Zebrafish paraxial protocadherin is a downstream target of spadetail involved in morphogenesis of gastrula mesoderm

During vertebrate gastrulation, movements of cohesive sheets of cells lead to the formation of a body plan composed of three germ layers. Although many of the regulatory transcription and growth factors involved are being identified (Lemaire and Kodjabachian, 1996; Harland and Gerhart, 1997), little is known about the structural molecules that mediate cell interactions and cell movements during gastrulation. In zebrafish spadetail (spt) mutants, cells of the lateral mesoderm fail to converge properly toward the dorsal midline, leading to decreased numbers of somitic cells in the trunk and to an accumulation of the corresponding cells in the tail region (Kimmel et al., 1989; Ho and Kane, 1990). At late blastula stages, spt⁻ mesodermal precursor cells on the dorsal side are less tightly packed than in wild-type embryos (Warga, 1996), an observation that suggests that altered cell adhesion may contribute to the cell movement defects.

The spt mutation affects predominantly the paraxial (somitic) mesoderm, whereas mutations in other genes, such as no tail (ntl, a T-box transcription factor homologous to mouse Brachyury) and floating head (flh, a homeobox gene homologous to the organizer-specific Xenopus gene Xnot), affect the axial (notochordal) mesoderm (Halpern et al., 1993, 1995; Schulte-Merker et al., 1994a,b; Talbot et al., 1995; Melby et al., 1996; Odenthal et al., 1996). An interesting gene interaction between spt and flh has been observed in double mutant studies (Amacher and Kimmel, 1998). flh single mutants lack a notochord, and instead axial mesoderm adopts a paraxial (muscle) fate (Talbot et al., 1995; Halpern et al., 1995; Melby et al., 1996). However, in spt;flh double mutants, this transfating event does not occur and, surprisingly, the formation of an anterior notochord, including expression of ntl, is restored. These results suggested that in normal development flh antagonizes spt function in axial mesoderm (Amacher and Kimmel, 1998).

In amphibians the mesodermal mantle involutes as a coherent sheet of cells, which is subsequently subdivided by the appearance of the notochord in the dorsal midline (Vogt, 1929; Keller et al., 1992). A molecule that could provide cohesion to the trunk mesodermal layer is paraxial protocadherin (papc), a Xenopus gene recently isolated during a screen for molecules expressed in Spemann’s organizer (Bouwmeester et al., 1996; S. H. Kim et al., unpublished observations). Protocadherins belong to a large family of transmembrane cell surface proteins from which the cadherins have evolved (Suzuki, 1996). The classical cadherins are strong homophilic cell adhesion molecules present in most solid tissues (Takeichi, 1995; Gumbiner, 1996). In comparison, protocadherins tend to have weaker cell adhesion activity in cell culture (Suzuki, 1996). In a recent study, neural fold protocadherin (NFPC) mRNA has been shown to mediate strong adhesion in ectoderm when microinjected into Xenopus embryos (Bradley et al., 1998).

Xenopus papc is expressed in the paraxial mesoderm of the trunk in regions comparable to those abnormal in spt⁻ zebrafish, and when overexpressed can mediate cell adhesion activity in aggregation assays of embryonic cells (S. H. Kim et al., unpublished observations). Thus, papc seemed a reasonable candidate for either encoding the spt gene or a close spt downstream target, prompting us to clone its zebrafish homologue. The observation that a deficiency allele of spt still contained papc DNA sequences (S. L. A. and A. Y., unpublished observations) eliminated the first possibility, and here we provide evidence that the second is the case. As this work was in progress, a cDNA encoding a T-box transcription factor was found to encode spt (Griffin et al., 1998).

In this paper we present studies on the role of zebrafish papc during mesodermal morphogenesis. During gastrulation, papc is initially expressed in dorsal mesodermal cells, but expression subsequently disappears from midline cells that will give rise to notochord. papc is not expressed in spt mutant gastrulae, demonstrating that papc lies downstream of spt. spt and papc transcripts are ectopically expressed in axial mesodermal cells in flh⁻ embryos, indicating that flh functions to repress expression of spt and papc in the midline. In microinjection studies, a dominant-negative secreted form of papc leads to phenotypes consistent with decreased convergence movements in the mesodermal layer, particularly in cells that give rise to somitic muscle. We propose that the PAPC cell adhesion molecule is a downstream effector of the spadetail transcription factor, and that PAPC may be a component of the molecular machinery that executes some of the morphogenetic cell movements during gastrulation.

