Su(Hw) enables the indirect binding of Combgap to chromatin

While loss of Su(Hw) binding to chromatin is not lethal for flies, it does lead to female sterility through a mid-stage arrest of oogenesis and egg chamber degeneration at approximately stage 9 [17, 18]. Thus, Drosophila ovaries are the most biologically relevant organ to study Su(Hw) functioning. In this study, we used su(Hw)v/E8 heteroallelic mutant flies (henceforth referred to as Su(Hw) loss of function or Su(Hw)LOF flies), which completely lack Su(Hw) binding to chromatin due to a point mutation in the su(Hw) coding region for zinc finger 7 [18]. Female flies carrying su(Hw)v/E8 alleles are sterile and experience egg chamber degeneration at stage 9 [17, 18]. To compare wild type and Su(Hw)LOF ovaries, we dissected ovaries from recently eclosed 15-hour-old flies that contain egg chambers no older than stage 8 (following the procedure in [19]).

In our previous study [19], we had shown that Su(Hw) ChIP-Seq peaks in the ovaries included a cluster of peaks that lacked the Su(Hw) DNA-binding motif (henceforth referred to as as indirect Su(Hw) peaks). Instead, these indirect Su(Hw) peaks contained a GTGT-motif, previously associated with the binding of Combgap, a protein involved in the recruitment of PcG group proteins [15, 16]. Here, we confirmed the binding of Combgap to these indirect Su(Hw) peaks both in wild-type and Su(Hw)LOF ovaries using ChIP-Seq with antibodies against Combgap (Fig. 1a-c). The majority of indirect Su(Hw) peaks are included in the set of Combgap peaks: 78% of indirect Su(Hw) peaks (515 out of 659 peaks) intersect with Combgap peaks (Fig. 1d). Additionally, we observed that Combgap also binds to a portion of direct Su(Hw) peaks (492 peaks out of 3166 direct Su(Hw) peaks), and this binding is Su(Hw)-dependent (Fig. 1a-b, Supplementary Fig. 1a). Using ChIP experiments coupled with qPCR, we confirmed the binding of Combgap to direct and indirect Su(Hw) ChIP peaks (Fig. 1c). We selected the 100 Su(Hw) peaks with the strongest Combgap binding (direct Su(Hw) peaks) and analysed the DNA motifs present in this set using the MEME-ChIP program [20]. We found a strong enrichment only for the Su(Hw) DNA motif, not the Combgap DNA motif, indicating a preference for Su(Hw)-mediated Combgap recruitment to these sites (Supplementary Fig. 1b).

Fig. 1

Combgap binds Su(Hw) ChIP-Seq peaks in Su(Hw)-dependant manner. (a) Heatmaps of Su(Hw) and Combgap ChIP/Inp signal on Su(Hw) ChIP-Seq peaks with direct and indirect Su(Hw) binding. Heatmaps are made for the wild type (WT, green) and Su(Hw)LOF (orange) Drosophila ovaries and are sorted by the strength of the median Su(Hw) ChIP/Inp signal in the wild type Drosophila ovaries. The DNA-motifs determined by MEME suite 5.4.1 [20, 55] are shown on the left from each group of peaks. (b) Genome browser (IGV) examples highlighting that Combgap binding to 62D and 50A insulators is Su(Hw)-dependent. (c) ChIP analysis of Su(Hw) and Combgap binding to direct and indirect Su(Hw) ChIP-Seq peaks in the wild type (green columns) and Su(Hw)LOF (orange columns) ovaries assessed by qRT-PCR. Well-known Su(Hw)-dependent insulators (62D, 1A2, 50A, 66E, 87E) were used as direct Su(Hw) ChIP-Seq peaks. The Y-axis represents the % of input chromatin fraction. The gray area on the diagrams indicates the Su(Hw) and Combgap binding levels on 1A1 negative control region in the wild type ovaries. The data are mean values from three independent experiments, error bars represent standard deviations. (d) Intersection of Combgap ChIP-Seq peaks with direct and indirect Su(Hw) ChIP-Seq peaks in the wild type Drosophila ovaries. (e) Immunoprecipitations (IPs) from nuclear protein extracts of Drosophila S2 cells. IPs were performed with antibodies against Combgap (Cg) and Su(Hw) (the total purified IgG antibodies of non-immunized rabbits (IgG) were used as a negative control), which is indicated on the top of the figure. Western blots were stained with the antibodies indicated on the left of the figure. Anti-lamin staining was used as a loading control for input protein extracts. All input and IP samples were loaded on a single western blot. The original Western blots are present in Supplementary Fig. 10. The numbers above the inputs and IPs represent a portion of a loaded fraction (in respect to the amount used for IPs)

