Combgap Promotes Ovarian Niche Development and Chromatin Association of EcR-Binding Regions in BR-C

The development of niches for tissue-specific stem cells is an important aspect of stem cell biology. Determination of niche size and niche numbers during organogenesis involves precise control of gene expression. How this is achieved in the context of a complex chromatin landscape is largely unknown. Here we show that the nuclear protein Combgap (Cg) supports correct ovarian niche formation in Drosophila by controlling ecdysone-Receptor (EcR)- mediated transcription and long-range chromatin contacts in the broad locus (BR-C). Both cg and BR-C promote ovarian growth and the development of niches for germ line stem cells. BR-C levels were lower when Combgap was either reduced or over-expressed, indicating an intricate regulation of the BR-C locus by Combgap. Polytene chromosome stains showed that Cg co-localizes with EcR, the major regulator of BR-C, at the BR-C locus and that EcR binding to chromatin was sensitive to changes in Cg levels. Proximity ligation assay indicated that the two proteins could reside in the same complex. Finally, chromatin conformation analysis revealed that EcR-bound regions within BR-C, which span ~30 KBs, contacted each other. Significantly, these contacts were stabilized in an ecdysone- and Combgap-dependent manner. Together, these results highlight Combgap as a novel regulator of chromatin structure that promotes transcription of ecdysone target genes and ovarian niche formation.


Introduction
The normal function of many adult organs depends on stem cells and their niches, which make functional and structural units. To make these units, niche precursors and stem cell precursors must coordinate their development. Understanding how this occurs is key to understanding organogenesis and regeneration. Here we show that in Drosophila, the chromatin binding protein Combgap (Cg) controls proliferation and differentiation of somatic niche and germline stem cell precursors, thereby promoting correct ovarian stem cell unit formation.
cg mutants were first isolated by Calvin Bridges in the 1930s [1]. The Cg protein contains 11 putative C2H2 zinc finger domains, a Sir2 domain, a Lambda-1 domain and a poly-glutamine stretch. These motifs suggest Cg acts as a chromatin-binding factor that can either promote or repress transcription. Indeed, Cg has been previously shown to control gene expression in several organs. In the visual cortex, Cg represses the Wingless target genes optomotor blind, decapentaplegic and aristaless [2]. In leg and wing discs, Cg represses cubitus interruptus (ci) expression in the posterior compartment and promotes ci expression in the anterior compartment [3,4]. The pleiotropic effects of Cg and its association with several major signaling pathways suggest that this nuclear protein performs a general function for gene transcription. We now show that Cg is required for correct transcription from the broad complex locus (BR-C), which is one of the major genes required for Germ Line Stem Cell (GSC) unit formation in the developing ovary of Drosophila melanogaster.
The Drosophila ovary makes a good model with which to study stem cell unit formation. During early larval development both Primordial Germ Cells (PGCs, the precursors of GSCs) and somatic cells proliferate [5,6,7]. Somatic niche formation initiates when a pool of proliferating somatic cells become non-proliferating Terminal Filament (TF) precursor cells (Fig 1A). TF precursors accumulate throughout third instar, acquire adult TF markers from mid-larval third instar (ML3), and then form filaments by convergent extension [8]. By Late Larval third instar (LL3) all 16-20 TF stacks, which are contained within a single ovary, have formed [9,10,11,12]. Following somatic niche formation, PGCs initiate their differentiation. However, PGCs that are close to the newly formed niches are spared and become the adult GSCs [6,7,13,14].
Our previous work underlined the importance of the hormone ecdysone as a switch signal between proliferation and differentiation and as a coordinator of somatic niche formation with GSC establishment [7,8,13]. At early larval stages, the ecdysone receptors EcR and Ultraspiracle (Usp) act as repressors of niche and PGC differentiation. This allows PGCs and somatic precursors time to proliferate. At third instar, consecutive ecdysone peaks promote niche formation and then PGC differentiation. Ecdysone signaling is known to activate many genes. One of its earliest targets is the BR-C Locus [15]. This locus encodes four related transcription factors that share a BTB domain and differ in their zinc finger moiety (Br-Z1, Z2, Z3 and Z4) [16,17,18]. In larval ovaries, ecdysone activation is associated with up-regulation of Br-Z1 [13]. BR-C is a large locus (over 100KBs) that contains many control elements, including two experimentally confirmed proximal and distal promoters [16,17]. In addition to its major role in GSC unit formation, BR-C transcripts are required for the correct development of Intestinal stem cells, imaginal discs, and the nervous system [19,20,21,22]. It also functions during immune response, metamorphosis and developmentally regulated cell death [18,22,23]. Understanding how transcription is controlled in this complicated locus will therefore provide us with a deeper insight into the transcriptional control of many developmentally regulated processes.
Here we show that ovarian development depends on Cg and its control of BR-C expression. Furthermore, we show that two EcR-enriched regions in BR-C are associated with each other and that this association increases in an ecdysone-and Combgap-dependent manner. We propose that Cg may act as part of a general machinery that modifies high-order chromatin associations.

