The DNA-binding Polycomb-group protein Pleiohomeotic maintains both active and repressed transcriptional states through a single site

Polycomb group (PcG) proteins associate with specific DNA regions, forming complexes that modify chromatin and maintain a transcriptionally silent state(Grimaud et al., 2006; Muller and Kassis, 2006; Ringrose, 2007; Schuettengruber et al., 2007; Schwartz and Pirrotta, 2007). Among the genetically identified Drosophila PcG proteins, only Pleiohomeotic (Pho) and its homolog Pho-like (Phol) have sequence-specific DNA-binding activity (Brown et al.,2003; Brown et al.,1998; Fritsch et al.,1999). They, together with other DNA-binding proteins(Americo et al., 2002; Brown et al., 2005; Muller and Kassis, 2006; Ringrose and Paro, 2007; Schuettengruber et al., 2007),have been implicated in targeting each of the known PcG complexes(Cao et al., 2002; Czermin et al., 2002; Kuzmichev et al., 2002; Shao et al., 1999; Wang et al., 2004a) to their target genes. In mammals, PcG proteins are involved in maintenance of the stem cell niche (Buszczak and Spradling,2006; Sparmann and van Lohuizen, 2006) and in oncogenesis(Sparmann and van Lohuizen,2006).

Trithorax group (trxG) proteins oppose the action of the PcG(Kennison, 2004), form complexes on target DNA (Badenhorst et al.,2002; Bantignies et al.,2000; Fyodorov et al.,2004; Papoulas et al.,1998; Petruk et al.,2001) and maintain an active state, by modifying histones and remodeling chromatin. PcG and trxG proteins are considered to be distinct from other regulators in that they can mediate heritable maintenance of the transcriptional state from one cellular generation to another(Grimaud et al., 2006; Muller and Kassis, 2006; Ringrose and Paro, 2007; Schuettengruber et al., 2007; Schwartz and Pirrotta, 2007). Target sequences that mediate the functions of these proteins are termed PcG-and trxG-responsive elements (PREs and TREs, respectively) or, collectively,maintenance elements (MEs) (Brock and van Lohuizen, 2001), because they often overlap. Dissection of these target elements has proven to be difficult, because of their complexity, and because the associated activities manifest themselves fully only in a developmental context. Consequently, how PcG and trxG complexes are targeted to DNA, and how this targeting varies with the transcriptional state, is just beginning to be elucidated.

Transgenes carrying PREs/TREs in Drosophila have been seen to share a property known as pairing-sensitive silencing (PSS)(Gindhart and Kaufman, 1995; Kassis, 2002). Transgenic flies carrying the mini-white gene typically have eye colors ranging from yellow to orange in a white mutant background. Normally, flies that are homozygous for such a transgene have a darker eye color than heterozygotes, as the genetic dose of mini-white is doubled. However,with transgenes carrying PRE/TREs, the eye color is often lighter in homozygotes than in heterozygotes. PSS is also often accompanied by a reduction in mini-white activity from heterozygous transgenes,causing a relatively faint eye color, or a variegated pattern of eye coloration. This phenomenon was first reported for a pairing-sensitive element(PSE) near the engrailed (en) transcription start site(Kassis, 1994; Kassis et al., 1991). It was also shown that the en PSE can function as a PRE in embryos, in that it can restrict expression of a transgenic reporter gene driven by the Ultrabithorax (Ubx) regulatory region bithoraxoid(bxd) to the posterior part of the embryo, where Ubx is normally expressed, in a PcG-dependent manner(Americo et al., 2002). More generally, although PREs usually exhibit PSS activity, PSEs may not be sufficient to function as a PRE (Kassis,1994; Kassis,2002).

