CoREST acts as a positive regulator of Notch signaling in the follicle cells of Drosophila melanogaster

The Notch signaling pathway plays important roles in a variety of developmental events. The context-dependent activities of positive and negative modulators dramatically increase the diversity of cellular responses to Notch signaling. In a screen for mutations affecting the Drosophila melanogaster follicular epithelium, we isolated a mutation in CoREST that disrupts the Notch-dependent mitotic-to-endocycle switch of follicle cells at stage 6 of oogenesis. We show that Drosophila CoREST positively regulates Notch signaling, acting downstream of the proteolytic cleavage of Notch but upstream of Hindsight activity; the Hindsight gene is a Notch target that coordinates responses in the follicle cells. We show that CoREST genetically interacts with components of the Notch repressor complex, Hairless, C-terminal Binding Protein and Groucho. In addition, we demonstrate that levels of H3K27me3 and H4K16 acetylation are dramatically increased in CoREST mutant follicle cells. Our data indicate that CoREST acts as a positive modulator of the Notch pathway in the follicular epithelium as well as in wing tissue, and suggests a previously unidentified role for CoREST in the regulation of Notch signaling. Given its high degree of conservation among species, CoREST probably also functions as a regulator of Notch-dependent cellular events in other organisms.


Introduction
The highly conserved Notch signaling pathway plays a crucial role in a broad array of developmental events, including the maintenance of stem cells, cell fate specification, control of proliferation and apoptosis (Borggrefe and Oswald, 2009;Artavanis-Tsakonas and Muskavitch, 2010). Misregulation of the Notch pathway is associated with a number of diseases, including different types of cancer (Koch and Radtke, 2010). The binding of the transmembrane ligands DSL (Delta, Serrate, LAG-2) to the extracellular domain of Notch, exposed on a neighboring cell, activates the signaling cascade by triggering a sequence of proteolytic cleavages of Notch protein. Extracellular cleavage (S2) leads to the formation of an intermediate membrane-bound Cterminal fragment of Notch, called NEXT. This event is followed by an intramembranous cleavage (S3) by the c-secretase complex (Struhl and Greenwald, 1999;Struhl and Greenwald, 2001). The intracellular domain of Notch (NICD) then translocates to the nucleus and binds to a transcription factor of the CSL family [CBF-1, Su(H), LAG-1], converting it from a transcriptional repressor to an activator. In the canonical Notch pathway, Su(H) directly activates Notch target genes in response to signaling (Bailey and Posakony, 1995;Furukawa et al., 1995;Lecourtois and Schweisguth, 1995). Despite the relative simplicity of the Notch transduction pathway, the presence of a large number of proteins that positively or negatively influence Notch signaling dramatically increases the complexity of the Notch pathway and its cellular responses (Panin and Irvine, 1998;Schweisguth, 2004;Borggrefe and Oswald, 2009). For instance, extracellular modulators, such as Fringe (Panin et al., 1997), alter ligandspecific Notch activation, whereas cytoplasmic modulators, such as Numb (Guo et al., 1996), restrict signal transduction. Nuclear modulators, for instance Mastermind (Wu et al., 2000), influence the transcriptional activity of the NICD-containing complex. In addition, there is increasing evidence of the importance of the epigenetic regulation of Notch targets, which can cause differential cellular responses upon Notch activation (Borggrefe and Oswald, 2009).
Drosophila melanogaster serves as an excellent model system to dissect the regulation of the Notch pathway. The Drosophila genome contains only a single Notch protein and two ligands [Delta (Dl) and Serrate (Ser)]. The Notch pathway is involved in several aspects of Drosophila development. The role of Notch in lateral inhibition during neurogenesis has been extensively studied; it restricts neural cell fates in the embryo, and leads to restriction of sensory-organ formation and induction of boundary formation in the wing discs (Micchelli and Blair, 1999;Portin, 2002;Cau and Blader, 2009). Notch activity is also required for many aspects of oogenesis, such as the establishment of egg chamber polarity, polar cell formation, control of follicle cell (FC) proliferation, differentiation, cell fate specification and morphogenesis (Deng et al., 2001;Grammont and Irvine, 2001;Lopez-Schier and St Johnston, 2001;Horne-Badovinac and Bilder, 2005;Grammont, 2007;Shyu et al., 2009;Klusza and Deng, 2011;Vachias et al., 2010). The Drosophila FCs are somatically derived epithelial cells that form a monolayer covering the germline cells during oogenesis. FCs divide mitotically from stage 2 to stage 6 of oogenesis, followed by the switch from the mitotic cycle to the endocycle (the M/E transition). Endocycles take place from stage 7 to stage 10A of oogenesis and include three rounds of DNA duplication without subsequent cell division (Klusza and Deng, 2011). The M/E switch is triggered upon Notch pathway activation. Dl produced in the germline binds to its receptor Notch, expressed in the FCs, and induces activation of the canonical Notch signaling pathway (Deng et al., 2001;Lopez-Schier and St Johnston, 2001;Lopez-Schier and St Johnston, 2002). Removal of Dl from germline cells, or of Notch from FCs, maintains follicle cells in the mitotic cycle throughout oogenesis. NICD complexed with Su(H) activates transcription of downstream target genes required for the M/E switch, such as Hindsight (Hnt) (Deng et al., 2001;Lopez-Schier and St Johnston, 2001). Hnt then mediates the Notch-dependent downregulation of Cut, String (Stg) and Hedgehog (Hh) signaling in the FCs, thus promoting the M/E switch (Sun and Deng, 2007).
We have identified the transcriptional cofactor Corepressor for element-1-silencing transcription factor (CoREST) as a positive modulator of Notch signaling in the FCs and during wing development. We show that CoREST is required for the promotion of the M/E switch during oogenesis. CoREST acts downstream of NICD release but upstream of Hnt activity, and it is a previously unidentified modulator of the Notch pathway. The genetic interactions between CoREST and Hairless (H), CtBP and Groucho (Gro), members of the Notch repressor complex, suggest that CoREST might influence the activity of either Notch transcriptional repressor or activator complexes. In addition, we found that CoREST specifically affects tri-methylation of lysine 27 of histone 3 (H3K27) and acetylation of H4K16 in FCs, because these chromatin modifications show elevated levels in the CoREST mutant cells. These findings point to a possible role of CoREST in regulation of the activity of the Notch repressoractivator complexes and/or epigenetic regulation of the components of the repressor-activator complexes or of factors involved in the transduction of the signaling or directly of target genes of the Notch signaling pathway.

