Mod(mdg4) and Snr1 are essential for dhd expression

We previously performed a female germline RNA interference screen to identify chromatin factors required for paternal chromosome incorporation into the zygote at fertilization [5]. As part of that screen, the histone H3K4 demethylase Lid, Sin3A and Rpd3, which participate in deacetylase complexes targeting various lysine residues in H3 and H4 [17,18], were identified as essential regulators of dhd expression. Because Lid and Sin3A can interact within a co-repressor complex [19,20], we asked whether other chromatin regulatory complexes might also be involved in dhd regulation. We therefore broadened our analysis to other knockdowns that caused maternal effect sterility associated with a dhd-like mutant phenotype, i.e. defective sperm nuclear decompaction at fertilization. Among these, we focused on two additional UAS-controlled small hairpin RNA (shRNA) constructs from the TRiP collection [21], respectively targeting mod(mdg4) and Snr1. Snr1 is an essential subunit of the Brahma chromatin remodeler that mediates protein-protein interactions within this complex as well as with external interacting partners [13,22]. The mod(mdg4) gene codes for up to 31 isoforms [23], all of which are targeted by the shRNA construct. Among these, the most well characterized, Mod(mdg4)67.2 is a common component of boundary insulators in the Drosophila genome [24], but other non-insulating isoforms exhibiting activator functions have also been identified [4,25,26]. These two candidates belonged to two classes of epigenomic effectors distinct from Lid and Sin3A, and we thus decided to investigate their function during the oocyte to zygote transition.

When activated by the Maternal Triple Driver (MTD) Gal4 source, these shRNAs efficiently reduced the levels of mod(mdg4) and Snr1 transcripts (S1A Fig). Previous studies reported defective oogenesis and diminished egg production in mod(mdg4) as well as Snr1 mutant females [22,25]. Interestingly, females with ovarian knockdown of mod(mdg4) or Snr1 (hereby referred to as mod(mdg4) KD or Snr1 KD females) were almost completely sterile (Table 1). Indeed, while KD females were able to lay more eggs than mutants, these almost systematically failed to hatch. Focusing on paternal chromatin organization at fertilization in these embryos, we found that both mod(mdg4) and Snr1 ovarian KDs systematically led to failure of male pronucleus decondensation, which remained elongated (Fig 1A). Concomitantly, these eggs exhibited retention of the protamine fluorescent marker Mst35Ba::GFP (ProtA::GFP) [27] in paternal chromatin, as observed in dhd loss of function mutants [9,12].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Mod(mdg4) and Snr1 are required for dhd expression.

A—Maternal Mod(mdg4) and Snr1 are required for protamine removal and sperm nuclear decompaction at fertilization. Top: Confocal images of pronuclear apposition in eggs from Control (MTD>+), dhd^J5, mod(mdg4) KD or Snr1 KD females mated with transgenic ProtA::GFP males. The sperm nucleus in dhd^J5, mod(mdg4) KD and Snr1 KD eggs retains ProtA::GFP (green) and has a needle-shape morphology. Bottom: zoom on the sperm nucleus. Scale bars: 5μm. B—dhd is strongly downregulated in mod(mdg4) KD and Snr1 KD ovaries. RNA-seq normalized reads per gene (in RPKM) are shown for mod(mdg4) KD vs Control (top) and Snr1 KD vs Control (bottom). Genes downregulated (green) or upregulated (red) in KD ovaries are highlighted. C—Genome Browser view of Control, dhd^J5, lid KD, Sin3a KD, mod(mdg4) KD and Snr1 KD ovarian RNA-seq signal at the dhd region showing dramatic downregulation in all KD conditions. Note that the Control track is represented at two different scales to fit the high read count for dhd (top track) or the low read count for its neighboring genes (bottom track). D—The DHD protein is undetectable in KD ovaries. Western blot analysis using an anti-DHD antibody on ovary extracts of the indicated genotypes. Alpha-tubulin is used as a loading control.

https://doi.org/10.1371/journal.pgen.1009615.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Embryo hatching rates.

The w¹¹¹⁸ strain is used as reference.

https://doi.org/10.1371/journal.pgen.1009615.t001

The above results suggest that Mod(mdg4) and Snr1 could regulate dhd expression. RNA-sequencing on mod(mdg4) and Snr1 KD ovaries indeed revealed that dhd is dramatically downregulated in both KDs, with a fold reduction of almost two orders of magnitude (Figs 1B, 1C, S1B and S1E). dhd was the first most strongly affected gene in mod(mdg4) KD ovaries in terms of fold-change in expression, and the 14^th most affected gene in Snr1 KD ovaries. Consistently, DHD protein levels assessed by Western Blot in KD ovaries were also dramatically reduced (Fig 1D). This was in contrast to a more modest impact of both KDs on the rest of the transcriptome and the limited overlap in their effects (S1B, S1C and S1D Fig). In particular, genes in the vicinity of dhd were not significantly affected by the KDs (Figs 1C and S1B). Therefore, despite the packed genomic organization of the dhd locus, its expression strictly and singularly depends on multiple epigenomic effectors.

Cut&Run with dedicated analysis reveals both the distribution of histone modifications and their associated regulatory elements

To more precisely characterize the chromatin landscape at the dhd locus, we next implemented the Cut&Run epigenomic profiling method [15]. In Cut&Run, histone modifications of interest are targeted in situ by a specific antibody following tissue permeabilization. Target-bound antibodies are subsequently coupled to a fusion between the bacterial Protein A and Micrococcal Nuclease (ProteinA-MNase) that cleaves exposed DNA in the vicinity of the antibody, releasing target nucleosomal particles into solution. Importantly, MNase is expected to also cleave exposed DNA in the immediate spatial vicinity of the nucleosome-bound antibody, causing the release of DNA particles bound by other proteins such as polymerases or DNA sequence-specific transcription factors (Fig 2A). In particular, DNA regulatory elements occupied by sequence-specific transcription factors are typically associated with MNase footprints distinctly shorter than nucleosomes [28–30]. Partially unwrapped dynamic nucleosomes typically associated with regulatory elements can also produce such distinctly short footprints [31]. Following paired-end sequencing, such released DNA fragments can be distinguished and separated by their size, yielding a map of nucleosomes (>146bp) and sub-nucleosomal particles, putatively corresponding to regulatory elements (<120bp). A single Cut&Run experiment should thus identify DNA regulatory elements that are in physical proximity of target histone modifications.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. A single Cut&Run experiment maps both histone modifications and their associated regulatory elements.

