nemo-like kinase is an essential co-activator of Wnt signaling during early zebrafish development

Wnt proteins comprise a family of secreted signaling molecules that perform essential roles in numerous processes during animal development. In early vertebrate embryos, Wnt signaling pathways have been implicated in axis specification, cell proliferation, mesoderm and neuroectoderm patterning, and in the regulation of cell movement during gastrulation (reviewed by Cadigan and Nusse, 1997; Peifer and Polakis, 2000; Veeman et al., 2003). Intensive study of the different roles of Wnt signals and their underlying molecular mechanisms has led to a general grouping of Wnt pathways into two classes: canonical (β-catenin dependent) and non-canonical(β-catenin independent) (reviewed by Veeman et al., 2003). Canonical Wnt signaling functions via the stabilization of cytoplasmic pools of β-catenin, which is constitutively degraded in the absence of Wnt signaling. Stabilized β-catenin then translocates to the nucleus and interacts with members of the Tcf/Lef family of transcription factors. Tcf/Lef proteins bound alone to DNA can repress transcription because of their binding of general repressors such as Groucho and CtBP. β-catenin contains transcriptional transactivation domains, so it is thought that by binding to Tcf/Lef proteins bound to DNA, β-catenin converts the repressive Tcf/Lef to an activator.

Recently, evidence that canonical Wnt signaling can be regulated by MAP kinase signaling has emerged. Genetic analysis of endoderm specification in the early C. elegans embryo demonstrates an essential requirement for both Wnt signaling and a MAPK-related pathway for proper specification of endodermal fate (Meneghini et al.,1999; Rocheleau et al.,1997; Rocheleau et al.,1999; Thorpe et al.,1997). Two of the genes involved in this process, lit-1and mom-4, encode homologs of the Drosophila MAPK-related gene nemo and vertebrate TGFβ-activated kinase (Tak1),respectively. Epistasis studies have shown that a C. elegans Tcf/Lef homolog, pop-1, represses endoderm fates, and that this repression is relieved by the combinatorial action of Wnt signaling and a parallel pathway involving lit-1/nemo and mom-4/Tak1, permitting endoderm development (Meneghini et al.,1999; Rocheleau et al.,1997; Rocheleau et al.,1999; Shin et al.,1999; Thorpe et al.,1997). Studies of a mouse homolog of lit-1/nemo,nemo-like kinase (Nlk), in mammalian cell culture also demonstrate that Nlk and Tak1 can regulate activation of Wnt targets(Ishitani et al., 2003b; Ishitani et al., 1999). Biochemical studies indicate that Tak1 activates Nlk, which can then phosphorylate Tcf, inhibiting the DNA-binding ability of β-catenin/Tcf complexes (Ishitani et al.,2003b; Ishitani et al.,1999; Rocheleau et al.,1999).

In different contexts, Nlk proteins can function as either inhibitors or activators of Wnt target genes. For example, in the early C. elegansembryo, the eventual outcome of the Wnt signal may be to simply derepress genes that are inhibited by Tcf proteins (for reviews, see Behrens, 2000; Thorpe et al., 2000). Following the elimination of the repressive activity of pop-1/Tcf by lit-1/Nlk and wrm-1/β-catenin, other elements within the promoter can drive transcription. For example, two GATA factors, med-1 and med-2, can initiate transcription of the earliest endoderm-specific genes when pop-1 repression is relieved(Maduro et al., 2001).

In other contexts, where a β-catenin/Tcf complex is required to directly activate transcription, phosphorylation of Tcf by Nlk and subsequent inhibition of DNA binding of the β-catenin/Tcf complex would block activation, as is shown in experiments where Nlk blocks activation of the Wnt-responsive TOPFLASH reporter (Ishitani et al., 2003b; Ishitani et al., 1999). Also, injection of mouse or Xenopus Nlk RNA into Xenopus embryos blocks duplication of the dorsal axis induced by Wnt or β-catenin (Hyodo-Miura et al.,2002; Ishitani et al.,1999).

A role for Wnt signaling in overcoming Tcf-mediated repression has recently been described in zebrafish (Kim et al.,2000; Dorsky et al.,2003). A Tcf3 homolog, encoded by the headless (hdl)gene, is required to repress Wnt signaling during neural patterning(Kim et al., 2000). wnt8 is required during gastrulation to induce posterior neural fates(Erter et al., 2001; Fekany-Lee et al., 2000; Lekven et al., 2001), and this posteriorizing activity is opposed in prospective anterior neuroectoderm by hdl/tcf3a (Kim et al.,2000). hdl/tcf3a mutant embryos lack anterior neural structures and have expanded posterior neural fates, consistent with ectopic Wnt signaling. Inhibition of tcf3a results in expanded posterior neural fates even in the absence of wnt8 function, supporting a model in which Wnt signaling functions solely to derepress Tcf-inhibited genes(Dorsky et al., 2003). Furthermore, a truncated Tcf3a protein that cannot bind β-catenin and can thus only act as a repressor completely rescues the hdl mutant(Kim et al., 2000). To determine whether the cooperative function of Nlk and Wnt signaling in derepressing Tcf-inhibited genes was conserved from nematodes to teleost fish,we undertook an analysis of Nlk function in zebrafish embryos.

