Development 127, 5093-5100 (2000) 5093 Printed in Great Britain © The Company of Biologists Limited 2000 DEV2519

Loss- and gain-of-function mutations show a Polycomb group function for Ring1A in mice

María del Mar Lorente1, Camelia Marcos-Gutiérrez1, Claudia Pérez1,2, Jon Schoorlemmer1, Angel Ramírez3, Thomas Magin4 and Miguel Vidal1,* 1Developmental and Cell Biology, Centro de Investigaciones Biológicas, Velázquez 144, 28006 Madrid, Spain 2Department of Animal Pathology: Animal Medicine, Faculty of Veterinary Science, University of León, 24071 León, Spain 3Cell and Molecular Biology, CIEMAT, Avda. Complutense, 28040 Madrid, Spain 4Department of Molecular Biology, Institute for Genetics, University of Bonn, Römerstrasse 164, Bonn 53117, Germany *Author for correspondence (e-mail: [email protected])

Accepted 14 September; published on WWW 2 November 2000

SUMMARY

The products of the Polycomb group (PcG) of act as skeleton, which indicates an unusual sensitivity of axial transcriptional repressors involved in the maintenance skeleton patterning to Ring1A dosage. Ectopic of homeotic patterns throughout expression of Ring1A also results in dose-dependent development, from flies to mice. Biochemical and molecular anterior transformations of vertebral identity, many of evidence suggests that the mouse Ring1A gene is a member which, interestingly, are shared by Ring1A−/− mice. In of the PcG of genes. However, genetic evidence is needed to contrast, the alterations of expression observed establish PcG function for Ring1A, since contrary to all in both type of mutant mice are subtle and involve a other murine PcG genes, there is no known Drosophila reduced number of Hox genes. Taken together, these results PcG gene encoding a homolog of the Ring1A . To provide genetic evidence for a PcG function of the mouse study Ring1A function we have generated a mouse line Ring1A gene. lacking Ring1A and mouse lines overexpressing Ring1A. − − + − Both Ring1A / and Ring1A / mice show anterior Key words: Polycomb, Mouse, Ring1A, Ring1B, Hox, Axial transformations and other abnormalities of the axial skeleton

