The Enhancer of Trithorax and Polycomb Corto Interacts with Cyclin G in Drosophila

Background Polycomb (PcG) and trithorax (trxG) genes encode proteins involved in the maintenance of gene expression patterns, notably Hox genes, throughout development. PcG proteins are required for long-term gene repression whereas TrxG proteins are positive regulators that counteract PcG action. PcG and TrxG proteins form large complexes that bind chromatin at overlapping sites called Polycomb and Trithorax Response Elements (PRE/TRE). A third class of proteins, so-called “Enhancers of Trithorax and Polycomb” (ETP), interacts with either complexes, behaving sometimes as repressors and sometimes as activators. The role of ETP proteins is largely unknown. Methodology/Principal Findings In a two-hybrid screen, we identified Cyclin G (CycG) as a partner of the Drosophila ETP Corto. Inactivation of CycG by RNA interference highlights its essential role during development. We show here that Corto and CycG directly interact and bind to each other in embryos and S2 cells. Moreover, CycG is targeted to polytene chromosomes where it co-localizes at multiple sites with Corto and with the PcG factor Polyhomeotic (PH). We observed that corto is involved in maintaining Abd-B repression outside its normal expression domain in embryos. This could be achieved by association between Corto and CycG since both proteins bind the regulatory element iab-7 PRE and the promoter of the Abd-B gene. Conclusions/Significance Our results suggest that CycG could regulate the activity of Corto at chromatin and thus be involved in changing Corto from an Enhancer of TrxG into an Enhancer of PcG.


Introduction
In Drosophila, the Bithorax-complex (BX-C) contains the three Hox genes, Ultrabithorax (Ubx), abdominal-A (abd-A) and Abdominal-B (Abd-B), that specify the identities of the third thoracic segment (T3) and the eight abdominal segments (A1 to A8) [1]. These genes are expressed in spatially regulated patterns during embryonic development thanks to maternal, gap and pair-rule proteins. Their large cis-regulatory sequences are modular and allow parasegmental regulation. These sequences contain different classes of elements such as initiation elements that respond to early segmentation gene products, insulators and promoter targeting sequences (reviewed in [2]).
Hox expression is maintained in the original pattern during later stages of development by the Polycomb-group (PcG) and trithorax-group (trxG) genes. In mutants of PcG or trxG genes, Hox patterns are established correctly but are not maintained. PcG proteins keep Hox genes silenced whereas TrxG proteins keep Hox genes activated thus counteracting PcG action [3,4]. PcG and TrxG proteins are required for the maintenance of many gene expression patterns [5]. These maintenance proteins form heteromultimeric complexes that bind to chromatin and alter its structure. Current models propose that PcG complexes lead to compact, transcriptionally inactive chromatin, whereas TrxG complexes maintain chromatin in an open conformation that facilitates transcription. In Drosophila, several PcG and TrxG complexes have been purified so far: the Polycomb Repressive Complex 1 (PRC1), the Polycomb Repressive Complex 2 (PRC2), the PhoRC complex, the Pcl-PRC2 complex, the Trithorax Activating Complex 1 (TAC1) and the Brahma Complex (BRM) also called SWI/SNF complex. They are extremely large complexes that contain several proteins including chromatin modifying enzymes such as histone methyl-transferases, acetyltransferases or deacetylases [5][6][7][8].
