dackel acts in the ectoderm of the zebrafish pectoral fin bud to maintain AER signaling

Tetrapod limb development relies on three signaling centers within the limb bud. They are established as a consequence of limb induction and confer developmental autonomy to the bud once it has been induced (reviewed in Duboule, 1994; Cohn and Tickle, 1996; Johnson and Tabin, 1997; Martin, 1998). At the boundary between dorsally and ventrally specified tissues, the apical ectodermal ridge (AER) is induced. It is the source of several fibroblast growth factors (Fgfs) and promotes proliferation in the underlying mesenchyme that constitutes the bud (reviewed in Martin, 1998). The AER becomes structurally distinct as the thickened ectodermal margin of the limb bud. Ablation of the AER in the chick causes an arrest in limb-bud outgrowth and results in a limb that is truncated along the proximodistal (P/D) axis (Saunders, 1948). In concert with mesenchymal factors from the lateral plate, the AER induces transcription of shh in the posterior limb-bud mesenchyme (Niswander et al., 1994; Laufer et al., 1994; Ros et al., 1996; Grieshammer et al., 1996). Shh mediates the functions of the zone of polarizing activity (ZPA; Riddle et al., 1993), a second signaling center of the bud (Saunders and Gasseling, 1968). It also maintains, polarizes and activates the expression of genes, among others, from the hoxa and hoxd clusters, and thus stabilizes and propagates anteroposterior (A/P) polarity in the bud (Riddle et al., 1993; Laufer et al., 1994; Neumann et al., 1999). In addition, Shh signals to the ectoderm to maintain Fgf-signaling from the AER (Martin, 1998; Zúñiga et al., 1999). In the zebrafish, Shh is necessary to activate fgf8 expression in the apical ectoderm (Neumann et al., 1999). A positive feed back loop between Shh and Fgfs coordinates growth and patterning along the A/P and P/D axes and is, at least in part, responsible for the developmental autonomy of the bud (Niswander et al., 1994; Laufer et al., 1994). Experimental or genetic ablation of either one of the two signaling centers causes the other one to stop signaling (Vogel and Tickle, 1993; Niswander et al., 1994; Chiang et al., 1996; Neumann et al., 1999). Classical experiments have suggested that a mesenchymal apical ectodermal maintenance factor (AEMF) acts to maintain signaling from the AER during limb development (Zwilling, 1961). Recent results indicate that AEMF activity consists of at least two independent signals. One is Shh, which maintains fgf expression in the AER (Zúñiga et al., 1999), the other is Fgf10, which is expressed in the distal limb bud mesenchyme, and activates and maintains fgf8 expression in the AER (Ohuchi et al., 1997).

The third signaling center resides in the non-ridge ectoderm, which imposes its dorsoventral (D/V) polarity onto the underlying mesenchyme (Poutou, 1977; Parr and McMahon, 1995). Wnt7a secreted from the dorsal ectoderm instructs the underlying mesenchymal cells to differentiate dorsal structures (Parr and McMahon, 1995), whereas En1 in the ventral ectoderm inhibits wnt7a expression and promotes development of ventral structures by default (Loomis et al., 1996).

The zebrafish (Danio rerio) pectoral fin bud is homologous to the tetrapod forelimb bud (Grandel and Schulte-Merker, 1998). In 26 h (hours postfertilization) embryos, a small pectoral fin bud has formed. The epidermis covering this bud thickens and forms the apical fold by 36 h in a position that is equivalent to the one taken by the tetrapod AER. After 48 h, the apical fold expands and is invaded by mesenchyme, thereby giving rise to the fin fold that will develop into the visible fin of the adult fish. Tetrapods do not develop any structure homologous to the fin fold. The mesenchyme of the bud gives rise to two endoskeletal elements, a proximal shoulder girdle and a distal endoskeletal disc from which the fin fold extends distally. Thus, zebrafish fin buds become structurally distinct from tetrapod limb buds only after 48 h of development when fin fold formation occurs.