To validate the ChIP-Seq findings, we conducted co-immunoprecipitation experiments (co-IPs) in nuclear protein extract from Drosophila S2 cells to test an interaction between Su(Hw) and Combgap. Antibodies against Su(Hw) successfully co-precipitated Combgap from the nuclear extract and vice versa (Fig. 1e). Moreover, recent immunoaffinity purification of the Combgap interactome coupled with high throughput mass spectrometry (IP/LC-MS) revealed statistically significant enrichment of Su(Hw) and its partners Mod(mdg4) and CP190, providing further support for the existence of protein-protein interaction between Su(Hw) and Combgap [21].

Su(Hw)-bound Combgap is associated with active chromatin rather that polycomb-directed repression

Originally, Combgap and Su(Hw) were characterized as proteins linked to repressed chromatin: Combgap has been identified as a protein capable of recruiting the Ph subunit of the PRC1 transcription repression complex to chromatin [15] and approximately 86% of Su(Hw) ChIP peaks were observed within repressed chromatin regions in the S2 cell line [22, 23]. However, subsequent research has shown that both Combgap and Su(Hw) also interact with proteins associated with active chromatin: Combgap ChIP-Seq peaks have been found to colocalise with RNA polymerase II pausing factors and the transcription start sites (TSSs) of active genes [21] while Su(Hw) has been found to co-immunoprecipitate with several transcription coactivators [19, 23]. This prompted us to investigate the chromatin features associated with Su(Hw)-Combgap interactions.

To determine whether the H3K27me3 Polycomb-dependent chromatin mark is associated with Su(Hw)-Combgap interactions in Drosophila ovary we utilised the H3K27me3 ChIP-Seq from 512c germline cells nuclei, FACS-sorted from the ovaries of 3–6 days old females of Drosophila melanogaster [24]. We examined the distribution of H3K27me3 on 492 direct Su(Hw) ChIP peaks that intersect with Combgap, 515 indirect Su(Hw) ChIP peaks that intersect with Combgap, and on H3K27me3 broad domains, annotated from this H3K27me3 ChIP-Seq (Fig. 2a). Results show no H3K27me3 enrichment, neither on direct nor indirect Su(Hw) peaks with Combgap, compared to H3K27me3 broad domains. Hence, the Su(Hw)-Combgap interaction does not appear to be associated with Polycomb-mediated heterochromatin establishment.