Although most PcG mutations suppress trxG mutations and vice versa, a large screen to identify modifiers of the trxG gene ash1 allowed isolation of enhancers that were previously identified as PcG [E(z), E(Pc), Asx, Scm, Psc and Su(z)2] [9]. These genes were then called Enhancers of Trithorax and Polycomb (ETPs). Further molecular data showed that some ETPs encode members of PRC complexes, such as E(Z), PSC or SCM, while some do not. Recently, Grimaud et al. proposed to reclassify these maintenance proteins, the label PcG being kept for members of PRC silencing complexes and the label TrxG for members of complexes that counteract PcG-mediated silencing [10]. A third class of proteins would be represented by PcG/TrxG DNA-binding recruiters or specific co-factors. We will keep here the term ETP for those maintenance proteins that play a dual role in PcG and TrxG functions without belonging to any PcG or TrxG complexes identified so far. The GAGA factor, Gaf, encoded by Trithorax-like (Trl), falls into this category. Indeed, it was first described as an activator of Hox genes, and later shown to play a role in the recruitment of PcG complexes without co-purifying with any PRC silencing complexes [11,12]. The HMG protein DSP1 also meets the criteria to be an ETP: dsp1 mutants exhibit Hox gene loss-offunction phenotypes but DSP1 is also important for PcG recruitment to chromatin [13,14]. We have previously shown that corto behaves genetically as an ETP. corto mutants present PcG as well as trxG phenotypes and enhance the phenotypes of some PcG, trxG and ETP mutants [15,16]. Corto directly interacts with Gaf and DSP1 suggesting that ETPs are involved in collaborative processes [16,17].
PcG, TrxG and ETP proteins bind DNA sequences called PRE/TRE that carry the information for the active or silent state of the gene they control (reviewed in [18]). Some PRE/TRE have been shown to maintain this transcriptional state throughout cellular divisions in absence of the initial activator or repressor [19,20]. Despite massive efforts towards identification of PcG complex targets at genome scale [21][22][23], the mechanism by which the active or inactive state of PRE/TRE is conserved throughout several cell cycles remains still largely unknown. Many PcG and ETP mutants [Asx, corto, E(z), Pc, ph, Psc, Su(z)2, Trl] exhibit proliferation defects as well as chromosome condensation and segregation defects. This suggests that maintenance proteins play a general role in cell cycle control [24][25][26][27][28]. An attractive hypothesis is that ETPs are critical to maintain the correct association of PcG or TrxG complexes with chromatin during the cell cycle.
In a two-hybrid screen using Corto as bait, we isolated Cyclin G (CycG), the Drosophila homologue of the mammalian Cyclin G1 and G2 (CycG1, CycG2). Vertebrate CycG1 is a transcriptional target of the tumor suppressor p53 [29,30]. It is possibly involved in cell proliferation as it is overexpressed in certain cancer cells [31,32]. However, CycG1 induces G2/M arrest and cell death in response to DNA damage [33][34][35]. Vertebrate CycG2 acts as a negative regulator of cell cycle, as shown by its high level in cells in which G1/S arrest has been induced by growth inhibitory signals [36,37].
Here, we address the interactions between Corto and CycG both in vitro and in vivo. We show that CycG plays an essential role during development. Moreover, we show that CycG is targeted to many sites on polytene chromosomes where it co-localizes partially with Corto and with the PcG factor PH. As an ETP, corto maintains Abd-B repression in embryos. This could be achieved by association between Corto and CycG since both proteins bind to Abd-B regulatory elements, including the iab-7 PRE and the promoter.

