Polycomb recruitment to DNA in vivo by the YY1 REPO domain

Abstract

Polycomb group (PcG) proteins are responsible for maintaining transcriptional repression of developmentally important genes. However, the mechanism of PcG recruitment to specific DNA sequences is poorly understood. Transcription factor YY1 is one of the few PcG proteins with sequence-specific DNA binding activity. We previously showed that YY1 can recruit other PcG proteins to DNA, leading to histone posttranslational modifications and stable transcriptional repression. Using Drosophila transgenic approaches, we identified YY1 sequences 201–226 as necessary and sufficient for PcG transcriptional repression in vivo. When fused to a heterologous DNA-binding domain, this short 26-aa motif was sufficient for transcriptional repression, recruitment of PcG proteins to DNA, and methylation of histone H3 lysine 27. Deletion of this short YY1 motif did not affect transient transcriptional repression but ablated PcG repression, PcG protein recruitment to DNA, and methylation of H3 lysine 27. We propose that this motif be named the REPO domain for its function in recruitment of Polycomb. The REPO domain is well conserved in YY1 orthologs and in related proteins.

Polycomb group (PcG) proteins are responsible for the heritable silencing of target genes in metazoans (1, 2) and are functionally conserved in Drosophila and mammals (2, 3). Mutations in PcG genes result in misexpression of target genes with resulting homeotic transformation of body parts (4). Studies in mammals showed that PcG proteins are important for normal skeletal, muscular, and hematopoietic development (5). A variety of studies support the existence of at least two PcG complexes, each composed of several polypeptides: PRC1, containing Polycomb (Pc), polyhomeotic (Ph), dRing, and posterior sex combs (Psc); and PRC2 containing enhancer of zeste [E(z)], extra sex combs (Esc), and suppressor of zeste 12 [Su(z)12] (1, 2). These complexes are recruited to chromatin, where they maintain transcriptional silencing. An unresolved issue is the mechanism of recruitment of PcG complexes to appropriate target genes.

The mechanism of PcG complex recruitment has remained elusive, especially in mammals. Drosophila Polycomb response elements (PREs) bind to PcG proteins and silence cis-linked genes in vivo (1, 2). The large size and poor sequence homology of characterized PREs has hampered the search for recruiting factors. Furthermore, nearly all characterized PcG proteins lack sequence-specific DNA-binding activity. Exceptions include the Drosophila PcG proteins Pleiohomeotic (PHO) and Pleiohomeotic-like (PHOL) and their mammalian counterpart, the vertebrate transcription factor Yin Yang 1 (YY1) (6, 7). Homology between these transcription factors is localized to the C-terminal zinc-finger DNA-binding domain (YY1 residues 298–414, 95% identical to PHO) and a short internal sequence (YY1 residues 205–226, 82% identical to PHO). We previously demonstrated that, similar to PHO, YY1 can function as a PcG protein to mediate silencing (8). Thus, human YY1 can silence a PcG-sensitive reporter in Drosophila embryos, correct the Drosophila pho mutant phenotype during development, and generate stable transcriptional repression complexes in vivo (8, 9).

PHO and PHOL are strong candidates for recruiting PcG complexes to DNA because many PRE sequences contain PHO/YY1-binding sites (10). Intact PHO-binding sites are necessary for silencing (11–13), and pho mutants enhance other PcG mutant phenotypes indicative of a functional link (14). Based on structural homology between YY1 and PHO, we hypothesized that YY1 might function to recruit PcG proteins to DNA. Indeed, we found that YY1 can recruit PcG proteins to DNA in vivo, leading to histone H3 deacetylation and H3 lysine 27 methylation (9, 15). RNAi knockdown of YY1 in mammals blocks EZH2 DNA binding to the skeletal muscle myosin promoter (16), consistent with our results showing that YY1 can recruit PcG proteins to DNA (15). Together, these observations indicate a PcG-recruiting role for YY1 in mammals.