Fig. 2

Su(Hw)-bound Combgap is associated with active chromatin factors. (a) Heatmaps and pile-up profiles of H3K27me3 ChIP/Inp signal from 512c FACS-sorted germline cells nuclei [24] on direct and indirect ovarian Su(Hw) ChIP peaks, intersecting with Combgap, and H3K27me3 broad peaks. Heatmaps are sorted by the strength of the median H3K27me3 ChIP/Inp signal. (b) Genome browser (IGV) example highlighting that NELF E, Paf1 and Rpb3 binding to direct Su(Hw) ChIP-Seq peaks coincides with Combgap binding and is Su(Hw)-dependent. The position of direct Su(Hw) ChIP-Seq peak is marked on top of the tracks with an arrow. (c) ChIP analysis of NELF E, Paf1 and Rpb3 binding to direct and indirect Su(Hw) ChIP-Seq peaks in the wild type (green columns) and Su(Hw)LOF (orange columns) ovaries assessed by qRT-PCR. Well-known Su(Hw)-dependent insulators (62D, 1A2, 50A, 66E, 87E) were used as direct Su(Hw) ChIP-Seq peaks. The Y-axis represents the % of input chromatin fraction. The gray area on the diagrams indicates the NELF E, Paf1 and Rpb3 binding levels on 1A1 negative control region in the wild type ovaries. The data are mean values from three independent experiments, error bars represent standard deviations. (d) Heatmaps of Combgap, NELF E subunit of NELF complex, Paf1 positive elongation factor and Rpb3 RNAPII subunit ChIP/Inp signals on direct Su(Hw) ChIP-Seq peaks, intersecting with Combgap, and on indirect Su(Hw) ChIP-Seq peaks, intersecting with Combgap. Heatmaps are made for the wild type (WT, green) and Su(Hw)LOF (orange) Drosophila ovaries and are sorted by the strength of the median NELF E ChIP/Inp signal in the wild type Drosophila ovaries

To investigate the potential link between Su(Hw)-Combgap interactions and RNAP II pausing factors, we conducted ChIP-Seq experiments on wild-type and Su(Hw)LOF ovaries using antibodies against the NELF E subunit of NELF, the Paf1 positive elongation factor, and the Rpb3 RNAP II subunit. We found significant Su(Hw)-dependent enrichment of NELF E, Paf1, and Rpb3 at direct Su(Hw) peaks containing Combgap (Fig. 2b-d, Supplementary Fig. 2a). The binding of NELF E, Paf1 and Rpb3 to direct Su(Hw) ChIP peaks was further verified through ChIP experiments coupled with qPCR (Fig. 2c). Notably, direct Su(Hw) peaks bound to NELF E (326 peaks), Paf1 (398 peaks), and Rpb3 (277 peaks) in wild-type Drosophila ovaries account for only 10.3%, 12.6%, and 8.7% of all direct Su(Hw) peaks (3166 peaks), respectively. On the other hand, 85% of direct Su(Hw)-NELF E, 80% of direct Su(Hw)-Paf1, and 94% of direct Su(Hw)-Rpb3 peaks also intersect with Combgap peaks (Supplementary Fig. 2b). These observations suggest that the binding of RNA polymerase II and pausing factors to direct Su(Hw) peaks is related to Combgap rather than being intrinsic to Su(Hw).

The majority of Su(Hw)LOF down-regulated promoters are located within 2 kb of Combgap peaks

It is well-known that disruption of Su(Hw) binding to chromatin leads to abnormal transcription in Drosophila ovaries [25, 26]. However, the underlying mechanism that mediates the impact of Su(Hw) on gene transcription remains unknown. To investigate whether the interactions between Su(Hw) and Combgap are associated with transcriptional regulation, we first identified all differentially expressed (DE) genes in Su(Hw)LOF ovaries compared with wild-type ovaries. We purified mRNA from the corresponding ovaries and performed RNA sequencing (RNA-Seq) on the cDNA libraries obtained (RNA-Seq was performed in two biological replicates for each genotype). We identified 636 transcripts that were significantly DE in Su(Hw)LOF mutants compared to WT (with adjusted p-value < 0.01 and fold-change > 2) (Supplementary Fig. 3a-b, Supplementary Table 1, Fig. 3a).

