Altered cellular proliferation and mesoderm patterning in Polycomb-M33-deficient mice

In Drosophila, the complex pattern of homeotic gene (HOM-C) expression is established early in development by the transiently expressed maternal and segmentation genes (Ingham, 1988). Late in development, two groups of genes are necessary to maintain the expression pattern of the homeotic genes. The trithorax group (trx-G) (reviewed by Kenisson, 1993) is required in maintaining the activity of the HOM-C genes in the appropriate segments, whereas the Polycomb group genes (Pc-G) are involved in the repression of the homeotic genes in the pertinent segments (reviewed by Paro, 1990; Bienz and Müller, 1995; Pirrotta, 1995). Loss-of-function mutations in members of the Pc-G genes result, as the expression of the segmentation genes decreases, in ectopic expression of the HOM-C genes late in embryogenesis whereas the early expression is not affected (Duncan and Lewis, 1982; Jürgens, 1985). This leads to shifts in the anterior limits of expression of the homeotic genes and, as a result, to posterior homeotic transformations. The binding of Pc to inactive HOM/lacZ transgenes (Zink et al., 1991) and the identification of a domain involved in chromatin binding shared with the heterochromatin protein HP1 (chromo-domain) in the PC protein (Paro and Hogness, 1991), together with the co-localization of several Pc-G proteins on polytene chromosomes (DeCamillis et al., 1992) to homeotic loci, has led to the proposition that Pc-G proteins are associated in large multimeric complexes which stably repress target genes by compacting the chromatin in a condensed structure inaccessible to specific transactivators (Paro, 1990). This compacted chromatin is thought to function as the basis for cellular memory during several cellular divisions maintaining HOM-C genes in an inactive state in specific cells (Paro, 1990).

Recently, mouse homologues related to the Pc-G genes were identified (Tagawa et al., 1990; Brunk et al., 1991; van Lohuizen et al., 1991), including the M33 gene which is considered as the structural (Pearce et al., 1992) and functional counterpart of the Pc gene since the M33 gene is able to partially rescue the Pc mutant phenotype (Müller et al., 1995).

To gain further insight into the functions of the M33 protein, we have generated a mutant mouse lacking the M33 protein. M33 mutant mice show homeotic transformations of the axial skeleton reminiscent of the bmi-1 and mel-18 mutant mice, as well as a failure to expand by several cell types, including lymphocytes and fibroblasts. Moreover, M33 mutants show an enhanced skeletal phenotype when treated with RA indicating that the M33 protein might play a role in defining access to retinoic acid response elements (RARE) defined in the regulatory regions of several Hox genes.

A XbaI-EcoRI fragment containing the pGK-neo gene flanked with two loxP sites, was subcloned in the EcoRV site of pBluescript (pBS) SK vector (Stratagene). Two DNA fragments from the M33 locus (Pearce et al., 1992) were isolated from a λ phage screened from a mouse 129/Ola genomic library: the 4.2 kb EcoRI-SalI fragment contains part of the 5′ non-coding region of the M33 gene, just upstream the start of transcription; the 3.5 kb XhoI-XbaI fragment, from the 3′ region, contains intron 4 and the last exon (exon 5). After filling the ends of restriction sites, the two M33 homologous fragments were cloned respectively in the SalI and XbaI sites of the modified pBS-neo vector, on each side of the loxP-neo-loxP cassette. This targeting vector pBS-neo-M33 was designed to allow the excision of exon 1 to exon 4 after homologous recombination in ES cells.

2×10⁷ E14 (129/Ola) strain ES cells were electroporated with 20 μg pBS-neo-M33 targeting construct DNA. 24 hours later, cells were positively selected with 300 μg/ml G418. Isolated G418-resistant colonies were picked after 7 days of selection. Homologous recombinants were tested by hybridization of HindIII-digested genomic DNA using a SmaI-EcoRI fragment (probe E) as an external probe and part of the exon 5 as an internal probe. A unique integration event was checked with a neomycin probe. For RT-PCR analysis, RNA was extracted from kidney and PCR amplification was performed using primers 1 and 2 on exon 5 (Fig. 1A).