In this report, we identify the specific YY1 sequences needed for PcG recruitment to DNA and for transcriptional repression in vivo. We found that the short internal YY1/PHO conserved domain (YY1 residues 201–226) is necessary and sufficient for silencing of transcription in vivo, PcG protein recruitment to DNA, and methylation of histone H3 on lysine 27. This domain is well conserved in YY1 orthologs and related proteins. We propose that this domain be named the REPO domain for its function in recruitment of Polycomb. Our studies define a protein domain needed for PcG recruitment to DNA and for subsequent PcG-mediated transcriptional repression.

Results

Identification of a YY1 Domain Necessary and Sufficient for PcG Repression.

We previously showed that YY1 can silence a PcG-dependent reporter gene, BGUZ, that consists of the LacZ gene under control of the ultrabithorax (Ubx) promoter and the BXD enhancer (8). This reporter contains five GAL4 DNA-binding sites to tether GAL-fusion test proteins to the promoter (see Fig. 1). The BGUZ construct becomes transcriptionally active in all cell types at ≈6 hours of embryogenesis, resulting in LacZ expression throughout the embryo (Fig. 1). Under conditions where PcG proteins are nucleated at the BGUZ reporter via fusion with the GAL4 DNA-binding domain, transcription is silenced, resulting in loss of LacZ staining (8, 17–19). When GAL-PcG fusion proteins are expressed from the hunchback promoter, they are delivered as a pulse of expression in the anterior halves of developing embryos, thus resulting in PcG-dependent repression in this portion of the embryo (see Fig. 1). Silencing in this system depends on PcG function because various PcG mutants abrogate repression (8, 17). This assay, therefore, provides a positive/negative score for PcG repression by any test protein. Anterior repression of staining indicates that the test protein mediates PcG repression.

Fig. 1.

Strategy for detecting PcG transcriptional repression by YY1 mutant transgenes. (Top) The BGUZ reporter with the BXD enhancer, GAL4-binding sites, Ubx promoter, and LacZ gene indicated. This reporter gives ubiquitous staining in embryos shown on the right of the figure. (Middle) The general structure of the mutant YY1 effector transgenes driven by the hunchback promoter. The GAL4 DNA-binding domain (residues 1–147) is indicated in red, and the location of the fused YY1 DNA segments is shown in black. These effector transgenes are expressed only in the anterior half of Drosophila embryos (black shading in the embryo at the right). (Bottom) Potential LacZ repression in the anterior end of embryos from crosses between the BGUZ reporter and the GALYY1 transgenes.

As described above, we previously showed that YY1 can mediate PcG repression and recruit PcG proteins to DNA (8, 9, 15). To determine the specific YY1 sequences needed for PcG repression, we prepared a variety of GALYY1 deletion mutants as Drosophila transgenes driven by the hunchback promoter (Fig. 2). We tested the ability of each of these YY1 mutant proteins to repress transcription of the PcG-dependent BGUZ reporter. YY1 amino acid sequences 1–200 contain the transcriptional activation domain (residues 16–99) as well as a transcription repression motif between residues 170–200 (20–22). This domain interacts with histone deacetylase (HDAC) RPD3 (23). However, we found that GALYY1 1–200 failed to silence transcription, indicating that this region is not sufficient for PcG repression (Fig. 2). In contrast, a GAL construct containing YY1 sequences 201–414 (GALYY1 201–414) was able to silence transcription similar to full-length YY1 (Fig. 2). This portion of YY1 contains both YY1 segments with high homology to PHO (YY1 sequences 205–226 and 298–414). The highest homology between mammalian YY1 and Drosophila PHO is within the zinc-finger region (95% overall and 100% within zinc fingers 2 and 3, which directly contact DNA) (7, 24). Previously, we identified a repression domain localized to residues 333–397 (20, 21). This motif overlaps with a region important for association with the nuclear matrix (21, 25). We tested a transgene that deletes nearly all of this repression domain defined by transient transfection assays and which deletes all of zinc fingers 3 and 4 and half of zinc finger 2 (GALYY1 1–340) (20, 21). This construct retained complete PcG repression activity (Fig. 2). Therefore, the repression domain identified within the YY1 zinc-finger domain is not necessary for YY1 PcG repression.

Fig. 2.

