Wapl antagonizes cohesin binding and promotes Polycomb-group silencing in Drosophila

Control of gene expression is vital to normal development. During Drosophila development, spatially and temporally restricted DNA-binding transcriptional activators or repressors establish early patterns of gene expression. Later, two groups of genes, the trithorax group (trxG) and the Polycomb group (PcG) are responsible for maintaining gene expression throughout development (Kennison, 1995). The PcG genes encode a group of transcriptional repressors that act in protein complexes to modify chromatin (reviewed by Simon and Kingston, 2009; Kerppola, 2009; Müller and Verrijzer, 2009). One hallmark of PcG function is the histone modification, H3K27me3, specified by the PcG protein complex PRC2. The trxG genes were identified by their ability to counteract the action of PcG genes (Kennison, 1995). Most trxG genes encode subunits of complexes that activate transcription through chromatin modification or remodeling. However, important for this study, the trxG gene verthandi (vtd) encodes the cohesin subunit Rad21 (Hallson et al., 2008).

Cohesin is made up of the proteins Smc1, Smc3, Rad21 and Stromalin (SA), and is important for sister chromatid cohesion and proper chromosome segregation during mitosis. In addition, cohesin and cohesin-associated proteins play an important role in regulating gene expression (reviewed by Peters et al., 2008; Xiong and Gerton, 2010; Dorsett, 2011). In a recent study, the cohesin subunits Smc1, Smc3 and Rad21 were found to co-purify with the PcG protein Polycomb (Strübbe et al., 2011), suggesting these protein complexes may physically interact at some loci.

In Drosophila, PcG proteins act through DNA elements called Polycomb group response elements (PREs) (reviewed by Müller and Kassis, 2006; Ringrose and Paro, 2007). In genome-wide studies, PREs are recognized as DNA fragments that are binding sites for multiple PcG proteins. The function of PREs can be tested in reporter assays in transgenic flies. One hallmark of PRE activity is their ability to repress the expression of the mini-white reporter gene, a commonly used marker for transgenic flies. The amount of repression is dependent on the number of PREs: more PREs equals more repression. Because PcG protein complexes interact with each other, this increase in repression is most likely due to an increase in the number or composition of PcG protein complexes that cooperate to repress or silence transcription. Homologous chromosomes are paired in Drosophila, and regulatory DNA on one chromosome can influence the expression of a gene on the homologous chromosome. Thus, PcG complexes bound to PRE-mini-white transgenes inserted near each other on homologous chromosomes can interact to silence the expression of the mini-white gene. This phenomenon is called pairing-sensitive repression or silencing (Kassis, 1994). Because pairing-sensitive silencing is mediated by PREs, it is reasonable to assume it depends on PcG function. In fact, mutations in some PcG genes suppress pairing-sensitive silencing (reviewed by Kassis, 2002).

In an effort to identify genes that are important for PcG silencing, particularly those that mediate interactions between PREs, we conducted a screen for dominant suppressors of pairing-sensitive silencing mediated by an engrailed PRE (Noyes et al., 2011). Here, we identify one of the suppressors we obtained in that screen as a mutation in the wapl gene, an unusual allele we named wapl^AG. Wapl is a cohesin-associated protein important for removing cohesin from chromosomes (Gandhi et al., 2006; Kueng et al., 2006; Shintomi and Hirano, 2009; Sutani et al., 2009; Gause et al., 2010). wapl^AG produces a truncated Wapl protein that acts in a dominant fashion to suppress pairing-sensitive silencing. Furthermore, wapl^AG pharate adults have an extra sex combs phenotype characteristic of mutations in PcG genes. Production of the truncated Wapl^AG protein in an otherwise wild-type background recapitulates the extra sex combs phenotype seen in the wapl^AG mutant and increases the stability of cohesin binding to polytene chromosomes. Our data suggest that increasing cohesin stability interferes with PcG silencing at genes that are co-regulated by cohesin and PcG proteins.

