. Extending this notion here, I worry that the authors may not be directly analyzing the downstream function of HDACs. Can the authors compare their differential expression data to existing gene profiling studies in embryos that have knockdown of FoxH1, Sox3, or HDAC1/2 to demonstrate similar profiles of dysregulation?

Reviewer #2 (Recommendations for the authors):

As a general reader, I do not know enough of the background knowledge in the field to know how novel the observations are, and it seems that most of these specific comparisons are made in the discussion, and it is not explicit what observations are really new, or whether the importance of the paper is in the documentation. The authors need to draw this out more clearly. Also, as a general reader without the ingrained knowledge of the activating or inhibitory functions of the different histone modifications, I initially found it impossible to hold onto the threads of the arguments but was able to do so by color coding the modification types as activating, inhibitory, or heterochromatin in the text. I would hope that modern journals might be able to use colored text to enable readers to read and evaluate the text more readily. Similarly, it is hard to hold on to the nature of different classes of response, so I suggest more descriptive names for the classes. Without these, the manuscript is very tough going for the general reader. Indeed, as a detailed examination of chromatin modification, it is inevitably dense. Again as a general reader, I would find it more useful to present the specific genes that convey the arguments in the main text and figures, so that there is something to get one's teeth into and move the venn diagrams and stacked bar graphs and violin plots, which I find uninformative for the specific arguments, but which are necessary for the generality of the claims, to the supplement. But these are suggestions to make the manuscript more accessible to the general reader. If the target audience is the specialized field of chromatin organization or detailed regulation of gene expression in developing animals (without going beyond inhibitor studies to the sequences responsible for the DNA recruitment or the enzymatic activity of the protein), then the paper would probably be more appropriate for a more focused journal. But if there are new insights that can be more clearly articulated, then the manuscript could be appropriate to eLife.

Major points, in no particular order.

1. While the Hdac genes are shown for mRNA it would be useful to see whether this matches the protein databases. Are they detectable, and does mRNA translate to protein abundance?

2. Line 140- were Vegt peaks found to be enriched in the maternal motifs? (Gentsch?)

Indeed, Was there enrichment of other motifs associated with transcriptional activation as found by Gentsch et al? (POU-SOX (Pou5f3-Sox3 heterodimer), Krüppel-like zinc finger (ZF; Sp1 and several Klf), POU (Pou5f3), SOX (Sox3), bZIP (Max), FOXH (Foxh1), ETS (Ets2), NFY (NFYa/b/c), SMAD (Smad1/2), VegT (mVegT)), A lack of enrichment would illustrate a strong difference in recruitment and be informative.

3. "Consistent with the overlapping binding of Ep300, Hdac1 peaks display a moderate level of H3K9ac (Cluster c and d), high levels of 164 H3K18ac (Cluster b, c, and d), H3K27ac (Cluster b, c, and d), and pan-H3 lysine acetylation (pan-H3Kac)."

Why is this consistent (ref required?) is Ep300 always an activating signal? The next section discusses activating signals, so maybe so, but it does not say so. What is the current understanding of Ep300? General enhancer binding? I fear that the paper is exceedingly hard to follow unless one has these identities imprinted in one's own brain, which I don't. Indeed, for people not in the immediate field, the nomenclature, with its acetylations, and multiple methylations is all too easy to forget. For ease of reviewing, I color-coded the histone marks according to Wikipedia. It made the manuscript much easier to follow, though I am sure my simplistic green or red color code is too simple. The color coding of the figures in the manuscript is more nuanced, and presumably reflects some deeper understanding. It would be useful to share this with the reader. Otherwise, the activating (green) repressive (red), and heterochromatin (brown, blue?- what are these?) and the main text should also match an informative color coding, which should be trivial in these days of digital typesetting.

4. "we speculate that these regions are differentially marked in space due to heterogeneous cell populations present in the whole embryo. Hdac1 Cluster I peaks are referred to as heterogeneous CRMs."

Surely the authors can give explicit examples here for some known heterogeneously expressed genes to reduce the speculative nature of this group? (though they do eventually show up in the supplements- but why not progress the argument from the specific to the general?).

5. Ideally, the figures and legends should be interpretable independent of the text, for example, when wondering what abcd are in figures 1 and 3, one should not have to search in the text for the relevant explanation.

"Overall, a minority of Hdac1 peaks are unique to each of the blastula stages (Cluster a and b) while a majority of Hdac1 peaks are present across multiple stages (Cluster c) and at the early gastrula stage (Cluster d) (Figure 1C, S1D).”

And in reading this, "unique to each blastula stage", could be expressed more clearly to indicate their uniqueness- ie at stage 8 (cluster a, likely driven by maternal factors?) and stage 9, (cluster b, likely driven by newly expressed zygotic factors or signals?).

Re the italics – where are the genes known to be in these classes? Ie the Nodals, vs genes driven by zygotically activated signals like Smad2? It is important for a simple argument to give the general reader some specific biological context.

6. Likewise "These observations suggest that Hdac1-mediated suppression is largely dictated by developmental programs" – perhaps the authors might give some examples in the text.

7. "First, an increase of pan-H3Kac is observed across Hdac1-bound CRMs, including active CRMs, consistent with the canonical enzymatic activity of Hdac1" remind us what H3 Kac is for – it seems to be in the purple, brown and red categories. Is it associated with transcription start sites? It's not clear what antibody was used. It seems it must be a pan H3Kac antibody?

8. What are the effects of Hdac inhibitors on the transcriptome? Seems that this is an essential piece of information to interpret the results on chromatin? Seems there is a lot of literature on this, and so far as I can tell it is not cited until the discussion. It would be nice to have a digested understanding to compare to the chromatin results in the Results section (some of this appears eventually, but late enough to leave the reader frustrated when the topic comes up).

9. I had great difficulty understanding the paragraph that begins "Given that Hdac1 CRM clusters (Clusters I-IV in Figure 3C) are subjected to both active and repressive epigenetic modifications presumably in different germ layers, we compared the general status of H3 acetylome (pan-H3Kac) between two distinct germ layers" I think that all of this is VPA and TSA independent information? But in the end, isn't it all circular? Ie higher enrichment of pan H3Kac associated with higher expression?

10. "Taken together, these data demonstrate that Hdac1 maintains differential H3 acetylation states between germ layers through its catalytic activity." Is a rather bland statement after trudging through the paragraph. I think that one can make the general statement that upon inhibition of HDAC, repressed genes have a larger increase in their pan H3Kac than do active genes. And is there evidence that it is through the catalytic activity rather than its binding?

"we found that the fold increases of pan-H3Kac signals after TSA treatment are negatively correlated with the levels of endogenous pan-H3Kac signals in both AC and VG explants" This is an unnecessarily complicated sentence. It's already hard to figure out the direction things are going in without having to think about what negative correlation means in these contexts. But I think it means what I just suggested?

11. I ask that the authors consider how much time would be spent by the reader looking back to see what Class1,2 and 3 genes are. Could they not be more readily described as "selectively activated", "constitutively repressed" and "Constitutively activated" if these were designated SA, CR, and CA that might be easier to remember/interpret, especially for me who can't remember class I and 2 from one paragraph to the next. Or consider the value of spelling them out each time, at least for the reviewer. Once accepted, one can return to uninformative abbreviations and let the reader suffer.