Results
GF60 mutant follicle cells fail to switch from the mitotic cycle to the endocycle The FCs of Drosophila divide mitotically from stage 2 to stage 6 of oogenesis, and then switch from the mitotic cycle to the endocycle, in the so called M/E transition (Klusza and Deng, 2011). In a genetic mosaic screen designed to identify genes involved in FC patterning, differentiation and morphogenesis (Denef et al., 2008;Yan et al., 2009), the GF60 allele was isolated. GF60 homozygous mutant posterior follicle cells (PFCs) are smaller and have much smaller nuclei than their neighboring wild-type cells, which is an indicator of a failure in the M/E switch ( Fig. 1A; supplementary material Fig. S1). The apical-basal polarity of the mutant cells, as tested by immunostaining for aPKC, dlg, DE-Cad and Arm, was, however, not affected (data not shown). To confirm that GF60 mutant FCs are blocked in the mitotic cycle, we stained the mosaic egg chambers for the mitotic markers phosphorylated histone H3 (PH3) and cyclin B (CycB). Although in wild-type egg chambers PH3 and CycB are only detected up to stage 6, in GF60 homozygous mutant clones both of these mitotic markers were still present after stage 6 ( Fig. 1B,C). Interestingly, this proliferation phenotype was more pronounced in the PFCs, and less frequently observed in the lateral or anterior follicle epithelial clones. Thus, in CoREST GF60 homozygous mutant FCs, PH3 expression after stage 7 was detected in 74% of PFC clones (n565), 24% of lateral (n546) and 18% of anterior clones (n555). We therefore concluded that the gene product disrupted in GF60 mutant cells is involved in the control of the M/E transition in FCs, and that its effect is largely restricted to PFCs.
The GF60 mutant allele encodes a truncated CoREST protein that lacks any functional domains We mapped the lethal GF60 mutation to the gene encoding CoREST, using a combination of meiotic recombination with visible recessive markers, P-element insertions and single nucleotide polymorphism (SNP) mapping approaches. The Drosophila genome contains only one gene encoding CoREST ( Fig. 2A), whereas three CoREST homologs are found in mammalian genomes (Dallman et al., 2004). The CoREST GF60 allele carries a guanine to adenine point mutation in the first nucleotide of the third intron ( Fig. 2A), leading to a non-functional splice donor site. The CoREST GF60 mutant retains the third intron, as we confirmed by RT-PCR analysis (data not shown). The presence of an in-frame stop codon at the end of the third intron results in a truncated CoREST protein lacking known functional  (Dallman et al., 2004).
Using an antibody that recognizes a sequence after the second SANT domain of CoREST present in two of the three splice variants (Dallman et al., 2004), we identified bands corresponding to the 97 kDa and 130 kDa isoforms on western blots from lysates of wild-type egg chambers (data not shown). This finding indicates that at least two splice variants of CoREST are produced in the ovaries. Furthermore, we found that CoREST is ubiquitously expressed in nuclei of wild-type FCs as well as in germline cells at all stages of oogenesis (supplementary material Fig. S2), whereas no CoREST expression was detected in CoREST GF60 mutant FCs (Fig. 2C). To confirm that the discovered mutation in the CoREST gene is indeed responsible for the phenotype observed in the GF60 mutant, we expressed an HA-tagged CoREST cDNA in GF60 mutant FCs, and rescued the mutant phenotype, restoring the PH3 staining after stage 6 of oogenesis and reverting other mutant phenotypes associated with the GF60 mutation, which are described below (supplementary material Fig. S3). Notably, misexpression of the CoREST cDNA did not lead to a premature M/E switch.