Whole-mount skeletons of newborns were stained as described (Lufkin et al., 1992). 14.5 d.p.c. embryos were stained with the same protocol except for the time in 1% KOH, which was reduced to 0.5 hour. Genotyping was done by Southern blotting using genomic DNA extracted from the tail (new born) or from foetal liver.

In situ hybridization on sagittal sections of 12.5 d.p.c. embryos was performed according to the procedure described (Gaunt et al., 1988). Hox probes (Hoxa-3, Hoxd-3, Hoxa-5, Hoxc-5, Hoxa-6, Hoxc6, Hoxc-8) were previously described (Gaunt et al., 1988).

Single cell suspensions from thymus were incubated at 1×10⁷ cells/ml in 100 μl of staining solution (PBS complemented with 0.2% foetal calf serum (FCS) and 0.2% mouse serum) for 30 minutes on ice with phycoerythrin (PE)-conjugated anti CD8a (53-6.7) and biotin-conjugated anti-CD4 (H129.19) monoclonal antibodies. After two washes in PBS/0.2% FCS solution, the cells were incubated with 50 μl of Streptavidin-Cy-Chrome reagent in staining solution for 30 minutes on ice. The stained cells were washed three times and then fixed in 150 μl of 1% paraformaldehyde in PBS. Analysis were performed on a Becton Dickinson FACScan. The monoclonal antibodies and the Streptavidin-Cy-chrome reagent were purchased from PharMingen.

To establish organ cultures, thymuses were isolated from 14 d.p.c. embryo’s, cultured in the presence of 1.35 mM dGuo to deplete endogenous lymphocytes and reconstituted in hanging drops with 1×10⁵ foetal liver cell suspensions from 12.5 d.p.c. M33+/+, M33+/− and M33−/− mice. These samples contained equivalent numbers of T cell precursors as determined by serial titration. Unbound cells were removed by washing the lobes transferred to floating filters and cultured for up to 3 weeks before being individually harvested and analysed.

The fibroblasts derived from 12.5 d.p.c. M33+/+ and M33−/− were cultured for no longer than four passages in Dulbecco’s modified medium (DMEM from Sigma) supplemented with 10% of heat-inactivated foetal calf serum. Cells were arrested by incubating them for 2 days in DMEM without serum. They were stimulated by adding 20% of serum and then plated in a 24-well plate for counting over a 2 week period. For activation assay, splenocytes were seeded at a concentration of 10⁶ cells/ml in a 96-well plate. Cells were incubated in 200 μl of DME, 10% FCS, 10 mM Hepes pH 7.4 solution supplemented with 10 μg/ml of LypoPolySaccharide (LPS, Sigma). After 48 hours of incubation at 37°C, [³H]thymidine was added at a final concentration of 5 μCi/ml. Incorporation of [³H]thymidine was measured after 24 hours by scintillation counting.

Retinoic acid (RA) experiments were done with animals from times matings. Animals were mated for 2 hours, and fertilisation was assumed to have occurred after 1 hour. RA in vegetal oil was administered by a single oral gavage, applying 40 mg/kilogram of body weight all trans RA (sigma).

The murine Polycomb gene (M33) was disrupted by homologous recombination deleting the four first exons of the gene and inserting a neomycin-resistance (neo) gene in reverse orientation to the M33 transcription (Fig. 1A). Loss of the chromodomain encoded by the first three exons abolishes Pc activity in Drosophila resulting in homeotic transformation (Messmer et al., 1992). Two independent, homologous recombinant E14 ES cell clones were obtained. Recombinant ES clones were injected into BALB/c blastocysts and reimplanted into pseudopregnant females to produce chimaeric offspring. Germ-line transmission was obtained after a backcross between chimaeric males and BALB/c females. The heterozygous and homozygous offsprings were genotyped as described above. Genomic Southern analysis (Fig. 1B) as well as reverse transcriptase polymerase chain reaction (RT-PCR), confirmed the loss of the M33 5′ region and the absence of the M33 transcript in the homozygous M33−/− mice. (Fig. 1C).

