The Drosophila pho-like gene encodes a YY1-related DNA binding protein that is redundant with pleiohomeotic in homeotic gene silencing

Polycomb-group (PcG) proteins are conserved transcriptional repressors with important roles during embryonic and post-embryonic development inDrosophila and vertebrates (for reviews, seeBrock and van Lohuizen, 2001;Francis and Kingston, 2001;Mahmoudi and Verrijzer, 2001;Simon and Tamkun, 2002). Among the best-characterized PcG target genes are the homeotic genes inDrosophila. At least 15 different PcG genes are required to repress homeotic genes in cells in which these genes must remain inactive during development. These PcG proteins exist in at least two distinct multiprotein complexes that do not appear to bind to DNA directly but bind to chromatin(Shao et al., 1999;Ng et al., 2000;Tie et al., 2001;Saurin et al., 2001). Antibody staining of polytene chromosomes revealed that PcG proteins are associated with about 80-100 loci in the Drosophila genome(Zink and Paro, 1989;DeCamillis et al., 1992;Martin and Adler, 1993;Rastelli et al., 1993;Lonie et al., 1994;Carrington and Jones, 1996;Peterson et al., 1997). Subsequently, chromatin-immunoprecipitation experiments showed that several PcG proteins can be crosslinked with specific DNA sequences, called Polycomb group response elements (PREs), in PcG target genes(Strutt and Paro, 1997;Strutt et al., 1997;Orlando et al., 1998). Recent progress towards understanding how PcG protein complexes may be anchored at PREs has come from the dissection of PREs and the identification of proteins that directly bind to PRE DNA.

PREs were first identified in reporter gene assays ascis-regulatory elements in homeotic genes that render expression of a reporter gene construct sensitive to mutations in PcG genes(Simon et al., 1993;Chan et al., 1994;Christen and Bienz, 1994). Furthermore, addition of a PRE causes unusual silencing of a mini-white reporter gene in transgenic flies and creates a new PcG band on polytene chromosomes at the transgene insertion site (reviewed byPirrotta, 1997a;Pirrotta, 1997b;Kassis, 2002). Characterization of a PRE from the engrailed gene led to the identification of the PcG protein Pleiohomeotic (Pho) as a PRE-binding factor(Brown et al., 1998). Pho is a DNA-binding protein related to the mammalian transcription factor YY1(Brown et al., 1998). Pho-binding sites are found in many different PREs and are required for PRE function in many different reporter constructs(Brown et al., 1998;Mihaly et al., 1998;Fritsch et al., 1999;Shimell et al., 2000;Mishra et al., 2001;Busturia et al., 2001). Despite this requirement for Pho-binding sites in reporter genes, pho mutants only weakly misexpress homeotic genes(Simon et al., 1992;Fritsch et al., 1999) and die as pharate adults with relatively weak homeotic transformations(Gehring, 1970). Thus, if Pho anchors PcG protein complexes on DNA, it most likely is not the only DNA-binding PcG protein that provides this function.

Because of the weak homeotic gene misexpression in pho mutants, we searched the Drosophila genome for pho-related sequences and identified a gene that we call pho-like (phol). The Phol protein binds to the same DNA sequence as Pho. The strong PcG phenotype ofphol, pho double mutants shows that Phol is another DNA-binding protein required for PcG repression. We examined the distribution of five different PcG proteins on polytene chromosomes in phol, pho double mutants. Our data show that binding of PcG proteins to a few chromosomal bands requires pho and phol, but that at most chromosomal locations PcG proteins remain bound in the absence of Pho and Phol.

Finally, we analysed the requirement for two other DNA-binding proteins that have recently been implicated in PcG repression: GAGA factor, which is encoded by the gene Trithorax-like (Trl); and Zeste(Horard et al., 2000; Mishra et al., 2000; Hodgson et al.,2001; Busturia et al.,2001; Hur et al.,2002). We report experiments aimed at testing the role of these two proteins in PcG repression in imaginal disks. Our results provide no evidence for a requirement of either Trl or zeste in PcG repression in imaginal disks.