INTRODUCTION adult males of ectopic sex combs characteristics of the first leg. Mutations in loci outside the HOM-C that cause an extra sex The Polycomb group (PcG) of genes were first identified as combs phenotype identify the PcG of genes (Jürgens, 1985). trans-acting regulators of homeotic gene function in In PcG mutants, the initial pattern of expression of HOM-C Drosophila (Kennison, 1995). Two clusters of homeotic genes, genes is normal, but later in development a generalized the Antennapedia complex (ANT-C) and the bithorax complex derepression occurs (Simon et al., 1992; Soto et al., 1995; (BX-C), collectively referred to as the homeotic complex Struhl and Akam, 1985). Thus, the PcG genes function to (HOM-C) are responsible for the determination of segmental maintain, rather than to determine, homeotic gene repression. identities in Drosophila. The ANT-C genes determine and The maintenance of homeotic gene expression is controlled by maintain the identity of the head and anterior thoracic another set of genes, the trithorax group (trxG), often identified segments, whereas the BX-C genes control the identity of as suppressors of the PcG-induced homeotic phenotypes posterior thoracic and abdominal segments (Kaufman et al., (Kennison, 1993). 1980; Lewis, 1978). The expression of homeotic genes in The PcG genes encode a group of structurally heterogenous specific overlapping domains along the anterior-posterior (AP) (Simon, 1995). Recently, plant and vertebrate genes axis of the embryo is correlated with the physical order of encoding proteins containing regions of homology with genes in the (Harding et al., 1985; Lewis, 1978). Drosophila PcG products have been identified (Gould, 1997; This principle of colinearity has been conserved in vertebrates Preuss, 1999; Schumacher and Magnuson, 1997). Mutations in (Duboule and Dollé, 1989; Graham et al., 1989). Loss- and these genes result in homeotic phenotypes and alterations gain-of-function mutations in the HOM-C genes cause the cells in the expression patterns of homeotic genes (Gould, 1997; in those regions where the concentration of the homeotic Schumacher and Magnuson, 1997). This indicates a products is altered, to form structures characteristic of a conservation of the PcG function throughout evolution. The different segment of the fly. One of these homeotic phenotypes PcG proteins form large complexes arising from their mutual is the transformation of the second and third legs into first leg, interactions through evolutionary conserved protein motifs as is evident from the presence in the second and third legs of (Kyba and Brock, 1998a,b). The genetic interactions and 5094 M. del Mar Lorente and others dosage effects of PcG genes in Drosophila (Adler et al., 1991; simplex virus thymidine kinase gene driven by the MC promoter Cheng et al., 1994; Jürgens, 1985) and in (Bel et al., (Thomas and Capecchi, 1987) (PyTK) for negative selection. The 1998) agree with PcG function being mediated by multiprotein targeting vector was linearized and electroporated into HM-1 complexes. Studies in Drosophila have shown that PcG embryonic stem (ES) cells as previously described (Porter et al., silencing occurs through Polycomb response elements (PRE), 1996). Colonies surviving the HAT/ganciclovir selection were which are regulatory DNA sequences harboring in vivo binding transferred into 24-well plates and approximately half of each was processed for PCR analysis. The targeted allele was identified sites for PcG proteins (Orlando et al., 1998; Strutt et al., 1997; with primers specific for the last exon of Ring1A Strutt and Paro, 1997). PREs have a modular structure and bind (TGGGGGCGGAGCGTTCACG, oligo 2) and for the HPRT PcG complexes of different composition (Strutt and Paro, minigene (AGCCTACCCTCTGGTAGATTGTCG, oligo 3). Colonies 1997). Targeting of some PcG complexes to DNA is mediated positive for the mutated Ring1A allele were expanded for further by, at least in part, the product of one of the PcG genes, genotyping by Southern blot analysis, using a 5′ external fragment pleiohomeotic (pho), which encodes a zinc finger protein that (probe in Fig. 1A) under stringent conditions. Chimeric mice binds DNA (Brown et al., 1998). However, the formation of generated by the injection of mutated ES cells into Balb/c blastocysts specific PcG complexes and how they affect gene expression were mated to Balb/c mice and heterozygous animals identified by ′ is mostly unknown. PCR, using oligos 2, 3 and an additional primer specific from the 3 Genetic evidence in Drosophila suggests that there are about end of the deleted sequences (TTTGAGCAAGGCTTGCATCG, oligo 1). Mice were also genotyped by Southern blot using a Ring1A 40 PcG genes (Jürgens, 1985), of which about 15 have been cDNA probe. Heterozygous mice were then backcrossed to Balb/c identified and only 13 are molecularly characterized. In an mice. effort to identify new mammalian PcG genes, we found Ring1A and Ring1B, the products of which interacted in a two hybrid Generation of transgenic mice assay with the product of the M33 (also called CBx2; Mouse The Ring1A expression construct was obtained by inserting the Genome Informatics) gene, one of the homologs of the Ring1A cDNA into a modified version of plasmid pCAGGS (Niwa et Drosophila PcG gene Polycomb (Pc) (Schoorlemmer et al., al., 1991). The DNA construct was separated from vector sequences 1997). We and others have shown that Ring1A interacts with by preparative electrophoresis and elution from Elutip columns (Schleicher and Schuell) prior to microinjection in the pronuclei of itself and with multiple PcG proteins: PC2 (also called Pcsk2; × another Pc homolog) and (Satijn et al., 1997; Satijn and fertilized oocytes from superovulated (C57BL/10 Balb/c) F1 females mated with (C57BL/10 × Balb/c) F1 males. Transgenic mice Otte, 1999; Schoorlemmer et al., 1997). Ring1B, however, were identified by Southern blot analysis using the Ring1A cDNA as has also been found to interact with several PcG members: a probe. PH2 (homolog of Drosophila polyhomeotic), PC2 and bmi1 (Hemenway et al., 1998; Satijn and Otte, 1999). Moreover, Skeletal preparations both Ring1 proteins are constituents of nuclear PcG complexes Skeletal preparations were stained with Alizarin Red S and Alcian and co-immunoprecipitate from cell extracts together with Blue 8GX as previously described (Lufkin et al., 1992). Briefly, PC2, PH1 and bmi1 (Garcia et al., 1999; Satijn et al., 1997). carcasses were skinned, eviscerated and fixed overnight in 90-100% Finally, we have found that both Ring1 proteins interact with ethanol. The cartilage was stained with 0.2 mg/ml Alcian Blue in 75% RYBP, a new zinc finger protein which, in turn, interacts with ethanol/25% acetic acid overnight and the skeleton cleared in 2% PHO and with its related mammalian protein YY1 (Garcia et KOH solution overnight. The bone was then stained with 0.075 mg/ml Alizarin Red in 1% KOH for 2 days. Stained skeletons were cleared al., 1999). in 1% KOH/20% glycerol and stored in 50% glycerol/50% ethanol. Although all these previous findings provide biochemical evidence indicating that Ring1A and Ring1B are physically Immunological procedures associated with PcG complexes, genetic evidence is needed to Western blot analysis of total protein extracts from 12.5 dpc embryos demonstrate a PcG function for these proteins. Such genetic were performed as described by Garcia et al. (1999). Antibodies evidence cannot arise from Drosophila studies since, in against Ring1A and Ring1B were described previously (Garcia et al., contrast to all other mammalian PcG genes, no known fly PcG 1999; Schoorlemmer et al., 1997). gene encoding a homolog of Ring1 proteins has yet been Ring1A immunohistochemistry was done on 7 µm cryosections identified. In this study we sought to establish a putative PcG loaded on poly-L-lysine coated slides, using a peroxidase Vectastain function for Ring1A by studying its biological activity in vivo. Elite ABC kit (Vector laboratories) and counterstaining with Hematoxilin and Eosin. For this, we generated mice lacking or overexpressing the Ring1A protein. The findings presented here provide genetic RNA in situ hybridization evidence indicating Ring1A is functionally a PcG member. Mouse embryos were fixed in 4% paraformaldehyde and hybridized to digoxigenin-labelled probes as described previously (Schoorlemmer et al., 1997). The following probes were used: Hoxa4, Hoxb8, Hoxc4, Hoxc5 and Hoxc6 cDNAs (provided by J. MATERIALS AND METHODS Deschamp), Hoxb4 and Hoxb5 cDNAs from R. Krumlauf, Hoxc8 − − cDNA from P. Gruss, Hoxd4 from D. Duboule and Hoxa3 and Gene targeting and generation of Ring1A / mice Hoxd11 (provided by N. Coré). After hybridization, embryos were Genomic Ring1A sequences were isolated from a λ phage obtained embedded in paraffin wax and sectioned. For Ring1A and PLZF from a mouse 129SVJ genomic library. The targeting vector (Fig. probes we used the cDNAs previously described (Schoorlemmer et 1A) consists of two DNA fragments, a 4.2 kb BamHI fragment al., 1997) and for Ring1B, a 1.4 kb cDNA fragment. In timed spanning 5′ flanking sequences and a 2.0 kb BstEII-KpnI fragment pregnancies, noon of the day of appearance of vaginal plug was taken containing the last exon, separated by the mouse phosphoglycerate- as 0.5 dpc (days post coitum). Pregnant females were killed at the kinase (PGK) promoter-driven HPRT minigene (PGK/pDWM1; desired gestation time and embryos were collected from the decidua. Porter et al., 1996). These sequences were then linked to a Herpes Amnion was used for genotype analysis. Ring1A is a PcG gene in mice 5095