M33−/− mutant mice develop to term and appear normal. They represent 24% of the offspring indicating that mutant mice do not die during embryogenesis. After a few hours following birth, half of all the M33−/− offspring die; after a few days, the surviving M33−/− mice can be easily recognised by their growth retardation (Table 1) and a high rate of lethality. Most of the M33 mutant mice (90%) die within 4 weeks and the life span does not exceed 6 weeks.

Whole-mount skeletal analysis of M33−/− mice reveals several malformations along the anteroposterior axis demonstrating the importance of the M33 gene in pattern formation (Table 1). Alizarin red and alcian blue staining (Lufkin et al., 1992) of new-born and 4-week-old animals indicates that all the M33−/− mice analysed show a malformation of the exoccipital (ex) bone and that the atlas (C1) is missing (Fig. 2B), resulting in six cervical vertebrae instead of seven in control mice (Fig. 2A). Partial transformation of C2 to C1, as indicated by the presence of an anterior arch of the atlas on the second cervical vertebrae was also observed in mutant mice. M33−/− animals show a posteriorisation of the thoracic vertebra T7 into T8, resulting in the presence of six verte-brosternal ribs instead of seven in M33+/+ mice (Fig. 2G,H). In addition, 15% of the mutant mice show a transformation of the lumbar vertebra L6 into a first sacral vertebra (S1) (Table 1). These transformations result in a C6/T13/L5 configuration in mutant mice in comparison to the C7/T13/L6 in wild-type mice.

Examination of 14.5 d.p.c. mutant foetuses indicates that the exoccipital/atlas malformation arises as a result of fusion between the ossification centres of these two bones (Fig. 2D). This shift induces a major remodeling of the craniocervical joint, as the first cervical vertebra is now fused to the basioc-cipital bone (Fig. 2F). In addition, the homozygous mice display an incomplete scapula formation resulting in a split ossification centre (Fig. 2J). At 14.5 d.p.c., the cartilage pri-mordium of the scapula is already affected and two ossification centres are present on each side of the scapula (Fig. 2L). The altered morphogenesis of the scapula in M33−/− mice is actually reminiscent of normal pelvic bone development where the ossification centres of the ischial and pubic bones fuse anteriorly, giving the pelvic bone a triangular shape with a central hole (obturator foramen).

Regulation of Hox gene expression boundaries is critical for specification of vertebral identities in mouse (McGinnis and Krumlauf, 1992). The Polycomb group genes, mel-18 (Akasaka et al., 1996) and bmi-1 (van der Lugt et al., 1994; Alkema et al., 1995), have been shown to regulate the anterior limits of expression of some of the Hox genes and to induce homeotic transformations in loss-of-function or gain-of-function experiments. Posterior or anterior transformations of vertebrae have been described in the case of Hox gene gain of functions or loss of functions in mice (McGinnis and Krumlauf, 1992). These data prompted us to examine the expression patterns of some representative Hox genes in the M33−/− mice. By RNA in situ hybridisa-tion, Hoxa-3 transcript was detected over the basioccipital bone anlage in mutant mice (Fig. 3B) but not in wild-type mice (Fig. 3A). This anterior shift in the Hoxa-3 boundary is widened by the fusion of the basioccipital and the first prevertebra (Pv1), which was already manifest at 12.5 d.p.c. of development (Fig. 3B). However, we found no significant differences in the anterior limits of expression of several other Hox genes in the mutant mice. As shown in Fig. 3C,D the boundaries for Hoxc-8 and Hoxc-6 are normal (Gaunt et al., 1988) if we assume that the first discernible prevertebra is Pv2 (Pv1 having become fused with the occipital). Similarly, the anterior boundaries for Hoxd-4 Hoxa-5, Hoxc-5 and Hoxa-6 (pv2, 3, 6 and 8, respectively) are not affected in the mutant mice (data not shown). Moreover the boundaries for all these Hox genes appear at apparently normal positions in the central nervous system (Fig. 4) (Gaunt et al., 1988). Fig. 4 also shows that there are no ectopic areas of Hox gene expression located anterior to the boundaries. This finding is in contrast to that made for Drosophila Polycomb^−/− embryos (Wedeen et al., 1986) and suggests that, unlike Pc in Drosophila, M33 protein in mice is not critical for the maintenance of all Hox expression domains.