12. What is the mechanism of Hdac inhibition by VPA and TSA? Does the Hdac dissociate from chromatin, allowing access to acetylases? Or does it stay on the chromatin in an inactive form, suggesting a more complex mechanism of the differential effects of inactivation? What do the data say about the potential mechanism?

Reviewer #3 (Recommendations for the authors):

My general concern regarding this work is that most of the conclusions come from indirect evidences:

– Maternal TF involvement is only supported by motif enrichment.

– Spatial difference in H3K27ac/me and HDAC1 marking between AC/V part of embryos based on whole embryos data.

– Functional test of HDAC1 requirement carried out through the use of broad HDAC inhibitor. Here I find that the authors investigate in great detail the behaviour of HDAC1-associated regions but do not provide evidence that these regions are preferentially affected by TSA treatment. HDAC1 knockdown would be a much more straightforward assay to interpret in my view.

Specific points

Figure 1

1C: St101/2 or stage 9-specific HDAC1 peaks seem to already show signal at stage 8. (Difficult to grasp relative level as no indication of what the blue scale is….). In any case, is this low level at stage 8 meaningful (how does it look in all enhancer peaks for example).

What is the temporal expression (early blastula to neurula stages for example) of the a-d set of HDAC1-associated genes? Any correlation between stage-specific binding and expression pattern would strengthen the proposed involvement of HDAC1 in the epigenetic regulation of genes around ZGA.

Figure 2:

2B What are the enriched motifs in stage 8 and stage 10 specific peaks?

2C: How does the HDAC1 peak intensity look if you integrate the HDAC1 chip a-amanitin data in the map from 1C (in complement to S2B).

Charney et al. document binding of TF pre/post ZGA.

To support the ID that TF binding is associated with HDAC1 binding it would be important to check if stage-specific binding of HDAC1 is associated with stage-specific binding of TF.

The title "HDAC1 binding is instructed maternally" is not appropriate. Interference with maternal TFs would be needed to go beyond correlation. HDAC1 binding after foxh1MO (as in Charney et al) will indicate if HDAc1 binding depends on the presence of this maternal TF

Figure 3

"moderate" level of K9ac etc….moderate compared to what??? I would think comparing for example level of p300+ HDAC+ to p300+HDAc- would give an actual idea if the level on HDAC associated p300 peak stands out.

The same goes for figure S3B.

Without such kind of comparison, it is difficult to conclude that HDAC1 peaks have biased epigenetic status compared to all p300 peaks for example.

Again in Figure 3B HDAc1 bound region exhibits a strong signal density of K27me3.

Line 183: I don't think cluster b fits the description.

Fig3CD why exclude HDAC peaks from stage 8?

It would be useful to see how HDAC peaks signal looks on Figure 3D map.

In 3D there seems to be a clear anti-correlation between K27me3 and K27ac signal (see cluster I) that does not really fit the description of line187-191. This is important because the concept proposed in the manuscript relies on these co-occurring K27ac/me sites. Are the signal anti-correlated in cluster I?

S3C K27ac and me do not seem to overlap much. Can we see the region where peaks are detected (add a box where peaks are detected)?

Line 203 – before testing of this hypothesis is surely again to compare acetylation in p300+ HDAC+ versus p300+ HDAC- (or at TSS) to evaluate if TSA primarily affects HDAc1 sites?

Figure 4A/B

Can the author use a TSA control experiment where TSA is applied only prior to ZGA or only around ZGA?? This will indicate what is the time window when TSA has this effect.

It would be also important to document the effect on the development of whole embryos of the TSA treatment scheme used on explant (Figure 4C). Are the embryos defective in these conditions?

4C:

What is the effect of TSA on the number of peaks detected?

Would be very important to have a heat map that summarizes the panAc-peak in control and TSA conditions as well as the HDAc1 peak. This would illustrate to which extent TSA primarily affects HDAC binding sites.

AC and VG-specific peaks seem to exhibit a much lower intensity of panAc signal than common (scale missing on the map…). What could be the reason for such a difference?

4E FC I assume is FC TSA / control.

Fig4E and S4G

Global increase in acetylation in embryos (4D) – immediate question whether acetylation on the genome disproportionately affects HDAC1 binding sites?

This is critical since the authors assume that TSA effect indeed reflects the effect on HDAC1 activity. Authors, unfortunately, focus analysis solely on HDAc1 binding sites so the conclusion line249 is not yet supported.

All observations could also fit the model whereby TSA leads to the homogeneous increase of ac level genome-wide – would obviously lead to lower fold change if starting point high (cluster III) and higher if low (cluster II).

Figure 5

When excluding genes with maternal transcripts how many genes are left in each category?

How do sets of genes with the same epigenetic configuration but HDAC- compare (i.e genes that are K27me3/K27ac but without HDAC1?).

VPA analysis cannot support the claim that TSA treatment has no off-target effect because this (i) is not RNA-seq analysis and (ii) 8/24 genes selected for RTqPCR are not showing the expected effect.

The first important question to address: are the TSA-affected genes disproportionally enriched for HDAC peaks associated genes?

5E Not sure to get the significance of observation.

Can it be that a gene can be detected as upregulated only if they have a low level in control conditions can be detected as downregulated only if they have a high level in control conditions? Which is the rationale for the selection of Ac or V-specific genes…?

https://doi.org/10.7554/eLife.79380.sa1

Author response

1) Direct versus indirect effects of deacetylation inhibition. Specifically, whether H3-acetylation peaks and gene expression changes from deacetylase inhibition can be directly assigned to HDAC1 activity. One reviewer suggested a knockdown of HDAC1 would be useful, as performed in a previous study by Rao and Labonne, Development 2018. Additionally, because the authors primarily performed their experiments with TSA, rather than HDAC1-specific VPA, it is suggested that they expand the comparison, at least for RNA-seq, and also consider directly comparing the effects on histone to separately assign chromatin binding and gene regulation individually to HDAC1 versus HDAC2.

We followed the advice and examined the effect of inhibiting Hdac activities using both TSA and VPA (Figure S4.1-3 Figure 5F, Figure S5A-F). Additionally, we also performed Hdac2 ChIP-seq to compare the chromatin binding behavior of Hdac1 and Hdac2 (Figure S1E-H).

2) Lack of evidence that HDAC1 recruitment is likely dependent on maternal factors. The authors relied on the correlation of HDAC1 ChIP-seq peaks from this manuscript to FoxH1 and Sox3 peaks in published ChIP-seq datasets. Although the correlation is supportive, reviewers felt that the conclusion lacked experimental support. They proffer two suggestions: (1) to temporally analyze when HDAC1 binding begins in the egg or early gastrula, or (2) to knockdown maternal FoxH1 and Sox3 and characterize the loss of HDAC1 peaks from ChIP-seq.

Inhibition of zygotic transcription using a-amanitin injection showed that the genomic binding profile of Hdac1 is largely unchanged at the late blastula stage. This inhibition of zygotic gene expression provides strong evidence supporting the conclusion that zygotic factors are not required for Hdac1 binding at this stage, implicating maternal recruitment. Additionally, we now include the ChIP-qPCR analysis of Hdac1 binding around Foxh1/Sox3 cobound regions in Foxh1 or Sox3 morphants. We found that Hdac1 binding is reduced, but not abolished in these morphants, suggesting that, in addition to Foxh1 and Sox3, other maternal TFs are likely to be involved in the Hdac1 recruitment to DNA.

