The concerted action of Meox homeobox genes is required upstream of genetic pathways essential for the formation, patterning and differentiation of somites

A key feature of the body plan of the vertebrate embryo is the presence of somites, which are transient segments of the paraxial mesoderm flanking the notochord and neural tube. Somites form as epithelial blocks of cells, which bud off in a highly coordinated fashion from the anterior end of the unsegmented presomitic mesoderm (PSM). The strict temporal and spatial regulation of somitogenesis is of fundamental importance, because segmentation of structures such as the peripheral spinal nerves, vertebrae, axial muscles and early blood vessels, develops according to the periodicity of somite segmentation. Each somite is subdivided into rostral and caudal compartments that differ in adhesive properties and gene expression, and this differentiation patterns the spinal nerves and ganglia and is also the mechanism that maintains borders between segments.

Experimental and gene-expression data strongly indicate that the generation of somite periodicity and the establishment of rostrocaudal polarity takes place before segment-border formation in the apparently homogenous PSM. Studies of knockout mice have confirmed that the Notch/Delta signaling pathway has a crucial role in establishing both temporal periodicity in the PSM and rostrocaudal polarity in somite primordial (reviewed by Pourquie, 2001; Saga and Takeda, 2001). Evidence of the molecular nature of the oscillator has emerged recently from studies that demonstrate lunatic fringe implements periodic inhibition of Notch signalling to establish a negative feedback loop, which controls the cyclic expression of genes in the presomitic mesoderm(Dale et al., 2003).

The correct formation, patterning and differentiation of somites requires the activity of at least several genetic pathways. Wnt signals from the surface ectoderm are implicated in somite epithelialisation(Borycki et al., 2000), and the paraxis bHLH transcription factor is necessary for the formation of epithelial somites (Burgess et al., 1996; Johnson et al., 2001).

Mutations in the Notch pathway disrupt not only the patterning of the PSM but also anteroposterior polarity of somites(Barrantes et al., 1999). Foxc1 and Foxc2 are both required for the formation of segmented somites, and may function by interaction with the Notch signalling pathway in the anterior presomitic mesoderm (Kume et al.,2001). Whereas mutations in Eph, Ephrin and cadherin genes in mice have not revealed phenotypes - in contrast to zebrafish(Durbin et al., 1998) and frog embryos (Kim et al., 2000) -affecting somite boundaries, a dominant-negative papc (cadherin) molecule disrupts the epithelial organization of cells at the segmental borders between somites in transgenic mice (Rhee et al.,2003).

A large body of evidence, from both in vitro and in vivo experiments(reviewed by Brent and Tabin,2002), indicates that antagonism between different signals from adjacent tissues is required to subdivide the somite into distinct compartments: the ventral mesenchymal sclerotome that generates the chondrogenic axial skeleton, and the dorsal epithelial dermomyotome that forms the skeletal muscles of the trunk, limbs and tongue. Sonic hedgehog (SHH) and noggin are thought to be the ventralising signals for sclerotome induction,and WNT proteins are involved in establishment of the dorsal domain of the somite. BMP signals, originating in the lateral mesoderm, negatively regulate the spatial and temporal activation of somitic myogenesis(Reshef et al., 1998) and positively regulate lateral somitic cell fates(Pourquie et al., 1996; Tonegawa et al., 1997). It is likely that noggin-mediated antagonism of BMP signaling is required for both myotomal and sclerotomal development(McMahon et al., 1998). There is also evidence that SHH changes the competence of target somitic cells to respond to BMPs to induce chondrogenesis(Murtaugh et al., 1999).

The induction of Pax1 and Pax9 gene expression by SHH is necessary for vertebral and rib formation(Peters et al., 1999), and Foxc2 is required for sclerotome proliferation(Winnier et al., 1997). Targeted mutagenesis of the MRF family of bHLH transcriptional activators(MYF5, MYOD1, myogenin, MRF4) in the mouse has revealed an essential, but different, role for members of this gene family in the formation of skeletal muscle (reviewed by Arnold and Braun,2000), and Pax3 and Myf5 are required for the expression of Myod1 in the trunk(Tajbakhsh et al., 1997). Long range signaling by SHH has a role in the induction of Myf5 gene expression in the dorsal somite(Gustafsson et al., 2002).