YY1 sequences 201–226 are necessary and sufficient for YY1 PcG repression. YY1 sequences in each effector transgene are diagrammed on the right. Embryos stained for LacZ expression from the BGUZ reporter are shown on the left with dorsal and lateral views of stained embryos. The name of each YY1 effector transgene is indicated on the left.

The above studies localized the YY1 sequences needed for PcG repression to between residues 200 and 340. Within this region is the small YY1 motif (residues 205–226) that bears sequence homology to Drosophila PHO (82% identity). This region of PHO was shown by others to interact with some PcG proteins (26–28). Therefore, we prepared a transgenic construct with a small internal deletion that encompassed this short YY1 segment (GALYY1Δ201–226). Interestingly, this YY1 internal deletion abolished YY1 PcG silencing ability (Fig. 2). Therefore, YY1 amino acids 201–226 are necessary for transcriptional silencing. To determine whether this short YY1 segment was sufficient for PcG repression, we tethered YY1 sequences 201–226 to the GAL4 DNA-binding domain. Strikingly, the GALYY1 201–226 transgene resulted in strong repression (Fig. 2). The GAL DNA-binding domain alone has no PcG repression activity (8). These results indicate that when the YY1 201–226 domain is fused to a DNA-binding domain, it is necessary and sufficient for PcG repression.

YY1 Residues 201–226 Do Not Silence a Non-PcG-Sensitive Transgene Reporter.

We wanted to explore the ability of our YY1 transgenes to repress other reporters that are not sensitive to PcG silencing. The GAL-NP6-LacZ transgene reporter is not sensitive to PcG-mediated silencing in Drosophila embryos but can be transiently repressed by either GAL-Pc or GALYY1 if the effector plasmids are induced to be expressed just before NP6 activation at 16 h of embryonic development (8, 17). Earlier effector transgene inductions are incapable of generating stable repression, but if hsp70 promoter-driven GALYY1 or GAL-Pc are induced by heat shock at 15 h of embryonic development, the GAL-NP6-LacZ reporter is transiently silenced (8, 17). Therefore, we prepared several of our GALYY1 mutant constructs as hsp70-driven transgenes.

Various hsp70-driven GALYY1 transgenic lines were crossed to the GAL-NP6-LacZ reporter strain. Embryos were heat shocked at 15 h, returned to 25°C until 18 h, and then processed for LacZ expression. As expected, full-length hsp GALYY1 repressed GAL-NP6-LacZ transcription (Fig. 3A). Similarly, the GALYY1Δ201–226 deletion protein showed potent ability to repress GAL-NP6-LacZ expression (Fig. 3A). Thus, YY1Δ201–226 can repress transcription by a transient mechanism, but not through a PcG-dependent mechanism (Fig. 2). Strikingly, GALYY1 201–226 failed to silence GAL-NP6-LacZ expression (Fig. 3A), indicating that its silencing ability is specific for PcG-responsive promoters.

Fig. 3.

YY1 sequences 201–226 are incapable of transient transcriptional repression. (A) YY1 sequences 201–226 are not needed for transient repression in fly embryos. Effector transgenes (indicated on the left) were crossed with the GAL-NP6-LacZ reporter, heat-shocked at 15 h after laying, and then stained for LacZ expression at 18 h. Repression was observed with full-length GALYY1 and GALYY1Δ201–226, but not with GALYY1 201–226. (B and C) YY1 sequences 201–226 are not needed for transient repression in cell culture. Quantities (in nanograms) are indicated for CMV expression (Exp.) plasmids cotransfected with the (YY1)₄TKCAT reporter in NIH 3T3 cells. Fold activation is plotted (C) with error bars representing the standard deviation from the mean. (D and E) The REPO domain does not support transient activation or repression. NIH 3T3 cells were cotransfected with GALTKCAT and various doses of either CMVGALYY1 or CMVGALYY1 201–226. CAT data in D, are presented in quantitative form in E.