Notch signaling is disrupted in CoREST mutant PFCs
The M/E switch in Drosophila FCs is under the control of the Notch pathway (Deng et al., 2001;Lopez-Schier and St Johnston, 2001;Sun and Deng, 2005;Sun and Deng, 2007). Germline cells express Dl (Lopez-Schier and St Johnston, 2001), which activates Notch signaling in FCs at stages 6-7 of oogenesis, resulting in the expression of the zinc-finger transcription factor Hnt, which in turn promotes the M/E switch through suppression of cut expression (Sun and Deng, 2007). To determine whether the failure of the M/E switch is a consequence of a disrupted Notch signaling in CoREST GF60 mutant cells, we stained mosaic egg chambers with antibodies specific for Hnt and Cut. Hnt is upregulated upon Notch pathway activation at stages 6-7 in the entire follicle epithelium (Fig. 3A) (Sun and Deng, 2007). In CoREST GF60 mutant PFCs, Hnt expression was not detected during mid-oogenesis, whereas in CoREST GF60 mutant lateral and anterior clones Hnt expression appeared comparable to that in the neighboring wild-type cells (Fig. 3B). Cut is expressed in FCs until stages 6-7 of oogenesis, but is afterwards repressed in all FCs, except the polar and stalk cells, through the activity of Hnt (Fig. 3D). The downregulation of Cut in FCs is necessary to induce the M/E switch (Sun and Deng, 2005;Sun and Deng, 2007). In CoREST GF60 homozygous mutant PFCs, Cut was not downregulated during mid-oogenesis, whereas Cut repression appeared mostly normal in anterior and lateral mutant clones (Fig. 3E). To determine whether CoREST affects expression of other Notch targets and not only the expression of Hnt, we used an E(Spl)-mb-CD2 transgenic line (de Celis et al., 1998) to assay Notch activity in CoREST GF60 mutant clones. Upon Notch activation, CD2 staining was detected in all follicle cells from White rectangles represent the 59 and 39 UTR, grey rectangles represent translated regions. The third intron is mutated in the GF60 allele leading to the substitution of A for G (highlighted in red), which abolishes its splicing (data not shown). The stop codon in the third intron is highlighted in yellow. (B) Schematic diagrams illustrating three Drosophila CoREST splice variants [adopted from Dallman et al. (Dallman et al., 2004)]. ELM2 domains are depicted as green rectangles, SANT domains are represented by blue ovals. The short, medium and long isoforms of CoREST are 273, 590 and 824 aa, respectively. The GF60 mutation leads to the truncation of CoREST before the ELM2 domains, and is indicated by a red star. (C) Stage 7 egg chamber containing CoREST GF60 homozygous mutant clones stained with anti-CoREST antibodies (red in C, white in C9) and Hoechst (blue in C) to label DNA. Mutant clones are identifiable by the absence of GFP (green in C, white in C99), and the borders between mutant clones and neighboring wild-type cells are marked by dotted lines (C9,C99). CoREST is present in wild-type follicle epithelial cells (open arrow in C9), and absent from mutant clones (filled arrow in C9). Egg chambers are oriented with the posterior to the right. Scale bar: 50 mm. stage 7 to stage 10A (Sun et al., 2008), but it was not present in CoREST GF60 mutant PFCs (supplementary material Fig. S4A). These data indicate that Notch signaling is disrupted in CoREST GF60 mutant PFCs. In addition to the failure of the M/E switch, prolonged Cut expression maintains follicle cells in an immature state, as indicated by the extended presence of FasIII (Lopez-Schier and St Johnston, 2001;Sun and Deng, 2005) and Eya (Bai and Montell, 2002;Sun and Deng, 2005) after stages 6-7 (supplementary material Fig. S4B,C). Thus, we conclude that CoREST controls the response to Notch signaling, and acts predominantly in the PFCs.

