Timing of establishment of PcG modifications at ZGA.

A) Heatmaps are plotted for ChIP-seq measurements in cellular blastoderm staged embryos for the indicated factors (top) over a set of E(z) peaks localized within PcG Domains, flanked by ± 10 kb. One representative E(z) peak per domain was selected for plotting. Heatmap rows are identically ordered from top to bottom by overall average H3K27me3 intensity per peak. E(z), Pho, Cg, and GAF ChIP-seq are all performed using an anti-GFP antibody on CRISPR-engineered embryos expressing GFP-tagged chimeric alleles of the target protein. Zld ChIP-seq was previously reported (33). Data were standardized prior to plotting and colorbars (bottom) indicate binding intensity in units of standard deviations. B) Heatmaps are plotted for standardized ChIP-seq measurements in embryos collected at the indicated stages (top) for either H3K27me3 (top row), or H3K27me1, bottom row. The average standardized ChIP-seq signal per time point is plotted at right. C) The average standardized ChIP-seq signal for timecourse measurements of H3K27me3, -me1, and H2Aub are plotted over entire PcG domains. The domains are first divided into 100 bins and standardized ChIP-seq signal was averaged per bin prior to plotting average signal across domains ± 13 kb flanking sequence. D) Visualization of the PcG domain containing the ind locus and counts-per-million (CPM) normalized ChIP-seq measurements for Cg, Pho, E(z) at cellular blastoderm, and timecourse measurements of H3K27me3 (NC13 - NC14 late), and H2Aub (NC13, NC14 late). ChIP target is indicated at left and the display range of the CPM normalized ChIP data is indicated at right. Scale bar: 2.5 kb E) Visualization of the PcG domain containing the bithorax complex for the same CPM-normalized ChIP tracks as in panel D. Scale bar: 10 kb.

E(z) peaks outside of PcG Domains generally associate with active promoters.

A: Heatmaps for the indicated ChIP targets were plotted centered over all 4576 E(z) peaks (± 10 kb) and subset by membership within a PcG domain. Data were z-scored prior to plotting, and the plotted intensity range is indicated by the colorbar beneath each respective heatmap. Peaks are ordered on the y-axis by the overall intensity of E(z) binding in the region shown. A negative control ChIP (IP for GFP on wild type chromatin) is shown at the far right. While E(z) peaks within domains are enriched for H3K27me3, peaks outside of domains show a high degree of correlation with marks of active transcription including RNA Pol2 (CTD pSer5), H3K4me1/2/3, and H3K27 acetylation. B: The mean enrichments of E(z) and DNA binding factors over E(z) peaks ± 10 kb are plotted for peaks inside (blue) and outside (yellow) of PcG domains.

Enrichment of Selected Transcription Factors with E(z) peaks that accumulate H3K27me3 by late NC14.

Peaks of transcription factors (column 1) overlapping with E(z) peaks were counted and scored for whether overlapping E(z) peaks acquire H3K27me3 by late NC14. 2x2 contingency tables (factor-by-K27me3 domain membership, for the set of all E(z) peaks) were calculated and subjected to Fisher’s Exact Test. The number of factor peaks overlapping any E(z) peak (n = 4576) is shown in column 2. The number of factor peaks overlapping E(z) peaks within PcG domains (n = 1264) is shown in column 3. Log2(Odds Ratios) are reported in column 4, and p-values are reported in column 5. Negative log2(odds ratios) indicate depletion (red) and positive log2(odds ratios) indicate enrichment (blue) of a factor associating with E(z) peaks that accumulate H3K27me3. Additional pairwise comparisons that did not yield significant p-values are not shown.

Nuclear localization patterns of E(z) and three PRE-binding factors over late cleavage divisions.

EGFP-E(z) (A), Pho-sfGFP (B), EGFP-Cg (C), and EGFP-GAF (D) were subjected to time-lapse confocal microscopy from nuclear migration (NC10) through the period of cellularization (NC14 + 60 minutes). All specimens also expressed Histone H2Av-RFP to visualize nuclei. The images at left show maximum-projected still images from representative time-lapse movies for the indicated GFP chimera (top) and His2Av-RFP (bottom) at interphase of the indicated nuclear cycles. The insets show a color merge of one nucleus at each respective nuclear cycle (yellow: GFP chimera, magenta: His2Av-RFP). The quantifications at right show the relative nuclear concentration of each GFP chimera from NC11 through NC14 (top x-axis). Grey vertical bars indicate the approximate mitotic periods for each nuclear cycle. Data represent averaged intensity measurements from n ≥ 3 embryos per chimera, mean normalized ± standard error of the mean. While E(z) readily localizes to nuclei throughout the cleavage divisions, the three PRE- binding factors Pho, Cg, and GAF increase nuclear concentration over the last nuclear divisions before cell cycle lengthening at NC14.