A—Schematic overview of the Cut&Run procedure for dissected Drosophila ovaries. After tissue permeabilization and antibody targeting, ProteinA-MNase cleaves nearby exposed DNA allowing the solubilization and retrieval of both nucleosomal particles carrying the targeted histone modification and DNA particles occupied by transcription factors in the immediate vicinity. B—Cut&Run reveals nucleosomes and transcription factor binding sites. Mid-point plot of ovarian H3K27me3 Cut&Run data centered at peaks identified by MACS2 from short fragments (<120bp) in the same experiment. This plot represents all paired-end sequenced fragments as their middle point coordinate in the X-axis, and their size in the Y-axis, revealing a class of clustered short fragments (50-130bp) flanked by nucleosome-sized fragments (>140bp). C—H3K27me3 Cut&Run at the bithorax complex (BX-C) in Drosophila ovaries reveals its regulatory architecture. Genome browser track displaying all Cut&Run fragments and <120 bp fragments separately. Multiple well-described Polycomb Response Elements (PRE) and insulators within the Bithorax complex detected as short fragment peaks are indicated (arrows). D—Cut&Run re-discovers regulatory elements associated with H3K27me3 genome-wide. Upper panels: short fragment peaks read density heatmaps of ovarian H3K27me3 Cut&Run (all fragments and <120bp fragments), ATAC-seq (from S2 cells, [35]), CP190 ChIP-seq (from Kc cells, [38]) and Polycomb ChIP-seq (Pc, from S2 cells, [37]) plotted at ±1kb around peak summit. Data is sorted by the ratio of H3K27me3 Cut&Run total reads at the 3’ versus 5’ flanks to reveal short fragment peaks at the borders or within H3K27me3 domains (dashed lines). Lower panels: average profiles corresponding to the top heatmaps, distinguishing 5’ border peaks, 3’ border peaks and peaks embedded within domains. Cut&Run short fragment peaks are enriched for ATAC-seq signal as well as CP190 (particularly at border peaks) and Polycomb (particularly at middle peaks).

https://doi.org/10.1371/journal.pgen.1009615.g002

With this in mind, we conducted H3K27me3 Cut&Run in Drosophila ovaries. Using only 12 pairs of ovaries per sample, we robustly revealed H3K27me3 domains. Remarkably, visualization of Cut&Run fragments shorter than 120bp (which excludes fully wrapped octameric nucleosomes) revealed that these were enriched at discrete peaks within H3K27me3 domains. Genome-wide analysis identified 679 peaks of fragments <120bp (hereon referred to as “short fragment peaks”) (Fig 2B and 2D). We hypothesized that short fragment peaks represented H3K27me3-associated regulatory elements occupied by transcription factors. Within H3K27me3 domains, we expected these to include Polycomb Response Elements (PREs) as well as insulators. For example, short fragment peaks corresponded to several well-described PREs and insulators in the Bithorax complex H3K27me3 domain [26,32,33] (Fig 2C), consistent with observations in larval tissue [34]. To ask whether this reflects a broader genome-wide trend, we compared short fragment peaks with PRE and insulator markers genome-wide. Although there is scarce genome-wide data available for Drosophila ovaries, H3K27me3 domains are generally present in most cell types. We thus exploited datasets from embryonic-derived S2 and Kc cell lines. Consistent with their occupancy by transcription factors, ATAC-seq peaks [35] -revealing hyper-accessible DNA- coincide with Polycomb regulatory elements in flies and mice [35,36]. Genome-wide, our small fragment peaks identified in ovaries were enriched for ATAC-seq signal, arguing that these indeed correspond to DNA regulatory elements (Fig 2D). Enrichment at these peaks of the Polycomb protein [37] and the insulator protein CP190 [38] further argues that these elements often correspond to functional PREs or insulators. Accordingly, at the borders of H3K27me3 domains, short fragment peaks were more frequently associated with CP190, confirming previous reports that this factor is associated with H3K27me3 domain boundaries [24,39] (Fig 2D). Instead, Polycomb was rather enriched at peaks localized internally within these domains. Our Cut&Run analysis strategy can therefore be used to reveal not only the breadth of histone modification domains in ovaries but also their associated DNA regulatory elements.

dhd lies within an H3K27me3/H3K9me3 mini-domain flanked by DNA regulatory elements

To gain insight on dhd regulation, we next sought to analyze its associated chromatin configuration. We previously showed that the active transcription modification H3K4me3 is enriched at the dhd promoter and that this mark is lost in lid KD ovaries [5] (S2A Fig). Using available ChIP-seq datasets from embryonic derived S2 cells, we further observed that that dhd lies within a ~5kbp mini-domain featuring two types of repressive histone modifications: H3K27me3 and H3K9me3 (S2A Fig) [40]. H3K27me3 is the hallmark of Polycomb-based repression [41,42], whereas H3K9me3 dictates Heterochromatin Protein 1 (HP1)-based repression [43,44]. Interestingly, this mini-domain was also found in ChIP-seq data from fly ovaries (S2A Fig) [45]. Potentially regulating this chromatin environment, Lid and Sin3A were described as participating in a co-repressor complex [20], but their global impact on repressive histone modifications is unclear. In turn, previous reports showed that depletion of insulator proteins Mod(mdg4), as well as CTCF, Su(Hw), CP190 or BEAF-32, did not affect the spread of Polycomb-associated domains but instead caused a general decrease in H3K27me3 levels [24]. In contrast, the Brahma/BAF complex is typically considered as counteracting Polycomb repression based on work in mammals [46], but this interplay has not been analyzed in Drosophila.

Based on these observations, we sought to better characterize the distribution of H3K27me3 and H3K9me3 in ovaries at the dhd locus. Using our Cut&Run approach, we confirmed that dhd is included in a ~5450bp heterochromatic H3K27me3/H3K9me3 mini-domain that extends from the promoter region of dhd to the promoter of the next gene active in ovaries, Sas10 (Fig 3A). While our whole-tissue Cut&Run approach cannot distinguish which cells harbor this domain, published H3K27me3 ChIP-seq profiles from either somatic or germline cells in ovaries indicate that this mark is present in both (S2A Fig). We segmented the Cut&Run H3K27me3 signal in control ovaries and identified 278 discrete H3K27me3 domains, ranging from 3 to 240kb in width. The dhd H3K27me3 domain stood out when considering its enrichment in this mark relative to its length, compared to other domains (Fig 3B). Our analyses indicate that despite its reduced size, the dhd locus is capable of accumulating proportionately high amounts of this repressive modification. While we also identified other previously described H3K9me3 domains lodged in euchromatic regions (S3 Fig), the signal is, as expected, largely dominated by pericentric heterochromatin. We could thus not robustly call such euchromatic H3K9me3 domains, preventing us from conducting an analogous analysis for dhd on this mark.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. dhd is embedded in an H3K27me3/H3K9me3 mini-domain flanked by regulatory elements.