To isolate zebrafish papc cDNA, we first tried RT-PCR cloning with degenerate primers using total RNA, isolated from 60-80% epiboly embryos. Degenerate primers were designed as in Sano et al. (1993) with a slight modification based upon the Xenopus papc sequence ((S. H. Kim et al., unpublished observations). The primers F5′-CCG-AATTCAA(A/G)(C/G)(C/G)NNTNGA(C/T)T(A/T)(C/T)GA-3 ′; R5 ′-CCGGATCCNNNGGNGC(A/G)TT(A/G)TC(A/G)TT-3′ generated three different protocadherin fragments (zPCR 1-3) that did not correspond to zebrafish papc homologues. To clone zebrafish papc a somite stage λgt10 cDNA library, a kind gift of H. Okamoto (Inoue et al., 1994), was screened at low stringency with a Xenopus papc probe (nucleotides 2186-3014), which encodes the transmembrane and cytoplasmic domains. A full-length clone was isolated and its 3156 nucleotides were sequenced in both strands.

Embryos were obtained from the AB or ABC lines at the University of Oregon zebrafish colony and raised in embryo medium as described by Westerfield (1995). The origin and maintenance of mutant lines, flhⁿ¹ and spt^b104, are described in Talbot et al. (1995) and Amacher and Kimmel (1998), respectively. Live embryos were photographed using Nomarski optics after orienting in 3% methylcellulose.

Digoxygenin-UTP-labeled or fluorescein-UTP-labeled RNA probes were synthesized using T7 polymerase from XbaI-linearized myoD template (Weinberg et al., 1996), an EcoRI-linearized spt template (Griffin et al., 1998), and an XbaI-linearized gsc template (Stachel et al., 1993) and using T3 polymerase from an ApaI-linearized papc template. Two-color in situ hybridizations were performed as described (Jowett and Yan, 1996); the fluoresceinated probe was detected using Fast Red (Sigma) or using BCIP alone (no NBT), which gives an aquamarine color. Single probe hybridizations, mounting and photography were done as described previously (Amacher and Kimmel, 1998). Embryos were sectioned after whole-mount in situ hybridization as described (Halpern et al., 1995).

Full-length papc cDNA was subcloned into pCS2+ expression vector (a gift of R. Rupp) and designated pCS2 FL-papc. To generate a dominant-negative form of papc mRNA (DN-papc), pCS2 FL-papc was digested with XhoI and religated (pCS2 DN-papc). pCS2 FL-papc and pCS2 DN-papc were linearized with ApaI and transcribed with SP6 RNA polymerase using the Ambion Message Machine Kit (Ambion). All mRNAs were quantified by gel electrophoresis and spectrophotometry.

Analysis of the adhesion activity of zebrafish papc in Xenopus embryos injected into single blastomeres at the 32-cell stage was carried out as described (S. H. Kim et al., unpublished observations). Zebrafish embryos were injected at the 2-4 cell stage using air pressure. After injection, embryos were raised in embryo medium (Westerfield, 1995) and fixed at the 8-9 somite stage. Injected embryos were analyzed morphologically (before fixation) and by in situ hybridization with myoD probe. In some experiments 500 pg of nuclear lacZ mRNA was co-injected with DN-papc mRNA. Embryos were fixed with 4% paraformaldehyde for 1 hour and processed to detect lacZ activity using salmon-gal (Biosynth International), before in situ hybridization.

FL-papc-HA and DN-papc-Flag were generated by PCR, adding the carboxy-terminal amino acid sequences YPYDVPDYA and DYKDDDDK, respectively. Embryonic human kidney cells (293T) were cultured in D-MEM containing 10% fetal calf serum and transfections carried out with Lipofectamin (GIBCO BRL). After 48 hours, condition medium and cell pellets were collected and analyzed using the ECL plus Western Blotting kit (Amersham) according to the manufacturer’s instructions.