Rabbit polyclonal antibodies (Covance) were raised against HIS-tag Wapl polypeptides that were purified with a HIS-affinity column (Enzymax). Polypeptide specific for Wapl-L was made from amino acids 1 to 300 and was used as the antigen to generate the α-Wapl-L antibody. The α-Wapl-SL antibody was generated against a polypeptide for amino acids 650 to 947 of Wapl-L (amino acids 1 to 300 of Wapl-S). Antibodies were affinity purified with the same polypeptide used for antibody production (Enzymax). For embryo immunoperoxidase staining, a 1:2500 dilution of crude anti-sera was used for both α-Wapl-SL and α-Wapl-L, according to standard procedures (Kwon et al., 2009). Similar results were obtained when using affinity-purified anti-sera (at a 1:50 dilution). Staining of polytene chromosomes was carried out using standard procedures (Eissenberg, 2006) with the following primary antibody dilutions: affinity-purified α-Wapl-L and α-Wapl-SL (1:20); α-Rad21 (1:100) and α-HA (1:800).

Embryos overexpressing Wapl-L and Wapl-S were collected from UAS-Wapl-L and UAS-Wapl-S crossed to en-Gal4. For larval samples, ten 3rd instar larvae were cut in half and the anterior region was inversed to remove its trachea, salivary gland and gut. The remaining tissues were homogenized in lysis buffer (1× PBS, 1% Triton X-100, 1 mM MgCl, 5 mg/ml RNase A, 5 mg/ml DNase, 1× Halt Protease Inhibitor (Thermo Scientific), 5 mM EDTA). The homogenate was centrifuged at 15,000 g for 10 minutes at 4°C. The supernatant was transferred to a new tube and mixed with equal volume of 2× SDS sample buffer, boiled for 5 minutes, then chilled on ice prior to loading on gel. Proteins were separated in 4-15% Mini-Protean TGX gels (Biorad). The ECL Plus Western Blotting Detection System (GE Healthcare) was used. Antibody dilutions were 1:1000 affinity-purified α-Wapl-L and α-Wapl-SL.

A partial wapl cDNA (LD16461) that contains the 5′ end of the large wapl cDNA to a XhoI site was obtained from Drosophila Genomic Resource Center. The 5′-end fragment of the Wapl-S cDNA was generated by PCR with LD16461 as template and two primers (GTCAGTCGGCCGATGATGATGACCATCGACGGAGC and GTCAGTCGGCCGCTCGAGGAAGGCGTCCGTGT). This PCR product was ligated into pCRII-Blunt-TOPO vector (Invitrogen) then transferred from the TOPO vector into pUAST vector by EagI digestion, resulting in pUAST-Wapl-S-5′. To make the wapl 3′-end cDNA fragment, PCR was carried out on genomic DNA from the iso-1 Drosophila strain with two primers (CAAGGTACTGATCAACCTGACG and ACGTACTCGAGGTAAACGTACACATATA). The PCR fragment was ligated into pCRII-Blunt-TOPO vector. This plasmid was cut with XhoI to obtain the wapl 3′-end fragment, which was ligated into XhoI-digested pUAST-Wapl-S-5′. Recombinant clones with the right orientation of the wapl 3′-end fragment were identified by EcoRI digestion. The final construct, pUAST-Wapl-S, was verified by DNA sequencing.

PCR was carried out with LD16461 as template and two primers (GTCAGTAGATCTATGTCGCGCTGGGGCAAGAA and a 3′ primer downstream of the BglII site). This PCR fragment was ligated into pCRII-Blunt-TOPO. The wapl-L-5′ fragment was released from this clone by BglII digestion and ligated into BglII-digested pUAST-Wapl-S backbone. The right clone was identified by PCR and DNA digestion with PstI. The final plasmid pUAST-Wapl-L was sequenced to confirm its identity.

The wapl^AG cDNA was generated by PCR with pUAST-Wapl-L as template. The two PCR primers are CAGAGATCTATGTCGCGCTGGGGCAAGA (with an added BglII site), and CATCTCGAGTTAGCGCCCCGCTTTGGGA (which has an added XhoI site and introduces the wapl^AG stop codon). The PCR product was digested with BglII and XhoI and ligated into the BglII and XhoI cut pUAST vector. Recombinant clones were selected and the wapl^AG cDNA insert was sequenced.