Fig. 1. Targeted disruption of the Ring1A locus. (A) Schematic representation of the Ring1A locus, the targeting vector used for mutagenesis and the mutated Ring1A locus. Exons and selectable genes are indicated by black and white boxes respectively. (B) Genotyping of representative offspring by Southern blot analysis. After digestion of tail genomic DNA with EcoRI, hybridization with an external probe (box labelled as probe in A) reveals a 10.8 kb fragment for the wild-type allele and a 9.5 kb fragment for the mutant allele. (C) PCR analysis of tail genomic DNA of the offspring from a heterozygous intercross using oligo 1, oligo 2 and oligo 3 amplifies a 2.2 kb product for the wild-type allele and a 2.5 kb product for the mutant allele. (D) In situ hybridization analysis of Ring1A and Ring1B expression in wild-type and 10.5 dpc Ring1A−/− embryos. Ring1A transcripts are mostly detected in the central nervous system of wild-type embryos while no expression is seen in mutant animals. The rather ubiquitous expression pattern of Ring1B remains unaffected by the Ring1A mutation. (E) Immunoblot analysis of total protein extracts from wild-type and Ring1A−/− embryos. Ring1A is detected as a 55 kDa band in wild-type extracts but not in extracts from Ring1A−/− embryos, whereas a 40 kDa band reactive with anti-Ring1B antibodies was present in extracts from both wild-type and Ring1A-deficient embryos. (F) Immunohistochemistry of Ring1A on tissue sections of 9.5 dpc wild-type and Ring1A mutant mice. The Ring1A signal is evident both in neural tube and mesoderm structures of wild-type embryo, but is absent in the mutant embryo. B, BamHI; E, EcoRI. Scale bar in F is 138 µm in left panel and 88 µm in right panel.

RESULTS mice appeared normal and were fertile. Because of the singular expression of Ring1A in the boundaries between rhombomeres Generation of Ring1A-deficient mice we examined their fate in mutant mice using PLZF, an To inactivate the Ring1A gene, a targeting vector was designed established marker for rhombomere boundaries (Cook et al., 1995). The expression pattern of PLZF mRNA in rhombomere to replace 2.7 kb of the locus with a HPRT minigene by − − homologous recombination (Fig. 1A). This deletes the majority boundaries remained unchanged in Ring1A / embryos (not of the coding region including the translation initiation start, shown). the RING finger, the homology domain 2 and part of the Since the product of the Ring1B gene is structurally related to Ring1A we analyzed its expression in wild-type and homology domain 3 (Schoorlemmer et al., 1997), and is − − expected to result in a null mutant. Ring1A / mice by in situ hybridization and western blot The targeting vector was electroporated into ES cells and analyses. Ring1B transcripts were found distributed among 198 clones analyzed 10 were found to have a Ring1A ubiquitously during embryonic development (Fig. 1D), thus targeted allele. Three of these clones were used to generate overlapping the pattern of Ring1A expression. Moreover, chimeric mice, which passed the targeted allele to their offspring. Ring1B expression remained unaffected in Ring1A-deficient Mice heterozygous for the mutated allele interbred to mice (Fig. 1D,E), thus making functional redundancy feasible. homozygosity close to the expected segregation ratio (24 +/+, 56 +/− and 20 −/− out of 100 mice born) suggesting that the Ring1A-deficient mice show skeletal abnormalities mutation caused no embryonic lethality. Mice homozygous for The analysis of skeletal whole mounts of newborn Ring1A−/− the Ring1A mutation showed no Ring1A transcripts detectable mice revealed a number of abnormalities along the AP axis by whole-mount in situ hybridization to Ring1A mRNA (Fig. (Fig. 2; Table 1). The most penetrant alteration affected the 1D). Moreover, no Ring1A protein was detected either by second cervical vertebra (C2) in all mice, which showed either western blot analysis of total protein extracts or in tissue sections a broadening of the neural arch, or ectopic points of ossification from mutant embryos (Fig. 1E,F). Male and female Ring1A−/− and aberrant cartilage growth (Fig. 2B,C). Also in the cervical 5096 M. del Mar Lorente and others

Table 1. Skeletal alterations in mice lacking and overexpressing Ring1A Ring1A deficient mice (%) Wild type Ring1A+/− Ring1A−/− (16) (14) (31) Newborns Abnormal C1 0 21 52 Abnormal C2 0 57 55 C2→C1 0 36 45 T3→T2 0 7 19 T8→T7† 0 36 42 L1→T13 6% 29 45

(9) (8) (8) 18.5 dpc Vertebrae abnormalities 0 62 100

Ring1A transgenic lines (%) wt 6111 6081 6109 6117 6117 DH (10) (11) (21) (7) (12) (3)* (17)** C2→C1 0 91 95 100 83 100 100 T8→T7‡ 0 0 24 0 0 33 12 T10→T90735271256688 L1→T1305448006670

*Homozygous 6117; **double hemizygous 6117, 6081. Number of mice analyzed are given in parentheses. ‡Unilateral and bilateral abnormalities were counted as being positive in the analysis. C1, atlas; C2, axis; T2, T3, T7, T8, T9, T10, T13, 2nd, 3rd, 7th, 8th, 9th, 10th, 13th thoracic vertebrae; L1, 1st lumbar vertebra. Penetrance of skeletal abnormalities is indicated as percentage of mice analysed.

