Pax3 acts cell autonomously in the neural tube and somites by controlling cell surface properties

Pax3 belongs to the family of paired-box-containing transcription factors. Mutations in the Pax3 gene lead to developmental defects in mouse and man. Thus, Pax3-deficient mice (Splotch) lack limb muscle, suffer from spina bifida and exencephaly, and exhibit defects in neural crest derivatives (Auerbach, 1954; Moase and Trasler, 1989; Franz and Khotary, 1993). In humans, PAX3 mutations lead to Waardenburg syndrome, an autosomal dominant disorder that consists of numerous defects in neural crest-derived tissues (Tassebehji et al., 1992; Tassebehji et al., 1993). The observed phenotypes exhibit similar defects that suggest loss-of-function mutations. In fact, a mutant mouse generated recently by knocking lacZ into the Pax3 locus by homologous recombination in embryonic stem (ES) cells, displays the same phenotype as Splotch mice (Mansouri and Gruss, 1998). Homozygous Splotch animals die at E14 of gestation and several defects are observed, such as spina bifida and exencephaly (Auerbach, 1954; Franz, 1989; Moase and Trasler, 1989).

Splotch mutants have been extensively characterized in order to elucidate Pax3 function, especially in skeletal muscle (Bober et al., 1994; Goulding et al., 1994; Daston et al., 1996; Tajbakhsh et al., 1997). It has emerged that Pax3 is an important regulator for the migration of myogenic precursors into the limb bud (Williams and Ordahl, 1994; Bober et al., 1994; Goulding et al., 1994; Daston et al., 1996; Mennerich et al., 1998). In addition, genetic and in vitro analysis revealed that MyoD (Myod1 – Mouse Genome Informatics) activation is dependent on either Pax3 or MyF5. Accordingly, two independent pathways have been postulated for the activation of myogenesis in the body (Tajbakhsh et al., 1997; Maroto et al., 1997).

Somite patterning is under the control of various signals provided by the dorsal neural tube, the notochord and the floor plate, the surface ectoderm, and the lateral plate mesoderm (Münsterberg et al., 1995; Pourquié et al., 1996; Hirsinger et al., 1997; Marcelle et al., 1997; Reshef et al., 1998; Tajbakhsh et al., 1998; Yamaguchi, 1997; Currie and Ingham, 1998). The multiple roles for Pax3 in the paraxial mesoderm and the neural tube do not indicate which tissues require Pax3 function. We therefore used Splotch^2H mice (Sp^2H) and lacZ-expressing Pax3 knockout mice (Sp^2G) to generate chimeras, composed of wild-type and Pax3^−/− cells. Chimeric embryos with Pax3^−/− cells could easily be identified by PCR through the presence of the Sp^2H allele (Epstein et al., 1991) and by X-gal staining. Embryos with a high proportion of Pax3-deficient cells display a similar phenotype to Splotch homozygous embryos, and have exencephaly and spina bifida. Moreover, these embryos die at embryonic day (E) 14. All the embryos that survived to term contained either the Sp^2H or the lacZ allele, indicating that high chimeric embryos containing homozygous cells do not survive to term. In order to trace more accurately the Pax3-deficient cells and to follow their behavior in a wild type environment we isolated three Pax3^−/− ES cell lines from embryos derived from the knockout mice generated by the knock-in of the lacZ gene into the Pax3 locus. Two homozygous and one heterozygous ES cell lines were used to generate chimeric embryos by aggregation with wild-type embryos. These chimeric embryos were analyzed at different stages of gestation (E9.0 to E13). The degree of ES cell contribution was monitored by the amount of pigmentation in the eye and by X-gal staining. Our analysis reveals that Pax3-deficient cells participate in forming all Pax3-expressing tissues. Chimeric embryos with high contribution of Pax3-deficient ES cells (>80%) often suffer, as do Splotch embryos, from spina bifida and exencephaly. At early stages of gestation (up to E10), mutant cells do not segregate from wild-type cells. However, later in development mutant cells are detected as isolated clones of blue cells, in the somites, spinal cord, brain and olfactory epithelium. Accordingly, Pax3^−/− cells may have lost or modified some cell surface properties. This is in agreement with earlier observations proposing that Pax genes may be involved in cell-cell adhesion or cell-cell interactions (Chalepakis et al., 1994; Quinn et al., 1996; Stoykova et al., 1997; Collinson et al., 2000, Duncan et al., 2000). In addition, we found that in chimeric mice and in grafts of Pax3^−/− ES cells into chick neural tube, Pax3-deficient neural crest cells are able to migrate. Pax3 function in the neural crest may therefore be related to the maintenance of other properties required post migration. Alternatively, normal neighboring tissue may be necessary for neural crest migration and/or survival.