Fig. 3

The Su(Hw)LOF mis-regulated TSSs are often located within 2 kb of Combgap peaks and possess distinct chromatin properties. (a) Heatmap of the genes, differentially expressed in Su(Hw)LOF ovaries (logFC > 1, adjusted P-value < 0.01). Up- and down-regulated genes are clustered in 3 categories: (1) with direct Su(Hw) peak within 2 kb from TSSs (these peaks contain directly bound Su(Hw) with and without Combgap) (2), with Combgap peaks within 2 kb from TSSs (excluding those which already appeared in the first group - these peaks contain Combgap with and without indirectly bound Su(Hw), but do not contain directly bound Su(Hw)), and (3) without direct Su(Hw)/Combgap peaks near TSSs. The log2 of normalised read counts are shown. (b) and (c) The box plots showing main statistical features of the expression and of the chromatin state (according to FAIRE signal on TSSs) for Su(Hw)LOF up- and down-regulated genes, correspondingly. Within each box plot, the thick line in the centre indicates median expression; boxes and whiskers around it represent 25–75 percentile interval and minimum/maximum expression values, respectively. The dots on the plots represent the features of individual genes. ****, *** and ** are for adjusted P-values of < 0.0001, < 0.001 and < 0.01, correspondingly, according to T-test

Previous studies have reported that 35% of genes mis-regulated in su(Hw)-/- mutant ovaries have Su(Hw) peaks located inside or within 2 kb upstream or downstream of these genes [25]. Using the same criteria, we found that 34.4% (218) of DE transcripts were located within 2 kb of direct Su(Hw) ChIP-Seq peaks. We also compared the proportion of DE TSSs that intersect with direct Su(Hw) and Combgap ChIP-Seq peaks by separately analysing Su(Hw)LOF up- and down-regulated TSSs (Fig. 3a). We observed that 17.1% (56 out of 328) of Su(Hw)LOF up-regulated TSSs were located within 2 kb of direct Su(Hw) peaks, whereas only 3.9% (12 out of 308) of Su(Hw)LOF down-regulated TSSs had direct Su(Hw) peaks within a distance of 2 kb (these direct Su(Hw) peaks contained directly bound Su(Hw) with and without Combgap). Surprisingly, 76.3% (235 out of 308) of Su(Hw)LOF down-regulated TSSs had Combgap ChIP-Seq peaks (these peaks contained Combgap with and without indirectly bound Su(Hw), but did not contain directly bound Su(Hw)) within 2 kb, while only 31.1% (102 out of 328) of Su(Hw)LOF up-regulated TSSs and 41.2% of all D. melanogaster TSSs localised within 2 kb of Combgap ChIP-Seq peaks. This suggests that Combgap may be involved in the transcription regulation of Su(Hw)LOF down-regulated TSSs. Interestingly, the median expression levels of Su(Hw)LOF down-regulated genes with TSSs within 2 kb of direct Su(Hw) and Combgap ChIP-Seq peaks were similar, while those of Su(Hw)LOF up-regulated genes varied significantly (Fig. 3b-c). Su(Hw)LOF up-regulated genes containing direct Su(Hw) peaks within 2 kb showed notably lower median expression compared with Su(Hw)LOF up-regulated genes whose TSSs colocalise with Combgap.

To further clarify the chromatin features of genes that are mis-regulated in Su(Hw)LOF ovaries, we analysed MNase-Seq data for follicular cells from egg chamber stages 1–8 [27] and FAIRE-Seq data for wild-type and Su(Hw)LOF ovaries [19]. Among Su(Hw)LOF down-regulated TSSs, TSSs with Combgap within 2 kb according to FAIRE-Seq typically had the most open chromatin state, which decreased in Su(Hw)LOF compared to WT and exhibited visible nucleosome phasing (Fig. 3c, Supplementary Fig. 3d). Su(Hw)LOF up-regulated TSSs with Combgap within 2 kb also exhibited the highest chromatin accessibility among other downregulated TSSs (Fig. 3b). In contrast, Su(Hw)LOF up-regulated genes with direct Su(Hw) peaks within 2 kb of their TSSs, exhibited features of genes that are repressed or inactive in wild-type ovaries, in general (Fig. 3b, Supplementary Fig. 3d). This suggests that this group of genes reflects the earlier described property of Su(Hw) to repress transcription [25].

