Wnt/β-catenin regulation of the Sp1-related transcription factor sp5l promotes tail development in zebrafish

Structures of the posterior vertebrate body are derived from the tailbud,an undifferentiated group of cells at the caudal end of the embryo. The tailbud has qualities of a stem cell population, in that differentiated cell types that contribute to posterior outgrowth are continuously generated, while a population of multipotent precursors is maintained to contribute to further caudal development (Cambray and Wilson,2002; Charrier et al.,1999; Gont et al.,1993). Within the tailbud, precise coordination of cell fate specification, proliferation and morphogenetic movements leads to caudal outgrowth of the embryo. A multitude of signaling pathways have been implicated in the regulation of these processes, among them the BMP, Nodal,FGF and Wnt pathways (Agathon et al.,2003; Griffin and Kimelman,2003; Mathieu et al.,2004; Vasiliauskas and Stern,2001).

Much of our current understanding of the role for Wnt/β-catenin signaling in posterior mesoderm formation in vertebrates comes from analysis of mouse mutants. For example, mice homozygous for a null allele of Wnt3a fail to form somites and notochord caudal to the forelimb bud,instead making ectopic neural tissue(Greco et al., 1996; Takada et al., 1994; Yoshikawa et al., 1997). These embryos form only the anterior-most seven to nine somites and completely lack a tailbud (Takada et al.,1994). Mice doubly homozygous for mutations in two downstream effectors of the Wnt pathway, Lef1 and Tcf1, or homozygous for a null allele of the Wnt co-receptor, LRP6, show a similar phenotype(Galceran et al., 1999; Pinson et al., 2000).

Studies in zebrafish have also implicated Wnt/β-catenin signaling at an early step in posterior mesoderm formation. Transplantation studies show that the ventral marginal region of the zebrafish gastrula functions as a tail organizer, able to induce the formation of an ectopic tail when transplanted to the animal pole (Agathon et al.,2003). In the same study, overexpression experiments indicated that a combination of high levels of Wnt/β-catenin, Nodal and BMP signaling can induce the formation of ectopic tail organizers(Agathon et al., 2003).

wnt8 is highly expressed at the ventral margin and has been shown to be required for specification of ventrolateral mesoderm, which contributes to formation of the tailbud at the end of gastrulation(Erter et al., 2001; Lekven et al., 2001). Strong loss of wnt8 function, either by mutation or by antisense morpholinos(MOs), results in tail formation defects(Agathon et al., 2003; Lekven et al., 2001). These defects are presaged during gastrulation by a dramatic loss of expression of ventrolateral mesoderm markers such as the T-box transcription factor tbx6, which is strongly expressed in the ventral margin and subsequently in the tailbud (Griffin et al., 1998; Hug et al.,1997), suggesting that wnt8 acts at a very early step of tail formation.

While Wnt3a in the mouse and wnt8 in the zebrafish clearly act at early steps in mesoderm induction and/or patterning that subsequently affect tail development, whether Wnts also act later during tail development is less clear. In zebrafish, the continued expression in the tailbud throughout somitogenesis of both wnt8 and another Wnt gene, wnt3a, as well as of a transgenic β-catenin-responsive reporter,TOPdGFP, suggests a role for continuous Wnt activity throughout tail formation(Buckles et al., 2004; Kelly et al., 1995; Lekven et al., 2001).

Relatively little is known about the identity of downstream effectors of Wnt/β-catenin signaling in tail development in any species. One direct transcriptional target of Wnt3a in the mouse is the T(Brachyury) gene, a T-box transcription factor that is also required for posterior development (Galceran et al., 2001; Yamaguchi et al.,1999). In zebrafish, the homolog of T, no tail, is required for tail formation (Halpern et al., 1993), but has not been shown to be a Wnt target. Conversely,while tbx6 is directly regulated by Wnt signaling in zebrafish(Szeto and Kimelman, 2004), it has not been demonstrated to have a functional role in tail development.

We demonstrate that Wnt/β-catenin signaling, activated by wnt3a and wnt8, is required not only during gastrulation for specification of the tail organizer, but also during early somitogenesis for the maintenance of expression of presomitic mesoderm markers within the tailbud. We then show that the Sp1-related zinc finger transcription factor sp5-like (sp5l), is expressed in response to wnt3aand wnt8 in the tailbud and acts downstream of these Wnts to regulate tail formation. Thus, Wnt signaling is required at multiple stages of tail development, acting at least in part by activating expression of sp5l.