During early pectoral fin development, the expression patterns of the orthologs of some of the genes described above show striking similarities to the tetrapod condition. shh (Krauss et al, 1993), fgf8 (Fürthauer et al., 1997; Reifers et al, 1998) and en1 (Ekker et al., 1992) are each active in the expected tissues of the bud, and thus probably exert an equivalent function as in the chicken or mouse. fgf8, however, which is the earliest AER-marker in tetrapods, becomes detectable only in the apical epidermis long after the fin bud has been established in zebrafish (Reifers et al., 1998). The analysis of the zebrafish syu mutant (Neumann et al., 1999), in which the zebrafish shh gene is mutated (Schauerte et al., 1998), has already shown important homologies to the tetrapod condition. As in the tetrapod limb bud, Syu mediates most aspects of A/P polarity and also is required for progressive development in the apical epidermis. As in the tetrapod limb bud, members of the hoxa and hoxd clusters are active in zebrafish fin buds (Sordino et al., 1995), and have been shown to be regulated by Syu signaling (Neumann et al., 1999).

Ten other mutants with a specific phenotypic alteration of pectoral fin development have been described (van Eeden et al., 1996). In this collection, the dackel (dak) mutant is unique for a specific proximodistal truncation of the fin and an early downregulation of syu expression (van Eeden et al., 1996).

In the present study, we investigate the fin phenotype of the dak mutant by combining mutant analysis and experimental embryology. We show that dak acts in the epidermis to initiate apical fold formation, to induce expression of fgf4 and fgf8, and to maintain expression of en1 in the ventral non-ridge ectoderm.

Fish maintenance was as described in Grandel and Schulte-Merker (1998). The dak alleles to273b and tw25e, the hypomorphic syu allele tq252, and the syu null allele t4 were used.

Whole-mount in situ hybridization was carried out and the same antisense probes were used as described in Neumann et al. (1999), with the exception of the fgf8 probe, which was from Fürthauer (1997). P. Sordino (1995) provided hoxd and hoxab probes. The zebrafish fgf4 gene was cloned by B. W. D. (Draper et al., 1999). A detailed characterization will be reported elsewhere. The amino acid sequence has been submitted to GenBank (accession number AF283555).

Fgf4 beads were prepared according to the method of Reifers et al. (2000). Zebrafish embryos were manually dechorionated and placed into a drop of Low Melting Point Agarose 1.5% (Gibco BRL), which was allowed to dry and subsequently fenestrated to manipulate the fin bud. The epidermis was digested away with a drop of Light White Mineral Oil (Sigma) at the ventral edge of the fifth somite. A bead was inserted into the lateral plate mesenchyme and pushed anteriorly below the epidermis into the fin bud mesenchyme with electrophoretically sharpened tungsten needles. Manipulation and further incubation was carried out in Ringer’s plus 20 μg/ml Gentamycin (Gibco BRL), which was replaced by E3 after 12 h.

Donor embryos were injected with 0.625% rhodamin-dextran and 0.625% biotin-dextran (Molecular Probes, Eugene, OR) at the one-cell stage. Transplantations were performed at late blastula stages between sphere stage to 30% epiboly, with stage matched donors and hosts – the manipulation is particularly easy at this stage and we were unable to perform transplantations when fin buds had formed. The choice of site for transplantation was guided by the fate map of Kimmel et al. (1990), which shows that ectodermal cell fates are located between 0% to 80% latitude at the 50% epiboly stage on the ventral side of the embryo. Because the dorsal and ventral sides cannot be distinguished during blastula stages, we implanted donor cells into two opposing areas roughly 30% latitude away from the animal pole to ensure that the implanted cells from one of the two locations would give rise to epidermis. In accordance with the map, we observed mainly epidermal and neuroectodermal clones. The frequency with which the transplanted cells populated the fin bud ectoderm in greater numbers was 6.5%. The manipulations were carried out in Ringer’s plus gentamycin. Embryos were allowed to develop until 60 h, fixed and prepared for staining. Biotin-labeled donor cells were detected with a rabbit anti-biotin primary antibody (kindly provided by Heinz Schwarz, MPI für Entwicklungsbiologie, Tübingen) and fluorescently labeled with a secondary goat anti-rabbit-Cy3 coupled antibody (Jackson).