Genome-wide predictions of possible Drosophila PREs using sequences common to known PREs, including binding sites for the PcG and trxG proteins Pho/Phol, GAGA factor (GAF) and Zeste, identified 167 candidates,including the even skipped (eve) promoter region(Ringrose et al., 2003). Other studies using two different techniques to determine in vivo protein binding(ChIP on chip and DamID) identified other candidate PREs with some, although not extensive, overlap (Negre et al.,2006; Schwartz et al.,2006; Tolhuis et al.,2006). These studies suggested that the eve locus is bound by the Polycomb (Pc) protein(Schwartz et al., 2006; Tolhuis et al., 2006). Earlier, mapping of binding sites for Pc and Polyhomeotic (Ph) on polytene chromosomes identified the cytological location of eve(Sinclair et al., 1998). Consistent with this, eve is ectopically expressed in the embryonic nervous system in mutants of the PcG gene ph(Smouse et al., 1988).

Here, we show that the Drosophila eve gene is regulated by a Pho-dependent maintenance element located 9 kb downstream of the promoter. This element maintains repression in cells where eve is turned off during early development, but maintains the active state in other cells. Both negative and positive maintenance depend on Pho binding, and on phogene activity. The element shares properties with other PREs, including binding sites for several factors. Dissection suggests that although GAF is involved, Zeste and Dsp1 may be dispensible, and that Grainyhead/Elf1(Bray and Kafatos, 1991), as well as unknown factors, may also contribute. Thus a `core' DNA-binding component, differentially modified or regulated by the existing transcriptional state, mediates epigenetic maintenance of that state.

All sequence coordinates in the eve locus are relative to the transcription start site (+1). The parental P-element plasmid containing Glass-binding sites has been described previously(Fujioka et al., 1999). To construct the PSS assay transgenes of Fig. 1, eve DNA from +8.9 to +9.2 kb (PRE300) was cloned upstream of the Glass-binding sites followed by mini-white. DNA from-275 bp to +116 bp was similarly used to make PSEpro. The mini-whiteP-element-based GFP-RR construct was made by fusing eve DNA from -275 bp to +99 bp with the EGFP-coding region, driven by the RR regulatory DNA(+7.9 to +9.2 kb), which consists of the neuronal RP2+a/pCC enhancer and PRE300 (Fujioka et al., 2003). For ΦC31 recombinase-mediated cassette exchange (RMCE)(Bateman et al., 2006; Groth et al., 2000), the piB-GFP construct (Bateman et al.,2006) was first modified to increase the number of unique cloning sites (attBΔ2; details available on request). Then, GFP-RR described above was transferred to this vector. For the PRE assays of Fig. 2, a P-element based bxd-Ubx-Z vector (with a bxd enhancer fragment inserted upstream of the -3.1Ubx-βgal fusion in Uβghz)(Muller and Bienz, 1991) was modified by inserting PRE300 (with or without the Pho site mutated) into the XbaI site, oriented such that the PRE300 3′ end was closer to the Ubx promoter. For RMCE, the PRE300 (with or without the Pho site mutated)-bxd-Ubx-Z region was transferred to attBΔ2. PRE300 was also replaced by a 500 bp λ phage DNA fragment as non-PRE negative control in this context (λ500-bxd-Ubx-Z). Binding site mutations used in this study are shown in Fig. 3.

P-element-based transgenic lines were established as previously described(Fujioka et al., 2000; Rubin and Spradling, 1982). RMCE was performed as previously described(Bateman et al., 2006). The attP target lines used were chosen at chromosomal locations that were not suggested to have PREs nearby (Negre et al., 2006; Schwartz et al.,2006; Sinclair et al.,1998; Tolhuis et al.,2006). The attP target lines 25C, 52D(Bateman et al., 2006) and 78C4(obtained in our laboratory) were used for inserting GFP-RR. For RMCE of bxd-Ubx-Z constructs, 25C, 52D, 78C4 and 95E5 (obtained in our laboratory) lines were tested. 52D and 95E5 were used for further analysis. For staining with anti-β-gal (ICN, 1:1000 dilution), embryos were collected for 6 hours at room temperature (RT), then incubated at 17°C for 15 hours. The antibody was visualized using biotin-conjugated anti-rabbit IgG followed by peroxidase-conjugated streptavidine, with DAB as a substrate. For Fig. 4, each central nervous system (CNS) was dissected out in phosphate-buffered saline, placed on a poly-L-lysine-coated cover slip, and covered with a non-coated cover slip using another cover slip as a spacer to prevent flattening.