Initial attempts to isolate the zebrafish homologue of Xenopus PAPC made use of a PCR strategy, and DNA fragments for at least three different zebrafish protocadherins were isolated (indicated as ZPCR 1-3 in Fig. 1). However, these sequences had amino acid deletions or insertions in the amplified region with respect to Xenopus PAPC, suggesting that they were not true orthologues. We isolated a zebrafish papc homologue using a low stringency hybridization approach with a probe spanning the cytoplasmic domain of Xenopus papc. The zebrafish papc cDNA clone is 3156 base pairs long and encodes a complete open reading frame of 950 amino acids (GenBank accession #AF042191). It was isolated by virtue of two short regions of high conservation to the probe in the cytoplasmic domain (Fig. 1A), one of which has a stretch of 67 nucleotides sharing 80% identity between Xenopus and zebrafish papc. In addition, unlike the sequences of our initial PCR fragments and of other protocadherins available in the database, the regions corresponding to extracellular cadherin repeat 4 in both zebrafish and Xenopus PAPC align without amino acid insertions or deletions, as shown in Fig. 1B. When the entire protein sequences of zebrafish and Xenopus PAPC are aligned, an overall identity of 43% (and 63% similarity using the GCG program) is observed. In comparison, zebrafish PAPC is 29% and 31% identical to human protocadherins 1 and 2 (Sano et al., 1993), respectively.

Several lines of evidence indicate that our cDNA clone corresponds to a protocadherin and not to a classical cadherin. First, papc is predicted to encode a transmembrane protein with six extracellular cadherin (EC) domains instead of the five domains present in classical cadherins. Second, PAPC lacks signature amino acid sequences present in EC3 and EC5 of classical cadherins (Suzuki, 1996). Finally, the cytoplasmic domain of PAPC lacks a β-catenin binding site present in all classical cadherins (Ozawa et al., 1989) but absent from all protocadherins (Suzuki, 1996). Zebrafish papc was cloned because of a highly conserved stretch of 26 amino acids in the middle of the cytoplasmic domain, which is rich in Ser and Asp (Fig. 1A). This intriguing conserved region might provide a binding site for intracellular components regulating papc function. To summarize, we have isolated a zebrafish cDNA clone that appears to be a close homologue of Xenopus papc (S. H. Kim et al., unpublished observations). This conclusion is based on sequence comparisons and on close parallels in expression patterns during development, as described in the following section.

Before gastrulation starts, at 30-40% epiboly, papc transcripts are found on the dorsal side of the marginal zone (Fig. 2A). During early gastrulation (6 hours, shield stage, all stages according to Kimmel et al., 1995), papc expression expands to encompass the entire blastoderm margin, including the ventralmost cells (Fig. 2B). As gastrulation proceeds, papc expression decreases in dorsal midline cells corresponding to the future notochord (Fig. 2C). By midgastrulation, papc transcripts are completely absent from the dorsal midline and the band of papc expression widens as the mesoderm continues to involute in the marginal zone (Fig. 2D, arrowhead). At the end of gastrulation a sharp anterior border becomes apparent, separating trunk mesoderm from the more anterior head domain (arrowhead in Fig. 2E), and the papc expression domain includes ventral mesodermal and future tailbud cells (Fig. 2F). Histological analyses show that papc expression is confined to gastrula mesoderm; at shield stage papc mRNA expression spans the dorsal midline region, but at later stages it is repressed in the notochord, persisting in paraxial mesoderm (Fig. 2K-M).

During somitogenesis, papc has a second, very dynamic, phase of expression. As each new somite forms, the anterior border of expression is displaced towards the posterior (Fig. 2G-I), until papc mRNA fades away as somite formation is completed in the tailbud (Fig. 2J, arrowhead). Four bilateral pairs of bands appear in paraxial mesoderm, and posterior to them the segmental plate (consisting of unsegmented somitic precursor cells) is stained uniformly and less intensely. Histological sections indicate that the first band is located in the anterior border of the newest somite formed and the second in the forming somite as it emerges from the segmental plate (Fig. 2N). The two posterior, stronger bands are located within the segmental plate mesoderm (Fig. 2N). To confirm that the papc bands of expression form before somite formation, we examined the extent of overlap of papc and myoD. myoD is prominently expressed in each formed somite, as well as one or two fainter bands in the newest forming somite and in the segmental plate (Weinberg et al., 1996). Double-label in situ hybridization shows that the two strong papc bands are indeed found within the segmental plate (Fig. 2O). Thus, papc bands are detected in regions that have been proposed to form somitomeres, preceding overt segmentation (Meier, 1979). Several genes expressed in striped patterns in the segmental plate have been described recently (Müller et al., 1996; Jen et al., 1997; Palmeirim et al., 1997) and papc can be added to this growing list. Although papc has an interesting late phase of expression during somitogenesis, this study will focus on the genetic regulation of papc during gastrulation by spadetail and floating head and on the potential role of PAPC in mediating morphogenetic movements.