process of T2 was greatly reduced, suggesting T3 had taken a more anterior identity (Fig. 2D,E). About 42% of Ring1A−/− mice had eight vertebrosternal ribs (on one or both sides) − − Fig. 2. Alteration of the axial skeleton of newborn Ring1A / mice. instead of seven, which is an anteriorisation of the thoracic Views of the cervical (A-E), thoracic (F,G) and lumbar (H,I) regions vertebra T8 into T7 (Fig. 2G,H). In the lumbar region, 45% of of cleared skeletons of newborn wild-type (A,D,F,H) and the mice showed on one or both sides either a rudimentary rib homozygous (B,C,E,G,I) mice. After staining with Alizarin Red or ectopic cartilage not attached to the vertebrae, or both (Fig. S/Alcian Blue 8GX, the bone and cartilage appear red and blue respectively. (A-C) Cartilage and bone abnormalities in the first (C1) 2I,J). This abnormality is consistent with an anteriorisation of and second (C2) cervical vertebra of Ring1A−/− mice. (A) Dorsal the lumbar vertebrae 1 (L1) into T13. −/− view of the cervical region of a wild-type mouse. The split neural When skeletal whole mounts of 18.5 dpc Ring1A fetuses arch of C1 found in about half of the mutant mice is shown in B, were examined (Fig. 3), all mice showed splitting of the neural together with abnormal ossification (arrowhead) and cartilage growth arches of the cervical vertebrae. Moreover, the foramen of C2. In another Ring1A-deficient mice, the C2 vertebra was transversum of C3 to C6 was open (Fig. 3). It is worth noting broadened (C). (D,E) Anterior transformation of the third (T3) into that the penetrance and expressivity of these defects in the axial the second (T2) thoracic vertebra. Lateral view of the cervical region skeleton of Ring1A+/− mice ranged from 40 to 100% of that of −/− of a wild-type (D) and a Ring1A mouse (E). The most prominent Ring1A−/− mice (Table 1). Thus, like the mouse PcG genes, spinous process in the mutant mice is on T3 instead of T2 as in the Ring1A plays a role in axial skeleton formation, although the wild type. (F,G) Anterior transformation of the eighth (T8) thoracic −/− vertebra. Ventral view of the thorax of a wild-type (F) and a mutant anterior homeotic transformation of Ring1A mice contrasts (G) mouse. On the right side of the sternum in the mutant skeleton with the posteriorization of skeletal structures observed in mice there are eight (T8) vertebrosternal ribs attached to it, instead of with mutation in PcG genes. seven (T7) as is the case in the wild type. (H,I) Anterior −/− transformation in the lumbar region of Ring1A−/− mice. Dorsal view Hox gene expression in Ring1A mice of the lumbar region of a wild-type (H) and a mutant (I) mouse The spatially restricted expression patterns of Hox genes are showing an ectopic rib (arrow) on one side of the first lumbar important for establishing positional identity along the AP axis. vertebrae (L1). In dorsal and ventral views anterior is to the top. In It has been previously shown that Hox gene expression is lateral views anterior is to the left. affected in loss-of-function mutations of PcG genes in mice (Akasaka et al., 1996; Coré et al., 1997; Takihara et al., 1997; region, the cartilage of the first cervical vertebra (C1) of about van der Lugt et al., 1996). These data prompted us to examine half of the mice failed to close (Fig. 2A,B). In the thoracic the pattern of mRNA expression of Hox genes from clusters A region of 19% of the mutant mice the spinous process of the (Hoxa3, Hoxa4), B (Hoxb4, Hoxb8) C (Hoxc4, Hoxc5, Hoxc6, third thoracic vertebra (T3) was the most prominent whilst the Hoxc8) and D (Hoxd4, Hoxd11) in 11.5 dpc Ring1A−/− Ring1A is a PcG gene in mice 5097