Reduction of Dl leads to higher penetrance of the CoREST phenotype in the follicle cells
The observation that CoREST GF60 mainly disrupts Notch activity in PFCs suggests either that CoREST functions specifically in those cells, or alternatively that Notch signaling in PFCs is particularly sensitive to the loss of CoREST. Therefore, we tested whether CoREST GF60 could produce a mutant phenotype in lateral and anterior cells of the follicle epithelium by creating mutant clones in a Dl 7 heterozygous background, which is more sensitive to Notch pathway disruption (Guo et al., 1999). In the control background, no Hnt expression was detected after stage 7 in 67% of CoREST GF60 homozygous mutant PFCs (n551), whereas absence of Hnt was only observed in 12% of lateral (n543) and 5% of anterior clones (n514; Fig. 3B). Generation of CoREST GF60 homozygous mutant clones in the Dl 7 heterozygous background increased the penetrance of this phenotype in posterior clones to 97%, (n537), in lateral clones to 40% (n535) and in anterior clones to 20%, (n515; Fig. 3C). Similarly, in the control background, in CoREST GF60 homozygous mutant FCs, the presence of Cut expression after stage 7 was detected primarily in the posterior mutant clones (97.5% of PFC clones were Cut positive; n5193), and less frequent in lateral (46%, n5187) and anterior clones (20%, n5110; Fig. 3E). CoREST GF60 homozygous mutant clones in the Dl 7 heterozygous background exhibited an enhanced phenotype with 100% of posterior mutant clones (n560), 79% of lateral (n567) and 80% of anterior clones (n520) with Cut-positive cells (Fig. 3F). Thus, the penetrance of the Notch phenotype was dramatically increased when the dose of Dl was reduced by half, indicating that the regulation of Notch signaling by CoREST is not solely restricted to the PFCs. Rather, PFCs are more sensitive to the loss of CoREST than lateral and anterior FCs.
CoREST regulates Notch signaling downstream of the Notch proteolytic cleavage but upstream of Hnt Notch, upon binding to Dl, undergoes a sequence of proteolytic cleavages. Extracellular cleavage (S2) leads to the formation of the intermediate membrane-bound C-terminal fragment of Notch, called NEXT. This event is followed by the second intramembranous cleavage (S3) resulting in the release of the Notch intracellular domain NICD (Lai, 2004;Schweisguth, 2004). To test whether any of these Notch fragments can rescue the CoREST GF60 mutant phenotype, we overexpressed full-length Notch (NFL), NEXT and NICD constructs in CoREST GF60 homozygous mutant clones using the MARCM system (Lee and Luo, 1999). As a readout of Notch signaling, we analyzed the expression of Cut in the PFCs after stage 7 of oogenesis. Cut expression was detected in 99% of NFL-(n578; Fig. 4A), in 94% of NEXT-(n563; Fig. 4B) and in 90% of NICD-overexpressing (n5137; Fig. 4C) CoREST GF60 mutant PFC clones. These results indicate that neither NFL, NEXT or NICD could rescue the CoREST mutant phenotype, which positions the effect of CoREST on Notch signaling downstream of NICD release.
The absence of Hnt expression in CoREST GF60 mutant PFCs during mid-oogenesis suggests that CoREST is acting upstream or at the level of Hnt transcriptional regulation. The prolonged expression of Cut in the CoREST GF60 clones is expected to result from the lack of its repression by Hnt. To confirm this relationship, we misexpressed Hnt in mutant clones using the MARCM system (Lee and Luo, 1999), and examined whether it was possible to rescue the CoREST GF60 mutant phenotype. The misexpression of Hnt in CoREST GF60 clones (Fig. 4D) resulted in Cut downregulation before stage 6 (Fig. 4E). The follicle cell nuclei in the clones were substantially larger than those of the neighboring wild-type cells even before stage 6, indicating a premature M/E switch (Fig. 4E0). Therefore, we were able to rescue the CoREST GF60 mutant phenotype (Cut expression, M/E transition) by Hnt misexpression. Together with the inability of Notch fragments to rescue the mutant phenotype, these results indicate that CoREST is involved in the regulation of Notch signaling downstream of NICD release and upstream of Hnt.