The spt gene is required for the convergence of mesodermal cells towards the dorsal side during gastrulation (Kimmel et al., 1989; Ho and Kane, 1990). To test whether PAPC, a cell adhesion molecule expressed in gastrulating mesoderm, might participate in these cell movements, we examined its expression in spt⁻ embryos that carry a null mutation (spt^b104, Amacher and Kimmel, 1998; Griffin et al., 1998). As shown in Fig. 3, expression of papc is strikingly affected in spt⁻ embryos. During early gastrulation, mesodermal papc transcripts are greatly reduced in 25% of embryos obtained by intercrossing spt heterozygous fish (Fig. 3A′,B′). At the completion of gastrulation and during early somitogenesis, a few papc-positive cells are found in spt⁻ embryos, located in regions where adaxial cells would be formed (adaxial cells are slow muscle precursors; Devoto et al., 1996); importantly, no expression is found in any other cells including the prospective tailbud region (compare Fig. 3C-C′ and D-D′). The bands of papc expression formed during somitogenesis are also under spt regulation (Fig. 3D-D′). However, as development proceeds, papc transcripts appear in the tail somites, paraxial mesoderm and in the tailbud (Fig. 3E-E′).

This genetic analysis indicates that papc is a downstream target gene of the spadetail T-box transcription factor. In spt mutants papc is not expressed in gastrulating trunk mesodermal cells affected by the mutation. However, in regions of the embryo where spt mutants have more normal development, such as in tail somites (Kimmel et al., 1989) and islands of putative adaxial cells (Weinberg et al., 1996), papc is transcribed. The late expression of papc during tail somitogenesis reflects the existence of a second, spt independent, pathway in the regulation of mesodermal development (Griffin et al., 1998).

If the spt transcription factor is an upstream regulator of papc, one expectation is that both genes should be expressed in similar regions. To test this, a detailed comparison of the expression patterns of spt (Griffin et al., 1998) and papc was performed. During gastrulation stages, the parallels between spt and papc expression are striking. Both are expressed in involuting mesoderm and later repressed in dorsal midline cells (notochord progenitors) with similar time courses (compare in Fig. 4A,B and C,D). At the end of gastrulation the patterns start to differ partially, with papc mRNA extending more anteriorly than spt (Fig. 4E,F). This difference in anterior expression borders becomes more marked when somitogenesis begins, as spt mRNA does not form bands of expression in paraxial mesoderm, but papc mRNA does (compare Fig. 4G,H). During gastrulation stages, when spt is required for convergence movements, the parallels between spt and papc expression patterns are remarkable. These observations, taken together with the genetic analyses shown in Fig. 3, suggest that papc encodes a molecule acting closely downstream of spt.

floating head is a homeobox gene expressed in the dorsal midline; flh mutants lack a notochord, and the axial mesoderm adopts a paraxial mesodermal (muscle) fate (Talbot et al., 1995; Halpern et al., 1995; Melby et al., 1996). The loss-of-function of a second gene, spt, prevents the formation of muscle in the midline and, surprisingly, restores anterior notochord formation in flh;spt double mutant embryos (Amacher and Kimmel, 1998). This finding suggested that in some way flh antagonizes spt function in axial midline cells. One possibility, not previously tested, is that spt is turned off in the axial midline by flh function. We found that this is indeed the case by examining spt expression in flh mutants and showing that spt is expressed ectopically in the flh⁻ midline (Fig. 4A′,C′,E′,G′). Since papc is a downstream target of spt, one would expect that papc would also be turned on by the ectopic spt product in the flh⁻ midline, and this is indeed the case (Fig. 4B′,D′,F′,H′). These findings show that both spt and papc are repressed by the flh organizer-specific homeobox gene in axial mesodermal cells, and further strengthen the view that papc is a close downstream target of spt.