Cg promotes niche formation and represses precocious PGC differentiation
In a screen to identify novel regulators of ovarian development [13] we identified the transcription factor Combgap (Cg) as required for proper niche formation and PGC maintenance. To follow the differentiation status of PGCs, we used anti-Hts antibodies that label the fusome, an intracellular organelle within germ cells. In wild type LL3 ovaries, PGCs do not yet differentiate to form germline cysts, and each PGC carries a spherical fusome (Fig 1B, arrowhead) [6,24,25]. By contrast,~50% of cg-mutant gonads from three different allelic combinations contained branched fusomes, suggesting PGCs differentiated precociously (Fig 1C and 1D, arrowheads, insets, for quantification see legend). Anti-Orb antibody, which robustly stains more developed 8-and 16 cell cysts [26], confirmed these observations. No Orb labeling was observed in WT ovaries ( Fig 1E). However, Orb was expressed in cg-mutant gonads (Fig 1F and 1G). Thus, both fusome morphology and Orb expression demonstrate that the wild-type function of Cg is required to prevent precocious PGC differentiation.
In addition to PGC maintenance, Cg also affected niche formation. To follow TF cell accumulation and stack formation we used anti-Engrailed (En), which specifically stains TF cells [27]. At LL3, TF cells in WT or heterozygous ovaries were already stacked to form the adult number of niches ( Fig 1H, Table 1) [8,12,28]. By contrast, TF formation was defective in cg mutants. This was particularly noted in the cg 1 /cg KG00882 combination ( Fig 1I, Table 1). However, fewer TFs were also noted in other allelic combinations (Table 1). These ovaries also displayed shorter TFs (Fig 1J, arrows), an additional indication for a difficulty in TF formation. These phenotypes suggest that Cg is required for the proliferation/survival of somatic precursors, for the process of TF specification or for its correct timing [7].
In accordance with a possible role for Cg in precursor cell proliferation, cg-mutant gonads were significantly smaller than WT (Table 1). Vital dye staining (Propidium Iodide, PI) indicated that cell death did not increase in cg-mutant ovaries (of 18 cg 1 /cg KG00882 ovaries, 16 were not labeled by PI at all, and 2 had 1-3 dead cells, a similar level to WT ovaries) [8]. On the other hand, phospho-Histone H3 (pH3) staining, which marks mitotic cells, was significantly decreased in cg-mutants (0.15±0.03 pH3-positive cells per 1 μm 3 in WT ovarian volume, n = 21, as compared to 0.08±0.02 per 1 μm 3 in cg 1 /cg KG00882 ovaries, n = 14, t-test p-value 1.35e-6). Thus, Cg affects gonadal size by promoting cell proliferation but not cell survival.
Cg is required in the soma for ovarian development Cg promotes somatic development and niche formation on the one hand, and represses precocious PGC differentiation on the other. To determine how might Cg affect both soma and germline, we first examined its expression using anti-Cg antibodies [4], and found that it was present in all nuclei of WT LL3 ovaries (Fig 2A). Similar results were obtained by examining the expression of a GFP protein trap inserted in the Cg locus ( Fig 2B) [29]. The ubiquitous ovarian Cg expression is in accord with its ubiquitous expression in other organs [2,4].
To determine the origin of the somatic and germline phenotypes of cg-mutants, we used mosaic analysis. Germline clones of either cg 2 or cg KG00882 did not develop as cysts within larval ovaries, and contained spherical fusomes (Fig 2C and 2D). Similarly, depletion of Cg specifically from PGCs by expressing UAS-cg RNAi with nos-Gal4 resulted in normal development of both somatic and germline lineages (S1 Fig). We conclude that Cg does not act within PGCs to maintain normal ovarian development.
Expression of UAS-cg RNAi with tj-Gal4, which is specific to the ovarian soma, also failed to produce phenotypes, possibly due to a low efficiency of the RNAi (S1 Fig, Table 1). Indeed, expression of an efficient RNAi line against GFP in Cg-GFP ovaries resulted in low Cg levels in most somatic ovarian nuclei (Compare Fig 2E and 2F and S1 Fig). GFP RNAi ovaries were significantly smaller and contained fewer TFs (Table 1). In addition, the formed TFs in the GFP RNAi ovaries were less developed and incorporated fewer TF cells per TF stack (Compare TFs under arrows in 2E, 2F, S1 and S2 videos). Thus, Cg reduction in the ovarian soma can phenocopy the somatic ovarian phenotypes of cg mutants. By contrast, PGC development was normal in the GFP RNAi ovaries (S1 Fig). One possibility is that PGC differentiation requires simultaneous depletion of Cg in both somatic and PGC nuclei. Alternatively, systemic effects could be causing PGC differentiation in cg mutant animals (see below).