Zebrafish were raised and maintained under standard conditions. In situ hybridization using digoxigenin-labeled mRNA probes was performed using standard methods (Oxtoby and Jowett,1993).

We used degenerate PCR on a 24-hour-stage cDNA library, followed by 5′ and 3′ RACE using the SMART RACE kit (Clontech), to amplify a cDNA containing the entire nlk ORF, plus 170 base pairs (bp) of 5′ UTR and 574 bp of 3′UTR sequence. This sequence was deposited in GenBank (Accession Number AY562552). For in situ hybridization, a fragment containing the complete ORF, plus 62 bp of 5′ UTR and 130 bp of 3′UTR was amplified and ligated into pGEM-T (Promega). For mRNA synthesis, we designed primers corresponding to the 5′ and 3′ ends of the ORF,and inserted the full-length product into pCS2+(Turner and Weintraub, 1994). The nlk point mutants, nlk (K117M) and nlk (C387Y),were constructed using standard PCR-based site-directed mutagenesis.

For mRNA injections, sense transcripts were synthesized using the mMessage mMachine kit (Ambion). For templates, we used full-length cDNA inserted into pCS2+. mRNA was resuspended in water or Danieau's buffer prior to injection. Morpholino antisense oligonucleotides (nlk:5′-GTGTGTGGTACCTTAAGCAGACAGT-3′) were obtained from Gene Tools(Philomath, OR). tcf3b, lef1, wnt8b, wnt11, wnt8 ORF1 and wnt8 ORF2 morpholinos have been previously described(Dorsky et al., 2003; Dorsky et al., 2002; Houart et al., 2002; Lekven et al., 2001; Lele et al., 2001). The standard control morpholino provided by Gene Tools was used in some experiments. Morpholinos were dissolved in Danieau's buffer(Nasevicius and Ekker, 2000)prior to use. For all injections, 2-3 nl of a 1 ng/nl stock was injected at the one-cell stage, except where otherwise noted.

We used degenerate PCR on a 24-hour-stage zebrafish cDNA library, then 5′ and 3′ RACE to amplify a full-length open reading frame for nlk. The predicted protein sequence is most similar to a recently identified Xenopus Nlk (76% identical, 83% similar). A protein sequence alignment between zebrafish, frog and human Nlks shows that all three proteins are highly homologous throughout the predicted kinase domains, but are very divergent over the first 100 residues(Fig. 1A).

RT-PCR analysis (data not shown) shows that nlk is expressed maternally and throughout the first day of development. By in situ hybridization, nlk is expressed strongly and ubiquitously throughout gastrulation stages (Fig. 1Band data not shown). At tailbud stage, nlk is still expressed throughout the embryo, but the presumptive notochord shows higher levels of expression (Fig. 1C,arrowhead). During somitogenesis (Fig. 1D,E), nlk expression is observed exclusively in the neural tube, including in the eyes by 24 hours post-fertilization (hpf). Expression of nlk is more restricted at 48 hpf, becoming localized to the otic vesicles (Fig. 1F,arrowheads) and the ganglion cell layer of the retina (arrow), as well as to unidentified regions of the ventral brain.

As a first test to assess the activity of nlk during early development, we injected in vitro-transcribed nlk mRNA into one-cell zebrafish embryos. Injection of 200 pg nlk RNA results in embryos with variable loss of eyes and forebrain(Fig. 2D and data not shown),and no other obvious morphological abnormalities at 24 hpf (59% of embryos have small or missing eyes; n=510). Changing a conserved cysteine residue to tyrosine at amino acid 425 in mouse Nlk eliminates its ability to bind Tcf (Ishitani et al.,1999). We made the analogous change in zebrafish nlk(C387Y) and found that injection of this mutated RNA had almost no effect on development (1% of embryos had small eyes; n=190), suggesting that the nlk overexpression phenotype is dependent on interaction with Tcf proteins. Furthermore, we constructed nlk (K117M), analogous to mouse Nlk (K155M), which is kinase dead and unable to block Tcf binding to DNA (Ishitani et al., 1999). Injection of nlk (K117M) RNA has no effect on development (2% of embryos had small eyes; n=109), indicating that the kinase activity of nlk is required to induce the loss of anterior neural structures.