For pho mutant analysis of PSS, six PRE300-carrying lines(including two GFP-RR lines) that showed PSS were crossed to generate PRE300;pho¹/ci^D. Larvae from these stocks were grown at room temperature, and pharate adults were dissected to analyze their eye color. For pho and phol mutant analysis of positive maintenance in the CNS, a mini-white P-element-based GFP-RR line was used to produce GFP-RR;phol^81A/TM6B,Tb,GFP-RR;pho¹/ci^D or GFP-RR;phol^81A/TM6B,Tb;pho¹/+, which were then self-crossed. Homozygous phol larvae were tentatively identified as non-Tubby, and confirmed post hoc by PCR. In all cases,homozygous pho and pho/+ larvae and pupae were identified post hoc by PCR analysis of their DNA(Brown et al., 2003). GFP-RR at 52D was crossed into mutant backgrounds to produce GFP-RR(52D);E(z)⁶¹/TM6B,Tb,GFP-RR(52D);trx¹/TM6B,Tb, and GFP-RR(52D);Trl^13C/TM6B,Tb. Homozygous mutant larvae from these lines were identified by their non-Tb phenotype. For E(z)⁶¹, eggs were collected at 18°C for 24 hours,kept at 18°C until near hatching, then moved to 29°C until the third larval instar. For trx¹, eggs were collected similarly,then moved to 25°C until the third larval instar. For Trl^13C, eggs were collected similarly, and kept at 18°C until the third larval instar.

To test effects of PcG mutations on eve PRE activity, lines homozygous for bxd-Ubx-Z-derived transgenes (at target site 52D) were each crossed with ph⁵⁰³/FM7,B or Pc⁴/TM3,Sb. Then, ph⁵⁰³/+;bxd-Ubx-Z/+(non-B) females were crossed with homozygous bxd-Ubx-Zmales, while bxd-Ubx-Z/+;Pc⁴/+ (non-Sb) males and females were intercrossed. Resulting embryos were collected as described above. Homozygous mutant embryos were identified by ectopic lacZexpression in anterior regions of embryos, which were not present in a wild-type background. For ph⁵⁰³, the ectopic lacZphenotype of hemizygous mutant embryos was confirmed by double-staining for ectopic eve expression in the CNS(Smouse et al., 1988).

Chromatin immunoprecipitation (ChIP) analysis was performed as follows. Dechorionated embryos were crosslinked with 2% formaldehyde for 15 minutes at room temperature, washed and stored at -80°C. About 200 μl of embryos were homogenized in 1 ml Buffer A (350 mM sucrose, 15 mM HEPES pH 7.6, 10 mM KCl, 5 mM MgCl₂, 0.1 mM EDTA, 0.5 mM EGTA). Homogenates were centrifuged at 600 g for 5 minutes to pellet nuclei, which were then washed gently with 1 ml Buffer A and repelleted. Pellets were resuspended in 1 ml Buffer A containing 0.2% Triton-X100, incubated for 5 minutes on ice, then layered on top of 3 ml 800 mM sucrose (in Buffer A) in 15 ml conical-bottom tubes, and centrifuged at 250 g for 10 minutes. The pellets were subjected to ChIP analysis essentially as described previously (Wang et al.,2004b), using Pho-specific antiserum(Brown et al., 1998). PCR primer sequences will be provided on request.