3) There are a number of comments from reviewer 2 about improving manuscript clarity, and specific notes from reviewer 3 about analyses and figures. These should be addressed as well. Reviewer 1 suggested highlighting further that TSA causes more extensive gene expression dysregulation in AC versus VG, and to specifically plot mis-expression of endoderm VG genes in AC.

We incorporated all suggestions raised by reviewers and also included additional gene expression dysregulation in AC vs VG (Figure S5E, Table S3).

Reviewer #1 (Recommendations for the authors):

This study provides a number of new insights on the role of histone acetylation-deacetylation in spatial regulation of gene expression that contributes to cell lineage regulation as germ layers begin to form in the gastrula embryos. The work is important and interesting to the field. However, my enthusiasm is partly limited by the correlative nature of the experiments and some missing analyses. My comments to the author are below, formatted as ways to improve the manuscript.

1. Extent of mis-regulation of germ layer-specific genes upon TSA and VPA expression. The authors provide volcano plots in Figure S5B, and differential gene expression analysis spatially within AC and VG in Figure 4F. They find that up-regulation of H3Kac peaks tends to occur for opposite lineage, ie. VG genes are misexpressed in AC. Further, that downregulated genes tend to be more lineage constrained. It would be helpful if they could also plot a smaller subset of the data for ectoderm and endoderm-specific gene lists.

We added a heatmap showing the germ layer specific gene expression changes upon HDAC inhibition by TSA and VPA (Figure S5E). Table S3 lists functions of these genes and their PubMed references.

Based on Figure 4F, one would expect that a large fraction of endoderm-specific genes would be inappropriately upregulated in the AC, and a large fraction of ectoderm-specific genes upregulated in the VG. To what extent is this true.

Figures 5F and S5F support the notion that a large fraction of endoderm-specific genes are inappropriately upregulated in the AC, whereas a large fraction of ectoderm-specific genes are upregulated in the VG upon HDAC inhibition.

To determine whether endodermal genes are preferentially upregulated in ectodermal cells upon TSA treatment (Figure 5F), we compared the following genomic data sets: (1) a list of genes preferentially expressed in ectodermal cells upon HDAC inhibition, (2) a list of endodermal genes based on dissected gastrula RNA-seq data (Blitz et al., 2017). We used Fisher’s exact test to determine genes are upregulated in ectodermal lineage cells upon HDAC inhibition.

The parameters to assign genes to spatial expression patterns is described in Materials and methods under Categorical Analyses. Genes associated with low expression levels (<1 TPM) are categorized as not expressed (NE), genes whose coefficient of variation is less than 10% among the 3 germ layers are considered expression not localized (NL), genes expressed highest in animal caps (ectoderm) are classified as animal cap enriched genes (AC), genes expressed highest in marginal zone (mesoderm) explants are classified as marginal zone enriched genes (MZ), and genes expressed highest in vegetal mass (endoderm) explants are classified as vegetal mass enriched genes (VG). For instance, the expression level of foxi1 is 107 TPM in the animal cap, 10.6 TPM in the marginal zone, and 1.2 TPM in vegetal mass. Since Foxi1 expression is highest in AC, it is assigned as an animal cap enriched (AC) gene. This assignment is not perfect but practical.

We also examined whether the mis-regulation of the ectodermal genes in endoderm, or endodermal genes in ectoderm is statistically significant using Fisher’s exact test. For instance, the first bar in Figure 5F indicates that 6,170 genes among 28,850 annotated genes are endodermal genes (VG, colored in green), which is 21.4%. The second bar shows that 43 out of 92 genes upregulated after HDAC inhibition in ectoderm (AC) are endodermal genes (VG, green), which is 46.7%. Fisher’s exact test indicates that the difference is statistically significant (p-value < 6.1e-8). Similar analysis showed that ectodermal genes are preferentially upregulated in the endoderm. We suggest that HDAC activity is required to safeguard the misactivation of “inappropriate” germ layer genes in each region of the embryo.

I also think the authors should highlight and discuss that TSA treatment has a much more profound effect on gene expression dysregulation in the AC compared to the VG (Figure S5B). There are many more genes up and downregulated in the AC. This is quite a novel and interesting result. Can the authors comment on why?

We have performed HDAC inhibition analysis using two HDAC inhibitors, TSA and VPA, and noticed that TSA does induce many more genes in AC than VG cells when the data are compared to each other (Figure S5B). We speculate that this may be a TSA-specific effect because TSA is known to inhibit non-Class 1 HDACs. Therefore, we focused on the genes that are affected by both TSA and VPA. However, even with this analysis, we still found many more genes to be affected in AC than VG. Perhaps, this is due to the large-scale difference in timing of zygotic gene activation occurring earlier in AC than VG (Chen and Good, 2022, PMID: 36007528). Since this is largely a speculation, we simply report on the finding.

2. Timing of HDAC1 binding to the genome. It is interesting that HDAC1 peaks are widespread in the genome by Stage 8, presumably during zygotic genome activation, and that these peaks largely strengthen over time into the early gastrula. This however begs the question of when HDAC1 begins to bind to the genome. Is it already bound in the egg, or upon fertilization, or are peaks only detectable starting around the time of genome activation? Given that the protein is present in the egg and throughout cleavage stages (Figure 1A), and that the authors propose FoxH1 and Sox3 may be contributing to recruitment, the manuscript would be strengthened by characterization of the time of recruitment.

We performed a time-course ChIP-qPCR analysis of selected regions, representing 9 identified Hdac1 bound regions, and 2 negative control regions. Figure 1G suggests that Hdac1 does not bind to the genome prior to st8 (ZGA).

We also note that the number of Hdac1 binding sites at st8 is low (n=1,340) compared to those of Foxh1 (Charney et al., 2017, n=28,611), Sox3 (Gentsch et al., 2019, n=8,368), Vegt (Paraiso et al., 2019, n=21,711), and Foxi2 (unpublished, n=13,158) at the same stage. Additionally, Hdac1 binding strength also seems to be weak (Figure 1F) at st8. Based on this evidence, we suggest that the timing of genomic engagement of Hdac1 differs from that of the maternal transcription factors.

3. Differential role for HDAC1 and HDAC2 in lineage maintenance? The TSA treatment makes it difficult to disentangle specific subsets of genes modulated by promoter-proximal deacetylation. Can the authors contrast TSA vs. VPA inhibition experiments to identify peaks that could likely be attributed to HDAC2 rather than HDAC1, and perform GO analyses? Alternatively, can the authors provide direct binding data with ChIP-seq of HDAC2 in blastula and gastrula.

New Hdac2 ChIP-seq results are summarized in Figure S1E-H. Since we performed

Hdac2 ChIP-seq experiment once, the peak numbers of Hdac2 are greater than that of Hdac1, which underwent IDR analysis. Despite this, overall, Hdac2 binding is observed in the regions where Hdac1 binds. Importantly, Hdac2 binding is notably stronger around Hdac1-Hdac2 co-bound peaks than around Hdac2 lone peaks (Figure S1H). This finding is consistent with the previous findings that Hdac1 and Hdac2 can assemble into the same HDAC complexes (Seto and Yoshida, 2014, PMID: 24691964; Milazzo et al., 2020, PMID: 32429325).

We attempted Hdac1 MO knockdown (see Source Data), but were only able to reduce the Hdac1 protein level in st9 and st10.5 embryos, but not in st8 embryos, which is likely due to the timing of degradation of stores of maternally deposited Hdac1 protein (https://www.xenbase.org/entry/gene/geneExpressionChart.do?method=drawProtein&g eneId=865283&geneSymbol=hdac1.S&addProteins=hdac1) in the early embryo. Therefore, we are unable to address the specificity difference between Hdac1 and 2 at present time.