Two zebrafish Tcf3-related genes, tcf3a and tcf3b, have been shown to cooperatively regulate AP patterning in the neurectoderm by repressing posterior neural fates (Dorsky et al., 2003). In tcf3a mutants, posterior neural fates expand at the expense of anterior fates, causing embryos to develop lacking telencephalon and eyes (Kim et al.,2000). Although tcf3b morphants do not show any defects in AP patterning of the neurectoderm, the tcf3a phenotype is enhanced when tcf3b is also inhibited, suggesting that the two genes function redundantly (Dorsky et al.,2003). Given the similarity between the nlkoverexpression and the tcf3a loss-of-function phenotypes, as well as the established role for the C. elegans Nlk homolog in derepression of tcf-inhibited genes, we reasoned that nlk was likely to be interacting with one or both of the zebrafish Tcf3 homologs. We therefore compared the effects of nlk overexpression and tcf3 loss of function on neural markers, and also examined the ability of nlkoverexpression to enhance the phenotypes due to loss of either tcf3homolog alone (Fig. 2).

Injection of tcf3a MO results in a loss of telencephalon and eyes at 24 hpf (Fig. 2G, compare with wild type in Fig. 2A), and caused an expansion of the midbrain/hindbrain boundary (MHB) domain of pax2 (Fig. 2H) and a decrease in the diencephalic expression of pax6(Fig. 2I), reflecting the increase in caudal character of the neurectoderm. Injection of tcf3bMO caused only a disorganization of the hindbrain(Fig. 2J); both pax2and pax6 expression was normal(Fig. 2K,L). Co-injection of both tcf3 morpholinos strongly enhanced the tcf3a phenotype,eliminating the morphological MHB at 24 hpf(Fig. 2M), greatly expanding pax2 expression, and eliminating pax6 expression(Fig. 2N,O).

Injection of 200 pg of nlk RNA results in anterior truncations like those seen in tcf3a morphants(Fig. 2D). Similar to tcf3a morphants, pax2 expression is expanded and pax6 expression is reduced (Fig. 2E,F). nlk overexpression, then, phenocopies inhibition of tcf3a function, suggesting that nlk can negatively regulate the repression of posterior neural genes by tcf3a. Co-injection of 100 pg tcf3a RNA resulted in significant rescue of the nlk overexpression phenotype [62% (n=63) of nlk/tcf3a RNA-injected embryos had normal pax2 expression at 3 somites versus 31% (n=51) of nlk/GFP RNA-injected embryos], further indicating that the nlk overexpression phenotype is due to downregulation of tcf3a. Although most nlk-injected embryos closely resemble tcf3a morphants, when we injected high doses(800 pg) of nlk RNA, we saw embryos with a slightly more severe phenotype, but not to the degree seen in tcf3a/tcf3b double morphants(data not shown). This suggests that although nlk may preferentially interact with tcf3a, it may also be able to inhibit tcf3b to some degree. To test this, we co-injected nlk RNA with either tcf3a or tcf3b morpholinos. If nlk can downregulate tcf3b in addition to tcf3a, then co-injection of nlk RNA and tcf3a MO should phenocopy tcf3a/tcf3b morphants, showing both the dramatic expansion of pax2 and the loss of diencephalic pax6 expression. Indeed, nearly every embryo shows a more severe phenotype than that seen by injection of nlk RNA alone, with most phenocopying tcf3a/tcf3b morphants (82%; n=73; Fig. 2P,Q,R, compare with Fig. 2M,N,O). Most nlkRNA/tcf3b MO embryos also show the more severe phenotype characteristic of tcf3a/3b morphants (89% of embryos completely lack pax6; n=54; Fig. 2S,T,U). These results indicate that nlk is capable of interacting with both tcf3a and tcf3b to relieve their repression of posterior neural genes.

The nlk phenotype is similar to that seen with injection of low levels (5-10 pg) of zebrafish wnt8 RNA(Kelly et al., 1995). wnt8 is required to induce posterior fates within the neurectoderm,and overexpression of wnt8 expands posterior neural identity at the expense of anterior fates (Erter et al.,2001; Lekven et al.,2001). To test whether nlk inhibits expression of anterior neural markers during gastrulation, we fixed nlk- or wnt8-injected embryos at 90% epiboly and performed in situ hybridization using probes for opl, a telencephalic marker, and tbx6, a marker of ventrolateral mesoderm(Fig. 3). Like wnt8-injected embryos, nlk-injected embryos show a dramatic reduction of opl expression, indicative of a loss of anterior neural identity (compare the loss of opl in Fig. 3B,C with wild-type expression in 3A, indicated by arrowheads).

