Su(z)12, a novel Drosophila Polycomb group gene that is conserved in vertebrates and plants

In both flies and vertebrates, HOX genes are expressed in spatially restricted patterns to control development of the head, trunk and limbs (Lewis, 1963; Lewis, 1978; McGinnis and Krumlauf, 1992). HOX genes have the potential to be active both within and outside of their correct expression domains, but each HOX gene is selectively silenced in cells where it must remain inactive. This silencing depends on the function of the Polycomb group (PcG) genes (Kennison, 1995; Simon, 1995; Pirrotta, 1998). PcG genes were first identified in Drosophila due to mutant phenotypes that suggested that their products function in repressing multiple HOX genes (Lewis, 1978; Struhl, 1981; Duncan, 1982; Ingham, 1984; Jürgens, 1985). Subsequent studies showed that mutations in 11 PcG genes indeed cause misexpression of HOX genes in embryos and in larvae. These genes are Polycomb (Pc), Polycomblike (Pcl), polyhomeotic (ph), Posterior sex combs (Psc), Sex combs on midleg (Scm), Sex combs extra (Sce), super sex combs (sxc), pleiohomeotic (pho), Enhancer of zeste (E(z)), extra sex combs (esc) and Additional sex combs (Asx) (Beachy et al., 1985; Ingham, 1985; Struhl and Akam, 1985; White and Wilcox, 1985; Cabrera et al., 1985; McKeon and Brock, 1991; Simon et al., 1992, Fritsch et al., 1999; Beuchle et al., 2001). Most PcG proteins are conserved in both sequence and function in vertebrates (Brunk et al., 1991; van Lohuizen et al., 1991; Pearce et al., 1992; van der Lugt et al., 1994; Müller et al., 1995; Schumacher et al., 1996; reviewed in van Lohuizen, 1998). However, only E(z) and esc appear to be more widely conserved. In C. elegans, sequence homologues of E(z) and esc function in germline development but, strikingly, they apparently have no role in the regulation of homeotic genes (Holdeman et al., 1998; Korf et al., 1998; Kelly and Fire, 1998). However, in Arabidopsis, the E(z)-like protein CURLY LEAF (CLF) functions as a transcriptional repressor of floral homeotic genes in vegetative tissues (Goodrich et al., 1997). This conservation of Polycomb group gene function in plants is particularly striking since HOX genes in animals and homeotic genes in plants are structurally unrelated (McGinnis et al., 1984; Yanofsky et al., 1990).

In Drosophila, all PcG genes are expressed in the female germline and maternally deposited wild-type protein often rescues homozygous mutant embryos to a considerable extent (Struhl, 1981; Breen and Duncan, 1986; Soto et al., 1995). Embryos that are doubly homozygous for mutations in two different PcG genes typically show strongly enhanced homeotic transformations, and the phenotype of such embryos is often similar to the null phenotype of the corresponding single mutants (i.e., lacking both maternal and zygotic gene function). Jürgens (1985) used this striking property and generated embryos that were doubly homozygous for PcG mutations and large chromosomal deficiencies. From these tests he estimated that the total number of PcG genes in the Drosophila genome would be in the range of 30 to 40 genes (Jürgens, 1985). Although this number is frequently cited, only two Drosophila genes with bona fide PcG mutant phenotypes have been described since Jürgens’ original proposal 15 years ago. These are multi-sex combs (mxc; Santamaria and Randsholt, 1995) and cramped (crm; Yamamoto et al., 1997).

We report the mutant phenotypes and molecular analysis of a new PcG member, Suppressor of zeste 12 (Su(z)12). Su(z)12 mutants show very strong homeotic phenotypes caused by widespread misexpression of HOX genes. The phenotypes of Su(z)12 mutants are comparable to those of the strongest PcG mutants. However, our analyses of Su(z)12 also reveal some striking properties that distinguish this gene from most other PcG genes; Su(z)12 function is needed for the development of germ cells and Su(z)12 loss-of-function mutations suppress PEV. Moreover, Su(z)12 is not only conserved in vertebrates, but is also related to Arabidopsis proteins that function as regulators of floral homeotic genes and other developmental processes.