Xenopus papc mRNA promotes cell adhesion activity in injected embryos (S. H. Kim et al., unpublished observations). Similarly, injection of zebrafish full-length papc mRNA (FL-papc) into single Xenopus animal blastomeres was sufficient to create adhesive patches of clonally related cells at gastrula stages (data not shown). When FL-papc mRNA was microinjected into wild-type zebrafish embryos no obvious malformations were observed (Fig. 6D, Table 1). Attempts to

rescue the spt⁻ phenotype by injecting FL-papc mRNA into mutant embryos failed to demonstrate recovery of trunk somites (data not shown). To test whether papc functions in mesodermal morphogenesis in zebrafish, we constructed a truncated form of papc mRNA that encodes only three amino-terminal extracellular cadherin domains. Transfection into 293T human kidney cultured cells showed that this truncated construct secretes a stable protein product into the culture medium (Fig. 5, lane 8). Secreted forms of extracellular domains of cadherins have been described previously (Brieher et al., 1996) and a similar construct of Xenopus papc was shown to act in a dominant-negative fashion to counteract PAPC homophilic adhesion (S. H. Kim et al., unpublished observations).

Injection of zebrafish dominant-negative papc mRNA (DN-papc) resulted in embryos in which somites were expanded laterally or disrupted in the trunk (Fig. 6A-C). These phenotypes were specific for paraxial mesoderm, because the notochord (axial mesoderm) usually had normal width and morphology (Fig. 6A-C). To analyze the DN-papc phenotype in detail, embryos were injected into a single site at the 2-4 cell stage with synthetic mRNA and monitored at the 8-9 somite stage by in situ hybridization with myoD. The myoD probe is useful because it marks each somite as well as the adaxial cells that flank the notochord on either side (Weinberg et al., 1996). As can be seen in Fig. 6D-F and Table 1, two phenotypes were observed. In 32% of injected embryos, somitic myoD expression was spread out laterally, suggesting reduced convergence of somitic precursors towards the midline (Fig. 6E,H,I). The second phenotype, observed in 14% of injected embryos, was the reduction or lack of trunk somites, usually on one side of the embryo (Fig. 6C). This phenotype is reminiscent of that of spt⁻ embryos, in which cell convergence movements to the dorsal midline are severely impaired and somites fail to form in the trunk. However, in the case of somite disruption by DN-papc mRNA, a few isolated myoD-positive cells were usually detected in paraxial mesoderm at a distance from the midline (Fig. 6F). These phenotypes were observed at 12 hours of development; however, by 24 hours paraxial cells were able to converge to the midline, forming somites of relatively normal width (data not shown). A similar regulation or recovery of the trunk somite defects is seen in spt⁻ embryos (Kimmel et al., 1989).

To investigate the specificity of phenotypes caused by injected DN-papc mRNA, several controls were performed. Injection of large amounts of lacZ mRNA (500 pg) had no phenotypic effects (Fig. 6A and Table 1). When lacZ mRNA was injected together with 50 pg of DN-papc mRNA, the lineage tracer was always detected on the side in which the myoD expression pattern was affected (n=53; Fig. 6H,I), whereas myoD expression was normal on the injected side in embryos receiving lacZ mRNA only (Fig. 6G). The specificity of the dominant-negative phenotypes of DN-papc mRNA was further investigated in co-injection experiments with FL-papc mRNA. The results showed that the full-length product could rescue the observed phenotypes; as seen in Table 1 (bottom two lines), the frequency of both the wide somite and the somite-loss phenotype caused by DN-papc mRNA was reduced by co-injection of FL-papc mRNA. These microinjection experiments using a dominant-negative approach to inhibit papc function suggest that papc encodes a transmembrane cell adhesion molecule involved in convergence movements of paraxial mesoderm during gastrulation.

We have isolated a zebrafish cDNA, papc, encoding a structural transmembrane protein that is expressed at the right time and place to play a role in mesodermal morphogenesis during gastrulation. papc encodes a typical member of the protocadherin family of cell adhesion molecules (Suzuki, 1996) expressed transiently in paraxial mesodermal precursors undergoing morphogenesis.