Phosphorylation of Tcf by Nlk specifically inhibits the DNA-binding ability of β-catenin/Tcf complexes, and not that of Tcf alone(Ishitani et al., 1999),suggesting that Nlk would only inhibit Tcf in the presence of β-catenin. To test whether the nlk overexpression phenotype is dependent onβ-catenin stabilized by Wnt signaling, we co-injected morpholinos to two wnt8 loci (hereafter referred to as wnt8 ORF1 and ORF2)known to be required for proper AP patterning of the neuroectoderm(Lekven et al., 2001). We injected nlk RNA with either wnt8 MOs or a control MO, and stained embryos at 90% epiboly with probes for opl and tbx6. We also scored for the presence of eyes at 24 hpf. Injection of wnt8MOs expands anterior neural fates, as shown by a posterior shift in the opl expression domain towards the margin(Fig. 3D, arrowhead; compare with wild type in 3A). Although nlk RNA/control MO embryos have dramatically reduced opl expression(Fig. 3E, arrowhead),co-injection of wnt8 MOs with 200 pg of nlk RNA resulted in substantial rescue of opl expression(Fig. 3F, arrowhead; opl expression domain is shifted posteriorly, as with wnt8MOs alone, see Fig. 3D). The formation of eyes is also rescued at 24 hpf[Fig. 3H versus 3G, arrows; 15%of nlk/wnt8 MO-injected embryos had small/reduced eyes(n=78) versus 56% of nlk/control MO injected embryos(n=52)]. We did not observe completely penetrant rescue of the nlk overexpression phenotype, possibly because of residual wnt8 activity. This suggests that zebrafish nlk, as has been postulated for its mammalian counterparts, requires wnt8, and thus probably β-catenin, for full activity.

To assess the function of endogenous nlk, we used morpholino antisense oligonucleotides to interfere with the splicing of nlk mRNA(Draper et al., 2001). We designed a morpholino to anneal to the splice donor site between exon 5 and intron 5 (see Fig. 4). Translation of the improperly spliced message would result in a truncated, probably non-functional, protein, with the 229 amino acids encoded by exons 1-5 appended to 19 non-conserved amino acids encoded by the 5′ end of intron 5, followed by a stop codon.

To test the efficacy of the nlk MO, we injected 1-3 ng nlk MO at the one-cell stage and isolated RNA from 24 hpf embryos(Fig. 4). We then performed RT-PCR using primers flanking intron 5 that are predicted to amplify a fragment of 233 bp from the spliced message and 320 bp from unspliced mRNA. We amplified only a band of approximately 230 bp from wild-type cDNA. By contrast, a larger product of approximately 320 bp was predominant in cDNA prepared from nlk MO-injected embryos. We confirmed that this larger band corresponds to the unspliced mRNA by DNA sequencing. Splicing of introns from unrelated genes is unaffected by injection of nlk MO, nor does injection of splice-blocking MO's targeted to other genes affect the splicing of nlk (data not shown). Although injection of higher doses of nlk MO caused a greater accumulation of unspliced mRNA (data not shown), it also led to an increase in common morpholino side affects, such as developmental delay. We therefore used a 1-3 ng dose of the splicing MO for the experiments reported here to minimize such effects. We also tested a translation-blocking nlk MO, which caused similar phenotypes (data not shown), but it was more toxic to embryos than the splicing MO. Confirming the specificity of the nlk MO phenotype, injection of nlkRNA was able to rescue nlk MO-injected embryos (see Fig. 6).

The principle phenotypes observed in 24 hpf nlk morphants are a disorganization of the head and a moderately short tail(Fig. 5I, compare with wild type in 5C). Notably, nlk morphants did not have the enlarged forebrain and eyes characteristic of wnt8 morphants (see Fig. 5F), suggesting that endogenous nlk does not play a role in posteriorization of the neurectoderm by wnt8, as suggested by the overexpression phenotype. To confirm this, we examined the expression of anterior neural markers during gastrulation in nlk morphants. Inhibition of wnt8 by mutation or morpholino injection results in the posterior expansion of the expression domains of otx2 and opl into the prospective posterior neural domain (Erter et al.,2001; Lekven et al.,2001). Injection of nlk MO had no effect on the expression patterns of these two genes, even when very high doses (10-15 ng)were used, nor did nlk MO injection enhance the phenotype of embryos injected with suboptimal doses of wnt8 MOs (data not shown).

We did detect, however, a significant enhancement of the ventrolateral mesoderm defect observed in wnt8 morphants(Fig. 5). In addition to its role in regulating AP polarity in the neurectoderm, wnt8 is also required to limit the size of the dorsal organizer and for formation of ventrolateral mesoderm, which contributes to somitic mesoderm in the trunk and tail (Erter et al., 2001; Lekven et al., 2001). Both wnt8 MO and nlk MO cause a modest reduction in the expression of tbx6 in the ventrolateral margin, with a small increase in the width of the dorsal domain from which tbx6 is excluded(Fig. 5D,G; compare to wild type in 5A; double-headed arrow indicates region of dorsal margin where tbx6 is absent). At 24 hpf,both nlk MO embryos (Fig. 5I) and wnt8 morphants(Fig. 5F) have a shortened tail compared with wild type (Fig. 5C), indicative of a mild defect in somitic mesoderm formation and consistent with the earlier reduction in tbx6 expression. In situ hybridization at the 7-somite stage with a cocktail of probes containing opl (telencephalon), pax2 (midbrain/hindbrain boundary, otic vesicles, pronephros), krox20 (rhombomeres 3 and 5) and myod(somites) shows that nlk morphants express all of these markers normally [compare the opl domain, marked with a bar, in Fig. 5H with that in wild type in Fig. 5B, and with the expanded domain seen in wnt8 MO-injected embryos(Fig. 5E)]. However, in nlk/wnt8 MO embryos, we observe a significant enhancement of the mesoderm phenotype to a degree of severity that is never seen in wnt8or nlk morphants alone. When both nlk and wnt8function is blocked, tbx6 expression is severely reduced in the ventrolateral margin, with a concomitant expansion of the axial domain lacking tbx6 (Fig. 5J). The severity of the effect on tbx6 expression resembles that seen in embryos homozygous for a chromosomal deficiency that completely removes the wnt8 locus (Lekven et al.,2001). nlk/wnt8 morphants have dramatically reduced trunk and tail mesoderm, and show a dramatically broadened expression of neural markers, reflecting the dorsalization of the embryos due to expansion of organizer fates (Fig. 5K). The most severely affected embryos completely lack the tail, and make only a few trunk somites (Fig. 5L). These data indicate that nlk functions with wnt8 to specify ventrolateral mesoderm and to limit the size of the dorsal organizer.