To confirm transient repression results observed with the GAL-NP6-LacZ reporter, we performed transient cotransfection assays in NIH 3T3 cells with a (YY1)₄TKCAT reporter containing multimerized YY1-binding sites and CMV-driven YY1 constructs. We previously showed that titration of the CMVYY1 expression plasmid first results in activation (low doses) and then repression (high doses) of the (YY1)₄TKCAT reporter in NIH 3T3 cells (20, 29). As expected, low amounts of YY1 activated and high amounts repressed (YY1)₄TKCAT expression (Fig. 3B). Similarly, YY1Δ201–226 activated and repressed (YY1)₄TKCAT expression (Fig. 3B). Quantitative comparison of data from three independent experiments indicated minimal differences in transient activity between full-length YY1 and YY1 Δ201–226 (Fig. 3C). Thus, deletion of YY1 sequences 201–226 had no impact on transient repression functions of YY1 (Fig. 3 A–C), but completely ablated PcG repression (Fig. 2). YY1 sequences 201–226, when fused to the GAL4 DNA-binding domain (GALYY1 201–226) were completely incapable of transient activation or repression of a GALTKCAT reporter (Fig. 3 D and E), although it should be noted that repression function by full-length GALYY1 was also not detectable in this assay system. We conclude that YY1 sequences 201–226 are necessary and sufficient for PcG transcriptional repression (provided they are tethered to a DNA-binding domain and targeted to a PcG-sensitive reporter), but are completely incapable of mediating non-PcG-dependent, transient transcriptional repression.

YY1 Sequences 201–226 Are Needed for PcG Recruitment to DNA.

We previously showed that YY1 can recruit PcG proteins to DNA in vivo in Drosophila embryos and larval wing imaginal discs (9, 15). This recruitment results in histone deacetylation of H3 and methylation on H3 lysine 27 (9, 15). Our observation that GALYY1 201–226 can silence transcription at a PcG-responsive promoter strongly suggested that this domain is responsible for YY1 PcG-recruiting activity.

We performed chromatin immunoprecipitation (ChIP) studies with embryos containing the BGUZ reporter and GALYY1Δ201–226 and GALYY1 201–226 fusion proteins under control of the hsp70 promoter (to drive ubiquitous expression after heat shock). Embryos were heat shocked at 2.25 h after the start of the egg lay and fixed for ChIP assay at 6 h. Sonicated chromatin was immunoprecipitated with various antibodies, PCR was performed with primers that flank the GAL4-binding sites in the BGUZ reporter, and products were detected by Southern blotting. Signals were normalized to the GAL antibody immunoprecipitated products (Fig. 4A, lanes 5 and 6) to provide an indication of the amount of transgenic fusion protein binding to DNA in vivo, thus serving as a control for comparable synthesis of functional nuclear protein.

Fig. 4.

YY1 sequences 201–226 are necessary and sufficient for recruitment of PcG proteins to DNA in vivo. (A Top) Schematic of the BGUZ reporter locus indicating the structural elements and the location of the PCR primers for ChIP experiments. (Middle and Bottom) Representative ChIP Southern blot from embryos expressing either the GALYY1Δ201–226, or the GAL 201–226 transgenes. Antibodies are shown above the lanes. Triangles indicate the increase in template concentration used for PCR (2 and 20 ng). (B) Quantitative ChIP data from three independent experiments with full-length GALYY1, GALΔ201–226, and GALYY1 201–226 transgenes. Values were normalized to the level of the GAL4 signal to normalize for equal amounts of GAL fusion protein bound to DNA in vivo. Error bars show the standard deviation from the mean. (C) YY1 recruits HDAC activity, but not H3K27 methyltransferase activity to the GAL-NP6-LacZ reporter. ChIP assays were performed with antibodies shown above the lanes. After PCR of either 2 or 20 ng DNA with primers specific for the GAL-NP6-LacZ reporter, samples were electrophoresed on agarose gels.