Zebrafish were raised and maintained under standard conditions. Our wild-type line is derived from AB. In addition, we used the previously published transgenic TOPdGFP line (Dorsky et al., 2002), and the hsΔTCF GFP line(Lewis et al., 2004).

In situ hybridizations using digoxigenin-labeled mRNA probes were performed using standard methods (Oxtoby and Jowett,1993). For assessing the mitotic index in the tailbud, we first processed the embryos for tbx6 expression using a Fast Red color reaction, which produces a fluorescent product. Subsequently, embryos were incubated overnight in a 1:500 dilution of polyclonal rabbit anti-phosphohistone H3 antibody (Upstate, Charlottesville, VA, USA) in PBT+10%calf serum, washed 5 × in PBT, and incubated overnight in a 1:1000 dilution of AlexaFluor 488 anti-rabbit secondary antibody (Molecular Probes,Eugene, OR USA). Mitotic cells within the tbx6 domain were counted and divided by the total number of tbx6-expressing cells.

wnt3a, wnt8 ORF1, wnt8 ORF2 and sp5l/spr2morpholinos (Gene Tools, Philomath, OR, USA) have been previously described(Buckles et al., 2004; Lekven et al., 2001; Zhao et al., 2003). For sense RNA injections, mRNA was synthesized using the mMessage mMachine kit (Ambion). Both mRNA and morpholinos were diluted in Danieau's buffer prior to injection. wnt3a MO was injected at a concentration of 2 mg/ml, except as noted in some co-injection experiments with sp5l MO, where the concentration was 1 mg/ml. wnt8 ORF1 and ORF2 MOs were injected at a concentration of 0.5 mg/ml each, and sp5l MO was injected at a concentration of 2.5 mg/ml. In all experiments, a volume of 3-5 nl was injected into the yolk of one-cell stage embryos.

As part of an earlier analysis of the role of zebrafish wnt3a in neural patterning, we noted a mild shortening of the tail when wnt3afunction was reduced using either of two antisense morpholino oligonucleotides(Buckles et al., 2004),suggesting that wnt3a might function in tail development. To determine if inhibition of wnt3a reduced β-catenin-mediated transcriptional activity in the developing tail, we studied the effects of knocking down wnt3a on expression of a transgenicβ-catenin-responsive reporter, TOPdGFP(Dorsky et al., 2002). For these and all subsequent experiments, we used a splice-blocking morpholino previously shown to completely block splicing of the first intron of wnt3a (Buckles et al.,2004). TOPdGFP is normally robustly expressed in the tailbud at the end of gastrulation (bud stage) and during somitogenesis(Dorsky et al., 2002). Expression is only modestly reduced when wnt3a is knocked down(Fig. 1D,E), and wnt3amorphants are essentially indistinguishable from wild-type at 24 hours post-fertilization (hpf) (Fig. 1F), although they are slightly shorter by 48 hpf (data not shown).

This result suggested that additional Wnt(s) expressed in the tailbud could be acting redundantly with wnt3a in tail formation. We therefore tested whether the mild phenotype observed in wnt3a morphants was due to functional redundancy with wnt8, which is also expressed in the developing tail (Kelly et al.,1995; Lekven et al.,2001). To specifically address a potential later role for wnt8 in tail development, we injected low doses of wnt8morpholinos that caused little or no effect on mesoderm patterning during gastrulation (see Fig. S1 in supplementary material, and C.J.T., unpublished results). In these experiments, we co-injected morpholinos against both open reading frames of the bicistronic wnt8 locus, referred to hereafter for simplicity as wnt8. These morpholinos have previously been shown to specifically inhibit wnt8 function(Lekven et al., 2001). Our partial knockdown of wnt8 moderately reduced TOPdGFP expression in the tailbud (Fig. 1G,H),although tail formation was not greatly affected(Fig. 1I). In contrast,knocking down both wnt3a and wnt8 results in a severe reduction in TOPdGFP expression at bud stage, and its complete absence in the tailbud by mid-somitogenesis (Fig. 1J,K). wnt3a/wnt8 double morphants are dramatically shorter, with most embryos lacking almost all tail structures(Fig. 1L, 68% of embryos make only 1-3 tail somites, and 18% make none, n=276). We conclude that both wnt3a and wnt8 are required for tail formation in zebrafish.