wnt8 likely functions through lef1 to specify ventrolateral mesoderm, as lef1 morphants also show decreased tbx6 expression during gastrulation(Dorsky et al., 2002). In contrast to tcf3a and tcf3b, lef1 appears to function predominantly as an activator of Wnt target genes(Kengaku et al., 1998; Merrill et al., 2001). We tested whether nlk and lef1 cooperate to activate tbx6 by co-injecting both morpholinos and examining tbx6expression (Fig. 6). When nlk MO is injected at a low dose (0.6 ng), no effect on tbx6is seen at 60% epiboly (Fig. 6B, compare with wild type in 6A). lef1 MO, as previously reported, reduces tbx6 expression(Fig. 6C). When 0.6 ng nlk MO is co-injected with lef1 MO, tbx6 expression is almost completely eliminated (Fig. 6D). It has recently been reported that tcf3a/tcf3bdouble morphants have elevated levels of lef1 transcripts during early gastrulation, suggesting that one role of tcf3a and tcf3b is to negatively regulate lef1 expression(Dorsky et al., 2003). Thus,one explanation for the synergy between nlk and lef1 could be that nlk is required for relieving tcf3-mediated repression of lef1 transcription. To test this hypothesis, we examined lef1 expression in nlk MO-injected embryos. At shield stage, lef1 expression is reduced in nlk morphants compared with wild-type embryos (compare Fig. 6F with 6E). If nlk activates lef1 expression by inhibiting tcf3aand tcf3b, then injection of tcf3a and tcf3b MOs should rescue lef1 expression in nlk morphants. Indeed, we found that nlk/tcf3a/tcf3b MO embryos(Fig. 6H), like tcf3a/tcf3b morphants (Fig. 6G), showed significantly elevated lef1 expression compared with wild type, suggesting that nlk is not directly required for lef1 transcription, but functions indirectly by modulating tcf3a/tcf3b activity. Expression of tcf3a and tcf3bis unaffected in nlk morphants (data not shown), indicating that the interaction between nlk and tcf3a/3b is not due to upregulation of tcf3a/3b transcription in the absence of nlk.

To confirm that these phenotypes were due to loss of nlk function,we attempted to rescue the reduction of tbx6 and lef1expression in nlk morphants by co-injecting nlk RNA. As a control RNA, we co-injected nlk (C387Y) RNA (shown as nlk^* RNA in Fig. 6). nlk RNA significantly rescues the reduction in tbx6 expression (compare Fig. 6K with 6J) induced by nlk MO injection [80% show wild-type levels of expression at 60% epiboly (n=65) versus 29% (n=63)when nlk (C387Y) RNA was co-injected). lef1 RNA was also able to restore normal tbx6 expression in nlk MO embryos(Fig. 6L; 77% show wild-type expression, n=60), further suggesting that the reduction in tbx6 expression in nlk morphants is due to a reduction in lef1. nlk RNA was able to restore normal lef1 expression in a significant fraction of nlk morphants. Although only 27%(n=62) of nlk MO/nlk (C387Y) RNA-injected embryos had wild-type levels of lef1 at 60-70% epiboly (compare Fig. 6N with wild type in 6M),70% (n=43) of nlk MO/nlk RNA-injected embryos(Fig. 6O) had normal levels of lef1. Taken together, these results show that nlk MO specifically interferes with nlk function, and that nlk is required for normal lef1 expression and subsequent ventrolateral mesoderm formation.