A—The dhd region features an H3K27me3 and H3K9me3 mini-domain. Genome browser snapshots showing the distribution of all fragments and <120 bp fragments in the dhd region, revealing that dhd lies within a ~5450bp heterochromatin mini-domain flanked by border regulatory elements associated with both H3K27me3 and H3K9me3. Arrows indicate direction of transcription. B-The dhd mini-domain is highly enriched in H3K27me3 relative to its size. Scatter plot of our 278 ovarian H3K27me3 domains identified in Cut&Run, representing their average read counts normalized to domain size in the Y-axis versus domain size in the X-axis. C–Effect of the KDs on H3K27me3 enrichment genome-wide. Average normalized counts of H3K27me3 Cut&Run (all fragments) in H3K27me3 domains (plotted as meta-domains and including ±3kb from domain borders) in Control (MTD>+), lid KD, Sin3a KD, mod(mdg4) KD, Snr1 KD and E(z) KD (arrows). D–lid KD, but not Sin3a, Snr1 or mod(mdg4), impacts H3K27me3 enrichment at the dhd mini-domain. Left: genome browser plots of normalized H3K27me3 Cut&Run signal (all fragments) at the dhd genomic region in Control and KD ovaries. Right: Quantification of normalized read counts for the same samples. Data in this figure is for one representative replicate: other replicates are shown in S5A Fig. E- Weak impact of KDs on heterochromatic marks at dhd. H3K27me3 and H3K9me3 Cut&Run-qPCR in Control and KD ovaries using the Sas10 gene as negative control and Ubx and CG12239 as positive controls for H3K27me3 and H3K9me3 respectively. Fold enrichment was calculated relative to Sas10. Error bars show technical variability from a representative replicate. Data in this figure is for one representative replicate: other replicates are shown in S5B Fig. P values indicate one-way ANOVA with Dunnett’s multiple comparisons test to a control (* P < 0.0001; n.s = not significant).

https://doi.org/10.1371/journal.pgen.1009615.g003

Surprisingly, short fragment peak analysis revealed two putative DNA regulatory elements associated with both histone marks, precisely at the mini-domain borders, with no internal peaks present (Fig 3A). ATAC-seq data from S2 cell lines confirmed presence of only these two border elements (S2B Fig). In addition, in Kc cells, dhd border elements are occupied by CP190 and Mod(mdg4), both of which can be found at the boundaries of Drosophila H3K27me3 domains [24,47]. However, the insulating isoform Mod(mdg4)67.2 is not found at the dhd locus (S2B Fig), suggesting that a different isoform playing an activator function is responsible for dhd regulation. Finally, the dhd-proximal 5’ border element featured a significant, although very modest enrichment for PRE markers Polycomb and Polyhomeotic (S2B Fig). This regulatory architecture was quite unusual, as we could not find any other H3K27me3 domain in the genome sharing this particular organization with two border elements and no internal elements. Together, these results revealed that dhd lies within a unique H3K27me3/H3K9me3 mini-domain featuring only border elements.

Sin3A, Snr1 and Mod(mdg4) control the regulatory architecture of the dhd mini-domain

We next aimed at evaluating the potential role for Lid, Sin3A, Mod(mdg4) and Snr1 in regulating this heterochromatin mini-domain. We used KD ovaries for these factors and included as a control a KD for the H3K27 methyltransferase Enhancer of zeste (E(z)), induced in germ cells by the MTD-Gal4 driver. While E(z) KD females were sterile as previously described [48,49] (Table 1), they were able to lay eggs and displayed only a moderate effect on dhd expression (a 25% reduction compared to controls) (S4A Fig). Immunofluorescence staining on dissected control ovaries showed that H3K27me3 marks follicle cell nuclei, the karyosome (i.e the oocyte nucleus) and nurse cell nuclei, although nurse cell staining was relatively weaker (S3B Fig), consistent with previous reports [48]. As expected, H3K27me3 was undetectable in the karyosome and in nurse cells of E(z) KD ovaries, whereas follicle cells (which do not express MTD-driven shRNAs) still carried this mark at normal levels. While lid, Sin3a and mod(mdg4) KD ovaries displayed normal H3K27me3 staining, we observed a moderate reduction in H3K27me3 levels in nurse cells in Snr1 KD ovaries, even while H3K27me3 levels were not affected on the karyosome (S4B Fig).

We next carried out H3K27me3 Cut&Run on ovaries from all KDs. Within our 278 identified H3K27me3 domains (see above), we compared the average enrichment in H3K27me3 signal in control and KD ovaries (Fig 3C). In E(z) KD ovaries, Cut&Run experiments revealed only a moderate loss of H3K27me3 signal (35% average reduction at these domains compared to controls) (Fig 3C), contrasting with the strong global reduction in H3K27me3 immunofluorescence signal. This difference is likely to reflect the fact that the H3K27me3 signal from Cut&Run experiments originates from both germline and somatic cells. Accordingly, E(z) KD completely abrogated H3K27me3 signal at the spen, Corto or ptc loci, all of which are decorated with H3K27me3 in nurse cells but not in follicle cells (S4C Fig) [50]. In contrast, the gl, dpp, or repo loci, which show stronger H3K27me3 in follicle cells compared to nurse cells, were only slightly affected in E(z) KD ovaries (S4C Fig). Together, these results show that our Cut&Run strategy detects H3K27me3 signal from both germline and somatic cells and is able to detect quantitative differences in the averaged signal when nurse cells are strongly affected. Consistent with immunofluorescence experiments, lid, Sin3a and mod(mdg4) KDs had only a modest global impact on average H3K27me3 levels (5% reduction compared to control), and no effect on the spread of H3K27me3 domains (Fig 3C). Also consistent with our immunofluorescence experiments, Snr1 KD led to a more severe average reduction of H3K27me3 Cut&Run signal compared to control (20%), although not as dramatic as E(z) KD.

In agreement with genome-wide observations, the levels of H3K27me3 in the dhd mini-domain were reduced in E(z) KD ovaries and unaffected in Sin3a or mod(mdg4) KD ovaries. More surprisingly, the domain was not measurably affected in Snr1 KD ovaries, despite the fact that H3K27me3 is globally impacted by this knockdown (Figs 3D and S5A). Within the sensitivity limits of our approach, these results indicate that Sin3a, Snr1 and mod(mdg4) KDs have little if any impact on H3K27me3 at the dhd locus. Conversely, in lid KD ovaries, in which global H3K27me3 levels were unaffected, we detected an increase in H3K27me3 levels at the dhd mini-domain (Figs 3C, 3D and S5A). This raised the possibility that Lid could facilitate dhd expression by counteracting Polycomb-mediated repression.

Since the dhd mini-domain also featured H3K9me3, we next turned to Cut&Run followed by qPCR to evaluate its status in KD ovaries. H3K27me3 Cut&Run-qPCR measures the expected enrichments at H3K27me3 domains and detects variations in the signal coherent with Cut&Run-seq results (Figs 3E and S5B). To validate the H3K9me3 Cut&Run-qPCR approach in ovaries, we exploited the CG12239 gene as a positive control [45], and detected an expected enrichment in H3K9me3 signal at this locus (Figs 3E and S5B). At the dhd locus, H3K9me3 was enriched as expected from ChIP-seq results. Importantly, knockdown of lid, Sin3a, mod(mdg4) or Snr1 had no effect on this enrichment (Fig 3E). The dhd heterochromatin mini-domain including H3K27me3 and H3K9me3 is thus independent of Sin3A, Snr1, whereas Lid counteracts H3K27me3.

We next evaluated the impact of different KDs on the dhd mini-domain short fragment peaks at border regions. We first analyzed the effect of our different knockdowns on the full set of 679 peaks previously defined (Fig 2B). Both E(z) and Snr1 KD led to a strong (~63%) decrease in short fragment peak average counts genome-wide (Fig 4A). Since these KDs also affect global H3K27me3 levels, this reduction could result from a general absence of histone modification-targeted MNase on chromatin. Remarkably, Sin3a KD led to a similarly strong effect on short fragment peak counts that could not be attributed to its global impact on H3K27me3. Instead, this data suggests that Sin3A is required to ensure proper occupancy and organization of transcription factors and/or nucleosomes at DNA regulatory elements associated with H3K27me3. In contrast, lid or mod(mdg4) KD did not globally affect short fragment peak counts, indicating that these factors do not play such a role (Fig 4A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Sin3A, Snr1 and Mod(mdg4) control the regulatory architecture of the dhd H3K27me3 mini-domain.