Cg is required for the correct expression of Broad-Complex
An additional phenotype of cg-mutant animals was their failure to pupariate at the end of larval development. cg-mutant larvae continued feeding and growing for several days after their WT siblings pupariated, eventually dying as giant larvae (S2 Fig). Such phenotypes suggest that Cg may somehow affect the ecdysone pathway, which induces molting and pupariation [30]. Since ecdysone signaling is required for ovarian development [8,13,31,32], we investigated whether Cg affected the ovarian ecdysone response, and in particular the Broad-complex (BR-C), its major ovarian target during larval development [8,13].  Table 1). In addition, many TFs are shorter than WT TFs Of the four Br isoforms, Br-Z3 is not expressed in the ovary and Br-Z4 is expressed at very low levels at the larval-pupal transition, after gonad morphogenesis has largely occurred. We therefore concentrated our analysis on Br-Z2, which is expressed in the somatic ovary from early larval stages, and on Br-Z1, which is induced by ecdysone at mid-3 rd instar [13]. Isoformspecific antibody staining in cg 1 /cg KG00882 ovaries revealed a reduction in Br-Z1 and Br-Z2 protein levels (For Br-Z1, n = 13, t-test p-value 1.44e-7, Fig 3A and 3B. For Br-Z2, n = 19, t-test pvalue 2.13e-10, Fig 3C and 3D). Since BR-C levels could be indirectly affected by the retarded development of cg-mutant animals, we also analyzed Br-Z1 and Br-Z2 in somatic mutant clones that were generated in a WT heterozygous background. Interestingly, cg-mutant somatic clones showed no reduction of Br-Z1 (S3 Fig), suggesting that gonadal Br-Z1 expression could be indirectly affected by the developmental retardation of cg mutants. By contrast, Br-Z2 levels were consistently reduced in cg somatic clones (Fig 3E-3F'). This demonstrates that Cg promotes Br-Z2 expression in a cell-autonomous manner, which is independent of changes in larval physiology or development.
To determine the extent to which reduced Br-Z2 expression contributes to gonadal phenotypes, we over-expressed Br-Z2 in cg 1 /cg KG00882 ovaries using a UAS promoter. Br-Z2 gonadal expression was rescued in these animals, and gonadal size and TF numbers increased, as compared to cg 1 /cg KG00882 mutants (

Cg over-expression represses Broad expression and blocks ovarian development
To further probe how Cg might affect BR-C expression, we over-expressed it using a UASinsertion line upstream of cg. Cg over-expressing LL3 ovaries were markedly smaller than WT ovaries and contained very few TFs (Fig 4A and 4B and Table 1). To further determine the developmental stage of Cg over-expressing ovaries, we stained them with anti-Tj. In young WT ovaries, Tj is expressed in all somatic gonadal nuclei, but during third instar, Tj expression is gradually restricted to ICs [5,33]. Indeed, only ICs were labeled by anti-Tj in WT LL3 ovaries ( Fig 4C). By contrast, all somatic cells were labeled in the Cg over-expressing gonads, and none of these cells intermingled with germ cells (Fig 4D). Thus, Cg-over expressing ovaries seem arrested at a younger developmental stage. Another feature that characterized both Cg-over expressing ovaries and young ovaries, was the absence of a posterior group of somatic cells (swarm cells), which migrate to the posterior during third instar. In Cg-over expressing ovaries, swarm cells failed to migrate and remained at the medial side (Compare Fig 4E to 4F, white star). Combined, these features suggest that Cg-over expressing ovaries failed to mature.
The gross morphological abnormalities of Cg over-expressing ovaries resembled the developmental defects in ovaries over-expressing the dominant negative EcR [13]. Indeed, examination of Broad expression using an antibody directed against the core region of BR-C, which labels all Broad isoforms, revealed a strong Broad suppression in Cg over-expressing cells. At ML3, all somatic WT nuclei express BR-C ( Fig 5A-5A''). tj-Gal4, which is expressed in all somatic ovarian nuclei at this stage [5], led to Cg over-expression in most somatic nuclei ( Fig  5B and 5B''). Consequently, hardly any BR-C expression was observed in these nuclei ( Fig 5B' and 5B'').
At LL3, due to the morphological defects caused by Cg over-expression, cells expressing high Cg levels were located at lateral ovarian regions ( Fig 5D). In these cells, BR-C levels were very low or non-existent ( Fig 5D' and 5D''). Interestingly, cells at the border of the region overexpressing Cg retained some BR-C labeling, suggesting a non-autonomous contribution to   BR-C expression (Fig 5D', arrowheads). Together, these results show that correct Broad expression requires wild-type levels of Cg, and any change in Cg amounts or function disrupts expression from this important locus.