Wnt signaling late in gastrulation has been proposed to regulate regional identity within the anterior neural plate, apart from the earlier action of wnt8 ORF1 and ORF2 (Heisenberg et al., 2001; Houart et al.,2002; Houart et al.,1998; Kim et al.,2002; van de Water et al.,2001). Knockdown of either wnt8b or fz8a results in a reduction in the expression of markers in the midbrain and posterior diencephalon, suggesting that these two genes play key roles in promoting development of these regions of the brain(Houart et al., 2002; Kim et al., 2002). We tested whether nlk plays a role in this process by injecting embryos with nlk MO, wnt8b MO, or both morpholinos together, and staining embryos with several neural markers (Fig. 7). In one experiment, we fixed embryos at tailbud stage and probed with a cocktail of opl, pax2 and tbx6 to mark the telencephalon, MHB and margin (Fig. 7A-C), or with en2 to mark the prospective MHB and midbrain (Fig. 7D-F). We observed no changes in opl expression in wnt8b MO(Fig. 7B) or nlk MO(Fig. 7C) embryos, confirming published results for wnt8b MO(Kim et al., 2002) and indicating that wnt8-dependent posteriorization was occurring normally. pax2 expression was normal in wnt8b morphants, but reduced in nlk MO embryos, similar to what is observed in fz8a MO embryos (Kim et al.,2002). en2 is substantially reduced in most wnt8b morphants (Fig. 7E, compare with control in 7D), as previously reported,and is eliminated in most nlk MO embryos(Fig. 7F). When we examined en2 expression at the 7-somite stage, en2 was present but at a much reduced level in wnt8bMO- (76%; n=45; Fig. 7H) or nlk MO(81%; n=42; Fig. 7I)-injected embryos. Co-injection of a high dose (1 ng each) of both morpholinos does not result in a stronger phenotype, but does increase the penetrance somewhat (95%; n=17). When low doses (0.5 ng) are injected, both nlk MO and wnt8bMO still cause a reduction of en2 staining at the 7-somite stage, but at a lower penetrance [48%(n=27) and 51% (n=31), respectively]. Co-injection of both MOs at the lower doses results in a significant increase in the penetrance of the en2 phenotype (91%; n=35), suggesting that nlkand wnt8b are functioning together to regulate en2expression.

Injection of wnt8b morpholinos has also been reported to cause a modest reduction in pax6 expression in the posterior diencephalon during early somitogenesis (Kim et al.,2002). When we examined pax6 expression at the 7-somite stage in wnt8bMO- or nlk MO-injected embryos, we found that the diencephalic domain of pax6 expression was reduced in nlk MO-injected embryos (Fig. 7L, compare with control, 7J), although we did not observe a significant reduction in wnt8bMO(Fig. 7K) embryos. Kim et al.(Kim et al., 2002) reported only a very mild reduction in the pax6 domain in wnt8bmorphants, but fz8a morphants showed a more severe reduction, similar to what we observe in nlk MO embryos. Taken together, our data implicate nlk in regulating AP patterning in the neural plate,possibly in conjunction with wnt8b and fz8a.

The data above indicate that nlk can function with canonical Wnts in regulating aspects of mesoderm and neurectoderm development. Recently published work using mammalian cell culture supports a potential role for Nlk as a downstream effector of non-canonical Wnts such as Wnt5a(Ishitani et al., 2003a). In vertebrates, non-canonical Wnts, together with homologs of Drosophilaplanar cell polarity (PCP) signaling pathway components such as strabismus and prickle, regulate convergence/extension movements during gastrulation (reviewed by Tada et al., 2002; Veeman et al., 2003). As nemo mutants in Drosophila exhibit some defects in PCP signaling, it is possible that the cell culture data describes a conserved role for Nlk proteins in non-canonical Wnt signaling(Choi and Benzer, 1994; Verheyen et al., 2001). Therefore, we attempted to detect genetic interactions between nlkand non-canonical Wnt signaling in zebrafish.

In zebrafish, mutations in two non-canonical Wnts, pipetail(ppt)/wnt5 and silberblick(slb)/wnt11, have been characterized. ppt/wnt5 embryos are defective in tail extension, whereas slb/wnt11 mutations cause partial cyclopia due to a defect in migration of the anterior axial mesoderm, the prechordal plate(Heisenberg and Nusslein-Volhard,1997; Heisenberg et al.,2000; Kilian et al.,2003; Rauch et al.,1997; Ulrich et al.,2003). To test for genetic interactions between nlk and ppt/wnt5, we injected nlk MO into embryos from a cross of ppt heterozygotes. nlk MO/ppt embryos display only an additive phenotype, with a moderately more severe tail defect(data not shown).