A–Sin3a and Snr1 KD, but not lid or mod(mdg4), impact H3K27me3-associated regulatory elements genome-wide. Left: H3K27me3 Cut&Run <120 bp fragments normalized counts in Control and KD ovaries, plotted at ±1kb around the summit of short fragment peaks. Right: Heatmaps displaying H3K27me3 Cut&Run short fragment peaks normalized read counts ±1kb around peak center in Control and KD ovaries. B- Sin3A, Snr1 and Mod(mdg4), but not Lid, impact the organization of regulatory elements at the borders of the dhd H3K27me3 mini-domain. Left: genome browser plots of normalized Control H3K27me3 Cut&Run signal (all fragments, top) and of normalized signal from <120bp fragments retrieved in H3K27me3 Cut&Run in Control and KD ovaries. Right: Quantification of <120bp fragments normalized read counts for the same samples. 5’ and 3’ border elements are plotted separately. Data in this figure is for one representative replicate: other replicates are shown in S4C Fig. Table recapitulating the effect of the different KDs on H3K27me3 and H3K27me3-associated regulatory elements (H3K27me3-RE) genome-wide and at the dhd locus. “≈” indicates modest or no change, “↗” indicates an increase and “↘” a decrease in average read counts compared to Control. “?” indicates inability to conclude.

https://doi.org/10.1371/journal.pgen.1009615.g004

Consistent with their effects genome-wide, short fragment counts at the dhd mini-domain border elements were strongly diminished upon E(z) and Sin3a KDs (Figs 4B and S5C). Intriguingly, mod(mdg4) KD led to a similar impact on these border elements (particularly the dhd-proximal one), even though it did not globally affect H3K27me3-associated elements genome-wide (Figs 4B and S5C). This observation could indicate that the dhd border elements become less frequently occupied by transcription factors, that these factors become less frequently associated with H3K27me3, and/or that their nucleosomal organization is compromised. In all cases, this suggests that Mod(mdg4) is required to ensure chromatin organization of the border DNA regulatory elements at the dhd mini-domain. Remarkably, Snr1 KD led to a similar effect on border elements without affecting H3K27me3 levels at the dhd mini-domain, suggesting that Snr1 is also required for the proper organization of the dhd border elements. In striking contrast, lid KD had no detectable effect on these regulatory elements (Figs 4B and S5C). We concluded that Lid, although essential for dhd expression, was not required to ensure the proper organization of dhd border elements.

Altogether our results, summarized in Fig 4C, indicate that Lid, Sin3A, Snr1 and Mod(mdg4), impact H3K27me3 or its associated regulatory elements genome-wide and/or at the dhd mini-domain in four distinct manners.

The dhd promoter-proximal DNA regulatory element is required for dhd expression independently of its heterochromatin mini-domain

We next performed sequence analysis of the dhd mini-domain border elements, screening against the flyreg.v2 transcription factor DNA binding motif database [51,52]. At the 5’ border element, which mapped to the dhd promoter region, we identified four perfect matches for the DNA replication-related element (DRE) motif, TATCGATA (Fig 5A). This motif is recognized by the insulator-associated factor BEAF-32 [53] and the core-promoter factor DREF [16]. These four DRE motifs overlap in the palindromic sequence TATCGATATCGATA, 37bp upstream of the dhd transcription start site. Consistently, BEAF-32 and DREF both occupy this element in Kc cells (S6 Fig) [38]. Previous studies showed that BEAF-32 null females are partially fertile (~40% hatching rate) [54], indicating that this factor is not essential for dhd expression. In turn, DREF is essential in a cell-autonomous manner and indeed dref mutations cause oogenesis defects [55]. Accordingly, we observed severe atrophy and failure to produce oocytes in dref KD ovaries. Because this precluded studying the role of DREF in dhd regulation, we instead sought to probe the importance of the DRE motifs themselves.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. The dhd promoter-proximal DRE motifs are required for its expression.

A–Schematic representation of the genotypes studied in this figure. Upper panels: genomic browser views recapitulating the Control distribution of H3K27me3 Cut&Run signal (all fragments and <120bp fragments) as well as RNA-seq signal from Figs 1 and 2 at the dhd locus and showing lack of signal at the transgene insertion locus in the absence of any transgenic construct. Middle panel: schematic representation of the genomic composition of w¹¹¹⁸ (reference strain), mutant and rescue flies, indicating the status of the dhd locus and the composition of the rescue transgene. Dashed lines indicate the targeted region by primer couples (primers R and 1–5) used for RT- and Cut&Run-qPCR in panels B and D. Bottom panel: sequence of the dhd promoter at the endogenous location (left) and in the ΔDRE mutant transgene where the 14bp containing the DRE motifs were deleted. B- The DRE motifs at the dhd promoter are necessary for its expression. RT-qPCR quantification of dhd mRNA levels in ovaries from wild-type w¹¹¹⁸ flies, dhd^J5 mutants or dhd^J5 mutants carrying either a WT (pW8-dhd^WT) or a mutant (pW8-dhd^ΔDRE) or a full domain (pW8-dhd^FD) transgene (measured using the R primers, normalized to rp49 and relative to expression in w¹¹¹⁸). Data from biological duplicates analyzed in technical duplicates are presented as mean ± SEM. P values indicate one-way ANOVA with Tukey’s multiple comparisons test (* P < 0.0001; n.s = not significant). C- The dhd heterochromatin domain is affected in dhd-containing transgenic constructs. Genome browser plots of normalized ovarian H3K27me3 and H3K9me3 Cut&Run signal at the dhd genomic region in the indicated genotypes. The H3K27me3 domain is abolished in all transgenic rescues. The H3K9me3 domain is partly affected in the dhd^J5 mutant, being lost the dhd-proximal end but maintained in the dhd-distal end. Dashed lines and numbers (1 to 5) indicate the targeted region by primer couples used for qPCR in panel D. D- A transgene containing the full dhd domain does not restore heterochromatin marks. H3K27me3 and H3K9me3 Cut&Run-qPCR in Control, dhd^J5, dhd^J5;;pW8-dhd^WT and dhd^J5;;pW8-dhd^FD ovaries. Fold enrichment was calculated relative to Sas10. Error bars show technical variability from a representative replicate. Data in this figure is for one representative replicate: other replicates are shown in S7 Fig.

https://doi.org/10.1371/journal.pgen.1009615.g005

The dhd^J5 null allele is a 1.4 kb deletion affecting the entire promoter region including the promoter-proximal regulatory element, and part of the coding region of dhd [7,9] (Fig 5A). A pW8-dhd^WT transgenic construct, bearing the entire dhd gene -including its promoter region-, restores dhd expression as well as fertility in dhd^J5 mutants [9] (Fig 5A, 5B and Table 1). We now constructed a second rescue transgene based on the pW8-dhd^WT, where the 14bp carrying the DRE motifs were deleted (pW8-dhd^ΔDRE) (Fig 5A). These constructs were inserted into the same genomic location as pW8-dhd^WT (62E1) and combined with the dhd^J5 deficiency. In striking contrast to pW8-dhd^WT, the pW8-dhd^ΔDRE construct was unable to rescue dhd expression, or substantially improve fertility in dhd^J5 deficient flies (Fig 5B and Table 1). The DRE motifs are thus essential to ensure dhd expression.