The spadetail gene controls movements of cells in lateral mesoderm. In spt mutants lateral mesodermal cells fail to converge towards the dorsal midline and at later stages somites are missing (Kimmel et al., 1989; Ho and Kane, 1990). One characteristic of spt mutants is that although somites are absent in the trunk, they are present in the tail. The molecular nature of the spt gene has been recently identified by Griffin et al. (1998); it encodes a transcription factor of the T-box family, cloned independently by Ruvinsky et al. (1998). The spt gene is a zebrafish homologue of a Xenopus gene isolated independently by four groups and named Xombi, VegT, Antipodean and Brat (Lustig et al., 1996; Zhang and King, 1996; Stennard et al., 1996; Horb and Thomsen, 1997). In microinjection studies, Xombi mRNA can induce the formation of ectopic blastopore lips, indicating that Xombi may have a morphogenetic function (Lustig et al., 1996). Another T-box gene, Brachyury, has also been implicated in the movement of mesodermal cells as they pass through the primitive streak of the mouse gastrula (Wilson et al., 1993). Thus, transcription factors of the T-box family stand out as major regulators of mesodermal cell movements but it is not known how they cause such profound effects on cell movements during gastrulation. papc now provides a possible link between spt, a transcriptional regulator, and the cell surface.

In zebrafish, loss of spt function extinguishes papc expression in the mesodermal mantle throughout gastrulation and until early segmentation stages (except for a few putative adaxial cells, which are known to be less affected in spt mutant embryos; Weinberg et al., 1996). Therefore, papc lies genetically downstream of spt during early gastrulation. At later stages, the expression of these two genes diverges: papc is expressed in a banded pattern in forming somites whereas spt is not (compare Fig. 4G,H), and papc is not expressed in Rohon-Beard neurons, which are a site of expression of the Xenopus spt homologue (Zhang and King, 1996). Thus, at later stages additional factors regulate papc expression. Preliminary attempts to rescue the spt⁻ phenotype by injection of full-length papc mRNA did not restore the formation of trunk somites, scored by myoD staining at the 8-somite stage. In future it will be interesting to test whether cell convergence movements can be rescued in spt⁻ cells by injection of papc mRNA followed by cell transplantation into wild-type embryos (Ho and Kane, 1990).

During gastrulation the expression patterns of papc and spt in wild-type embryos are remarkably similar, suggesting that papc is activated closely downstream of spt. papc transcripts are markedly decreased in the trunk of spt mutants, but are present in the tail, a pattern that mirrors the regions affected in spt mutants. During late segmentation stages papc is expressed in the tailbud, and it has been proposed (Griffin et al., 1998) that a good candidate to take on spt function in the tail is tbx6, a T-box gene expressed in tail mesoderm in zebrafish (Hug et al., 1997) and mouse (Chapman and Papaionnou, 1998).

Studies in the Xenopus animal cap system have shown that a secreted dominant-negative form of xPAPC inhibits the convergence and extension movements induced by high doses of activin without affecting cell differentiation (as determined with the myoD, α-actin, Xbra, collagen 2 and Xwnt-8 molecular markers; S. H. Kim et al., unpublished observations). In addition, full-length xPAPC enhances convergence and extension movements and characteristic changes in cell shape in animal caps treated with low doses of activin. This cell movement-promoting activity does not affect the differentiation of tissue types (S. H. Kim et al., unpublished observations). Thus, Xenopus PAPC is not merely a cell adhesion molecule but is also able to trigger cell movements. During convergence and extension mesodermal cells associate closely with their neighbors, but cell attachments must be formed and subsequently released as intercalating cells move past each other in directional movements towards the dorsal midline (Keller et al., 1992). It has been proposed that Xenopus PAPC may promote cell convergence by mediating transient associations between neighboring cell surfaces (S. H. Kim et al., unpublished observations).