By contrast, we observe a strong interaction between nlk and wnt11/slb. We quantified the severity of the phenotype by calculating the Cyclopia Index (CI)(Marlow et al., 1998) (see Fig. 8). Injection of 0.5 ng wnt11 MO causes a mild phenotype, with the large majority of embryos being either unaffected or having slightly more closely set eyes (CI=1.59). nlk MO has no cyclopia phenotype on its own, but strongly enhances the wnt11 MO phenotype, resulting in many embryos with complete eye fusion (CI=3.98). As a control, we saw no effect on the wnt11 MO phenotype upon co-injection of wnt8 ORF1 MO (CI=1.54 versus 1.59 in wnt11 MO alone). We directly analyzed migration of the prechordal plate by fixing embryos at 95-100% epiboly and staining with probes for tbx6 (to mark the margin) and gsc (to mark the prechordal plate). Although all nlk MO (n=45), and most wnt11MO (85%, n=41) embryos showed normal migration of the prechordal plate to near the animal pole (Fig. 8C,D, arrowheads; compare with wild type in B), most nlk/wnt11 MO embryos showed a dramatic defect in migration of the prechordal plate (Fig. 8E). This effect of nlk MO on prechordal plate migration is specific to wnt11, as co-injection of nlk MO does not enhance the mild cyclopia induced by injection of low doses of strabismus/trilobite MO(Fig. 8), nor does it enhance the phenotype of knypek/glypican4 mutants (data not shown), which very rarely exhibit partial cyclopia, but which do strongly enhance the cyclopic phenotype of both wnt11 and strabismus mutants(Henry et al., 2000; Jessen et al., 2002; Park and Moon, 2002; Sepich et al., 2000; Topczewski et al., 2001).

Previous work has indicated that although wnt8 ORF1 and ORF2 function redundantly in mesoderm and neurectoderm patterning, they may possess different signaling capabilities (Lekven et al., 2001). Whereas wnt8 ORF1 seems to signal strictly canonically, in some overexpression assays wnt8 ORF2 behaves similarly to non-canonical Wnts. We therefore tested whether wnt8ORF2 interacted with wnt11. Injection of wnt8 ORF2 MO significantly enhanced the phenotype of wnt11 morphants (CI=2.61),suggesting that wnt8 ORF2 may have a non-canonical signaling role. Like nlk, knockdown of wnt8 ORF2 did not enhance the weak cyclopia in tri/strabismus morphants. Although nlk and wnt8 ORF2 MO both enhance wnt11 loss of function, knocking out nlk and ORF2 together does not reveal any redundancy between the two genes (CI=1.00). Although our data do not definitively place nlkdownstream of a non-canonical Wnt, they clearly indicate that nlkgenetically interacts with wnt11/slb to regulate cell movements during gastrulation, possibly in conjunction with wnt8 ORF2.

Our results support a role for nlk in the activation of Wnt targets during zebrafish embryogenesis. Overexpressed nlkdownregulates two tcf3 homologs, tcf3a and tcf3b,that repress activation of Wnt target genes during neural patterning. This functional interaction with Tcf3 homologs requires wnt8 signaling,and thus probably β-catenin, consistent with previous data indicating that Nlk specifically interferes with the DNA-binding ability ofβ-catenin/Tcf complexes, not that of Tcf alone. Interference with endogenous nlk function reveals important roles in two processes that are regulated by canonical Wnts, mesoderm patterning by wnt8, and patterning of midbrain and forebrain by wnt8b. As loss of nlk enhances or phenocopies loss of function of these two Wnts, we conclude that nlk functions as an activator of some canonical Wnt targets in zebrafish. nlk also interacts, directly or indirectly,with non-canonical Wnt pathways, as inhibition of nlk strongly enhances convergent extension phenotypes associated with loss of wnt11 function. We also uncover a role for an unusual wnt8homolog, wnt8 ORF2, in regulating cell movements during gastrulation.

The interplay between caudalizing wnt8 and rostrally localized tcf3 homologs functioning as inhibitors of Wnt signaling helps establish AP polarity within the neural plate. Overexpression of nlkin zebrafish phenocopies the effects of ectopic wnt8 signaling on AP patterning, and enhances the phenotype associated with loss of tcf3function. In light of the well-established ability of Nlk homologs in other species to phosphorylate Tcfs and inhibit the DNA binding ability ofβ-catenin/Tcf complexes, we suggest that Nlk and β-catenin can function together to derepress genes inhibited by Tcf3a and Tcf3b. This provides the first evidence in vertebrates that Nlk can function as a positive regulator of canonical Wnt target genes.

Mouse Nlk has been shown to directly interact with both mammalian Lef1 and Tcf4, as well as Xenopus Tcf3(Ishitani et al., 2003b; Ishitani et al., 1999; Meneghini et al., 1999). Mutational analysis has demonstrated that Nlk phosphorylates conserved serine and threonine residues on Lef1 and Tcf4, and that these residues are required for Nlk to inhibit the DNA-binding ability of Tcf/β-catenin complexes(Ishitani et al., 2003b). Zebrafish tcf3a, tcf3b and lef1 all possess serine or threonine residues at analogous sites to their mammalian homologs, suggesting that they are all potential substrates for nlk. Interestingly,overexpression of zebrafish nlk only affects AP patterning in the neurectoderm, consistent with an inhibition of only tcf3a and tcf3b, not of lef1. Also, overexpression of nlkdoes not cause ventralization in either zebrafish or Xenopus (J. Waxman, unpublished), suggesting that nlk does not interact with a(as yet unknown) Tcf/Lef protein required for transducing maternalβ-catenin signaling. By contrast, mouse Nlk can block dorsal axis formation in Xenopus (Ishitani et al., 1999), and can block formation of dorsal mesoderm when injected into zebrafish (data not shown), suggesting that it can inhibit maternal β-catenin signaling. Although little is known regarding which regions of Nlk are important for interaction with β-catenin or Tcf/Lef proteins, it is possible that the substantial differences between mammalian and zebrafish nlk over the first 90-100 amino acids could contribute to the differences in phenotypes when overexpressed. In early zebrafish development, nlk seems to interact specifically with two Tcf3 genes known to play a role in the repression of Wnt target genes, but not with lef1, which has been proposed to function primarily as an activator.