Drosophila Cyclin G interacts with Corto
To further investigate Corto function, we performed a twohybrid screen for potential Corto partners. As bait, we used the amino-terminal half of Corto containing a chromodomain [16]. A positive clone spanning almost the full-length CG11525 cDNA (positions 43 to 2263; Accession number NM 079870) encoding Cyclin G (CycG) was isolated. Subsequent two-hybrid assays showed that the chromodomain was not sufficient for interaction with CycG, and that CycG did not interact with the C-terminal half of Corto ( figure 1A).
Then, we performed GST pull-down assays. In vitro translated CycG protein was retained on GST-Corto beads containing the full-length protein and on GST-C1/324 beads containing the amino-terminal half of Corto, but not on the other GST-Corto fusion proteins tested ( figure 1B). Reciprocally, in vitro translated Corto protein was retained on GST-CycG beads which contained the full-length CycG protein. Corto was not retained on GST-CycG-215/566 which contains the cyclin domain but was bound by the N-terminal CycG-region ( figure 1C). These results corroborate the two-hybrid results and indicate that the aminoterminal half of Corto interacts with the amino-terminal end of CycG.

CycG is an essential gene in flies
The cyclin domain of Drosophila CycG is highly similar to the cyclin domains of vertebrate CycG1 and CycG2 (42% and 46% identity, respectively; figure 2A). In agreement with genome annotations, Northern blot analysis revealed 5 different mRNAs ranging between 2.0 and 3.5 kb. All transcripts were found throughout development although notably less abundant in third instar larvae (figure 2B). Antibodies were raised against the Nterminal part of CycG. Two isoforms of 68 kDa and 60 kDa were revealed in total embryonic extracts whereas only the 68 kDa species was found in chromatin extracts (figure 2C). Three translation start sites are predicted in CycG using ATGpr software [38], resulting in putative proteins with molecular weights of 63, 50 and 30 kDa, respectively (figure 2A). Our antisera do not allow testing for the presumptive 30 kDa isoform. However, the two isoforms we detect probably correspond to the two larger predicted proteins.
Since no mutant of CG11525/CycG was available, we designed a P{UAS::dsCycG} construct to inactivate the gene by RNAinterference (RNAi) as described previously [39]. Tissue specific RNAi using various Gal4 driver lines resulted in a considerable downregulation of CycG activity as visualized by reduction of CycG mRNA and CycG protein levels (figure 3), respectively. Ubiquitous downregulation of CycG (da::Gal4; UAS::dsCycG or Act::Gal4; UAS::dsCycG) animals led to lethality of late third instar larvae or pharates. However, the percentage of dead animals varied depending on the transgenic line: lethality of da::Gal4.UAS::ds-CycG2 was estimated to 31% in females and 56% in males whereas no da::Gal4.UAS::dsCycG3 or Act::Gal4.UAS::dsCycG3 adults were obtained (table 1). Lethality was complete in Act::Gal4.UAS::ds-CycG2 males and reached 86% in Act::Gal4.UAS::dsCycG2 females. We observed that males of this genotype never undergo metamorphosis and stop their development as third instar larvae, dying after a few days, whereas most females die as late pharates. Thus, these two lines interfere with CycG activity to different degrees, indicative of partial inactivation. Overexpression of CycG with an ubiquitous driver (da::Gal4.UAS::CycG or Act::Ga-l4.UAS::CycG) suppressed the lethality induced by UAS::dsCycG3 suggesting that it was linked to specific inactivation of CycG and not to Off-target effects (table 1). Taken together, these results show that CycG plays an essential role during development.

Corto and CycG interact in vivo and co-localize at multiple sites on polytene chromosomes
In vivo physical interactions between Corto and CycG were first analysed by co-immunoprecipitation of total embryonic protein extracts ( figure 4A). Only the 68 kDa and not the 60 kDa CycG isoform co-immunoprecipitated with Corto. We cannot exclude that the latter may be hidden by co-migrating IgG. To confirm this result, we co-transfected Schneider S2 cells with pAct::Corto-Flag and pAct::Myc-CycG and probed for interaction using anti-Flag and anti-Myc antibodies. Myc-CycG co-immunoprecipitated with Corto-Flag and conversely (figure 4B-C).
Corto was previously shown to bind polytene chromosomes at multiple discrete loci suggesting that it might participate in the regulation of many genes [17]. Using CycG antisera, we first showed that CycG is ubiquitously expressed in embryos and larvae (not shown). To test whether interactions between Corto and CycG could take place on chromatin, we explored the binding of CycG to polytene chromosomes (figure 5A). We detected CycG at multiple discrete sites. 30 to 40% of these sites overlapped with Corto binding sites suggesting that Corto and CycG could indeed interact on chromatin.
We next asked whether Corto was essential for CycG recruitment to chromatin, and analysed CycG fixation on polytene chromosomes derived from corto 07128 /Df(3R)6-7 larvae. On the whole, no modification of the CycG binding pattern was observed (data not shown) suggesting that CycG recruitment does not depend on Corto. We have previously shown that Corto shares many sites on polytene chromosomes with PcG proteins ( [16,17]  and figure 5C). We observed that CycG also shares many sites with PH (figure 5B). Moreover, some sites were simultaneously occupied by Corto, CycG and PH, suggesting that the interaction between Corto and CycG could be related to PcG function (figure 5D, E).

corto participates in the regulation of Abd-B expression in embryos
Previous findings indicate that corto is involved in the regulation of Hox genes such as Scr or Ubx in larvae [15,17]. Interestingly, in corto germinal clone embryos, Ubx was strongly down-regulated in   figure 6A). In wild-type embryos, Abd-B expression domain extends from PS 10 to 13 [40,41]. In light of the posterior prevalence phenomenon, Ubx down-regulation could be due to up-regulation of Abd-B in the same segments. Indeed, in corto germinal clone embryos, the expression of Abd-B not only increased in the normal Abd-B expression domain but also extended more anteriorly than PS10 ( figure 6B). Moreover, discontinuous ectopic expression of Abd-B occurred in some anterior parasegments indicative of homeotic transformation towards more posterior identities. These results suggest that corto participates in maintenance of Abd-B repression in embryos.