4. Direct evidence that HDAC1/2 binding contributes to gene expression regulation. The authors utilize inhibition of deacetylation (TSA) or HDAC1 specifically (VPA) to analyze changes to pan-H3K-acetylation and gene expression, regionally within the AC and VG. In general, I like this line of experiments, however, I wonder whether the TSA inhibition might be affecting other deacetylase enzymes, and what the effect would be of MO knockdown of the HDACs (or their recruitment via FoxH1, Sox3). Other groups have found that HAT-HDAC binding to promoters and enhancers does not always correlate to gene expression regulation. Extending this notion here, I worry that the authors may not be directly analyzing the downstream function of HDACs. Can the authors compare their differential expression data to existing gene profiling studies in embryos that have knockdown of FoxH1, Sox3, or HDAC1/2 to demonstrate similar profiles of dysregulation?

At present, Sox3 knockdown RNA-seq data is not available. Therefore, our analysis is focused on Foxh1 KD data. We compared the differentially expressed genes in Foxh1 knockdown and TSA-treated animal cap explants (Zhou et al., 2022, PMID: 35848281), and found that only a small subset of genes are co-regulated by Hdac1 (defined by TSA sensitive and Hdac1 genomic binding) and Foxh1. However, we found that Sox3 and Foxh1 double MO KD results on modest reduction (~2 to 3 fold on average) of Hdac1 binding (Figure 2F). Therefore, the recruitment of HDACs to individual genomic regions is likely to be complex, involving multiple maternal TFs. The result is not surprising as there could be multiple TFs that recruit Hdacs.

Reviewer #2 (Recommendations for the authors):

As a general reader, I do not know enough of the background knowledge in the field to know how novel the observations are, and it seems that most of these specific comparisons are made in the discussion, and it is not explicit what observations are really new, or whether the importance of the paper is in the documentation. The authors need to draw this out more clearly. Also, as a general reader without the ingrained knowledge of the activating or inhibitory functions of the different histone modifications, I initially found it impossible to hold onto the threads of the arguments but was able to do so by color coding the modification types as activating, inhibitory, or heterochromatin in the text. I would hope that modern journals might be able to use colored text to enable readers to read and evaluate the text more readily. Similarly, it is hard to hold on to the nature of different classes of response, so I suggest more descriptive names for the classes. Without these, the manuscript is very tough going for the general reader. Indeed, as a detailed examination of chromatin modification, it is inevitably dense. Again as a general reader, I would find it more useful to present the specific genes that convey the arguments in the main text and figures, so that there is something to get one's teeth into and move the venn diagrams and stacked bar graphs and violin plots, which I find uninformative for the specific arguments, but which are necessary for the generality of the claims, to the supplement. But these are suggestions to make the manuscript more accessible to the general reader. If the target audience is the specialized field of chromatin organization or detailed regulation of gene expression in developing animals (without going beyond inhibitor studies to the sequences responsible for the DNA recruitment or the enzymatic activity of the protein), then the paper would probably be more appropriate for a more focused journal. But if there are new insights that can be more clearly articulated, then the manuscript could be appropriate to eLife.

We have incorporated the suggested changes.

Major points, in no particular order.

1. While the Hdac genes are shown for mRNA it would be useful to see whether this matches the protein databases. Are they detectable, and does mRNA translate to protein abundance?

Currently, there is no database that would allow us to compare the protein levels between different HDACs directly in Xenopus. While mass spectrometry datasets (Peshkin et al., 2019; Nguyen et al., 2022) available in Xenopus laevis are quantitative, the data for each protein is relative abundance over developmental time points and doesn’t permit comparisons between the levels of different HADC proteins.

2. Line 140- were Vegt peaks found to be enriched in the maternal motifs? (Gentsch?)

Indeed, was there enrichment of other motifs associated with transcriptional activation as found by Gentsch et al? (POU-SOX (Pou5f3-Sox3 heterodimer), Krüppel-like zinc finger (ZF; Sp1 and several Klf), POU (Pou5f3), SOX (Sox3), bZIP (Max), FOXH (Foxh1), ETS (Ets2), NFY (NFYa/b/c), SMAD (Smad1/2), VegT (mVegT)), A lack of enrichment would illustrate a strong difference in recruitment and be informative.

We now include a list of the top 50 motifs (ranked by FDR) in our Supplementary File 1 for all 3 stages. At st8, POU motifs are highly enriched. Among various FOX motifs, Foxh1 motif is the most enriched. Some POU-SOX, SOX motifs are also present. At st9, SOX, POU, POU-SOX, and FOX motifs are highly enriched. ZIC (including a few KLF) and HOX motifs are also detected, but ranked below the top 100 enriched motifs. T-box (Eomes) motifs are captured below the top 200. At st10, FOX motifs are rare.

Instead, ZIC motifs (including ZIC, KLF, and SP) are highly enriched. SMAD, SOX and T-box (Eomes) are present within the top 150 enriched motifs. ETS, MAX, and NFY motifs are present below the top 200. The appearance of new motifs such as ZIC, Tbox, and ETS, and the disappearance of POU and FOX motifs at the later developmental stages is consistent with the model that different TFs dynamically recruit Hdac1 to the genome.

3. "Consistent with the overlapping binding of Ep300, Hdac1 peaks display a moderate level of H3K9ac (Cluster c and d), high levels of 164 H3K18ac (Cluster b, c, and d), H3K27ac (Cluster b, c, and d), and pan-H3 lysine acetylation (pan-H3Kac)."

Why is this consistent (ref required?) is Ep300 always an activating signal? The next section discusses activating signals, so maybe so, but it does not say so. What is the current understanding of Ep300? General enhancer binding?

Ep300 is a histone acetyltransferase (HAT) that catalyzes histone acetylation at H3K9, K18, K27 (Zhang et al., 2018, PMID: 30150647). Therefore, the presence of Ep300 is consistent with the presence of histone acetylation. In general, Ep300 is highly correlated with active enhancers, and so is histone acetylation (Heintzman et al., 2007, PMID: 17277777; Ong and Corces, 2011, PMID: 21358745). We clarified this point in the text (p9).

I fear that the paper is exceedingly hard to follow unless one has these identities imprinted in one's own brain, which I don't. Indeed, for people not in the immediate field, the nomenclature, with its acetylations, and multiple methylations is all too easy to forget. For ease of reviewing, I color-coded the histone marks according to Wikipedia. It made the manuscript much easier to follow, though I am sure my simplistic green or red color code is too simple. The color coding of the figures in the manuscript is more nuanced, and presumably reflects some deeper understanding. It would be useful to share this with the reader. Otherwise, the activating (green) repressive (red), and heterochromatin (brown, blue? - what are these?) and the main text should also match an informative color coding, which should be trivial in these days of digital typesetting.

Thank you for your suggestion. We added descriptive words, such as “activating” and “repressing” throughout the text to improve the clarity of the text. The ChIP-seq data under a specific color panel (e.g., Figure 3) represents specific ChIP-seq data sets (e.g., H3K9ac, H3K27ac, etc) from different developmental stages. This makes it easier to compare the ChIP-seq results across different experiments.

4. "we speculate that these regions are differentially marked in space due to heterogeneous cell populations present in the whole embryo. Hdac1 Cluster I peaks are referred to as heterogeneous CRMs."

