Function of the Polycomb protein is conserved in mice and flies

The design of animals depends on a special class of homeotic selector genes (Lewis, 1963, 1978; Garcia-Bellido, 1975; Kaufman et al., 1980). Even between creatures as different as mice and flies, these genes are conserved in their arrangement on the chromosome, in their homeobox sequences, in some aspects of function and in their order of deployment from head to tail (e.g. Duboule and Dollé, 1989; Graham et al., 1989; McGinnis and Krumlauf, 1992). Early in development, a unique set of selector genes becomes activated in each group of primordial cells, the particular combination of active genes in a group determining which developmental pathway it will follow (Hadorn, 1965; Garcia-Bellido et al., 1979). The descendants of each primordial cell then maintain their determined state by keeping the same set of selector genes continuously active throughout development (Morata and GarciaBellido, 1976; Morata and Lawrence, 1977). Since ectopic expression of selector gene products, even at a late developmental stage, can change the determined state of a cell (e.g. Lewis, 1963; Schneuwly et al., 1987), it is equally important to keep selector genes stably inactive where their products are not wanted. This is achieved by repression: selector genes contain enhancers that are potentially active in many primordia, but whose activity is blocked or “silenced” in some (Bieberich et al., 1990; Müller and Bienz, 1991; Püschel et al., 1991). In Drosophila, heritable silencing of homeotic genes is achieved by a two-step process (Garcia-Bellido and Capdevila, 1978). The first step, initiation, provides the spatial specificity: gap gene products that are localised in the early embryo repress homeotic genes in sets of primordia (White and Lehmann, 1986; Harding and Levine, 1988; Irish et al., 1989; Qian et al., 1991; Zhang et al., 1991; Shimell et al., 1994). The second step, maintenance, keeps the homeotic genes silenced even though the gap proteins have gone and even though the cells divide. This heritable silencing depends on a set of more than 10 genes (Lewis, 1978; Struhl, 1981; Jürgens, 1985), called the Polycomb-group (Pc-G), which includes the Polycomb (Pc) gene itself. The Pc-G gene products are necessary for the proper regulation of homeotic genes in many parts of the body and are almost uniformly distributed in the embryo (Paro and Hogness, 1991; Franke et al., 1992; Martin and Adler, 1993); they therefore do not seem to carry any intrinsic positional information. The Pc-G proteins maintain cell determination, they function as a molecular memory.

Contemporary studies on bmi-1, a murine homologue of the Pc-G gene Posterior sex combs (Psc), provide evidence that it acts as a regulator of Hox gene expression in mice (van der Lugt et al., 1994; Alkema et al., 1995). Here we show that the M33 protein, a mouse homologue of Polycomb (Pearce et al., 1992, Paro and Hogness, 1991), can substitute functionally for the Pc protein in transformed flies. These two complementary lines of evidence suggest that at least the maintenance mechanism of cell determination has been widely conserved.

A cn;ry⁴² strain was used as host for transformation.

The following mutant alleles were used to test for rescue activity of the transgenes: Pc^23937.4A, Pc^26865.4, Pc^XT109 are three alleles that lack antigen for the Pc protein (R. Paro, personal communication). For some experiments, we used Pc³, an allele with an antimorphic effect (Duncan and Lewis, 1982) which we chose to minimise any rescue of the larval phenotype due to Pc⁺ product given by the mother (Denell, 1982; Lawrence et al., 1983). In some experiments, an Abd-B^M1 allele was used for marking purposes.