Interestingly, GALYY1Δ201–226 failed to recruit the Polycomb protein to DNA, resulting in very weak methylation of H3 lysine 27 (Fig. 4A Upper, lanes 7, 8, 11, and 12). On the contrary, GALYY1 201–226 resulted in efficient Polycomb recruitment to DNA and subsequent high levels of methylation of H3 lysine 27 (Fig. 4A (Lower, lanes 7, 8, 11, and 12). We performed quantitative ChIP studies to compare the level of Pc recruitment and H3 lysine 27 methylation by full-length GALYY1 compared with GALYY1Δ201–226 and GALYY1 201–226 proteins (Fig. 4B). These studies again showed that GALYY1Δ201–226 was incapable of recruiting Pc, whereas the GALYY1 201–226 protein was even more efficient than full-length GALYY1 (Fig. 4B). Similarly, compared with full-length GALYY1, GALYY1 201–226 was more efficient at recruiting H3 lysine 27 methyltransferase activity, and GALYY1Δ201–226 failed to recruit this activity. H3 methyl-lysine 27 results from PcG protein EZH2 methyltransferase activity and is considered a relatively specific mark for PcG function (30). Thus, the YY1 201–226 motif is necessary and sufficient to recruit PcG proteins to DNA. Because this short segment of YY1 is necessary and sufficient for PcG transcriptional repression (Fig. 2), and is necessary and sufficient for recruiting PcG proteins and PcG enzymatic activities to DNA (Fig. 4), we propose that this sequence be referred to as the REPO domain (recruitment of Polycomb).

Some PcG complexes also contain HDAC activity (31), and we previously showed that full-length GALYY1 DNA binding in vivo can result in H3 deacetylation (15). Therefore, we tested the ability of our mutant YY1 transgenes to result in H3 lysine 9 deacetylation. Interestingly, the GALYY1Δ201–226 protein resulted in loss of H3 acetylation on lysine 9 (Fig. 4 A top panel, lanes 9,10, 4B). Therefore, recruitment of HDAC activity does not depend on the presence of the YY1 REPO domain. This is perhaps not surprising because the HDAC protein, RPD3, physically interacts with YY1 sequences 170–200 (23) and may thus be responsible for reduction in H3K9 acetylation, and for transient repression functions (Fig. 3). Interestingly, the GALYY1Δ201–226 protein resulted in more significant H3 deacetylation than the full-length GALYY1 protein itself (Fig. 4B). This finding might be due to steric hindrance with the full-length GALYY1 protein, which can recruit both PcG and HDAC proteins. The GALYY1Δ201–226 protein cannot recruit PcG proteins and therefore may be more efficient at recruiting HDAC proteins. In contrast to these results, DNA binding by GALYY1 201–226 failed to result in H3 deacetylation (Fig. 4A Lower, lanes 9 and 10; Fig. 4B). Apparently, the PcG complexes recruited by the isolated YY1 REPO domain do not contain HDAC activity. Alternatively, a subset of embryos under these conditions may contain H3 acetyl-K9 at the BGUZ reporter, whereas the remaining embryos do not (discussed more fully below).

Interestingly, recruitment of H3K27 methyltransferase activity correlated with YY1 PcG function. Thus, although YY1 resulted in recruitment of H3K27 methyltransferase activity to the BGUZ reporter (Fig. 4 A and B), it failed to recruit this activity to the non-PcG-dependent GAL-NP6-LacZ reporter (Fig. 4C, lanes 7 and 8). However, YY1 was able to recruit H3K9 HDAC activity to the GAL-NP6-LacZ reporter (lanes 5 and 6), consistent with transient YY1 repression activity on this reporter. As expected, levels of H3 acetylation and methylation at the endogenous PRE_D site were unaffected by the various GALYY1 transgenes (data not shown) apparently because of binding by endogenous PHO and PHOL proteins as well as to the inability of GALYY1 201–226 to bind to endogenous PRE sequences because of absence of the YY1 zinc fingers.

REPO Domain Homologies.