EcR binding to polytene chromosomes depends on correct Cg levels
To determine how direct the control of Cg on BR-C expression is, we asked whether Cg is localized to the BR-C locus. At LL3, increased transcription from BR-C can be directly observed as puffing of the locus in polytene chromosomes of salivary glands [34]. Staining with anti-Cg revealed that Cg was localized to the BR-C puff in 84.5% of all chromosomes tested (n = 58, Fig  6A' and 6A''', green arrow). Moreover, co-labeling with anti-EcR revealed that in all these cases, Cg co-localized with EcR ( Fig 6A" and 6A''', white arrow). This suggests that transcription from BR-C may depend directly on Cg. EcR and Cg co-localized in some loci but not in others (Fig 6A'-6A''', green, magenta and white arrowheads). This suggests that Cg may not be an obligatory component of the EcR transcriptional complex. Alternatively, Cg may be present in some EcR-containing polytene chromosome bands at levels that are undetected by antibody staining, or associate with EcR complexes in a temporally-restricted manner.
As EcR and Cg affect BR-C expression, and both co-localized at the BR-C locus, we investigated whether EcR localization may depend on Cg. In the strong cg 1 /cg KG00882 allelic combination, lower Cg levels were detected on polytene chromosomes (100% of spreads, n = 18, Fig 6B' and 6B'''). On such chromosomes, EcR was still present, but at reduced levels (89% of spreads, n = 18, Fig 6B'' and 6B'''). In addition, the chromatin-banding pattern was loose, suggesting a global effect on chromatin by Cg (Fig 6B). Other allelic combinations displayed similar, albeit weaker phenotypes: The reduction in Cg levels in the weaker allelic combinations was not as severe, resulting in a smaller reduction in EcR binding to chromatin (S5 Fig).
Our data show that in the somatic cells of the ovary, a reduction in Cg protein results in lower Br-C induction, while over-expressing Cg results in an almost complete block of Br-C expression (Figs 3 and 5). We therefore investigated the status of EcR binding to polytene chromosomes of salivary glands that over-expressed Cg. In 45% of chromosome spreads, Cg was indeed over-expressed, coating the entire chromatin (n = 58, Fig 6C' and 6C'''). In such spreads, EcR binding was almost completely undetected (92% of Cg over-expressing spreads, Fig 6C'' and 6C'''). The effects of Cg levels on EcR binding to chromatin (reduction in the mutant background and severe reduction upon Cg over-expression), match the manner by which changes in Cg levels affect BR-C expression. Together, these results confirm that normal EcR activity depends on tight regulation over Cg levels and function.

Cg promotes long-range chromatin contacts at the BR-C locus
The changes in chromatin structure observed in polytene chromosome of cg 1 /cg KG00882 animals, suggested that Cg may affect the chromatin structure at the BR-C locus. To test this possibility we used circularized chromosome conformation capture with high-throughput sequencing (4C). To compare chromatin conformation prior to-and following exposure to the ligand, and to ask how may Cg affect this response, we used Kc167 cells, which respond well to the ligand (see below). Since our observations in polytene chromosomes suggested a connection between EcR and Cg, we selected as viewpoints into the BR-C locus two regions that are associated with EcR. The first viewpoint, B1 (at 1609904), was located between the distal and proximal promoters, 2 Kilo-Bases (KB) away from EcR binding sites. The second viewpoint (B2, at 1619647) was close to the proximal promoter, in a region within the first intron, which binds EcR in-vivo (S6 Fig) [35,36,37].
We observed a domain-like contact landscape around the B1 viewpoint, with asymmetric contacts extending some~250KB downstream (Fig 7A, color coded domainogram), and 125KB upstream. Analysis of B2 indicated several long-range contacts that were enriched over the typical domain architecture (Fig 7A, right). We selected the three loci with the strongest evidence for long-range contacts for further quantitative analysis (X1-X3). Significantly, these three contact points constituted the three sites with the highest EcR-binding content within this~370KB region (marked as red bars underneath X1-X3, Fig 7A, right, S6 Fig) [36,37]. Thus, the EcR binding region within BR-C is engaged in long-range contacts with other EcR-binding regions.
We compared systematically the 4C contact profiles around the B1 and B2 viewpoints between conditions, using differential contact analysis at multiple scales (Fig 7B). Quantitative follow up of the contact intensity at the X1, X2 and X3 loci (Fig 7C) showed that the overall EcR is bound at the same puff (magenta arrow in A"). In the composite (A"'), co-localization of the two antibodies appears in white. Two such bands are marked (white arrows). In addition numerous bands demonstrate unique EcR or Cg binding (magenta and green arrows, respectively). (B-B"') Chromosomes spread from cg 1 / cg KG00882 mutant larva. Chromatin bands are not as tight, very little Cg is observed and EcR binding to chromatin is reduced. (C-C"') Cg was over-expressed in salivary glands using sgs3-Gal4. A large amount of Cg, which spreads throughout the chromatin is observed. Little EcR can be detected on such chromosomes.  contact architecture was stable. However, ligand exposure and Cg presence modified some contacts in significant ways: The closest contact (X1) was located~30KBs upstream, next to the distal BR-C promoter. Our data indicated about 27%-22% increase in the number of contacts between X1 and either B1 or B2, respectively, following ecdysone treatment (Fig 7C). Notably, a significantly weaker increase was detected when cells were pre-treated with cg ds-RNA. Thus, exposure to ecdysone increases BR-C transcription and contact stability. Both these phenomena are Cg-dependent. Reciprocal 4C confirmed the increased number of contacts between B1 and B2 to X1 upon ecdysone treatment and strengthened the dependency of this interaction on Cg (S7 Fig).
To further substantiate the effects of Cg-depletion on the BR-C locus, we studied its conformation in wing imaginal discs derived from wild type and from cg 1 /cg KG00882 late third instar animals. At this developmental stage, ecdyosne titers are high, and BR-C transcription peaks. Contacts between X1 and B1 or B2 were significantly lower in cg-mutant animals as compared to wild type (Fig 7D). This conforms to the reduced BR-C expression and reduced EcR-binding to the locus that is observed in mutant animals. Thus, in-vivo studies confirm the dependency of contacts at the BR-C locus on normal Cg levels.
The second contact region (X2) was~228KB downstream of B2, within CG42666, and showed an ecdysone -dependent decrease in contacts with B1 and B2 (Fig 7C). Finally, The most distant contact (X3) was~327KB downstream of B2, within the first intron of DHR4. X3 showed a significant increase in contacts with B2 upon ecdysone treatment. Contacts between B1/B2 and X2 or X3 were not affected by cg knockdown (Fig 7C).
In summary, our 4C analysis showed a complex chromatin structure at the BR-C locus, featuring long-range contacts between EcR-binding regions across~370KB. These connections are differentially modulated by exposure to ligand, but only the contacts within the BR-C locus are Cg-dependent. 4C profiles around BR-C serve as a preliminary indication to the diversity of changes both ligand and Cg confer upon the BR-C chromatin landscape.