A Pax3 knockout mouse was generated by replacing the first exon and part of the first intron with the lacZ gene from Escherichia coli (data not shown; Mansouri and Gruss, 1998). This knockout mouse line was designated Sp^2G. The pattern of expression of the β-galactosidase protein from the mutated allele recapitulates the previously described expression of the Pax3 gene (Fig. 1; Goulding et al., 1991). Highly chimeric males that had transmitted the mutated allele to the germline were mated to 129/Sv mice, and born animals carrying the Pax3 lacZ allele were used for the efficient derivation of ES cells. Females with vaginal plugs were opened at E3.5 of gestation (considering day of vaginal plug as E0.5). About 60 blastocysts were prepared as described (Hogan et al., 1994), and cultured for few days on mitomycin treated embryonic fibroblasts in the presence of 500 U/ml of LIF and 20% fetal calf serum (FCS). Sixteen ES cell lines were derived and three of them (Pax3/3, Pax3/9 and Pax3/15) were Pax3-deficient. Genotyping of the ES cells was performed by genomic Southern blot with an external probe used for screening the Pax3 knockout mice (Mansouri and Gruss, 1998 and data not shown).

In preliminary experiments, we generated chimeric embryos by using Sp^2H and the knockout mice (Sp^2G). Using PCR, lacZ and Sp^2H alleles could be traced. Embryos with a genotype including both alleles indicated that homozygous cells are present in the chimeric embryo (Fig. 2). Embryos with heterozygous cells exhibit either the Sp^2H or the lacZ allele. Briefly, embryos at the morulae stage were prepared and the zona pellucida removed by passage through acidic Tyrode solution (Hogan et al., 1994) and aggregated according to the scheme shown in Fig. 2.

Chimeric embryos were also generated by aggregating Pax3^+/− and Pax3^−/− ES cells with morulae from albino NMRI embryos (Table 1). The resulting degree of chimerism could easily be monitored by the amount of pigmentation in the eye. Embryos were prepared at different stages and whole-mount X-gal staining was performed to determine the degree of chimerism in Pax3-expressing tissues.

For this specific experiment, the Pax3/6 heterozygous and Pax3/9 homozygous ES cells were grown on 0.1% (w:v) gelatine in the absence of feeder layers in standard medium containing 2000 U/ml mouse LIF (GIBCO-BRL). All the following procedures have been described previously (Beauvais-Jouneau et al., 1999). Briefly, ES cells were labelled with 10 μM of the vital fluorescent CFSE marker (C-1157, Molecular Probes) to localize them in the chicken embryo. Before grafting, ES cells were stained in DMEM containing 30 μg/ml Neutral Red (N-2889, Sigma) to visualize the cells during the manipulation. Chicken embryos were incubated until they had developed 20-25 somite pairs. Aggregates of 15-30 ES cells were pipetted onto the chicken embryo and gently inserted between a somite and the neural tube through a slit in the ectoderm. The graft was performed at the axial level where neural crest cells had just migrated off the dorsal surface of the neural tube (i.e., the fifth to last somite pairs). After grafting, the eggs were sealed with tape and incubated in humid atmosphere at 37°C for about 18 hours. Migrating ES cells were localized either by confocal scanning fluorescence or X-gal staining.

For whole-mount X-gal staining, embryos were briefly washed in cold phosphate buffered saline (PBS), fixed in a solution containing formaldehyde, glutaraldehyde and NP40, and stained overnight at 30°C for lacZ activity as described (Allen et al., 1988). Stained embryos were cleared in glycerol solution. For histological analysis X-gal-stained embryos were briefly fixed in 4% PFA, washed in PBS and saline, and dehydrated in ethanol and isopropanol before embedding in paraffin. Sections (10 μm) were used for the analysis. X-gal-stained sections were counterstained with nuclear Fast Red (Vector Laboratories). Additionally, vibratome sections (50 μm) were prepared after embedding X-gal-stained embryos in gelatin.