Surely the authors can give explicit examples here for some known heterogeneously expressed genes to reduce the speculative nature of this group? (though they do eventually show up in the supplements- but why not progress the argument from the specific to the general?)

We now provide specific examples in the text as we discuss the data. We thank the viewer for this suggestion as this change has improved the clarity of the manuscript.

5. Ideally, the figures and legends should be interpretable independent of the text, for example, when wondering what abcd are in figures 1 and 3, one should not have to search in the text for the relevant explanation

Heatmaps in Figure 1C and Figure 3A and B, all include ChIP-seq data of Hdac1 segregated into 4 clusters (a, b, c and d). This was intended to reference all other relevant data against that of Hdac1 (a, b, c and d) clusters. We now more explicitly explain these panels in the text, figures, figure legends.

"Overall, a minority of Hdac1 peaks are unique to each of the blastula stages (Cluster a and b) while a majority of Hdac1 peaks are present across multiple stages (Cluster c) and at the early gastrula stage (Cluster d) (Figure 1C, S1D). "

And in reading this, "unique to each blastula stage", could be expressed more clearly to indicate their uniqueness- ie at stage 8 (cluster a, likely driven by maternal factors?) and stage 9, (cluster b, likely driven by newly expressed zygotic factors or signals?).

Re the italics – where are the genes known to be in these classes? Ie the Nodals, vs genes driven by zygotically activated signals like Smad2? It is important for a simple argument to give the general reader some specific biological context.

We now added the temporal expression profiles of genes associated with Hdac1 binding (Figure 1D). We focused our analysis on the expression of zygotically expressed genes (TPM <1 between 0 and 1 hour post-fertilization, and TPM >1 at any time in the window of 1- 23 hours post-fertilization), and excluded maternally expressed genes from this analysis.

In general, the timing of Hdac1 binding is correlated with the activation of genes. Cluster a (peaks uniquely present at st8) associated genes are activated at 4hpf (p=0.0035). Cluster b (peaks uniquely present at st9) associated genes are activated at 5hpf (p=8.56e-5). Cluster d (peaks uniquely present at st10.5) associated genes are also activated at 5hpf (p<2.2e-16). Interestingly, cluster c (peaks that are present in st8 and/or st9 and/or 10.5) associated genes exhibit activation at 4hpf (p=5.93e-8). In terms of gene expression levels, cluster a genes are least activated whereas cluster c genes remain activated at the highest level after 5hpf. The expression levels of cluster d

genes gradually exceed that of cluster b as development proceeds. We have briefly summarized these findings in the text p6.

6. Likewise "These observations suggest that Hdac1-mediated suppression is largely dictated by developmental programs" – perhaps the authors might give some examples in the text.

We modified this sentence and state, “Hdac1 binds to facultative heterochromatic regions facilitating the repression of genes”. This modification is needed as we do not have a molecular understanding of how it works and therefore decided the above change would be appropriate.

7. "First, an increase of pan-H3Kac is observed across Hdac1-bound CRMs, including active CRMs, consistent with the canonical enzymatic activity of Hdac1" remind us what H3 Kac is for – it seems to be in the purple, brown and red categories. Is it associated with transcription start sites? It's not clear what antibody was used. It seems it must be a pan H3Kac antibody?

We now state in the text that pan-H3Kac antibody detects histone 3 K9, K14, K18, K27, and K56 acetylation. Regarding the question about H3Kac being associated with open chromatin that is transcriptionally active, we cannot conclude that all these acetylation are associated with transcription start sites. This is because we have used pan-H3Kac antibody, which recognizes various forms of acetylation on H3.

8. What are the effects of Hdac inhibitors on the transcriptome? Seems that this is an essential piece of information to interpret the results on chromatin? Seems there is a lot of literature on this, and so far as I can tell it is not cited until the discussion. It would be nice to have a digested understanding to compare to the chromatin results in the Results section (some of this appears eventually, but late enough to leave the reader frustrated when the topic comes up).

We stated on p4, ln 75-77, that HDAC inhibition in cell lines results in both up- and down-regulation of genes, with several papers cited.” We have also modified sentences (p4, ln 75-80) to better describe these experiments.

9. I had great difficulty understanding the paragraph that begins "Given that Hdac1 CRM clusters (Clusters I-IV in Figure 3C) are subjected to both active and repressive epigenetic modifications presumably in different germ layers, we compared the general status of H3 acetylome (pan-H3Kac) between two distinct germ layers" I think that all of this is VPA and TSA independent information? But in the end, isn't it all circular? Ie higher enrichment of pan H3Kac associated with higher expression?

In Figure 3C, we focus on H3K27 modification. Lysine 27 can be either acetylated or methylated, but cannot accommodate both modifications simultaneously on the same histone polypeptide. However, we found a cluster (cluster I) of regions marked with both modifications. For instance, Figure S3C shows that the surrounding regions of ectodermally expressed lhx5 (Houston and Wylie, 2003, PMID: 12736213) are marked with both active H3K27ac and repressive H3K27me3. Our hypothesis is that lhx5 is marked by active H3K27ac in the ectoderm cells, but lhx5 in endoderm cells is marked by repressive H3K27me3. This ensures that lhx5 is specifically expressed in ectoderm, but not in endoderm. This hypothesis is confirmed by the analysis of active pan-H3Kac data using dissected AC and VG explants (Figure S4.2F). CRMs of lhx5, kcdt15, and foxi1 are acetylated (pan-H3Kac) in AC tissue, but not in VG tissue. We also find foxa4, mixer, sox17a, vegt are acetylated in VG, but not in AC. We have modified p12 and added a new sentence (ln 248-251) and a figure (FigS4.2C) to clarify the statement.

10. "Taken together, these data demonstrate that Hdac1 maintains differential H3 acetylation states between germ layers through its catalytic activity." Is a rather bland statement after trudging through the paragraph. I think that one can make the general statement that upon inhibition of HDAC, repressed genes have a larger increase in their pan H3Kac than do active genes. And is there evidence that it is through the catalytic activity rather than its binding?

Thank you for the suggestion, and we have modified the text (see p14, ln815-816). In addition, we investigated the binding of Hdac1 upon TSA inhibition. Figure S4.1C and

D show that the genome binding ability of Hdac1 is largely insensitive to TSA treatment.

Hence, the changes of histone acetylation patterns is due to the loss of the enzymatic activity of HDAC and not the result of binding alone.

"we found that the fold increases of pan-H3Kac signals after TSA treatment are negatively correlated with the levels of endogenous pan-H3Kac signals in both AC and VG explants" This is an unnecessarily complicated sentence. It's already hard to figure out the direction things are going in without having to think about what negative correlation means in these contexts. But I think it means what I just suggested?

We have modified the sentence and it now states,

“Interestingly, CRMs with low endogenous levels of histone acetylation tend to be more responsive to HDAC inhibition, than CRMs with high acetylation” (p 13, ln 607-608).

11. I ask that the authors consider how much time would be spent by the reader looking back to see what Class1,2 and 3 genes are. Could they not be more readily described as "selectively activated", "constitutively repressed" and "Constitutively activated" if these were designated SA, CR, and CA that might be easier to remember/interpret, especially for me who can't remember class I and 2 from one paragraph to the next. Or consider the value of spelling them out each time, at least for the reviewer. Once accepted, one can return to uninformative abbreviations and let the reader suffer.