We further analysed the DNA motifs enriched in Combgap ChIP peaks, flanking DE TSS (Fig. 3a), to test whether they contain the Combgap DNA-binding motif. Interestingly, the Combgap ChIP-Seq peaks located within 2 kb of Su(Hw)LOF down-regulated TSSs were enriched with BEAF32 and M1BP DNA motifs, in addition to Combgap’s own DNA-binding motifs (Supplementary Fig. 3e). Previous studies have shown that both M1BP and BEAF32 flank promoters [28, 29] and active chromatin domains [8], hence the enrichment of their motifs aligns with our observation that Su(Hw)LOF down-regulated TSSs are characterised by more open chromatin then Su(Hw)LOF up-regulated TSSs (Fig. 3B-C). For Combgap ChIP-Seq peaks located within 2 kb of Su(Hw)LOF up-regulated TSSs we found enrichment of Combgap and GAF DNA motifs. Taken together, these observations suggest that, in addition to Combgap and Su(Hw), other DNA binding proteins may be involved in the regulation of Su(Hw)LOF DE genes.

Su(Hw) ChIP-Seq peaks form long-range chromatin interactions with Combgap ChIP-Seq peaks

A previous study has shown that the absence of the BEAF32 DNA-binding motif in BEAF32 IBP ChIP-Seq peaks indicates the presence of LRIs between chromatin-bound BEAF32 and its partner proteins [30]. To determine whether similar LRIs exist between Su(Hw) and Combgap ChIP-Seq peaks, we conducted in situ Hi-C on ovaries dissected from wild-type and Su(Hw)LOF flies using the four-cutter restriction enzyme MboI. Two biological replicates were conducted for each genotype. The Hi-C data obtained totalled 266–276 million raw pair-end reads per genotype with 66–70 million of valid Hi-C pairs per genotype (the statistics for the read alignments, mapped reads and valid Hi-C pairs can be found in Supplementary Fig. 4 and Supplementary Table 2).

We used the coolpup.py program to estimate the average LRIs between different pairs of ChIP-Seq peak [31]. As a positive control, we examined the average spatial interactions between Rpb3 (a subunit of RNAPII) peaks and found a strong enrichment of Rpb3-Rpb3 LRIs (Supplementary Fig. 5a). We also observed significant enrichment of Combgap-Combgap LRIs (Fig. 4a). This is consistent with previous finding that Combgap can recruit the Ph subunit of the PRC1 complex, which is often found at the anchors of chromatin loops in Drosophila [15, 32]. Importantly, both Rpb3-Rpb3 and Combgap-Combgap LRIs remained intact in Su(Hw)LOF flies, indicating that these LRIs are independent of Su(Hw). Next, we estimated the average spatial interactions within the set of direct Su(Hw) ChIP-Seq peaks. Although we found clear enrichment of LRIs between direct Su(Hw) ChIP-Seq peaks, we did not detect any differences in these LRIs in the Su(Hw)LOF background (Supplementary Fig. 5a).

Fig. 4

Long-range interactions between Su(Hw) and Combgap ChIP-Seq peaks. (a) Averaged spatial interactions between the different sets of Su(Hw) and Combgap ChIP-Seq peaks in the wild-type (WT) and Su(Hw)LOF ovaries, estimated with a coolpup.py program [31]. The minimal and maximal distances of interactions were set at 200 kb and 1000 kb, correspondingly, the pad size is ± 40 kb around the central pixel. (b) Profile plots of insulation score at direct Su(Hw) FAIRE+, Su(Hw) FAIRE- and Combgap ChIP-Seq peaks in the wild-type (WT) and Su(Hw)LOF ovaries. The pile-up profiles were generated as a median of insulation score signal. The standard error is displayed on the profiles as semi-transparent area around the main line of the profiles. (c) Hi-C matrices from the wild-type (WT) and Su(Hw)LOF ovaries, showing one genomic region with insulator scores and occupancies of Su(Hw), Combgap (ChIP-Seq) and open chromatin regions (according to FAIRE-Seq) in the wild-type (WT) and Su(Hw)LOF ovaries. The image was generated using pyGenomeTracks [60]. This particular region was selected to illustrate Su(Hw)-dependence of long range interactions (LRIs) between direct Su(Hw) FAIRE + and Combgap ChIP-Seq peaks (position of direct Su(Hw) FAIRE + peak in these LRIs is marked with the vertical line)