We have previously described the isolation of the MEOX sub-family of homeobox transcription factors (Candia et al., 1992; Candia et al., 1996). Both Meox1 and Meox2 genes have characteristic expression in the somites of the paraxial mesoderm in vertebrate embryos. Meox1 mutant mice display defects restricted to sclerotomal derivatives, the vertebrae and ribs are fused (S.S., B.M., C.W., V.P. and H.A., unpublished). By contrast, the Meox2 mutation produces a phenotype that affects the development of the limb muscles (Mankoo et al.,1999). Meox2 is required for the expression of Pax3 RNA in migrating limb myoblasts; and also for the induction of Myf5 gene expression, but not that of Myod1, in limb myoblasts. As each single Meox gene mutation affected only a subset of somitic derivatives, despite a largely overlapping expression pattern, this raised the possibility the two genes have overlapping functions and are capable of compensating for each others absence. To investigate the combined function of the MEOX subfamily of homeoproteins, we crossed mutations for both Meox1 and Meox2. The complete absence of Meox gene activity resulted in unexpected and severe defects in somite development. The axial skeleton and most skeletal muscles were not formed. Somite epithelialisation and rostrocaudal somite patterning were also disrupted, as was the maintenance of somite boundaries. Both Meox1 and Meox2 genes were also required for the normal differentiation of cells derived from both the sclerotome and dermomyotome. We propose that the concerted activity of the two Meox genes is an essential component of the genetic circuitry that regulates somitogenesis.

Details of the generation of the two Meox1 mutations are to be published separately (S.S., B.M., C.W., V.P. and H.A., unpublished). Molecular analysis of a transgenic line identified a recessive insertional mutation led to the conclusion that a DNA fragment consisting of 1.8 kb of the interferonα/β-inducible mouse Mx1 promoter, the 1.1 kb HTLV1 Tax cDNA, and part of intron 2 and noncoding exon 3 of the mouse α-globin gene, had inserted fortuitously in first intron of the Meox1 gene and ending 31.2 kb downstream of the stop codon, deleting a total of 45.4 kb of DNA. The allele was given the full designation Meox1^TgN2627ARN and is here referred to as Meox1^im (insertional mutant). A knockout allele in Meox1 was generated by established procedures in ES cells. Upon homologous recombination, the transcription start site, the entire exon 1,including the translational start codon, and 2.3 kb of the flanking intron 1 of Meox1 were deleted. Chimaeric males from four independent clones transmitted the mutation and generated heterozygous animals. Homozygous animals derived from these clones had identical phenotypes. According to nomenclature rules, this allele carries the full designation Meox1^tm1BSM and is here referred to as Meox1^-. The targeted disruption of the Meox2 gene (Meox2^-) been described previously(Mankoo et al., 1999). The Meox1^- and Meox2^-alleles were kept on a mixed C57BL/6//129/Ola background and the Meox1^im allele on a mixed C57BL/6//C3H/HeJ background.

Details on the molecular identification of the transgene insertion of Meox1 and the determination of all mutant genotypes can be obtained on request.

For histology, embryos or tissues were fixed in Bouin's fixative,dehydrated and embedded in paraffin wax. Serial sections (8 μm) were stained with Haematoxylin and Eosin. For semi-thin sections the tissues were embedded in epoxy resin and sections were cut with a glass knife. Whole mount in situ hybridization was performed as previously described(Mankoo et al., 1999). For cryosectioning, embryos were postfixed in 4% paraformaldehyde, equilibrated in 30% sucrose and embedded in OCT. Skeletal preparations of newborn pups were produced using a combination of Alcian Blue and Alizarin Red staining.