Several proteins are implicated in PcG recruitment to DNA in Drosophila. These proteins include GAGA, PHO, PHOL, Zeste, and DSP1 (11, 13, 26, 32–34). Of these Drosophila proteins, only PHO and PHOL contain sequences homologous to the YY1 REPO domain. Many organisms, however, contain genes apparently orthologous to YY1, and the encoded proteins often show high levels of homology. We aligned the YY1 REPO domain with comparable sequences in these apparent YY1 orthologs from 13 different species (Fig. 5). The level of conservation was extraordinary, with complete identity between human and chimpanzee, sheep, rat, mouse, dog, and puffer fish. In the case of the puffer fish–human comparison, these proteins are overall only 72% identical (data not shown), but the REPO domain shows complete identity. Puffer fish YY1 shows much lower conservation in sequences amino terminal to the REPO domain, with the REPO domain marking the boundary between lower homology in the amino-terminal half of the protein and higher homology in the carboxyl-terminal half, which also contains the zinc fingers. In chicken, frog, zebra fish, and honey bee, only a single amino acid was substituted in the 22-aa REPO domain (95% identity). Honey bee and human YY1 show extensive divergence overall (54% identity), but show high conservation in only two regions; the zinc fingers and the REPO domain. This pattern of homology is reminiscent of human YY1 and Drosophila PHO (7). Sea urchin contains two substitutions compared with human YY1, and Drosophila PHO shows four substitutions. An additional REPO-containing protein, YY2, was recently identified (35). Here again, the homology is strong. Human YY1 and YY2 are only 56% identical overall but are 86% identical within the REPO domain (Fig. 5). YY2 proteins are only observed in placental mammals, and YY2 proteins diverge rapidly outside of the zinc-finger region (36). Interestingly, the YY2 REPO domains from several species show much higher variation than the YY1 REPO domains, consistent with more extensive sequence divergence of YY2 among placental mammals (36) (Fig. 5). Finally, Drosophila PHOL shows the highest divergence (Fig. 5).

Fig. 5.

Homology of YY1 REPO domain to similar domains in other proteins. The REPO domain is highly conserved. (Upper) The human YY1 REPO domain (residues 205–226) is shown at the top with homologous sequences from YY1 and YY2 proteins from other species as well as from Drosophila PHO and PHOL shown below. Dashes indicate amino acid identity, and asterisks represent deletions. (Lower) Human YY2 REPO domain is compared with the corresponding YY2 REPO domains from chimpanzee, mouse, rat, and dog. Accession numbers are as follows: human YY1 (AAA59926), chimpanzee YY1 (XP_510162), sheep YY1 (AAT74924), rat YY1 (AAR14688), mouse YY1 (AAH55899), dog YY1 (XP_854514), chicken YY1 (NP_510162), frog YY1 (CAA54777), zebra fish YY1 (AAH71351), puffer fish YY1 (CAG01508), sea urchin YY1 (XP_790188), honey bee YY1 (XP-397280), fly PHO (AAL48765), fly PHOL (NP_648317), human YY2 (AAS68634), chimpanzee YY2 (XP_529235), dog YY2 (XP_868531), and rat YY2 (AAZ38710). The mouse YY2 sequence was taken from Luo et al. (36).

Discussion

We have shown here that the YY1 REPO domain is necessary and sufficient for PcG transcriptional repression, Polycomb recruitment to DNA, and methylation of histone H3 on lysine 27. The strong homology of the REPO domain in numerous species suggests that this domain carries out similar functions in multiple proteins. We suggest that the homologous REPO domains from PHO, PHOL, and YY2, along with the REPO motifs from predicted YY1 orthologs from other species, carry out the same PcG functions. However, in a slightly different system, the Drosophila PHO REPO domain (amino acids 140–172) fused to the GAL4 DNA-binding domain failed to support repression of the BGUZ reporter (37). In this case, the GAL-PHO REPO construct was expressed from the hsp70 promoter after embryos were repeatedly heat shocked, rather than from the hunchback promoter as in our repression assays. This finding suggested a heat-sensitive step in REPO domain repression compared with full-length YY1 repression. This might not be surprising because other potentially stabilizing YY1 protein interactions would be absent with the YY1 201–226 construct. To test this hypothesis we subjected BGUZ hbGALYY1 and BGUZ hbGALYY1 201–226 embryos to either no heat shock, a single heat shock, or to multiple heat shocks similar to the system used by Klymenko et al. (37). Although none of the heat shock regimens affected full-length GALYY1 repression, the multiple heat-shock regimen partially ablated GALYY1 201–226 repression in ≈30% of the embryos [see supporting information (SI) Fig. 6]. Similarly, we tested BGUZ repression by human GALYY1 201–226 when driven by the hsp70 promoter after heat shock. Although most embryos showed efficient BGUZ repression after a single heat shock, many GALYY1 201–226 embryos failed to repress transcription after multiple heat shocks. Therefore, there appears to be a heat-sensitive step in REPO domain repression. Activation of BGUZ expression in a subset of heat-shocked GALYY1 201–226 embryos may also explain the high level of H3 acetyl-K9 we observed at the BGUZ reporter in hsp70-driven GALYY1 201–226 embryos (Fig. 4 A and B). Thus, the failure of repression by the Drosophila PHO GAL-REPO fusion (37), is likely to be due to heat-sensitivity of the REPO repression mechanism. It will be interesting to test the Drosophila PHO REPO domain in the hunchback-driven system.