PCR was performed on the on the EST clone, LD42095 using the rphoBam(5′GCTGGTAATGCCATGGGATCCGCTGGCGCGGCCGGC3′) and rphoXhoR(5′CGGTGTCTTGTTGTTCACGAGTTACGGTGCGCGCGGCCA3′) primers. rphoBam introduces an inframe BamHI site upstream of the first zinc finger(at amino acid 490) of Phol. rphoXhoR introduces an in-frame stop codon immediately downstream of the fourth zinc finger of Phol (amino acid 624)followed by a new XhoI site. The PCR product was subcloned downstream of the β-globin 5′ UTR and ATG in pT7link (provided by R. Treisman) to generate Pho-like_490-624T7. The integrity of the inserted PCR product was confirmed by sequencing.

³⁵S-labeled full-length Pho protein was translated in vitro from the Pho_2-520pT7 construct(Fritsch et al., 1999) and Phol from the Pho-like_490-624T7 construct using the TNT reticulocyte lysate system (Promega). Gel mobility shift assays were performed as described previously (Americo et al.,2002) using 3 μl of the in vitro translation reaction.

Virgin females from the stock EP0559 were crossed to males containing the immobilized transposase insertion, P[ry⁺(Δ2, 3)]99B(Robertson et al., 1988) to generate deletions of the P-element and flanking DNA. Males of the genotypew; EP0559/ry⁵⁰⁶ SbP[ry⁺(Δ2, 3)]99B were crossed to w; TM3,Ser/Sb males. Individual w^-, Ser male flies resulting from loss of the w⁺ marker present in the EP were crossed to w; TM3, Ser/Sb virgins and stocks were established that were EP0559(w^-)/TM3. DNA was made from homozygous EP0559(w^-) flies from 120 lines and checked for loss of the EP element using primers to the 5′ and 3′ ends of the P-element and flanking sequences. Lines that appeared to be missingphol DNA were subjected to further molecular analyses.

phol homozygous (Tb⁺) male larvae were collected from a stock that was pho¹/ci^D;phol^81A/TM6B, Tb. After removal of the salivary glands, DNA was prepared from the remaining carcass. The pho¹ mutation is associated with a DOC element insertion at codon 272 upstream of the zinc finger domain of the Pho protein (Brown et al., 1998). To identify homozygous pho¹ mutant larvae, we used primers 5′TTTGGCATTGATGGCTTCACG3′ and 5′GCATTGCAGATGAATCTCTGA3′ in a Long PCR reaction(Brown et al., 1998) with the DNA from individual larvae. The ci^D chromosome gives a 618 bp PCR product, while the pho¹ chromosome (which carries the DOC element insertion) generates a fragment in excess of 5 kb.pho¹ homozygotes produce only the larger PCR product. Thepho-like mutation was confirmed by the absence of a PCR product using the rpho14 (5′CGGTAGCCTCATCATCCTC3′) and rpho23(5′AGGGTTGCATTGTGG3′) primers.

Squashes were performed as described elsewhere(Zink et al., 1991), except incubation in solution 1 was for 30 seconds and in solution 2 was for 4.0-4.5 minutes. Slides were washed in PBS 10 minutes, incubated in blocking buffer(PBS, 5% BSA, 5% dried milk, 0.4% Tween 20) for 30 minutes, and incubated overnight at 4°C with primary antibody (αPc, 1:100; αScm,1:50; αPsc, 1:50; αE(z), 1:25; αPh, 1:500; αMs13,1:200; and αPho, 1:100) in blocking buffer. Slides were rinsed 30 minutes in PBS (adjusted to 300 mM NaCl with 0.4% Tween 20) then incubated with secondary antibody for 1 hour at room temperature. Secondary antibodies were FITC-, Cy2- or Cy3-labelled affinity-purified F(ab′)₂fragments (Jackson ImmunoResearch Labs). The slides were washed twice for 30 minutes in PBS (plus 0.1% Tween 20), stained with DAPI for 5 minutes, rinsed with PBS and mounted in 1 mg/ml phenylenediamine/70% glycerol in PBS. In some experiments, antibodies against Ms13, a male-specific protein that coats the male X (reviewed by Kelley and Kuroda, 2000), were used to identify unambiguously the X-chromosome.

Rabbit polyclonal antibodies were raised against a gel-purified HIS-tag/Pho full-length fusion protein. The crude polyclonal antisera specifically super-shifted a Pho/DNA complex in gel shift assays (data not shown). In western blots the antibody detects a strong band corresponding to Pho and a very weak band corresponding to Phol in 0-12 hour nuclear embryonic extracts(data not shown).