Our original intention was to provide the reader the opportunity to interpret the data without our preferences. However, we realize the difficulty of keeping track of these uninformative names. We now added a short description for each gene class when it is helpful.

12. What is the mechanism of Hdac inhibition by VPA and TSA? Does the Hdac dissociate from chromatin, allowing access to acetylases? Or does it stay on the chromatin in an inactive form, suggesting a more complex mechanism of the differential effects of inactivation? What do the data say about the potential mechanism?

TSA chelates the Zn2+ cation in the active site of histone deacetylase enzymes (Finnin et al., 1999, PMID: 10490031) and inhibits the enzymatic activity. The mechanism by which VPA inhibits HDAC activity still remains obscure. VPA has been shown to block the interaction between Sp1/Sp3 zinc finger TFs and Hdac1/2 (Davie 2003, PMID: 12840228), and to induce ubiquitin-mediated Hdac2 proteolytic degradation (Kramer et al., 2003, PMID: 12840003). ChIP-seq analysis of Hdac1 after TSA inhibition suggests that TSA generally does not interfere with its association to the genome (Figure S4.1).

Reviewer #3 (Recommendations for the authors):

Figure 1

1C: St101/2 or stage 9-specific HDAC1 peaks seem to already show signal at stage 8. (Difficult to grasp relative level as no indication of what the blue scale is….). In any case, is this low level at stage 8 meaningful (how does it look in all enhancer peaks for example).

We have now assigned scale bars to all heatmaps. At present, we are unable to conclude whether the weak signals (clusters b and d at stage 8) are meaningful as the peaks were not statistically significant. However, we note that these weakly bound regions gradually gain binding strength over time. We, therefore, described the binding dynamic of Hdac1 to be progressive (page 7, ln 162).

What is the temporal expression (early blastula to neurula stages for example) of the a-d set of HDAC1-associated genes? Any correlation between stage-specific binding and expression pattern would strengthen the proposed involvement of HDAC1 in the epigenetic regulation of genes around ZGA.

Figure 1D describes the correlation between the Hdac1 binding and the temporal expression profiles of Hdac1-associated zygotic genes in cluster a-d. In general, the timing of Hdac1 binding correlates with the activation of associated genes. Genes associated with cluster a are activated first (4 hpf), and then cluster b and d associated genes (5 hfp). Cluster c associated genes are typically activated between 4-5 hpf.

Figure 2:

2B What are the enriched motifs in stage 8 and stage 10 specific peaks?

We now include a list of the top 50 motifs (ranked by p-value) found in stages 8, 9 and 10.5 (see the source data).

2C: How does the HDAC1 peak intensity look if you integrate the HDAC1 chip a-amanitin data in the map from 1C (in complement to S2B).

Pearson correlation analysis comparing Hdac1 peak intensity between uninjected control and a-amanitin injected embryo shows that there is no significant difference between these samples (Figure 2A).

Charney et al. document binding of TF pre/post ZGA.

To support the ID that TF binding is associated with HDAC1 binding it would be important to check if stage-specific binding of HDAC1 is associated with stage-specific binding of TF.

Figure S2I shows heatmaps of the Hdac1 peaks at stage 8, 9 and 10.5 with Foxh1 and

Sox3 peaks. Hdac1 peaks overlap well with Foxh1 and Sox3 ChIP-seq peaks. The

Foxh1 and Sox3 peaks at cluster b, c and d precede that of Hdac1. For instance, while Foxh1 and Sox3 peaks are present in stage 8 and 9 samples, Hdac1 speaks appear at stage 9. Similarly, Foxh1 and Sox3 peaks in cluster c are present in stage 8, Hdac1 peaks appear at stage 9 and persist. Lastly, the peaks of Hdac1 at stage 10.5 appears after the appearance of Foxh1 and Sox3 peaks at stage 9. These data is consistent with the view that Hdac1 binding occurs after the appearance of Foxh1 and Sox3 binding.

The title "HDAC1 binding is instructed maternally" is not appropriate. Interference with maternal TFs would be needed to go beyond correlation. HDAC1 binding after foxh1MO (as in Charney et al) will indicate if HDAc1 binding depends on the presence of this maternal TF.

Here we provide two pieces of evidence that support the statement. First, inhibition of RNA pol II activity by a-amanitin proves that Hdac1 binding to the genome at blastula stage is independent of zygotic factors (Figure 2A, Figure S2A-C), and therefore its recruitment to the DNA must be instructed by factors supplied in the egg/embryo maternally. Based on this evidence alone, we think the current heading of this section is appropriate. Second, Hdac1 binding after Foxh1 and/or Sox3 MO injection is reduced in foxh1 and sox3 morphants (Figure 1F). The partial knockdown of Hdac1 recruitment in the absence of maternal Foxh1 and Sox3 is consistent with the model but additional TFs are likely involved in its genomic recruitment.

Figure 3

"moderate" level of K9ac etc….moderate compared to what??? I would think comparing for example level of p300+ HDAC+ to p300+HDAc- would give an actual idea if the level on HDAC associated p300 peak stands out.

The same goes for figure S3B.

Without such kind of comparison, it is difficult to conclude that HDAC1 peaks have biased epigenetic status compared to all p300 peaks for example.

Again in Figure 3B HDAc1 bound region exhibits a strong signal density of K27me3.

We agree with the referee’s comments and have removed quantitative words in our description and simply state the presence of epigenetic marks at Hdac1 binding sites.

Line 183: I don't think cluster b fits the description.

We modified the sentence. Instead of stating, “Hdac1 peaks (Cluster b, c, d, but not a) are marked by both active and repressive epigenetic signatures…”, we changed the sentence to “a majority of Hdac1 peaks (cluster c and d) are marked by both active and repressive epigenetic signatures…”

Fig3CD why exclude HDAC peaks from stage 8?

Since these peaks are free of notable epigenetic signatures (e.g., Ep300, RNA pol2, and histone acetylations and methylations), it is unclear what the physiological relevance/function of these peak regions during germ layer specification. Therefore, we excluded them from the current analysis.

It would be useful to see how HDAC peaks signal looks on Figure 3D map.

We now include Hdac1 peak data (st9 and st10.5) in Figure 3D. We have also modified the text to better describe the behavior of cluster 1.

In 3D there seems to be a clear anti-correlation between K27me3 and K27ac signal (see cluster I) that does not really fit the description of line187-191. This is important because the concept proposed in the manuscript relies on these co-occurring K27ac/me sites. Are the signal anti-correlated in cluster I?

Regarding the deposition patterns and density of H3K27me3 and H3K27ac being different, we offer the following explanation. First, H3K27ac peaks are called using a NarrowPeak format, which identifies narrow sharp peaks resembling those of transcription factors. On the other hand, H3K27me3 marks are found using a BroadPeak format because methylation signal on H3K27 is generally broader (Zhou et al., 2011, PMID: 21116306). The difference in peak calling will impact both the peak position and the overall peak intensity of H3K27ac and H3K27me3. This may result in peaks that look different in the heatmap. However, when these regions are examined (Figure S3C), the overlapping regions are readily visible (see response below).

S3C K27ac and me do not seem to overlap much. Can we see the region where peaks are detected (add a box where peaks are detected)?

Figure S3C shows the gene browser view of various regions (specific peak overlaps are now boxed). There are significant peak overlaps between H3K27ac and H3K27me3 around lhx5 (Class I), but not in Class II, III and IV genes.