To test a role for this regulatory element and its DRE motifs in regulating the H3K27me3/H3K9me3 mini-domain, we performed Cut&Run-seq and Cut&Run-qPCR on homozygous dhd^J5 ovaries, as well as rescue dhd^J5;;pW8-dhd^WT and non-rescued dhd^J5;;pW8-dhd^ΔDRE ovaries. Strikingly, the 5.4kbp dhd H3K27me3 mini-domain was completely lost in dhd^J5 ovaries (Figs 5C and 5D and S7), despite the fact that 90% of this mini-domain were intact in the deficient chromosome. This indicates that the dhd-proximal border of this mini-domain is essential for establishment and/or maintenance of H3K27me3. Furthermore, H3K27me3 signal was absent within the mini-domain in dhd^J5;;pW8-dhd^WT rescue ovaries (Figs 5C and 5D and S7), suggesting that the 5’-most 2.8kbp of the domain are also insufficient to establish and/or maintain H3K27me3. This result further confirms that dhd can be expressed at high levels in the absence of H3K27me3, consistent with results from E(z) KD ovaries (S4A Fig). Finally, the H3K27me3 mini-domain was also completely absent in dhd^J5;;pW8-dhd^ΔDRE ovaries (Fig 5C), indicating that the DRE motifs are required for dhd expression independently of H3K27me3.

Based on these observations, we hypothesized that the complete mini-domain sequence, including both border regulatory elements, might be necessary for restoring heterochromatin marks. We thus constructed a transgene containing the full domain sequence of the dhd heterochromatin domain (pW8-dhd^FD) (Fig 5A), inserted at the same genomic location as the pW8-dhd^WT transgene. Interestingly, this transgene restored dhd expression (Fig 5B) and rescued fertility (Table 1) but was unable to restore H3K27me3 (Figs 5D and S7). These results indicate that the border-to-border mini-domain is not autonomous and suggest that its genomic location impacts its chromatin configuration.

We next focused on the H3K9me3 mark. In contrast to H3K27me3, Cut&Run analysis in dhd^J5 mutants showed that H3K9me3 was lost at the dhd-proximal half of the domain, while this mark was maintained at the dhd-distal part (Fig 5C). Cut&Run-qPCR using primers across the dhd domain revealed that H3K9me3 was not restored at the dhd-proximal part of the H3K9me3 domain in dhd^J5;;pW8-dhd^WT or dhd^J5;;pW8-dhd^FD ovaries (Figs 5D and S7). Together, these results indicate that the border-to-border mini-domain is not autonomous to establish its own heterochromatin configuration, and that dhd expression can proceed at near normal levels independently of these marks.

Lid, Sin3A, Snr1 and Mod(mdg4) activate dhd independently of its heterochromatin mini-domain

The fact that the pW8-dhd^WT transgene restored most of dhd expression without re-establishment of the heterochromatin mini-domain at this locus provided an opportunity to clarify the role of our set of dhd regulators. KD of lid is associated with increased H3K27me3 at the dhd mini-domain, suggesting that Lid may operate as an anti-repressor by counteracting heterochromatinization of the locus. However, we have previously found that dhd expression is not re-established in lid KD ovaries carrying a pW8-dhd^WT rescue transgene [5]. Lid is thus required for dhd expression not only at its endogenous locus but also from the rescue transgene not decorated by H3K27me3 (Fig 5). Therefore, Lid activates dhd independently of heterochromatin, suggesting that it does not operate strictly as an anti-repressor.

To discriminate between anti-repressive or activating roles of Sin3A, Mod(mdg4) and Snr1, we generated flies combining a dhd^J5 deficiency, the pW8-dhd^WT transgene and an shRNA targeting lid, Sin3a, Snr1 or mod(mdg4), driven in germ cells by a nos-Gal4 driver (Fig 6A). We confirmed by RT-qPCR that knockdowns were still efficient when using this driver (S8A Fig). Remarkably, all of these flies were almost completely sterile, and showed strong downregulation of dhd revealed by RT-qPCR (Fig 6B and Table 1). Using Cut&Run-qPCR at the dhd locus, we further confirmed that these knockdowns had no effect on H3K27me3, which remained depleted in all conditions (Figs 6C and S8B). Lid, Sin3A, Snr1 and Mod(mdg4) therefore stimulate dhd transcription in the absence of its heterochromatin mini-domain.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Lid, Sin3A, Mod(mdg4) and Snr1 are necessary for dhd expression in the absence of its heterochromatin domain.

A–Schematic representation of the genomic composition of w¹¹¹⁸ (reference strain) and mutant flies carrying a rescue transgene and shRNA constructs controlled by the female germline specific nanos-Gal4 driver, respectively inserted at the platforms attP 3B- 62E1 and attP 2-68A4. B-The rescue transgene does not restore dhd expression in KD ovaries. RT-qPCR quantification of dhd mRNA levels in ovaries of the indicated genotypes (normalized to rp49 and relative to expression in w¹¹¹⁸ ovaries). Data from biological duplicates analyzed in technical duplicates are presented as mean ± SEM. P value indicates one-way ANOVA with Tukey’s multiple comparisons test (* P < 0.0001), C-H3K27me3 is absent from the dhd rescue transgene. H3K27me3 Cut&Run-qPCR in the indicated genotypes. The Sas10 gene was used as negative control and Ubx as positive control. Fold enrichment was calculated relative to Sas10. Error bars show technical variability from a representative replicate. Error bars show technical variability from a representative replicate. Data in this figure is for one representative replicate: other replicates are shown in S8 Fig. D- dhd is not ectopically expressed in adult tissues in the absence of its heterochromatin domain. RT-qPCR quantification of dhd mRNA levels in dissected ovaries or corresponding female carcasses as well as testes or corresponding male carcasses, in all indicated genotypes (normalized to rp49 and relative to expression in ovaries in w¹¹¹⁸). Data from biological duplicates analyzed in technical duplicates are presented as mean ± SEM. P value indicates one-way ANOVA with Tukey’s multiple comparisons test (* P < 0.0001).

https://doi.org/10.1371/journal.pgen.1009615.g006

Our results indicate that the dhd heterochromatin mini-domain does not play a repressive role in ovaries, but do not exclude that it might maintain dhd silent in other tissues. RT-qPCR analysis on dissected ovaries, testes and male and female carcasses from transgenic lines expressing dhd^J5;;pW8-dhd^WT revealed dhd expression uniquely from ovaries (Fig 6D). Because this transgene rescues dhd expression without restoring heterochromatin marks, these results suggest that the dhd heterochromatin mini-domain is not essential to repress ectopic dhd expression in adults. We note, however, that we cannot exclude that dhd was weakly and/or transiently expressed in certain cell types in these conditions, or that the rescue transgene could accumulate repressive marks in tissues other than ovaries.