Among the 11 genes that have been identified by mutant phenotype to be involved in zebrafish pectoral fin development, recessive alleles of dackel (dak) lead to the loss of functional pectoral fins in homozygous mutant larvae (van Eeden et al., 1996; Fig. 1A,B). Four morphologically indistinguishable alleles of dak have been found (van Eeden et al., 1996). The two alleles chosen in the present study also behave alike with regard to molecular markers. Despite the dramatic loss of the larval fin, pectoral fin buds form normally in dak mutants. During the first 10 hours of fin development (26-36 h), it is not possible to identify living dak embryos by phenotype. Only after formation of the apical fold in sibling embryos, can dak mutants be identified by the lack of this structure (Fig. 1C,D) as well as by the smaller size of the fin buds, which have stopped growing after 32 h. The vast majority of dak mutant embryos (94%; 185 of 196 buds from 11 independent egglays) fail to form an apical fold, even at later stages. dak homozygotes also do not form a fin fold. In those cases where a small apical fold can be observed at 60 h (6%, 11 of 196), the underlying mesenchyme has not invaded the fold. Skeletal preparations of hatched dak larvae show that not all pectoral fin structures are lost in the mutant (Fig. 1F). dak mutants retain the pectoral girdle, the proximalmost pattern element of the fin. The girdle itself appears mildly affected as well, as the postcoracoid process is shorter than in siblings and is sometimes absent. All dak alleles are larval lethals (van Eeden et al., 1996).

As the dak mutant fin phenotype is suggestive of a defect caused in apical ridge signaling, we tested the expression of four apical ectodermal markers. At 28 h, dlx2 and bmp2b were both detected in dak mutants (Fig. 2A,B,E). But while bmp2b was expressed at wild-type levels, the expression of dlx2 was already reduced. At 38 h, after the apical fold has formed in sibling embryos, the activity of both markers was downregulated severely in dak⁻ fin buds, becoming undetectable at even later stages in the case of dlx2 (Fig. 2C,D,F,G and data not shown)

As the apical fold forms, expression of fgf8 starts in the apical ectoderm at 36 h (Reifers et al., 1998). This contrasts with tetrapods where fgf8 is the earliest AER marker. We report expression of fgf4 in the apical fold at the same time (Fig. 2H). The onset of fgf4 expression correlates well with the situation in tetrapods, where fgf4 marks the apical ridge at later developmental stages (Niswander and Martin, 1992; Laufer et al., 1994). However, we could not detect a posterior bias of the fgf4 expression domain typical of tetrapods. Another difference concerns its later expression in the myogenic mesenchyme of the fin at 48 h (Fig. 2J), which has not been reported in tetrapods. (A manuscript on fgf4 function and expression is in preparation by B. W. D.).

At 38 h, fgf8 was not detectable in most dak⁻ embryos (42 of 47), or was present at very weak levels only (5 of 47; Fig. 2K,L). At the same stage, fgf4 remained inactivated in a similar manner (41 of 45 showed no detectable expression, Fig. 2H,I). At 48 h and at 60 h, fgf4 could not be detected at all in 16 and 62 cases, respectively, and fgf8 was seen in minor amounts in only 8 of 42 and 2 of 18 cases, respectively (data not shown). The lack of an apical fold in dak mutant fin buds is a structural difference to syu null mutant fin buds, which are able to form an apical fold with some delay. However, syu null mutant buds, like dak⁻, do not express either fgf (Neumann et al., 1999; Fig. 2M,N).

The analysis of epidermal gene expression in dak⁻ fin buds clearly shows a requirement for a functional dak gene product already at the earliest developmental stage tested. This contrasts with the wild-type appearance of living dak⁻ fin buds at the same stage. Clearly, at later stages dak is indispensable for normal growth, epidermal morphogenesis and for initiation of fgf4 and fgf8 expression, two genes proposed to play key roles in mediating AER function in tetrapods.