The four EMS-induced Su(z)12 alleles, Su(z)123, Su(z)122, Su(z)124 and Su(z)125 were originally called l(3)76BDo1, l(3)76BDo2, l(3)76BDo4 and l(3)76BDo5, respectively, and their isolation as mutations that fail to complement Df(3L)kto2 has been described (Kehle et al., 1998). The Su(z)121 allele was isolated in a P-element mutagenesis screen for mutations that modify the eye color of z1 mutants. Mutagenized males were mated to z1 Dp(1;1)w z+R61e19 virgin females and the male offspring were screened for suppression or enhancement of the yellow eye color. One red-eyed male was isolated and subsequently crossed with balancers to establish a stock. Of several revertants of Su(z)121, obtained by mobilizing the P element, we isolated five that were viable and fertile in trans to Su(z)121.

All Su(z)12 alleles were recombined onto an FRT2A chromosome to obtain the following strains:

w; Su(z)121 FRT2A/ TM6C, cu Sb e Tb ca

w; Su(z)122 FRT2A/ TM6C, cu Sb e Tb ca

w; Su(z)123 FRT2A/ TM6C, cu Sb e Tb ca

w; Su(z)124 FRT2A/ TM6C, cu Sb e Tb ca

w; Su(z)125 FRT2A/ TM6C, cu Sb e Tb ca

Germline clones of each mutant were generated using the standard ovoD technique. No eggs from germ-line clones were obtained in the case of Su(z)123 and Su(z)124. To ‘clean’ the left arm of the Su(z)123 and Su(z)124 chromosomes from other potential lethal mutations, we substituted most of the chromosome arm distal to the Su(z)12 locus with DNA from a homozygous viable ru h th st cu sr e ca chromosome. Four independent ru h th Su(z)123 FRT2A and six independent ru h th Su(z)124 FRT2A recombinant chromosomes were isolated and tested for production of germline clones but no eggs were obtained in either case.

Imaginal disc clones were generated by crossing the appropriate Su(z)12 FRT2A mutant strains with either yw flp122; hs-nGFP FRT2A or yw flp122; M(3)i55hs-nGFP FRT2A/TM6B flies and heat-shocking the F₁ larvae. Heat shock treatment to induce clones was done in vials for 1 hour in a 37°C water bath, and the larvae were then allowed to develop for the appropriate time 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.

Effects on PEV were analysed by crossing Su(z)12 mutant males to In(1)wm4 females and comparing the eye phenotypes of the In(1)wm4; Su(z)12/+ and In(1)wm4; Balancer/+ male progeny.

Antibody staining of embryos with antibody against Ubx protein was done following standard protocols. Imaginal discs were stained with antibodies against Ubx or Abd-B and GFP proteins as described (Beuchle et al., 2001).

A LAMBDA library of EcoRI digested genomic DNA was generated from Su(z)121 heterozygotes and screened using P-element sequences as probe. A subclone containing a 2.2 kb insert was isolated; this insert contained P-element sequences and 1.6 kb of flanking genomic DNA. This genomic fragment was used as a probe to isolate a larger genomic fragment from an EMBL 4 library and, with that as a probe, cDNAs for three different transcription units were isolated from an embryonic cDNA library (Clontech). Northern blot analysis revealed that one of these transcripts showed an altered pattern in Su(z)121 mutants. Two EST clones with 5′ sequences identical to this cDNA were obtained from the Berkeley Drosophila Genome Project (LD13365 and LD02025). LD02025 was sequenced and LD13365 was partially sequenced. Introns were mapped by use of internal primers, PCR amplification and sequencing. The Su(z)12 gene was mapped to 76D using a digoxigenin-labeled probe for in situ hybridisation on polytene chromosomes from salivary glands.

Sequencing of the EMS-induced Su(z)12 alleles was done as follows. Genomic DNA was isolated from Su(z)123 or Su(z)124 heterozygotes or, in the case of Su(z)122, from Su(z)122 homozygous larvae that were identified by the red marker mutation on the mutant chromosome. In each case, the genomic DNA spanning the Su(z)12 open reading frame was amplified by PCR. Three overlapping subfragments covering this interval were amplified, subcloned into bluescript and several independent clones were sequenced. In each mutant, only a single base change was found in several independent clones (Fig. 5). For each mutant allele the identified base changes were confirmed by sequencing clones obtained from a second, independent PCR amplification.