Cell proliferation analysis was studied by the incorporation of BrdU (Sigma) into embryos and labeled cells were detected by the immunohistochemical procedure on paraffin-embedded sections using anti-BrdU antibody (Bioscience, USA). Briefly, foster mothers were injected with 100 μg/g bodyweight bromodeoxyuridine (BrdU) at 12.00pm and 1 hour later the concepti were recovered. Embryos from E10 to E11.5 were processed for BrdU immunohistochemistry. They were fixed in 4% paraformaldehyde, infiltrated and embedded in paraffin. Serial sections (10 μm) were cut and processed for staining. Endogenous peroxidase activity was blocked with H₂O₂, the sections were treated with intervening washes in PBS with HCl, and then Na₂B₄O₇, pepsin, anti-BrdU antibody (Bio-Science Products), biotynlated anti-mouse/rabbit IgG (Vector Laboratories), avidinbionylated horseradish peroxidase complex (Vector Laboratories) and diaminobenzidine.

For TUNEL assay, E10 and E11.5 embryos were fixed in paraformaldehyde and processed as for paraffin-embedded sections. Cells undergoing apoptosis were analyzed by TdT-mediated dUTP-biotin nick end-labeling using in situ apoptosis detection kit (ApoTag, Oncor) according to the manufacturer’s instructions.

Pax3-deficient embryos die at E14 of gestation so that the generation of chimeric embryos containing Pax3^−/− cells requires the use of heterozygous crossings. We used Sp^2H and Sp^2G in order to genotype the chimeric embryos unambiguously. Sp^2H has a 34bp deletion in the Pax3 gene (Epstein et al., 1991). Sp^2G, the knock-in allele of the bacterial lacZ gene into the Pax3 locus recapitulates the expression of Pax3 in the embryo (Fig. 1; Mansouri and Gruss, 1998). Both alleles could be monitored by PCR. In addition, X-gal staining was used to trace the expression of the Sp^2G locus corresponding to Pax3-deficient cells in chimeric embryos. Matings and aggregation of morulae were performed as shown in Fig. 2. We have generated a total of 65 chimeras, from which 19 were +/+ ↔ Sp^2H/Sp^2G; 19 were +/+ ↔ +/Sp^2H; 27 were +/+ ↔ +/Sp^2G and 29 were +/+ ↔ +/+. Chimeras were analyzed between E10 and E12.5 (Table 1). From the chimeras with a compound genotype (+/+ ↔ Sp^2H/Sp^2G), two embryos were necrotic (E12.5), 11 exhibited exencephaly, spina bifida or both defects, and six had no obvious malformation. The observed defects in the compound chimeras are consistent with those described previously in homozygous Splotch embryos (Auerbach, 1954).

In order to monitor Pax3^−/− cells in chimeric embryos directly without the need to make compound embryos, we used the Sp^2G knockout mice to derive ES cells from blastocysts provided by crossing Sp^2G/+ males with Sp^2G/+ females. Three Pax3^−/− and several Pax^+/− ES cell lines were isolated. More than 100 chimeras were produced using Pax3^−/− ES cells aggregated with NMRI morulae. With the Pax3^−/− ES cells Pax3/9, several high chimeric (>80%) embryos were obtained, indicating that mutant Splotch cells are able to colonize all Pax3-expressing tissues such as the spinal cord, the mesencephalon and the dermomyotome, as revealed by X-gal staining (Fig. 3). In addition, a high number of these embryos displayed a Splotch phenotype with defects in the neural tube and dermomyotome (Figs 3, 4). Thus, chimeras generated with two independent methods indicate that Pax3 function is required in the neural tube and the dermomyotome.