In a previous study, we reported that a region between +8.6 and +9.2 kb(relative to the eve transcription start site), at the 3′ end of the eve locus, caused PSS of mini-white expression in transgenes (Fujioka et al.,1999). This region also caused repression of mini-whiteexpression in heterozygotes, making it difficult to use a standard mini-white-containing P-element vector to generate transgenic lines,because complementation of the white^- phenotype was extremely weak at most chromosomal locations (data not shown). In such a case,it is likely that insertions at many sites are missed, resulting in a biased sampling of potential insertion sites. To overcome this problem, three tandem Glass activator-binding sites, which are active specifically in the eye disc,were introduced just upstream of the mini-white promoter to increase expression (Fujioka et al.,1999). This Glass site vector was also used in the analysis of the en PRE (Americo et al.,2002), where it was shown that the presence of Glass sites does not affect the percentage of PRE-carrying transgenic insertions that exhibit PSS. As expected, with the Glass-binding sites, the range of eye colors of heterozygous transgenic flies shifted toward the dark end of the spectrum(Fig. 1, compare `with Glass'with `without Glass'; note that none of these lines carry a PSE). Among 53 independent lines carrying the Glass vector without the PSE, 43 were homozygous viable and none showed PSS (Fig. 1, `with Glass').

We dissected the original 600 bp PSE (+8.6 to +9.2 kb, data not shown) down to an active region of 300 bp (+8.9 to +9.2 kb). The resulting element caused repression of Glass-augmented mini-white expression in heterozygous transgenic flies, as evidenced by the resulting shift toward lighter eye colors (`PRE300' in Fig. 1;compare with `with Glass'). Among homozygous viable lines, 73% of them showed PSS activity (Fig. 1, Fig. 2A,B). We refer to this element as a PRE because of these results together with further results described below.

The eve promoter region (from about -700 to -100 bp), but not the eve 3′ region, was previously identified in a genome-wide prediction of PRE/TREs based on a clustering of Pho-, GAF- and Zeste-binding sites (Ringrose et al., 2003). We discovered that the eve promoter region from -275 to +166 bp also has PSS activity. Fifty-three percent of transgenic lines carrying this region showed PSS (PSEpro, Fig. 1). Although the combination of PRE300 with the eve promoter caused more repression of heterozygous mini-white expression than did either one alone, PSS activity itself was not increased relative to that of PRE300(PRE300 + PSEpro, Fig. 1).

As the eve 3′ PRE showed stronger activity than the promoter, we focused on the 3′ PRE in further experiments. Previous dissections of PSS regions from the en gene(Americo et al., 2002), the MCP silencer (Busturia et al.,2001) and the iab-7 PRE(Hagstrom et al., 1997; Mishra et al., 2001) (both from the bithorax complex) showed that activity depended on both Pho-and GAF-binding sites. PRE300 contains one perfect match to the Pho-binding site consensus and three close matches to the GAF consensus. Mutating the single Pho site (Fig. 3)strongly reduced PSS activity from 73% to 11% (PRE300ΔPho, Fig. 1). This was accompanied by a shift in the spectrum of heterozygous eye colors back to that obtained without the PSE, suggesting that the cis-repressive activity of PRE300 is entirely dependent on the Pho site. Mutating two or all three of the GAF sites(Fig. 3) caused a progressive decrease of PSS activity, but only to 53%(Fig. 1: PRE300Δgaga1,Δgaga12 and Δgaga123). Mutating GAF sites also did not have as much of an effect on heterozygous eye color as did mutating the Pho site. These data suggest that the PSS activity of PRE300 depends only weakly on GAF binding sites, whereas, like other PSEs examined in detail, it is strongly dependent on the Pho-binding site.

In order to test whether this Pho-binding-site dependence reflects an actual dependence on Pho binding, or whether it might be due to some overlapping DNA-binding activity, we placed several PRE300-carrying lines in a pho¹ mutant background. As pho¹mutants die as pharate adults, it was necessary to dissect pupae in order to examine their eye color. Compared with a wild-type background, the homozygous eye color was darker in pho¹ mutants in all six lines examined. This effect was very strong in one line(Fig. 2E,F), whereas the other lines showed less complete, albeit still clear, derepression(Fig. 2C,D; data not shown). Incomplete derepression is expected in the pho mutants, because the pho-like gene (phol) provides a partially redundant function with that of pho. It is not possible to carry out this analysis in phol^81A; pho¹ double mutants, because they fail to pupariate (Brown et al.,2003). The convergence of these genetic data with the Pho-binding-site dependence of PSS strongly suggests that Pho is directly involved in silencing through PRE300.

