A role for Pax-1 as a mediator of notochordal signals during the dorsoventral specification of vertebrae

The axial skeleton of vertebrates is composed of a metameric series of vertebral bodies and intervertebral discs. Both of these structures are derived from the sclerotome. Sclerotome cells are generated as a consequence of the differentiation of somites into the mesenchymal sclerotome and epithelial dermomyotome. After deepithelialization of somites, medial sclerotome cells migrate and proliferate towards the notochord to form the anlagen of vertebral bodies and intervertebral discs, whereas the lateral sclerotome cells form the pedicles and laminae of neural arches and ribs (Christ and Wilting, 1992).

The notochord has an important role in the dorsoventral specification of not only the neural tube (reviewed by Jessell et al., 1989; van Straaten and Heckking, 1991; Placzek et al., 1990; Yamada et al., 1991), but also the paraxial mesoderm (Watterson et al., 1954; Avery et al., 1956; Cooper, 1965; Youn and Malacinski, 1980). The influence of the notochord on somite differentiation has been analyzed mainly in avian and amphibian embryos treated surgically or chemically to remove the notochord and/or the neural tube. These results suggest that the notochord is required for proper formation of the cartilaginous axial skeleton from somites.

The availability of mouse notochord mutants allows a genetic approach to analyze the function of the notochord in skeletal development. Danforth’s short-tail (Sd) is a skeletal mutant in which the notochord is affected (Dunn et al., 1940). The gene responsible for the Sd mutation maps to the proximal part of mouse chromosome 2 and has not yet been identified. Grüneberg (1958) and Gluecksohn-Schoenheimer (1945) indicated that the abnormalities of the vertebral column in Sd embryos were a result of structural changes of the notochord, particularly, in the tail region of day 10.5 embryos. Histochemical changes of the notochord and the perichordal sheath were shown to occur earlier, suggesting that the primary abnormality is indeed in the notochord (Paavola et al., 1980; Center et al., 1982). Therefore, the various morphological defects of the vertebrae are considered to reflect the absence of the notochord as a primary defect in Sd. Our interest in Sd is based on a genetic interaction between Sd and undulated (un). Crossing the two mutants resulted in an increase in the severity of the skeletal phenotype, particularly in the vertebral bodies of cervical vertebrae (Wright, 1947; Grüneberg, 1953). Since un is caused by a mutation in the Pax-1 gene (Balling et al., 1988; Dietrich, Gruss and Balling, unpublished data), this suggests an underlying interaction between the notochord and Pax-1-expressing sclerotome cells.

Pax-1 is a DNA-binding transcriptional activator identified as a result of its sequence similarity with Drosophila paired-box containing genes (Bopp et al., 1986; Deutsch et al., 1988; Chalepakis et al., 1991). Pax-1 is expressed in the developing sclerotome, pharyngeal pouches, facial mesenchyme, limb buds and thymus during embryogenesis (Deutsch et al., 1988; Koseki and Balling, unpublished data). The Pax-1 locus has been demonstrated to be altered in three undulated alleles; undulated (un), undulatedextensive (un^ex) and Undulated-short tail (Un^s) (Balling et al., 1988; Dietrich, Gruss and Balling, unpublished data). In all un alleles, the development of the vertebral column is affected, resulting in the lack or deformity of vertebral bodies and intervertebral discs (Balling et al., 1991). As the un mutation affects the differentiation of the sclerotome cells into the axial skeleton (Grüneberg, 1954), Pax-1 seems to play a key role in this process.

The purpose of this study was to analyze the role of the notochord and Pax-1 during differentiation of the sclerotome. To this end we analyzed the expression of Pax-1 in Sd mutant embryos. We tried to correlate three aspects; the skeletal phenotype, Pax-1 expression and histochemical changes of the notochord in Sd. Our studies provide evidence that Pax-1 is required for the specification of the ventral parts of the vertebrae in response to notochordal signals.