Transcription of the shh gene in the posterior mesenchyme requires a functional AER signal in the tetrapod limb bud (Niswander et al., 1994; Laufer et al., 1994; Ros et al., 1996; Grieshammer et al., 1996). We thus probed dak mutants for the presence of syu transcript in the fin buds in order to test for the presence of ridge activity. van Eeden et al. (1996) described the early reduction and later loss of syu transcript in the pectoral fin buds of dak mutants. We have confirmed these results (Fig. 3A-F). At 28 h, only 2 hours after the first appearance of the bud, dak mutant buds showed reduced levels of syu expression. The reduction was mild at first, but enhanced at later stages, with syu mRNA levels finally falling below the level of detection. This behavior points to the early presence of apical ridge activity in dak⁻ fin buds as the induction of syu expression does occur.

As Syu positively feeds back on its own expression (Neumann et al., 1999), we assayed the activity of the pathway downstream of syu in dak mutants, and analyzed the expression of patched1 (ptc1), a target gene of Syu signaling (Lewis et al., 1999; Fig. 3G-L). While ptc1 expression was not detected in syu null mutants, its expression was detected at 28 h and 38 h in dak mutants, although at reduced levels. At 48 h, ptc1 expression had been lost along with syu transcript. hoxa13 and hoxd13 are likewise targets of Syu signaling in the fin bud, as they are not activated in syu null mutant buds (Neumann et al., 1999). As in the case of ptc1, their expression was initiated in dak mutant buds but lost at 38 h (see Figs 5R-U, 6R-U). These results suggest that the pathway downstream of syu is functional and that dak directly or indirectly maintains syu expression. There are three additional observations indicating that downregulation of the Syu signaling pathway in the fin bud is insufficient to cause the dak mutant phenotype.

When we tested dak mutants for en1 expression (Fig. 4A-C), we found wild-type levels of expression at 28 h, but a reduction of the expression level and the size of the expression domain at 38 h and later. As en1 expression was not reduced in syu null mutant fin buds at 38 h and 48 h (Neumann et al., 1999; Fig. 4D,E), this aspect of the phenotype cannot be explained by reduction of syu signaling.

Structural comparisons were made of the endoskeletal elements of the fins in dak and syu mutants. In larvae carrying the syu null allele, no endoskeleton formed at all, while in carriers of the hypomorphic syu allele, both endoskeletal elements, girdle and disc were formed but were variably reduced (Neumann et al., 1999; Fig. 1G,H). As dak mutants do not phenocopy either syu mutant, the phenotypic difference indicates the involvement of the two genes in different processes. Another observation to indicate that dak does not act solely via the syu pathway comes from the analysis of hox expression patterns.

At 28 h, wild-type levels of expression of the 3′ located member of the hoxd cluster, hoxd10, could be detected in dak mutants (Fig. 5A). At the same time, reduced expression of hoxd11 and hoxd12 was seen in dak mutants (Fig. 5F,G,L,M). hoxd13 was not stably expressed at this stage in wild-type fin buds, but a strong reduction of its expression was seen at 32 h in dak⁻ fin buds (Fig. 5R,S). At later stages, the expression of all four hoxd genes was further downregulated, having dropped below detection level by 38 h (hoxd13; Fig. 5T,U) and 48 h (hoxd10, hoxd11 and hoxd12; Fig. 5D,E,J,K,P,Q). Though the late loss of the three more 5′ located hoxd genes can be explained by the loss of syu signaling in dak mutant fin buds, downregulation of hoxd10 at 38 h is not seen in syu null mutants at the same stage (Neumann et al., 1999).