Immunostaining of E(z) and three PRE-binding factors.

Embryos expressing EGFP-E(z) (A), Pho- sfGFP (B), EGFP-Cg (C), and EGFP-GAF (D) chimeras were formaldehyde-fixed and subjected to immunostaining with an anti-GFP antibody to facilitate measurement of nuclear concentration prior to nuclear migration. All images are maximum z-projections. For each chimera, the top row of images show GFP immunofluorescence signal from representative specimens staged at the nuclear cycle indicated at top. Samples were co-stained with DAPI to visualize DNA/nuclei. The insets show an enlarged single nucleus indicated by the magenta arrowhead (green: GFP chimera, magenta: DAPI). Displayed image intensities are consistent for post-migration nuclear cycles (NC10-14) and are linearly adjusted based on DAPI intensity for pre-migration nuclear cycles, when nuclei are found deep within the embryo. For EGFP-GAF (D), one example of a pre-migration mitosis is shown (NC7 anaphase) to demonstrate the ready detection of GAF on mitotic chromatin in contrast to much lower signal during pre-migration interphases. E(z) localizes to interphase nuclei from the beginning of the cleavage divisions and is readily detected. Pho, Cg, and GAF do not localize to interphase nuclei until after nuclear migration. The scale bar is shown in yellow in the bottom right of the top-left image in each panel, size = 10 µm.

Immunostaining of H3K27me3 and H2Aub during cleavage stages.

Embryos (w1118) were fixed and subjected to immunostaining for H3K27me3 (A) and H2Aub (B). Samples were co-stained with DAPI to highlight DNA/nuclei. Each panel shows maximum z-projected immunostaining for the indicated histone modification in the top row of images at the indicated nuclear cycle (top). The bottom row shows the corresponding DAPI image. Image intensities are normalized as described for Figure 4 to allow for semi-quantitative comparison of signal intensities across nuclear cycles. The inset shows a single enlarged nucleus (indicated by a magenta arrowhead) with a color merge of the modification (green) and DAPI image (magenta). The magenta asterisk in panel B, NC10 indicates the position of the polar body. H2Aub is not detectable in somatic nuclei above background until late cleavage divisions. The scale bar is shown in yellow and corresponds to 10 µm in all cases.

GAF knockdown has limited impact on accumulation of H3K27me3 at NC14.

A) Control (EGFP-GAF) or GAF-knockdown (Jabba-Trap/+; EGFP-GAF) embryos were formaldehyde fixed and immunostained for GFP to detect GAF at NC14. Specimens were counterstained with wheat germ agglutinin (WGA) to highlight nuclear membranes. A single z-slice of individual channels (grey) and merged channels (GAF: green, WGA: magenta) shows that in the presence of Jabba-trap, EGFP-GAF is undetectable within NC14 nuclei. B) ChIP-seq was performed against EGFP-GAF in either positive control (EGFP-GAF), Jabba-trap GAF (JT-GAF), or negative control (w1118) samples with an anti-GFP antibody. Following peak calling, average GAF ChIP-seq signal in counts per million (CPM) over GAF peaks was measured and plotted, with ± 5 kb flanking sequence. The plot shows that whereas GAF (purple) is strongly enriched over peaks, Jabba-trap (orange) reduces GAF occupancy to near-background (grey) levels. C) A volcano plot is shown for the results of a DESeq2 differential enrichment analysis for H3K27me3 ChIP-seq between control (EGFP-GAF) and JT-GAF. Enrichment was measured over 2 kb bins and ‘runs’ of consecutive bins with p- values < 0.05 were merged. The volcano plot shows on the x-axis the average log2 fold change (FC) of the merged bins and the -log10 value of adjusted p-values merged using Fisher’s method. The size of a plotted point indicates the relative mean counts for the run of bins. Blue dotted lines indicate traditional cutoffs for significance testing (abs(log2(FC)) > 1 & p-adj < 0.05). Runs of bins exceeding these thresholds are colored blue. Regions of interest are labeled in red. While few regions show large-magnitude, GAF-dependent changes in K27me3, the general trend is for this modification to increase within PcG domains in the absence of GAF. D) Detail of ChIP-seq coverage over the escargot PcG domain. GAF ChIP seq for control or JT-GAF is shown at top (purple) and H3K27me3 for both genotypes is shown at bottom. The log2(FC) from the DESeq analysis over 2 kb bins is plotted in the bottom track. Scale bar is shown at top left, and y-axis range for coverage plots are indicated in brackets, with units of counts per million. E-G) Detail of ChIP-seq coverage over the bithorax complex (E) and zoomed-in views of the bx (F) and iab7 (G) PREs are shown. Tracks are as described for panel D. The location of the PREs in F and G are indicated with a grey arrow. Scale bars are shown at top left for each respective plot. Y-axis range for coverage plots are indicated in brackets with units of counts per million.