Among known PREs/TREs, binding sites for Pho/Phol, GAF/Pipsqueak, Sp1/KLF,Zeste and Dsp1 were found to be clustered(Brown et al., 2005; Muller and Kassis, 2006; Ringrose, 2007; Schuettengruber et al., 2007). PRE300 contains all of these binding sites, including two Dsp1 sites, three GAF sites, two Sp1/KLF sites, one Zeste site and one Pho site(Fig. 3). It has also been shown that the grainyhead (grh) gene product(Bray and Kafatos, 1991), which was originally identified as a transcriptional activator termed Elf-1(Bray et al., 1989), physically interacts with RING, a PcG gene product(Tuckfield et al., 2002). Importantly, Grh/Elf-1 was shown to cooperatively interact with Pho on the iab-7 PRE in vitro, and to genetically interact with pho(Blastyak et al., 2006). PRE300 also contains a reasonable match to the Grh consensus site(Fig. 3). In order to test whether these binding sites are required for its PSS activity, PRE300 was split into two overlapping parts (PRE300-5′ and PRE300-3′). As the Pho site is necessary for the activity, this site was included in both parts(Fig. 3). Surprisingly,PRE300-5′, which includes all of these sites except for one Sp1/KLF, one GAF and the Grh site, showed PSS activity in only 5% of transgenic lines(Fig. 1). By contrast,PRE300-3′, which contains only the Pho site, one Sp1/KLF site, the Grh site and one of the GAF sites, showed almost full PSS activity (69% versus 73%for PRE300, Fig. 1). These data suggest that only a subset of known binding sites are required for the PSS activity of PRE300, and further suggest that other, unidentified factors may well be involved.

Although PSS activity is often seen with transgenes carrying a PRE, not all PSEs can act as a PRE (Kassis,2002). We directly tested for PRE activity of eve PRE300 by combining it in a transgene with a bxd-Ubx-Z reporter gene. This reporter is expressed throughout the ectoderm and nervous system in late-stage embryos (Muller and Bienz,1991). However, when a PRE is inserted into this reporter, it can prevent activation of lacZ expression in parasegments 1-5, similar to the endogenous Ubx PRE from the bxd region(Americo et al., 2002; Brown et al., 2005; Busturia and Bienz, 1993; Hagstrom et al., 1997; Horard et al., 2000; Muller and Bienz, 1991). In two out of four P-element-based transgenic lines carrying PRE300 in this reporter, we observed an anteriorly restricted lacZ pattern (similar to Fig. 2H, data not shown). Similarly, studies of other PREs in this reporter found that not all transgenic lines showed repression (Americo et al., 2002; Brown et al.,2005; Busturia and Bienz,1993; Hagstrom et al.,1997; Horard et al.,2000; Muller and Bienz,1991), presumably owing to the effects of different chromosomal contexts. In order to further analyze the PRE activity of PRE300, we incorporated the bxd-Ubx-Z reporter into a vector for RMCE(Bateman et al., 2006; Groth et al., 2004), which allows the analysis of multiple transgenic constructs at the same chromosomal location. We initially tested PRE300-bxd-Ubx-Z inserted at four different chromosomal target sites, two of which were used in a previous study(Bateman et al., 2006) and two others that we developed. All four targeted lines showed an anteriorly restricted pattern (Fig. 2H,N,compare with Fig. 2G,M, which carry non-PRE DNA as a negative control, and data not shown). We tested the effect of mutating the Pho site in PRE300 in three of these chromosomal contexts, and all three showed significant anterior derepression(Fig. 2I,O; data not shown). We further tested one of these anteriorly restricted lines to see whether the repression depends on PcG genes, using the alleles Pc⁴ and ph⁵⁰³. Repression was lost or significantly reduced in each of these mutants (compare Fig. 2K,L with 2J). In addition to a strong effect in ph hemizygotes(Fig. 2), based on the number of embryos in the population showing anteriorly expanded expression, there appears to be a partial loss of repression in ph⁵⁰³heterozygotes (from ph⁵⁰³ heterozygous mothers) as well(data not shown). So, the PRE activity of eve PRE300 depends on both the Pho site and on PcG gene function, and thus has similar characteristics to other known PREs.

