Establishment of developmental gene silencing by ordered polycomb complex recruitment in early zebrafish embryos

Vertebrate embryos achieve developmental competency during zygotic genome activation (ZGA) by establishing chromatin states that silence yet poise developmental genes for subsequent lineage-specific activation. Here, we reveal the order of chromatin states in establishing developmental gene poising in preZGA zebrafish embryos. Poising is established at promoters and enhancers that initially contain open/permissive chromatin with ‘Placeholder’ nucleosomes (bearing H2A.Z, H3K4me1, and H3K27ac), and DNA hypomethylation. Silencing is initiated by the recruitment of polycomb repressive complex 1 (PRC1), and H2Aub1 deposition by catalytic Rnf2 during preZGA and ZGA stages. During postZGA, H2Aub1 enables Aebp2-containing PRC2 recruitment and H3K27me3 deposition. Notably, preventing H2Aub1 (via Rnf2 inhibition) eliminates recruitment of Aebp2-PRC2 and H3K27me3, and elicits transcriptional upregulation of certain developmental genes during ZGA. However, upregulation is independent of H3K27me3 – establishing H2Aub1 as the critical silencing modification at ZGA. Taken together, we reveal the logic and mechanism for establishing poised/silent developmental genes in early vertebrate embryos.

To identify candidate TFs that might bind selectively at the promoters of particular clusters, we analyzed the DNA sequences flanking the transcription start site (TSS) (500 bp) at each cluster using the motif finding program, HOMER ( Figure 1E; Heinz et al., 2010). Largely non-overlapping motifs were identified for TF-binding sites at clusters linked to developmental vs. housekeeping genes ( Figure 1E, clusters 1 and 2 vs. 3 and 4; partitioned by H2Aub1/Rnf2 levels). Here, the strong enrichment of motifs for homeodomain-containing TFs (and other families) in clusters 1 and 2 provides (histone modifications/variant) at promoters (UCSC refseq). For DNAme, red color indicates regions that lack DNAme. (D) Genome browser screenshots of ChIP-seq enrichment at representative genes from the clusters in panel (C). (E) Transcription factor motif enrichment from HOMER (Heinz et al., 2010) at promoter clusters from (C).
The online version of this article includes the following source data and figure supplement(s) for figure 1: Source data 1. Uncropped western blots for panel A.       H3K4me1 peak summits located outside of promoters). DNAme heatmap displays WGBS fraction-methylated scores (note: red color indicates regions that lack DNAme). ChIP-seq heatmaps display log2(ChIP/input) scores. (B) Genome browser screenshots of ChIP-seq enrichment at representative loci from the indicated K-means clusters in (A). (C) Transcription factor motif enrichment from HOMER (Heinz et al., 2010) at enhancer clusters from (A). (D) Figure 2 continued on next page candidate factors that may help recruit PRC1 to developmental loci. DNA-methylated promoters at ZGA (cluster 5) represent a large and heterogeneous set of genes, which are activated in particular cell types later in development.

Enhancer poising parallels features and factors at promoters
Analysis of enhancer regions revealed features that were similar to those at developmental promoters ( Figure 2). Specifically, enhancers with high H2Aub1 and Rnf2 during preZGA and ZGA acquired robust H3K27me3 during postZGA (Figure 2A Figure 2C), and these sites overlapped partly with those enriched at promoters, consistent with the expectation that the factors that recruit histone modifiers to promoters and enhancers partially overlap. Taken together, enhancers acquire H3K27me3 during postZGA in proportion to their levels of Rnf2 and H2Aub1 during preZGA and ZGA, consistent with our observations at promoters.

A unique enhancer class with high H3K27ac and DNA methylation
Curiously, enhancer cluster 5 ( Figure 2A) was unique at postZGA -displaying high H3K4me1, very high H3K27ac, and open chromatin (via ATAC-seq analysis; Figure 2-figure supplement 1C,D)but bore DNA methylation -an unusual combination given the typical strong correlation between high H3K4me1 and DNA hypomethylation. Notably, Nanog-binding sites were highly enriched solely at cluster 5, and Nanog ChIP-seq during ZGA and postZGA (Xu et al., 2012) showed Nanog occupancy highly and selectively enriched at cluster 5 relative to other enhancer clusters ( Figure 2C-E). Thus, cluster 5 enhancers may utilize Nanog and H3K27ac to open and poise these DNA-methylated enhancers for later/subsequent transition to an active state. Consistent with this notion, GO analysis of cluster 5 enriches for terms related to developmental and signaling processes (p-value; 5.1E-18) (