The wapl^AG cDNA was generated by PCR with primers CACCATGTCGCGCTGGGGCAAGA and GCGCCCCGCTTTGGGAATG using pUAST-Wapl-L as template. The PCR product was ligated into pENTR-D-TOPO vector (Invitrogen). Plasmid from the recombinant clones was then transferred into the pTWH vector (developed by T. Murphy's lab, Baltimore, MD; http://emb.carnegiescience.edu/labs/murphy/Gateway%20vectors.html) with Gateway LR Clonase Enzyme Mix (Invitrogen). The final construct was named pTWH-wapl^AG and its wapl^AG cDNA insert was verified by DNA sequencing. Transgenic lines were made using standard techniques (Genetics Services and BestGene).

FRAP was conducted with late 3rd instar salivary glands of the indicated genotypes as previously described using an EGFP-Smc1 expressing transgene (Gause et al., 2010). This transgene uses the Chip gene promoter to express EGFP-Smc1 at lower levels than endogenous Smc1. EGFP-Smc1 rescues Smc1 null mutants to pharate adulthood and is incorporated into cohesin (Gause et al., 2010).

y wapl^AG w UAS-Wapl-S/FM0 virgin females were crossed to arm-Gal4 homozygous males. Progeny were 14 males of the genotype y wapl^AG w UAS-Wapl-S; arm-Gal4/+; 50 FM0; arm-Gal4/+ males; 51 FM0/+; arm-Gal4/+ females; and 49 y wapl^AG w UAS-Wapl-S/+; arm-Gal4/+ females (14 rescued males out of 50 expected). We did not obtain any rescued males from a similar cross of y wapl^AG w UAS-Wapl-L/FM0 virgin females to arm-Gal4 males (about 300 progeny scored). Because the UAS-Wapl-L was inserted in a different chromosomal location than UAS-Wapl-S, we do not know whether the lack of rescue with UAS-Wapl-L is the result of either too much or too little Wapl-L protein made from this insertion site, or from a difference in the activity of the two proteins. Similar crosses with wapl² showed that neither UAS-Wapl-S nor UAS-Wapl-L could rescue this null mutant.

To generate germline clones, y wapl^AG w P{FRT(w^hs)}101/FM0 or y¹ w v²⁴ P{FRT(w^hs)}101 virgin females were crossed to w ovo^D1 v²⁴ P{FRT(w^hs)}101/Y; P{hsFLP}38 males. After 1 day, parents were removed. The larvae were aged for 1 day and then heat-shocked at 37°C for 1 hour on two consecutive days (Chou and Perrimon, 1996). Virgin females of the genotypes y wapl^AG w P{FRT(w^hs)}101/w ovo^D1 v²⁴ P{FRT(w^hs)}101 and y¹ w v²⁴ P{FRT(w^hs)}101 /wovo^D1 v²⁴ P{FRT(w^hs)}101 were crossed to C(4)RM, ci ey^R/0. Non-disjunction and chromosome loss were detected as y+ sons (X chromosome non-disjunction and loss) and ci ey progeny (4th chromosome non-disjunction and loss).

We conducted a screen for dominant suppressors of pairing-sensitive silencing by a 181 bp engrailed PRE and recovered nine mutants (Noyes et al., 2011). One of the dominant suppressors mapped to the X chromosome and was hemizygous and homozygous lethal. We mapped the mutation between y and w by recombination mapping. We then used small deficiencies and gene-specific mutations to determine that the mutation was in the wapl gene, and named the allele wapl^AG. The wapl gene encodes two transcripts, a large transcript and a small transcript, generating a large and small protein – Wapl-L and Wapl-S (Fig. 1). The transcription start site for the small transcript is within an intron of the large transcript (McQuilton et al., 2012). The large transcript codes for a protein of 1741 amino acids; wapl^AG is a nonsense mutation at Gln²⁷¹ (CAA to TAA) of the large protein. This suggests that it truncates Wapl-L protein but does not interfere with the production of Wapl-S. In addition, we determined the molecular lesion of wapl², the null mutation we used in our experiments. wapl² was reported to contain an insertion of about 1.6 kb of DNA into the middle of the wapl gene, eliminating the correct production of both the large and small wapl mRNA transcripts (Vernì et al., 2000). We determined that wapl² contains a 970 bp insertion of DNA from the epidermal growth factor receptor gene into an exon common to both the large and small proteins, and eliminates the production of both proteins (see below).