Reduction of H, CtBP and Gro dosage leads to a mild suppression of the CoREST mutant phenotype
In the absence of Notch activation, expression of Notch target genes is repressed by the transcription factor Su(H) acting in a complex with H. H recruits two corepressors, CtBP and Gro, which are required for H-mediated repression (Nagel et al., 2005). The translocation of NICD to the nucleus displaces these corepressors from Su(H), converting it from a transcriptional repressor to a transcriptional activator. Subsequently, the transcriptional coactivator protein Mastermind (Mam) binds to NICD-Su(H) forming a coactivator complex (Fryer et al., 2002;Kovall, 2007). Because CoREST is required for the Notch pathway downstream of NICD release and upstream of Hnt in the FCs, it could be directly involved in the regulation of hnt expression. It might act through either negative regulation of repressor complex members, such as H, CtBP or Gro, or by positive regulation of activator complex members, such as Mam, thus modulating their ability to transcriptionally regulate hnt expression. To test for genetic interactions with these regulatory proteins, we examined whether we could modulate the penetrance of the CoREST GF60 mutant phenotype by changing the dosage of H, CtBP or Gro. We generated CoREST GF60 homozygous mutant follicle epithelial clones in H 1 (Maier et al., 1992), CtBP P1590 (Poortinga et al., 1998) or gro e47 (Orian et al., 2007) heterozygous backgrounds. In the CoREST GF60 homozygous mutant clones generated in the control background, Cut was detected after stage 7 of oogenesis mainly in the posterior mutant clones (97.5%, n5193; Fig. 3E), and less frequently in lateral (46%; n5187) and anterior clones (20%, n5110; Fig. 3E). The reduction of H dosage in the CoREST GF60 homozygous mutant cells significantly (P,0.001) suppressed the CoREST mutant phenotype: 51% of PFC clones (n573), 6% of the lateral (n564) and 0% of the anterior clones (n552) were Cut positive after stage 7 of oogenesis (Fig. 5A,D). Similarly, the reduction of CtBP dosage in the CoREST GF60 homozygous mutant cells slightly but statistically significantly (P,0.001) suppressed the CoREST mutant phenotype: 73% of PFC clones (n5224), 19% of the lateral (n5206) and 3% of the anterior clones (n5135) were Cut positive after stage 7 of oogenesis (Fig. 5B,D). In comparison with the CoREST GF60 mutant clones in the control background, CoREST GF60 homozygous mutant clones in the gro e47 heterozygous background (Fig. 5C,D) exhibited a similar penetrance of the CoREST mutant phenotype in the PFCs, with 90% of Cut-positive PFC clones (n573). However, the penetrance of the CoREST mutant phenotype was weakly but statistically significantly reduced for the CoREST GF60 homozygous mutant lateral follicle cell clones (27%, n555; P,0.001) and anterior follicle cell clones (5%, n538; P,0.001) in the gro e47 heterozygous background (Fig. 5D). Notably, we observed a stronger suppression of the CoREST mutant phenotype in the H 1 and CtBP P1590 heterozygous backgrounds in comparison with the gro e47 heterozygous background. In summary, we were able to modulate the strength of the CoREST GF60 mutant phenotype by changing the dosage of components of the Notch repressor complex, H, CtBP or Gro, indicating that H-CtBP-Gro and CoREST play antagonistic roles in Notch signaling. This is consistent with the conclusion that CoREST acts as a positive modulator of the Notch pathway in the FCs. The genetic interactions between H and CoREST, CtBP and CoREST, and gro and CoREST further suggest that CoREST might be involved in the regulation of the Notch repressor-activator complex.
CoREST does not act in concert with Chn in Notch signaling regulation in FCs CoREST was first identified in humans as a primary cofactor for RE1 silencing transcription factor (REST; also known as neuralrestrictive silencing factor; NRSF) (Andres et al., 1999), and has been implicated in transcriptional and epigenetic regulation of neural cell fate decisions . Interestingly, no REST homolog has been identified in Drosophila (Bruce et al., 2004). However, Drosophila Charlatan (Chn), which encodes a C2H2-type zinc-finger protein, has several structural and functional similarities to human REST and can associate with CoREST in cultured S2 cells (Tsuda et al., 2006). Chn was shown to repress Dl expression and play a role in the initiation of eye development (Tsuda et al., 2006). To determine whether CoREST plays a Chn-dependent role in the Notch pathway regulation in FCs, we generated chn ECJ1 (Escudero et al., 2005) homozygous mutant FC clones, and looked at the expression patterns of the Notch downstream genes, hnt and cut. We observed wild-type expression of Hnt and Cut in chn mutant cells (supplementary material Fig. S5), and therefore concluded that CoREST plays a Chn-independent role in Notch signaling regulation in FCs.
CoREST is a chromatin regulatory protein that affects H3K27 tri-methylation and H4K16 acetylation in the follicle cells The CoREST transcriptional complex was shown to be involved in regulation of neuronal gene expression in mammals where it acts to modify the acetylation and methylation status of histones through recruitment of histone-modifying enzymes (Ballas et al., 2001;Lunyak et al., 2002;Lakowski et al., 2006). In a twohybrid interaction screen, Drosophila CoREST was shown to interact with Su(VAR)3-3, the Drosophila homolog of LSD1 (lysine-specific histone demethylase 1), and with Rpd3, a Drosophila class-I HDAC (Dallman et al., 2004). LSD1 is able to remove activating mono-and dimethyl groups on lysine 4 of histone 3 (H3K4) (Shi et al., 2004), and HDAC1 removes activating acetyl groups from lysine residues on histones, leading to transcriptional repression.
To determine whether CoREST is involved in the epigenetic regulation of gene expression in FCs, for instance, of components of Notch repressor-activator complexes or of Notch target genes, we analyzed histone modifications in mosaic egg chambers. Using specific antibodies recognizing different histone modifications, we noticed that the state of histones is dynamic in FCs and changes at different stages of oogenesis. Thus, we observed strong staining for H3K4me2, H3K9me3, H3K27me3, H3K9, H4K8, H4K12 and H4K16 acetylation until stages 6-7. All these signals weakened after the M/E transition (data not shown). To determine whether any histone modifications are affected in CoREST GF60 mutant cells specifically as a result of the absence of the functional CoREST protein, and not as a secondary effect of a block in the mitotic cycle, we used hnt FG47 mutant follicle cell clones as a control. Abnormal levels of histone modifications observed in CoREST GF60 mutant FCs, but not in hnt FG47 mutant cells, after stages 6-7 of oogenesis were considered as CoREST specific. Interestingly, we found an increased level of staining for H3K27me3 and acetylated H4K16 in CoREST GF60 mutant FCs after stage 7 of oogenesis, compared with their wild-type neighboring cells or hnt FG47 mutant clones (Fig. 6). Furthermore, this phenotype was observed irrespective of clone position and was not restricted to overproliferating CoREST GF60 mutant PFCs. In addition, we identified abnormally high levels of H3K4me2, H3K9, H4K8 and H4K12 acetylation in both CoREST GF60 and hnt FG47 mutant clones; this phenotype was always coupled to overproliferation of the FCs (data not shown). Therefore, CoREST activity specifically affects the downregulation of H3K27me3 and H4K16 acetylation, whereas other chromatin modifications remain at high levels as a secondary consequence of the failure to promote the M/E switch.