Corto and Cyclin G bind the iab-7 PRE and the promoter of Abd-B
Co-localization of Corto and CycG proteins on polytene chromosomes raises the possibility that CycG and Corto belong to a complex that regulates Abd-B expression. To address this possibility, we investigated whether both proteins bind to Abd-B cisregulatory sequences in embryos. We performed immunoprecipitation on formaldehyde cross-linked chromatin (XChIP) from 0-14h embryos. The co-immunoprecipitated DNA was amplified using primer pairs corresponding either to the promoter region of Abd-B (generating fragment p10), to the iab-7 PRE (generating fragments p9, p8 and p7) or to rp49 as a negative control. We found that Corto and CycG were both present on the promoter and on the iab-7 PRE (figure 7). As previously described for Polycomb (Pc), we observed prominent binding for the p9 fragment that contains the minimal PRE [42,43]. These results strongly suggest that Corto and CycG directly regulate Abd-B expression.

Discussion
We have identified Cyclin G as a new binding partner of the ETP Corto in Drosophila melanogaster. CycG inactivation leads to lethality showing that this gene is essential in flies. Mammalian genomes encode two G-type cyclins, CycG1 and CycG2, the first one being mainly nuclear whereas the second is mainly cytoplasmic [46]. Drosophila has a single homologue, however, it produces at least two different protein isoforms, only the larger being associated with chromatin. These isoforms could combine CycG1 and CycG2 functions. In Drosophila, large scale two-hybrid screens suggested binding of CycG to various Cyclin-Dependent Kinases (CDK) (Cdc2 and Cdk4) [47,48]. Corto and CycG interact in vitro as well as in vivo and form a complex in embryos and presumably also on chromatin. Moreover, Corto interacts with the amino-terminal domain of CycG, which is compatible with the simultaneous binding of CDK and cell-cycle control function of CycG.
Requirement of PcG, trxG and ETP genes in cell-cycle control has already been shown in Drosophila [49,50]. Interestingly, PcG and trxG genes are also involved in self-renewal and proliferation of hematopoietic stem cells in vertebrates [51,52]. One way they might control cell proliferation is by an epigenetic regulation of genes involved in cell cycle and cell proliferation. Indeed, homologues of Drosophila E(z) and Brm participate in the transcriptional regulation of Cyclin A and E in vertebrates, and in Drosophila, Cyclin A is a PcG target [53][54][55]. Alternatively, PcG, TrxG or ETP proteins may interact directly with cell cycle   regulatory proteins. Indeed, it has been shown that Brm interacts with Cyclin E, that Mel-18, a human homologue of Posterior Sex Combs, interacts with Cyclin D2 possibly blocking its interaction with Cdks [56,57] and we show here that the ETP Corto interacts with CycG. These interactions reveal a potential role for these maintenance proteins in regulating the cell cycle independently of transcriptional regulation. This could be a widespread mechanism by which PcG, TrxG and ETP coordinate the chromatin activity status. CycG and Corto co-localize on many sites on polytene chromosomes suggesting that they may have regulated associations. Our data show that Corto represses Abd-B in embryos and although we were not able to test the role of CycG in regulating Abd-B expression in embryos, we observed that both Corto and CycG bind the iab-7 PRE and the promoter of Abd-B suggesting that they could cooperate in this function. Nevertheless, neither Corto nor CycG were detected on the BX-C locus in salivary glands suggesting that they regulate Abd-B in a tissue-specific manner. The role of the CycG-Corto interaction needs to be further investigated. CycG could regulate Corto activity directly on chromatin by recruiting other factors like kinases or phosphatases thus modifying the phosphorylation status of Corto itself, of histones or other proteins at PRE/TRE and promoters. It has been shown that binding of the PcG protein Bmi1 to chromatin correlates with its phosphorylation status [58,59]. It will be interesting to investigate whether Corto and CycG bind the iab-7 PRE and promoter of Abd-B simultaneously, to examine their phosphorylation status when bound to chromatin, and to determine if their presence correlates with Abd-B transcriptional activity. One interesting possibility would be that CycG is involved in changing Corto from an Enhancer of TrxG into an Enhancer of PcG.