As mentioned above, the degree of chimerism is documented by the percentage of blue cells (lacZ expression). In older embryos (E11) this is corroborated by the amount of pigmentation of the eye (Fig. 3; data not shown). In all chimeras, we found a tendency of mutant cells, documented by the lacZ expression pattern, to segregate from wild-type cells. In high chimeric (>80%) embryos, the somites, the neural tube and the olfactory epithelium exhibited only a few wild-type cells trapped in a mass of blue cells (Fig. 5 and data not shown). These embryos also have spina bifida and/or exencephaly, and are reminiscent of Pax3 knockout homozygous embryos (Figs 1, 3). In moderate or low chimeric embryos, however, somites, neural tube and olfactory epithelium display a patchy lacZ-expression pattern, pointing to a failure of mixing capability between mutant (blue) and wild-type cells (white) (Fig. 5). This segregation behavior of mutant and wild-type Pax3-expressing cells is not detected before E10 (Fig. 5). In addition, it is interesting to note that in all chimeras with an open neural tube, no patchy expression of lacZ is observed in the spinal cord or mesencephalon, even in embryos older than E10. This indicates that mutant Pax3 cells still contribute efficiently to the neural tube of the developing embryo.

The limbs of Splotch embryos are devoid of muscle cells and Pax3 was therefore suggested to be responsible for the migration of muscle precursors into the limb (Franz et al., 1993; Bober et al., 1994; Goulding et al., 1994; Daston et al., 1996; Mennerich et a., 1998). We analyzed our chimeras for the capability of Pax3 mutant cells in the lateral dermomyotome to colonize the limb. In high chimeric embryos, the lateral dermomyotome is disorganized and has a ‘fuzzy’ structure. Pax3-deficient cells do not seem to migrate into the limb, where no blue cells can be detected when compared with chimeras generated from Pax3^+/− ES cells (Fig. 4 and data not shown). Our analysis indicates that Pax3 function is required for the structural organization of the dermomyotome and for the migration of muscle progenitors into the limb. In addition, Pax3 mutant cells are not able to mix with wild-type cells (Fig. 5). Accordingly, Pax3 mutant cells have lost or modified surface properties so that a rescue with wild-type cells does not occur. Thus, Pax3 acts cell autonomously in the dermomyotome and the neural tube.

Pax3 is already expressed in the presomitic mesoderm and is found in the newly formed somites. At later differentiation stages, it is confined to the dorsal somite compartment, the dermomyotome (Goulding-etal-1991;-Mansouri et al., 1996a). The dermomyotome is an epithelial structure from which the myotome and the dermatome form. It differentiates through an epithelial-to-mesenchymal transition (Christ and Ordahl, 1995). The integrity of the epithelium is crucial for this process. The analysis of our chimeric embryos revealed that at E9.0 the development of the epithelium from Pax3^−/− cells occurs normally (Fig. 6). At later stages, however, embryos with strong X-gal staining in the somites display severe defects in the structure of the dermomyotome. In these embryos only, the epithelium rather consists of a loosely packed cell mixture where some epithelial organization can be still recognized (Fig. 7D-I). Embryos with strong lacZ expression in the spinal cord but almost no blue staining in the somites have a well-formed dermomyotome (Fig. 7B). This indicates that during differentiation, Pax3^−/− cells cannot maintain the epithelial architecture of the dermomyotome. Strikingly, the analysis using several markers (Myf5, MyoD, myogenin) revealed that only the expression of the gene for the fibroblast growth factor Fgf8 is abolished in the dermomyotome of Sp^2H or highly chimeric embryos (Fig. 8). Fgf8 is normally detected in the rostral and caudal dermomyotomal lip (Fig. 8C); it has been shown to play an important role in cell proliferation and/or migration (Trumpp et al., 1999; Sun et al., 1999). All generated chimeras exhibit remarkable levels of lacZ expression in the neural tube at all stages of development analyzed so far. In contrast, this is not true for the somites. The contribution of Pax3^−/− ES cells to the somites is detected very frequently from E9 to E10 when compared with E10.5 or older embryos. In contrast, control Pax3^+/− ES cells contribute equally to the somites at all stages analyzed (data not shown). Therefore, proliferation and cell survival studies have been performed on chimeric embryos. BrdU labeling and TUNEL assay revealed that only cell survival is affected. In fact, apoptosis is detected in Pax3^−/− cells of the lateral dermomyotome but not in the neural tube (Fig. 9A,G; data not shown).

Several highly chimeric embryos suffer from spina bifida and/or exencephaly, indicating that Pax3 function is required in the neural tube. Our analysis demonstrates that Pax3 mutant cells contribute efficiently to the neural tube. Although Pax3-deficient cells do not intermingle with wild-type cells, they are not excluded from the neural tube during later development. As mentioned above, cell survival is not affected.