To investigate the role of Meox genes during mammalian embryogenesis, we generated mice carrying null mutations at the loci of the known members of this family: Meox1 and Meox2. Two null alleles of the Meox1 locus were used. The Meox1^imallele was produced by the fortuitous insertion of an HTLV-1 Taxtransgene into first intron of the Meox1 locus that deleted the rest of the gene (S.S., B.M., C.W., V.P. and H.A., unpublished), while the Meox1^- allele was generated by targeted mutagenesis in embryonic stem (ES) cells, deleting exon 1. Animals homozygous for either of the two alleles did not express Meox1 mRNA (data not shown) and have abnormalities of the axial skeleton characterised by the presence of hemi-vertebrae, as well as rib, vertebral and craniovertebral fusions, but no apparent defects in skeletal myogenesis (data not shown). Compound heterozygotes for the two Meox1 mutant alleles(Meox1^im/-) had the same phenotype as the single mutant homozygotes (data not shown), supporting the conclusion that both mutations have produced null alleles at this locus. Meox2-deficient animals (Meox2^-/-) have no skeletal defects but are characterised by deficiencies in limb myogenesis(Mankoo et al., 1999). Compound heterozygotes (Meox1^+/im; Meox2^+/- or Meox1^+/-; Meox2^+/-) displayed no abnormalities and were intercrossed to produce mice carrying various combinations of wild-type and mutant Meox alleles. Animals with a single or no wild-type Meox allele were born at the expected Mendelian ratio (of 1/8 and 1/16, respectively) but were severely malformed and died shortly after birth. More specifically, the trunks of such mutants were drastically reduced in length, while the skin was loose and cyanotic. The tail in double homozygous mutants (hereafter referred to as Meox1^-/-;Meox2^-/-) was reduced to a rudimentary stump lacking skeletal elements (data not shown).

Skeletal preparations of animals carrying a single wild-type Meox1allele (Meox1^+/-, Meox2^-/-) displayed defects affecting the axial skeleton; the ribs were normal, and vertebral defects were only apparent at posterior levels(Fig. 1C,G). Animals with a single wild-type Meox2 allele (Meox1^-/-,Meox2^+/-) were more seriously affected; ribs were present but with fusions, and vertebral bodies were split at the lumbar level, while posterior to the pelvic girdle poorly differentiated cartilaginous elements were seen in place of vertebrae (Fig. 1B,F). Skeletal preparations of double mutants(Meox1^-/-,Meox2^-/-) revealed a striking phenotype, these animals lacked a normal vertebral column, which was largely replaced by two strips of fused cartilage, corresponding in position to the neural arches. There was no cartilage or bone present in the ventral midline the location of vertebral bodies; and, although centres of ossification were observed at the cervical and thoracic level of the axial skeleton, neither normal vertebrae nor ribs were observed(Fig. 1D). In addition, no skeletal or cartilaginous elements were detectable at or posterior to the pelvic girdle (Fig. 1D,H) The occipital skull bones, which are somite derived, were hypoplastic, whereas other cranial bones were unaffected (not shown). These observations demonstrate that strong dosage dependent interactions between Meox1and Meox2 are essential for the formation of the axial skeleton; each gene can compensate, to a differing extent, for the absence of the other.

In addition to the skeletal abnormalities, Meox-deficient animals had major defects in the development of the somite derived skeletal musculature. Thus,double homozygous mutants (Meox1^-/-,Meox2^-/-) had a severe depletion of the prevertebral muscles of the head and neck, and also in the epaxial (paraspinal) and hypaxial muscles of the trunk (abdominal wall and intercostal) and limb(Fig. 2C,D). As a consequence,it was no longer possible to identify individual muscles in the mutants. Interestingly, the intrinsic muscles of the tongue, despite originating in somitic mesoderm, were relatively unaffected in the double mutants(Fig. 2A,C). Cranial muscles that do not originate in the somitic mesoderm were generally unaffected (data not shown), including the masseter and extra-ocular muscles that do express Meox genes during embryonic development(Candia et al., 1992). The failure of the epaxial skeletal musculature to develop normally was associated with the absence of the overlying brown adipose tissue(Fig. 2D).

The rib, vertebral and occipital bone defects and the muscle abnormalities of mice lacking MEOX proteins indicate that the combined function of Meox genes is required for normal somitogenesis. To analyse the processes of somite patterning and differentiation in Meox-deficient embryos, we examined histological sections from E9.5-10.5 double mutants. Contrary to control embryos, in which newly formed somites appeared as well defined epithelial spheres that differentiated into a dorsal epithelial dermomyotome and a ventral mesenchymal sclerotome, the newly generated somites of mutant embryos were irregular in shape and failed to epithelialise. The basal lamina that normally surrounds each somite and separates it from its neighbours was no longer detectable between somites, although it was clearly present both dorsally and ventrally to the somites of double mutant embryos(Fig. 3A,B). Moreover, no evidence of differentiation into morphologically identifiable compartments was observed in more mature somites of mutant embryos(Fig. 3C,D). In addition, the segmented organisation of the ventral sclerotome (which normally gives rise to the vertebral bodies, neural arches and pedicles of vertebrae), resulting from the alteration of anterior and posterior sclerotome halves (composed of loose and dense mesenchyme, respectively), was lost in mutants and replaced by a uniform unsegmented mesenchyme (Fig. 3E,F). The alteration of anterior and posterior properties of each somite patterns the dorsal root ganglia (DRGs) and spinal nerves. As a result of the Meox double mutation, DRGs, which normally differentiate in the anterior half of the sclerotome of each somite, were irregular in shape and size and fused together (Fig. 3G,H) and similarly spinal nerves were irregular in spacing and direction (data not shown).