The following strains were used in this study (FRT2A is an abbreviation for P{w^+mW.hs=FRT(w^hs)}2A. FRT82Bis an abbreviation for P{ry^+t7.2=neoFRT}82B):

• pho¹/ci^D

• w; phol^81A FRT 2A/TM6B

• w; phol^81A/TM6B; pho¹/ci^D

• w; Trl^R85/TM6B

• z^v77h

• z^v77h; pho¹/ci^D

• w; FRT82B trx^E2/TM6C

• w; FRT82B trx^B11/TM6C

• yw hs-flp; M(3)ⁱ⁵⁵ hs-nGFP FRT2A

• ywhs-flp; FRT82B hs-nGFP

For analysis of phol^81A or Trl^R85mutant clones in pho¹ homozygotes, yw hs-flp/+;M(3)ⁱ⁵⁵ hs-nGFP FRT2A/TM6B; pho¹/+ virgins were crossed to w; phol^81A FRT2A/TM6B; pho¹/+ males or w; Trl^R85 FRT2A/TM6B; pho¹/+ males,respectively. In both cases, clones are marked by the absence of GFP andpho¹ homozygotes were identified by their misexpression of Ubx protein.

Mitotic clones were generated by crossing the appropriate fly strains listed above and heat-shocking the F₁ larvae. Heat shock treatment was for 1 hour in a 37°C water bath; larvae were then allowed to develop for 96 hours at 25°C. Prior to dissection, larvae were subjected to another 1 hour heat shock, followed by a 1 hour recovery period, to induce expression of the GFP marker protein.

Inverted larval carcasses were fixed and labelled with antibodies against Ubx or Abd-B or β-gal and, in the case of clonal analyses, double labelled with GFP antibody, followed by incubation with fluorescently labelled secondary antibodies as described (Beuchle et al., 2001). X-Gal staining was carried out as described(Christen and Bienz, 1994). Embryos were stained with β-gal antibodies and biotinylated secondary antibodies, followed by DAB staining using standard protocols.

BP01 females (Müller and Bienz, 1991) were crossed topho¹/ci^D males and BP01/+;pho¹/+ flies were inbred. pho¹ homozygotes were identified by the misexpression of the BP01 reporter gene in a quarter of the embryos. Mcp725-P[T8], an insertion on the third chromosome (Busturia et al.,1997), was recombined onto a Trl^R85 FRT2Achromosome; recombinants were selected and tested for the presence ofMcp725-P[T8], Trl^R85 and FRT2A. Males from aTrl^R85 FRT2A Mcp725-P[T8]/TM6B stock were crossed withyw flp122; M(3)i5 5hs-nGFP FRT2A virgins and clones in their progeny were analysed for misexpression of the Mcp725-P[T8] transgene using antibodies against β-gal and GFP.

The PRE_DGAGAmut reporter gene was obtained by mutating two to three nucleotides per GAGAG motive as indicated on top of the sequence in Fig. 7. Sixteen independent transformant lines were obtained and analysed.

phol is located on chromosome 3 in polytene subdivision 67B and is designated by the Drosophila genome project as CG3445. This sequence is predicted to encode a protein of 669 amino acids with four zinc fingers that share 80% sequence identity with the four zinc fingers of Pho(Fig. 1A). Although this is less conservation than between Pho and human YY1, which have 96% sequence identity over the zinc-finger region (Brown et al., 1998), all amino acids involved in making important DNA contacts (Houbaviy et al.,1996) are conserved in Phol. In addition, a short region of the spacer, conserved between Pho and YY1(Brown et al., 1998), is also conserved in Phol, although to a lesser extent(Fig. 1B). No other regions of conservation between Pho and Phol or between Phol and YY1 were detected.

As the amino acids contacting the DNA are identical in Pho and Phol, we expected Phol to have the same DNA-binding specificity as Pho. Gel shift assays with the Phol zinc-finger domain showed that this protein specifically bound an oligonucleotide containing a Pho-binding site. Binding was not competed by an oligonucleotide containing a mutated Pho-binding site(Fig. 1C,D). A gel shift using full-length Pho is shown for comparison. Pho and Phol can bind to the same DNA sequence with the same apparent binding specificity.