The Arm-M33 and Arm-Pc transgenes were assembled in bluescript from the following starting plasmids: Arm bs, containing a 1.8 kb fragment of the armadillo promoter (Vincent et al., 1994); M33bs, a subclone containing the M33 coding region (Pearce et al., 1992); hb-GAL-Pc bs, a subclone that contains the Pc coding region, a fragment with the hb translation initiation codon and 5′ untranslated leader sequences (Müller, 1995). Convenient restriction sites were engineered by site-directed mutagenesis to link the different fragments isolated from these plasmids and obtain Arm-M33 bs [KpnI-PvuII-(arm promoter nucleotides 11-1792)-BamHI-ACTAGA-(hb 5′ untranslated leader nucleotides 4675-4683)-ATGAAGCTT-(M33 codons 2-519)-TGAGACCAT-SalI] and Arm-Pc bs [KpnI-PvuII-(arm promoter nucleotides 11-1792)-BamHI-ACTAGA-(hb 5′ untranslated leader nucleotides 4675-4683)-ATGAAGCTT-(Pc codons 3-390)-TGAGC-SalI]. Each transgene was then subcloned as a KpnI-SalI fragment into the backbone of a modified transformation vector ‘pry’ to obtain ARM-M33 pry and ARM-Pc pry, respectively; ‘pry’ contains RNA trailer sequences from the Drosophila hsp70 gene downstream of the SalI site (Müller and Bienz, 1991). Detailed maps of the plasmids used are available on request.

Three (Arm-Pc) and three (Arm-M33) independent lines of transformed flies were examined and gave almost identical results.

All the three Arm-Pc lines that we used are viable when homozygous, but the homozygotes frequently show defects in A4 of adults (ventrally only in most cases). First instar larvae show a hierarchical series of defects. (1) A gap in the midline in the region of ps 9 of the CNS of developing embryos and in the cuticle where the A4 denticle band is split and displaced laterally. (2) Complete lack of A4. (3) Loss of A4 plus reduction of A2 and A8. Since these effects are not observed in zygotes carrying only one paternally derived copy of the transgene, we wondered if they might be a consequence of very high levels of Pc protein donated by the mother to young embryos (staining of Arm-lacZ constructs suggests that the armadillo promoter is very active during oogenesis) and whether an abnormally high level could affect the expression of genes that are not otherwise known to be controlled by Pc. However, the armadillo promoter is not always so active: our rescue data of adults suggest the armadillo promoter functions less well than the endogenous Pc promoter (see below).

Note that larvae and adults carrying two copies of the Arm-Pc transgene plus two copies of the endogenous Pc⁺ gene do not show any homeotic transformations. Note also that in the three Arm-M33 lines, homozygous larvae and flies show no defects.

We used different trans-combinations (Pc^26865.4/Pc^XT109; Pc^XT109/ Pc^XT109; Pc^23937.4A/Pc^XT109; Pc^26865.4/Pc^26865.4) and either one or two copies of the transgene to test for adult rescue and all gave survivors with Arm-Pc but not with Arm-M33. The survivors fell into two classes: flies carrying two doses of the transgene were either wild type in phenotype or showed slight defects in the wings but had no other Pc phenotype; flies carrying one dose showed sexcombs on all three legs, antennae partially transformed to legs, abnormal wings, abdominal cuticle and bristles showing slight transformation towards analia, that is fewer cell hairs and longer finer bristles. These flies are sick and sterile. These results suggest that, in later development, the armadillo promoter is somewhat less active than the wild-type Pc promoter, which in one dose gives more Pc function (although the flies are still not wild type).

To analyse Ubx expression in putatively rescued embryos, a cross was made in which the second chromosome was heterozygous for the transgene and the third chromosome carried Abd-B^M1Pc³/TM2. As expected one quarter of the embryos were Abd-B^M1Pc³/Abd-B^M1Pc³ and of these only one quarter showed the complete Pc⁻ phenotype, as would be predicted if one dose of either transgene substantially rescued the Pc⁻ phenotype. The same result was obtained in a similar cross using the Pc^XT109 allele; as the chromosome in this case was not marked, we counted all the embryos and found that approximately 1 in 16 showed the complete Pc⁻ phenotype.

Embryos were fixed and stained with a monoclonal antibody against Ubx protein (White and Wilcox, 1984), as previously described (Lawrence and Johnston, 1989). Preparations of larval and adult cuticles were by standard methods (Roberts, 1986).