embryos showed ectopic expression (nearly ubiquitous) and at much higher levels than in wild-type embryos (not shown). Anterior transformations of the axial skeleton in mice overexpressing Ring1A Analysis of the axial skeleton of Ring1A transgenic neonates revealed several alterations along the AP axis in all the lines. These malformations are summarized in Table 1. In the cervical region, the neural arch of C2 was broadened resembling the appearance of the C1 vertebra (Fig. 4A-D). In the thoracic region, anterior transformations were also observed including the transformation of T8 into T7 as indicated by the presence of eight ribs attached to the sternum, instead of seven (Fig. 4E,F). The transitional vertebra of transgenic animals was also transformed anteriorly. In wild-type mice this vertebra is T10 and is defined as the most anterior vertebra to show a lumbar rather than a thoracic articulation. In a large proportion of newborn transgenic animals the dorsal cartilage and the neural arch of T10 took the appearance of T9, thus making T11 the new transitional vertebra (Fig. 4G,H). In the lumbar region, the first vertebra (L1) was anteriorly transformed toward the identity of T13. Either extensive ribs or rib heads were found attached to one or both sides of L1 (Fig. 4I,J). As seen in Table 1, the penetrance of the anterior transformations observed in Ring1A transgenic animals was highly variable. These variations suggested a dosage dependence resulting from differences in the expression levels of the transgene. To test this we generated mice homozygous for the 6117 line and double hemizygous for the 6117 and 6081 lines. In both cases, the penetrance and expressivity of the phenotype became more pronounced when compared to that of the 6117 and/or the 6081 hemizygous animals. The anterior transformations of the axial skeleton resulting from the ectopic expression of Ring1A are similar to those observed in bmi1 Fig. 3. Morphology of cervical vertebrae of 18.5 dpc wild-type and transgenic mice (Alkema et al., 1995). Ring1A targeted mice. Skeletons were stained with Alizarin Red S/Alcian Blue 8GX. Vertebrae from mutant mice show splitting of Hox gene expression in Ring1A transgenic mice the neural arch. Examples of the two types of abnormalities seen in We analyzed the distribution of Hox transcripts in 11.5 dpc C2 are shown. Arrowheads point at the opened foramen transversum embryos of the transgenic lines 6081 and 6111. As for the of vertebrae in mutant fetuses. First to seventh cervical vertebrae are Ring1A−/− embryos, we found subtle changes in the expression indicated by C1 to C7. domains of a reduced number of Hox genes. The changes consisted of a reduction in the signal intensities in the anterior embryos by whole-mount in situ hybridization analysis. We boundaries of expression in the paraxial mesoderm. These found a subtle anterior shift in the anterior border of expression included Hoxb8 and Hoxc8 in embryos of the line 6081, but of Hoxc8 in neuroectoderm, but not in the mesoderm, of all six not in embryos of the line 6111, which only showed alterations Ring1A-deficient embryos analyzed (data not shown). In in Hoxc6 expression in a subset of embryos (data not shown). addition we found, although only in one out of 5 embryos, a Overall, the analysis of Hox gene expression in Ring1A−/− mice subtle anteriorization of the rostral boundary of expression of and in Ring1A transgenic mice suggests that Ring1A can Hoxd4 in mesoderm (data not shown). The expression pattern contribute to the regulation of Hox gene expression. of the other Hox genes analyzed remained unchanged in both the paraxial mesoderm and in the neuroectoderm. DISCUSSION Ectopic expression of Ring1A in transgenic mice We sought to gain further insight into the in vivo function of Cumulative biochemical and molecular evidence suggests that Ring1A by ectopically overexpressing its protein product using the mammalian Ring1A and Ring1B genes are members of the a transgenic mouse model system. For this, we constructed a PcG of genes. However, genetic evidence is needed to establish β-actin/Ring1A plasmid in which the Ring1A coding sequence PcG function for the Ring1 proteins. Here, we have addressed was placed under the control of a CMV enhancer and chicken such question by generating a mouse line with a null mutation actin promoter cassette (Niwa et al., 1991). Four founder mice in the Ring1A locus and transgenic mice lines that overexpress were obtained from which transgenic lines were derived. Ring1A. An additional interest in these mutants is the lack of Analysis of Ring1A transcripts in 9.5-12.5 dpc transgenic a Drosophila PcG gene encoding a homolog of the Ring1 5098 M. del Mar Lorente and others

Fig. 4. Anterior transformations of the axial skeleton in Ring1A transgenic mice. Views of the cervical (A-D), thoracic (E,H) and lumbar (I,J) regions of skeletal preparations of wild-type (A,C,E,G,I) and transgenic (B,D,F,H,J) mice. (A-D) Anterior transformation of C2 towards the identity of C1. Dorsal (A,B) and lateral (C,D) views of the cervical region of a wild-type (A,C) and a transgenic (B,D) newborn mouse. A thickening of the cartilage (B) and bone (D) portion of the C2 neural arch is observed in the transgenic animal. (E,F) Anterior transformation of the eighth (T8) thoracic vertebra. Ventral view of the thorax of a wild-type (E) and a transgenic (F) mouse. In the transgenic mouse there are eight pairs (T8) of vertebrosternal ribs attached to the sternum instead of seven (T7) as is the case of the wild type. (G,H) Anteriorisation of the transitional vertebra in Ring1A transgenic animals. Dorsal view of the thoracic region of a wild-type (G) and a transgenic (H) mouse. In the transgenic skeleton, the dorsal cartilage of the tenth thoracic vertebra (T10) becomes squared in shape thus resembling that of the ninth thoracic vertebra (T9). (I,J) Anterior transformation of the first lumbar vertebra (L1) towards the thirteenth thoracic vertebra (T13). Ventral view of the thoracic to lumbar region intersection of a wild- type (I) and a transgenic (J) skeleton. On the left side of the transgenic L1 there is a rib head consisting exclusively of bone, while on its right side a full rib is seen.