During gastrulation, wnt8 is required for dorsoventral mesoderm patterning and for promoting posterior neural fates in the neurectoderm(Erter et al., 2001; Lekven et al., 2001). Since wnt3a, like wnt8, is expressed in the margin during gastrulation (data not shown), we tested whether wnt3a also plays a role in patterning of the mesoderm and neurectoderm by co-injecting wnt3a and wnt8 MOs and examining the expression of marker genes. We found that while wnt3a is not required by itself for either of these processes, wnt3a knockdown modestly enhances the effects of partial loss of wnt8 function in both dorsoventral patterning of the mesoderm and anterior-posterior patterning of the neurectoderm (see Fig. S1 in supplementary material). Thus, while wnt8 is principally responsible for these two early patterning events, wnt3a also plays a more minor,redundant role. Consistent with this possibility, when higher doses of both wnt3a and wnt8 morpholinos were injected, nearly all embryos were strongly dorsalized (data not shown). Since we wanted to specifically address a later role for wnt3a and wnt8 in tail development,all further experiments were done using a hypomorphic dose of wnt8MOs to minimize early mesoderm patterning defects.

To gain further insight into the mechanism by which wnt3a and wnt8 promote tail development, we examined the expression of several markers expressed in the tailbud at multiple time points from bud stage through early somitogenesis (Fig. 2). In both wnt3a(Fig. 2B,F) and wnt8(Fig. 2C,G) morphants,expression of ntl and fgf8 is unaffected relative to wild-type (Fig. 2A,E) at bud stage. tbx6 is expressed normally in wnt3a morphants(Fig. 2J, compare with wild-type in I), and is slightly reduced in wnt8 knockdown embryos(Fig. 2K). Despite the much more severe tail phenotype in wnt3a/wnt8 morphants, ntl,fgf8 and tbx6 are all expressed at bud stage(Fig. 2D,H,L), with tbx6 slightly reduced, as seen in wnt8 morphants. The relative lack of effect of the wnt3a and wnt8 MOs on the expression of these genes was somewhat surprising, since expression of the TOPdGFP reporter is dramatically reduced at this stage in wnt3a/wnt8morphants (Fig. 1J), suggesting that even low levels of Wnt signaling are sufficient for initial specification of the tailbud. We conclude that the block in tail development apparent by mid-somitogenesis in wnt3a/wnt8 morphants is not due to a failure to specify early tailbud fates.

Since initial cell fates were properly specified, we then asked whether Wnt signals were required for maintenance of cell fates. As somitogenesis proceeds, ntl and fgf8, as well as bmp4, continue to be expressed in the tailbud of wnt3a, wnt8 and wnt3a/wnt8morphants, although ntl is moderately reduced by the 10-somite stage in wnt3a/wnt8 embryos (data not shown). In contrast, markers of presomitic mesoderm such as tbx6, spadetail (spt)(Griffin et al., 1998) and pipetail/wnt5 (ppt)(Krauss et al., 1992; Rauch et al., 1997) fade during early somitogenesis and are lost or drastically reduced by the 10-somite stage. In the case of tbx6, expression in wnt3a/wnt8 morphants, is dramatically reduced at 5 somites(Fig. 2P; 34/41 embryos have significantly reduced tbx6 expression relative to that seen in wnt8 morphants, Fig. 2O), and tbx6 expression is completely absent at 10 somites (Fig. 2T; 38/42 embryos lack expression). spt and ppt/wnt5 are similarly affected in wnt3a/wnt8 morphant embryos. At the 10-somite stage, expression of spt is reduced in both wnt3a MO and wnt8 MO embryos compared with wild type (Fig. 2U-W), but is absent in wnt3a/wnt8 morphants(Fig. 2X; 26/34 embryos lack spt expression). ppt/wnt5 is expressed at reduced levels in wnt3a (Fig. 2Z) and wnt8 (Fig. 2A′)morphants, and is even more strongly reduced in the double morphants(Fig. 2B'; 20/27 embryos). As ppt/wnt5 is itself required for tail extension movements(Hammerschmidt et al., 1996; Rauch et al., 1997), it is possible that the reduction in ppt/wnt5 expression could contribute to the tail truncation we observe in wnt3a/wnt8 morphants. However,while we do observe some compression of somites along the anterior-posterior axis (see Fig. 1K), we do not observe the lateral expansion of posterior somites typical of ppt/wnt5 mutants, nor do we observe any accumulation of cells in the tailbud as somitogenesis proceeds, but rather a pronounced reduction in size(see Fig. 1L). Thus, the effect of the reduction in ppt/wnt5 expression is likely to be minor. Taken together, these results suggest that wnt3a/wnt8 signaling is required for maintenance of presomitic mesoderm fates within the tailbud during early somitogenesis.