To test whether papc participates in convergence movements in zebrafish, we adopted a dominant-negative strategy using a secreted form of papc. A striking phenotype caused by microinjection of zebrafish DN-papc mRNA into the 2 or 4-cell embryo is the formation of laterally widened somites (Fig. 6B,E). A similar wide-somite phenotype has been found in zebrafish gastrulation mutants such as knypek and trilobite, which are proposed to be defective in convergence movements (Solnica-Krezel et al., 1996; Hammerschmidt et al., 1996). A second phenotype observed with DN-papc-injected embryos was the lack of somites and myoD expression (except for isolated myoD positive cells in paraxial mesoderm) on the injected side of the embryo (Fig. 6C,F). A lack of somites in the trunk region is seen in spt mutants (Kimmel et al., 1989; Ho and Kane, 1990). Both phenotypes may be explained by decreased cell convergence movements toward the dorsal midline. Together with the remarkable correspondence between the regions in which cell movements and papc expression are affected in spt⁻ embryos, the dominant-negative results suggest that papc may play an important role in mesodermal morphogenesis.

Protocadherins belong to a large gene family (Suzuki, 1996). In zebrafish we have isolated five additional protocadherins that are expressed in distinct patterns during early development (A. Y., unpublished observations). This raises the question of whether the dominant-negative effects are specific for papc. Although one can not eliminate cross-reaction with other protocadherins, three lines of evidence indicate that the effects may be specific for papc. First, the area affected by DN-papc correlates with the region of endogenous papc expression. Second, the dominant-negative effects can be rescued by full-length papc mRNA. Lastly, in Xenopus PAPC-mediated cell adhesion can be inhibited by DN-PAPC, but not by a dominant-negative form of another protocadherin expressed in Axial mesoderm (AXPC). Conversely, adhesion patches mediated by Xenopus AXPC can be disrupted by DN-AXPC but not by DN-PAPC, indicating that they affect distinct homotypic protocadherin interactions (S. H. Kim et al., unpublished observations). The definitive loss-of-function phenotype of papc must await the identification of a zebrafish mutant or the targeted mutation of its mouse homologue.

spadetail was the first zebrafish gene proposed to regulate cell movements in mesoderm (Kimmel et al., 1989). Since then, the genetic pathways that determine axial versus paraxial mesoderm and their epistatic relationships have been studied extensively (Halpern et al., 1993, 1995, 1997; Talbot et al., 1995; Odenthal et al., 1996; Amacher and Kimmel, 1998). Fig. 7 summarizes these genetic interactions. In flh mutants the notochord is absent (Talbot et al., 1995) and in gain-of-function studies Xnot recruits additional mesoderm into notochord fates at the expense of paraxial fates in Xenopus (Gont et al., 1996). We now show that spt and papc are ectopically expressed in the midline when flh function is lost (Fig. 4). Thus, an important function of the wild-type flh homeobox gene is to repress transcription of spt in the midline. In flh mutants, expression of spt in axial cells leads to the adoption of paraxial (myoD-positive) cell fates and consequently to the loss of notochord. When both flh and spt functions are removed in double mutants, ectopic muscle can no longer form in the midline and instead anterior notochord, including ntl-positive cells, can develop (Amacher and Kimmel, 1998). spt is required to activate myoD (Weinberg et al., 1996) and papc expression. As with any genetic model, the interactions indicated in Fig. 7 may be direct or indirect; in future it would be interesting to investigate whether the papc promoter contains T-box binding sites.

paraxial protocadherin introduces a new player in the pathway of mesodermal patterning. Many genes expressed specifically in the vertebrate gastrula, particularly in Spemann’s organizer, have been described in recent years (reviewed by Lemaire and Kodjabachian, 1996; Harland and Gerhart, 1997). Since these genes encode regulatory transcription or secreted factors, one must wonder at which point the building of an embryo through morphogenetic movements of coherent cell sheets begins. Unlike these regulatory transcription and growth factors, PAPC is a transmembrane protein that may be part of the actual structural machinery that effects cell movements during gastrulation. The PAPC cell adhesion molecule may provide a link between transcription factors and morphogenesis.

We are indebted to Drs Kevin Griffin and David Kimelman for generously sharing unpublished results and H. Okamoto for the cDNA library. We thank T. Bouwmeester, L. Leyns, R. Cornell, B. Appel and Y. Yan for advice, A. Ungar (University of Washington) for microinjection suggestions, K. Larison and R. Bremiller for technical assistance and the University of Oregon zebrafish facility for excellent fish care. A. Y. was supported by the Nagasaki University School of Dentistry, S. L. A. by an ACS fellowship (#PF-4087), and E. M. D. R. is an Investigator of the Howard Hughes Medical Institute. This work was supported by NIH grants HD22486 and HD21502.