1994) mice show almost no difference from wild-type mice whereas the penetrance of skeletal alterations in bmi1+/− is small and restricted to a subset of defects (van der Lugt et al., 1996). This suggests that the dosage of the Ring1A gene is far more critical for axial skeleton development than that of other mouse PcG genes studied so far. Another singularity of the axial skeleton phenotype of Ring1A−/− mice is that the skeletal transformations are anterior, whereas the ones observed in mice with mutations in the PcG genes bmi1, mel18 and rae28 are posterior (Akasaka et al., 1996; Takihara et al., 1997; van der Lugt et al., 1994). In M33−/− mice, although most skeletal transformations are posterior, they also show the anterior transformation of C2 into C1 (Coré et al., 1997; Katoh-Fukui et al., 1998). Given that PcG genes are thought to be negative regulators of Hox gene expression, the phenotype of PcG mouse mutants has been explained by the posterior prevalence (Duboule, 1991) and by the Hox code models (Kessel and Gruss, 1991) developed to proteins. Both mouse models lead to homeotic transformations interpret the skeletal transformations of mice with mutations in and (weak) deregulation of Hox genes, strongly suggesting that Hox genes. The former states that the effects of Hox genes Ring1A is functionally a member of the PcG of genes. expressed more anteriorly are suppressed by the action of Hox genes with more posterior, overlapping, boundaries of − − Skeletal phenotype of Ring1A / mice expression. The latter states that the combination of Hox genes The mouse line lacking Ring1A we have generated displays a expressed in a region is what determines its identity. According variety of abnormalities in the axial skeleton. It is not evident to these models, the derepression of Hox genes in PcG mutants how such a phenotype arises, as Ring1A transcripts appear to would lead to posterior transformations of those body parts that be restricted to the embryonic nervous system of 9.5-11.5 dpc now mimic the pattern of Hox expression of more posterior embryos (Schoorlemmer et al., 1997). However, we show that regions. However, the above models may not be optimal to Ring1A protein is readily detectable in the paraxial mesoderm interpret PcG phenotypes. First, they often fail to predict of 9.5 dpc embryos. This suggests that Ring1A is expressed in vertebral identities and the direction of homeotic mesoderm structures at levels low enough to be easily detected transformation of vertebrae in Hox mutant mice (Boulet and by whole-mount in situ hybridization. The presence of Ring1A Capecchi, 1996; Jeannotte et al., 1993; Ramirez-Solis et al., in the paraxial mesoderm, although at low levels, could explain 1993; Small and Potter, 1993). Second, mutations in individual its role in axial skeleton formation. PcG genes can affect several Hox genes simultaneously, adding The penetrance and expressivity of the axial skeleton to the complexity of the interpretation of skeletal alterations. alterations in Ring1A+/− mice are much higher than those seen The skeletal alterations in Ring1−/− mice include a delay in in mice heterozygous for mutations in PcG genes. Thus, the development of cervical vertebrae, and in about half of the M33+/− and rae28+/− (Takihara et al., 1997; van der Lugt et al., mice aberrant neural arches in the C2. Interestingly, the latter Ring1A is a PcG gene in mice 5099 are reminiscent of those seen in mice heterozygous for a null to T7 and L1 to T13 (Le Mouellic et al., 1992; Pollock et al., mutation in the Mll locus (Yu et al., 1995), a mammalian 1992). According to an antipodal model, Hoxc8 would interact homolog of the Drosophila trx gene. Mll heterozygous mice with one or more cofactors to form a complex controlling axial show bidirectional transformation of vertebrae along the axial patterning. A functional complex would require a specific skeleton and posteriorization of Hox gene expression. A concentration of Hoxc8 protein so that its absence or its possibility is that the loss of Ring1A function affects the presence at higher levels would result in a non-functional proliferation rate of the growth plate cartilage and/or the complex, and ultimately lead to the same aberrant phenotype. replacement of chondrocytes by osteoblasts. We are currently Whether the subtle alterations in the regulation of Hoxc8 or the examining this possibility. deregulation of another unidentified factor(s) acting as in the antipodal model are responsible of the common features in the Ring1A in the regulation of Hox gene expression Ring1A phenotypes is not known. The skeletal phenotypes of Ring1A-deficient mice are Finally, the phenotype of mice lacking Ring1A, in the associated with minor alterations of Hox gene expression in presence of the more ubiquitous and highly related Ring1B these mice. This observation is most likely due to protein, is an indication that in some cases, different PcG compensation of the Ring1A deficiency by the product of the genes, including paralogs, play specific roles within a given Ring1B gene. A similar compensation effect by PC2 may be regulatory pathway. operating in M33−/− mice, which also show weak alterations of Hox gene expression (Coré et al., 1997). The role of Ring1A