Mi-2 binding on chromatin correlates with EcR binding sites

Since an interaction with the nascent RNA transcript appeared not to play an important role in dMi-2 recruitment I investigated the potential involvement of EcR in Mi-2 recruitment to chromatin. NRs have been demonstrated to function by recruiting co-regulatory proteins to chromatin. Since Mi-2 and EcR interact physically, I hypothesised that EcR can recruit Mi-2 to specific binding sites in the genome. EcR binding sites were mapped genome-wide in Drosophila in two studies to date. A publication by Gauhar and colleagues identified EcR/USP binding upon six hours of 20HE treatment in Kc167 cells using DamID (Gauhar et al., 2009). This study identified 502 binding sites for the EcR/USP complex. Due to technical reasons, the binding regions identified by DamID are rather broad with an average size of 4 to 5kb. A recent study by the Stark lab mapped EcR binding sites by ChIPSeq in S2 cells (Shlyueva et al., 2014). This technique identified 9148 EcR binding sites in untreated and 9305 EcR binding sites in cells that were treated for 24 hours with 20HE.

In order to test if Mi-2 and EcR reside in close proximity on chromatin, the overlap between Mi-2 and EcR binding sites was analysed on a genome-wide level. Therefore, Mi-2 binding sites from ChIPSeq were compared to ChIPSeq data for EcR published by the Stark lab. To determine the correlation between Mi-2 and EcR binding sites, the strongest 10% Mi-2 peaks (in terms of tag counts) were selected and used to calculate the overlap with all EcR binding sites (Figure 4.22A). I selected the strongest 10% Mi-2 peaks since these obtained the most reads from sequencing and were therefore the

110 most confident. Mi-2 and EcR peaks had to overlap by at least one base in order to be defined as “overlapping”. In case two or more Mi-2 binding sites overlapped with the same EcR peak, this co-occupancy was only counted once. In untreated S2 cells, 215 out of 1643 Mi-2 binding sites overlapped with EcR binding sites. This overlap was comparable in 20HE treated cells, where 198 out of 1499 Mi-2 peaks coincided with EcR binding sites. In both conditions, approximately 13% of Mi-2 binding sites overlapped with EcR peaks. This overlap was significantly higher than expected if the binding sites were randomly distributed across the genome (Monte-Carlo-method, personal communication R. Pahl and F. Finkernagel). These results were further supported by a comparison between 20HE induced Mi-2 binding sites (103 peaks, Table 7.1) and all EcR binding sites in the presence of 20HE from the Stark data. Here, 35 of 103 induced Mi-2 peaks overlapped with EcR binding sites in the presence of hormone (Figure 4.22B). Even though this comparison considered only a small subset of Mi-2 binding sites that are more than 2.3fold induced upon 20HE treatment, 33.7%

of these binding sites show an overlap with EcR binding sites. In summary, genome-wide analysis identified a significant overlap between Mi-2 and EcR binding sites in both untreated and 20HE treated S2 cells.

Figure 4.22: Genome-wide overlap of 2 and EcR binding sites. (A) Overlap between Mi-2 and EcR binding sites in untreated and +Mi-20HE treated SMi-2 cells is depicted in a Venn diagram.

The 10% Mi-2 bindings sites with the highest tag count in ChIPSeq were considered. (B) Overlap between 20HE induced Mi-2 binding sites (Appendix, Table 7.1) and EcR binding sites in 20HE treated S2 cells is depicted in a Venn diagram. (A) and (B) EcR binding sites were taken from (Shlyueva et al., 2014). Overlap is defined as a co-occupancy of Mi-2 and EcR binding sites of at least 1bp at the same genomic region.

111 For detailed investigation of the co-occupancy of Mi-2 and EcR at the broad and vrille loci, the EcR binding profile, identified by the two studies described above, was analysed (Figures 4.23 and 4.24). Gauhar and colleagues identified several EcR/USP binding sites upon 20HE treatment at the broad locus (Figure 4.23, grey boxes). The Mi-2 peak in the 5’ UTR (Figure 4.5, region A) as well as the region within the first intron (Figure 4.5, region B and C) showed an overlap with 20HE induced EcR/USP binding mapped by DamID. However, no overlap between 20HE induced EcR/USP binding at region D within the exon of the broad (Figure 4.5) gene was identified. With respect to vrille the study found two broader regions of 20HE induced EcR/USP binding both of which overlap with either one of the regions where Mi-2 is recruited upon 20HE induction (Figure 4.24, grey boxes).

Figure 4.23: Comparison of Mi-2 and EcR binding sites at the broad locus. Mi-2 binding profile at the broad locus as demonstrated in Figure 4.5. Regions A-D (red boxes) depict sites of increased Mi-2 binding upon 20HE treatment as calculated from ChIPSeq (Zoom-in Figure 4.5). EcR/USP binding sites identified by DamID upon six hours of 20HE treatment taken from (Gauhar et al., 2009) are depicted as grey boxes. EcR ChIPSeq tracks from (Shlyueva et al., 2014) in untreated and 20HE treated cells are illustrated.