In the chick and mouse, Fgf signaling has been proposed to play a similar anti-differentiation role, such that higher levels of Fgf in the caudal tailbud inhibit the differentiation of presomitic mesoderm, and as cells move more anteriorly within the tailbud, they escape the influence of Fgf,differentiate, and form somites (Dubrulle et al., 2001; Dubrulle and Pourquie, 2004; Mathis et al.,2001). Since these data support a role for Fgf signaling in maintaining posterior cells in an undifferentiated state, we asked whether the failure to maintain presomitic fates in wnt3a/wnt8 morphant embryos was due to a defect in Fgf signaling. Although we had already observed normal expression of fgf8 in the tailbud (see above), at least three additional Fgf ligands, fgf17b, fgf3 and fgf24, are also known to be expressed in this region (Cao et al., 2004; Draper et al.,2003; Phillips et al.,2001).

To test whether Fgf signaling is compromised in the tailbud of wnt3a/wnt8 embryos, we examined expression of sprouty4(Fig. 3), a Fgf-induced inhibitor of the Fgf receptor that is expressed in the tailbud in response to Fgf activity (Furthauer et al.,2001). At both 5- and 10-somite stages, we observe robust expression of sprouty4 in the tailbuds of wnt3a/wnt8morphants (Fig. 3D,H). In particular, at the 10-somite stage, when markers of presomitic mesoderm are completely absent or severely reduced (above, Fig. 2), sprouty4expression, and thus Fgf signaling, is still observed. These data indicate that a loss of Fgf signaling is not responsible for the failure to maintain presomitic mesoderm in wnt3a/wnt8 morphants, and suggests a direct role for Wnt/β-catenin signaling in maintaining expression of presomitic mesoderm markers in the tailbud.

We next tested whether a decrease in cell proliferation within the tailbud could account for the loss of presomitic mesodermal fates by staining embryos with an antibody to phosphohistone H3, a marker of mitotic cells, at the 4-somite stage. We observed no apparent reduction in cell proliferation in the tailbuds of wnt3a, wnt8, or wnt3a/wnt8 morphant embryos (see Fig. S2 in supplementary material). We also measured cell death within the tailbud at several stages during somitogenesis using the TUNEL assay, and observed no obvious differences in levels of cell death between wild-type embryos and any of the morpholino-injected embryos (data not shown). In sum,our data indicate that Wnt signaling is required for maintenance of the expression of presomitic mesoderm markers, but not for regulation of cell proliferation or cell death in the tailbud.

The mouse Wnt3a knockout results in a severe early phenotype, with a loss of notochord and somites caudal to the forelimb bud, such that most mutant embryos make only 7-8 somites(Takada et al., 1994). Cells involuting through the primitive streak that would normally contribute to somitic tissue instead adopt a neural fate and form an ectopic neural tube(Yoshikawa et al., 1997). These results in the mouse raise the question of whether Wnt loss of function in zebrafish might result in similar fate transformations. Close examination of wnt3a morphants at 28 hpf shows an absence of notochord near the end of the tail (Fig. 4B,arrow; 91% of embryos have premature truncation of the notochord, n=253, compared with wild-type; Fig. 4A), indicating that wnt3a is required for late notochord development. In contrast, we observe no notochord defects in the tail of wnt8 morphants(Fig. 4C). While most wnt3a/wnt8 morphants completely lack or make only a rudimentary tail,the occasional weakly affected embryos have an even more penetrant and severe loss of notochord in the tail than that seen in wnt3a morphants(Fig. 4D, arrow; 100% of embryos have a notochord phenotype, n=78). We conclude that wnt3a and wnt8 both function to specify notochord during tail outgrowth, with wnt3a playing the central role.