Splotch mice and individuals with Waardenburg syndrome suffer from defects of neural crest derivatives. Hence, Pax3 is an important regulator of neural crest cell migration. We therefore wanted to know how Pax3^−/− neural crest cells behave in a chimeric environment with wild type cells. The analysis of several chimeras at E9 and E11 of development revealed that the neural crest cells of the rostral neural tube (hindbrain) seemed to migrate more efficiently than those located caudally. In addition, more caudal neural crest cells were able to migrate and blue Pax3^−/− cells could be detected in dorsal root ganglia (DRG) and spinal ganglia (SG) (Fig. 10A-D; data not shown). Furthermore, DRG of highly chimeric embryos were affected and smaller in size at early stages of development, but were later rescued by wild-type cells. In fact, older embryos (E11.5) always exhibited normally formed SG and DRG (Fig. 5I,J; data not shown). A normal contribution of blue cells to the trigeminal ganglion was also observed (data not shown). Accordingly, in contrast to other tissues, Pax3 does not act cell autonomously in neural crest cells. The contribution of mutant neural crest cells to the formation of melanocytes were therefore analyzed by grafting Pax3^+/− and Pax3^−/− ES cells into chick embryos.

Heterozygous Pax3/6 and homozygous Pax3/9 ES cells were labelled with a vital fluorescent marker (CFSE) and grafted into early trunk neural crest cell migratory pathways. The migratory behavior of the cells was assessed by fixing the host embryos 18 hours after receiving the graft. ES cells were detected in the embryos either by confocal scanning fluorescence microscopy or by histochemistry after staining for β-galactosidase. In the presence of Pax3, wild-type and heterozygous ES cells were able to migrate in both pathways simultaneously (data not shown and Beauvais-Jouneau et al., 1999). In the absence of Pax3, ES cells were also able to migrate in both the dorsoventral and dorsolateral pathways (Fig. 10E,F). Moderate β-galactosidase activity was revealed by whole-mount X-gal staining of Pax3 homozygous and heterozygous ES cells when migrating in chicken embryo (Fig. 10F; not shown). Paraffin sections revealed that cells migrating dorsolaterally and dorsoventrally were expressing lacZ (data not shown).

In the presence or absence of Pax3, the percentage of positively grafted embryos over the grafted embryos was similar. The proportions of embryos in which ES cells were not migrating or migrating exclusively dorsoventrally were similar in heterozygous and in homozygous ES cells (Fig. 10G). In embryos in which ES cells were migrating dorsolaterally and dorsoventrally, the number of cells was determined on each pathway. The embryos were classified as a function of the cells migrating in majority on the dorsolateral (DL) or the dorsoventral (DV) pathway (Fig. 10H). In this respect, in the absence of Pax3, the proportions of DL and DV embryos were equal to 50%. In the presence of wild-type or heterozygous Pax3, the proportion of DL embryos was equal to 80%. These results suggest that ES cells that lack Pax3 have a stronger tendancy to migrate dorsolaterally. Finally, the average distance of migration is similar for wild type, heterozygous and homozygous Pax3 ES cells.

Loss-of-function mutations of the paired-box-containing gene Pax3 lead to developmental defects documented by the Splotch phenotype in mice and the Waardenburg syndrome in humans. These abnormalities include a failure of neural tube closure, dysgenesis of neural crest-derived tissues and the incapability of myogenic precursors of the lateral dermomyotome to migrate into the limb (Mansouri et al., 1996a). In this study we used different Splotch alleles and also Pax3^−/− embryo-derived stem cells to generate chimeric embryos consisting of wild-type and Pax3^−/− cells. Pax3-deficient cells were monitored by the expression of the lacZ gene inserted into the Pax3 locus by homologous recombination. Thus, the amount of Pax3^−/− cells in the affected tissues can easily be traced in the individual embryo. The chimeric embryos provide a system in which Pax3-deficient cells are allowed to mix and interact with surrounding wild-type cells, enabling them to be rescued in the affected tissues. A failure to rescue mutant cells would suggest that Pax3 acts cell autonomously in that specific tissue (Rossant and Spence, 1998; West, 1999).