It has been suggested that the normal boundaries of Hox gene expression might normally be regulated by retinoic acid and that ectopic expression of the Hox genes induced by RA administration leads to morphological transformations (Kessel and Gruss, 1991; Conlon and Rossant, 1994). The recent identification of retinoic response elements (RARE) in the regulatory region of Hox genes (Langston and Gudas, 1992) has provided direct evidence for an involvement of RA in the regulation of Hox genes during development. In order to test whether the M33 protein plays a role in the accessibility of the RARE, we administered RA to 7.5 d.p.c. pregnant M33+/− females bred with M33+/− males. Pregnant females were killed at 17.5 d.p.c. and the morphological changes were scored in M33+/+, M33+/− and M33−/− embryos and compared to the untreated M33−/− embryos. As shown in Fig. 5, the skeletal alterations seen in the M33−/− mice are largely amplified in the M33−/− RA-treated embryos. Althought the effects of RA in the cervical region in the M33 mutants are difficult to interpret (since RA induces by itself in wild-type mice strong malformations in the cervical region; Kessel and Gruss, 1991), the additional effects of RA can be more easily seen in the thoracic region of the M33 mutant mice. The M33−/− RA-treated mice present three sternebrae ossification centres instead of five in the M33−/− untreated animals (Fig. 5A). The xiphoid process of the treated mutant is split in two parts. Moreover the treated mutant mice display only five sternebrae; the first two being fused before joining the manubrium sterni. In addition, the scapula defect detected in the M33−/− mutant (Fig. 5D) is also severely enhanced by the RA treatment (Fig. 5C). These changes were not induced in the M33+/+ mice and were induced only weakly in M33+/− mice (Fig. 5B). Heterozygotes did not show these changes in the absence of retinoic acid (Table 1).

Examination of 3- and 4-week-old M33−/− animals showed pronounced involution of the thymus and spleen with a drastic reduction in the total number of nucleated cells in these organs (>100 fold and >10 fold, respectively; Table 1). To assess which cell types were affected, flow cytometric analysis (FACS) was performed on thymocytes using standard lymphoid markers (Fig. 6A). The presence of single positive cells (SP) (CD4+ or CD8+) expressing normal level of T-cell receptor (data not shown) showed that T cell development can progress in the absence of M33. However the relative proportions and number of each T cell subset were disturbed in M33−/− mice as compared with controls; immature CD4−/CD8− thymocytes account for 25% of cells in M33−/− thymuses suggesting that these precursors are either retarded in their normal differentiation pathway as a direct effect of the absence of M33 or alternatively that the failure to thrive and general poor health of the M33−/− animals accounts for the diminishing population CD4+/CD8+ DP cell population. To establish the cause of reduced thymocyte numbers and CD4+8+ subsets in M33−/− mice, thymic organ cultures were used to estimate the frequency and the differentiation potential of T cell precursors in vitro. Foetal liver cells from 12.5 d.p.c. embryos were used to recolonise alymphoid (dGuo-treated) thymic lobes (Cohen and Duke, 1984), and the development of T cells was monitored by FACS. Serial titration experiments showed that the relative abundance of pro-T cells in foetal liver of M33−/−, heterozygotes and wild-type mice, was equivalent at days 12-14 d.p.c., (data not shown). A marked reduction in T cell expansion was observed in cultures derived from M33−/− mice as illustrated in Fig. 6C. Over a two weeks culture period, the number of T cells recovered per lobe increased from 1×10⁴ to 2.6×10⁵ in wild-type and heterozygote samples, compared with an increase from 1×10⁴ to 5.1×10⁴ in samples established from M33−/− mice. This reduction was accompanied by a slight retardation in T cell maturation (Fig. 6B), although representatives of all stages of T cell development were observed. These data show that pro-T cells in M33−/− mice appear at normal levels in the embryonic liver and can transit between the immature DN (CD4−8− IL2R+), DP (CD4+8+) and mature SP(CD4+ or CD8+) stages, but that they generate far fewer mature T cells. Furthermore, they indicate that this failure to expand is an inherent property of the M33−/− thymocytes rather than a consequence of compromised health of the mice or generalized impairment of the thymic environment.