The ovarian hyperactivation of dhd

Here, we sought to understand how the genomic and epigenomic environments of dhd contributed to its remarkable regulation, its expression being both among the highest in Drosophila, and absolutely specific to adult ovaries [5,9]. Lid, Sin3A, Snr1 and Mod(mdg4) all shared a critical and rather specific role in ensuring dhd expression. Yet, these four broadly expressed proteins play multiple roles other than dhd regulation. For example, transcriptomic analyses following individual depletion of Lid, Sin3A or Snr1 in S2 cells, wing discs or pupae shows activation or repression of hundreds of targets [19,22,56]. ChIP-seq data further indicates that Mod(mdg4), Sin3A and Lid each target several thousand sites in the genome [24,26,57,58]. Consistently, our RNA-seq analyses did reveal that each of these knockdowns were associated to up- or down-regulation of 407 to 2020 genes in ovaries, with dhd being in every case among the most strongly dependent on these factors. We propose that dhd is a hypersensitive gene that reacts radically to epigenome imbalances.

The key question is therefore what is the formula for dhd ovarian hyperactivation. One reasonable hypothesis was that dhd could be highly regulated by distal enhancers. This would be notably consistent with the previously described role of Mod(mdg4) in organizing 3D contacts between regulatory elements and promoters [26]. It would also be consistent with recent findings that H3K27me3 micro-domains may reflect such contacts [59]. However, no interaction between dhd and any other locus can be found in Hi-C data, and our rescue transgene experiments show that a small, ectopic genomic segment almost fully recapitulates its expression, arguing that the genomic and epigenomic environment at the endogenous dhd region may play only a minor role in its ovary-specific hyperactivation.

We indeed found a key regulatory element containing a tandem DRE motif, known to recruit the DREF core promoter factor. The minimal DRE motif (TATCGATA) is found in thousands of gene promoters [60], while multiple genes were individually shown to require this motif for proper activation. These include genes with ovarian expression, and, accordingly, DREF mutations cause oogenesis defects and female sterility [55]. In contrast, the particular tandem DRE motif in the dhd regulatory sequence is uncommon, being only found in 9 other gene promoters. Yet, among these 9 genes, only 4 displayed an expression bias in ovaries, and none were nearly as highly transcribed as dhd. Therefore, this motif does not seem to be autonomously sufficient for ovarian hyperexpression.

Another unusual feature of dhd is its surrounding heterochromatin mini-domain bearing both H3K27me3 and H3K9me3 marks, as well as H3K4me3. The co-occurrence of these active and repressive modifications at an ensemble of developmentally regulated genes in mammals led to the concept of bivalent promoters [61]. It is speculated that such promoters may be poised for rapid activation or repression upon differentiation. In Drosophila, bivalent chromatin is associated with genes that can be strongly activated in a tissue-specific manner [62,63]. Our experiments showed that dhd is expressed at ~60–70% of its normal levels in E(z) KD ovaries, as well as in rescue transgenes—conditions in which the H3K27me3/H3K9me3 mini-domain is impaired. We thus cannot exclude that these heterochromatin marks play a positive role in dhd activation to ensure its transcription at maximum capacity, perhaps via establishment of a bivalent configuration. We also note that our whole-tissue experiments leave the possibility open that these histone modifications may decorate dhd in different cell types and/or at different times during gametogenesis.

Altogether, we uncovered multiple unusual genomic and epigenomic characteristics at the dhd locus, but failed to identify any single feature that was truly defining. The dramatic regulation of dhd may rely not on any individual trait but rather on a unique combination of such rare features. Further work will be needed to elucidate how these different components may together achieve ovarian hyperexpression.

A non-canonical chromatin domain

The unique properties of dhd led us to uncover interesting features of its epigenomic regulators. First, dhd is embedded in an H3K27me3 mini-domain flanked by regulatory elements. A recent report suggested that H3K27me3 domain borders may be established independently of PREs or border elements, provided that an immediately neighboring active gene instead delimits H3K27me3 spreading [47]. The case of dhd is however peculiar in that the H3K27me3 domain border overlaps with this highly active gene, a scenario that was not found in other domains. The coincidence of H3K27me3 and H3K9me3 is also uncommon and, in fact, we could not find any other such dual domain in ovaries. Our result favors the view that this heterochromatin domain does not silence dhd expression. Nonetheless, H3K9me3 was always maintained at the dhd-distal portion of the domain. We therefore cannot exclude that this mark represses dhd neighbors.

An intriguing question is then how this heterochromatin mini-domain is formed. We found that a transgene containing the full mini-domain sequence is unable to restore H3K27me3 or H3K9me3, suggesting that genomic location of this domain is a critical determinant of dhd heterochromatin. From this perspective, the border elements may act as weak PREs, as described in other contexts [64]. Interestingly, we also showed that H3K9me3 can be partially maintained at the distal part of the domain in dhd^J5 mutants while H3K27me3 is completely lost. This indicates that these marks are not necessarily inter-dependent at this locus, and H3K9me3 may benefit from additional mechanisms ensuring its deposition. While not much is known on highly localized euchromatic deposition of H3K9me3, Smolko and colleagues suggested that Setdb1-dependent accumulation of H3K9me3 at certain target genes is dependent on the RNA-binding protein Sxl [45]. At the dhd domain, different mechanisms could thus ensure H3K27me3 and H3K9me deposition, both of which would depend on the endogenous genomic location.

Along these lines, another recent study reported the existence of H3K27me3 micro-domains (typically 2–8 nucleosomes wide) that depend on 3D contacts with larger H3K27me3 domains, mediated, in particular, by BEAF-32 and CP190 [59]. The dhd mini-domain is wider and much more strongly enriched in H3K27me3 than typical micro-domains. Nonetheless, our data is consistent with a model whereby H3K27me3 could be deposited via such looping interactions. First, BEAF-32 and CP190 are indeed found at the border elements of the dhd mini-domain. Second, this mini-domain does not feature internal PREs and border elements are only weakly if at all bound by Polycomb proteins, arguing against an autonomous recruitment of E(z). Finally, a deletion of the BEAF-32/CP190-bearing regulatory element in the dhd^J5 mutant, or its displacement to an ectopic genomic location in rescue transgenes both abrogate H3K27me3 deposition. Consistent with such a model, data from Heurteau et al. show a modest reduction of H3K27me3 enrichment at the dhd mini-domain upon BEAF-32 depletion. Of note, BEAF-32 was also previously shown to facilitate H3K9me3 deposition at sites featuring multiple instances of the CGATA motif, analogous to those found at the dhd promoter [65]. Other studies found that ATCGAT motifs recognized by BEAF-32, also found at the dhd promoter, are more broadly enriched at the promoters of Lid-activated genes [66], which is the case of dhd. Thus, it is possible that a BEAF-32-mediated looping mechanism is responsible for H3K27me3 enrichment at the dhd mini-domain. However, our results also show that this mark is not strictly required to repress nor to activate dhd in adults, and that Lid, Sin3A, Snr1 and Mod(mdg4) activate dhd independently of it.