The ntl gene is required for notochord formation in the tail(Amacher et al., 2002; Halpern et al., 1997; Halpern et al., 1993; Schulte-Merker et al., 1994),and its murine homolog, T, has been shown to be a direct transcriptional target of Wnt3a(Yamaguchi et al., 1999). We therefore tested whether wnt3a and/or wnt8 regulate notochord formation in the tail by regulating ntl expression in the tailbud. While wnt3a morphants have a gap of ntl expression where the notochord has failed to form(Fig. 4F, arrow), expression within the tailbud is not affected at 28 hpf(Fig. 4F, arrowhead). wnt8 morphants have normal notochord development(Fig. 4G, arrow) and also show normal levels of ntl expression in the tailbud(Fig. 4G, arrowhead). In contrast, wnt3a/wnt8 morphants make no notochord tissue within their truncated tails (Fig. 4H,arrow) and also lack ntl expression in the tailbud(Fig. 4H, arrowhead; compare with wild type in Fig. 3E). The truncated notochords of wnt3a/wnt8 embryos also display a characteristic sharp bend or kink at the terminus, leading to a broadened patch of ntl expression (Fig. 4H, arrow). This probably does not represent an anteriorward displacement of the tailbud domain of ntl expression, as we observe a similar pattern of notochord expression of collagen2α, which is not expressed in the tailbud (data not shown). Thus, while ntlexpression is dependent on wnt3a and wnt8, loss of ntl expression does not account for the absence of notochord, since wnt3a morphants lack the caudal notochord, but still express ntl in the tailbud.

Since mouse Wnt3a is required for production of somites as well as notochord (Takada et al.,1994), we next investigated whether somites were properly formed in the tail of wnt3a, wnt8 or wnt3a/wnt8 morphants. Since most wnt3a/wnt8 morphants make essentially no tail, we examined more weakly affected embryos that made a small tail. To assess the production of somites, we stained 26 hpf embryos for α-tropomyosin(Fig. 4I-L). Whileα -tropomyosin staining extends nearly to the tip of the tail in both wild-type (Fig. 4I) and wnt8 morphants (Fig. 4K), we observe a slight reduction of expression in wnt3amorphants (Fig. 4J; 17/24 embryos), indicating a deficit in formation of somitic mesoderm. In wnt3a/wnt8 MO embryos (Fig. 4L; 29/29 embryos), α-tropomyosin staining is completely absent from the caudal tail; absence of somites and premature termination of the notochord occur at a similar rostral/caudal level (compare Fig. 4H,L). Thus, wnt3a and wnt8 are also required for production of somitic mesoderm in the tail.

In addition to the truncation of the notochord and loss of somites found in these embryos, we also observed an apparent expansion of neural tube tissue posterior to the end of the notochord in these more mildly affected wnt3a/wnt8 embryos (4D, arrowhead indicates enlarged lumen of neural tube). To confirm this with molecular markers, we stained embryos at 26 hpf for F-spondin, which is expressed in the floor plate of the neural tube (Fig. 4M-P)(Higashijima et al., 1997). We observe a significant expansion of F-spondin expression in the posterior of wnt3a/wnt8 morphants(Fig. 4P). Expression of collagen 2α in the floor plate is also expanded, confirming the expansion of floor plate tissue (data not shown). In contrast, the panneural marker ngn1 (Korzh et al.,1998) is not expanded in wnt3a/wnt8 morphants(Fig. 4T), suggesting that the loss of Wnt signaling results in an expansion only of the floor plate, and not of other fates within the neural tube.

Lastly, we examined the expression of markers of other tail tissues,including ventral fin epidermis (msxb)(Akimenko et al., 1995), blood(gata1) (Stainier et al.,1995), and vasculature (fli1)(Thompson et al., 1998), as well as the tailbud marker (eve1)(Joly et al., 1993). These data are presented in Fig. S3 in supplementary material. Briefly, while eve1 expression in the tailbud is absent in wnt3a/wnt8morphants, blood and vasculature is specified normally, suggesting that wnt3a and wnt8 are not required for more lateral posterior mesodermal fates. Also, expression of msxb in the ventral tailfin was absent, possibly reflecting a mild dorsalization of wnt3a/wnt8embryos. Taken together, our data show that wnt3a and wnt8are required for the formation of notochord and somitic mesoderm in the tail,as well as for inhibiting production of floor plate cells.

To identify genes that function downstream of Wnt signaling during early development, we performed a microarray screen for Wnt-responsive genes. Briefly, RNA from early gastrula stage embryos that had been injected with either wnt8 RNA or GFP RNA was used to probe a microarray chip containing 8,000 zebrafish cDNAs from a mixed stage cDNA library. One gene,the RNA levels of which were significantly upregulated by overexpression of wnt8, was sp5-like (sp5l), a member of the Sp1 family of zinc-finger transcription factors. sp5l has previously been described as spr2, and has been implicated in mesoderm induction,acting downstream of FGF signaling (Zhao et al., 2003).