Eve is expressed in several sets of neurons within each hemisegment of the embryonic central nervous system (CNS), including RP2, a/pCC, CQ/U and the EL clusters. Each of these aspects of expression is maintained to the third larval instar. When the RP2+a/pCC enhancer (+7.9 to +8.6 kb, previously referred to as `RN') (Fujioka et al.,2003) was used to drive heterologous gene expression (GFP or lacZ), expression was normal in embryos, but was not maintained to the third larval instar (data not shown). However, when the 3′ end point was extended to +9.2 kb (to generate the `RR' construct)(Fujioka et al., 2003),expression was fully maintained (Fig. 4A). The region required for this maintenance contains PRE300. Although Pho, as a PcG gene, is known only to be involved in maintaining the repressed state, its ortholog in mammals, YYI, is known to be involved in activation of some target genes (Gordon et al., 2006). We tested whether Pho affects positive maintenance activity. When the same Pho-binding site mutation that knocked down PSS(Fig. 3) was introduced into the RR construct, CNS maintenance activity was reduced or eliminated. Out of the 11 independently inserted transgenes tested, eight gave normal expression in the embryonic CNS. Of these eight, three completely failed to maintain expression to the third larval instar (data not shown, similar to Fig. 4F), and the other five showed relatively weak and variable expression of GFP in third instar larvae(Fig. 4B and data not shown). As the Pho dependence of positive maintenance seemed to be affected somewhat by the chromosomal context, we confirmed these results by examining expression in a pair of lines, one with the Pho site mutated, inserted at the same chromosomal location using RMCE. At this target site, although GFP expression in third instar larvae was somewhat weaker than in the random insertion line shown in Fig. 4A, expression in third instar larvae was nonetheless abolished when the Pho site was mutated(Fig. 4E,F). Thus, the maintenance of CNS expression by PRE300 requires the Pho-binding site.

GAF has been shown to affect the activity of other PREs(Busturia et al., 2001; Hagstrom et al., 1997; Hodgson et al., 2001; Horard et al., 2000; Mishra et al., 2001), and to facilitate binding of Pho to a chromatinized PRE(Mahmoudi et al., 2003). Moreover, mutations in GAF sites affect the PSS activity of PRE300(Fig. 1). Therefore, we tested the effects of GAF site mutations using both conventional P-element transformation and RMCE. When all three GAF sites were mutated, GFP expression depended strongly on the insertion site. Although two out of 11 random insertion lines showed similar levels of expression in third instar larvae as transgenes carrying wild-type PRE300, four out of 11 lines showed weak expression, and five out of 11 lines showed ectopic expression throughout the CNS, such that we could not discern the intensity in RP2 and a/pCC neurons. When RMCE was used to compare directly the activity of wild-type and triply GAF-mutated PRE300, there was no apparent change in GFP expression (data not shown). Thus, mutating sites for GAF can have either a positive or a negative effect on CNS maintenance by PRE300. The ectopic expression observed with the GAF-mutated element also suggests that wild-type PRE300 may have a negative maintenance activity in cells of the CNS where eve is not normally expressed, in addition to its role in positive maintenance.