RSV mice carrying the Danforth’s short-tail (Sd) mutation were obtained from the Jackson Laboratories. Undulated-short tail (Un^s) and undulated-extensive (un^ex) mutant mice were kindly provided by Dr A. M. Malashenko, Krosnogorsk, Russia and Dr J. L. Cruick-shank, Leeds, England, respectively. These mutants were backcrossed onto C57BL/6 background. In the case of Sd mutants, N3 or N4 heterozygotes were used to obtain homozygous embryos. N1 to N4 mice were utilized in Un^s and un^ex mutants. Embryos were recovered on days 9-19 of gestation where day 0.5 is 12 a.m. on the day of detection of the vaginal plug. For the genotyping of Sd embryos, embryonic DNA was extracted from the yolk sac and subjected to a genotype analysis using the microsatellite PCR primers for the D2Mit7 locus (Dietrich et al., 1992). D2Mit7 is localized 6.2±1.7 cM distal from the Sd locus and is polymorphic between C57BL/6 and RSV (Koseki et al., 1993). Genotypes of Un^s and un^ex newborn mice were determined by DNA blot analysis using a Pax-1 probe.

Skeletal preparations were made by an alcian blue-alizarin red staining procedure, using the protocol described by Kessel et al. (1990). Photographs were taken under the stereomicroscope using a flashlight unit originally prepared for the duplication of slides. The number of vertebrae was scored starting with the atlas as the first vertebra.

Total cellular RNA was isolated from day-10.5 p.c. and day-13.5 p.c. embryos by the guanidine/cesium chloride method (Glisin et al., 1974). RNase protection assays were performed as described (Melton et al., 1984). The Pax-1 HincII-SacI paired box fragment which was cloned into the SmaI site of pSPT18 (Deutsch et al., 1988) was linearized with HindIII and was transcribed with T7 RNA polymerase to generate a 377 nucleotide antisense riboprobe. β-tubulin (M 5) was used as an internal control to quantify the amount of Pax-1 transcripts (Wang et al., 1986). pBluescriptIIKS vector carrying the BamHI-SacII fragment was linearized with HindIII and subjected to subsequent in vitro transcription using T7 RNA polymerase to generate a 306 nucleotide riboprobe. Radioactivities of the protected RNA fragments were quantified using a Bio Image Analyzer (Fujix BAS2000, Fuji film).

Embryos were fixed overnight in 4% paraformaldehyde in phosphate-buffered saline (PBS), and then embedded in paraffin wax. Sections were cut at 6 εm and placed onto 3-aminopropyltriethoxysilane (TESPA)-coated slides. The sections were subjected to in situ hybridization as described (Kessel and Gruss, 1991). The Pax-1 probe was transcribed from the same construct as used for RNase protection. For Pax-3 hybridization, a HindIII-PstI 516 nucleotide fragment of the mouse Pax-3 cDNA was used after alkaline hydrolysis (Goulding et al., 1991). M-twist expression was investigated with the riboprobe generated from a XbaI-EcoRI 425 nucleotide fragment (Wolf et al., 1992).

Whole-mount in situ hybridization was performed according to the protocol devised by Rosen and Beddington (1993) using digoxygenin-11-UTP labeled riboprobes. Embryos were cleared using a mixture of benzbenzoate and benzylalcohol (2:1) after equilibration with absolute methanol, and scored and photographed under the stereomicroscope.

The antiserum against the T-protein was raised by immunization of rabbits with bacterially expressed protein and was affinity purified (Kispert and Herrmann, 1993). After fixation in 4% paraformaldehyde in PBS, embryos were soaked in absolute methanol, treated with 1 M Glycine (pH 7-8) and 3% H₂O₂ in PBS, washed briefly with PBST (0.2% BSA and 0.1% Triton X-100 in PBS), incubated in PBSTN (0.2% BSA, 0.1% Triton X-100 and 5% fetal calf serum in PBS) and subjected to staining reaction with affinity-purified T antiserum diluted 1:500 to 1:1000 in PBSTN at 4°C overnight. After extensive washing with PBST, they were incubated with goat-anti-rabbit Ig conjugated with peroxidase diluted 1:400 in PBSTN for 2 hours at room temperature. Peroxidase coupled secondary antibody was detected by incubation with 0.3 mg/ml diaminobenzidine (DAB) and 0.03% H₂O₂ in PBS. The reaction was stopped by washing with PBS twice. Embryos were cleared using a mixture of benzbenzoate and benzylalcohol (2:1) after equilibration with absolute methanol, and then scored and photographed under the stereomicroscope.