Cg is required for proper EcR response in Kc167 cells
Our results demonstrate that EcR binding regions in BR-C are associated with two other genes containing EcR-binding regions, and that EcR binding to chromatin depends on Cg. This prompted us to ask whether Cg and EcR might physically interact, and whether Cg may also affect the transcription of DHR4 and CG24666. Our attempts to co-immunoprecipitate EcR and Cg from either adult ovaries or Kc167 cells were unsuccessful. Thus, either the two proteins do not physically interact, or they are part of a temporally restricted or labile complex. To examine this last option, we measured EcR-Cg interactions using a proximity ligation assay (PLA), which can detect physically associated proteins without cell disruption.
Control, β-Gal ds-RNA, cells contained an average of 5.09 PLA spots located at their nuclei (n = 461 cells, Fig 8A and 8B). Treatment of control cells with 1μM ecdysone for 2 hours increased the average number of spots to 7.26 (n = 490 cells, Fig 8A and 8B). A Mann-Whitney U test determined that this change is highly significant (p vaue 3.54E-13). cg dsRNA cells profile, typically based on 1-3M sequenced reads. (B) Comparison of 4C profiles in control cells treated with ecdysone (red) to untreated cells (black, top) or treated cells in which Cg was knocked down (blue, bottom). Each comparison shows overlaid contact intensity trends on top. The domainogram is showing the differential contact intensity in increasing genomic window sizes. Red bands on the dominogram indicates contacts in the region are stronger following ecdysone treatment. (C) Bar-plots represent the mean contacts for windows X1, X2, 200 fragments per each and 30 fragments for X3. The number of contacts per window is color-coded. (D) Bar-plots representing the mean contacts for X1 with viewpoints B1 and B2 in wing imaginal discs derived from wild type and cg 1 /cg KG00882 animals, 40 fragments. **P<0.001, *P<0.05 (Chi-Square pair-wise test).
doi:10.1371/journal.pgen.1006330.g007 treated with ecdysone contained only 3.27 spots on average (n = 320, Fig 8B, p value 1.06E-35, compared to ecdysone treated control cells), attesting to the specificity of the assay. Together, the PLA and polytene chromosome studies suggests that only a fraction of Cg and EcR may reside within the same complex. Some of these complexes may be preformed, while others either form de-novo, or are stabilized upon exposure to ecdysone.
We next tested how may Cg affect transcription of genes in contact with BR-C. We first measured isoform-specific induction of BR-C mRNA levels upon ecdysone treatment. Control Kc167 cells treated with dsRNA against β-Gal, responded to ecdysone exposure by up-regulating both BR-Z1 and BR-Z2 transcripts, with BR-Z2 induction consistently higher than BR-Z1 (Fig 8C). When cg levels were reduced by cg-dsRNA (S8 Fig), up-regulation of either transcript was attenuated by about half (Fig 8C). The dependence of BR-C expression on Cg in both gonadal and Kc167 cells, which originate from embryonic blood cells, suggests that control of this locus by Cg is a ubiquitous phenomenon.
The first associated gene, CG42666, lying next to X2, was not induced by ecdysone (Fig 8C). Since contacts between BR-C and CG42666 were reduced by ligand treatment (Fig 7C), this suggests that following exposure to ligand, some long-range contacts with inactive EcR targets may be diminishing.
On the other hand, transcription of DHR4 (which encompasses X3) was greatly increased by ecdysone treatment. Induction of this gene was an order of magnitude higher than that of BR-C (Fig 8C). Contacts between BR-C and DHR4 increased upon ligand exposure (Fig 7). Thus, the contacts with the un-induced gene diminish, while those with the induced gene increase. Significantly, DHR4 transcriptional induction by ecdysone was reduced in cg-RNAi cells. Thus, Cg can affect other ecdysone-induced genes besides BR-C.