To investigate the molecular mechanisms that underlie the morphological defects in somite patterning and differentiation observed in Meox-deficient animals, we analysed the expression of molecular markers of somitic cell lineages by in situ hybridisation. Meox1 is expressed in the most rostral part of the presomitic mesoderm and throughout newly formed somites,while Meox2 is first expressed concomitant with the formation of epithelial somites. The initial formation of somites is known to be dependent on the function of the Notch signalling pathway in the presomitic mesoderm to establish boundaries and anteroposterior somite patterning(Barrantes et al., 1999; Saga and Takeda, 2001). We examined the expression in the rostral presomitic mesoderm of members of the Notch signalling pathway in Meox double mutants. The presomitic mesoderm expression of neither Mesp2 (Fig. 4A,B) nor Lfng (data not shown) was affected, indicating that Meox gene function is not required for the expression of these genes in the most anterior presomitic mesoderm when segmentation is specified; and supports the histological finding that segmental units are formed in the paraxial mesoderm of Meox double mutants.

The formation of somites emerging at the most anterior end of the presomitic mesoderm was also analysed using paraxis (Tcf15 - Mouse Genome Informatics), a gene that is normally expressed throughout the epithelial somites and in the epithelial dermomyotome of mature somites(Burgess et al., 1995)(Fig. 4C,E). In double Meox mutants, paraxis expression in newly formed somites, albeit segmental, was restricted to a narrow dorsal domain (Fig. 4D,F), similar to that seen with Dll1 (not shown). No signal was detected in more anterior somites indicating that the dermomyotome-specific expression of paraxis was absent. These data suggest that Meox1 and Meox2 are important for epithelialisation events in the newly formed somites and of the dermomyotome. Paraxis is essential for somite epithelialisation(Burgess et al., 1996) and,therefore, the reduced paraxis expression could explain the absence of epithelialisation of newly formed mutant somites and the absence of an epithelial dermomyotome in older somites. Paraxis is also implicated in maintaining the anteroposterior polarity of somites(Barnes et al., 1997; Johnson et al., 2001), and the disruption of paraxis gene expression in the absence of Meox proteins could contribute to the observed defects in somite patterning.

We also examined the expression of Dll1, a member of the Notch signalling pathway, which is expressed in the presomitic mesoderm and the posterior half of newly formed somites and is required for epithelialisation of somites and establishment of their anteroposterior polarity(Hrabe de Angelis et al.,1997). In E9.5 Meox1; Meox2 double mutants, expression of Dll1 in the presomitic mesoderm was unaffected, but expression in the newly formed somites was greatly reduced and restricted dorsally(Fig. 4G,H; data not shown). As visualised by Dll1 expression, the myotomes of anterior somites in Meox1; Meox2 double mutants were fused ventrally but separated dorsally (Fig. 4I,J). This analysis is consistent with a defect in epithelialisation of somites in double mutants. Furthermore, it indicates that in the absence of Meox proteins, the specification of the posterior somitic halves is not maintained resulting in somitic fusions and abnormal patterning of DRGs and spinal nerves.

To determine whether the observed abnormalities in somite polarity were a consequence only of defects in specification of the posterior half of somites,we examined the expression of Epha4. At E9.5, Epha4expression is located in two stripes: a broad posterior stripe at the most anterior border of the presomitic mesoderm and an anterior stripe in the anterior half of the most newly formed somite(Barrantes et al., 1999)(Fig. 4K). In all mutants, the posterior stripe of Epha4 expression in the presomitic mesoderm was not altered. In mutants with only one wild-type Meox allele, the anterior stripe of expression was present but less refined than in controls(Fig. 4L,M); however, in the Meox1; Meox2 double homozygotes, the expression corresponding to the anterior half of the prospective somite was ablated(Fig. 4N), indicating that patterning of the anterior half of the somite is also defective in the Meox mutants.