In a previous study, we demonstrated that direct Su(Hw) ChIP-Seq peaks in Drosophila ovaries can be classified into two clusters based on their enrichment with the FAIRE signal and active chromatin features, such as chromatin remodelers and Gcn5 histone acetyltransferase (henceforth referred to as Su(Hw) direct FAIRE + peaks), or lack of such enrichment (Su(Hw) direct FAIRE- peaks) [19]. We analysed the average LRIs for these two clusters separately. We observed the LRIs between Su(Hw) direct FAIRE + peaks disappeared in the absence of Su(Hw) (Fig. 4a). Interestingly, for Su(Hw) direct FAIRE- peaks, the LRIs appeared to be independent of Su(Hw) binding to chromatin. When we analysed the average insulation profiles for these groups of peaks, we observed that Su(Hw) direct FAIRE + peaks showed strong insulation scores (Fig. 4b). In Su(Hw)LOF, the insulation score for Su(Hw) direct FAIRE + peaks decreased, indicating that Su(Hw) can define local insulation at the domain borders where it binds. On the other hand, Su(Hw) direct FAIRE- peaks did not show any enrichment in insulation score, suggesting that they do not colocalise with the domain borders.

To investigate the spatial interactions between Combgap and Su(Hw), we excluded all peaks from the Combgap peak set that intersected with direct Su(Hw) ChIP-Seq peaks. Results showed the presence of LRIs between Combgap and Su(Hw) direct FAIRE + peaks, but not with Su(Hw) direct FAIRE- peaks (Fig. 4a). In Su(Hw)LOF flies, these LRIs disappeared, indicating the importance of Su(Hw) for their formation. These Su(Hw)-dependent LRIs between Su(Hw) direct FAIRE + peaks and Combgap are also evident in Hi-C maps (Fig. 4c, Supplementary Fig. 5c-g).

We also investigated the interactions between Su(Hw) direct FAIRE + and FAIRE- peaks and the ChIP peaks of M1BP and BEAF32 proteins, for which we detected motifs in the Combgap ChIP peaks within 2 kb of Su(Hw)LOF down-regulated TSSs (Supplementary Fig. 3e). We observed that similarly to the interaction with Combgap, Su(Hw) direct FAIRE + peaks show Su(Hw)-dependent LRIs with M1BP and BEAF32 (Supplementary Fig. 5b). This observation demonstrates that other architectural proteins, such as M1BP and BEAF32, may also be involved in the establishment and maintenance of Su(Hw)-dependent LRIs.

To validate the findings of our Hi-C experiments, we examined the existing Hi-C datasets [33] for the OSC cell line, which has been derived from somatic cells of the Drosophila ovary. Results showed long-range interactions in the Hi-C data from OSC cells, similar to our observations in wild-type ovaries (Supplementary Fig. 6).