We used several independent assays to show that sp5l is regulated by Wnt signaling (Fig. 5). First, we injected wnt8 RNA and examined sp5l expression at early gastrula stage. In embryos overexpressing Wnt8, we observed ectopic sp5l expression at the animal pole (arrowhead in Fig. 5A, right panel),confirming that activation of the Wnt pathway can activate sp5lexpression. Conversely, we used a dominant-negative, heat shock-inducible,ΔTCF GFP transgenic line to block the activation of Wnt/β-catenin target genes (Lewis et al.,2004) to test whether sp5l was down-regulated. We found that expression of ΔTCF GFP can repress sp5l expression very rapidly: 15 minutes after induction of ΔTCF GFP expression at tailbud stage, sp5l RNA levels were already substantially reduced(Fig. 5B, middle panel) and hardly detectable after 30 minutes (Fig. 5B, right panel). This rapid downregulation suggests thatΔTCF GFP directly represses sp5l transcription and that sp5l may be a direct target of Wnt/β-catenin signaling. In support of this, the sp5l promoter contains six consensus Tcf/Lef binding sites within the 519 bp 5' of exon 1, which bind Lef1 protein in vitro and are required for Wnt responsiveness in reporter assays conducted in zebrafish embryos (Weidinger et al.,2005). Thus, sp5l is a direct Wnt/β-catenin target gene.

As the above experiments did not reveal which endogenous Wnt ligand(s)regulate sp5l, we examined sp5l expression in wnt3a,wnt8 and wnt3a/wnt8 morphant backgrounds. During early gastrulation, wnt8, and not wnt3a, is required for full sp5l expression (Weidinger et al., 2005) (C.J.T., unpublished). At the 3-somite stage, when sp5l is robustly expressed in the tailbud of control embryos(Fig. 5C), its expression is significantly reduced in wnt3a morphants(Fig. 5D), and even more substantially reduced in wnt8(Fig. 5E) and wnt3a/wnt8 (Fig. 5F)morphant embryos. Inhibition of wnt3a and wnt8 does not completely eliminate sp5l expression, in contrast to what we observe following induction of ΔTCF GFP. This may be due to residual Wnt activity in our double morphant embryos. We conclude that by early somitogenesis, sp5l expression in the tailbud is largely dependent on Wnt signaling, principally wnt8.

Since sp5l expression is regulated by wnt3a and wnt8 in the tailbud, we examined whether knockdown of sp5lcould enhance the defects seen in wnt3a or wnt8 morphants. We used a translation blocking morpholino previously shown to specifically inhibit sp5l function (Zhao et al., 2003). We observed no enhancement of the wnt8morphant phenotype when sp5l MOs were co-injected (data not shown),perhaps because sp5l expression is already so dramatically reduced in wnt8 morphants. In contrast, co-injection of sp5l MO strongly enhances the phenotype of wnt3a MO embryos. While sp5l morphants have no apparent tail defects at 48 hpf(Fig. 6C), and wnt3aMO embryos are only slightly shorter than wild-type embryos(Fig. 6E, compare with wild type in Fig. 6A), wnt3a/sp5l morphants are dramatically shorter(Fig. 6G; 93% of embryos, n=104). Similarly, at the 10-somite stage, wnt3a/sp5lmorphants (Fig. 6H) have greatly reduced expression of tbx6 relative to sp5l MO(Fig. 6D) or wnt3a MO(Fig. 6F) alone (32/34 embryos with significantly reduced staining relative to either single MO injection). Also, sp5l MO enhances the penetrance of the notochord truncation phenotype observed in wnt3a morphants [67% of embryos injected with 1 mg/ml wnt3a MO plus control MO have truncated notochords(n=73) vs 100% when co-injected with sp5l MO(n=75)]. Lastly, like wnt3a/wnt8 morphants, wnt3a/sp5l embryos lack somites posterior to the truncated notochord(data not shown). These data indicate that sp5l functions redundantly with wnt3a in tail development, both in regulation of presomitic mesoderm markers such as tbx6 and in promoting mesodermal fates in the caudal tail.

Since sp5l expression in the tailbud is regulated by Wnt/β-catenin signaling, and sp5l functionally interacts with wnt3a in tail formation, we next asked whether sp5l could be placed functionally downstream of wnt3a. We assessed this by co-injecting sp5l RNA with the wnt3a MO, and scoring for rescue of the notochord truncation phenotype at 48 hpf(Fig. 7). Co-injection of 200 pg sp5l RNA (Fig. 7B),but not control RNA (Renilla luciferase; Fig. 7A) resulted in significant rescue (see graph in Fig. 7C). Taken together, our observations indicate that sp5lis a functional target of Wnt signaling that plays key roles in development of the posterior body in zebrafish.