Scrutiny of dhd regulation further uncovered how its four regulators have convergent yet distinct roles. This was particularly intriguing for Lid and Sin3A, which can be found in a co-repressor complex [20], at odds with their positive impact on dhd. Indeed, their dual depletion in cultured cells causes the misregulation of hundreds of genes [19]. Interestingly, in that study, only 55 out of 849 affected genes were similarly impacted by individual and dual knockdowns, indicating that Lid and Sin3A functionally cooperate only at a minor subset of their common targets. This seems to be the case at the dhd locus, where individual KD of these factors caused an equally catastrophic collapse of transcriptional activity, suggesting a cooperative activity. Yet, Lid, but not Sin3A, acted as a negative regulator of H3K27me3 at the dhd locus, revealing at least partially independent functions. In contrast, Sin3A, but not Lid, controlled the stability of regulatory elements associated with this H3K27me3, not only at the dhd domain but also genome-wide.

Our results further show a critical role for Mod(mdg4) as a transcriptional activator. In cell lines, ChIP-seq experiments specifically mapping the insulating Mod(mdg4)67.2 isoform or total Mod(mdg4) showed that additional isoforms are recruited to DNA [24]. Isoforms other than the 67.2 were found in particular at gene promoters in ovaries and female heads [4]. Such is the case at the dhd promoter, where total Mod(mdg4) is found but not the 67.2 isoform (S2B Fig). Non-insulating roles of Mod(mdg4) were previously discussed in the context of the Polycomb-repressed Bithorax complex where the close binding of Mod(mdg4) to Abd-B transcription start sites suggested a role in transcription activation [26]. While we cannot rule out indirect effects, these observations argue that an activating isoform of Mod(mdg4) operates directly at the dhd promoter. In agreement, Mod(mdg4) appears to be essential to activate dhd within its H3K27me3 mini-domain, seemingly by stabilizing the dhd promoter regulatory element, although its function is equally essential in the absence of heterochromatin marks in the dhd transgenic rescue construct.

The Snr1-containing Brahma complex is required for activation of target genes in Drosophila in vivo, notably during immune responses [60] and tissue regeneration [56]. In ovaries, while Snr1 has a global impact on nuclear integrity and architecture, previous immunostaining experiments interestingly showed that this factor is only expressed during a restricted time in early oogenesis [22]. This underlines the fact that dhd may be dynamically regulated during oogenesis, with different regulatory components intervening at particular times. Considering that Snr1 KD causes a disruption of the dhd promoter-proximal regulatory element associated with H3K27me3, this would suggest that its associated DNA-binding transcription factors also intervene during a restricted time in oogenesis. A precise dissection of the timing of dhd transcription, and determining whether these factors target dhd directly and simultaneously, would be essential to understand the cascade of events leading to its massive expression.

The case of dhd indeed illustrates the complexity of understanding the chromatin landscape at cell type-specific genes, when the starting material is a complex tissue. In this context, the Cut&Run analysis implemented in our study allowed us to reveal the co-occupancy of H3K27me3 nucleosomes and associated transcription factors. While this approach cannot identify the cell of origin of each individual DNA molecule, it can be used to make important deductions on the combinatorial co-occupancy on DNA of different chromatin components. This approach joins other recent methods comparable in their principle, namely the DNA methyl-transferase single-molecule footprinting (dSMF) method [67] and the low-salt antibody-targeted tagmentation (CUTAC) approach [68]. Together with single-cell methodologies, these approaches hold the potential to begin uncovering complex epigenomic regulation processes, such as that of dhd, that were until recently inaccessible.

Drosophila strains

Flies were raised at 25°C on standard medium. The following stocks were obtained from the Bloomington Drosophila Stock Center (simplified genotypes are given): P{TRiP.HMS00849}attP2 (mod(mdg4) shRNA; #33907), P{TRiP. HMS00363}attP2 (Snr1 shRNA; #32372), P{TRiP.GL00612}attP40 (lid shRNA; #36652), P{TRiP.GLV21071}attP2 (lid shRNA; #35706), P{TRiP.HMS00359}attP2 (Sin3a shRNA; #32368), P{TRiP.HMS00066}attP2 (E(z) shRNA; #33659), P{y[+t7.7] = CaryP}attP2 (Control line for TRiP RNAi lines; #B36303), P{otu-GAL4::VP16.R}1; P{GAL4-nos.NGT}40; P{GAL4::VP16-nos.UTR}MVD1 (Maternal Triple Driver or “MTD-Gal4”; #31777), P{GAL4::VP16-nos.UTR}MVD1 (“nos-Gal4”; #4937). Other stocks are: w¹¹¹⁸, Df(1)J5/FM7c [7], P[Mst35Ba-EGFP] [27], pW8-dhd^WT [9]. TRiP lines target all predicted isoforms of their respective target genes. “Control” in shRNA experiments refers to the offspring of the control line for TRiP lines crossed with the MTD-Gal4 line.

For the pW8-dhd^ΔDRE mutant, two fragments were amplified by PCR from w¹¹¹⁸ genomic DNA using the primers ΔDRE-1-for/ ΔDRE-1-rev and ΔDRE-2-for/ ΔDRE-2-rev (S1 Table). PCR products were assembled and cloned into the pW8-dhd^WT vector [9] previously digested by KpnI and BamHI using the NEBuilder HiFi DNA Assembly Cloning Kit (NEB, #E5520S). dhd^ΔDRE transgene was integrated in the PBac{attP-3B}VK00031 platform (62E1) using PhiC31-mediated transformation and flies were generated by The Best Gene (TheBestGene.com).

Germline knock-down and fertility tests

To obtain KD females, virgin shRNA transgenic females were mass crossed with transgenic Gal4 males at 25°C and females of the desired genotype were recovered in the F1 progeny. All RNAi experiments were carried at 25°C. To measure fertility, virgin females of different genotypes were mated to males in a 1:1 ratio and placed for 2 days at 25°C. They were then transferred to a new vial and allowed to lay eggs for 24 hours. Embryos were counted and then let to develop for at least 36 hours at 25°C. Unhatched embryos were counted to determine hatching rates.

Gene expression analysis by RT-QPCR

Total RNA was extracted from ovaries of 3-day-old females using the NucleoSpin RNA isolation kit (Macherey-Nagel), following the instructions of the manufacturer. 1μg of total RNA was reverse transcribed using the SuperScript II Reverse Trancriptase kit (Invitrogen) with oligo (dT) primers. RT-qPCR reactions were performed in duplicates as described previously [5]. Primer sets used are provided in S1 Table. Statistical tests were performed using GraphPad Prism version 9.2.0 for Mac OS (GraphPad Software).

Immunofluorescence and imaging

Early (0–30 min) embryos laid by females of the indicated genotypes were collected on agar plates. Embryos were dechorionated in bleach, fixed in a 1:1 heptane:methanol mixture and stored at -20°C. Embryos were washed three times (10 min each) with PBS 0.1%, Triton X-100 (PBS-T) and then incubated with primary antibodies in the same buffer on a wheel overnight at 4°C. They were then washed three times (20 min each) with PBS-T. Incubations with secondary antibodies were performed identically. Embryos were mounted in DAKO mounting medium containing DAPI.