The mechanism of REPO domain function is not yet clear. A number of potential phosphorylation sites exist within the REPO domain, and a number of residues are highly conserved (Fig. 5). Detailed mutagenesis studies will be required to precisely define the residues needed for PcG function. The ability of the YY1 REPO domain to recruit PcG proteins to DNA strongly suggests that the REPO domain functions, at least in part, by direct protein interactions. Indeed, Wang et al. (28) showed that this segment of YY1 and PHO could physically interact with Drosophila EZ. Similarly, Mohd-Sarip et al. (26, 27) showed that the PHO REPO domain influenced PcG interaction with DNA by in vitro electrophoretic mobility-shift assays. YY1 also interacts with the PcG protein RYBP (38), the Xenopus Esc homolog, XEED (39), and the PRC2 protein EZH2 (16), similar to interactions described for the homologous fly proteins PHO and E(z) (28). Although YY1 can physically interact with several PcG proteins, it is not typically purified as a stoichiometric component of either the PRC1 or PRC2 complexes (2, 3, 31, 37, 40–44). YY1 is likely to either bind DNA separately then recruit other PcG proteins, or transiently interact with PcG complexes to target them for recruitment to DNA. Recent work by Klymenko et al. (37) showed that Drosophila PHO can be purified in complexes containing either the INO80 complex or a PHO-repressive complex containing PcG protein dSfmbt. Whether this PcG protein serves as a bridge between PHO and other PcG complexes waits to be seen.

Although the YY1 REPO domain can clearly mediate PcG repression and can function in PcG recruitment to DNA, other proteins must be involved in these processes. Drosophila pho/phol double mutants die as third-instar larvae and lack PcG binding to several sites on polytene chromosomes (6). However, PcG binding at most genomic sites is unaffected (6). Therefore, additional proteins are most likely able to maintain PcG DNA binding in the absence of PHO and PHOL.

YY1 is a ubiquitously expressed protein with multiple functions. It is likely that specific YY1 functions are regulated by either a variety of protein interactions, or specific posttranslational modifications. It will be interesting in the future to determine the specific protein interactions that are mediated by the REPO domain for carrying out PcG function.

Materials and Methods

Fly Strains.

BGUZ and GAL-NP6-LacZ reporter strains were kindly provided by Juerg Muller (European Molecular Biology Laboratory, Heidelberg, Germany) (17). Strain ry⁵⁰⁶ and various balancer strains (Acc+Vg2/FM7C and BcE1p/Cyo; Dr/Tm3Sb) were kindly provided by N. Bonini (University of Pennsylvania). YY1 expression constructs were coinjected into dechorionated ry⁵⁰⁶ embryos with phsπ. Eclosed flies were crossed back to ry⁵⁰⁶, and progeny were screened for transgene incorporation by appearance of ry⁺ eyes.

Plasmids.

YY1 cDNA fragments were prepared by standard PCR methods and cloned into pCDNA3.1+ (Invitrogen). GALYY1 DNA fragments were cloned into either hunchback- or hsp70-driven pry-derived vectors (8). Primers and strategies for generating plasmid constructs are available upon request.

Embryo-Staining Reporter Assay.