Reduced levels of expression of hoxa9, hoxa10 and hoxa13 were detected at 28 h in dak⁻ embryos, while transcript levels of hoxa11 were not affected at this time (Fig. 6A,B,G,H,M,R,S). During further development, mRNA of hoxa9 and hoxa10 dropped below the detection level, while hoxa11 and hoxa13 could show very low levels of expression in some dak mutant buds at 48 h (Fig. 6E,F,K,L,P,Q,V,W). We further noticed the loss of hoxa10 expression in the posterior and distal mesenchyme at 38 h, which leaves the bud with an apolar patch of hoxa10 at that time, reminiscent of the situation found in the syu null mutant fin buds (Fig. 6I,J). But again, the early reduction of the 3′ member in the cluster, hoxa9, distinguishes dak mutant fin buds from those of the syu null mutant.

We have presented several differences in the behavior of dak and syu mutants that all suggest that dak is involved in other processes besides maintaining syu signaling. In view of the fact that the apical epidermis is more strongly affected in dak mutants than in syu mutants, our results suggest that the primary defect is a loss of apical ridge function in dak mutants.

Beads soaked in either Fgf4 or Fgf2 are equally potent in replacing an ablated AER in the chick embryo when stapled onto or transplanted into the ridge denuded bud (Niswander et al., 1993; Fallon et al., 1994). We have implanted Fgf4-soaked beads into dak mutant buds to investigate whether dak interferes with the bud’s reception of the Fgf signal from the apical fold.

The implantation was carried out within a time window between 35 h and 37 h, which surrounds the timepoint of endogenous fgf4 and fgf8 activation in wild-type fin buds. Of the 62 mutant fin buds that received an Fgf4 bead, all showed further growth than usually seen. 24 hours after the operation the treated buds were about two to three times the size of the untreated buds on the contralateral side (Fig. 7). We found expression of syu, hoxd11 and hoxa13 to be reactivated in the operated buds, while no expression or only slight expression was seen on the contralateral side (Fig. 7A-F). As indicated in Table 1, detection of slight expression in the control side buds of some of the marker genes tested was expected from in situ analyses of untreated dak mutants at the same stage. All three mesenchymal markers showed the posterior bias characteristic of syu, hoxd11 and of the early phase of expression of hoxa-13, irrespective of the position of the bead. In a total of 42 cases, PBS-soaked beads were implanted into dak mutant fin buds as controls. PBS-beads had no effect on bud size nor did they lead to activation of marker gene expression above the usual level at 60 h (Fig. 7L,M).

We also examined expression of ectodermal markers. The early apical ectodermal markers bmp2b and dlx2 were activated. They showed a posterior bias in 6 of 7 cases and in 7 of 9 cases, respectively (Fig. 7G,H; Table 1, and data not shown). The late apical ectodermal markers fgf4 and fgf8, however, could not be detected in Fgf4-treated fin buds (Fig. 7I,J; Table 1, and data not shown). Apical fold formation was not readily detected in vivo. We therefore sectioned six of the bead-treated fin buds and found two individuals with a small apical fold, while the other four did not show any sign of apical fold formation. Somewhat surprisingly, the expression of en1 was upregulated again in its normal domain upon the Fgf signal (Fig. 7K).

As the fin bud mesenchyme is able to respond to the Fgf4 signal by activating syu, hoxd11 and hoxa13, we conclude that dak does not act downstream of the Fgf signal. The dak ectoderm, however, is only partially able to respond to Fgf4 as it activates only early epidermal markers but neither fgf expression nor significant apical fold formation are observed. dak thus behaves as if it were required to promote progression of apical ectodermal development.

Recombination of ectoderm and mesenchyme of mutant and wild-type donor and host limb buds has been carried out successfully in the chick limb bud in order to better define the capacities of each tissue in mutants (Ros et al., 1996). We have carried out a similar experiment by injecting lineage-labeled wild-type cells into dak mutant embryos during early epiboly stages and screened for their presence in the fin buds at 60 h. 46 dak mutant embryos survived the procedure, which allowed for inspection of 92 fin buds. 16 buds showed wild-type cells in the fin bud ectoderm. Among these, 10 buds had received only one or a few wild-type cells and were phenotypically dak⁻. Six buds showed larger clusters of transplanted cells in the epidermis. These buds were larger than unmanipulated dak⁻ buds and all had developed an apical fold distally.