The PRC2 component Aebp2 is coincident with Rnf2 and H2Aub1 at developmental loci
The pre-marking of developmental genes with H2Aub1 and Rnf2-PRC1 prior to H3K27me3 establishment raised the possibility of a 'non-canonical' (nc) mode of recruitment, involving ncPRC1 action (H2Aub1 addition) followed by the recruitment of PRC2 (ncPRC2) -via H2A/Zub1 recognition -to deposit H3K27me3. This mode and order of recruitment has precedent in Drosophila and in mammalian embryonic stem (ES) cell cultures, with the Aebp2 and Jarid2 protein components of ncPRC2 recognizing H2Aub1 and both targeting and facilitating H3K27me3 addition Blackledge et al., 2020;Cooper et al., 2014;Cooper et al., 2016;Kalb et al., 2014;Kasinath et al., 2021;Tamburri et al., 2020). We then addressed whether establishment of H3K27me3 during postZGA is mediated by the non-canonical Aebp2-Jarid2-PRC2 complex at loci pre-marked with H2Aub1. Interestingly, Aebp2 protein levels were very low during preZGA and ZGA stages, but robustly detected postZGA ( Figure 3A), without a large increase in aebp2 transcript levels ( . Taken together, these results suggest that translational upregulation and/or protein stability enables Aebp2 protein accumulation postZGA -enabling the 'reading/binding' of H2Aub1, and H3K27me3 deposition during postZGA by PRC2. Features of an enhancer cluster with exceptionally high H3K27ac and Nanog binding. K-means clusters generated in (A) were utilized to plot heatmaps of Nanog and H3K27ac. (E) A genome browser screenshot depicting Nanog and H3K27ac ChIP enrichment at a DNA-methylated enhancer from cluster 5 in (D).
The online version of this article includes the following figure supplement(s) for figure 2:     To investigate the possible additional contribution of Jarid2 to 'reading/binding' of H2Aub1 (leading to H3K27me3 deposition), we performed ChIP-seq of Jarid2 during postZGA (Figure 3figure supplement 2A). Here, profiling of Jarid2 occupancy by ChIP-seq resulted in chromatin maps with only modest enrichment and dynamic range, providing 448 bound promoters (Figure 3-figure  supplement 2B). Comparison of Aebp2 and Jarid2 occupancy at promoters revealed that the majority of Jarid2-binding sites (295/448) overlap with Aebp2-bound sites (Figure 3-figure supplement 2B). Only a minority of Aebp2-bound promoters overlapped with Jarid2 -an observation that may reflect our modest ChIP efficiency, but could also reflect the presence of Aebp2-bound promoters that lack Jarid2 binding. Notably, promoters bound by both Aebp2 and Jarid2 had enrichment of GO-term categories corresponding to developmental genes ( Figure 3-figure supplement 2C, D). Conversely, promoters bound solely by Jarid2 were not associated with developmental functions and instead enriched for ribosomal genes (Figure 3-figure supplement 2C). Analysis of Aebp2 and Jarid2 occupancy at enhancer loci revealed similarities to our analysis at promoters. Here, the majority of Jarid2bound enhancers (84/163) were also bound by Aebp2 (Figure 3-figure supplement 2E, F). However, this overlap accounted for only a minority of the Aebp2-bound enhancers. As our Aebp2 ChIP-seq exhibited greater robustness and dynamic range than Jarid2, we hereafter utilized Aebp2 occupancy as the primary functional marker for non-canonical PRC2 complex in the remainder of our analyses.