In a screen for zygotic-lethal mutations in the 76D region (Kehle et al., 1998), we previously identified a lethal complementation group, l(3)76BDo, that turned out to be allelic to Su(z)121, a P-element-induced mutation that was isolated in a screen for modifiers of the zeste-white interaction (see Materials and Methods). Since Su(z)121 fails to complement any of the four EMS-induced l(3)76BDo alleles we have renamed l(3)76BDo1, l(3)76BDo2, l(3)76BDo4 and l(3)76BDo5 (Kehle et al., 1998) as Su(z)123, Su(z)122, Su(z)124 and Su(z)125, respectively.

Animals that are homozygous or hemizygous for Su(z)121, Su(z)122, Su(z)123 or Su(z)124 die during the first or second larval instar, whereas several transheterozygous combinations with Su(z)125 develop into pharate adults with strong posteriorly directed homeotic transformations (Fig. 1). These homeotic transformations are consistent with inappropriate activation of several HOX genes in the Antennapedia and bithorax complexes. For example, the additional sex combs on meso- and metathoracic legs suggests misexpression of Sex combs reduced (Scr) in these primordia (Pattatucci and Kaufman, 1991), whereas the antenna to leg transformation is consistent with inappropriate activation of Antennapedia (Antp) in the eye-antennal disc (Struhl, 1981) and the wing to haltere transformations most likely reflects misexpression of BXC genes in the wing disc (Cabrera et al., 1985; Fig. 1). This suggests that Su(z)12 acts as a repressor of several homeotic genes and is a member of the PcG.

Since Su(z)121, Su(z)122, Su(z)123 and Su(z)124 homozygotes die as larvae without any obvious Pc-like phenotypes, we next analyzed Su(z)12 mutant embryos for misexpression of HOX genes (Fig. 2). Embryos that are homozygous for any of the four EMS-induced Su(z)12 alleles (Su(z)122-5) show very subtle misexpression of Ubx; similar misexpression is observed in embryos that are homozygous for Df(3L)kto2, a deficiency that deletes the Su(z)12 locus (Fig. 2 and data not shown). Interestingly, Su(z)121 homozygotes show substantially more misexpression than embryos that are homozygous for the deficiency or for any of the other alleles (Fig. 2; see below). A likely explanation for these subtle Pc-like phenotypes is that maternally deposited wild-type Su(z)12 product rescues these Su(z)12 mutant embryos.

We therefore attempted to generate embryos from Su(z)12 mutant germ cells. Females with Su(z)12 mutant germ cells were crossed to males heterozygous for Df(3L)kto2. We found that Su(z)12 hemizygous embryos derived from Su(z)122 or Su(z)125 germ cells showed very extensive misexpression of Ubx already at the extended germ band stage (Fig. 2 and data not shown). These animals showed severe homeotic phenotypes with all abdominal, thoracic and several head segments transformed into copies of the eight abdominal segment. This phenotype is consistent with Abd-B being misexpressed in all segments (Fig. 2 and data not shown). The strong PcG phenotype of these Su(z)12 mutant embryos is comparable to that of embryos lacking esc or Pc function (Struhl, 1981; Lawrence et al., 1983). We find that zygotically provided Su(z)12 function is sufficient to prevent the inappropriate activation of HOX genes; Su(z)122/+ heterozygotes obtained as the progeny of Su(z)12 mutant germ cells and a wild-type sperm develop into wild-type-looking adults.

