Differential stability of β-catenin along the animal-vegetal axis of the sea urchin embryo mediated by dishevelled

β-Catenin plays multiple, important roles in patterning animal embryos. This protein is an essential component of the highly conserved,canonical Wnt signaling pathway, which has been studied intensively in embryonic and cancer cells (reviewed by Huelsken and Behrens, 2002; Moon et al., 2002). According to the current view of the canonical pathway, Wnt signaling leads to the stabilization of cytosolic β-catenin. In the absence of a Wnt signal,β-catenin is phosphorylated at several N-terminal serine and threonine residues, first by a priming kinase, casein kinase I (CKI), and then by glycogen synthase kinase-3-β (GSK3β). Phosphorylation by GSK3βtargets β-catenin for ubiquitination and proteasome-mediated degradation. Phosphorylation of β-catenin occurs in a multiprotein complex that includes GSK3β, Axin (a scaffolding protein), the tumor suppressor gene product adenomatous polyposis coli protein (APC), and several other proteins. Binding of secreted Wnt ligands to frizzled receptors inhibits the degradation of β-catenin via the activation (possibly by phosphorylation) of dishevelled (Dsh). Activated Dsh inhibits GSK3β by a mechanism that remains unclear but that might involve recruitment of GBP/FRAT1 to the degradation complex (Ferkey and Kimelman,2002; Hino et al.,2003). Increased cytoplasmic pools of β-catenin are thought to lead to accumulation of the protein in the nucleus, where β-catenin interacts with LEF/TCF transcription factors and activates the transcription of Wnt-responsive genes. Although it has sometimes been argued that signaling may not be regulated by total or nuclear levels of β-catenin (e.g. Chan and Struhl, 2002), a large body of evidence from many experimental systems has shown a correlation between levels of nuclear β-catenin and signaling (see discussion in Guger and Gumbiner, 2000; Henderson and Fagotto, 2002; Tolwinski et al., 2003).

Recent studies have demonstrated that β-catenin has a highly conserved function in patterning early metazoan embryos. In all deuterostome embryos that have been closely examined in this regard, including those of amphibians,fish, chicks, ascidians and sea urchins, β-catenin becomes localized to cell nuclei preferentially at one pole of the cleavage-stage embryo(Schneider et al., 1996; Larabell et al., 1997; Rowning et al., 1997; Logan et al., 1999; Roeser et al., 1999; Imai et al., 2000). In these various organisms, nuclear activity of β-catenin, through its interaction with LEF/TCF proteins, is required for early axis specification and the establishment of critical signaling centers in the early embryo(Heasman et al., 1994; Kelly et al., 1995; Wylie et al., 1996; Pelegri and Maischein, 1998; Imai et al., 2000; Kelly et al., 2000). Recent studies with cnidarian embryos suggest that this mechanism of axis specification might be very ancient; i.e. its origins may pre-date the divergence of radially and bilaterally symmetrical metazoans(Wikramanayake et al.,2003).

In the sea urchin, β-catenin is required for the formation of endoderm and mesoderm. Overexpression of proteins that interfere with nuclear localization and/or function of β-catenin, including cadherins,GSK3β and a dominant negative form of TCF/LEF, lead to the development of`dauerblastula' embryos, which lack mesenchyme cells and gut(Emily-Fenouil et al., 1998; Wikramanayake et al., 1998; Logan et al., 1999; Vonica et al., 2000).β-Catenin appears to have multiple functions in endomesoderm specification. In the large micromere-primary mesenchyme cell (PMC) lineage,β-catenin is an essential activator of a network of early zygotic transcription factors, including Pmar1, Ets1, Alx1 and T-brain, that regulate the powerful signaling properties and later morphogenesis of these cells(Kurokawa et al., 1999; Fuchikami et al., 2002;Oliveri et al., 2002, 2003; Sweet et al., 2002; Ettensohn et al., 2003). In the macromeres, β-catenin is required for the activation of genes involved in endoderm and non-skeletogenic mesoderm specification, including those that render the cells responsive to micromere-derived signals(McClay et al., 2000; Davidson et al., 2002).β-Catenin also plays an indirect role in ectoderm specification, through its influence on vegetally derived signals that pattern the overlying ectoderm(Wikramanayake et al.,1998).