In the original studies (Wright, 1947; Grüneberg, 1953) the un-allele was used to show a genetic interaction between Sd and un whereas in the present study we used the Un^s-allele to analyze the basis of this genetic interaction. Un^s is a complete deletion of the Pax-1 locus. The skeletal phenotype of day-17.5 fetuses from matings between Un^s and Sd heterozygotes was investigated. Double heterozygous mice carrying both mutations (Sd/+, Un^s/+) were much more severely affected than either of the single heterozygotes (Fig. 1A-D). Although in the single heterozygotes the vertebral bodies were only slightly affected in the thoracolumbar region, almost a complete loss of vertebral bodies was observed in the thoracolumbosacral region of the double mutants. The cervicothoracic region of these double heterozygotes was less affected than the more caudal part, but an interaction between the two genes was confirmed at all levels of the vertebral column by determination of the dorsoventral diameter of the vertebral bodies (data not shown). The length of the tail was shorter in the compound heterozygotes compared to single heterozygotes.

In order to explain the molecular basis of this genetic interaction we analyzed whether the expression of Pax-1 was changed in Sd embryos. First, the amount of Pax-1 mRNA was compared to that of M 5, one of the β-tubulin gene isotypes, using RNase protection analysis (data not shown). The M 5 gene is a component of the cytoskeleton, which is expressed in a ubiquitous manner (Wang et al., 1986). While no significant change was observed in Sd homozygotes at day 10.5, Pax-1 expression was reduced to 40% of the wild-type level in day-13.5 p.c. Sd homozygotes. Compound heterozygotes between Sd and Un^s express less Pax-1 than either of the single heterozygotes (data not shown). Therefore, an interaction between the Sd and Un^s mutation leads to a reduction in the level of Pax-1 expression.

In order to determine whether Pax-1 expression is affected in a tissue-specific or in a widespread manner, we carried out an in situ RNA hybridization analysis. Based on the results of the RNase protection analyses, Pax-1 expression was first investigated in day-13.5 embryos (Fig. 2). In wild-type embryos Pax-1 was expressed along the entire axis in the region surrounding the anlagen of the vertebral bodies, the pedicles of the neural arches and the proximal parts of the ribs. In addition, Pax-1 expression was observed in the thymus, the mesenchyme of the lower jaw, the sternum and limb buds (data not shown). Parasagittal sections indicated the existence of more than 40 Pax-1-expressing segmental domains along the developing vertebral column. However, in the axial skeleton of Sd homozygotes, Pax-1 expression stopped abruptly in the thoracic region (Fig. 2C). In three embryos analyzed, only 14, 17 and 18 Pax-1-expressing domains, respectively, were found. In Sd homozygotes no cartilaginous tissues were seen in the more caudal region, correlating with the absence of Pax-1 expression (Fig. 2E,F).

The initial changes of Pax-1 expression in Sd homozygotes were traced back to the period between day-9.5 to -10.5 p.c.. At day 9.5 (with 20-25 somites) only a few Sd homozygotes were affected (data not shown). However, at day 10.5 (with 35-40 somites) Pax-1 expression in the posterior part of Sd homozygotes was affected without exception (Fig. 3). In wild-type embryos, Pax-1 expression was observed along the entire body axis from the occipital bone anlage to the newly generated somites in the tail region (Fig. 3 and Koseki and Balling, unpublished observations). Pax-1 expression was stronger in the caudal half and a small area in the most cranial part of each segmental unit (distinguished by bars in Fig. 3). In the cervical region of Sd homozygous and wild-type embryos, no differences in the expression of Pax-1 could be seen (Fig. 3). Pax-1 expression was seen up to the thoracic region of Sd homozygotes and could not be observed in the lumbar, sacral or caudal regions (Fig. 3). Other Pax-1 expression domains, including the forelimb and hindlimb bud and pharyngeal pouches, were not affected (Fig. 3). In seven Sd homozygotes the number of Pax-1-expressing sclerotomes starting from the proatlas anlage was 17 on average. This is similar to what was observed in day-13.5 embryos.