A Drosophila strain (EP0559) containing a P element insertion in the untranslated leader region of the phol transcription unit was obtained from the Drosophila genome project(Fig. 2A). Flies that are homozygous or hemizygous for this P-element insertion are viable and fertile. Two phol deletion alleles were generated by imprecise excision of the P-element (see Materials and Methods). In the phol^81A null mutation, part of the P-element was deleted along with the entirephol-coding region. In phol^106C, the EP0559 element was completely deleted along with 1389 bp of the pholtranscription unit, but leaving the phol promoter and the zinc-finger region intact. Therefore, it is possible that phol^106Cencodes a truncated Phol protein. Importantly, the flanking transcription unit, CG3448, is left intact in both alleles. In phol^81A,the deletion ends 846 bp upstream of the CG3448 mRNA.phol^106C and phol^81A are both homozygous and hemizygous viable; males are fertile but females are sterile. The female sterility of both mutant alleles is rescued by a pholtransgene (data not shown). Homozygotes for either phol allele look phenotypically normal and the mutants show no obvious homeotic phenotypes. Eggs laid by mothers that are homozygous for either phol allele look normal, are fertilized, but do not develop(Fig. 2B,C). Embryos derived from germline clones from heterozygous phol^106C andphol^81A mothers have the same phenotype showing thatphol is required in the germ cells (data not shown). The requirement for phol in the germline did not allow us to generate embryos that lack Phol protein and we therefore could not examine the role of pholin regulation of homeotic genes in embryos.

pho homozygotes die as pharate adults with weak homeotic transformations (Gehring,1970), while phol homozygotes survive and are phenotypically normal adults. By contrast, phol, pho double mutants die as third instar larvae and fail to pupate. Examination of phol,pho larvae showed that the brain is smaller than normal, the disks are misshapen and smaller than wild-type disks, and the salivary gland polytene chromosomes were enlarged (data not shown). The larger salivary gland polytene chromosomes may be due to additional rounds of endoreplication in the double mutants. To test whether phol functions in PcG repression, we examined Ubx and Abd-B expression in wing imaginal disks from single and double mutants of phol and pho(Fig. 3A). As expected, no Ubx or Abd-B expression was observed in wild-type or phol mutant wing disks. pho mutants showed misexpression of Ubx in a few cells in the wing pouch, but did not misexpress Abd-B(Fig. 3A)(Fritsch et al., 1999). By contrast, phol, pho double mutants strongly misexpress Ubx and Abd-B in the wing disk (Fig. 3A). This suggests that Phol and Pho redundantly repress homeotic genes in imaginal disks and can partially substitute for each other. We note that Ubxmisexpression is confined to the wing pouch in phol, pho double mutants; the lack of Ubx misexpression in more peripheral areas of the disk possibly reflects downregulation by Abd-B, which is strongly misexpressed in these regions of the disk(Fig. 3A).

We next tested whether removal of phol during larval development would also cause derepression of Ubx and Abd-B by generating clones of phol mutant cells in imaginal wing disks of phomutant larvae. In these experiments, phol mutant cells were identified by the absence of a GFP marker. Strong misexpression of Ubx and Abd-B was observed in double mutant cells in the wing pouch similar to the misexpression observed in wing disks from the phol, pho double mutant larvae (Fig. 3B and data not shown). These observations suggest that either Phol or Pho is required throughout development to keep homeotic genes repressed.

It has recently been suggested that pho may not play a role in PRE function in embryos (Poux et al.,2001a). This was surprising to us given previous reports showing misexpression of engrailed, abd-A and Abd-B in phomutant embryos (Moazed and O'Farrell,1992; Simon et al.,1992). In addition, the severe defects observed in embryos lacking maternal pho function suggested a strong requirement for Pho during oogenesis and/or embryogenesis (Breen and Duncan, 1986; Girton and Jeon,1994). We therefore re-examined the role of pho in embryos by testing the requirement for pho and Pho binding sites in the regulation of PRE-containing reporter genes.