Loss of H2Aub1 via Rnf2 inhibition prevents Aebp2 localization and H3K27me3 deposition
To functionally test whether H2Aub1 recruits Aebp2-PRC2 for de novo establishment of H3K27me3 at developmental genes, we utilized the RNF2 inhibitor, PRT4165 (Chagraoui et al., 2018;Ismail et al., 2013;Zhu et al., 2018). PRT4165 is a small molecule inhibitor of Rnf2 that has previously been shown to strongly reduce H2Aub1 modification, but not to affect the activity of related H2A E3 ligases such as Rnf8 and Rnf168 (Chagraoui et al., 2018;Ismail et al., 2013;Zhu et al., 2018). In each of three biological replicates, PRT4165 treatment (150 μM) from the one-cell stage onward largely eliminated H2Aub1 by 4 hpf (ZGA) ( Figure 4A and B), and conferred a developmental arrest that resembled untreated 4 hpf embryos (Figure 4-figure supplement 1A). To determine whether the loss of H2Aub1 conferred loss of Aebp2 genomic targeting, we performed ChIP experiments on Aebp2 in PRT4165-treated and DMSO-treated embryos (three biological replicates per condition). Remarkably, developmental loci that normally display high Aebp2 in untreated or DMSO-treated embryos lost Aebp2 binding following PRT4165 treatment ( Figure 4C, clusters 3 and 4; Figure 4figure supplement 1B, C). Curiously, PRT4165 treatment also conferred many new/ectopic Aebp2 peaks ( Figure 4C, clusters 1 and 2), however our profiling of H3K27me3 following PRT4165 treatment (three biological replicates) revealed that new/ectopic Aebp2 sites did not acquire H3K27me3 ( Figure 4C, clusters 1 and 2; Figure 4-figure supplement 2B), consistent with prior observations that other modifications (such as H2Aub1; Kalb et al., 2014) may be needed to stimulate H3K27me3 addition by PRC2. Importantly, treatment with PRT4165 eliminated or strongly reduced H3K27me3 at virtually all loci normally occupied by Aebp2 and H3K27me3 ( Figure 4C, clusters 3 and 4, Figure 4figure supplement 1D), a conclusion supported by our use of a 'spike-in' control involving Drosophila nuclei bearing H3K27me3-marked regions in all three replicates per condition. Finally, new ectopic H3K27me3 peaks were very rare in the PRT4165 treatment, and none of these loci were bound by Aebp2 (Figure 4-figure supplement 2B). Taken together, H2Aub1 deposition by Rnf2 during preZGA is required for the recruitment of Aebp2 and subsequent de novo deposition of H3K27me3 postZGA at developmental loci.
To determine whether H2Aub1 impacts transcriptional repression of developmental genes at postZGA, we performed RNA sequencing (three biological replicates per condition) on 4 hpf embryos that were either vehicle-treated (DMSO) or PRT4165-treated from the one-cell stage onward      ( Figure 4-figure supplement 1E). Here, we identified and characterized both up-and downregulated genes ( Figure 4E). PRT4165-upregulated and -downregulated genes (>3-fold, p-value < 0.01) were both enriched in developmental factors, but the number of genes associated with upregulated GO-terms was substantially greater than downregulated GO-terms (Figure 4-figure supplement  1F). Here, ~16.6% of H2Aub1-marked protein coding genes were upregulated, which may reflect the availability at postZGA of an opportunistic activator, following H2Aub1 loss (Figure 4-figure  supplement 2C). Affected genes include those in clustered loci (e.g. Hox genes) where the effect was moderate, as well as non-clustered/solo formats where the effect of PRT4165 was more pronounced (Figure 4E and F; Figure 4-figure supplement 2A).
Having observed precocious developmental gene upregulation in response to PRT4165 treatment during postZGA (4 hpf), we were curious whether their upregulation could also be observed during the preZGA and ZGA stages. To test, we repeated our PRT4165 treatment regimen and isolated embryos at preZGA (2.5 hpf) and ZGA (3.5 hpf) stages, and performed RNAseq (with biological triplicates; Figure 4-figure supplement 3). Here, analysis of upregulated genes meeting our threshold criteria (fold change ≥1.5; p-value ≤ 0.01) by GO-term analysis uncovered no enrichment of developmental genes upon RNF2 inhibition during preZGA or ZGA. Instead, enriched GO-terms from upregulated transcripts corresponded to genes encoding RNA-binding proteins and ribosomal proteins. Thus, chromatin de-repression (via H2Aub1 removal) does not cause transcriptional activation of large numbers of developmental genes during preZGA or ZGA -not even the developmental genes that are activated postZGA following Rnf2 inhibition. Here, we note that general/housekeeping transcription does not occur until ZGA; only the miRNA-430 locus and a very limited number of genes are transcribed during preZGA. Therefore, H2Aub1 removal does not, by itself, lead to the activation of developmental genes that are normally marked by H2Aub1 during preZGA. Taken together, these results strongly suggest that H2Aub1 represses developmental genes during preZGA and ZGA, independent of H3K27me3, and that the absence of H2Aub1 renders developmental genes susceptible to precocious activation following ZGA, which confers developmental arrest prior to gastrulation.