As the expression of Pax-1 is affected in both mutants, Sd and un, we were interested to analyze the phenotypic consequences of the loss or the reduction of Pax-1 expression. A remarkable change in Sd homozygotes was the truncation of the axial skeleton (compare Fig. 1A and E). The severity of defects along the anterior-posterior (A-P) axis of the vertebral column increased in the thoracolumbar region. Sd homozygotes had, on average, 18 vertebral bodies (ranging from 14 to 20), 23 neural arches (ranging from 21 to 25) and 12 ribs (ranging from 10 to 13) indicating that the formation of the vertebral bodies stopped around the 11th thoracic vertebra, whereas neural arches were formed up to the 3rd lumbar vertebra (Fig. 1F). In the caudal part of the truncated axial skeleton of Sd homozygotes, ventral structures such as the vertebral bodies, the intervertebral discs and the pedicles were always much more affected than the dorsal structures, for instance, the laminae of the neural arches (Fig. 1G).

The skeletal phenotype of the thoracolumbar region in Sd homozygotes revealed a significant similarity to that produced by the different un alleles (Fig. 4A). From the 8th to the 10th thoracic vertebra (bold closed arrows in Fig. 4A), the ossification centers of each vertebral body were smaller or absent and occasionally bilaterally separated. These alterations could also be observed in Un^s heterozygotes (Fig. 1B). Between the 11th and 13th thoracic vertebrae the vertebral bodies were not formed and the pedicles of the neural arches protruded towards the midline and eventually fused (open arrows in Fig. 4A). Ribs often directly attach to the pedicle without forming a joint. This phenotype closely resembles abnormalities observed in un^ex and un homozygotes (Fig. 4C; Wallin and Balling, unpublished data). In the lumbar region of the truncated vertebral column of Sd homozygotes, vertebral bodies and intervertebral discs were totally lost and only the neural arches were developed; this closely mimics the phenotype in Un^s homozygotes (arrow-heads in Fig. 4A).

Un^s is the result of a complete deletion of the Pax-1 locus with a semidominant genetic trait (Blandova and Egorov, 1975; Dietrich, Gruss and Balling, unpublished data). undulated-extensive (un^ex) is a recessive mutation caused by the deletion of the last exon of the Pax-1 gene, leading to a reduced amount of Pax-1 transcripts (Wallace, 1985a; Dietrich, Gruss and Balling, unpublished data). The similarity of the phenotype produced by the Sd and un alleles suggests that the abnormalities of the vertebral column of Sd mutants might be the result of a reduction in the amount of Pax-1 expression to a level similar to that found with the different un alleles.

Differentiation of the somite into sclerotome and dermomyotome is known to correlate with changes in the expression of several genes. In wild-type embryos, the expression of Pax-1 and M-twist colocalize in the sclerotome whereas Pax-3 is exclusively expressed in the dermomyotome. M-twist codes for a protein with a helix-loop-helix domain, which was isolated by sequence similarity to the Drosophila twist gene (Wolf et al., 1992). We used M-twist as an additional sclerotome marker. Pax-3 is another member of the Pax gene family which is expressed in the epithelial somite and later in the dermomyotome (Goulding et al., 1991). To assess the sclerotome formation from the somite in the thoracolumbar region of Sd embryos, we analyzed the expression of Pax-1, Pax-3 and M-twist at day 10.5 p.c. in adjacent sections. In Sd homozygotes, expression of Pax-1 and Pax-3 was compared using frontal sections. The expression of Pax-1 in the paraxial mesoderm of the caudal part of Sd-embryos could not be observed at all while the expression in the cranial part was almost normal (Fig. 5A). Pax-3 expression in adjacent sections localized only in the dermomyotome similar to what is seen in wild-type embryos (Fig. 5B; Goulding et al., 1991). The expression of M-twist in the sclerotome extended up to the 20th sclerotome while Pax-1 expression had already stopped at the 18th sclerotome (Fig. 5D,E). This implies that the differentiation of the somites into the dermomyotome and the sclerotome has occurred in this region and suggests that deepithelialization of the somite is independent of the expression of Pax-1.