Cell labeling experiments within the caudal presomitic mesoderm of zebrafish indicate that individual cells and their descendants reside in this region for variable amounts of time before exiting and beginning to differentiate (Muller et al.,1996). Thus, even as cells exit the tailbud and form somites, a population of undifferentiated precursors that will give rise to more caudal structures is maintained. Our data show that tailbud fates are initially specified correctly in wnt3a/wnt8 morphants, followed by a rapid decline in the expression of presomitic mesoderm markers. As we observe no significant changes in cell death or gross levels of cell proliferation in the tailbud of wnt3a/wnt8 morphants, what then is the cause of the loss of presomitic mesoderm and failure in tail development? One possible explanation to account for our observations is that Wnts could be acting as anti-differentiation factors in the presomitic mesoderm, where Wnt activity is localized, as determined by expression of the TOPdGFP reporter(Dorsky et al., 2002) (C.T.,unpublished). In the absence of Wnt activity, presomitic mesoderm cells might prematurely differentiate and form somites, leaving no undifferentiated cells behind to maintain the tailbud. In this way, the cells initially allocated to the tailbud are quickly exhausted and tail development is dramatically truncated. Intriguingly, in spadetail mutants, presomitic cells fail to differentiate into somites and instead accumulate in the tailbud, and it has been proposed that this phenotype may be due to a failure to turn off genes such as wnt8 that are normally highly expressed only in the caudal tailbud (Griffin and Kimelman,2002). Although an exciting possibility, it will be necessary to carefully map the fates of presomitic precursors in wnt3a/wnt8embryos to directly test this model.

Alternatively, as Wnt3a has recently been shown to regulate oscillation of the segmentation clock (via regulation of axin2) in mouse presomitic mesoderm (Aulehla et al.,2003), it is also possible that a defect in segmentation underlies the failure of tail formation observed in wnt3a/wnt8 morphants. We do not favor this hypothesis for two main reasons. First, the data from the mouse experiments indicates that repression of Wnt signaling in the presomitic mesoderm leads to larger somites being formed, while in our wnt3a/wnt8 embryos, the posterior somites appear somewhat smaller than normal (see Fig. 4L). Secondly, in zebrafish segmentation mutants, as well as in the mouse Wnt3a mutant, vestigial tail, defects in segmentation are associated with a failure of somite boundary formation, leading to diffuse,unsegmented expression of somitic markers such as myoD(Aulehla et al., 2003; Holley et al., 2000). We do not observe this phenotype in wnt3a/wnt8 morphants (see Fig. 1K,L; Fig. 4L), suggesting that segmentation is grossly normal. A detailed analysis of the expression of oscillating clock genes, such as her1, will be necessary to definitively address this issue, although it is noteworthy that axin2expression is not reported to cycle in zebrafish as it does in the mouse(Aerne and Ish-Horowicz, 2004),raising the possibility that regulation of segmentation in these two vertebrates may not be strictly orthologous.

Zebrafish wnt3a is required for formation of caudal notochord,which is missing in nearly all wnt3a MO embryos. Since Wnt3adirectly activates the transcription of T in the mouse(Yamaguchi et al., 1999), and the zebrafish T homolog, ntl, is required for notochord formation (Halpern et al.,1993; Schulte-Merker et al.,1994), it is tempting to speculate that the notochord defect in wnt3a morphants is due to a loss of expression of ntl. Although ntl continues to be expressed in the tailbud of wnt3a morphants (Fig. 3), ntl expression at the caudal end of the notochord– the chordoneural hinge – is lost, beginning at approximately the 20-somite stage (C.J.T., unpublished results). Fate mapping studies have shown that cells in this region can adopt a notochord or floor plate fate, and ntl has been shown to be important in promoting notochord fates; ntl mutants lack notochord and produce ectopic floor plate cells(Halpern et al., 1997). Although we observe a clear loss of notochord in the caudal tail of wnt3a morphants, we do not observe any increase in floor plate,suggesting that notochord progenitors may not be adopting floor plate identity. In the mouse Wnt3a mutant, the ectopic neural tissue observed in the absence of notochord and somites is derived from prospective somitic mesodermal cells (Takada et al.,1994; Yoshikawa et al.,1997). Our observation that ectopic floor plate is not produced when only notochord is missing (in the wnt3a morphant), but only when somitic mesoderm is also lacking (the wnt3a/wnt8 double morphant),suggests that the same fate transformation could be occurring in the zebrafish.