We did not see any additional mis-expression of Ubx orAdbB in phol, pho double mutant embryos over what was seen in pho single mutants. Thus, we have conducted our embryonic experiment in pho single mutants. We looked at lacZexpression from a Pbx-Bxd-Ubx-lacZ (BP01) reporter gene(Müller and Bienz, 1991). This reporter is derepressed in Pc mutant embryos(Müller and Bienz, 1991). Similarly, we found that it is derepressed in a pho mutant(Fig. 3C). This shows that Pho protein is required for the silencing of this reporter gene in the embryo. Next, we looked at whether mutation of the Pho-binding sites within a PRE disrupts silencing. We used a construct containing PRE_D, a 567 bp fragment from the Ubx gene. We have previously shown that mutation of Pho-binding sites in PRE_D inactivated its silencing capability in imaginal disks(Fritsch et al., 1999). Poux et al. (Poux et al., 2001a)reported that mutation of Pho-binding sites did not cause a loss ofPRE_D silencing in embryos. However, we obtained different results using the same lines. We looked at expression from three wild-typePRE_D lines and five PRE_DPhomut. All wild-type PRE_D lines gave the expression pattern shown. Two out of five PRE_DPhomut lines gave expression similar to that shown in Fig. 3C,including two out of three lines examined by Poux et al.(Poux et al., 2001a). A thirdPRE_DPhomut line also showed unrestricted expression in embryos but the levels were lower compared with the other lines. A fourth line showed no silencing in the embryonic epidermis and in discs, but maintained restricted expression in the embryonic CNS. A fifth line showed restricted expression similar to the wild-type PRE_D control lines. These results show that Pho protein and Pho-binding sites do play a role in repression during embryogenesis.

The experiments described above suggest that the DNA-binding proteins Pho and Phol play important and redundant roles in PcG repression. One possible role of these two proteins may be to anchor other PcG proteins to PREs. To test this hypothesis, we analysed binding of five different PcG proteins to polytene chromosomes in phol, pho double mutants.

First, we examined the localization of Pho proteins on polytene chromosomes of wild-type larvae. We previously reported binding of Pho to about 35 chromosomal sites (Fritsch et al.,1999). Using a new Pho antiserum, combined with immunofluorescent techniques, we now detect Pho binding to about 100 sites on polytene chromosomes (Fig. 4). Psc colocalizes with Pho at about 65% of these sites(Fig. 4). Psc has also been reported to bind to 65% of the Pc sites(Rastelli et al., 1993).

Next, we looked at the distribution of the PcG proteins Pc, Psc,Polyhomeotic (Ph), Sex combs on midleg (Scm) and Enhancer of zeste (E(z)) on polytene chromosomes. Pc, Ph and Psc are all core components of the PcG protein complex called PRC1 (Shao et al.,1999; Saurin et al.,2001; Francis and Kingston,2001). Scm has also been reported to co-purify with PRC1(Shao et al., 1999). Scm and Ph may also be present in protein complexes other than PRC1(Roseman et al., 2001;Hodgson et al., 2001). E(z) is a component of the Esc-E(z) complex, which is distinct from PRC1(Ng et al., 2000;Tie et al., 2001). We focused our analysis on PcG protein binding sites on the X chromosome and on the right arm of chromosome 3, which includes the bithorax and Antennapedia gene complexes (BXC and ANTC).

Fig. 5A shows binding of Pho, Pc, Psc, Ph and Scm to chromosomal sites in the distal region of the X chromosome in wild type and phol, pho double mutants. Three sites that are bound by all five PcG proteins in wild-type chromosomes are indicated. As expected, in phol, pho double mutants, no Pho protein is detected (Fig. 5A). Binding of Pc, Psc and Scm is lost at polytene subdivision 2D(Fig. 5A, arrow) in phol,pho double mutants; binding of these proteins to all other sites on the X chromosome is unaffected (Fig. 5A and data not shown). Binding of Ph is completely unaffected inphol, pho double mutants. In particular, the Ph signal at 2D is present, suggesting that Ph can bind at this site even if other PcG proteins are removed. We also find that Pc binding to 2D is not lost in eitherpho or phol single mutants(Fig. 5C), suggesting that the presence of either of these two proteins is sufficient for Pc to bind to this site.

We were particularly interested in knowing whether E(z) protein distribution would change in the double mutants because the vertebrate homologues of Pho and Esc interact in in vitro binding experiments (Satijin et al., 2001) and Pho co-immunoprecipates with Esc in early embryos(Poux et al., 2001b). One attractive hypothesis is that Pho might be required for the binding of E(z)/Esc protein complexes to chromatin. However, we did not detect changes in any E(z) chromosomal sites on either the X chromosome(Fig. 5B) or on 3R in phol,pho double mutants (data not shown). It was reported that E(z) bound to chromosomal subdivision 2D (Carrington and Jones, 1996); however, we were able to detect E(z) at this site on only about 20% of the wild-type chromosomes. Although we never saw E(z) at 2D on phol, phol double mutant chromosomes, we cannot definitely conclude there is a difference between this and wild type.