CoREST affects Notch signaling in wings
To determine whether CoREST also regulates Notch signaling in tissues other than the follicular epithelium, we crossed the CoREST GF60 allele to several alleles of Notch (N) and Dl, and examined whether any dominant dosage-sensitive interactions could be detected in the wings of transheterozygous adult flies, because the wing area is very sensitive to defects in Notch (D) Percentage of the egg chambers with Cut-positive cells in the CoREST GF60 homozygous mutant posterior (black), lateral (dark gray) and anterior (light gray) FC clones, in the control, H 1 , CtBP P150 or Gro e47 heterozygous backgrounds. Percentage of egg chambers with Cut-positive cells in CoREST GF60 homozygous mutant clones in wild-type background: 97.5% of PFC clones (n5193), 46% of lateral clones (n5218), 20% of anterior clones (n5110); in CoREST GF60 homozygous mutant clones in the H 1 heterozygous background: 51% of PFC clones (n573), 6% of lateral clones (n564), 0% of anterior clones (n552); in CoREST GF60 homozygous mutant clones in the CtBP P150 heterozygous background: 73% of PFC clones (n5224), 19% of lateral clones (n5206), 3% of anterior clones (n5135); in CoREST GF60 homozygous mutant clones in the gro e47 heterozygous background: 90% PFC clones (n573), 27% lateral clones (n555), 5% anterior clones (n538). The x 2 test was applied for statistical analysis. ***P,0.001 and **P,0.01 for the comparison between CoREST GF60 homozygous mutant clones in the wild-type background and all other groups.
signaling (Guo et al., 1999). Initially, we tested the ability of the heterozygous CoREST GF60 mutant to interact with the Dl 7 allele, which, as we showed previously, dramatically increases the penetrance of CoREST mutant phenotype in the FCs (Fig. 3C,F). The Dl 7 allele is characterized by thickened wing veins and small deltas located near cross-veins (Fig. 7B), and we found that a heterozygous CoREST GF60 allele enhanced this phenotype. In transheterozygotes between Dl 7 and CoREST GF60 mutants, we observed a greater extent of thickening and deltas at the wing veins and even in some cases a severe blistering of the wings (Fig. 7C,C9). Similarly, we investigated whether the CoREST GF60 mutant interacts with Notch alleles, N FU42 and N XK11 (Wieschaus et al., 1984;Hoppe and Greenspan, 1986). Both of these Notch alleles produce the dominant notching of the wing blade (Fig. 7D, and data not shown), and we found that transheterozygotes of Notch and CoREST exhibited slightly thicker wing veins and deltas (Fig. 7E). We also observed a few cases of wing blistering (Fig. 7E9). The occasional wing blistering observed in this study indicates a cell adhesion defect (Zusman et al., 1990;Zusman et al., 1993;Zhang et al., 2010), and this is probably a Notch-independent phenotype. Overall our findings indicate that CoREST acts as a positive modulator of Notch signaling not just in the FCs, but also in wing tissue, indicating a common role for CoREST in the regulation of Notch signaling.

Discussion
Initially, CoREST was identified in humans as a corepressor with REST (RE1 silencing transcription factor) in mediating repression of the proneuronal genes, and thus as an important factor in the establishment of non-neural cell specificity (Andres et al., 1999;Lunyak et al., 2002). Subsequently, CoREST was identified in a variety of vertebrate and invertebrate species, and was shown to  play a functionally conserved role in neurogenesis (Andres et al., 1999;Tontsch et al., 2001;de la Calle-Mustienes et al., 2002;Jarriault and Greenwald, 2002;Dallman et al., 2004). Recent studies show that CoREST regulates a very broad range of genes by both REST-dependent and REST-independent means, including genes encoding members of key neural developmental signaling pathways, such as BMP, SHH, Notch, RA, FGF, EGF and WNT (Abrajano et al., 2009a;Abrajano et al., 2010;Qureshi et al., 2010). Analysis of CoREST downstream target genes and their developmental expression profiles suggested that the liberation of CoREST from gene promoters is associated with both gene repression and activation depending on the cell context (Abrajano et al., 2009a;Abrajano et al., 2009b;Abrajano et al., 2010). In the work reported here, we isolated a lethal allele of Drosophila CoREST, and analyzed the contribution of CoREST to the development of FCs, a process that involves cell proliferation and differentiation. We demonstrated that CoREST is implicated in the regulation of Notch signaling, and acts as a positive modulator of the Notch pathway in Drosophila FCs.
CoREST acts as a positive modulator of Notch signaling in the follicle cells In this study, we identified a previously unidentified function of CoREST in the Notch-mediated regulation of the M/E switch during stage 6 of oogenesis. We showed that the loss of CoREST activity in FCs primarily disrupts the Notch signaling pathway. We further demonstrated that CoREST regulates the Notch pathway downstream of NICD release and upstream of Hnt. The misexpression of Hnt in the CoREST mutant clones rescues the failure in the M/E switch. Furthermore, the role of CoREST in Notch pathway regulation is not restricted to FCs: CoREST also interacts with Notch during wing development. Interestingly, CoREST was identified as a negative modulator of Notch signaling in Caenorhabditis elegans in a genetic screen for suppressors of the developmental defects in sel-12 presenilin mutants (Eimer et al., 2002;Jarriault and Greenwald, 2002;Lakowski et al., 2006). Presenilin is a component of the c-secretase complex that performs the S3 cleavage of Notch (Struhl and Greenwald, 1999;Struhl and Greenwald, 2001;Lai, 2004;Steiner et al., 2008). Mutations in spr-1, the C. elegans homolog of CoREST, suppress the developmental defects observed in sel-12 animals by derepressing the transcription of the other functionally redundant presenilin gene, hop-1 (Jarriault and Greenwald, 2002;Lakowski et al., 2006). Therefore, CoREST acts as a negative regulator of the csecretase complex in C. elegans, and hence proteolytic cleavage of Notch and release of NICD. By contrast, Drosophila CoREST does not affect the processing of the Notch receptor in the follicle cells, and instead acts as a positive modulator of the Notch pathway functioning downstream of NICD release.