Somite patterning is under the control of different signals provided by the neural tube and other tissues such as the surface ectoderm, and the axial and the lateral mesoderm (Münsterberg et al., 1995; Pourquié et al., 1996; Hirsinger et al., 1997; Marcelle et al., 1997; Reshef et al., 1998; Tajbakhsh et al., 1998). The multiple roles of Pax3 in the paraxial mesoderm and the neural tube do not disclose which tissue requires Pax3 function.

In chimeric embryos, where only the neural tube exhibits a high proportion of mutant cells, the dermomyotome develops normally. Thus, Pax3 function in the neural tube is not related to signals exerted on the somites. However, the contribution of mutant cells to the dermomyotome leads to disorganized somites, as has been reported for Splotch embryos. In addition, Pax3 mutant cells in the lateral dermomyotome are not able to migrate into the limb. (Franz et al., 1993; Bober et al., 1994, Goulding et al., 1994; Daston et al., 1996; Mennerich et al., 1998). Accordingly, Pax3 acts cell autonomously in the lateral dermomyotome. Furthermore, mutant cells in the neural tube are not able to interact with wild-type cells, resulting in an open neuroepithelium and suggesting a cell autonomous role for Pax3 in this tissue.

The analysis of chimeric embryos further revealed that in the neural tube, somites and olfactory epithelium, mutant and wild-type cells do not intermingle. Pax3^−/− and wild-type cells fail to mix and clear segregation is readily detectable in the affected tissues in whole-mount X-gal-stained embryos. The affected tissues consist exclusively of patches of blue cells (mutant) or white cells (wild type) revealing a distinct boundary. However, this abnormal participation of mutant cells is not observed before E10 of gestation. In addition, the neural tube alone displays segregation of heterozygous mutant and wild-type cells in some chimeras generated by the aggregation of Pax3^+/− ES cells with wild-type embryos. This is in close correlation with the often observed failure of neural tube closure at the posterior neuropore (spina bifida) in heterozygous Splotch embryos (Mansouri et al., 1996a). The failure of mutant cells to mix and interact normally with surrounding wild-type cells suggests that Pax3 may control cell surface properties. Similar observations have been made in chimeras generated with mutant Pax6 Sey cells (Quinn et al., 1996; Collinson et al., 2000). Differences in cell-cell adhesion documented by the expression of various cadherins may confer segregation behavior and thus define cell identity (Takeichi, 1991). In fact, it has been suggested that some Pax genes act on cell surface molecules (R-, N-cadherin), members of the immunoglobulin superfamily (N-CAM, L1) or integrins (Stoykova et al., 1997; Brand-Saberi et al., 1996; Moase and Trassler, 1991; Mansouri and Gruss, 1998; Chalepakis et al., 1994; St Onge et al., 1997; Quinn et al., 1996; Collinson et al., 2000; Kozmik et al., 1992; Duncan et al., 2000). Alternatively, molecules involved in modulating cell surface properties may also act downstream of Pax genes. Such a protein may be the c-met tyrosine kinase, the receptor for HGF/SF (hepatocyte growth factor/scatter factor), which has been proposed to act downstream of Pax3 and initiate the migration of myoblasts from the lateral dermomyotome (Bladt et al., 1995; Daston et al., 1996; Epstein et al., 1996; Yang et al., 1996). Accordingly, the tendency of Pax-deficient cells to segregate from wild-type cells in the affected tissues indicates a common mechanism, reflecting a similar role for Pax genes in various organs.