Changes in chromatin architecture in Su(Hw)LOF background correlate with transcription mis-regulation

To investigate whether changes in chromatin architecture correlate with transcriptional mis-regulation in Su(Hw)LOF ovaries, we analysed the average spatial interaction between Su(Hw) direct FAIRE + peaks and Combgap localised within 2 kb of Su(Hw)LOF DE genes (Supplementary Fig. 7). For distances ranging from 50 to 200 kb, we did not observe any significant interactions between these regions, likely due to the shorter average ranges of transcriptional regulatory interactions (such as enhancer-promoter interactions). Indeed, it was shown that the majority of characterised enhancers are within 10 kb of their target gene, with only a few capable of acting at distances beyond 50 kb [34]. Unfortunately, the analysis of interactions at distances below 50 kb is not possible with the obtained Hi-C matrices due to the high basal level of contacts at such short distances.

Analysing Hi-C maps for wild-type and Su(Hw)LOF ovaries, we observed that the disruption of Su(Hw)-Combgap LRIs correlated with the transcriptional mis-regulation of nearby genes in a Su(Hw)LOF background (Supplementary Fig. 8). It is already known that changes in LRIs can affect enhancer-promoter communication within interacting domains [6, 7]. In the OSC cell line, derived from somatic ovary cells, all DNA elements capable of acting as enhancers have been annotated [35]. Following the approach from Cavalheiro et al. [36] to analyse enhancer hijacking, we observed that Su(Hw)-Combgap boundaries, which are disrupted in Su(Hw)LOF, often localise near “hijacked” OSC enhancers (Fig. 5). This suggests that changes in Su(Hw)-Combgap LRIs may impact the enhancer-promoter interaction networks of DE genes in the somatic cells of the ovary.

Fig. 5

LRIs between Su(Hw) and Combgap ChIP-Seq peaks correlate with transcription mis-regulation in Su(Hw)LOF background. (a-b) Two loci displaying enhancer hijacking. Top of the figure: Hi-C and differential Hi-C matrices (log2FC) of the bun (a) and CG3104 (b) loci. Bottom of the figure: zoom-in of Su(Hw), Combgap ChIP-Seq and FAIRE-Seq in WT (green) and Su(Hw)LOF (orange) ovaries and RNA-Seq separately for 2 replicates from WT (green) and Su(Hw)LOF (orange) ovaries. Known OSC STARR enhancers are shown; Su(Hw)-dependent boundaries weakened in Su(Hw)LOF are marked with grey dashed lines; potential newly formed enhancer-promotor interactions are shown in dotted arrows. Genes highlighted in red are up-regulated in Su(Hw)LOF (bun, CG3104), while the possible enhancer’s original targets (eRF3, CG31694) are in green. RNA-Seq normalized levels are displayed in bar plots; n.s. is for adjusted P-value > 0.05 and * is for adjusted P-value < 0.05

To identify changes in the binding of transcriptional regulators to mis-regulated genes in Su(Hw)LOF, we analysed the ChIP-Seq distribution of Combgap, chromatin remodelers [19], the Rpb3 subunit of RNA polymerase II, the NELF E subunit of NELF complex, and the Paf1 positive elongation factor binding along mis-regulated genes (Supplementary Fig. 9). These transcriptional regulators have been shown to bind Su(Hw) ChIP peaks in Su(Hw)-dependent manner [19]. We found no significant changes in the distribution of these proteins along Su(Hw)LOF up-regulated genes (Supplementary Fig. 9b). However, for Su(Hw)LOF down-regulated genes that have Combgap located within 2 kb of their TSSs, we observed an increase in the binding of Combgap, Rpb3, and ISWI ATPase of the ISWI chromatin remodeler family, along with a decrease in the binding of Brm ATPase of the SWI/SNF family and CHD1 ATPase of the CHD family (Supplementary Fig. 9a). For Su(Hw)LOF down-regulated genes that have direct Su(Hw) peaks within 2 kb of their TSSs, we observed a substantial decrease in the binding of Brm ATPase and Mi2 ATPase of the CHD chromatin remodeler family, along with a reduction in Paf1 binding (Supplementary Fig. 9a). Previous studies have shown that ISWI and Mi-2, in addition to their role in transcription activation [37, 38], can contribute to the establishment of an inactive chromatin state [39,40,