Discussion
Our work reveals that developmental gene silencing in early zebrafish embryos is established through sequential recruitment and activity of PRC1 and PRC2 complexes, respectively, to otherwise open/ permissive loci bearing Placeholder nucleosomes ( Figure 5). Placeholder nucleosomes containing H3K4me1 and the histone variant H2A.Z(FV) are installed by the chromatin remodeler SRCAP during preZGA (Murphy et al., 2018), and are focally pruned to small regions by the chaperone Anp32e (Kobor et al., 2004;Krogan et al., 2003;Mao et al., 2014;Obri et al., 2014;Wang et al., 2016). Here, our reanalysis of published data confirms that H3K27ac is an additional component of Placeholder nucleosomes during preZGA (Zhang et al., 2018). Functional studies reveal that Placeholder nucleosomes prevent DNAme where they are installed and are utilized to reprogram the DNAme patterns during cleavage stage. Therefore, from a 'permissive' Placeholder platform during preZGA, two very different chromatin/transcriptional states are attained at ZGA: active or poised (Murphy et al., 2018).
Our work suggests that the initial poised state at developmental genes and enhancers involves the imposition of polycomb-based silencing upon the permissive states established by Placeholder nucleosomes. Specifically, we observe H2Aub1 addition by Rnf2/PRC1 during preZGA and ZGA to confer initial transcriptional silencing at developmental loci -which is subsequently read by the Aebp2/ PRC2 complex to add H3K27me3 after ZGA ( Figure 5). Prior work in ES cells has provided an in vitro parallel in which PRC1 activity (H2Aub1 addition) can occur independently of PRC2-mediated recruitment at certain loci Blackledge et al., 2020;Cooper et al., 2014;Cooper et al., 2016;Kalb et al., 2014;Tamburri et al., 2020;Tavares et al., 2012), which has been termed 'non-canonical' order of recruitment, to contrast with prior data showing the reverse/canonical order (Wang et al., 2004a). Furthermore, and consistent with our work, human AEBP2 and JARID2 have recently been shown to directly bind H2Aub1 and stimulate PRC2 activity in the presence of H3K4 methylation (Kasinath et al., 2021), and a similar mechanism may be utilized to establish bivalency after ZGA in zebrafish.
Our work also clarifies and extends prior work in zebrafish which showed that an incross of zebrafish heterozygous for an rnf2 loss-of-function truncation mutation yielded a pleitropic terminal phenotype at 3 days post fertilization (dpf), coincident with the loss of rnf2 RNA. Notably, loss of rnf2 at 3 dpf was associated with partial upregulation of certain developmental genes, but H3K27me3 remained fully present, showing that Rnf2 is not required for H3K27me3 maintenance . However, progression of rnf2 mutants to 3 dpf may have relied on maternally inherited WT rnf2 RNA or protein to provide the initial establishment of gene silencing and H3K27me3, raising the possibility that rnf2 is actually essential at a much earlier developmental stage. Our use of the Rnf2 inhibitor, PRT4165, during preZGA and ZGA stages reveals the necessity for Rnf2 activity for the establishment of developmental gene silencing, for subsequent H3K27me3 addition, and for progression of zebrafish development beyond the ZGA stage. Importantly, developmental gene upregulation is not attributable to H3K27me3 loss, as maternal zygotic ezh2 mutant zebrafish embryos do not precociously activate developmental genes during ZGA, and they progress through gastrulation without H3K27me3 San et al., 2016;San et al., 2019). Notably, although upregulation of developmental genes in the presence of PRT4165 is clear, this involves ~16.6% of the developmental gene repertoire occupied by H2Aub1/Rnf2. Here, we suggest that tissue-specific activators for the majority of developmental genes are not present at ZGA. Maternal loading of mRNA and proteins present challenges for functional experiments during early embryonic stages of development. Our use of the Rnf2 inhibitor, PRT4165, enabled us to test whether H2Aub1 impacts gene expression at ZGA, and whether H2Aub1 is required for the recruitment of Aebp2-PRC2 and for the subsequent establishment of H3K27me3 at developmental loci during postZGA. The observable impact of PRT4165 includes the loss of H2Aub1, the loss of Aebp2 recruitment to all loci that formerly bore H2Aub1, and the loss of H3K27me3 deposition at all developmental loci that are normally marked by H3K27me3 during postZGA. Our interpretation of these observations is that they are linked to, and dependent on, Rnf2 inhibition. However, it is important to note that PRT4165 may have off-target effects aside from inhibition of Rnf2 which may contribute to our observations. While we cannot formally account for possible off-target effects of PRT4165, we note that ZGA (which involves the activation of thousands of genes) progresses relatively normally, except for the upregulation of a cohort of developmental genes which were formerly H2Aub1 marked. This provides a measure of confidence that the upregulation of developmental genes is largely a direct consequence of H2Aub1 loss in the early embryo. However, as PRT4165 was the only approach we employed that successfully removed the vast majority of H2Aub1, future studies involving orthogonal approaches to elicit Rnf2 loss in zebrafish will be needed to further validate the consequences of H2Aub1 loss in early zebrafish embryos.