To make clones in the adult abdomen, y; Arm-Pc/+; Dp(1;3)sc^J4, y⁺M(3)67C/Pc^XT109ory; Arm-M33/+; Dp(1;3)sc^J4, y⁺M(3)67C/Pc^XT109 males were generated and irradiated as larvae to induce somatic recombination (1000 R of X rays). Clones from these and their Sternopleura (Sp) control siblings (y; Sp/+; Dp(1;3)sc^J4, y⁺M(3)67C/ Pc^XT109 males) were compared. Because of the positions of the yellow marker gene and the Pc mutation on the third chromosome, only about half of the yellow clones are Pc⁻/Pc⁻, the other half are Pc⁻/Pc⁺ (Busturia and Morata, 1988) and due to recombination distal to the Pc locus. In those flies carrying no transgene, 37% of the abdominal clones (n=38) showed a complete Pc mutant phenotype. In flies carrying one copy of Arm-Pc all of the abdominal clones (n=27) were wild type in phenotype. In flies carrying one copy of Arm-M33, 51% of the clones (n=47) were rescued to an intermediate phenotype with both wild type and Pc⁻ character but some gave an apparent Pc⁻ phenotype. As expected, the remaining half of the clones (49%) were wild type.

The mouse M33 gene (Pearce et al., 1992) was isolated by its homology to the Pc chromobox (Paro and Hogness, 1991), a sequence motif of about 40 amino acids also found in a variety of other genes (James and Elgin, 1986; Delmas et al., 1993; Tschiersch et al., 1994). The chromodomain of Pc is thought to be responsible for recognising specific targets on the chromosome (Messmer et al., 1992), probably via one or more other proteins — the Pc protein does not seem to bind to DNA on its own (Franke et al., 1992). In addition to the chromodomain, M33 contains a C-terminal stretch of about 30 aminoacids that is well conserved in Pc but identified so far in no other proteins (Pearce et al., 1992); however, over the remaining 85% of their sequence the two proteins share no significant homology (Fig. 1). Both the chromodomain and the C-terminal domain are essential for Pc function in flies (Messmer et al., 1992; Franke et al., 1995; Müller, 1995); the conservation of these two regions between the Pc and M33 genes suggests that the two proteins might be functionally homologous.

Here we ask whether the M33 protein functions in flies by testing its ability to substitute for the fly Pc protein. As the promoter of the fly gene is ubiquitously active but has not been defined, we used the armadillo promoter which functions in the ovary, the embryo and imaginal discs (Vincent et al., 1994). The staining of Arm-LacZ (Vincent et al., 1994) and our own results suggest that this promoter is stronger than the Pc promoter in oogenesis but weaker in the imaginal discs (see Materials and Methods). Using a construct in which the armadillo promoter directs expression of the fly Pc protein (Arm-Pc), we find that two doses can rescue Pc⁻ mutants and give flies that are wild type in appearance. We then turned to the mouse Polycomb protein, and asked whether, in transformed flies, the Arm-M33 transgene would also be able to rescue Pc⁻ mutants.

First, we looked at the cuticle pattern of larvae. In the wild type, each metamere bears a characteristic belt of denticles, the pattern being specified by the combination of homeotic genes that are active at each locale (Figs 1, 2). In Pc mutants, silencing fails and all homeotic genes tend to become universally expressed. As a consequence, and because of a functional hierarchy amongst homeotic proteins (Lewis, 1978; Struhl, 1981; Struhl and White, 1985; González-Reyes and Morata, 1990), all denticle belts become transformed towards that of the eighth abdominal segment (Fig. 1; see legends for details). Both the control Arm-Pc and the experimental Arm-M33 transgenes rescue the Pc⁻ cuticle phenotype completely (Figs 1, 2) and the rescued larvae hatch. However, unlike the situation with Arm-Pc, we saw no rescue to the adult by Arm-M33.