Ovaries were dissected in PBS-T and fixed at room temperature in 4% formaldehyde in PBS for 25 minutes. Immunofluorescence was performed as for embryos. Ovaries were then mounted as described above.

Antibodies used are provided in S2 Table. Images were acquired on an LSM 800 confocal microscope (Carl Zeiss). Images were processed with Zen imaging software (Carl Zeiss) and ImageJ software.

Western blotting

Ovaries from 30 females were collected and homogenized in lysis buffer (20mM Hepes pH7.9, 100mM KCl, 0.1mM EDTA, 0.1mM EGTA, 5% Glycerol, 0.05% Igepal and protease inhibitors (Roche)). Protein extracts were cleared by centrifugation and purified with Pierce GST Spin Purification Kit (ThermoFisher Scientific, #16106). Western analysis was performed using standard procedures and used antibodies and concentrations are presented in S2 Table.

Ovarian RNA sequencing and analysis

Samples were processed as previously described [5].

Sequencing was completed on two biological replicates of the following genotypes:

mod(mdg4) KD (MTD-Gal4>shRNA mod(mdg4)), i.e

P{w[+mC] = otu-GAL4::VP16.R}1, w[*]/y[1] sc[*] v[1]; P{w[+mC] = GAL4-nos.NGT}40/+; P{w[+mC] = GAL4::VP16-nos.UTR}CG6325[MVD1]/P{y[+t7.7] v[+t1.8] = TRiP. HMS00849} attP2

Snr1 KD (MTD-Gal4>shRNA Snr1), i.e

P{w[+mC] = otu-GAL4::VP16.R}1, w[*]/y[1] sc[*] v[1];P{w[+mC] = GAL4-nos.NGT}40/+; P{w [+mC] = GAL4::VP16-nos.UTR}CG6325[MVD1]/P{y[+t7.7] v[+t1.8] = TRiP. HMS00363}attP2

Chromatin profiling by CUT&RUN

Cut&Run in Drosophila tissues was previously described [37]. Briefly, ovaries from 3-day-old flies were dissected in Wash+ Buffer (20 mM HEPES pH 7.5, 150 mM NaCl, 0.9 mM spermidine, 0.1% BSA with cOmplete protease inhibitor, Roche) and were bound to BioMag Plus Concanavalin-A-conjugated magnetic beads (ConA beads, Polysciences, Inc). Tissues were then permeabilized for 10min in dbe+ Buffer (20 mM HEPES pH 7.5, 150 mM NaCl, 0.9 mM spermidine, 2 mM EDTA, 0.1% BSA, 0.05% digitonin and protease inhibitors). Samples were then incubated with gentle rocking overnight at 4°C with primary antibody solution in dbe+ buffer (see S2 Table for antibody concentrations). Protein A fused to micrococcal nuclease (p-AMNase) was added in dbe+ buffer and samples were incubated with rotation at room temperature for 1 hour. Cleavage was done in WashCa+ buffer (20 mM HEPES pH 7.5, 150 mM NaCl, 0.9 mM spermidine, 0.1% BSA, 2 mM CaCl2 with and protease inhibitors) at 0° for 30 minutes. Digestion was stopped with addition of 2XSTOP Buffer (200mM NaCl, 20mM EDTA, 4mM EGTA, 62.5μg/mL RNaseA). Samples were incubated at 37°C for 30 min to digest RNA and release DNA fragments. Cleaved DNA was then recovered with Ampure XP beads (Beckman Coulter) immediately after protease treatment. Antibodies used for CUT&RUN are listed in S2 Table. Retrieved DNA was used either for qPCR or for library preparation followed by deep sequencing. Sequencing libraries for each sample were synthesized using Diagenode MicroPlex Library Preparation kit according to supplier recommendations (version 2.02.15) and were sequenced on Illumina Hiseq 4000 sequencer as Paired-End 100 base reads following Illumina’s instructions (GenomEast platform, IGBM, Strasbourg, France). Image analysis and base calling were performed using RTA 2.7.7 and bcl2fastq 2.17.1.14. Adapter dimer reads were removed using DimerRemover.

Cut&Rut-qPCR

0,1 ng of retrieved DNA in Cut&Run were used as template in a real time quantitative PCR assay using SYBR Premix Ex Taq II (Tli RNaseH Plus) (Takara). All qPCR reactions were performed in duplicates using Bio-Rad CFX-96 Connect system with the following conditions: 95°C for 30s followed by 40 cycles of denaturation at 95°C for 5s and annealing and extension at 59°C for 30s. As a normalization control, we processed ovary samples from each studied genotype as for Cut&Run, except the antibody and pA-MNase incubation steps were omitted and instead we incubated tissue with 10U of Micrococcal Nuclease for 30min at 37°C (ThermoFisher Scientific, #88216). Fold change in histone mark enrichment was determined relative to this whole MNase control and relative to the Sas10 gene, which was depleted in the histone marks tested in this study. Primer sets used are provided in S1 Table.

Sequencing data processing

Paired-end reads were mapped to the release 6 of the D. melanogaster genome using Bowtie2 (v. 2.4.2). To compare samples with identical readcount for genome coverage quantifications, we employed Downsample SAM/BAM (Galaxy Version 2.18.2.1). To obtain short fragment datasets for DNA regulatory element identification, peak calling and visualization, we selected fragments shorter than 120 bp from SAM files. These were typically a small minority of all fragments, as our Cut&Run datasets were largely dominated by nucleosome-sized fragments (150-250bp). We therefore could separately analyze modification-bearing nucleosome coverage (for which the complete Cut&Run dataset, “All fragments”, was a good approximation) or putative regulatory element coverage (<120bp fragments). For genomic track visualization, we used bamCoverage from deepTools 2.0 (Galaxy Version 3.3.2.0.0) to calculate read coverage per 25bp bin for all fragments or 10bp bin for short fragments, with paired-end extension. Peak calling was done on sorted short fragments (<120 bp) with MACS2 (v. 2.1.1.20160309) with the following parameters:–nomodel,–p-value = 0.0001,–keep-dup = all and the rest by default. To establish a high-confidence short fragment peak list we retained peaks that were present in biological replicates from the control genotype. Genome browser views screenshots were produced with the IGV software, for Cut&Run we used a 25bp bin for all fragments and a 10bp bin for short fragments (<120bp). For the midpoint-plot of fragment sizes around short fragment peaks, the length of each fragment was plotted as a function of the distance from the fragment midpoint to the summit of the peak identified by MACS2. For signal quantification at the dhd locus, normalized read counts were counted within the domain (Coordinates [5,312,54–5,317,465]) or within border element peaks (5’ element: [5,312,120–5,312,211], 3’ element: [5,317,300–5,317,447])

Heatmaps were generated with RStudio (RStudio Team (2016). RStudio: Integrated Development for R. RStudio, Inc., Boston, MA URL) and the packages ‘gplots’ (v.3.1.1) and ‘plyr’ (Wickham, 2011).

Motif scanning

Motif scanning on the pW8-dhd^WT transgene sequence was done with FIMO (v. 5.3.3) [69] using the flyreg v.2 motif database with default parameters.