While our work was in review, work on H2Aub1 and H3K27me3 dynamics in the preimplantation mouse embryo was reported by two separate groups (Chen et al., 2021;Mei et al., 2021). Both studies in mouse demonstrated that H2Aub1 temporally precedes H3K27me3 at developmental genes during early embryonic stages, and both studies also lowered H2Aub1 via one effective approach: either through maternal genetic loss of non-canonical PRC1 subunits (Pcgf1/6) or through overexpression of an H2A de-ubiquitinase (PR-DUB). Here, both approaches led to precocious transcription of a moderate subset of developmental genes at or shortly after ZGA. Thus, these sets of results align well with our observations in zebrafish.
Notably, each method of H2Aub1 perturbation in mice was unique (with attendant advantages and caveats) and conferred different impacts on embryogenesis and establishment of H3K27me3 at developmental loci. In mice, zygotic overexpression of PR-DUB resulted in rapid and significant erasure of H2Aub1, precocious transcription of certain developmental genes, and growth arrest at the four-cell stage. However, reduction of H2Aub1 by zygotic PR-DUB overexpression had minimal impact on H3K27me3, likely due to the fact that H3K27me3 is very low at the four-cell stage. In contrast, Mei et al. disrupted H2Aub1 via maternal genetic ablation (paternal genes remained intact) of two genes, Pcgf1/6, which encode subunits of noncanonical PRC1. Here, Pcgf1/6 mutant oocytes were fertilized with wild-type sperm to produce maternal-null Pcgf1/6 embryos. In contrast to PR-DUB overexpression, maternal-null Pcgf1/6 embryos displayed global reductions in both H2Aub1 and H3K27me3 during cleavage stages -a finding that aligns well with our findings. Notably, the reductions of H2Aub1 and H3K27me3 from maternal Pcgf1/6 loss largely returned to wild-type levels by the morula stage, possibly due to zygotically (and paternally) expressed Pcgf1/6. Here, we postulate that the more severe loss of H3K27me3 from Pcgf1/6 deletion may stem from the temporal difference of disrupting H2Aub1 in the female germline vs. the early mouse embryo, or potentially the developmental stage in which H3K27me3 was assayed ( two cell vs. four cell). A major issue not addressed in these recent papers in mouse is how mechanistically H2Aub1 directs H3K27me3. Notably, our study addresses this mechanism by demonstrating that non-canonical PRC2 complex containing the H2Aub1 'reader' Aebp2 occupies all loci with high H2Aub1 during postZGA, which then receive H3K27me3.
Interestingly, Chen et al. and Mei et al. utilized hybrid mouse strains to trace parental asymmetries of H3K27me3 and H2Aub1 in the early embryo. Here, both groups uncovered rapid erasure of H2Aub1 from the paternal genome shortly after fertilization which is re-established by the two-cell stage. Our data also demonstrate that the paternal genome is initially devoid of H2Aub1, and rapidly acquires H2Aub1 shortly after fertilization. However, zebrafish sperm are entirely devoid of H2Aub1 ( Figure 1B; Figure 1-figure supplement 1C), whereas mouse sperm bear H2Aub1 coincident with H3K27me3. Thus, the paternal genome in both mammals and zebrafish attain a transient state lacking H2Aub1 soon after fertilization -but arrive at that state by alternative routes.
Here, we observe that developmental or housekeeping gene promoters attract either a high-or low-level PRC1 binding, respectively. How ncPRC1 is recruited to CpG islands in ES cells is partially understood, as the ncPRC1 subunit Kdm2b helps recruit ncPRC1 to CpG islands of developmental genes. In this context, Kdm2b binds to hypomethylated CpGs via the CxxC motif Farcas et al., 2012;He et al., 2013;Wu et al., 2013). However, ncPRC1 is recruited robustly to only a minority of Kdm2b-bound CpG islands, implying that additional factors are needed to specify recruitment of ncPRC1 to CpG islands of developmental genes. Here, we speculate that particular TFs, such as the candidates in Figures 1E and 2C, are likewise utilized in cooperation with zebrafish Kdm2b to enable strong focal recruitment of Rnf2-PRC1 to Placeholder-occupied developmental loci. In contrast, the candidate TFs at housekeeping genes would recruit MLL complexes to implement H3K4 methylation, and not Rnf2-PRC1. Thus, future work will explore which particular TFs recruit SRCAP (for Placeholder/H2A.Z installation), PRC1 (for H2Aub1 addition at developmental genes), or MLL complexes (for H3K4me3 modification at housekeeping genes) to define which specific genes are subjected to silencing or activation at ZGA.
One curiosity arising from our work is why zebrafish utilize the non-canonical PRC1 complex, which adds monoubiquitination to H2A/H2A.Z(FV), rather than canonical PRC1 complex, which compacts chromatin for initial developmental gene silencing. Here, we speculate that the rapid (~16 min) cell cycles that characterize the preZGA cleavage state -coupled to the need for continual DNA replication during cleavage stage -are not compatible with the compaction conferred by canonical PRC1 (Gao et al., 2012;Grau et al., 2011;Lau et al., 2017). Furthermore, the necessary substrate to recruit canonical PRC1 to chromatin, H3K27me3, is absent in preZGA embryos. Instead, the use of the repression modes conducted by non-canonical PRC1 addition of H2Aub1 -which antagonizes RNA Pol II transcriptional initiation or bursting -may help confer silencing without conferring a compaction that might impede DNA replication (Dobrinić et al., 2020;Stock et al., 2007). However, once embryos exit cleavage stage, the cell cycle greatly lengthens, and Aebp2-PRC2 complexes add H3K27me3 to loci -which may then enable canonical PRC1 to localize to H3K27me3-marked loci, and conduct compaction. Indeed, prior work has shown that later stages of zebrafish development utilize canonical PRC complexes . Here, future studies may reveal more precisely the logic and timing underlying the transition from non-canonical to canonical utilization of PRC complexes in zebrafish development.