We examined the distribution of wild-type cells in the six fin buds, which had received larger contributions of wild-type cells (Table 2). In all cases, wild-type cells could be observed in the apical fold (Fig. 8). All buds had also received different amounts of cells in the ventral epidermis (Fig. 8). In all but one case, wild-type cells were located in the dorsal epidermis as well.

We also have carried out in situ hybridizations in the experimental fin buds and found, among the six larger buds, fgf8 expression in the apical fold in three cases (Fig. 8) and syu expression in the other three cases (Table 2; data not shown). Thus, all mutants that had received wild-type cells in the epidermis and were phenotypically scored to show a limited ‘rescue’ of the dak mutant phenotype showed activation of the marker gene tested. All other dak⁻ embryos had neither activated fgf8 nor syu.

We conclude from these observations that a functional dak gene product is required in the epidermis of the fin bud for induction of fgf8 and for apical fold morphogenesis. The nonautonomous activation of syu in the posterior mesenchyme provides evidence for the presence of ectodermal signaling to the mesenchyme during fin development.

The formation of a boundary between dorsally and ventrally specified territories is a prerequisite for AER induction in the limb bud (Laufer et al., 1997; Rodriguez-Esteban et al., 1997). The early wild-type expression of en1 in the ventral ectoderm of dak mutant buds indicates that the initial specification of cell fates along the D/V axis occurs normally in the mutant. The presence of bmp2b and dlx2 in the apical epidermis shows that ridge induction has occurred in dak mutant fin buds at 28 h, though ridge activity is reduced already at this early stage. This is indicated by reduced levels of dlx2 expression and by lower amounts of syu transcript. Despite the reduced signal from the apical epidermis, living dak mutants cannot be distinguished from wild-type siblings by their phenotype prior to 36 h – up to 10 hours after bud formation. This indicates that during the first 10 hours of fin development, the fin bud behaves normally with respect to growth and shape. Likewise, the expression of mesenchymally expressed genes can be detected in their wild-type domains at 28 h, though many of them show lower levels of expression. In the case of ptc1, hoxa10, hoxa12, hoxa13, hoxd11, hoxd12 and hoxd13 the observed reduction of expression can be understood in terms of their dependence on Syu and thus as an indirect consequence of the reduced ridge signal. hoxa9 and hoxd10, however, show a more direct dependence on dak/AER function. At 38 h, expression of dlx2 and bmp2b have been strongly downregulated in dak mutant fin buds and induction of fgf4 and fgf8 expression, as well as apical fold formation, do not occur properly. The failure of progressive development in the apical epidermis and the successive loss of syu expression leads us to conclude that the early weak AER signal has been lost by 38 h in dak mutant fin buds. The early presence and later loss of the AER signal also is reflected by the later truncation of the fin along the proximodistal axis. This behavior points to a qualitative change between 28 h and 38 h in the mechanisms controlling the function of the apical epidermis and promoting its developmental progression into the apical fold phase: although the dak gene is necessary at both stages for normal development, it is not until later that its product becomes indispensable.

As a feedback loop interconnects Fgf signaling from the AER and Shh signaling from the ZPA in the tetrapod limb bud (Niswander et al., 1994; Laufer et al., 1994), it could be possible that the cause of the observed loss of AER signal in dak mutant buds lies in the involvement of the dak gene in this feedback loop. However, our analyses of gene expression, the phenotypic comparison of dak mutants with the syu mutants and the Fgf-bead implantation experiment suggest that this is not the case.