Discussion
The BR-C locus controls a variety of processes ranging from stem cell establishment, cell death, neural network organization to animal behavior [13,18,19,20,21,22,23]. Despite its central role in fly biology, little is known about how this complex locus is being controlled. Our work shows that correct levels and function of the chromatin binding protein Combgap are required for Br-C expression. We further show that EcR-enriched regions within BR-C are engaged in long-range associations with each-other and with other EcR-enriched loci. Lastly, our data indicate Cg affects transcription of EcR-induced genes by controlling the access of EcR to chromatin.

Cg and EcR-mediated BR-C expression
Cg is associated with the BR-C locus and its RNA and protein products in three different cellular systems: somatic ovarian cells, Kc167 cells and in salivary glands. One major inducer of transcription from this locus is EcR [15,17,34]. Several lines of evidence suggest that Cg effects BR-C via EcR. First, EcR and Cg co-localize at the BR-C locus. Second, we observe reduced EcR binding to polytene chromosomes in cg mutants. Third, EcR binding to chromatin is almost completely abolished when Cg is over-expressed in salivary glands. These changes in EcR binding to chromatin fit very well with how Cg levels affect BR-C in somatic ovarian cells. We observe a reduction in Br-Z2 levels in cg-mutant clones, and a complete block of all BR-C isoforms in cells over-expressing Cg. Cg levels affect EcR in a global manner, suggesting other EcR targets are likely affected. Indeed, the transcription of another EcR target, DHR4, is also reduced in cg-RNAi Kc167 cells.
Lastly, the PLA approach suggests that at least some nuclear complexes contain both EcR and Cg. It is likely that EcR and Cg do not directly bind each other, as co-IP approaches failed to establish a direct connection. In addition, Cg is not dedicated exclusively to the ecdysone pathway, and also affects Wingless and Hedgehog target genes [2,3,4]. Indeed, the amount of PLA spots and co-stained bands in salivary glands suggest that only a minor fraction of Cg is engaged in such stable complexes. We hypothesize Cg may have pleiotropic roles, and engage in different or more general complexes. Indeed, a recent publication identified Cg as a PREbinding protein, and as co-localizing with pleiohomeotic and polyhomeotic (Ph) [38]. It is unclear if this interaction overlaps with how Cg affects EcR and the BR-C locus, or if it represents an entirely independent function. The exact biochemical function of Cg as a more general mediator of protein binding onto chromatin requires further investigation.

Association between EcR-enriched loci
4C analysis shows that EcR-enriched regions within the first intron of BR-C contact EcRenriched binding site close to the distal BR-C promoter,~30KBs upstream. The interactions between these EcR-enriched regions existed in Kc167 cells prior to exposure to ecdysone, but increased following ligand binding. Our results support a model where the chromatin is pre-programmed for specific transcriptional responses, and ligand exposure can modulate the intensity of these pre-formed connections. The ecdysone response is rapid, and would benefit from a pre-arranged chromatin structure that enables a coordinated transcriptional response for the hormone.
The contacts of EcR-enriched regions between BR-C, CG42666 and DHR4 suggest that BR-C is transcribed in association with other genes, or in a 'multigene' complex. Such complexes are prevalent in mammalian genomes [39]. Interestingly, contact intensity increases upon ecdysone treatment for BR-C and DHR4, the two genes displaying ecdysone-induced transcription. Contacts intensity is reduced for CG42666, which does not respond to ecdysone treatment. At this stage we cannot determine whether the changes in chromosome conformation participate in regulating ecdysone-induced transcription.
Within the Br-C locus, increased contact intensity is Cg-dependent. The decreased intensity following cg-reduction is observed in both Kc167 cells, and to a larger extent, in cg-mutant animals. The consistency of the effect in cells of different origin suggests that Cg affects the conformation of DNA at the BR-C locus in a global manner. The larger effect in the mutant animals is likely due to the inefficiency of the dsRNA technique in tissue culture.
Unlike the contacts between the distal BR-C promoter and the BR-C first intron, the contacts with DHR4 are Cg-independent. As Cg affects the transcription of both BR-C and DHR4, this adds to the complexity of the regulatory interactions in this region. In a similar vein, qPCR measurements show an order of magnitude difference in ecdysone induction levels between BR-C and DHR4. Thus, induction strength is not determined in this case by the association per-se, but by individual promoters and/or the existence of other elements such as insulators, which participate in ecdysone-induced chromatin looping [35,40]. Whether Cg may participate in these complexes or even mediate some or their differences await biochemical analysis.