Zebrafish husbandry
Wild-type Tübingen zebrafish were maintained as described (Westerfield, 2007). All experiments involving zebrafish were approved by University of Utah IACUC (Protocol 20-04011). Embryos were scored for developmental staged as described (Kimmel et al., 1995).

Acid extraction of nuclear proteins
Two-hundred embryos were added to 1.5 ml tubes and washed twice with cold PBS; 800 μl of Mild Cell Lysis Buffer (10 mM Tris-HCl pH 8.1, 10 mM NaCl, 0.5 NP-40, 2× protease inhibitors) was applied to embryos and incubated on ice for 5 min. Embryos were homogenized by passing through a 20-gauge syringe several times. Samples were briefly centrifuged to bring down chorions. Supernatants were transferred to fresh tubes and centrifuged at 1300× g for 5 min at 4°C. Pelleted nuclei were washed twice with cold Mild Cell Lysis Buffer. Nuclei were resuspended to a final volume of 800 μl in cold Mild Cell Lysis Buffer and supplemented with 10 μl of sulfuric acid (18.4 M). Samples were sonicated for 10 s (1 s ON, 0.9 s OFF) at 30% output using a Branson sonicator. Proteins were extracted for 30 min at 4°C on a rotator, and 160 μl of 100% trichloroacetic acid was added and proteins were allowed to precipitate for 30 min on ice. Samples were centrifuged at 13,000 rpm for 5 min at 4°C. Protein pellets were washed with 800 μl of cold acidified acetone, and centrifuged again at 13,000 rpm for 5 min at 4°C. Protein pellets were washed with 800 μl of cold acetone, and centrifuged again at 13,000 rpm for 5 min at 4°C. Supernatant was discarded and pellets were dried at 37°C for 5 min. Dried protein pellets were resuspended in 2× Laemmli sample buffer and boiled for 8 min. Samples were then used for western blotting. Bands were quantified in ImageJ (Schneider et al., 2012).

Immunohistochemistry and DAPI staining
Standard protocol for immunohistochemistry was followed as described (Fernández and Fuentes, 2013;Zhang et al., 2018). Three biological replicates were performed for each immunohistochemistry experiment. Briefly, 30 embryos were collected at appropriate timepoints and fixed with fresh 4% paraformaldehyde (Electron Microscopy, Cat# 50980487) in 1× PBS at room temperature for 12 hr. Droplets of glacial acetic (100%, Merck, Cat# 1000560001) or DMSO (final concentration 0.5%, Sigma) were added 5-10 s after initiation of the fixation. Chorions were manually removed from fixed embryos with forceps and dechorionated embryos were dehydrated in methanol and stored at -20°C. For immune-staining, embryos were rehydrated into PB3T (1× PBS with 0.3% TritonX-100, and then incubated in blocking agent 1% BSA, 0.3 M glycine in PB3T). Embryos were incubated with primary antibodies diluted in blocking agent overnight at 4°C. Primary antibodies were removed and embryos were washed extensively with PB3T. Embryos were next incubated with appropriate secondary antibodies in the dark followed by extensive washes in PB3T. Primary antibodies used for immune-staining are listed below. Secondary antibodies used were donkey α-rabbit IgG-488 at 1:500 (Life Technologies, Cat# A-21206). DAPI was used at 1:1000 as a nuclear counterstain. The yolk cells were removed from embryo and embryo was mounted on glass slide with ProLong Gold Antifade mounting media (Thermo Fisher, Cat# P-36931) and a 2.0 mm square coverslip sealed with nail polish. Samples were stored at 4°C until imaged.