We thank Drs J. Deschamps, R. Krumlauf, P. Gruss, D. Duboule and M33 as regulators of Hox genes, therefore, is not as and M. Djabali for providing Hox probes. This work was influential as that of bmi1, Mel18 and RAE28 proteins supported by grants PB94-0089 and PB97-1238 from CICYT, (Akasaka et al., 1996; Takihara et al., 1997; van der Lugt et al., Ministerio Educación y Ciencia of Spain (M. V.). T. M. M. acknowledges financial support through SFB 284 (German Research 1996). These results suggest that even small changes in the Council) and the Bonner Forum Biomedizin. M. L. and J. S. were levels of Hox proteins may lead to changes in the AP patterning supported by fellowships from the Comunidad de Madrid and CSIC, of the axial skeleton. Thus, it would be possible that respectively. quantitative differences in Hox gene expression, with no alterations of the corresponding spatial expression patterns, could take place in Ring1A mutant mice. Consistent with this REFERENCES suggestion is the observation of a reproducible delay in the appearance of in situ hybridization signals for Hoxc8 in Adler, P. N., Martin, E. C., Charlton, J. and Jones, K. (1991). Phenotypic Ring1A−/− embryos compared to wild-type embryos. consequences and genetic interactions of a null mutation in the Drosophila Conversely, in transgenic embryos the hybridization signals for Posterior Sex Combs gene. Dev. Genet. 12, 349-361. Akasaka, T., Kanno, M., Balling, R., Mieza, M. A., Taniguchi, M. and Hoxb8 and Hoxc8 appear later than in wild-type embryos (M. Koseki, H. (1996). A role for mel-18, a Polycomb group-related vertebrate L. and M. V., unpublished). A semiquantitative analysis of the gene, during the anteroposterior specification of the axial skeleton. expression of these genes in mutant and wild-type mice is Development 122, 1513-1522. being carried out. Alkema, M. J., van der Lugt, N. M., Bobeldijk, R. C., Berns, A. and van Lohuizen, M. (1995). Transformation of axial skeleton due to In addition, there is the possibility that Hox gene expression overexpression of bmi-1 in transgenic mice. Nature 374, 724-727. is deregulated in Ring1A mutant mice very early in Bel, S., Core, N., Djabali, M., Kieboom, K., Van der Lugt, N., Alkema, M. development. It is known that early alterations in the J. and Van Lohuizen, M. (1998). Genetic interactions and dosage effects expression of some Hox genes can lead to skeletal of Polycomb group genes in mice. Development 125, 3543-3551. abnormalities which cannot be compensated for at a later Boulet, A. M. and Capecchi, M. R. (1996). Targeted disruption of hoxc-4 causes esophageal defects and vertebral transformations. Dev. Biol. 177, developmental stage, when an appropriate expression pattern 232-249. of Hox genes is observed. An example is the expression of Brown, J. L., Mucci, D., Whiteley, M., Dirksen, M. L. and Kassis, J. A. Hoxd11 and Hoxd10 genes in mice with a mutation in the (1998). The Drosophila Polycomb group gene pleiohomeotic encodes a HoxD cluster (Zakany et al., 1997). A similar transient DNA binding protein with homology to the YY1. Mol. Cell 1, 1057-1064. alteration of Hox gene expression is the anteriorization of the Cheng, N. S. N., Sinclair, D. A. R., Campbell, R. B. and Brock, H. W. rostral boundary of Hoxc9 in the neuroectoderm at early, but (1994). Interactions of polyhomeotic with polycomb group genes of − − − − not later stages, of development in M3 / bmil / mice (Bel et Drosophila melanogaster. Genetics 138, 1151-1162. al., 1998). Cook, M., Gould, A., Brand, N., Davies, J., Strutt, P., Shaknovich, R., Licht, J., Waxman, S., Chen, Z., Gluecksohn-Waelsch, S., Krumlauf, R. Partial similarities of skeletal phenotypes in loss and Zelent, A. (1995). Expression of the zinc-finger gene PLZF at rhombomere boundaries in the vertebrate hindbrain. Proc. Natl. Acad. Sci. and gain of Ring1A function USA 92, 2249-2253. The overexpression of Ring1A leads to the anteriorization of Coré, N., Bel, S., Gaunt, S. J., Aurrand-Lions, M., Pearce, J., Fisher, A. the identities of a number of vertebrae, which is the type of and Djabali, M. (1997). Altered cellular proliferation and mesoderm patterning in Polycomb-M33-deficient mice. Development 124, 721-729. skeletal alterations conventionally expected for PcG gain-of- Duboule, D. (1991). Patterning in the vertebrate limb. Curr. Opin. Gen. Dev. function mutations. Thus, an intriguing finding of this study is 1, 211-216. the partial resemblance between the phenotypes of Ring1A- Duboule, D. and Dollé, P. (1989). The structural and functional organization deficient and Ring1A-overexpressing mice, as indicated by the of the murine HOX gene family resembles that of Drosophila homeotic common C2 to C1, T8 to T7 and L1 to T13 transformations. genes. EMBO J. 8, 1497-1505. Garcia, E., Marcos-Gutierrez, C., del Mar Lorente, M., Moreno, J. C. and A similar scenario is found in Hoxc8 loss- and gain-of-function Vidal, M. (1999). RYBP, a new repressor protein that interacts with experiments, which lead to remarkably similar alterations in components of the mammalian Polycomb complex, and with the the axial skeleton including the anterior transformations of T8 transcription factor YY1. EMBO J. 18, 3404-3418. 5100 M. del Mar Lorente and others