For experiments with the BGUZ reporter, overnight egg lays were used. For experiments with the GAL-NP6-LacZ reporter, 1-h egg lays were aged for 15 h at 25°C, followed by heat shock at 37°C for 45 min, and then an additional incubation at 25°C until hour 18. The collected embryos were processed for LacZ staining as described (8).

Chromatin Immunoprecipitation.

Drosophila embryos were harvested from 1-h egg lays, heat shocked at 3 h for 45 min, and allowed to age until 6 h from the start of the egg lay (BGUZ) or were heat shocked at 15 h after laying and harvested at 18 h (GAL-NP6-LacZ). Embryos were dechorinated in 50% bleach and fixed with 2% formaldehyde in 50 mM Hepes (pH 7.6) 100 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, and 3 volumes n-heptane with shaking for 15 min. Embryos were washed with PBS containing 0.125 M glycine and 0.01% Triton X-100, buffer A (10 mM Hepes, pH 7.6/10 mM EDTA/0.5 mM EGTA/0.25% Triton X-100), buffer B (10 mM Hepes, pH 7.6/200 mM NaCl/1 mM EDTA/0.5 mM EGTA/0.01% Triton X-100), and then sonicated in 10 mM Hepes (pH 7.6), 1 mM EDTA, and 0.5 mM EGTA with 106-μm glass beads (G-4649; Sigma, St. Louis, MO).

Fifty micrograms of chromatin in 16.7 mM Tris·HCl (pH 8.1), 0.01% SDS, 1.1% Triton X-100, 1.2 mM EDTA, 167 mM NaCl, and mammalian protease inhibitor mixture (Sigma) were precleared with protein A agarose beads in 200 μg/ml salmon sperm DNA. Antibodies [GAL4, sc-577; Santa Cruz Biotechnology (Santa Cruz, CA); PC, sc-25762; Santa Cruz Biotechnology; acetyl-K9, 07–352; Upstate Biotechnology (Lake Placid, NY); and trimethyl K27, 07–449; Upstate Biotechnology] were added and rocked overnight at 4°C. Protein A agarose beads were added, collected, washed, and resuspended in TE (10 mM Tris·HCl, pH 8.0/1 mM EDTA) containing RNase A (50 μg/ml) at 37°C for 30 min. TE was adjusted to 0.5% wt/vol SDS and 0.5 mg/ml proteinase K at 37°C for 4–6 h and then placed at 65°C overnight. DNA was recovered by phenol–chloroform extraction and ethanol precipitation with glycogen as carrier. PCR was done with 2 and 20 ng of the recovered DNA, 200 mM dNTPs, and 1 unit of AmpliTaq DNA polymerase (Applied Biosystems, Foster City, CA) in the provided reaction buffer. PCR for BGUZ consisted of 20 cycles (95°C for 30 s; 55°C for 30 s; and 72°C for 30 s) with primers F2BXDGGATTTTTTCTGGCTCGTTCAACC and UBXR2 GTTTCCTAAAACTAATAATCCATTTTA, followed by Southern blotting. GAL-NP6-LacZ PRCs were performed at 95°C for 2 min; followed by 29 cycles of 95°C for 30 s, 69°C for 30 s, and 72°C for 30 s and then a final extension at 72°C for 3 min with primers: forward, CCGACGGCACGCTGATTGAAG and reverse, ATACTGCACCGGGCGGGAAGGAT.

Transient Transfections.

NIH 3T3 cells were transfected by either the calcium phosphate method using 3 μg of (YY1)₄TKCAT reporter, 1 μg of CMV-βGal, and indicated amounts of CMV expression plasmids and CMV vector to bring to a total of 10 μg or by the Effectine method (Qiagen, Valencia, CA) using 200 ng of GALTKCAT, 100 ng of CMVβGal, and indicated amounts of CMVGAL expression plasmids and CMV vector to a total of 4 μg.

Abbreviations

HDAC

histone deacetylase

Pc

Polycomb

PcG

Pc Group

PHO

pleiohomeotic

PHOL

pleiohomeotic-like

PRE

Polycomb response element

REPO

recruitment of Polycomb.