Imagining of zebrafish embryos
Images were acquired on a Leica SP8 White Light laser confocal microscope. Image processing was completed using Nikon NIS-Elements multi-platform acquisition software with a 40×/1.10 Water objective. Fiji (ImageJ, V 2.0.0-rc-69/1.52p) was utilized to color DAPI channel to cyan, GFP color remained green. Confocal images are max projections of Z stacks taken 0.5 μm apart for a total of the embryo ~7-12 μm.

Primary antibodies
The following antibodies were utilized in for the present study: anti-H2Aub1 (

ChIP-seq in zebrafish embryos
Embryo fixation Approximately 1.5 million cells were used for each ChIP replicate. Embryos were allowed to progress to the desired developmental stage and then transferred to 1.5 ml microcentrifuge tubes (~200 embryos per tube). Chorions were removed enzymatically by treatment with pronase (1.25 mg/ml in PBS). Dechorionated embryos were gently washed twice with PBS to remove pronase. Samples were fixed with 1% formaldehyde (Electron Microscopy Sciences, Cat# 15712) for 10 min at room temperature with end over end rotation. Fixation was quenched with 130 mM glycine for 5 min at room temperature. Samples were centrifuged for 5 min at 500× g at 4°C. Supernatant was discarded and cell pellets were washed twice with ice-cold PBS. Cell pellets were frozen with liquid nitrogen and stored at -80°C.

Chromatin sonication
Nuclear lysates were sonicated with a Branson Digital Sonifier with the following settings: 10 s duration (0.9 s ON, 0.1 s OFF), 30% amplitude. Seven sonication cycles were performed. Samples were placed in an ice bath for at least 1 min between each sonication cycle. Sonicated samples were centrifuged at 14,000 rpm, 4°C, for 10 min to pellet insoluble material. Supernatants were transferred to new tubes. A portion of the sample was set aside to confirm optimal chromatin shearing by agarose gel electrophoresis.

Preclear
Twenty µl of Dynabeads (Invitrogen) were blocked with 0.5 mg/ml BSA in PBS. Blocked Dynabeads were subsequently applied to each sonicated sample and rotated for 1 hr at 4°C. Samples were placed on a magnet stand for 1 min and precleared supernatant was transferred to a new tube; 5% of the sample was removed and stored at -80°C as input. Antibody and fresh 1× protease inhibitors were added to each sample. Samples were rotated overnight at 4°C.

Pulldown
Samples were centrifuged at 14,000 rpm, 4°C, for 5 min to pellet insoluble material. Supernatants were transferred to new tubes; 50 µl of Dynabeads (Invitrogen) were blocked with BSA 5 mg/ml in PBS. Blocked Dynabeads were subsequently applied to each sample and rotated for 6 hr at 4°C.

Elution and reversing crosslinks
One-hundred µl of Elution Buffer (10 mM Tris-HCl pH8.0, 5 mM EDTA, 300 mM NaCl, 0.1% SDS) was added to beads. Two µl of RNase A (Thermo Fisher, Cat#EN531) was added to each ChIP and input sample and incubated at 37°C for 30 min with gentle agitation. Ten µl of Proteinase K (Thermo Fisher, Cat# 25530049) was added to each sample and incubated at 37°C for 1 hr with gentle agitation. Crosslinks were reversed overnight at 65°C with gentle agitation. ChIP DNA was purified with a Qiagen MinElute PCR Purification kit (Cat#28004).

ChIP-seq library construction and sequencing
ChIP-seq libraries were prepared using NEBNext ChIP-Seq Library Prep Reagent Set for Illumina (New England BioLabs, Cat# E6240). High-throughput sequencing was performed on Illumina HiSeq 2500 for single-end 50 bp reads or Illumina NovaSeq 6000 for paired-end 50 bp reads.

ChIP-Rx-seq in zebrafish embryos
ChIP-Rx was adapted from Orlando et al., 2014, for H3K27me3 ChIP in zebrafish embryos treated with DMSO or PRT4165. Crosslinked Drosophila melanogaster S2 cells (ATCC Cat# CRL-1963) were spiked into resuspended zebrafish embryo pellets at a ratio of 5:1 (zebrafish cells: S2 cells). ChIP-Rx was subsequently performed in the same way as described above.

ChIP in zebrafish sperm
ChIP in zebrafish sperm was conducted as described (Murphy et al., 2018). qPCR was carried out using 2× SsoAdvanced Universal SYBR Green Supermix (Biorad Cat# 1725270) and a Biorad CFX real-time thermal cycler.

Oligonucleotides
See Table 1 for oligonucleotides used for ChIP-qPCR.