To further examine somitic differentiation in Meox-deficient animals, we analysed the expression of Pax1 and Pax9, two genes that are expressed in the sclerotome and are critical for normal axial skeleton development (Peters et al.,1999). In E9.5 Meox1; Meox2 double mutants, Pax1mRNA was absent from the paraxial mesoderm throughout the anteroposterior axis, while branchial arch and limb bud expression were unaffected(Fig. 5A-F). In addition, Pax9 expression was dramatically reduced, particularly in the posterior somite halves, resulting in a continuous, albeit anteriorly truncated, band of reduced expression along the anteroposterior axis of the mutant embryo (Fig. 5G-J),further supporting a defect in somite compartmentalisation.

Pax1 mutant animals are characterised by vertebral abnormalities that are milder compared with those observed in Meox1; Meox2 double mutants (Wilm et al., 1998),whereas Pax9-deficient animals have normal axial skeletons(Peters et al., 1998). Although absence of both functional Pax1 and Pax9 genes in one animal results in a more severe defect of the axial skeleton(Peters et al., 1999), such a mutant is less severely affected compared with Meox1;Meox2double mutants. This suggests that additional, non PAX-mediated genetic pathways involved in sclerotome differentiation are disrupted in the Meox double mutants. This is supported by the observed loss, in the paraxial mesoderm of double mutant Meox animals, of expression of Twist(Fig. 5K-N), a gene that is implicated in myogenic and chondrogenic differentiation(Fuchtbauer, 1995; Spicer et al., 1996). To investigate whether the effect on Pax1, Pax9 and Twistexpression was due to a failure of sclerotome specification, we examined the expression of Foxc2, a gene that is expressed in ventral somites and plays a key role in the proliferation of sclerotomal cells(Winnier et al., 1997). We observed that in the mutants uniformly low levels of Foxc2 mRNA throughout the anteroposterior axis replaced the normal segmental pattern of expression (Fig. 5O-R). This presence of Foxc2 transcripts indicates that sclerotome is specified in Meox double mutants; however it fails to differentiate further into anterior and posterior compartments, and their derivatives. Overall, multiple defects in sclerotome differentiation are likely to explain the profoundly severe defects in the formation and patterning of the axial skeleton of Meox1; Meox2 double mutants.

To study the mechanisms underlying the skeletal muscle defects, we examined the expression of essential regulators of myogenesis in the dermomyotome and myotome, such as Pax3 and Myf5(Maroto et al., 1997; Tajbakhsh et al., 1997). In E9.5 Meox1; Meox2 double mutant embryos, the expression of Pax3 was severely reduced in paraxial mesoderm with only a weak signal observed in the ventrolateral region of somites at the level of the forelimb bud (Fig. 6A,B). This signal is likely to correspond to the precursors of limb myoblasts that colonise the limbs of Meox1; Meox2 mutants. This finding indicates that despite the failure of differentiation of an epithelial dermomyotome, at least a certain degree of myoblast specification takes place in Meox double mutants. A similar reduction was observed in the expression of Pax7 (Fig. 6C,D), an additional marker of dermomyotome (Jostes et al., 1990). The dramatic attenuation of Pax3 and Pax7 expression in Meox1; Meox2 mutants indicates that Meox genes function as crucial regulators of genetic pathways upstream of Pax3 and Pax7 gene activation.

We then examined the formation of myotome using Myf5 and myogenin as molecular markers (Tajbakhsh et al.,1997; Smith et al.,1994). In Meox1^-/-;Meox2^-/- mutants, the most caudal and rostral somites did not express Myf5, although inter-limb somites showed low levels of Myf5 expression that did not extend as far dorsally and ventrally as in control embryos (Fig. 6E,F). Interestingly, the expression of Myf5 showed a dependency on Meox gene dose, which was not observed with any of the other markers of somite differentiation described above. Although embryos with one wild-type Meox1 allele showed some loss of Myf5 expression compared with controls (Fig. 6G), those embryos possessing only one wild-type Meox2 allele showed a significant downregulation of Myf5 expression(Fig. 6H). The expression of myogenin, which identifies differentiated myocytes, was affected in a manner that paralleled that of Myf5, low levels were detected only in a ventral domain of inter-limb somites of Meox double mutants(Fig. 6I,J); and Myogenin expression levels also demonstrated a variation that was dependent on Meox gene dosage (data not shown). Overall, our analysis indicates that the combined action of Meox1 and Meox2 genes is a crucial regulator of genes involved in skeletal muscle specification and differentiation.