Careful fate mapping within the tailbud to determine the origin of the ectopic floor plate will be required to directly address this possibility.

Sp1-related transcription factors are characterized by having multiple zinc-finger DNA binding domains related to the kruppel gene from Drosophila (Pieler and Bellefroid, 1994). Sp1 proteins bind to GC-rich promoter regions,and while some members of this protein family are expressed ubiquitously and are thought to be required generally as enhancers of transcription, others,including members of the Sp5 subfamily, are expressed in restricted domains and are thought to participate in specific processes during development(Bell et al., 2003; Briggs et al., 1986; Harrison et al., 2000; Tallafuss et al., 2001; Treichel et al., 2001; Treichel et al., 2003).

Our data strongly suggest that sp5l functions downstream of Wnts in zebrafish tail development. What role might sp5l play in this process? One function could be to bind the promoters of downstream Wnt targets and enhance their activation. For example, in the mouse T promoter,both Tcf/Lef sites and regions containing Sp1 binding sites are important for normal expression of T in the primitive streak and tailbud(Clements et al., 1996; Yamaguchi et al., 1999), and mutations in the Sp1 family member Sp5 enhance the tail truncation phenotype of heterozygous T/+ mice(Harrison et al., 2000),suggesting that Sp5 could be functioning with Wnt3a to activate T transcription. Also, in vitro analysis indicates that LEF-1 can activate transcription from the HIV-1 promoter only when an Sp1-containing fraction, or purified Sp1 protein, is added to the transcription reaction (Sheridan et al.,1995).

The synergistic loss of tbx6 expression observed when both sp5l and wnt3a are inhibited suggests that full activation of the tbx6 promoter may also require the binding of bothβ-catenin/Tcf complexes to Tcf/Lef sites(Szeto and Kimelman, 2004) and Sp5l to putative Sp1 binding sites found in the tbx6 promoter(C.J.T., unpublished). Additional deletion analysis of the tbx6promoter will be required to more directly address a requirement for sp5l in its activation.

Since the GC-rich promoter elements bound by Sp1 proteins are found upstream of many genes, it is conceivable that activation of multiple Wnt targets could be potentiated by sp5l. Since sp5l morphants,like the mouse Sp5 knockout, have no discernable tail phenotype, this function may not be essential, or may be redundantly encoded. The latter is a distinct possibility, as in both mouse and zebrafish, another Sp5 homolog is largely co-expressed in the same tissues(Harrison et al., 2000; Tallafuss et al., 2001). Interestingly, the other identified zebrafish Sp5 homolog, called buttonhead/Sp-related 1(bts1), is also a target of Wnt signaling(Harrison et al., 2000; Tallafuss et al., 2001; Weidinger et al., 2005). Also,a recent report suggesting that a mouse buttonhead homolog, mBtd,also called SP8, may play a role in maintaining expression of Wnt targets in the limb bud (Bell et al.,2003; Treichel et al.,2003) is an additional link between Sp1 family members and Wnt signaling during vertebrate development.

We suggest that in addition to a previously characterized role during gastrulation for specification of future tail fates, Wnt/β-catenin signaling is also required during somitogenesis for maintenance of presomitic mesoderm in the tailbud, and also to promote mesodermal fates and inhibit floor plate formation in subsequent tail outgrowth(Fig. 8). The early functions in dorsal-ventral patterning and specification of the tail organizer are principally carried out by wnt8. During early somitogenesis, both wnt3a and wnt8 function to maintain presomitic mesoderm fates in the tailbud, while the later mesoderm/floor plate distinction is more sensitive to loss of wnt3a function. Thus, Wnt/β-catenin signaling, functioning in part through sp5l, is required throughout tail development to properly specify, pattern and maintain a precursor population in the tailbud. Further examination of the defects in mesoderm formation in Wnt-inhibited embryos and identification of additional Wnt/β-catenin targets will be important for a more complete understanding of tail development.

We thank Thuy Tran for excellent fish care, members of the Moon laboratory for comments on the manuscript, and Kevin Griffin for helpful discussions. C.J.T. and G.W. are Associates and R.T.M. an Investigator of the HHMI, which supported this research. G.W. also acknowledges the Austrian Academy of Science for an APART fellowship.