Next, we studied expression of a homeotic gene in embryos (Fig. 3). In Pc mutant embryos carrying either the Arm-Pc or the Arm-M33 transgenes, homeotic gene expression is restricted – that is, silenced – until the end of embryogenesis (Fig. 3). No ectopic expression is seen in the epidermis or mesoderm, but, in the central nervous system of mutant embryos carrying Arm-M33, a few cells begin to show ectopic expression at late stages (Fig. 3). Thus it appears that the M33 protein can substitute for fly Pc although it is somewhat less efficient.

The function of Pc is necessary throughout development (Struhl, 1981; Duncan and Lewis, 1982; Busturia and Morata, 1988) and we therefore asked if the transgenes can rescue clones of Pc⁻ adult cells. We found that one dose of the Arm-Pc transgene rescues Pc⁻ cells and gave clones with a completely wild-type phenotype, while one dose of Arm-M33 also rescues but gave clones with an intermediate phenotype (Fig. 4). Thus the M33 protein can emulate Pc both early and late in development.

The fly Pc protein has two functional domains: the chromodomain and the C-terminal region. The chromodomain is found in at least one other protein that binds to chromosomes with a characteristic distribution. Whereas the Pc protein can be detected at about 100 sites on the polytene chromosomes (Zink and Paro, 1989), the HP1 chromodomain protein is mainly located in the centromeric heterochromatin (James and Elgin, 1986). Pc protein with a defective chromodomain can still work if artificially tethered to the DNA (Müller, 1995) — suggesting that the normal function of the chromodomain is to bind to a chromosomal target. This binding is presumably via Polycomb-group proteins, which themselves recognise particular base sequences of DNA (Chan et al., 1994; Christen and Bienz, 1994; Müller, 1995). The other functional part of the protein, the C-terminal domain, most likely acts to recruit further members of the Polycomb-group of proteins to form a silencing complex (Müller, 1995, Franke et al., 1995). Note that complexes containing the Pc protein have been isolated from Drosophila embryos and contain at least one other Polycomb-group protein (Franke et al., 1992).

Our finding that the M33 protein, having these two domains but no other homology with Pc, can rescue Pc mutants makes the following points: First, it is striking that even though the two conserved domains constitute a small fraction of the protein, the mouse protein is effective in flies. Second, if the fly Pc chromodomain does recognise proteins on the DNA, it would follow that the mouse homologue can bind to these fly proteins, suggesting that they have been conserved between mice and flies. Third, if the C-terminal domain is indeed there to recruit one or more proteins to form a multimeric complex, then the proteins that are recruited may also be conserved. This universalist view has other support: mice lacking the bmi-1 gene, a homologue of the Drosophila Polycomb-group gene Posterior sex combs, show “homeotic” transformations along the whole body axis (van der Lugt, 1994). As in Drosophila, these transformations change patterns characteristic of anterior segments towards those characteristic of more posterior ones.

Determination results in “a reproducible cell state propagated ……by cell heredity” (Hadorn, 1965). This requires the transmission of the silenced state of homeotic selector genes through many cell divisions. In flies this heritable silencing depends on the Polycomb-group of genes. Our results, taken together with studies of the bmi-1 gene (van der Lugt et al. 1994; Alkema et al., 1995) suggest that cell determination in vertebrates depends on a group of genes that is homologous to the Polycomb-group of flies. Thus, although mice and flies might use different molecules and mechanisms to initiate determination, the genetic and molecular mechanisms that maintain the silenced state of homeotic genes through cell divisions may be the same.

We thank Renato Paro for advice and the Pc alleles as well as Michael Levine, Gines Morata and Gary Struhl for encouragement. Mariann Bienz and Jean-Paul Vincent have helped us improve the manuscript. J. M. is supported by a fellowship from the Swiss National Science Foundation. The illustrations are witness to the skill of Paul Johnston, who died as this work was completed.