Drosophila strains and genetics
Details on Drosophila strains can be found in Flybase [60]. corto 420 and corto 07128 are strong hypomorphic alleles; corto 420 /TM6B was a gift from Roland Rosset [27]. corto-deficient germline clones were obtained as previously described [15]. Other strains were obtained from either Bloomington or Kyoto Drosophila stock centers. CycG transgenic lines were established by standard P-element mediated transformation.
P{UAS::CycG} was constructed by cloning the entire cDNA as EcoRI/XhoI fragment into pUAST [62]. UAS::dsCycG was constructed as outlined before [39]. About 600 nucleotide coding sequence (codons 72 to 268) was chosen to prepare the RNAiconstruct. This segment shows only limited identity to other Drosophila genes and none of them conforms to an optimal siRNA. Two possible ''Off-Targets'' were found, LvpL encoding a larval protein with a predicted role in glucose metabolism and CG15639 encoding an unknown product. In both cases, 21 nucleotides are identical with a GC content of 55-58% instead of the optimal 43-53% [63]. Cloning details are available upon request.

Antibodies
Corto and Polyhomeotic (PH) antibodies were used as described previously [17]. Antibodies against Abd-B (clone 1A2E9) were obtained from the Developmental Studies Hybridoma Bank. Polyclonal antibodies against CycG were raised against the Nterminal 276 amino-acids of CycG fused to maltose binding protein in rabbit, rat and guinea pig. Their specificity was checked on CycG protein generated by in vitro transcription/translation. Monoclonal anti-Flag M2 and anti-HA was from Sigma (F-3165, H-3663) and anti-Myc clone 9E10 from Santa-Cruz Biotechnology.
For co-immunoprecipitations, 1 g of 0-14 h w 1118 embryos were homogenized in RIPA buffer (50 mM Tris pH 7.5, 150 mM NaCl, 0.5% NP40, 0.1% SDS, 1 mM PMSF) with protease inhibitors (Roche Diagnostics). 500 ml of total extract (about 1 mg) were pre-cleared with protein A plus protein G agarose beads (for polyclonal antibody IP) or protein G beads (for monoclonal anti-Myc, anti-Flag or anti-HA). Input was 20 ml of this mixture. Incubation was with 10 ml of either rabbit pre-immune or Corto antiserum, mouse anti-Flag, anti-Myc or anti-HA overnight at 4uC. The appropriate beads were added and further incubated for  [44]. Corto or CycG XChIP was performed using total chromatin from 0-14 h embryos. Fragments around 500 bp from the iab-7 PRE (p9, p8) and overlapping the iab-7 PRE (p7) or the Abd-B B promoter (p10) were amplified with specific primers using the immuno-precipitated DNA [42,45]. A fragment from the rp49 gene was used as a negative control. The 31 st , 33 rd and 35 th PCR cycle samples are shown. Relative enrichment was estimated for the 33 rd PCR cycle sample from the ratio between Corto or CycG immunoprecipitations and mock signals from three independent experiments. The input track shows amplification of DNA from total chromatin with the same primers (Mock: rabbit preimmune serum, Corto: rabbit anti-Corto, CycG-R: rabbit anti-CycG, CycG-GP: guinea-pig anti-CycG). The iab-6, iab-7 and iab-8 cis-regulatory domains are indicated. doi:10.1371/journal.pone.0001658.g007 2 h at 4uC. The supernatant was kept; the beads were washed five times with RIPA buffer and finally resuspended in 40 ml of Laemmli buffer. 20 ml of input (4%), 20 ml of supernatant and half of the beads (20 ml) were loaded. Immunoprecipitates were detected with respective antisera developed in rat. Drosophila S2 cells were cultivated at 25uC in Schneider medium supplemented with 10% fetal calf serum and antibiotics. Cells were transfected using EffectenH transfection reagent according to the manufacturer (Qiagen). Commonly, 2610 6 cells were transfected with 1 mg of each DNA. Cells were collected after 48 h of incubation and homogenized in 500 ml of RIPA buffer.