Although the regulation of nuclear localization and function ofβ-catenin is a critical feature of early metazoan development, the underlying mechanisms are not well understood. It has been proposed that the differential nuclear accumulation of β-catenin is a consequence of regulated proteolytic degradation along the embryo axis. This hypothesis is based on experiments demonstrating that overexpression of proteins predicted to interfere with or enhance β-catenin degradation lead to corresponding changes in nuclear localization and axis specification (see reviews by Moon and Kimelman, 1998; Sokol, 1999). Although overexpression studies point strongly to differential degradation, there has been no direct demonstration that the stability of β-catenin varies along an early embryo axis. One study compared the half-life of β-catenin on the dorsal and vegetal sides of the cleavage-stage Xenopus embryo using biochemical methods but reported that β-catenin was highly and equally stable (t_1/2=3.6-3.8 hours) on both sides of the embryo (Guger and Gumbiner,2000). Mechanisms other than regulated proteolysis have also been put forward to account for changes in levels of nuclear β-catenin,including regulated nuclear import/export and interactions with cytoplasmic and nuclear anchoring proteins (Henderson and Fagotto, 2002). One recent study, for example, has argued that Wnt signaling increases levels of nuclear β-catenin not by regulating GSK3β-mediated degradation, but by triggering the degradation of a cytoplasmic anchoring protein (Tolwinski et al., 2003).

In the present study, we have exploited the optical transparency of the sea urchin embryo to measure the half-life of β-catenin in specific cell lineages in vivo. We report a gradient in β-catenin stability along the animal-vegetal (A-V) axis during cleavage and show that degradation ofβ-catenin in animal blastomeres is dependent on GSK3β-mediated phosphorylation of the protein. We find that overexpression of a dominant negative form of Dsh blocks nuclear accumulation of β-catenin in vegetal cells and suppresses mesoderm and endoderm formation. Finally, we report that a GFP-tagged form of Dsh shows a striking, specific localization to the vegetal cortex of the fertilized egg and early vegetal blastomeres, where we propose it is locally activated and subsequently functions to stabilizeβ-catenin. Through a detailed mutational analysis of Dsh, we have identified several regions that are required for vegetal targeting, including a short phospholipid-binding motif within the DIX domain of the protein.

Adult Lytechinus variegatus were obtained from the Duke University Marine Laboratory (Beaufort, NC) and Carolina Biological Supply (Burlington,NC). Spawning was induced by intracoelomic injection of 0.5 M KCl. For most experiments, embryos were cultured at 23°C in temperature-controlled waterbaths.

A 600-bp fragment of LvDsh was cloned by RT-PCR from mesenchyme blastula-stage cDNA using degenerate primers. 5′- and 3′-RACE were used to isolate clones containing the remainder of the coding sequence and the 5′ and 3′ untranslated regions. A single cDNA clone encoding a fragment of Axin from Strongylocentrotus purpuratus was identified in an EST sequencing project (Zhu et al.,2001). PCR primers designed from this sequence were used to amplify a fragment of LvAxin from unfertilized egg-stage cDNA. 5′- and 3′-RACE were used to isolate clones containing the remainder of the sequence. LvGSK3β was cloned by RT-PCR using the published sequence of GSK3β from Paracentrotus lividus and primer sequences described by Emily-Fenouil et al. (1998). Sequences of LvDsh, LvAxin, and LvGSK3β have been deposited in GenBank(Accession numbers AY624074, AY624075 and AY624076, respectively).

For mRNA injections, cDNAs were subcloned into pCS2+MT or pCS2+GFP vectors. Injections of in-vitro transcribed, capped mRNAs were carried out as described(Sweet et al., 2002). Immunostaining with anti-β-catenin antibody was performed following the protocol of Logan et al.(1999), except that embryos were fixed for 4 hours. Point mutations in LvDsh were introduced using the Quick-Change Site-Directed Mutagenesis Kit (Stratagene) and deletions were generated by PCR.