In contrast to Su(z)122 or Su(z)125, we found that germ cells mutant for any of the other three Su(z)12 alleles failed to develop (Su(z)123 and Su(z)124) or developed into highly abnormal eggs (Su(z)121). Two observations suggest that the failure to obtain embryos in the case of Su(z)121 and Su(z)124 is not caused by second-site mutations on the Su(z)12 mutant chromosomes but can be attributed to a requirement for Su(z)12 function in germ-cell development. First, we found that revertants obtained by excision of the P element in the Su(z)121 allele are viable and fertile. Second, Su(z)124 mutant germ cells still failed to develop even after “cleaning” the chromosome from other potentially lethal mutations by replacing the chromosomal DNA flanking this Su(z)12 allele with unmutagenized wild-type DNA (see Materials and Methods). Hence, these results suggest that Su(z)12 function is essential for the development of germ cells. Furthermore, Su(z)121, Su(z)123 and Su(z)124 are strong alleles and Su(z)122 and Su(z)125 are weaker alleles (see below). The fact that Su(z)121 homozygous embryos show more severe misexpression than Df(3L)kto2 homozygotes suggests that Su(z)121 is not a simple loss-of-function allele but is an antimorphic allele that encodes a product that interferes with the function of maternally deposited, wild-type Su(z)12 protein. We note that Su(z)121/+ embryos show no misexpression of homeotic genes in the embryo (not shown).

We next tested the requirement for Su(z)12 at later developmental stages by generating Su(z)12 mutant clones in imaginal discs. We assayed for HOX gene silencing in such clones by monitoring the expression of the HOX genes Ubx and Abd-B in the imaginal wing disc (where they are normally stably repressed) using antisera against their protein products. In these experiments, the Su(z)12 mutant cells were identified by the absence of a GFP-expressing marker gene (see Materials and Methods). In addition, we used the Minute technique to generate Su(z)12-/Su(z)12-clones that carry two copies of a wild-type Minute allele (i.e., Su(z)12- M+/ Su(z)12- M+), which gives them a growth advantage relative to their Su(z)12- M+/ Su(z)12+ M- neighbours.

In a first set of experiments, we analyzed cell clones of the different Su(z)12 alleles 96 hours after clone induction. We found that Su(z)121 and Su(z)124 mutant clones showed strong misexpression of both Ubx and Abd-B in most mutant cells (Fig. 3A). Su(z)122 mutant clones also showed misexpression of Ubx 96 hours after clone induction but misexpression is confined to the pouch and hinge region in the posterior compartment of the wing disc (Fig. 3A). No misexpression of Abd-B was detected in Su(z)12 2 mutant clones and neither Ubx nor Abd-B were misexpressed in Su(z)125 mutant clones (Fig. 3A). We also found no misexpression in Su(z)123 mutant clones but we found that these clones were much smaller than those obtained with the other Su(z)12 alleles (data not shown). We do not know whether the cell proliferation/survival defect associated with the Su(z)123 chromosome is caused by a second mutation in a closely linked gene (see Materials and Methods) or is a unique property of this particular allele. In summary, the PcG phenotypes observed with several Su(z)12 alleles suggest that Su(z)12 is needed throughout development to keep HOX genes repressed. Moreover, these results support the allele classification obtained by the analysis of germ-line clones; namely, that Su(z)122 and Su(z)125 are hypomorphic alleles whereas Su(z)121 and Su(z)124 appear to be stronger alleles.

We next examined the kinetics of HOX gene derepression in Su(z)121 and Su(z)124 mutant clones by analyzing Ubx expression 24, 48 and 72 hours after clone induction. We again used the Minute technique in these experiments. 24 hours after clone induction, Ubx is still tightly repressed. 48 hours after clone induction, Su(z)121 mutant clones show misexpression of Ubx protein in the wing pouch but Ubx is still stably silenced in other parts of the wing disc (Fig. 3B). In Su(z)124 mutant clones, Ubx is still stably silenced 48 hours after clone induction except in a few clones in the center of the pouch where we detect weak Ubx signal (Fig. 3B). Finally, 72 hours after clone induction, repression of Ubx is lost in most Su(z)121 and Su(z)124 mutant clones in the pouch, in the latter case Ubx is still silenced in some parts of the disc (Fig. 3B). This slow and gradual loss of silencing is comparable to the kinetics of HOX gene derepression in Pc, Pcl, Scm or Sce mutant clones (Beuchle et al., 2001).