Line 203 – before testing of this hypothesis is surely again to compare acetylation in p300+ HDAC+ versus p300+ HDAC- (or at TSS) to evaluate if TSA primarily affects HDAc1 sites?

To demonstrate that Hdac1 bound regions are preferentially affected by TSA, we randomly sampled 23,442 genomic regions (an equivalent Hdac1 peak number) and compared the changes of pan-H3Kac in these randomized regions vs. Hdac1 peaks. First, we observed that the change of pan-H3Kac is lower in randomized regions than that of Hdac1 peaks (Figure 4E, S4.3A). We next examined the levels of pan H3Kac across clusters (Figure S4.3E, S4.3F) as well as fold changes of pan-H3Kac after TSA treatment (Figure 4F, S4.3D). TSA affected Hdac1 bound regions more than randomized genomic regions in both cases. Based on these analyses, we conclude that TSA preferentially targets Hdac1-bound regions to alter the levels of pan-H3Kac.

Figure 4A/B

Can the author use a TSA control experiment where TSA is applied only prior to ZGA or only around ZGA?? This will indicate what is the time window when TSA has this effect.

It would be also important to document the effect on the development of whole embryos of the TSA treatment scheme used on explant (Figure 4C). Are the embryos defective in these conditions?

The suggested experiment is difficult to conduct because we do not know how quickly (or how deeply) HDAC inhibitors diffuse into Xenopus embryos and can be washed away after the treatment. Since the Xenopus tropicalis embryos progress from stage 8 to 10.5 in about 3 hours, it would be very challenging to perform treat-and-wash-away experiments in such a tight developmental time window.

4C:

What is the effect of TSA on the number of peaks detected?

Would be very important to have a heat map that summarizes the panAc-peak in control and TSA conditions as well as the HDAc1 peak. This would illustrate to which extent TSA primarily affects HDAC binding sites.

Based on the TSA-induced global hyperacetylation (shown by westerns in Figure 4B, and quantitative ChIP-seq in Figure 4E, S4.3A), we expect to detect more peaks in TSAtreated samples than in DMSO-treated samples. Indeed, spike-in normalized peak calling revealed that TSA treatment not only induced new peaks in AC or VG explants (Figure S4.3B, S4.3C, cluster b), but also elevated common pan-H3Kac signal (a).

AC and VG-specific peaks seem to exhibit a much lower intensity of panAc signal than common (scale missing on the map…). What could be the reason for such a difference?

We are a bit confused with this question. If this referee refers to the intensity difference, yes he/she is correct to notice that. Consistently, the common cluster (cluster A) displayed a stronger H3K9ac and H3K27ac signal density than the germ layer specific clusters B and C. We also found that genes associated with cluster A are expressed significantly higher (p< 2.2e-16) than genes associated with germ layer specific clusters B and C, shown in Figure S4.2C. This is consistent with the observation that histone acetylation levels generally positively correlate with gene expression levels. We briefly state this observation in the text (p 12, line 555). Perhaps, the genes belonging to cluster A have higher acetylation density because these genes need to be expressed at high levels. A scale bar is now added to Figure 4C. Alternatively, if the referee is asking why the AC and VG intensities, presumably in cluster A, are significantly weaker that in whole embryos, we do not know the reason why this is the case. Additionally, this is not a major point of this paper, and we prefer not to comment on this issue in the paper.

4E FC I assume is FC TSA / control.

You are correct, and the heading of Fig4F is better labeled now.

Fig4E and S4G

Global increase in acetylation in embryos (4D) – immediate question whether acetylation on the genome disproportionately affects HDAC1 binding sites?

This is critical since the authors assume that TSA effect indeed reflects the effect on HDAC1 activity. Authors, unfortunately, focus analysis solely on HDAc1 binding sites so the conclusion line249 is not yet supported.

We performed Hdac1 ChIP-seq in TSA-treated embryos at st9 and st10.5 and found that TSA treatment has little effect on the Hdac1 genomic binding profiles (Figure S4.1). Therefore, the elevated acetylation by TSA does not alter the DNA binding ability of Hdac1.

All observations could also fit the model whereby TSA leads to the homogeneous increase of ac level genome-wide – would obviously lead to lower fold change if starting point high (cluster III) and higher if low (cluster II).

Our data do not support a model of homogeneous increase. We examined a fold change as shown in Figure 4F and Figure S4.3D. CRMs of Hdac1 clusters I-IV respond differentially upon HDAC inhibition in both animal and vegetal explants. We also plotted the difference (subtraction) of normalized counts of pan-H3Kac (Figure S4.3E and S4.3F). Compared to randomized genomic regions, Hdac1 all peaks displayed a much higher level of differences in pan-H3Kac, suggesting that TSA inhibition does not lead to a homogenous increase of histone acetylation level genome-wide; instead, it had a more profound effect at genomic regions bound by Hdac1.

Figure 5

When excluding genes with maternal transcripts how many genes are left in each category?

We added the numbers of genes examined in both Figures 5B and 5C.

How do sets of genes with the same epigenetic configuration but HDAC- compare (i.e genes that are K27me3/K27ac but without HDAC1?).

This is an interesting question, but somewhat difficult to execute because a large number of genes bear multiple CRMs, of which each CRMs can have different epigenetic signatures. Therefore, we first excluded genes that have multiple CRMs bearing various epigenetic signatures (e.g., H3K27me3 and H3K27ac), and examined the expression of the remaining genes. Class I and III genes that are bound by Hdac1 are preferentially expressed at the blastula and gastrula stage, compared to the class I’ and III’ genes that are devoid of Hdac1 peaks (Figure S3E, 4-8 hpf).

VPA analysis cannot support the claim that TSA treatment has no off-target effect because this (i) is not RNA-seq analysis and (ii) 8/24 genes selected for RTqPCR are not showing the expected effect.

We now include RNA-seq results of VPA-treated embryos. Overall, a large number of genes affected by TSA are also affected by VPA (Figure S5A). When we compared the expression levels of genes affected by TSA alone, VPA alone, or both VPA and TSA, similar trends of gene expression changes are observed (Figure S5B). For the spatial gene expression change analysis upon HDAC inhibition, we focused on genes coaffected by TSA and VPA to minimize non-specific effects.

The first important question to address: are the TSA-affected genes disproportionally enriched for HDAC peaks associated genes?

In Figure S5D, we show the frequency of genes that are affected by TSA alone, VPA alone, and both VPA and TSA treatments. Overall, a majority of genes (>50%) affected by Hdac1 inhibition have functional Hdac1 binding site nearby. HDAC inhibitors had the most impact on the genes that are affected by both VPA and TSA.

5E Not sure to get the significance of observation.

Can it be that a gene can be detected as upregulated only if they have a low level in control conditions can be detected as downregulated only if they have a high level in control conditions? Which is the rationale for the selection of Ac or V-specific genes…?

In this figure (now Figure 5F), we aim to address whether the expression of genes differentially regulated by TSA shows spatially restricted expression patterns. For example, are endodermally expressed genes preferentially upregulated in ectodermal lineage cells upon TSA treatment? To tackle this question, three components are needed: a list of genes classified as upregulated in ectodermal cells upon TSA treatment (Up in AC in Figure 5F), a list of genes expressed in endoderm (labeled as VG) in normal embryos (based on Blitz et al., 2017), a list of genes that are preferentially expressed in AC or VG (see the description of statistical treatment below).