We have shown by in situ hybridization that syu signal transduction is functional in dak mutants as the mesenchymal targets of syu signaling ptc1, hoxa13 and hoxd13 are induced in dak mutant fin buds. The phenotypic difference of the fin endoskeletons in dak and syu mutants likewise suggests the involvement of dak and syu in different genetic pathways. By implantation of Fgf4-soaked beads, we further showed that the reception of the Fgf-signal from the apical fold by the dak⁻ fin bud mesenchyme was not hampered as syu, hoxa13, and hoxd11 are induced. In addition to these results that suggest functional syu and fgf signal transduction cascades in dak mutants, we have made observations, which are not in agreement with a possible function of dak in the Syu/Fgf feedback loop. The expression patterns of hoxd10 and en1 in dak mutant fin buds differ from those observed in syu null mutant fin buds in which the Syu/Fgf feedback loop is abolished. Without Syu signal, hoxd10 expression is equivalent to wild-type levels at all stages investigated. In dak mutant fin buds, however, hoxd10 is downregulated at late stages, which again implies that dak and syu exert their effects through different pathways. A similar observation is made for en1 expression, which is unaffected at early and late stages in syu null mutant embryos but is downregulated in dak mutant buds at 38 h and later. If dak were to act solely as a component of the Syu/Fgf feedback loop, downregulation of hoxd10 and en1 should not be observed in dak mutant fin buds.

bmp2b and dlx2 become targets of Syu signaling at later stages of fin development (Neumann et al., 1999). Upon implantation of an Fgf4 bead into dak mutant fin buds, late expression of both markers is detected at high levels with a posterior bias. This suggests that the dak apical ectoderm is competent to respond to the prolonged Syu signal by inducing the early AER markers. Nevertheless, progressive ectodermal development is still hampered. The late apical fold markers, fgf4 and fgf8, which are also targets of Syu signaling (Fig. 2M,N; Neumann et al., 1999) are not activated upon Fgf-4 bead implantation, despite the Syu signal in the experimental buds. Likewise, apical folding which is independent of syu signaling does not take place significantly in these buds.

The differences in the activation of early and late developmental events in the apical epidermis upon Fgf4 bead implantation in experimental dak mutant fin buds are probably due to a specific function of dak that is needed at later stages for the progression of development in the apical fold.

hoxd10 is downregulated in dak mutant fin buds at 38 h and hoxa9 is already expressed at lower levels at 28 h. In contrast, the syu null mutant does not influence the expression of hoxd10 in the fin bud at 32 h and 38 h, nor of hoxa9 at 32 h. This specific difference indicates the presence of a dak-dependent maintenance signal for these two 3′ located hox genes that is independent of Syu. In the chick limb bud, an AER signal is needed in concert with Shh to initiate ectopic hoxd expression in the anterior limb bud mesenchyme, thus indicating an influence of the ridge on hox expression (Laufer et al., 1994). As will be discussed below, dak acts in the epidermis to maintain AER functions. It therefore seems likely that the downregulation of expression of both 3′ hox genes reflects the dependence of their expression on the AER.

The early wild-type expression of en1 and the normal positioning of the early apical ectodermal markers indicate the correct initial specification of cell fates along the D/V axis in dak mutants. At 38 h and later, however, en1 has disappeared from the ventral ectoderm except for a small distal patch. Additionally, the ridge markers are downregulated and syu, the mesenchymal ridge-target, is lost. These findings prompted the question whether the loss of en1 reflects the cause for the loss of ridge activity or whether a ridge signal is needed for sustained en1 expression. There is no evidence from chicken or mouse for either of the two possibilities. The rescue of en1 expression by implantation of an Fgf4 bead shows that a Fgf signal is sufficient to maintain levels of En1 in the ventral ectoderm. dak thus reveals the later dependence of en1 on a maintenance signal for ventral ectodermal identity. The fact that Fgf4 can provide the maintenance signal strongly suggests, that the ridge provides the signal to the ventral ectoderm to maintain en1 expression. The loss of en1 expression in dak mutants is thus due to the loss of ridge function.

This finding uncovers a new interaction among two of the three signaling centers in the bud, namely the dependence of the D/V organizer on the AER. The mouse en1 mutant clearly shows that En1 is required for restricting the expression of the dorsalizing signal Wnt7a to the dorsal ectoderm (Loomis et al., 1996). We therefore conclude that in the zebrafish, the AER maintains the proper dorsoventral organization of the ectoderm and thus its proper patterning activity along the D/V axis via dak and fgf.