We note that the loss of silencing occurs more rapidly in Su(z)121 clones than in Su(z)124 clones (Fig. 3). The molecular characterization of Su(z)124 suggests that this is most likely a null allele (see below). Our analysis of Su(z)121 homozygous embryos suggested that Su(z)121 is not a simple loss-of-function allele but is an antimorphic allele (see above). It is possible that the more rapid loss of silencing in Su(z)121 mutant clones again reflects an interference of the mutant Su(z)121 product with wild-type Su(z)12 molecules (i.e., during the depletion of persisting wild-type Su(z)12 protein after clone induction).

To test whether Su(z)12 may also participate in other processes of transcriptional silencing, we tested whether Su(z)12 mutations suppress position-effect variegation (PEV). PEV is observed in chromosomal rearrangements in which a euchromatic gene is placed near heterochromatin. The translocated gene may then become inactivated in a fraction of cells, presumably because transcription of the gene is silenced by heterochromatin-associated proteins. A number of mutations have been identified that suppress or enhance PEV in a dosage-dependent fashion (reviewed by Wakimoto, 1998). Mutations that suppress PEV are generally referred to as Su(var)s; some Su(var) gene products have indeed been shown to be components of heterochromatin (Eissenberg et al., 1990). One well-studied reporter for PEV is wm4, a chromosomal inversion juxtaposing the white gene to centromeric heterochromatin (Muller, 1930). As illustrated in Fig. 4, mutations in Su(z)12 strongly suppress PEV at the wm4 locus; in animals that are heterozygous for any of the five Su(z)12 alleles, the white locus is transcriptionally active in a higher proportion of ommatidia than in control animals. We note that suppression of PEV was observed with four different EMS-induced Su(z)12 alleles but not with most other mutations that were isolated in the same EMS-mutagenesis experiments, suggesting that suppression of PEV is indeed due to the mutations at the Su(z)12 locus and not due to other PEV modifiers on the mutagenized chromosomes (data not shown). These results suggest that Su(z)12 can be classified as a suppressor of PEV.

To identify the Su(z)12 gene we isolated and cloned the chromosomal DNA flanking the P-element insertion in the Su(z)121 allele (see Materials and Methods). Using this genomic DNA as probe we isolated several cDNAs by screening a cDNA library and by searching an EST database for cDNAs that match this genomic DNA. Sequence analysis of the longest cDNA (LD 02025) revealed a single open reading frame of 900 amino acids. The P-element in Su(z)121 is inserted into codon 564 of this open reading frame and would therefore result in a C-terminal truncation of the predicted protein (Fig. 5). Further proof that the identified open reading frame encodes the Su(z)12 protein was obtained by sequencing the coding region of three EMS-induced Su(z)12 alleles. Su(z)122, Su(z)123 and Su(z)124 each show single base changes in this open reading frame; Su(z)123 and Su(z)124 both show base substitutions that result in predicted premature termination codons after codon 218 and codon 298, respectively, and in Su(z)122, a base substitution changes the codon for Gly²⁷⁴ into a codon for Asp (Fig. 5; this glycine is conserved in the human Su(z)12 homologue described below). The molecular characterizations of these lesions prove that we identified the Su(z)12 open reading frame. Furthermore, these lesions support our classification of the different alleles based on phenotypic criteria; Su(z)122 and Su(z)124 both have stop codons in the N-terminal third of the protein and therefore may represent null (or at least strong loss-of-function) alleles whereas the amino-acid substitution in Su(z)122 is consistent with the idea that this is a hypomorphic allele.