The patterns of binding of Psc, Ph, Scm and E(z) proteins on chromosome arm 3R were indistinguishable in wild type and phol, pho double mutants(data not shown). In particular, these PcG proteins were still bound to regions that include the BXC and ANTC loci in phol, pho double mutants. The binding of Pc to the BXC and ANTC, and most other loci was also unaltered in the double mutant, but we found that binding to two specific chromosomal sites was lost (Fig. 6). Interestingly, Psc, Scm and E(z) were not detected at these sites on wild-type chromosomes (data not shown)(Rastelli et al., 1993;Carrington and Jones, 1996),suggesting that Pc binds independently of these proteins at these sites. Ph was present at one of these two sites, but its binding was not altered inphol, pho double mutants (data not shown).

Taken together, the immunolocalization data suggest that binding of PcG proteins to most sites is unaltered in the absence of Pho and Phol protein,but that these two proteins are redundantly required for PcG protein binding at a few specific sites. Intriguingly, it appears that all PcG proteins tested here are still associated with the BXC and ANTC loci. Nevertheless, we found that the BXC genes Ubx and Abd-B were derepressed inphol, pho double mutant wing disks. We propose several different explanations for this paradox. First, derepression of homeotic genes and binding of PcG proteins were not assayed in the same tissues. We were not able to detect derepression of Ubx in salivary gland cells of phol,pho double mutants (data not shown). Second, Pho and Phol may only be required for anchoring PcG proteins at some PREs in the BXC. Different DNA-binding proteins may provide this function at other PREs. This is supported by our finding that binding of PcG proteins is lost at some sites inphol, pho double mutants (see Figs5,6). Moreover, several different PREs have been identified in the Ubx gene(Chan et al., 1994;Christen and Bienz, 1994;Müller, 1995;Chang et al., 1995;Orlando et al., 1998;Hodgson et al., 2001). The resolution of antibody signals on polytene chromosomes is not refined enough to resolve distinct PREs in a single gene and, hence, loss of only a fraction of PcG protein complexes may not be detectable. Finally, Phol and Pho may not be necessary for the anchoring of PcG protein complexes to the DNA, but may confer the actual transcriptional repression mediated by PREs in imaginal disks, while the PcG protein complexes might function in the propagation and memory of the repression. Thus, PcG protein complexes might serve to recruit Phol and Pho or their co-repressors to the DNA.

The GAGA factor protein is encoded by the Trithorax-like(Trl) gene. A hypomorphic Trl allele was originally isolated due to mutant phenotypes that suggested a requirement for activation of homeotic gene expression (Farkas et al.,1994). The proposal that Trl functions in PcG repression was based on the observations that Trl protein bound to PRE sequences, that it co-immunoprecipitated with Pc, and that mutation of Trl-binding sites caused a loss of mini-white silencing and PRE function in reporter genes(Horard et al., 2000;Mishra et al., 2001;Busturia et al., 2001;Americo et al., 2002;Poux et al., 2002). These conflicting data prompted us to analyse the role of Trl in homeotic gene regulation by generating clones of Trl mutant cells in imaginal disks. In these experiments, we used Trl^R85, a null allele(Farkas et al., 1994), and the mutant cells were again marked by the absence of a GFP marker protein.

We first analysed Trl mutant clones in the wing disk for misexpression of Ubx and Abd-B and found no evidence for such misexpression (Fig. 7A and data not shown). As PREs often contain Pho- and Trl-binding sites in close proximity, and Busturia et al. (Busturia et al., 2001) reported a weak genetic interaction betweenpho and Trl heterozygous mutants, we tested whether removal of Trl in pho mutant wing disks would exacerbate the misexpression of Ubx observed in pho mutants. This was not the case. pho mutant wing disks with clones of Trlhomozygous cells showed no additional misexpression of Ubx compared with pho single mutants (compareFig. 7A withFig. 3A). Thus, we find no evidence for a genetic interaction between Trl and pho.