Transcriptional and epigenetic functions of CoREST
CoREST plays transcriptional and epigenetic regulatory roles: it can promote gene activation in addition to repression, as well as being able to modify the epigenetic status of target gene loci distinct from its effects on transcription ). Below we discuss several possible scenarios of how CoREST could be involved in the regulation of Notch signaling, based on the previous knowledge about CoREST and considering our data.
We showed that hnt, the downstream target gene of Notch signaling in FCs, fails to be properly upregulated upon Notch activation in the CoREST mutant cells. CoREST might therefore act as a transcriptional repressor for an unknown factor, which is in turn involved in the transcriptional repression of hnt. Alternatively, CoREST could be directly involved in the transcriptional regulation of hnt and act as an activator. hnt was shown to be a putative direct target of Notch signaling in DmD8 cells (Krejci et al., 2009) from the analysis of genes for which mRNA levels increase within 30 minutes of Notch activation, and which contain regions occupied by Su(H). If hnt is a direct target of Notch in FCs, its transcription would be regulated by the balance between Notch repressor and activator complexes, and CoREST might be involved in the regulation of stability or activity of either of these. Interestingly, CoREST was shown to interact with CtBP1 in mammals (Kuppuswamy et al., 2008), and to bind to the SIRT1-LSD1-CtBP1 complex, which is required for the repression of certain Notch target genes (Mulligan et al., 2011). Thus, Drosophila CoREST might similarly directly bind to the repressor complex containing CtBP and modify its activity or destabilize it. However, CoREST could be involved in the transcriptional regulation of the components of Notch repressor or activator complexes. In this scenario, in CoREST mutant FCs, upregulation of negative regulator(s) would lead to greater activity of negative than positive regulators, resulting in disruption of Notch signaling. Both suggested models of the direct and indirect transcriptional role of CoREST are consistent with our results, given that the CoREST mutant phenotype could be suppressed by removal of one copy of H, CtBP or Gro, components of the Notch repressor complex.
More recently, epigenetic mechanisms have emerged as an important interface regulating context-dependent and stagespecific gene regulation. Mammalian CoREST acts as a scaffold for recruitment of transcriptional regulators such as REST, and epigenetic factors such as the enzymes HDAC1, HDAC2 and LSD1 (Lakowski et al., 2006;Qureshi et al., 2010). In Drosophila, using two-hybrid interaction, CoREST was also shown to interact with Su(VAR)3-3 (Drosophila homolog of LSD1) and Rpd3 (HDAC1) (Dallman et al., 2004). In this study, we showed that the levels of H3K27me3 and H4K16 acetylation are significantly and specifically increased in the CoREST mutant FCs. Recently, the H3K27me3 demethylase UTX was shown to act as a suppressor of Notch-and Rb-dependent tumors in Drosophila eyes (Herz et al., 2010), and in addition to increased level of H3K27me3 staining, an excessive activation of Notch was detected in Utx mutant eye discs. The observation of increased levels of H3K27me3 coupled to cell overproliferation and modified Notch signaling in both of these cases [Hertz et al. (Herz et al., 2010) and this study] suggests that the increased H3K27me3 results in epigenetic regulation of genes involved in Notch signaling and/or of Notch target genes. However, in the eye tumor system, this increase in H3K27me3 promotes Notch signaling, whereas in the follicle cells, it reduces Notch signaling. This indicates a strong context-dependent effect on Notch signaling by certain chromatin modifications. Thus, these chromatin modifications might be involved in cell-contextdependent Notch target gene silencing and/or activation (Schwanbeck et al., 2011). Interestingly, many Notch-regulated genes are highly enriched in a characteristic chromatin modification pattern, termed a bivalent domain, consisting of regions of H3K4me3, a marker for actively expressed genes, and H3K27me3, a marker for stably repressed genes; and Notch signaling could be involved in resolving these domains, leading to gene expression (Schwanbeck et al., 2011). Therefore, the increased level of H3K27me3 in CoREST mutant FCs might lead to a repression of certain Notch target genes, for instance hnt.

CoREST regulates Notch signaling independently of the function of known CoREST transcription cofactors, Chn and Ttk
To further understand the function of the Drosophila CoREST in Notch pathway regulation, identification of other CoREST essential and specific binding partners would be useful. One previously identified partner for CoREST is Chn (Tsuda et al., 2006). Given that we observed wild-type expression of Hnt and Cut in chn mutant cells, this factor does not appear to partner CoREST in regulation of Notch signaling in FCs. Using yeast two-hybrid analyses and an embryonic cDNA fusion protein library, it was shown that all three splice variants of Drosophila CoREST interact with the unique C-terminus of Tramtrack88 (Ttk88), a known repressor without homology to REST (Dallman et al., 2004). In addition, a Ttk69 splice variant can form a complex with CoREST and Ttk88 (Dallman et al., 2004). However, Ttk88 was not detected in the ovary by immunofluorescence or western blot analysis, and disruption of Ttk88 does not have any impact on oogenesis (French et al., 2003). Conversely, Ttk69 is steadily expressed in FCs before stage 10 (French et al., 2003;Sun and Deng, 2007) and it is required for the M/E transition (Jordan et al., 2006). However, in contrast to CoREST, which acts upstream of Hnt, Hnt expression is not affected in ttk 1e11 mutant FCs, indicating a role of Ttk69 downstream of Hnt in the control of the M/E switch (Sun and Deng, 2007). Additionally, Ttk69 is not required for cell differentiation, as expression of FasIII, a cell fate marker for immature follicle cells, is normal in ttk 1e11 mutant FCs (Jordan et al., 2006). From these important phenotypic differences between Ttk69, Ttk88 and CoREST, it appears that CoREST plays a Ttk-independent role in Notch pathway regulation in the FCs. Future work to identify transcription regulators that act as binding partners of CoREST will help in determining the precise biochemical role of CoREST in modulating Notch signaling.