Four-dimensional microscopy was carried out using a Bio-Rad MRC-600 laser-scanning microscope equipped with 40× and 60× water immersion objectives. Z-stacks (20-40 images/stack) of 256×256 pixel images were collected every 2 minutes with a step size of 4 μm, using Kalman filtering and the Fast1 scan rate. Acquisition of the stacks was automated using the SOM program (dlapse) accessible through the CoMos software package (Bio-Rad). A two-dimensional projection (maximum-intensity method) was generated from each z-stack and transferred to a Macintosh G4 computer. NIH Image was used to generate time-lapse sequences of the two-dimensional projections. Representative (non-time-lapse) images of embryos expressing various GFP-tagged proteins were obtained by confocal microscopy as described above, except that z-stacks were collected at a slow scan rate with a step size of 1-2 μm and an image size of 512×512 pixels.

To calculate the half-life of GFP-tagged proteins, fertilized eggs were injected with mRNA encoding the protein of interest and allowed to develop for a period of time sufficient for accumulating levels of the GFP-tagged product detectable by confocal microscopy. At various times, the translational inhibitor emetine was added to the cultures (final concentration=100 μM)and the embryos were cultured continuously in the presence of the inhibitor. After 30 minutes, when emetine was maximally effective, time-lapse sequences were collected as described above. Each sequence encompassed 45-60 minutes of real time. To ensure that the intensity of all fluorescence images was below saturation, the CoMos software was used to examine histograms of pixel intensities in the 8-bit images and settings on the MRC-600 were adjusted such that the brightest pixel in the raw images in each sequence had an intensity value of <256. From the maximal-intensity projections of each z-stack, NIH Image was used to measure the mean pixel intensity within selected regions of each frame. Average pixel intensities were measured in specific cell lineages by hand-selecting the cells using a free-form tool. These values represented average pixel intensities over the entire cellular region, i.e. including both nuclear and cytoplasmic pools of protein. Initial average pixel values (30 minutes after addition of emetine) were set to 1 and the natural log (ln) of average pixel intensity was plotted versus time using Microsoft Excel. These plots showed a linear decrease in pixel intensity over time, as expected for an exponential decay (Fig. 2D). A trend line was added to each curve and used to determine the R²value and the equation of the line (y=mx +b). The slope (m) was used in the equation t_1/2=0.693/m to calculate the half-life (t_1/2) of the fluorescence in each cell lineage.

Cleavage-stage embryos were incubated in emetine for 25 minutes and 20μCi/ml ³⁵S-methionine was then added to the medium for 15 minutes. After TCA precipitation, incorporation of the radioactive label was measured with a scintillation counter. Levels of incorporation were plotted relative to sibling embryos not exposed to emetine.

Using the optically transparent embryos of Lytechinus variegatus,we expressed GFP-tagged Xenopus β-catenin(Xl-wt-β-catenin-GFP) and followed the dynamics of protein expression using 4-D confocal laser-scanning microscopy(Fig. 1). Microinjection of in-vitro transcribed, capped mRNA encoding Xl-wt-β-catenin-GFP resulted in detectable protein expression by the 2-cell stage. Initially, the protein was expressed at similar levels in all blastomeres(Fig. 1A); in the cytoplasm and nucleus, and in association with cell membranes. Later in cleavage, levels of GFP fluorescence (both cytoplasmic and nuclear) in animal blastomeres declined dramatically (Fig. 1B-E). Over 1-2 cell cycles, nuclear Xl-wt-β-catenin-GFP became restricted to a region at the vegetal pole of the embryo that included the micromere territory and the vegetal-most portion of the macromere territory. This restricted pattern of Xl-wt-β-catenin-GFP nuclear localization closely matched the normal vegetal nuclear accumulation of β-catenin in L. variegatus embryos, visualized using an antibody against the endogenous protein (Logan et al.,1999).

The precise stage at which the loss of β-catenin-GFP in animal blastomeres was first detectable, and the final boundary of the vegetal domain of nuclear β-catenin-GFP, varied with the amount of mRNA injected. At relatively low doses, clearing could be detected as early as the 16-cell stage, and nuclear β-catenin-GFP became restricted to the micromere territory and small numbers of veg2-derived cells. At higher doses, loss of fluorescence in the animal blastomeres was delayed until the 64-128-cell stages and larger numbers of cells of the veg2 territory retained nuclearβ-catenin. Overexpression of Xl-wt-β-catenin-GFP produced no apparent phenotype, despite the transient nuclear localization of the protein in animal blastomeres.