As stated above in relation to PSS, it is important to determine whether the Pho-binding site dependence reflects an actual dependence on Pho binding,or whether it might be due to some overlapping DNA-binding activity. In order to test this, the GFP-RR transgene was placed in a phol^81A,pho¹ double mutant background. Indeed, relative to a wild-type background, GFP expression was clearly reduced(Fig. 4C, compare with 4A). In pho¹ single mutants, there was a reduction in most larvae(Fig. 4D), but the difference versus wild type was not as clear-cut as in the double mutant. In phol^81A single mutants, there was no apparent effect (not shown). This is consistent with the fact that pho¹ mutants die as pharate adults, whereas phol^81A mutants survive to adulthood (Brown et al.,2003). These data indicate that Pho and Phol are indeed involved in the maintenance not only of repression, but also of activation. Consistent with PRE300 having a negative maintenance activity in the CNS in addition to its role in positive maintenance, as suggested above, we observed widespread ectopic expression (seen as a higher background of GFP fluorescence) in some PRE300-carrying lines both in pho/phol double mutants and when the Pho site was mutated (e.g. Fig. 4C and Fig. 4F,respectively).

As a preliminary test of the involvement of other PcG genes in positive maintenance by PRE300, we tested the allele E(z)⁶¹, as homozygous larvae survive to the third larval instar. However, we did not see any effects on positive maintenance (data not shown). We also tested some trxG alleles that survive to the third larval instar. Larvae homozygous for trx¹ and Trl^13C did not show any effects on CNS expression of our GFP reporter (data not shown). Larvae heterozygous for trx^B11, Trl^R65 and Trl^R85 were also tested, and, again, no clear effects were observed (data not shown). However, because these alleles have significant maternal and zygotic contributions that allow them to survive long enough to be assayed, further analysis will be required to determine definitively whether these genes are involved along with pho and phol in positive maintenance by PRE300.

The convergence of genetic and biochemical evidence presented above strongly suggests that Pho acts directly through the binding site in PRE300. To confirm that Pho interacts with this region, and with this site, in vivo,we performed ChIP assays. First, we surveyed the eve locus for Pho binding using a series of primer sets for detection of immunoprecipitated fragments by PCR. As shown in Fig. 5 (B and E, primer set 13), Pho was detected specifically in the region of PRE300, both in early and later-stage embryos. Pho was also detected in the promoter region where binding sites were previously predicted,primarily in early embryos (Fig. 5B, primer set 6), as well as in the region of enhancers driving mesodermal and anal plate expression (Fig. 5B,E, primer set 8). This confirms that Pho interacts with the endogenous eve locus. Furthermore, it does so specifically with the region that our transgenic analysis implicated in maintenance of both the repressed and activated state. To test the involvement of the functionally important Pho-binding site, we tested transgenes carrying PRE300, either with the normal sequence, or with the Pho site mutated. Again, ChIP analysis indicated that Pho interacts with the transgenic copy of PRE300(Fig. 6A). Furthermore, the signal is reduced with the mutant Pho site, relative to both the endogenous eve signal and to the background controls(Fig. 6D, compare to Fig. 6A). This indicates that the site is important for Pho binding in vivo, and the loss of binding correlates with the loss of both repression and activation that occurs when the site is mutated.

In this study, we dissected a strong pairing sensitive silencing element from the 3′ boundary of the eve locus. We found that silencing activity depends on a single Pho-binding site, whereas sites for a number of other proteins found in such elements are less important. The element is genetically responsive to PcG-group activity, as it depends on phogene function. This eve 3′ PRE has bona fide PRE activity,which can maintain a silenced state established in embryos.

Previous studies have suggested that PRE-containing P-element-based transgenes have a tendency to insert near endogenous PREs, and that this can bias reporter gene expression. Here, we applied both P-element analysis and the RMCE system to compare the effects of mutating binding sites. Our data reveal that there is also variation in PRE effects using RMCE into different target sites. Therefore, it would seem important to test several target sites when using RMCE, to ensure that results are not specific to one chromosomal location. Furthermore, where sensitivity to position effects is high, such as with our GAF site-mutated PRE, it remains valuable to use the standard methodology to probe a variety of insertion sites.