The sequence similarity between Drosophila Wapl and human, C. elegans and yeast Wapl proteins is contained within the C-terminal region of the Drosophila proteins, extending from amino acids 488 to 1018 of the short protein and amino acid 1137 to 1667 of the long protein (Kueng et al., 2006). We examined the sequence conservation of the 649 amino acids unique to the long protein. The entire 649 amino acids are well conserved in the genus Drosophila (data not shown). In addition, ~100 amino acids at the N terminus are conserved in all insects (data not shown) but not in other organisms. This suggests that this part of the protein has an important function in insects.

wapl^AG suppresses pairing-sensitive silencing of mini-white by a 181 bp PRE (Fig. 2). wapl^AG w/FM0; P[181PRE]8-10C/P[181PRE]8-10C flies have yellow eyes (similar to what was seen with our other suppressors), whereas w; P[181PRE]8-10C/P[181PRE]8-10C have white eyes. Interestingly, wapl² w/FM0; P[181PRE]8-10C/P[181PRE]8-10C also have white eyes, showing that loss of one copy of wapl is not sufficient to generate the phenotype observed (data not shown). This shows that wapl^AG is a gain-of-function mutation. wapl^AG, unlike the other dominant suppressors of pairing-sensitive silencing that we isolated (Noyes et al., 2011), has no effect on the heterozygous eye color of P[181PRE]8-10C. This suggests that wapl^AG may interfere with the interaction between PREs, which enhances the silencing activity of these elements.

Given the molecular nature of the wapl^AG mutation, we reasoned that the gain-of-function phenotype could be due to either the production of a truncated protein, or to a disruption in the ratio of Wapl-S versus Wapl-L protein. We therefore generated antibodies to Wapl to test whether wapl^AG could generate a stable protein.

We generated rabbit antibodies against two antigens: one specific for the large protein (α-Wapl-L) and one that will detect both the small and the large proteins (α-Wapl-SL) (Fig. 1). Western blots show the specificity of α-Wapl-L for the large protein, whereas α-Wapl-SL detects two bands (Fig. 3). Both Wapl proteins are present in embryos at similar levels, whereas the predominant form in larvae is Wapl-L (Fig. 3B,C). This is consistent with RNA-seq data that shows that the Wapl small transcript is expressed at a very low level in larvae (McQuilton et al., 2012). The western blots also show the complete absence of both Wapl proteins in wapl² larvae, and the absence of Wapl-L in wapl^AG larvae (Fig. 3C). Both antibodies detect nuclear antigens in embryos (Fig. 3D-G).

Interestingly, α-Wapl-L detects a protein of about 30 kDa in wapl^AG embryos (Fig. 3A) and larvae (supplementary material Fig. S1), but not in wild-type embryos. The expected molecular weight for the truncated mutant protein is 28 kDa. In order to test whether this protein is also nuclear, we made a UAS-Wapl-AG construct encoding the predicted truncated mutant protein and expressed it under the control of engrailed-GAL4 so that the protein would be expressed in stripes in embryos (Fig. 3H). Two versions of UAS-Wapl-AG were made, one HA-tagged to distinguish the ectopic protein from wild-type Wapl-L and one without a tag. Both Wapl-AG and Wapl-AG-HA are nuclear proteins (Fig. 3H and data not shown).