In order to determine whether proliferation defects are present in other cell types in mutant mice, proliferation assays were performed on splenocytes and fibroblasts. Thymidine incorporation assays with LPS-activated splenocytes derived from three weeks old mutant mice indicate that these cells also fail to proliferate (Fig. 7A). Similarly, cellular counts over a two week period reveal that the M33−/− derived fibroblasts obtained from 12.5 d.p.c. embryos are severely impaired in their capacity to expand (Fig. 7B).

In this study, we have generated a mouse mutant line in which the M33 gene, a homologue of Drosophila Polycomb, has been deleted. The M33-deficient mice present severely retarded growth, perturbations of cellular proliferation for several cell types and homeotic transformations of vertebrae. In Drosophila, the early pattern of expression of the homeotic genes established by the maternal and segmentation genes is later maintained by the trx-G and the Pc-G genes. These genes are involved in the maintenance respectively of the active or repressed state of expression of the homeotic genes in the appropriate segments. Loss of function in these groups of genes results in ectopic (Pc-G) or the loss (trx-G) of expression of the homeotic genes and, consequently, in homeotic transformations. In mice, as in Drosophila, Hox genes regulate organisation of the body plan (reviewed by McGinnis and Krumlauf, 1992). Similarly, it has been shown that mutant mice heterozygous for the trx-G gene homologue (Mll) present anterior and posterior transformation of the axial skeleton and that null mutant mice fail to maintain the expression of the Hox genes at day 10 of development (Yu et al., 1995). Mutant mice for the Pc-G genes mel-18 and bmi-1 show axial posterior transformations and an anterior shift in the expression boundary of a subset of Hox genes (Akasaka et al., 1996; van der Lugt et al., 1994, 1996).

The homeotic transformations observed in M33−/− mice (Table 1; Fig. 2) are reminiscent of homeotic transformations described for several mouse Hox gene mutants. Among loss-of-function mutants, for example, the Hoxc-8 mutant mice present an anterior transformation as seen by the generation of an extra pair of ribs from the first lumbar vertebra (Le Mouellic et al., 1992); Hoxb-4 mutant mice show partial posterior transformation of the atlas to the axis and sternal malformations (Ramirez-Solis et al., 1993). In Hoxa-5 mutant mice, there is a posterior transformation of the cervical vertebra (C7) into the first thoracic vertebra (T1). Similarly homeotic transformations of the axial skeletal structures occur in gain-of-function mutants generated for Hox genes. The ectopic expression of Hoxd-4 induces posterior skeletal transformations as seen by ectopic neural arches and the absence of the supra-occipital and exoccipital bones as well as a fusion of the atlas and the axis (Lufkin et al., 1992). Transgenic mice overexpressing the Hoxa-7 gene show also posterior transformations of the axial skeleton as the atlas and the axis present characteristics of more posterior vertebra (Kessel et al., 1990).