In order to analyze how and when the notochord of Sd homozygotes displays degenerative changes, the expression of the T (Brachyury) gene product was investigated, using a specific antiserum against a bacterially expressed fusion protein by whole-mount antibody staining (Kispert and Herrmann, 1993). As the first changes in Pax-1 expression can be found at day 10.5, T-protein expression was investigated at day 9.5 and 10.5 (Fig. 6). Wild-type embryos at day 9.5 with 24 somites showed specific staining in the notochord from the infundibulum to the tailbud (Fig. 6A). The staining of the notochord was homogeneous and continuous. Sd homozygotes at the 21 somites stage also had notochord staining along the body axis including the tailbud (Fig. 6B). However, this staining was heterogeneous in the head and the cervical region and also discontinuous in the cervical region (Fig. 6B). Histological analysis revealed that T-protein-expressing cells are more dispersed in Sd homozygotes than in the wild type (data not shown). At day 10.5, wild-type embryos with 37 somites showed clear staining of the notochord and the tailbud as observed in day-9.5 embryos (Fig. 6C,E). Sd homozygotes with 35 somite pairs showed considerable changes of the notochordal staining (Fig. 6D,F). In the cervicothoracic region, where the notochord was already stained at day 9.5, only rudimentary notochordal staining was occasionally seen (open arrow in Fig. 6D). In the caudal region, although the tailbud staining was similar to that found in wild-type embryos, expression of the T-protein in the notochord could not be observed at all. Not even a trace of the notochord seemed to be generated from the tailbud. The same observations were made in histological semithin sections (data not shown). Therefore the critical phase at which the notochord phenotype manifests itself in Sd embryos is between day 9.5 and 10.5 p.c.

In this study, we have begun a genetic and molecular dissection of the events leading to axial skeleton development by use of two mouse mutants, Sd and un. These mutants display primary abnormalities in the notochord and in the sclerotome, respectively (Grüneberg, 1954, 1958). The importance of the notochord for vertebral column development has been documented (Watterson et al., 1954; Grobstein and Holtzer, 1955). Recently Pourquie et al. (1993) have shown that a supernumerary notochord can induce additional vertebral bodies surrounding a transplanted notochord, but the molecular mechanisms of this process are poorly understood. We have analyzed the morphological changes of the notochord in Sd homozygous embryos using antibodies against the T protein, as a specific molecular marker. T is expressed in the entire notochord along the axis and in the tailbud at day-9.5 and -10.5 p.c. of normal mouse embryogenesis (Wilkinson et al., 1990; Kispert and Herrmann, unpublished data ). Changes in the notochord in Sd homozygotes first appear between day 9.5 and 10.5. During this period, the notochord is affected in two different ways in Sd embryos. One is a degenerative change of the notochord (Paavola et al., 1980; Fig. 6). This change manifests itself primarily in the prospective cervical and thoracic region. Another problem involves the tailbud from where the new notochord is generated. In Sd mutants, the tailbud apparently still has the ability to generate a notochord similar to wild-type embryos at day 9.5 p.c. However, this function of the tailbud is lost at day 10.5 p.c. (Grüneberg, 1958; Fig. 6E,F). This means that the notochord is never formed in the prospective lumbosacrocaudal region of Sd embryos. Nevertheless, formation of bilaterally segmented paraxial mesoderm from the tailbud can be seen at least up to day 11.5 p.c.

These observations imply that somites in the cranial part of Sd embryos, generated before day 9.5 of gestation, can differentiate in the presence of the notochord. However, somewhere between day 9.5 and 10.5 p.c. the mutant embryos lose the ability to generate a notochord so that more caudal somites are formed in the absence of any notochordal signals.