Our results extend other findings that indicate the signaling centers mutually interact to establish pattern coordinately along the three axes of the limb (Laufer et al., 1994; Niswander et al., 1994; Parr and McMahon, 1995; Yang and Niswander, 1995). Yet we would also like to point out that the functions of the zebrafish apical epidermis are probably maintained by different genetic pathways. The apical ectoderm of syu null mutant fin buds retains the capacity to form an apical fold and en1 expression is maintained in the ventral ectoderm (Neumann et al., 1999). Together with our data from the present analysis of the dak mutant fin bud, these findings suggest that the syu signal from the ZPA maintains only some aspects of AER integrity and signaling but not others. If our proposal is correct, that it is an Fgf signal from the AER that maintains en1 expression in vivo, this might indicate that not all Fgfs are under the control of Syu in the zebrafish fin bud.

Upon transplantation of wild-type cells into the epidermis of dak mutant fin buds, fgf8 expression and apical fold formation were observed. A wild-type copy of dak is thus required in the fin bud ectoderm to promote developmental progression of the apical epidermis. As wild-type cells were found in the apical fold, but also in the ventral non-fold ectoderm and often also in the dorsal non-fold ectoderm, we cannot unambiguously localize the site where the dak gene is required. Importantly, syu was transcribed in the mesenchyme of dak fin buds, which had received wild-type cells in the ectoderm. As syu expression is not observed in untreated dak fin buds at this stage, this result confers experimental proof of epidermal-mesenchymal signaling in the fin bud. The analysis of the syu null mutant (Neumann et al., 1999) has shown that the Syu signal from the ZPA is necessary to maintain strong expression of dlx2 and bmp2b, and to induce expression of fgf4 and fgf8 in the apical ectoderm, thus demonstrating mesenchymal-epidermal signaling in the fin bud. It is thus likely that there exists, as in tetrapods, a positive feedback loop between epidermal signals and Syu.

Our results provide evidence that in the zebrafish fin bud, as in the tetrapod limb bud, different mechanisms regulate the early establishment of AER activity and its later maintenance during development. dak functions in the ectoderm to activate fgf4 and fgf8 synergistically with Syu. In addition it is required, independently of Syu, to induce apical fold formation, as well as to maintain en1 expression in the ventral ectoderm. Although we cannot exclude a permissive function of the dak gene in these processes, we favor a hypothesis in which dak acts to transduce the signal of a Syu independent apical ectodermal maintenance factor. This interpretation is suggested by the fact that the AER relies on such a signal as it cannot maintain its activity autonomously (Zwilling, 1961).

We propose that dak⁻ fin buds arrest development because the fin bud ectoderm fails to transduce an apical ectodermal maintenance signal. This signal is independent of syu signaling, as different aspects of the dak mutant phenotype are not in agreement with a function of dak downstream of Syu. Owing to the lack of AER maintenance, the feedback loop between the ZPA and the apical ectoderm is not productively established, which causes both organizers to lose activity over time. The nonautonomous effect of ectodermal wild-type cells to maintain syu expression in dak mutant mesenchyme is a clear indication that the AER activity has been restored in these buds. An attractive hypothesis for the function of dak is that it acts in the apical epidermis in a cell-autonomous manner to activate late apical ectodermal gene expression and to promote apical fold formation upon reception of an AEMF signal.

H. G. thanks Carl Neumann for fruitful discussions while the work presented here was in progress and for critically reading of an earlier version of this manuscript. We also thank Alexander Picker, Frank Reifers and Michael Brand for making the bead-implantation protocol available to us prior to publication. Bernard Thisse kindly provided the fgf8 probe. B. W. D. was supported by the Cancer Research Fund of the Damon Runyon-Walter Winchell Foundation Fellowship, DRG-1420. H. G. thanks Christiane Nüsslein-Volhard for support. This work was supported by a grant from the DFG (Schwerpunktsprogramm: Evolution entwicklungsbiologischer Prozesse, SCHU 1228).