Wapl plays an important role in removing cohesins from chromosomes for cell division, and regulates cohesin chromosome-binding dynamics during interphase, but its role in gene expression has not been explored. Similar to its interaction partner Pds5, we found that Wapl colocalizes with Rad21 and Nipped-B on polytene chromsomes (Fig. 4 shows Rad21 colocalization). No Wapl protein was seen with either the α-Wapl-L or the α-Wapl-SL antibodies on wapl² polytene chromosomes, confirming the specificity of our antibodies (supplementary material Fig. S2 and data not shown). In addition, we observed that Rad21 and Nipped-B were both bound to polytene chromosomes in the absence of Wapl (data not shown), consistent with in vivo fluorescence recovery after photobleaching (FRAP) experiments (Gause et al., 2010). Wapl-AG-HA binds to the same bands as Rad21, as well as to additional bands (Fig. 4B). We suggest that Wapl-AG could be mediating its phenotypic effects via an interaction with cohesin.

wapl² mutants die at the larval-pupal boundary and have small imaginal discs (Vernì et al., 2000). Although most wapl^AG mutants die during the pupal stage without the development of cuticle, a few make it to pharate adults. We examined the phenotype of wapl^AG hemizygous males and found they have an extra sex combs phenotype that is characteristic of males with mutations in PcG genes (Fig. 5A). We suspected that this phenotype might be due to the truncated Wapl-AG protein. In fact, overexpression of UAS-Wapl-AG or UAS-Wapl-AG-HA by the daughterless-GAL4 driver caused an extra sex combs phenotype (Fig. 5B). In addition, several other phenotypes were observed with overproduction of Wapl-AG or Wapl-AG-HA: extra and abnormal sternopleural bristles (Fig. 5C), a failure in dorsal-ventral fusion in the most posterior abdominal segment (Fig. 5D) and twisted genitalia (Fig. 5B). UAS-Wapl-AG or UAS-Wapl-AG-HA inserted in different chromosomal insertion sites showed similar phenotypes, but to different extents (Table 1). This probably reflects a difference in expression levels of the transgenes. Extra sternopleural bristles and a failure in dorsal-ventral fusion were also observed in some wapl^AG pharate adults (data not shown). Overexpression of Nipped-B greatly reduces viability, but surviving adults display a failure in dorsal-ventral fusion (Fig. 5E) and twisted genitalia (not shown), similar to phenotypes seen by Wapl-AG. Similar to Wapl-AG (see below), overexpression of Nipped-B increases the stability of cohesin binding to polytene chromosomes (Gause et al., 2010).

We examined the effect of mutations in the genes pds5 and Nipped-B on the suppression of pairing-sensitive silencing by wapl^AG and the extra sex combs phenotype seen in UAS-Wapl-AG-HA; da-GAL4 males. Pds5 forms the releasin complex with Wapl to unload cohesin from the chromosome, whereas Nipped-B complexes with the Mau-2 protein to form the kollerin complex that loads cohesin (Nasmyth, 2011). The heterozygous null pds5^e3 mutation further darkened the wapl^AG w/w; P[181PRE]8-10C eye color, whereas Nipped-B²⁹ lightened it (Fig. 6A). Furthermore, heterozygous mutations in Nipped-B suppressed and pds5 enhanced the extra sex combs phenotype seen in UAS-Wapl-AG-HA; da-GAL4 flies (Fig. 6B). It has been shown that heterozygous pds5^e3 increased the amount of stable cohesin binding to salivary chromosomes, and a heterozygous Nipped-B mutation decreased the amount of stable cohesin (Gause et al., 2010). Because Wapl is thought to act with Pds5 to remove cohesin, we predicted that Wapl-AG would increase the stability of cohesin binding. The dynamics of cohesin binding were compared in EGFP-Smc1/+; da-GAL4/+ control and EGFP-Smc1/UAS-Wapl-AG-HA; da-GAL4 salivary glands using FRAP to determine the effects of Wapl-AG on cohesin binding (Fig. 7). As before, we found cohesin in three pools: unbound, weak and strong binding (Gause et al., 2010). Wapl-AG increases the half-life of the stably bound population threefold (from 237 to 772 seconds) without affecting the half-life of the weakly bound population. Furthermore, the amount of stably-bound cohesin increased two-fold in UAS-Wapl-AG-HA; da-GAL4 polytenes, from 13% to 26% of the total cohesin population. This increase in cohesin stability was similar to that seen in the loss-of-function wapl² hemizygotes. We suggest that Wapl-AG interferes with the unloading of cohesins from the chromosome causing an increase in binding stability. A decrease in the amount of Pds5 further increases stable cohesin leading to enhanced phenotypes, whereas a decrease in the amount of Nipped-B reduces cohesin loading and alleviates the wapl^AG phenotypes.