The loss of Meox1 and Meox2 function from the somitic tissue generates a synthetic phenotype that could not be predicted from the phenotype of either single mutant, demonstrating a strong interaction between the two Meox genes during mammalian embryogenesis. Furthermore, the loss of Meox gene function impacts somite specification and development at several different levels, providing further insight into the gene regulation hierarchies operating during crucial phases of paraxial mesoderm development. A number of gene pathways are affected, involving genes known to have essential functions in somite patterning and differentiation.

Multiple functions of the Notch pathway have been proposed(Barrantes et al., 1999; Takahashi et al., 2000)initially in the presomitic mesoderm (prior to any Meox gene function) to segment the mesoderm and establish rostrocaudal polarity of presumptive somites; and subsequently in nascent somites to regulate somite patterning and boundary formation. The polarity of the somites was disrupted in Meox1; Meox2 mutants as a consequence of defects in patterning of both anterior and posterior halves, with an associated downregulation of Dll1 and Epha4 expression, a phenotype shared with the Dll1 mutants. Our observations indicate that aspects of the phenotype of the Meox1; Meox2 double mutants are partly (via Dll1) explained by a perturbation in the Notch signalling pathway. The irregular and different size of somites is similar to that observed in Notch1 mutants(Conlon et al., 1995), which also supports our interpretation that the Meox genes are involved in the correct transition of cells from the presomitic mesoderm into somites.

The phenotype of Meox1/Meox2 mutants resembles aspects of the Foxc1/Foxc2 compound mutants(Kume et al., 2001); however,the effect of the Foxc mutations seems to impact at the anterior presomitic mesoderm prior to Meox gene expression, indicating that Foxc genes do not function directly to activate Meox expression.

The most dramatic aspect of the Meox double mutant phenotype is the severe loss of both ventral (vertebrae and ribs) and dorsal (skeletal muscles) somite derivatives. The sclerotomal defect in Meox1; Meox2 mutants can be traced back to early defects in the specification and patterning of the ventral somite, as revealed by reduced expression of Pax1, Pax9 and Twist. This interpretation is further supported by the abnormal dorsal restriction of paraxis expression in the newly formed somites. The absence of Pax1 induction in the somite phenocopies the effect seen in the absence of hedgehog signalling in the mouse embryo(Zhang et al., 2001) and suggests that a function of Meox genes may be to mediate the response of somitic cells to hedgehog signals. The severe defects in skeletal muscles were also due to early defects in the patterning and differentiation of the myogenic derivatives of the dorsal somite, as revealed by alterations in Pax3, Pax7, Twist, Myf5, myogenin and paraxis expression.

Our data indicate an essential requirement for Meox activity in both chondrogenesis and myogenesis in the somite, differentiation pathways that have been considered to be mutually exclusive. One mechanism by which this may occur is based on the requirement for Pax gene activity in chondrogenesis(Pax1 and Pax9) (Peters et al., 1999) and myogenesis (Pax3 and Pax7)(Tajbakhsh et al., 1997; Seale et al., 2000) and our observation that both Meox genes are co-expressed with all four of these Pax genes in somitic mesoderm. We have previously demonstrated that in migrating limb myoblasts, which express only Meox2 and not Meox1,Meox2 is upstream of Pax3(Mankoo et al., 1999). In the present study, we have shown that the absence of Meox gene activity from somitic mesoderm disrupts the expression of all four somite-expressed Pax genes. Therefore, the differentiation of somite derivatives into cartilage and muscle requires the Meox-dependent expression of Pax gene function. Interestingly, we have also observed that Meox proteins can interact with Pax1 and Pax3, indicating that there may be cooperativity in the action of these proteins during somitogenesis (Stamataki et al., 2001).