Database searches show that the Su(z)12 protein is highly conserved in vertebrates and, strikingly, that Su(z)12-related proteins also exist in plants (Fig. 5). In contrast, the worm and yeast genomes do not seem to encode Su(z)12-related proteins. The function of the highly conserved human homologue of Su(z)12, HsSU(Z)12 (Fig. 5), is not known but EMF2, FIS2 and VRN2, the three Su(z)12-related proteins in Arabidopsis, have been identified as regulators in plant development (Yang et al., 1995; Luo et al., 1999; N. Yoshida, personal communication; A. R. Gendall, personal communication). One characteristic feature of all these proteins is a single classical C₂H₂ zinc finger similar to the fingers found in sequence-specific DNA-binding proteins (Fig. 5). Attempts to show any DNA-binding activity of a polypeptide containing the Su(z)12 zinc finger have failed so far (A. K. S. and J. M., unpublished). A second stretch of amino acids that is conserved between Su(z)12, HsSU(Z)12, EMF2, VRN2 and FIS2 is located C-terminal to the zinc finger (Fig. 5). We term this part of the protein VEFS box (VRN2-EMF2-FIS2-Su(z)12 box). We note that the predicted protein products encoded by Su(z)123 and Su(z)124 lack both the zinc finger and the VEFS box, whereas the protein encoded by Su(z)121 would contain the zinc finger but lack the VEFS box.

We report the mutant phenotypes and cloning of Su(z)12, a novel Drosophila gene. The most notable phenotypes of Su(z)12 mutants are strong homeotic transformations caused by the widespread misexpression of several HOX genes in embryos and in larvae. These phenotypes clearly classify Su(z)12 as a PcG gene and our clonal analyses show that Su(z)12 plays an essential role in HOX gene silencing throughout development. However, a number of properties distinguish Su(z)12 from most other Drosophila PcG genes and we shall discuss these in turn.

Our genetic and molecular analyses suggest that Su(z)123 and Su(z)124 are most likely null or at least strong loss-of-function alleles, whereas Su(z)122 and Su(z)125 appear to be hypomorphic alleles. Su(z)121 is also a strong loss-of-function allele but in addition, it also shows properties of an antimorphic allele. Since Su(z)123 may contain a second, cell-lethal mutation, we will omit this allele for discussion of the Su(z)12 mutant phenotype and presume that the phenotype of Su(z)124 represents the Su(z)12 null phenotype. The analysis of Su(z)121 and Su(z)124 germline clones suggests that Su(z)12 function is essential for germ-cell development and only germ cells carrying hypomorphic Su(z)12 mutations develop into embryos. By contrast, germ cells mutant for most other PcG members complete oogenesis (Struhl, 1981; Lawrence et al., 1983; Breen and Duncan, 1986; Soto et al., 1995) and only E(z), crm and mxc seem to be required for germ cell development (Phillips and Shearn, 1990; Yamamoto et al., 1997; Saget et al., 1998). Although we do not know which processes in germ-cell development require Su(z)12 function, the requirement for Su(z)12 in the germline clearly distinguishes Su(z)12 from most other PcG genes.

A second distinction between Su(z)12 and most other PcG mutants is suggested by the suppression of PEV in Su(z)12 mutants. Heterochromatin-mediated silencing has often been compared to HOX gene silencing (e.g. Paro, 1990; Pirrotta and Rastelli, 1994). Although the two processes may use similar molecular mechanisms, they require two distinct sets of proteins; Su(var) mutants show no PcG phenotypes and most PcG mutations do not suppress wm4 variegation (Kennison, 1995; Sinclair et al., 1998). Among the exceptions (besides Su(z)12), mutations in the PcG gene crm suppress wm4 variegation (Yamamoto et al., 1997). E(Pc) mutations also suppress wm4 variegation (Kennison, 1995; Sinclair et al., 1998), but it is not clear whether E(Pc) is a PcG gene (Soto et al., 1995; Sinclair et al., 1998). Finally, E(z) mutations have been reported to weakly suppress wm4 variegation (Laible et al., 1997) or to enhance it (Sinclair et al., 1998). Although we favour the interpretation that Su(z)12 protein functions directly in heterochromatin-mediated gene silencing, (e.g. as a component of heterochromatin), we cannot exclude the possibility that the effect on PEV is indirect.