Cg, BR-C transcription, and gonad morphogenesis
While qPCR measurements in Kc167 cells show cg-dsRNA reduces transcription of both BR-Z2 and BR-Z1, protein measurements in somatic ovarian clones show a marked reduction only for BR-Z2 protein levels. One option to explain this difference involves the smaller reduction of BR-Z1 transcripts in cg-dsRNA as compared to BR-Z2. This smaller difference may still be detected by qPCR, but not by antibodies. A second, not mutually exclusive explanation may be connected to post-transcriptional control. The reduced BR-Z1 levels may trigger some compensation at the level of splicing or protein translation. Alternatively, the manner by which Cg controls transcription from the BR-C locus in ovarian and Kc167 cells may be somewhat different.
BR-Z1 protein levels are reduced in whole-body mutants but not in clones, suggesting a systemic component for cg-depletion on ovarian BR-C transcription. During late larval stages, BR-C acts in a feed-forward loop, augmenting ecdysone production [41]. Cg may therefore affect ovarian BR-C expression both autonomously and systemically. An antagonistic effect may be caused by DHR4, which is a negative regulator of ecdysone production [42] and is also reduced by cg-depletion. How these two opposing effects balance out in the prothoracic gland, which synthesizes ecdysone, remains to be investigated. However, cg-mutant larvae, which do not pupariate, suggest that Cg depletion results in insufficient ecdysone production.
Finally, our previous work showed that ecdysone signaling promotes three processes during gonadal development: gonadal growth, niche differentiation, and PGC differentiation. ecdysone signaling in the larval ovary promotes BR-Z1 expression exactly at the time when niches and PGCs differentiate. This suggests that BR-Z1 is necessary for stem cell unit differentiation. Indeed, expression of a dominant negative EcR abolishes both BR-Z1 expression and stem cell unit differentiation [13]. BR-Z2 is expressed even prior to stem cell unit differentiation, and therefore cannot be sufficient to induce differentiation. However, BR-Z2 may still be necessary for the correct differentiation of niches and PGCs, together with BR-Z1. Indeed, BR-Z2 could partially rescue both ovarian growth and niche differentiation.
One important difference between the effects of BR-C and Cg on gonad morphogenesis is that while both factors promote ovarian growth and niche formation, Cg retards PGC differentiation while BR-C promotes it. This suggests that Cg affects other targets important for gonadal growth. Previous work uncovered many signaling pathways that are important for niche formation and GSC establishment [7]. Understanding how these pathways converge to form the organized organ is crucial for basic understanding of organogenesis and for applications in regeneration. Uncovering how pleiotropic proteins such as Cg contribute to coordinated transcription and hence to organogenesis will be part of this effort.

Larval dissections and staining
Larval developmental timing was determined as described previously [13]. In short, crosses were performed in bottles, to ensure a sparse larval population. In these strict under-crowded conditions, larval development is uniform. Laying was performed for two hours, and eggs were allowed to mature at 25°C until late third instar (120 h after laying). cg mutants, which do not pupariate, were dissected 5-7 days after laying. Females were dissected and stained as previously described [13,43].

Antibodies
Generation of BR-Z2-specific antibodies was according to [44]: The Br-Z2 isoform-specific region was PCR amplified (forward primer: GGATCCAAGGAAACTCGCCCAAGAAACTC, reverse primer: GAATTCTGCTGTGGCTGTTGGCTTTGC).The amplified region was cloned using BamHI and EcoRI, into pGEX-3X bacterial expression vector in E. coli BL21 strain. Bacterial cultures were grown until OD 600nm 0.5 and induced with 2mM IPTG for 4h. GST fusion proteins were purified on glutathione agarose (Sigma). The purified GST-BrZ2 was used to immunize New Zealand female rabbits. The first injection was in Freund's complete adjuvant and the boost, spaced for at least 3 weeks, was in Freund's incomplete adjuvant. A total of 3 boosts were performed and sera taken 2 weeks after the last boost was tested for immunoreactivity against the injected proteins. The resulting Ab reacts specifically with Br-Z2 (S9 and S10 Figs).

Tissue culture
Kc167 cells were maintained at 25°C in Shields and Sang M3 insect medium (Sigma) containing 10% fetal bovine serum (Gibco) and 1% penicillin-streptomycin (Gibco). For RNAi, dsRNA synthesis was according to DRSC protocol using Readymix (Sigma), MEGAscript T7 kit (Ambion) and RNAeasy (Qiagen). Amplicons from the DRSC database DRSC32301 for combgap and DRSC24562 for β-Gal were used. 2 μg of dsRNA against either combgap or LacZ together with 2 μl DharmaFECT4 (Dharmacon) were applied to 0.4x10 6 cells in 12-well plates. Cells were exposed to dsRNA for 72 hours and then 1μM ecdysone or control 95% Ethanol were added to the medium for 2 hours. Following ecdysone exposure, cells were harvested for qPCR.