In the cranial part of the vertebral column, up to the 18th vertebra on average, the skeletal abnormalities are mild, showing decreased size of vertebral bodies. In this part of the body the somites have differentiated in the presence of a (degenerating) notochord. In the sacral-caudal region, cartilaginous structures are completely absent. The lack of a notochord in this region presumably results in a complete absence of sclerotome formation. In a transition zone between these two regions, comprising around five segments, we observed several vertebrae lacking ventral structures and being composed of neural arches only. Interestingly, this is the region where we can detect sclerotome formation but little or no Pax-1 expression. It seems that in this particular region, where the notochord must have been present only transiently, we can separate the initial event of sclerotome formation and the induction, or maintenance, of Pax-1 expression.

The observed genetic interaction between Sd and both un (Wright, 1947; Grüneberg, 1953) and Un^s (Fig. 1D) prompted us to analyze Pax-1 expression in Sd mutant embryos. The results of that analysis support the notion that the notochord is required for normal Pax-1 expression. First, a quantitative analysis of day-13.5 embryos revealed a significant reduction of Pax-1 expression in Sd homozygotes and heterozygotes. Compound heterozygotes between Sd and Un^s expressed less Pax-1 than either of the single heterozygotes (data not shown). By in situ hybridization of day-10.5 embryos, Pax-1 expression could be observed on average up to the 17th sclerotome in Sd homozygotes. In addition, in the last few segments we could detect a gradual loss of expression (Fig. 5D). We speculate that in the cervicalthoracic region, where we also have a clear interaction between the mutants, Pax-1 expression might be slightly reduced but escaping detection for technical reasons. The notochord is more affected in the caudal part than in the cranial part. This indicates that Pax-1 expression levels in sclerotome cells are tightly correlated with the changes of the notochord. These observations provide a molecular basis of the genetic interaction between Sd and un. In contrast to the Pax-1 expression in the sclerotome, the expression in the hindlimb bud is apparently independent of the notochord.

In another mutant of the notochord, T-curtailed (T^c; Searle, 1966), a similar correlation between notochord changes and Pax-1 expression can be observed in the caudal region (data not shown). Furthermore, recent experiments in the chick support the stimulatory property of the notochord on Pax-1 sclerotome expression. A supernumerary notochord transplanted in a dorsolateral position to the paraxial mesoderm, induces a widened Pax-1 expression domain while suppressing dermomyotome differentiation (Brand-Sabieri et al., 1993). In summary, these data strongly support the idea that notochordal signals induce Pax-1 expression. Whether this signal(s) is a diffusible molecule or is mediated via cell-cell or cell-matrix contacts is currently unknown.

The skeletal phenotypes of Sd and the different un alleles display striking similarities. An analysis of Un^s, un^ex and un homozygotes demonstrates the requirement of Pax-1 expression for proper formation of the ventral part of the vertebra in a dosage-dependent manner while the neural arches are apparently independent of the expression of Pax-1 (Wallin and Balling, unpublished data). Observations on Sd homozygotes and compound heterozygotes between Un^s and Sd reinforce this idea. The number of vertebral bodies seen in Sd homozygotes correlates well with the number of segmental Pax-1 expression domains. Cumulative reduction of Pax-1 expression by Un^s and Sd mutations, predominantly affects the ventral part of vertebrae in the double heterozygous mutants.

The expression of Pax-1 is normally tightly correlated with the formation of the sclerotome during differentiation of somites (Deutsch et al., 1988; Koseki and Balling, unpublished observations). Since the number of sclerotomes can be represented by that of the neural arches, sclerotome formation and Pax-1 expression do not correlate strictly in Sd homozygotes. Sclerotomes are formed even if Pax-1 is not expressed. A comparative analysis of Pax-1, Pax-3 and M-twist also suggests that the process of sclerotome formation itself does not require Pax-1. Findings in Un^s homozygotes further demonstrate this point. Normal sclerotome formation was observed in Un^s homozygotes on day 10.5 p.c. embryos although Pax-1 is completely deleted in this allele (Wallin and Balling, unpublished data).