We also analysed the effects of removing Trl on the silencing capabilities of two different PRE-containing Ubx-LacZ reporter transgenes; PRE_1.6 contains a PRE from the Ubxgene (Fritsch et al., 1999)and MCP725 contains a PRE from the Abd-B gene(Busturia et al., 1997). In wild-type flies, expression of both transgenes was confined to the posterior compartments of the haltere and third leg disks, and both transgenes were misexpressed in a variety of PcG mutants(Busturia et al., 1997;Fritsch et al., 1999) (M. Bakala, Diploma thesis, University of Tübingen, 2001). By contrast, we observed no misexpression of either transgene in Trl mutant clones in wing imaginal disks (Fig. 7B).

We also tested whether mutation of Trl protein binding sites (i.e. GAGAG sequences) in a PRE from the Ubx gene would compromise its silencing capability. For this experiment we used a previously described reporter gene,PRE_D, that is stably silenced in the wing imaginal disk due to the presence of the 567 bp long PRE core fragment(Fig. 7C)(Fritsch et al., 1999). Previous studies showed that mutation of Pho protein-binding sites withinPRE_D abolished repression of this reporter transgene in wing imaginal disks (Fig. 7C)(Fritsch et al., 1999). By contrast, mutation of all five GAGAG motifs in PRE_D caused no misexpression of this reporter transgene(Fig. 7C). Sixteen lines were obtained, five produced expression caused by positional effects and could not be analysed. The other eleven all showed silencing in the wing disk similar to that shown in Fig. 7C.

Finally, we tested the requirement for Trl in maintaining expression of Ubx and Abd-B in their normal expression domains. Intriguingly, we observed no obvious reduction of Ubx orAbd-B expression in Trl mutant clones in the haltere and third leg disk (Ubx) or in the genital disk (Abd-B)(Fig. 7D and data not shown). By contrast, clones of trithorax (trx) mutant cells showed a dramatic reduction in Ubx protein levels(Fig. 7D).

These results fail to support a role for Trl in PcG repression in imaginal disks. However, we cannot exclude the possibility that Trlis playing a role in the establishment of PcG repression in the embryo. The requirement for Trl function in the germline and the early embryo(Liaw et al., 1995;Bhat et al., 1996) does not allow an analysis of embryos lacking Trl protein.

Another protein that has been proposed to function in PcG repression is Zeste (Hur et al., 2002).zeste (z) null mutants are viable and fertile and show no obvious homeotic phenotypes (Goldberg et al., 1989). However, Zeste co-purifies with the PcG protein complex PRC1 (Saurin et al.,2001) and Zeste protein binding sites have been implicated in PcG function of an embryonic reporter gene(Hur et al., 2002). To test whether z might interact genetically with pho in repression of homeotic genes, we examined the expression of Ubx inz^v77h and in z^v77h, pho¹double mutant wing disks. Both z^v77h andpho¹ are presumed null mutants(Goldberg et al., 1989;Brown et al., 1998). zmutant wing disks showed no misexpression of Ubx(Fig. 8) and z, phodouble mutants showed no more misexpression of Ubx than that seen inpho mutant wing disks (compareFig. 3A withFig. 8).

Our results show a strong requirement for the DNA-binding proteins Pho and Pho-like in homeotic gene silencing in imaginal disks. In fact, the strong misexpression of homeotic genes observed in phol, pho double mutant imaginal cells is comparable with that seen in imaginal disk clones mutant forPc, Scm, Sce or Pcl(Beuchle et al., 2001). The loss of PcG protein binding at only a small number of sites in phol,pho polytene chromosomes is consistent with the idea that Phol and Pho are required to recruit PcG protein complexes at only a subset of PREs. Alternatively, Phol and Pho may be required for transcriptional repression mediated by PREs, but not for anchoring of PcG protein complexes.

We are indebted to many people for the generous gifts of antibodies: Pat O'Farrell and Renato Paro for anti-Pc; Jeff Simon for anti-Scm; Donna Arndt-Jovin for anti-Ph; Paul Adler for anti-Psc; Rick Jones for anti-E(z);Tim Karr for anti-sperm tail (DROP 1.1); and Richard Kelley and Mitzi Kudora for anti-Msl3. We thank Judy Leatherman and Tom Jongens for an open exchange of information and reagents on phol. We thank Jim Kennison for stimulating discussions during the course of this work and comments on this manuscript. We also thank a reviewer for helpful suggestions.