A third, striking feature of Su(z)12 is its conservation not only in vertebrates but also in plants. Most Drosophila PcG proteins have vertebrate homologues and studies on PcG mutant mice showed that these proteins are needed to repress HOX gene transcription outside of the normal HOX expression domains (reviewed by van Lohuizen, 1998). The Su(z)12 protein is highly conserved in humans and hence, it seems likely that vertebrate Su(z)12 homologues are also needed for silencing of HOX genes. Of the other Drosophila PcG genes, only E(z) and esc are also conserved in plants and previous studies showed that the E(z) homologue CURLY LEAF (CLF) is needed for repression of floral homeotic genes in leaves (Goodrich et al., 1997). Su(z)12 shows sequence similarity with three Arabidopsis proteins; FIS2, VRN2 and EMF2. Each of these proteins functions as a regulator to suppress a particular developmental process during plant development. FIS2 is needed to repress seed development in the absence of fertilization, a process that also requires the E(z)- and esc-related proteins FIS1/MEA and FIS3/FIE (Grossniklaus et al., 1998; Luo et al., 1999). VRN2 is needed for the stable repression of FLC, a key regulator that controls flowering (Sheldon et al., 2000, A. R. Gendall, personal communication). Particularly intriguing is the similarity between Su(z)12 and EMF2 (N. Yoshida personal communication). EMF2 acts as a floral repressor by suppressing the onset of reproductive development; EMF2 mutants show misexpression of the floral homeotic genes APETALA1 (AP1) and AGAMOUS (AG) in germinating seedlings (Chen et al., 1997). Thus, it appears that repression of HOX genes in Drosophila and repression of floral homeotic genes in Arabidopsis both depend on a conserved set of PcG proteins, Su(z)12 and E(z) in flies and EMF2 and CLF in plants.

The hallmarks of Su(z)12, EMF2, FIS2 and VRN2 are a single C₂H₂ zinc finger and a conserved stretch of amino acids that we named the VEFS-box. In all four genes, the VEFS-box is located C-terminal to the zinc finger. In DNA-binding assays, we have found no evidence that the Su(z)12 zinc finger by itself binds to DNA (A. K. S. and J. M., unpublished data). However, most other PcG proteins also do not bind to DNA directly but bind to chromatin as multiprotein complexes that contain different PcG members (Franke et al., 1992; Strutt and Paro, 1997; Shao et al., 1999; Ng et al., 2000; Tie et al., 2001). It is possible that the Su(z)12 protein also functions in a chromatin-binding protein complex and that in the context of such a complex, the zinc finger is needed for making DNA or protein contacts. As discussed in the following, the comparison of Su(z)124 and Su(z)121 mutant phenotypes suggests that the zinc finger and the VEFS box are probably two distinct functional domains. In embryos, Su(z)121 homozygotes show more extensive misexpression of HOX genes than Su(z)124 or Df(3L)kto2 homozygotes, and in imaginal discs, Su(z)121 mutant clones show a more rapid loss of HOX gene silencing than Su(z)124 mutant clones. As already discussed, the stronger phenotype of Su(z)121 mutants may be attributed to the interference of an aberrant Su(z)12¹ product with persisting Su(z)12⁺ protein molecules. The lesion in Su(z)121 may result in the expression of a truncated polypeptide that contains the zinc finger but lacks the VEFS box, whereas the short polypeptide encoded by the Su(z)124 allele lacks both the zinc finger and the VEFS box. One possible molecular explanation for the stronger phenotype of Su(z)121 mutants would therefore be that the truncated Su(z)12¹ protein, containing the C₂H₂ zinc finger, competes with wild-type Su(z)12 protein for binding to its natural target (i.e., a DNA sequence or another protein) but is not functional since it lacks the VEFS box and the C terminus. It is possible that the VEFS box is needed for interaction with other (PcG) proteins or, alternatively, that it is a catalytic domain providing an enzymatic activity needed for silencing.

A very recent study (Koontz et al., 2001) reports that endometrial stromal tumors in humans show chromosomal rearrangements in which the human homologue of Su(z)12, HsSU(Z)12, is fused to a zinc-finger protein.

We are grateful to Nobumasa Yoshida and Anthony Gendall for communicating the EMF2 and VRN sequences prior to publication. We thank Christiane Nüsslein-Volhard for continuous encouragment, support and critical reading of the manuscript. The work of A. B., J. L. and A. R.-L. was supported by grants from the Swedish Natural Science Research Council, the Royal Physiographic Society and the Lars Hiertas Fund.