In the present study, we provide evidence that formation of the vertebral column from the somites should be divided into two distinct processes as suggested previously (Cooper, 1965). The first is the formation of sclerotomes. This process depends on signals from the notochord and/or the ventral half of the neural tube, probably the floor plate (Watterson et al., 1954; Grobstein and Holtzer, 1955), but does not require Pax-1. A second process is the specification of sclerotome cells to obtain ventral or dorsal characteristics. We present evidence that without Pax-1 expression sclerotome cells are unable to form ventral structures of the developing vertebrae. A striking similarity of the skeletal phenotype between embryos with Sd and un alleles suggests that Pax-1 is one of the major responding elements of the notochordal signals within the sclerotome (Fig. 7).

The formation of the dorsal part of vertebrae not only requires sclerotome formation but also a subsequent interaction between the dorsal half of the neural tube, including neural crest cells, and sclerotome cells (Watterson et al., 1954; Detwiler and Holtzer, 1956). This process seems to be independent of Pax-1, but recent studies using the chickquail chimera system indicate that cellular interactions between the surface ectoderm and the neural tube are important for the formation of the spinous process (Takahashi et al., 1992).

Based on the present study, we suggest that both the notochord and Pax-1 expression in sclerotome cells are necessary for formation of the ventral parts of vertebrae. At present the molecular mechanisms by which the notochord determines the dorsoventral (DV) polarity of the paraxial mesoderm and the neural tube are unknown. During mouse development, not only Pax-1 but also at least eight other Pax genes are expressed with a distinct spatiotemporal pattern in the paraxial mesenchyme and/or the spinal cord (reviewed by Gruss and Walther, 1992; Wallin et al., 1993). Notochord transplantation and extirpation experiments in the chick have shown that Pax-3 and Pax-6 expression in the neural tube is dependent on the presence and precise spatial position of the notochord (Goulding et al., 1993). In the case of Pax-3, however, the notochord seems to have an inhibitory effect on Pax-3 expression. In the prospective sacral region of Sd homozygotes, Pax-3 expression becomes extended to the ventral half of the neural tube and of the paraxial mesoderm, as compared to normal embryos (Fig. 5B). Therefore, expression of Pax-3 might be necessary for the specification of the neural cells localized in the dorsal half of the neural tube including the neural crest cells and for the specification of the dermomyotome. Since the mouse mutation Splotch (Sp) is a result of mutations in the Pax-3 locus, an embryological analysis of Sp might be important to understand the role of Pax-3 in the DV specification, not only in the neural tube, but also in the paraxial mesoderm (Epstein et al., 1991; Franz et al., 1993). A genetic interaction between Sd and Sp mutants might occur in a similar manner as seen for un and Sd. Interestingly, one of the modifiers of the Sd mutation maps close to the Sp locus (Wallace, 1985b).

In the lumbosacral region of Sd embryos the loss of the floor plate and the disorganization of motor axon guidance has been attributed to the loss of notochordal signals (Bovolenta and Dodd, 1991). For the specification of the floor plate, not only signals from the notochord (Yamada et al., 1991) but also the competence of the neural cells to respond to these signals are important. The formation of the floor plate is intrinsically affected in the zebrafish mutant, cyclops, where the floor plate is not formed although the signals from the notochord are intact (Hatta et al., 1991). The receptor for notochordal signals or downstream elements of this signal transduction pathway might be disturbed in this mutant, resulting in a loss of competence to form a floor plate. The role of notochordal signals and the competence of sclerotome cells for these signals might be comparable to the specification of the floor plate in the spinal cord. In this system, Pax-1 would be required for interpretation of a notochordal signal in the process of the DV specification of the vertebral column. Therefore, similar mechanisms might exist for the DV specification of the neural tube and the vertebral column (Yamada et al., 1993).

We are grateful to Peter Gruss and Fabienne Perrin-Schmitt for the gift of the Pax 1-8 and M-twist cDNA clones, respectively. We thank Michael Kessel, Wendy Pascoe, Klaus Schughart, Davor Solter, Claudio Stern and Steven Wood for comments on this manuscript and Johannes Schimers for technical assistance. H. K. was supported by an H. F. S. P. fellowship and J. W. was supported in part by an EMBO long-term fellowship and in part by the Swedish Medical Research Council. This work was made possible by financial support from the Max Planck Society and the DFG (Ba 869/3-1).