Zelda is necessary during ZGA for the establishment of H3K27me3 at a subset of PcG domains.

A) A volcano plot is shown for DESeq2 analysis of H3K27me3 ChIP-seq comparing wild type and zelda germline clone embryos. Plotting methodology is as described in Figure 6C. On average, Zelda activity is required for establishment rather than antagonism of H3K27me3 at a subset of PcG domains. B) H3K27me3 is predominantly lost at sites requiring Zelda for RNA Pol2 recruitment. The log2(FC) of H3K27me3 ChIP-seq is plotted for 2 kb bins containing TSS that depend on Zelda for Pol2 recruitment. In the absence of Zelda, the TSS included in this analysis have major reductions in Pol2 occupancy. TSS were grouped into two categories: those within a PcG domain (top) and those outside a PcG domain (bottom). In the absence of Zelda, TSS within PcG domains that lose Pol2 (n = 66) also typically see losses in H3K27me3, with the exception of a few loci (eve, zen) that see increases. Loss of Pol2 at Zelda-sensitive TSS outside of PcG domains (n = 203) does not correspond to de novo accumulation of H3K27me3, except at the ocho locus. C-E) Coverage plots for loci representing classes of PcG domain responses to zelda loss of function. ChIP for factors indicated at left were performed for wild type (top track) and zelda germline clone (bottom track) embryos. ATAC data are from (Hannon 2017) and Zelda ChIP-seq is from (Harrison 2011). Scale bar is at top left and the y- axis range for each displayed track is indicated in brackets, with units of counts per million. Panel C shows a class 1 locus, even skipped. Class 1 are PcG domains where increases in K27me3 are seen in the absence of zelda. Panel D and E show Class 2 loci, which are PcG domains where decreases in K27me3 are seen in the absence of zelda. Class 2A loci (panel D) see reductions in K27me3 accompanied by loss of E(z) recruitment. Class 2B loci (Panel E) see reductions in K27me3 without significant loss of E(z) recruitment.

The number of E(z) peaks inside and outside of PcG Domains are shown in the bar graph, subset by whether a peak directly overlaps with an annotated transcription start site.

We note that in comparison to the heatmaps this overlap measurement likely under-estimates the degree of overlap outside of PcG domains. This could be due to under-counting of active TSS, and/or a peak not strictly overlapping a TSS, but instead being near-by. Nevertheless, E(z) peaks outside of domains are significantly enriched for annotated active TSS regions with a log2 odds ratio of 2.5 and a Fisher’s exact p-value of 5.31e-105.

DESeq on 2 kb bins for the GAF knockdown experiment and making runs out of bins.

A) A volcano plot is shown for DESeq2 analysis of H3K27me3 ChIP-seq for EGFP-GAF versus JT-GAF. Overall, very few 2 kb bins show effects that surpass the significance thresholds for p-value and log2(fold change). The bin corresponding to the escargot transcription start site is labeled. B) Reproduced from Figure 6 with additional annotations indicating the location of the highlighted point in panel A and a broader region surrounding the bin with a more moderate effect on H3K27me3 following JT-sequestration of GAF. C) Strategy for generating consecutive 2 kb runs of bins with p-values < 0.05, and how fold-change and significance values are propagated. D) Reproduced from Figure 6 to show how making consecutive runs transforms the data.

DESeq on 2 kb bins for the zelda mutant experiment and correlation between H3K27me3 and H2Aub.

A) A volcano plot is shown for DESeq analysis performed on H3K27me3 ChIP seq between zelda mutants and wild type embryos, staged mid NC14. 226 bins have significant reductions in H3K27me3, and 62 bins have significant increases. Significant bins containing a zygotic transcription start site are labeled with the gene symbol (red). B) Log2(fold change) values from DESeq analysis of H3K27me3 (x-axis) and H2Aub (y-axis) ChIP-seq between zelda and wild type embryos are plotted. The correlation coefficient is 0.78. Linear regression was performed, and the fit model is plotted in red (slope = 0.71). There is a strong positive correlation between changes in both modifications following loss of zelda, although the trend is for H2Aub to lag slightly in its magnitude of response compared with H3K27me3.