The craniocervical transformations seen in the M33-deficient mice are, however, most similar to those reported in Hoxd-3 null mutant mice (Condie and Capecchi, 1993). To explain the defects observed in Hoxd-3−/− mice, Condie and Capecchi suggested a model in which Hoxd-3 regulates the proliferation rate of precursor cells. In agreement with this model, it has been shown that several Hox genes are involved in controlling cell proliferation (Care et al., 1994; Sauvageau et al., 1994; Sordino et al., 1995). Our results on the altered proliferation potential of different cell types as well as the cervical defects detected in the M33−/− mice are consistent with this model and support the notion that the relative dosage of the HOX proteins is critical for both proliferation and patterning processes (Condie and Capecchi, 1993). It has been shown that compound mutant mice for the Hoxd-3/Hoxa-3 genes show a complete loss of the atlas vertebra revealing synergistic interactions between these two genes (Condie and Capecchi, 1994). Since the malformations observed in the cervical region of the M33 mutant mice correspond more to those of the Hoxd-3 mutant than to those of the compound Hoxd-3/Hoxa-3 mutants suggest that the loss of the M33 protein mostly acts on the rate of expression of the Hoxd-3 gene in the craniocervical region (Condie and Capecchi, 1994).

The findings that homeotic transformations are seen over the entire A-P axis indicates that M33 may control the level of expression of several other Hox genes. The transformations observed in the anterior part of the skeleton are consistently seen in all the M33−/− mice whereas the transformations located in the posterior part are less penetrant (Table 1); this suggests that the M33 protein might primarily affect the Hox genes located in the 3′ end of the complex. Alternatively, since Pc-G genes in Drosophila are involved in the regulation of gap genes (Pelegri and Lehmann, 1994), it is possible that in M33−/− mutant mice the initiation phase of expression of some Hox genes is altered leading to posterior transformations. Although a direct regulation of the Hox genes by M33 has yet to be established, the alterations observed in the specification of structures along the A-P axis in M33 mutant mice are in accordance with the Pc gene function in Drosophila.

bmi-1 and mel-18 mutant mice present posterior transformations of the axial skeleton as well as an anterior shift in the expression boundaries of several Hox genes. Some of these posterior transformations are very similar to those observed in the M33 mutant mice. Similarly, in the M33-deficient mice, the Hoxa-3 gene expression domain is shifted anteriorly. However, we have not detected modifications in the anterior boundaries of several other Hox genes in the M33 mutant mice. This might reflect the different mechanisms used to maintain the Hox gene expression and vertebral identity in mice. It has been shown that the two Pc-G genes, bmi-1 and mel-18, regulate common Hox targets. It is possible that, in the absence of the M33 product, those genes still sustain some of the Hox gene domain of expression in their normal boundaries but that their level of expression is not properly maintained. Since bmi-1 and mel-18 also regulate specific Hox genes, it is possible that M33 controls the anterior limit of expression of Hox genes unaffected in bmi-1 and mel-18 mutant mice (Akasaka et al., 1996; van der Lugt et al., 1996). Alternatively, since we have used F₁ mice from a cross between 129/ola and BALB/C to obtain M33−/− mice, it is possible that the hybrid background might influence the Hox boundaries.

Pc-G gene products are thought to participate in the formation of large protein complexes, which promote the formation of condensed chromatin along specific chromosomal regions leading to a heritable repression of the HOM-C genes (Paro, 1990). In this model, the Pc-G proteins are thought to form a compacted chromatin structure that prevent interactions between DNA and DNA-binding proteins. This chromatin model implies that all transcription factors should be excluded from DNA in this repressed state. However, it has been recently shown in Drosophila (McCall and Bender, 1996) that, while the yeast GAL4-dependent transcription is inhibited by Pc, T7 RNA polymerase is not, implying that the compacted chromatin model for Pc-G repression is not completed by a simple exclusion of all the transcription factors. This repression could be achieved through a screening of the transcription factors by size or shape (McCall and Bender, 1996) or by nuclear compartmentalization whereby Pc-G repressed chromatin is maintained in an inactive region of the nucleus (Paro, 1993).

In mouse, the proper expression of several Hox genes is dependent on multiple RA response elements. For example, RAREs have been found in the Hoxa-1 (Langston and Gudas, 1992) and Hoxd-4 (Popperl and Featherstone, 1993) genes and mediate an up-regulation in response to ectopic doses of RA in cultured cells. Two enhancers 3′ of the mouse Hoxb-1 have been identified, which are required for the proper expression of Hoxb-1 (Marshall et al., 1994; Studer et al., 1994) and mediate the early ectopic response to RA. The enhancer that controls the RA response and regulates the expression of Hoxb-1 in the neuroectoderm contains a RARE: point mutations in the RARE abolish the expression of the Hoxb-1 gene in the neuroectoderm demonstrating that this RARE is essential for the correct expression of the Hoxb-1 gene.