We used UAS-Wapl-S and UAS-Wapl-L in an attempt to rescue the lethality of wapl^AG. We found that driving the expression of UAS-Wapl-S and UAS-Wapl-L by the da-GAL4 driver caused lethality in both wild-type and wapl^AG flies. Therefore, we used the arm-GAL4 driver and observed that UAS-Wapl-S could rescue the lethality of wapl^AG (see Materials and methods for details). Some of the rescued wapl^AG males looked phenotypically normal, showing that Wapl-S could rescue the phenotypes normally seen in wapl^AG males. Our results suggest that both the lethality and the abnormal phenotypes seen in wapl^AG mutants are caused by a decrease in Wapl function and that the truncated Wapl^AG protein interferes with the normal function of Wapl. wapl^AG produces half of the N terminus of Wapl-L (the location of most of the insect-conserved amino acids), and wapl^AG; UAS-Wapl-S; arm-Gal4 rescued flies would have both Wapl-S and Wapl-AG. Therefore, we cannot conclude that the N terminus of Wapl is dispensable for viability in flies. Attempts to rescue wapl^AG by UAS-Wapl-L or wapl² by UAS-Wapl-L and UAS-Wapl-S were not successful (for details see Materials and methods). We suspect this could have been due to difficulties obtaining the correct level of expression of the transgenes.

Embryos that lack maternal Wapl proteins, generated by making germline clones in wapl² heterozygous mothers, arrest before the cellular blastoderm stage with many abnormal nuclei (Perrimon et al., 1985; and data not shown). By contrast, embryos derived from maternal germline clones of wapl^AG can be paternally rescued by a wild-type wapl allele. The wapl^AG zygotic mutants from these maternal germline clones die at the same stage as wapl^AG zygotic mutants from heterozygous mothers. Wapl-S proteins expressed from the wapl^AG mutant chromosome must be sufficient to rescue the early embryonic requirement for wapl function. Females heterozygous for wapl mutations show an increase in non-disjunction (Vernì et al., 2000). Therefore, we examined non-disjunction of the X and 4th chromosome from the germline clones of wapl^AG. Like other wapl mutants, wapl^AG causes an increase in non-disjunction (Table 2). All of our results suggest that wapl^AG reduces the amount of functional Wapl protein, leading to zygotic lethality, extra sex comb teeth in males and non-disjunction in oocytes.

Our results suggest a role for the cohesin-associated protein Wapl in the control of gene expression and regulation of PcG silencing at genes co-regulated by cohesin and PcG proteins. This is supported by three results: Wapl colocalizes with the cohesin subunit Rad21 on polytene chromosome; wapl^AG interferes with the function of PcG-silencing at specific targets; and wapl^AG phenotypes are modified by mutations in Nipped-B and pds5, two cohesin-associated factors.

wapl-null mutants die as third instar larvae and are germline lethal. By contrast, wapl^AG mutants live until the pharate adult stage and germline clones are viable. wapl^AG is a mutation that introduces a stop codon in the Wapl-L transcription unit, leaving the Wapl-S transcription unit intact. Wapl-L is transcribed throughout development, whereas Wapl-S is transcribed at all times except in larvae. We suggest that the lethality of wapl^AG results from a loss in total Wapl levels owing to the absence of expression of Wapl-S in larvae, and that Wapl-S is sufficient for germline development. What about the dominant effect of wapl^AG? We suggest that the N-terminal region of the Wapl-L protein plays a regulatory role that attenuates the activity of Wapl and that the presence of Wapl-AG, coupled with the loss of the Wapl-L C terminus, leads to an attenuation of Wapl activity in wapl^AG mutants. The rescue of both the lethality and the extra-sex combs phenotype of wapl^AG by UAS-Wapl-S is consistent with this hypothesis. Overproduction of Wapl-AG using the Gal4 system would similarly attenuate Wapl activity leading to disrupted gene expression of cohesin-regulated genes.