In addition, the contributions of mutant cells to the neural tube and the somites displayed temporal differences. All chimeric embryos showed a remarkable contribution of blue cells to the neural tube at all stages analyzed, indicating that the Pax3-deficient cells are not excluded from the neural epithelium while development proceeds. In contrast, this is not true for the paraxial mesoderm. Somite contribution of Pax3^−/− ES cells becomes less frequent in older embryos, when compared with controls with Pax3^+/− ES cells. This suggests that although in the neural tube and somites cell surface properties are related to Pax3 function, Pax3 may play different roles in both tissues. In the somites, Pax3 may be necessary for cell proliferation and/or survival as suggested previously (Bernasconi et al, 1996; Amthor et al., 1999; Borycki et al., 1999). Although natural cell death was described in the somites (Cotrina et al., 2000), we think that in the lateral dermomyotome of Sp^2H embryos, observed apoptosis is significant. Our results support the idea that Pax3 function in the dermomyotome is related to cell survival (Borycki et al., 1999). Furthermore, our findings suggest that Pax3 is also required for maintenance of the integrity of the dermomyotome, where it may drive a proper epithelial-to-mesenchymal transition process that is necessary for the formation of the myotome (Christ and Ordahl, 1995). This is in close correlation with earlier observations in Splotch embryos, where after E9.5 the lateral dermomyotome becomes truncated, leading to a loss of epithelial morphology (Daston et al., 1996). The lack of Fgf8 expression in the dermomyotome of Splotch embryos provides further evidence for the role of Pax3 in the cytoarchitecture of this structure. In the absence of Fgf8 cell survival is affected in the first branchial arch. During gastrulation, the lack of Fgf8 causes a failure of cell migration (Trumpp et al., 1999; Sun et al., 1999). Fgf8 may therefore be one of the factors that mediates cell proliferation and/or migration in the dermomyotome. In addition, the expression of Fgf8 in the rostral and caudal dermomyotomal lips points to a role in the epithelial-to-mesenchymal transition during the differentiation of the dermomyotome. In the absence of the Fgf8 receptor, Fgfr1, it was proposed that the primary defect is a deficiency in the ability of cells to make the transition from an epithelial-to-mesenchymal morphology (Ciruna et al., 1997). Altogether, these results suggest an important role for Pax3 in the morphogenesis of the epithelium of the lateral dermomyotome. Fgf8 acts downstream of Pax3 to achieve this function. Pax7 possibly restores the integrity of the medial dermomyotome, as in Pax3/Pax7 double mutants the whole epithelium is truncated (A. M. and P. G., unpublished).

In the neural tube, Pax3 has been also suggested to be necessary for the migration of neural crest cells (Moase and Trasler, 1990). However, grafting of dorsal neural tube tissue from Splotch mice into chick host embryo results in normal neural crest migration (Serbedzija and McMahon, 1997). Our studies clearly provide further evidence that Pax3-deficient neural crest cells are able to migrate from the neural tube and that wild-type cells always rescue neural crest derivatives (DRG and SG). The role of Pax3 in neural crest cells may be related to the maintenance of other properties required post migration, such as proliferation and/or survival. Similar findings have recently been described for cardiac neural crest cells (Epstein et al., 2000). Alternatively, neural crest migration may require interaction with neighboring tissues (Serbedzija and McMahon, 1997; LaBonne and Bronner-Fraser, 1999). As stated above, it is conceivable that these neighboring cues are also related to cell surface molecules, such as extracellular matrix proteins (Duncan et al., 2000). Strikingly, overexpression of the extracellular matrix protein versican has been described and associated with defective neural crest migration in Splotch embryos (Henderson et al., 1997). Our results however, do not correlate with earlier studies that suggest a cell autonomous role for Pax3 in neural crest cells (Li et al., 1999). Our analysis of ES cell grafts into the chick neural tube confirms the experiments reported previously (Serbedzija and McMahon, 1997) and support the idea that Pax3 does not act cell autonomously in neural crest migration. In addition, these results indicate that Pax3 is not necessary for ES cells to migrate in chicken embryo via the DV and DL pathways. A slight difference in the behavior of the cells that lack Pax3 should be noticed. By an as yet unexplained mechanism, the absence of Pax3 seems to favor slightly the DL migration.

We cannot exclude the possibility that Pax3 function in neural crest is related to intrinsic properties, which are required post migration. The difference in the migration potential of mutant neural crest cells between rostral and caudal neural tube of Splotch embryos is most likely related to a redundant function of Pax7 in the hindbrain (Mansouri et al., 1996b; Serbedzija and McMahon, 1997).

In summary, the chimeric analysis using Pax3-deficient ES cells revealed a cell-autonomous function of Pax3 in the somites and neural tube. A common denominator of Pax3 function may be the modulation of cell surface properties, although distinct roles are enacted in various tissues.

We thank D. Treichel, T. Franz, T. Braun, M. Hallonet and P. Collombat for fruitful discussions and suggestions; D. Ferrari for critically reading the manuscript; R. Altschäffel for excellent photowork; and C. Zeden and the animal house crew for taking care of the mice. The excellent technical assistance of J. Krull is greatly appreciated. This work was supported by the Max-Planck Society.