Gould, A. (1997). Functions of mammalian Polycomb group and trithorax (1993). Hoxb-4 (Hox-2.6) mutant mice show homeotic transformation of a group related genes. Curr. Opin. Gen. Dev. 7, 488-494. cervical vertebra and defects in the closure of the sternal rudiments. Cell 73, Graham, A., Papalopulu, N. and Krumlauf, R. (1989). The murine and 279-294. Drosophila clusters have common features of organization and Satijn, D. P., Gunster, M. J., van der Vlag, J., Hamer, K. M., Schul, W., expression. Cell 57, 367-378. Alkema, M. J., Saurin, A. J., Freemont, P. S., van Driel, R. and Otte, A. Harding, K., Wedeen, C., McGuinnis, W. and Levine, M. (1985). Spatially P. (1997). RING1 is associated with the polycomb group protein complex regulated expression of homeotic genes in Drosophila. Science 229, 1236- and acts as a transcriptional repressor. Mol. Cell. Biol. 17, 4105-4113. 1242. Satijn, D. P. and Otte, A. P. (1999). RING1 interacts with multiple Polycomb- Hemenway, C. S., Halligan, B. W. and Levy, L. S. (1998). The Bmi-1 group proteins and displays tumorigenic activity. Mol. Cell. Biol. 19, 57-68. oncoprotein interacts with dinG and MPh2: the role of RING finger Schoorlemmer, J., Marcos-Gutierrez, C., Were, F., Martinez, R., Garcia, domains. Oncogene 16, 2541-2547. E., Satijn, D. P., Otte, A. P. and Vidal, M. (1997). Ring1A is a Jeannotte, L., Lemieux, M., Charron, J., Poirier, F. and Robertson, E. J. transcriptional repressor that interacts with the Polycomb- M33 protein and (1993). Specification of axial identity in the mouse: role of the Hoxa-5 is expressed at rhombomere boundaries in the mouse hindbrain. EMBO J. (Hox1.3) gene. Genes Dev. 7, 2085-2096. 16, 5930-5942. Jürgens, G. (1985). A group of genes controlling the spatial expression of the Schumacher, A. and Magnuson, T. (1997). Murine Polycomb- and trithorax- bithorax complex in Drosophila. Nature 316, 153-155. group genes regulate homeotic pathways and beyond. Trends Genet. 13, Katoh-Fukui, Y., Tsuchiya, R., Shiroishi, T., Nakahara, Y., Hashimoto, N., 167-170. Noguchi, K. and Higashinakagawa, T. (1998). Male-to-female sex Simon, J. (1995). Locking in stable states of gene expression: Transcriptional reversal in M33 mutant mice. Nature 393, 688-692. control during Drosophila development. Current Opinion Cell Biology 7, Kaufman, T. C., Lewis, R. and Wakimoto, B. (1980). Cytogenetic analysis 376-385. of chromosome 3 in Drosophila melanogaster: the homeotic gene complex Simon, J., Chiang, A. and Bender, W. (1992). Ten different Polycomb group in polytene chromosome interval 84 A-B. Genetics 94, 115-133. genes are required for spatial control of the abdA and AbdB homeotic Kennison, J. A. (1993). Transcriptional activation of Drosophila homeotic products. Development 114, 493-505. genes from distant regulatory elements. Trends Genet. 9, 75-79. Small, K. M. and Potter, S. S. (1993). Homeotic transformations and limb Kennison, J. A. (1995). The Polycomb and trithorax group proteins of defects in Hox A11 mutant mice. Genes Dev. 7, 2318-2328. Drosophila: trans- regulators of homeotic gene function. Annu. Rev. Genet. Soto, M. C., Chou, T. B. and Bender, W. (1995). Comparison of germline 29, 289-303. mosaics of genes in the Polycomb group of Drosophila melanogaster. Kessel, M. and Gruss, P. (1991). Homeotic transformations of murine Genetics 140, 231-243. vertebrae and concomitant alteration of Hox codes induced by retinoic acid. Struhl, K. and Akam, M. E. (1985). Altered distributions of Ultrabithorax Cell 67, 89-104. transcripts in extra sex combs mutant embryos of Drosophila. EMBO J. 4, Kyba, M. and Brock, H. W. (1998a). The Drosophila polycomb group protein 3259-3264. Psc contacts ph and Pc through specific conserved domains. Mol. Cell. Biol. Strutt, H., Cavalli, G. and Paro, R. (1997). Co-localization of Polycomb 18, 2712-2720. protein and GAGA factor on regulatory elements responsible for the Kyba, M. and Brock, H. W. (1998b). The SAM domain of polyhomeotic, maintenance of homeotic gene expression. EMBO J. 16, 3621-3632. RAE28, and scm mediates specific interactions through conserved residues. Strutt, H. and Paro, R. (1997). The polycomb group protein complex of Dev. Genet. 22, 74-84. Drosophila melanogaster has different compositions at different target Le Mouellic, H., Lallemand, Y. and Brulet, P. (1992). Homeosis in the genes. Mol. Cell. Biol. 17, 6773-6783. mouse induced by a null mutation in the Hox-3.1 gene. Cell 69, 251-264. Takihara, Y., Tomotsune, D., Shirai, M., Katoh-Fukui, Y., Nishii, K., Lewis, E. B. (1978). A gene complex controlling segmentation in Drosophila. Motaleb, M. A., Nomura, M., Tsuchiya, R., Fujita, Y., Shibata, Y., Nature 276, 565-570. Higashinakagawa, T. and Shimada, K. (1997). Targeted disruption of the Lufkin, T., Mark, M., Hart, C. P., Dolle, P., LeMeur, M. and Chambon, P. mouse homologue of the Drosophila polyhomeotic gene leads to altered (1992). Homeotic transformation of the occipital bones of the skull by anteroposterior patterning and neural crest defects. Development 124, 3673- ectopic expression of a homeobox gene. Nature 359, 835-841. 3682. Niwa, H., Yamamura, K. and Miyazaki, J. (1991). Efficient selection for Thomas, K. R. and Capecchi, M. R. (1987). Site-directed mutagenesis by high-expression transfectants with a novel eukaryotic vector. Gene 108, 193- gene targeting in mouse embryo-derived stem cells. Cell 51, 503-512. 199. van der Lugt, N. M., Alkema, M., Berns, A. and Deschamps, J. (1996). The Orlando, V., Jane, E. P., Chinwalla, V., Harte, P. J. and Paro, R. (1998). Polycomb-group homolog Bmi-1 is a regulator of murine Hox gene Binding of trithorax and Polycomb proteins to the bithorax complex: expression. Mech. Dev. 58, 153-164. dynamic changes during early Drosophila embryogenesis. EMBO J. 17, van der Lugt, N. M., Domen, J., Linders, K., van Roon, M., Robanus- 5141-5150. Maandag, E., te Riele, H., van der Valk, M., Deschamps, J., Sofroniew, Pollock, R. A., Jay, G. and Bieberich, C. J. (1992). Altering the boundaries of M., van Lohuizen, M. and et al. (1994). Posterior transformation, Hox3.1 expression: evidence for antipodal gene regulation. Cell 71, 911-923. neurological abnormalities, and severe hematopoietic defects in mice with Porter, R. M., Leitgeb, S., Melton, D. W., Swensson, O., Eady, R. A. and a targeted deletion of the bmi-1 proto-oncogene. Genes Dev. 8, 757-769. Magin, T. M. (1996). Gene targeting at the mouse cytokeratin 10 locus: Yu, B. D., Hess, J. L., Horning, S. E., Brown, G. A. and Korsmeyer, S. J. severe skin fragility and changes of cytokeratin expression in the epidermis. (1995). Altered Hox expression and segmental identity in Mll-mutant mice. J. Cell Biol. 132, 925-936. Nature 378, 505-508. Preuss, D. (1999). Chromatin silencing and Arabidopsis development: A role Zakany, J., Gerard, M., Favier, B. and Duboule, D. (1997). Deletion of a for polycomb proteins. Plant Cell 11, 765-768. HoxD enhancer induces transcriptional heterochrony leading to Ramirez-Solis, R., Zheng, H., Whiting, J., Krumlauf, R. and Bradley, A. transposition of the sacrum. EMBO J. 16, 4393-4402.