We do not know how Wapl-AG interferes with Wapl function. The colocalization of Wapl-AG with Rad21 on polytene chromosomes suggests that it might interact with either wild-type Wapl or its interacting partner Pds5. The amino acid sequence FGF is important for the interaction of human Wapl with human Pds5 (Shintomi and Hirano, 2009), but this sequence is not present in either Wapl-S or Wapl-L. Protein structure and domain prediction programs did not find any distinguishing features in the N-terminal sequence of Wapl-L. Comparison of the amino acid sequence of Wapl-AG and Wapl-S showed a number of short stretches of sequence identity, and, curiously, one of our polyclonal α-Wapl-L antibodies, which was not used in these studies, recognized overexpressed Wapl-S protein weakly, suggesting that the antigens used for the two immunizations might share some epitopes. Clearly, more work needs to be done to determine how Wapl-AG disrupts cohesin function.

If Wapl is important for removing cohesin from the chromosomes, why does it colocalize with it on polytene chromosomes? Cohesin binding is dynamic (Gause et al., 2010). Our results show that interfering with Wapl function causes an increase in cohesin-binding stability. Our results suggest that Wapl probably cycles on and off the chromosomes with cohesin. This is consistent with the finding that Pds5, the Wapl interaction partner, has virtually identical chromosome-binding dynamics as cohesin, including a stable-binding form with a long chromosomal residence time (Gause et al., 2010).

Cohesin binding occurs mainly at transcriptionally active genes with paused polymerase, and with rare exceptions, is not detectable at PcG-silenced genes marked by H3K27me3 (Fay et al., 2011; Misulovin et al., 2008). Our data show that Wapl-AG increases both the stability and extent of cohesin binding, and therefore suggest that stronger cohesin binding interferes with the function of PcG proteins at silenced genes. We theorize, therefore, that cohesin may be actively excluded from silenced genes. In the rare cases where cohesin binding overlaps H3K27me3, the genes are not fully silenced (Schaaf et al., 2009). Although this is consistent with the idea that cohesin counteracts silencing, in fact cohesin negatively regulates expression in these unusual cases, in part by blocking transition of paused RNA polymerase to elongation (Fay et al., 2011). The complex interplay between cohesin and PcG proteins revealed by these and prior studies show that cohesin regulates gene expression by additional mechanisms beyond its long-recognized role in facilitating enhancer-promoter interactions (Dorsett, 2011).

It is worth noting that the phenotypes we see with wapl^AG are very specific, and overlap those seen with reduction in the dose of PcG proteins, suggesting that the target genes that are the most sensitive to PcG activity are also the most sensitive to cohesin. The extra sex combs phenotype seen in wapl^AG mutants is the defining feature of PcG group genes (Kennison, 1995). It is caused by the loss of Scr silencing, which is exquisitely sensitive to the level of PcG proteins. Strictly speaking, wapl^AG can be classified as a PcG mutation. Interestingly, mutations in both Nipped-B and vtd, which codes for Rad21, have been classified as trxG genes. In fact, vtd mutations were first isolated as suppressors of the extra sex combs phenotype caused by Polycomb mutations (Kennison and Tamkun, 1988). Therefore, trxG phenotypes can be caused by mutations in cohesin subunits, whereas PcG phenotypes can be caused by mutations that increase cohesin binding. Our data further suggest that increasing cohesin stability can inhibit PcG function by interfering with interactions between PREs. A 181 bp minimal engrailed PRE can repress mini-white expression in a transgene when present as a single copy (in flies heterozygous for the transgene), but it is much more effective when present in two copies (in flies homozygous for the transgene) (Noyes et al., 2011). wapl^AG affects only the eye color of homozygous flies, suggesting that it only acts on paired PREs. Interestingly, control of Scr also depends on interactions between regulatory elements, most probably PREs (Southworth and Kennison, 2002). These data, combined with data showing a biochemical interaction between cohesin and PcG proteins (Strübbe et al., 2011), suggest that cohesin might directly inhibit PcG function.

We thank Payal Ray and Sandip De for comments on the manuscript and the Bloomington Stock Center for fly stocks.