Higher sensitivity of the PFCs to loss of CoREST
Our results demonstrate an unexpected role for CoREST in positively regulating Notch signaling. The effect of the loss of CoREST is particularly strong in the PFCs and relatively mild in the lateral and anterior follicle cells. This implies that CoREST is crucially required in cells that are more sensitive to loss of Notch signaling. The difference between the PFCs and the other follicle cells is established at approximately stages 6-7 of oogenesis by EGF receptor activation in response to Gurken produced by the oocyte (Gonzalez-Reyes et al., 1995;Roth et al., 1995). EGF signaling, therefore, is active around the same time as the Notch pathway and hence it is probable that downstream effector(s) of EGFR signaling result in the increased sensitivity of PFCs to the loss of CoREST. In our model of CoREST negatively affecting a repressor of Notch signaling, we would expect EGFR signaling to act positively to enhance expression and/or activity of a Notch repressor. Thus, loss of CoREST from the PFCs would occur in a cell type where repressor activity is already augmented, which would explain our observation of differential loss of Notch signaling in the PFCs.
In summary we have shown that CoREST, a component of transcriptional repressor complexes, acts positively in Notch signaling in the ovarian follicle cells of Drosophila. The results also show that different cell types are differentially sensitive to loss of this repressor. Future identification of partners and targets of CoREST in the follicle cells should further elucidate how activity of EGFR and other signaling pathways are integrated in this process.

Drosophila stocks and genetics
The CoREST GF60 , hnt FG47 and N FU42 mutations were isolated in a mosaic screen for EMS-induced mutations in y w FRT19 flies (Denef et al., 2008). The UAS-Notch line was a gift from Gary Struhl (Struhl et al., 1993). UAS-NEXT and UAS-NICD lines were gifts from Spyros Artavanis-Tsakonas (Rebay et al., 1993), to assay Notch activity E(Spl)-mb-CD2 line was used (de Celis et al., 1998). The CtBP P1590 (Poortinga et al., 1998) and gro e47 (Orian et al., 2007) lines were gifts from Susan Parkhurst. We obtained the N XK11 line from Eric Wieschaus (Wieschaus et al., 1984;Hoppe and Greenspan, 1986). chn ECJ1 FRT42D was a gift from Juan Modolell and Sonsoles Campuzano (Escudero et al., 2005). Dl 7 , hnt EP55 , H 1 and various duplication and P-element lines used for mapping were obtained from the Bloomington Drosophila Stock Center. CoREST GF60 was recombined with the hnt EP55 mutation to create the double mutant chromosome.

Mapping of the CoREST GF60 mutation
We used recombination with visible recessive markers to map the lethal mutation in the CoREST GF60 mutant to the region proximal to f (proximal to 15F7). The lethal phenotype was rescued by duplications covering the genomic region 18B6-18C2; 19A2 (Dp(1;Y)BSC134 and Dp(1;Y)BSC135). Subsequent SNP mapping (Berger et al., 2001;Hoskins et al., 2001;Chen et al., 2008;Chen et al., 2009) placed the lethal mutation between 18C8 and 18D3. We sequenced PCR products of at least two independent genomic isolations covering the region, and compared the sequences with those of the FRT19A control.

Constructs for transgenesis
To generate the pTIGER-CoREST and pTIGER-CoREST-39HA rescue constructs, the coding region of CoREST was obtained by amplification of LD26250 (Drosophila Genomic Resource Center) cloned into pTIGER in frame with or without a C-terminal HA tag. pTIGER was derived from pUASp with enhancements that confer compatibility with WC31 integration strategies while retaining P-element compatibility and enhanced restriction enzyme selection (constructed by Scott Ferguson in T. Schupbach's laboratory; unpublished; details available upon request).

Immunofluorescence staining and microscopy
Ovaries were dissected in PBS, fixed for 20 minutes in 4% paraformaldehyde at room temperature, and stained according to standard procedures (Ashburner, 1989). Primary antibodies used were rabbit polyclonal anti-CoREST (1:1000; a gift from Gail Mandel) (Dallman et al., 2004) Barcelo for technical help, Joe Goodhouse for support with confocal microscopy, and members of the Schupbach and Wieschaus laboratories for feedback and suggestions. We also thank Olivier Mauti, Yi Sun and Shawn Little for critical reading and helpful comments on the manuscript.

Funding
This work was supported by the Howard Hughes Medical Institute and by the US Public Health Service [grant number RO1 GM077620 to T. S.]. Deposited in PMC for release after 6 months.