The patterning and differentiation of somites is governed by complex interacting signals that originate in adjacent tissues: neural tube, lateral plate mesoderm and surface ectoderm(Borycki and Emerson, 2000; Correia and Conlon, 2000; Gossler and Hrabe de Angelis,1998). It is clear that the competition between antagonistic signals is largely responsible for the patterning of somites and the subsequent fate of the cells in the different somitic domains (reviewed by Brent and Tabin, 2002). These signals include SHH, noggin, WNT and BMP proteins. Signalling by WNT and SHH molecules, which have been shown to act at a distance greater than the length of a somite in vitro (Fan et al.,1997; Fan et al.,1995), appears to be responsible for the subdivision of the somite into dorsal and ventral subdomains respectively. Furthermore, Sfrp2 is a SHH-inducible WNT antagonist that can block the dermomyotome-inducing properties of WNTs in explants (Lee et al., 2000) and, conversely, GAS1 may function as a WNT—induced inhibitor of SHH activity in the dorsal somite(Lee et al., 2001). BMP signals can negatively regulate the spatial and temporal activation of somitic myogenesis (Reshef et al.,1998) and sclerotome induction by SHH(McMahon et al., 1998), and positively regulate lateral somite fates(Pourquie et al., 1996; Tonegawa et al., 1997). The suppression of BMP signals by noggin is probably required for both myotomal and sclerotomal development (McMahon et al., 1998). There is also evidence of interactions of these signals, for example, SHH regulates competence of cells to respond to BMP. In the absence of SHH, BMP signals result in lateral plate gene expression, but following prior exposure to SHH cells respond to BMP by inducing chondrogenesis in explant cultures(Murtaugh et al., 1999). As the dorsoventral and mediolateral subdivision of the somite is affected profoundly in the Meox1;Meox2 mutants, it suggests that these genes may function to provide competence to the somitic cells to respond to one or more of these signals.

Whereas the formation of somite boundaries and the initial establishment of rostrocaudal polarity in the presomitic mesoderm are genetically separable(Nomura-Kitabayashi et al.,2002), and take place prior to the expression of both Meox genes,it is clear from our studies that the maintenance of boundaries and polarity in newly formed somites are not separate events and require the activity of both Meox genes. Evidence that interactions between compartments occur during somitogenesis is provided by the observed vertebral defects in Myf5;paraxis double mutants, which indicate an indirect role for Myf5 in the development of the axial skeleton (A. Rawls, personal communication). Furthermore, the defects in the lateral sclerotome derivatives in Pax3 mutant mice may result from a disruption in the interaction between Pax3-expressing dermomyotome and the non-expressing sclerotome (Henderson et al.,1999). Whereas the initiation of expression of Pax3 is independent of paraxis, the maintenance of Pax3 expression in the dermomyotome requires paraxis(Wilson-Rawls et al., 1999). Therefore, in the dorsoventral dermomyotome paraxis may function as an intermediate in the regulation of Pax3 expression by Meox genes.

The extreme nature of the Meox1; Meox2 double mutant phenotype may be explained by one of two models: (1) The perturbation of a single early event in somite formation that results in failure of somitic cells to respond to one or more inductive signals from surrounding tissues; or (2) a synergistic perturbation of several somite patterning and differentiation pathways, with a compounding effect on the defects occurring in individual somite compartments. The expression pattern of the Meox genes is consistent with both hypotheses - which are not mutually exclusive, in any case - and much work will be required to resolve the complex combinatorial effect of these genes. Overall our studies demonstrate that Meox homeobox genes function in a co-ordinated manner to regulate critical processes that effect the development of somites.

We thank the Medical Research Council UK for supporting this work and H. Boyes for animal husbandry. We also thank Drs P. Rigby, T. Heanue, I. Rodrigo-Castro and K. Imai for critical reading of the manuscript and many helpful suggestions. Dr A. Rawls kindly provided unpublished data. We thank the following for generously providing probes: Dll1 (D. Henrique), EphaA4 (D. Wilkinson), Foxc2 (B. Hogan), Mesp2 (Y. Saga), Myf5 and myogenin (P. Rigby), paraxis (A. Rawls), Pax1 and Pax9 (R. Balling), Pax3 and Pax7(P. Gruss), and Twist (F. Conlon).