The identification of EcR binding sites by ChIPSeq from the Stark lab allowed to directly compare binding profiles of Mi-2 and EcR with a better resolution than the EcR/USP binding sites identified by DamID. Interestingly, at the broad locus, EcR displayed a similar binding pattern as Mi-2 in untreated cells (Figure 4.23). Both proteins showed enriched binding at the 5’ UTR and a broad binding region within the first intron (see Figure 4.5). EcR binding in 20HE treated cells was not altered significantly in the study from the Stark lab. No clear enrichment of EcR upon hormonal stimulation was detected across the broad gene and at the 20HE induced Mi-2 binding

112 sites A to D. This difference may be due to the fact that the Stark lab identified EcR binding sites by ChIPSeq after 24 hours of 20HE treatment. However, broad is an early induced gene that is transcribed shortly after 20HE treatment (Figure 4.9C). Therefore, I hypothesised that 20HE induced EcR binding at the broad locus can be detected shortly after hormonal treatment, but can not be seen after 24 hours, the time point analysed by the Stark lab. This may explain why no overlap between 20HE induced Mi-2 and EcR binding at the broad gene could be detected. Also, EcR and Mi-2 ChIPSeq tracks were compared for the vrille gene (Figure 4.24). In contrast to the observations at the broad gene, the binding of Mi-2 and EcR identified by ChIPSeq in untreated cells did not display a comparable pattern. Interestingly, when comparing the EcR and Mi-2 binding profile in 20HE treated cells, EcR binding as determined by the Stark study was clearly enriched at the second 20HE induced Mi-2 binding site (region Y) at the vrille locus. No such overlap was observed for the first Mi-2 peak (region X) within the vrille gene. In conclusion, a comparison of the Stark ChIPSeq data set has revealed that Mi-2 and EcR coincide at several binding sites at the broad and vrille genes in the absence of hormone. Additionally, 20HE induced Mi-2 binding identified in this thesis and EcR binding upon hormonal treatment determined by the Stark lab were shown to coincide at one binding site within the vrille gene.

Figure 4.24: Comparison of Mi-2 and EcR binding sites at the vrille locus. Mi-2 binding profile at the vrille locus as demonstrated in Figure 4.5. Regions X and Y (red boxes) depict sites of increased Mi-2 binding upon 20HE treatment as calculated from ChIPSeq. EcR/USP binding sites identified by DamID upon six hours of 20HE treatment taken from (Gauhar et al., 2009) are depicted as grey boxes. EcR ChIPSeq tracks from (Shlyueva et al., 2014) in untreated and 20HE treated cells are illustrated.

In order to verify binding of EcR/USP within the broad and vrille genes at the binding sites identified above, EcR ChIP followed by qPCR was performed. However, anti-EcR

113 antibody did not precipitate EcR efficiently in ChIP (data not shown). To circumvent this problem, binding sites were analysed by ChIP with an anti-USP antibody since EcR is recruited to ecdysone responsive genes in complex with its heterodimerisation partner USP (Figure 4.25) (Thomas et al., 1993).

Figure 4.25: Identification of USP binding sites at vrille and broad by ChIP qPCR. (A) ChIP was performed using an antibody ag41ainst USP. Chromatin was prepared from untreated (white bars) and 20HE treated (grey bars) S2 cells. Values are expressed as % input. Purified DNA was quantified by qPCR with oligos at binding sites indicated in Figure 4.5 (broad) and Figure 4.6 (vrille). “Intergenic” refers to an unrelated intergenic region on chromosome arm 2R (see Material and Methods). Error bars denote standard deviation of technical triplicates.

Experiment was performed as biological duplicate and one representative experiment is shown here.

Unfortunately, precipitation of USP was also inefficient as only 0.01% input was precipitated at most of the sites tested in untreated and 20HE treated cells. However, a few binding sites showed higher enrichment of USP upon 20HE treatment. For the broad gene, binding site 1 and 4 showed two- to threefold increased USP signal upon hormonal stimulation. Both of these regions overlapped with sites of Mi-2 recruitment upon ecdysone induction (Figure 4.23). However, binding site 5 that overlapped with a

114 predicted EcR/USP region, did not show enrichment of USP in ChIP (Gauhar et al., 2009). At the vrille gene two binding sites (2 and 6) showed twofold increased recruitment of USP upon 20HE stimulation. In fact, these two sites also showed enrichment of Mi-2 in ecdysone treated cells (Figure 4.6 and Figure 4.7, regions X and Y), and overlapped with the predicted EcR/USP binding sites in the study by Gauhar and colleagues (Figure 4.24). In summary, USP binding sites at the broad and vrille locus were identified by ChIP. These are likely to be also bound by EcR since EcR and USP heterodimerise in the presence of ecdysone. This may further confirm the correlation between 2 and EcR/USP binding on chromatin. The co-occupancy of Mi-2 and EcR at ecdysone dependent genes together with the ability of both proteins to interact physically (Figure 4.12 and 4.15) may indicate a functional interaction of both proteins in the regulation of hormone regulated genes.

4.4.3 Depletion of EcR decreases recruitment of Mi-2 to broad and vrille in 20HE