4C-Seq
The preparation of 4C template (five biological repeats) was performed as described previously [45], with some modifications. 5X10 6 Kc167 cells were fixed and following cross-linking, the DNA was digested by the 4 cutter restriction enzyme DpnII (NEB), ligated by high concentrated T4 ligase (NEB). The ligation products were sonicated in to fragments at the average length of 500 bp, end repaired by fill-in and exonuclease reaction and then ligated to an adaptor sequence for the Illumina Trueseq process. 150-200 ng of the resulting 4C template were used for 2 sequential PCR reactions. The first PCR was performed by primers designed to target the region of our interest and illumina Truseq enrichment PCR primer 2. In the second PCR, we used primers that include the sequence illumina Truseq universal adapter AATGA TACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCTTCCGATCT followed by specific internal sequence together with illumina Truseq enrichment PCR primer 2 (see below). The primers for the first PCR were designed to target region that is far by 80 to 98 bp from the DpnII site. The primers (internal primes) for the second PCR were designed to target the region that is far by around 3-80 bp from the same DpnII site.
The primers that were used for two BR 4C viewpoints (B1, B2) and for the primers that were used for three contact regions (X1, X2, X3) are listed in the following table 2.
Sequenced products were mapped to the genome using an in house pipeline (available upon request), and the precise sonication sites were used to eliminate PCR duplicates. 4C domainograms [46] show a smoothed contact intensity profile with gray error band representing 2 SD on the estimated mean contact intensity in a running window. It also shows a color-coded visualization of contact intensities in a series of increasing scales (from 10 restriction fragments to 300 restriction fragments). As the modified protocol provide precise control over PCR duplicates, linear means of the number of ligation evens per restriction fragment within windows can be used. Similarly, statistical tests comparing the total number of ligations in a given window between two conditions (following normalization for total sequence coverage) are robust. This was reconfirmed for the current experiments by quantifying technical variance on triplicate experiments.

Imaginal Discs harvesting for 4C analysis
Inverted head segments of 400 OR (wild type) or 200 cg KG000882 /cg 1 animals were fixed in 2% formaldehyde in PBS (20 min., Room Temperature). Fixed tissue was then washed twice with PBS/125mM Glycine/0.01% Triton-X-100, and a third time with PBS. Samples were washed again with PBS/1% protease inhibitor cocktail (P8340, Sigma). Wing Imaginal discs were then dissected out and placed in an eppendorf tube. Excess fluid was removed and tissues were snap-freezed in Liquid Nitrogen and stored at -80°C. To prepare DNA, discs were thawed on ice and centrifuged at 13,000 rpm for 20 sec. Excess fluid was discarded. Discs were resuspended in 50 ul lysis buffer (10 mM Tris-HCl pH 8.0, 10 mM NaCl, 0.2% Igepal CA360 (Sigma I8896)), and 10 ul/ml of protease inhibitors (Sigma P8340). Discs were homogenized with a plastic motorized pestle (3X2min). Following a brief centrifugation, 500 ul of lysis buffer and 50 ul of protease inhibitors were added and the suspension was centrifuges at 5000 rpm for 5 min, at RT. Lysate was washed twice with ice-cold 1.2X NEBuffer 3 at 5000 rpm, 5 min, at RT, and resuspended in 500 ul NEBuffer 3, supplemented with 7.5 ul 20% SDS. Mixture was incubated at 37°C, rotating at 900 rpm for 1h in a Thermomixer, followed by an addition of 50 ul of 20% T-X-100 and an additional incubation at 37°C, rotating at 900 rpm for 1h in a Thermomixer. Lysate was digested with 400 U of DpnII at 37°C, 900 rpm overnight in a Thermomixer. Enzyme inactivated at 65°C, 20 min. Ligation and continuation of the 4C protocol was as performed for KC cells.

Polytene chromosome squash
Larvae were grown at 18°C in a non-crowded culture. Wandering 3 rd instar larvae were washed and dissected in Ringer's buffer. Extracted salivary glands were spread according to published protocol [47]. Rabbit-anti-Cg (1:1000) and mouse-anti-EcR (1:20) were used to localize EcR and Cg to the broad locus, and DAPI staining was used to highlight the banding pattern. The br locus was identified by the banding pattern that is specific to the tip of the X chromosome and by the puffing pattern specific to the ecdysone response at late third instar. Spreads of polytene chromosomes from salivary glands were stained with Dapi (white), anti-Cg (green) and anti-EcR (magenta). Glands were stained and imaged using the same confocal settings and on the same day as those of WT (Compare to Fig 6). Cg staining is weaker than WT in the cg 1 / cg 2 (A-A"') and cg 2  S2 Video. Third instar gonads of the genotype dicer2; tj-Gal4, Cg-GFP; UAS-GFP RNAi , in which cg is reduced, were dissected and stained with anti-En (magenta), which labels TFs. Gonads are smaller than WT (compare to S1 video). Less TFs are present than in WT ovaries. While some TFs are long and mature, others are shorter.

Supporting Information
(AVI)