Our results with RA treatment on skeletal transformations in M33 mutant mice suggest an interaction between RA activation pathway and M33. It is possible that the M33 protein might control the accessibility to the RAREs defined in the regulatory regions of some Hox genes during normal development. The absence of M33 might allow those elements to be inappropriately accessible for transcription leading to misexpression or ectopic Hox genes expression. Whether RA-treated M33−/− embryos present extended shifts in the anterior limit of expression or overexpression of some Hox genes will require extensive analysis of several Hox genes that have been found to be directly responsive to RA via their RARE at different time points of development.

During gastrulation, vertebrate Hox genes are transcribed in a temporal sequence that is correlated to their position in the complex (temporal colinearity) (Dollé et al., 1989, Duboule and Morata, 1994; Gaunt and Strachan, 1996). It has been proposed that this 3′ to 5′ progressive opening of the Hox complexes might be achieved through a change in the chromatin configuration (Dollé et al., 1989). Our results support this proposal and suggest that this change might be mediated by Pc-G function. In this view, transition from an inactive chromatin state to an active state would allow critical Hox promoter regions (i.e. RAREs) to be accessible to specific activators or repressors. This suggests a fundamental difference between vertebrate and fly in activation of the Hox genes: in vertebrates, the Hox complexes are progressively opened to transcription while, in Drosophila, it is progressively closed during development (van der Hoeven, 1996).

Mutant mice for the bmi-1 and for mel-18 genes show proliferation abnormalities and posterior transformations similar to the M33−/− mice indicating that this group of genes might interact in regulating the Hox complexes in mice. However, the specific axial transformations observed in M33−/− mice demonstrate that the M33 gene has distinct, as well as common, Hox targets. It has been shown on polytene chromosomes in Drosophila that the Pc-G proteins are co-localized at many sites (DeCamillis et al., 1992). In Xenopus embryos, XPOLYCOMB and XBMI-1 proteins are able to interact with each other (Reijnen et al., 1995). Furthermore, stronger posterior transformations are observed in Drosophila mutant for two or more Pc-G genes (Jürgens, 1985). Consequently, one would expect extended posterior transformations and extended shifts in the Hox boundaries in double Pc-G mutant in mice. bmi-1+/− and M33+/− mice are currently being intercrossed in order to test this hypothesis.

Our results on thymocyte differentiation demonstrate that the M33 protein is not required for T cell maturation since CD4+ and CD8+ single positive cells expressing normal level of T-cell receptor are detected in the thymus of mutant mice. However, we have found that the M33 gene is necessary for T cell precursors proliferation. Recently, another Pc-G protein (Hobert, 1996), ENX-1, has been isolated in mouse by virtue of its association with the protein encoded by the proto oncogene vav. vav mutant mice (Tarakhovsky et al., 1996) also display involuted thymi and an impaired proliferative response to lymphoid activation emphasizing the possibility that Pc-group proteins have a pivotal role in lymphocyte proliferation. Since Pc-G proteins are thought to form large protein complexes, it is possible that the M33 protein could exert its effects on lymphoid proliferation through the VAV protein in the thymus. On the other hand, it has been demonstrated that misregulation of some Hox genes leads to lymphocytes proliferative alteration (Perkins et al., 1990, Sauvageau et al., 1994). Then, it is possible that the thymocytes defects induced in the M33 mutant mice could be due to an altered expression of some Hox genes.

This work was supported by grants from CNRS, ARC, GEFLUC and British Council. N. C. is supported by ARC and Fondation de France Fellowships. We thank M. Merkenschlager for helping with FACS analysis, P. Golstein, P. Naquet and P. Dollé for critical reading of the manuscript, M. Malissen and A. Gillet for instruction in ES cell culture and blastocyt injection techniques.