RNA-seq
Total RNA was harvested from zebrafish embryos with a Qiagen Allprep kit (Cat# 80204). The Invitrogen DNA-free DNA removal kit (Cat# AM1906) was subsequently used to remove contaminating DNA from RNA samples. Intact poly(A) RNA was purified from total RNA samples (100-500 ng) with oligo(dT) magnetic beads and stranded mRNA sequencing libraries were prepared as described using the Illumina TruSeq Stranded mRNA Library Preparation Kit (RS-122-2101, RS-122-2102. Purified library quality was assessed on an Agilent Technologies 2200 TapeStation using a D1000 ScreenTape assay (Cat# 5067-5582 and 5067-5583). The molarity of adapter-modified molecules was defined by qPCR using the Kapa Biosystems Kapa Library Quant Kit (Cat# KK4824). Individual libraries were normalized to 5 nM and equal volumes were pooled in preparation for Illumina sequence analysis. High-throughput sequencing for RNAseq was performed on an Illumina HiSeq 2500. RNA-seq data displayed in Figure 3-figure supplement 1A was collected from http://www.ebiac.uk/gxa/experiments/E-ERAD-475. We would like to thank the Busch-Nentwich lab for providing RNA-seq data used in Figure 3-figure supplement 1A.

ChIP-seq analysis
ChIP-seq Fastq files were aligned to Zv10 using Novocraft Novoalign with the following settings: -o SAM -r Random. SAM files were processed to BAM format, sorted, and indexed using Samtools (Li et al., 2009). ChIP-seq replica correlation was assessed with deepTools (Ramírez et al., 2016). Briefly, BAM files were read normalized with deeptools bamCoverage with the --normalizeUsingRPKM flag. Deeptools multiBigwigSummary bins and plotCorrelation were used to generate genome-wide correlation matrices for assessing replica correlation. ChIP-seq peak calling was accomplished using MACS2 with the following settings: callpeak -g 1.4e9 -B -q 0.01 -SPMR (Zhang et al., 2008). ChIP-seq peak calling for comparisons between Aebp2 and Jarid2 as well as ChIP-seq involving drug treatments was performed by utilizing the Multi-Replica Macs ChIPSeq Wrapper. Called peaks were annotated in R with the ChIPseeker package (R Development Core Team, 2020;Yu et al., 2015). Output bedgraph files from MACS2 were processed into bigwig files with UCSC Exe Utilities bedGraphToBigWig  (Kent et al., 2010). Resulting bigwig files were loaded into IGV for genome browser snapshots of ChIP-seq enrichment (Thorvaldsdóttir et al., 2012). Heatmaps of ChIP-seq enrichment at promoter and enhancer regions were made with deepTools (Ramírez et al., 2016). A bed file of Zebrafish zv10 UCSC RefSeq genes from the UCSC Table Browser was utilized for plotting heatmaps of ChIP enrichment at promoters (Karolchik et al., 2004). Genes residing on unmapped chromosomal contigs were excluded. Enhancer heatmaps utilized a bed file of postZGA H3K4me1 ChIP-seq (Bogdanovic et al., 2012) peak summits that had been filtered to exclude promoters and unmapped chromosomal contigs. Gene ontology analysis was performed with DAVID (Huang et al., 2009).
Violin plot of ChIP-seq enrichment at promoters Log2(ChIP/input) data for promoter regions (±1 Kb from TSS) of interest was collected from processed bigwig files by utilizing the program Bio-ToolBox 'get_datasets.pl'. Collected data was plotted in violin format. Unpaired t-tests with Welch's correction were utilized to determine statistical differences in ChIP enrichment between promoter K-means clusters. Violin plot and statistical analysis (Figure 1-figure supplement 2J) were performed in GraphPad Prism version 8.3.1 using GraphPad Prism version 8.3.1 for MacOS, GraphPad Software, San Diego, CA, https://www.graphpad.com.

DNA motif analysis
HOMER was utilized for identifying putative TF-binding motifs present at promoters and enhancers (Heinz et al., 2010). The following parameters were used on bed files of promoters and enhancers of interest: findMotifsGenome. pl danRer10 -size -250,250. Known motifs (as opposed to de novo motifs) from HOMER were presented in Figures 1 and 2.

ChIP-seq analysis involving drug treatments
Analysis of ChIP-seq experiments involving drug treatments was performed by utilizing the Multi-Replica Macs ChIPSeq Wrapper.

Data access
All sequencing datasets generated in this study have been deposited at the Gene Expression Omnibus under the accession number GSE168362.

Drug treatments
PRT4165 (Tocris Cat#5047) was dissolved in DMSO at a concentration of 50 mM. PRT4165 was further diluted to a working concentration of 150 µM in embryo water and mixed vigorously. Zebrafish embryos were collected and immediately placed in embryo water containing 150 µM PRT4165 (or DMSO) and allowed to develop to the desired developmental stage.