Mutational analyses in Drosophila and in mice have been inconclusive in distinguishing whether Nlk functions as an activator or an inhibitor of the Wnt pathway. For example, nemo mutant flies show,among other phenotypes, a variable segment polarity phenotype that can include both excess naked cuticle, indicative of excess wingless(wg)/Wnt signaling, and loss of denticle diversity, which is suggestive of a reduction of wg signaling(Mirkovic et al., 2002; Verheyen et al., 2001). Similarly, the mouse knockout causes a pleiotropic phenotype that may or may not reflect an underlying role in the regulation of Wnt signaling(Kortenjann et al.,2001a).

Our loss-of-function studies support a role for nlk in regulating canonical Wnt signaling. The enhancement of wnt8 and lef1loss-of-function phenotypes suggests that nlk is functioning as an activator of Wnt targets in mesoderm patterning. As nlk is required for lef1 expression in the presence of tcf3, our data support a model in which nlk relieves tcf3-mediated repression of lef1 transcription during gastrulation. β-catenin stabilized by wnt8 could then interact with Lef1 to activate tbx6 and other genes.

Our observation that nlk MO causes defects in en2 and pax6 expression similar to those seen in wnt8b and fz8a morphants suggests that nlk may be functioning in this later AP patterning event. A simple hypothesis is that nlk relieves the repression of wnt8b targets by tcf3a and/or tcf3b. Supporting a possible role for tcf3a in regulating wnt8b signaling, injection of lower doses of tcf3a MO (0.5 ng) or nlk RNA (100 pg) results in embryos with normal pax2expression and an expanded pax6 domain, as would be expected from ectopic wnt8b signaling (Kim et al., 2002) (and data not shown). This phenotype indicates that tcf3a may act to limit wnt8b signaling within the rostral neurectoderm, and suggests that nlk could function in wnt8b-mediated patterning by inhibiting tcf3a.

Given that overexpression of nlk expands posterior neural fates in a wnt8-dependent manner, we were somewhat surprised that we did not observe any affects of nlk MO on the expression of neural genes known to be regulated by wnt8, such as opl. Although we note that a comprehensive search of sequenced mammalian and invertebrate genomes indicates that only one Nlk homolog is present, an ancient gene duplication event in the ancestry of teleost fish has led to the presence of multiple copies of many genes that are present only once in tetrapods, raising the possibility that additional Nlk homolog(s) in zebrafish could be functioning in this process (Kortenjann et al.,2001b; Taylor et al.,2001).

The strong enhancement of the slb/wnt11 phenotype by nlkMO indicates that nlk has an important role in regulating cell movements during gastrulation. Our data do not necessarily place nlkdownstream of a non-canonical Wnt, as has been demonstrated in cell culture with mouse Nlk and Wnt5a. Although components of non-canonical Wnt pathways are clearly important for controlling cell movements, canonical Wnt signaling can also play a role. For example, in zebrafish, maternal β-catenin signaling is known to activate Stat3, which is required for normal convergence/extension movements (Yamashita et al., 2002). Maternal β-catenin signaling has also been shown to regulate gastrulation movements in Xenopus, probably through activation of Xenopus nodal-related 3 (Xnr3)(Kuhl et al., 2001). Thus, the observed interaction between nlk and wnt11 could reflect a defect in the regulation of a canonical Wnt target caused by loss of nlk, particularly in light of the observation that nlkmorphants do not show any obvious convergence-extension defects.

Given the potent affects on cell proliferation and fate specification by Wnts, it is essential to restrict activation of Wnt target genes to only the appropriate time and place within an organism. It is likely, then, that many Wnt targets are actively repressed in the absence of Wnt ligand, and a significant function of Wnts in these contexts is to derepress them,permitting activation by other factors. Nlk is an important component of the mechanism by which some Wnt targets are activated in C. elegans, and this activity appears to be conserved in zebrafish.

We thank Rich Dorsky for providing the lef1 and tcf3bmorpholinos and the lef1 plasmid; Thuy Tran for fish care; members of the Moon Laboratory for comments on this manuscript; and Josh Waxman for Xenopus injections. C.J.T. is an Associate, and R.T.M. an Investigator of the HHMI, which supported this research.