Surprisingly, we found that the eve PRE is also required for positive maintenance of expression in the larval CNS, and that this activity requires both the Pho-binding site and pho gene function. Together,these data strongly suggest that Pho is directly involved in positive maintenance of gene activity. This is surprising because Pho has heretofore been associated only with direct repression of target genes, by recruiting the PRC2 complex and other PcG proteins (Wang et al., 2004b). However, recent studies have blurred the distinction between PcG genes and trxG genes, as some members of each class appear to have dual functions (Grimaud et al., 2006). Furthermore, PREs usually reside in close proximity to TREs (Ringrose and Paro,2007), and an element from the promoter region of engrailed that mediates PSS and can act as a PRE was recently shown to have an activating role in its natural context(Devido et al., 2008). Recent studies of the Ubx locus have indicated that PcG proteins are present at PREs in both the off and the on state, and that binding of Ash1 prevents silencing by the PRC complex in cells where Ubx is expressed,suggesting that silencing is actively prevented(Papp and Muller, 2006). A similar situation may pertain to Pho function in the eve locus.

Because trxG proteins are known to be involved in positive regulation by other maintenance elements, we were interested in whether they are involved in pho-dependent positive maintenance by the eve PRE. We were also interested in whether other PcG proteins are involved. Because our positive maintenance assay requires survival to the third larval instar, we have so far been able to test only weak alleles of trx, Trl and E(z), neither of which showed discernable effects in our assays. At this point, we can not definitively say whether other trxG or PcG proteins are involved in the positive maintenance function of the eve PRE. However, our observation that a consensus Grh binding site is present in the more active half of PRE300 suggests the involvement of Grh. Indeed, Grh has been shown to interact genetically with Pho, and to facilitate cooperative interaction with Pho in vitro (Blastyak et al., 2006).

Consistent with the broad overexpression of eve seen in the CNS of ph mutants (Smouse et al.,1988), the eve PRE may silence expression in many cells by forming a silencing complex. In wild-type embryos, in the subset of CNS cells where eve is expressed, the same Pho-dependent DNA binding platform may recruit a distinct complex that maintains the active state. Consistent with this model, we have found that expression driven by the eve RP2+a/pCC enhancer fades prematurely in late stage embryos in ph mutants, at the same time that endogenous eve is broadly overexpressed (M.F. and J.B.J., unpublished). It will be interesting to determine the composition of Pho-dependent complexes in cells where eve is on, and in those where eve is off.

How can a region 9 kb away from the basal promoter affect the state of gene expression? There are accumulating data suggesting that locus-wide regulation occurs through direct interactions of the promoter with enhancers and locus control regions. For example, a recent study showed that silencing by the bxd PRE directly affects the activity of the transcriptional machinery at the promoter (Dellino et al.,2004). In the eve locus, there are PSEs both at the 3′ end of the locus and at the promoter. Both contain clusters of binding sites typical of a PRE/TRE. It has been suggested that PRE-containing transgenes have a tendency to insert near endogenous PREs(Chiang et al., 1995; Fauvarque et al., 1995), which might be expected if they mediate long-range interactions. Putting these ideas together, the eve 3′ PRE may physically interact with the promoter region in a Pho-dependent manner. This may serve to keep eveon in some cells and to keep it off in others, depending on whether activating or repressive complexes mediate the association.

We thank Xian Wu for excellent technical assistance, Judy A. Kassis for plasmids and Pho antiserum; Michele P. Calos, C.-Ting Wu and Jack R. Bateman for reagents for the ΦC31 system; and Hugh W. Brock, Paul Schedl, Judy A. Kassis and the Bloomington Stock Center for mutant strains. We also thank Judy A. Kassis, Alex Mazo and Hugh W. Brock for comments on the manuscript. This work was supported by NSF-IOB0416760, NSF-MCB0818118 and NIH-R01GM050231 awards to J.B.J. and M.F.