The method of assigning genes to spatial expression patterns was published previously

(Paraiso et. al 2019; PMID: 31167141) and also described in Materials and methods- Categorical Analyses. Briefly, genes expressed less than 1 TPM are categorized as not expressed (NE), genes whose coefficient of variation is less than 10% among 3 germ layers are considered not locally expressed (NL), genes whose expression is highest in animal cap (ectoderm) explants are classified as animal cap enriched genes (AC), genes whose expression is highest in marginal zone (mesoderm) explants are classified as marginal zone enriched genes (MZ), and genes whose expression is highest in vegetal mass (endoderm) explants are classified as vegetal mass enriched genes (VG). For instance, the expression level of foxi1 is 107 TPM in the animal cap, 10.6 TPM in the marginal zone, and 1.2 TPM in vegetal mass. Hence, foxi1 is an animal cap enriched (AC) gene (and is in the salmon red area in Figure 5F). This assignment is not perfect, but practical as foxi1 is a known regulator for ectoderm specification (MatsuoTakasaki et al., 2005, PMID: 16079156; Mir et al., 2007, PMID: 17229765).

We applied Fisher’s exact test to examine the proportional significance between two groups of genes. For instance, the first bar (All genes) of Figure 5F shows among 28,850 annotated genes, 6,170 genes (VG, 21.4%) are endodermally expressed genes at this stage of development. The second bar shows among 155 genes that are upregulated after TSA treatment, 71 genes (VG, 45.8%) are endodermal specific genes (VG). Fisher’s exact test shows that the enrichment of endodermal gene upregulation in anima cap (ectodermal cells), that is 45.8% (test group), is statistically significant over that from all annotated genes (21.4%, reference group), resulting in a p-value of 1.32e11. We conclude that endodermal genes are preferentially upregulated in the animal cap cells upon HDAC inhibition, and the data is consistent with the view that HDAC activity is required to safeguard misactivation of spatially restricted genes.

https://doi.org/10.7554/eLife.79380.sa2

Article and author information

Author details

  1. Jeff Jiajing Zhou

    Department of Developmental and Cell Biology, University of California, Irvine, Irvine, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing - original draft
    Competing interests
    No competing interests declared
  2. Jin Sun Cho

    Department of Developmental and Cell Biology, University of California, Irvine, Irvine, United States
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  3. Han Han

    Department of Developmental and Cell Biology, University of California, Irvine, Irvine, United States
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  4. Ira L Blitz

    Department of Developmental and Cell Biology, University of California, Irvine, Irvine, United States
    Contribution
    Resources, Writing - original draft, Writing - review and editing
    Competing interests
    No competing interests declared
  5. Wenqi Wang

    Department of Developmental and Cell Biology, University of California, Irvine, Irvine, United States
    Contribution
    Resources, Investigation
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-4053-5088
  6. Ken WY Cho

    1. Department of Developmental and Cell Biology, University of California, Irvine, Irvine, United States
    2. Center for Complex Biological Systems, University of California, Irvine, Irvine, United States
    Contribution
    Conceptualization, Supervision, Funding acquisition, Investigation, Writing - original draft, Writing - review and editing
    For correspondence
    kwcho@uci.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-7282-1770

Funding

National Institute of General Medical Sciences (R01GM126395)

  • Ken WY Cho

National Institute of General Medical Sciences (R35GM139617)

  • Ken WY Cho

National Science Foundation (1755214)

  • Ken WY Cho

National Institute of General Medical Sciences (R01GM126048)

  • Wenqi Wang

American Cancer Society (RSG-18-009-01-CCG)

  • Wenqi Wang

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

Acknowledgements

We thank Drs. Kyoko Yokomori (University of California, Irvine), Yongsheng Shi (University of California, Irvine), and current Cho lab members for insightful comments on this study. We thank the Genomic High Throughput Facility at the University of California, Irvine for sequencing services. We also thank the Research Cyberinfrastructure Center at the University of California, Irvine for the ongoing support of High Performance Community Computing Clusters. This work is supported by NIH R01GM126048 and ACS RSG-18-009-01-CCG to WW and NIH R01GM126395, R35GM139617 and NSF 1755214 to KWYC.

Ethics

All of the animals were handled according to approved Institutional Animal Care and Use Committee (IACUC) protocols (#AUP-21-068) of the University of California, Irvine.

Senior Editor

  1. Marianne E Bronner, California Institute of Technology, United States

Reviewing Editor

  1. Matthew C Good, University of Pennsylvania Perelman School of Medicine, United States

Version history

  1. Received: April 9, 2022
  2. Preprint posted: May 6, 2022 (view preprint)
  3. Accepted: March 24, 2023
  4. Accepted Manuscript published: March 27, 2023 (version 1)
  5. Version of Record published: April 6, 2023 (version 2)

Copyright

© 2023, Zhou et al.

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.

Metrics

  • 731
    Page views
  • 131
    Downloads
  • 1
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Jeff Jiajing Zhou
  2. Jin Sun Cho
  3. Han Han
  4. Ira L Blitz
  5. Wenqi Wang
  6. Ken WY Cho
(2023)
Histone deacetylase 1 maintains lineage integrity through histone acetylome refinement during early embryogenesis
eLife 12:e79380.
https://doi.org/10.7554/eLife.79380

Share this article

https://doi.org/10.7554/eLife.79380

Further reading

    1. Developmental Biology
    Chhavi Sood, Md Ausrafuggaman Nahid ... Sarah E Siegrist
    Research Article

    Neuroblasts in Drosophila divide asymmetrically, sequentially expressing a series of intrinsic factors to generate a diversity of neuron types. These intrinsic factors known as temporal factors dictate timing of neuroblast transitions in response to steroid hormone signaling and specify early versus late temporal fates in neuroblast neuron progeny. After completing their temporal programs, neuroblasts differentiate or die, finalizing both neuron number and type within each neuroblast lineage. From a screen aimed at identifying genes required to terminate neuroblast divisions, we identified Notch and Notch pathway components. When Notch is knocked down, neuroblasts maintain early temporal factor expression longer, delay late temporal factor expression, and continue dividing into adulthood. We find that Delta, expressed in cortex glia, neuroblasts, and after division, their GMC progeny, regulates neuroblast Notch activity. We also find that Delta in neuroblasts is expressed high early, low late, and is controlled by the intrinsic temporal program: early factor Imp promotes Delta, late factors Syp/E93 reduce Delta. Thus, in addition to systemic steroid hormone cues, forward lineage progression is controlled by local cell-cell signaling between neuroblasts and their cortex glia/GMC neighbors: Delta transactivates Notch in neuroblasts bringing the early temporal program and early temporal factor expression to a close.

    1. Developmental Biology
    Nicolas G Brukman, Clari Valansi, Benjamin Podbilewicz
    Research Article Updated

    The fusion of mammalian gametes requires the interaction between IZUMO1 on the sperm and JUNO on the oocyte. We have recently shown that ectopic expression of mouse IZUMO1 induces cell-cell fusion and that sperm can fuse to fibroblasts expressing JUNO. Here, we found that the incubation of mouse sperm with hamster fibroblasts or human epithelial cells in culture induces the fusion between these somatic cells and the formation of syncytia, a pattern previously observed with some animal viruses. This sperm-induced cell-cell fusion requires a species-matching JUNO on both fusing cells, can be blocked by an antibody against IZUMO1, and does not rely on the synthesis of new proteins. The fusion is dependent on the sperm’s fusogenic capacity, making this a reliable, fast, and simple method for predicting sperm function during the diagnosis of male infertility.