The loss of β-catenin in animal blastomeres was dependent on GSK3β, a serine-threonine kinase that phosphorylates β-catenin on several N-terminal residues and targets the protein for ubiquitination and degradation. We expressed a GFP-tagged form of Xenopus β-catenin(Xl-pt-β-catenin-GFP) in which four N-terminal serine and threonine residues were converted to alanines. Three of these residues are phosphorylated by GSK3β and one by a priming kinase, casein kinase 1(Yost et al., 1996; Liu et al., 2002). This variant of β-catenin shows increased stability in biochemical assays and in vivo (Yost et al., 1996; Guger and Gumbiner, 2000). Overexpression of Xl-pt-β-catenin-GFP resulted in persistent nuclear localization of the protein in all blastomeres throughout cleavage and blastula stages (Fig. 1F-I). Similarly, coexpression of Xl-wt-β-catenin-GFP and a kinase-dead,dominant negative form of Xenopus GSK3β (Xl-dn-GSK3β)prevented the loss of β-catenin from animal blastomeres(Fig. 1J). Overexpression of Xl-pt-β-catenin-GFP or Xl-dn-GSK3β resulted in a vegetalized phenotype, as previously reported(Emily-Fenouil et al., 1998; Wikramanayake et al., 1998). These studies indicated that the restriction of nuclear β-catenin-GFP to the vegetal region is dependent on GSK3β-mediated degradation of the protein in animal blastomeres.

To measure the half-life of β-catenin in specific cell lineages, we injected mRNA encoding Xl-wt-β-catenin-GFP into fertilized eggs, allowed the protein to accumulate to levels detectable by confocal microscopy, then blocked further protein translation with emetine. Control experiments showed that 100 μM emetine inhibited >90% of new protein synthesis within 20 minutes (Fig. 2E). Levels of fluorescence were quantified over the subsequent 45-60 minutes and the rate of fluorescence decay was used to calculate the half-life of the protein in different blastomeres (Fig. 2). Additional control experiments showed that fluorescence decay was dependent on GSK3β-mediated phosphorylation of β-catenin and that photobleaching of GFP was negligible over the time course of our experiments(Fig. 2F-G).

These measurements provided direct evidence of a gradient in β-catenin stability along the A-V axis during early cleavage. The average half-life ofβ-catenin-GFP in the mesomere, macromere and micromere territories was 0.24, 0.59 and 1.60 hours, respectively(Fig. 2H). In our emetine experiments, we detected differential degradation of β-catenin as early as the 8-cell stage (approximately 2 hours postfertilization), the first stage at which cleavage divisions separated animal from vegetal blastomeres. We also made half-life measurements at the 16- and 64-cell stages and consistently observed a gradient in β-catenin stability at these stages. We found no striking increase or decrease in β-catenin stability within any single cell lineage between the 8- and 64-cell stages, and therefore data from these various cleavage stages were pooled in the table shown in Fig. 2H.

The differential stability of β-catenin along the A-V axis suggested that positive or negative regulators of degradation might be localized (or activated) in the animal or vegetal regions, respectively. Localized degradation of GSK3β has been proposed as a mechanism for regulating nuclear accumulation of β-catenin in the early Xenopus embryo(Dominguez and Green, 2000). We cloned GSK3β from L. variegatus (GenBank Accession number AY624076) and expressed a GFP-tagged form, but half-life measurements showed that the protein was highly and equally stable in all blastomeres(Fig. 3). Dorsal translocation of dishevelled (Dsh) following fertilization has also been proposed to play a role in Xenopus (Miller et al.,1999). We cloned Dsh from L. variegatus (GenBank Accession number AY624074) and found that LvDsh mRNA was ubiquitously expressed in eggs and early embryos (data not shown). We expressed a GFP-tagged form of LvDsh (Lv-wt-Dsh-GFP) and observed a striking, vegetal cortical localization (VCL) of the protein(Fig. 4). At the light microscope level, Lv-wt-Dsh-GFP accumulated in punctate structures associated with the cortical cytoplasm in the vegetal region(Fig. 4F). We also observed punctate localization to the perinuclear region of all cells, including animal blastomeres. In some embryos injected with high concentrations of mRNA, VCL could be detected even before first cleavage(Fig. 4A). Continuous observation of such embryos showed that the first and second cleavage planes bisected the domain of Lv-wt-Dsh-GFP localization. The third cleavage unambiguously identified this region as the vegetal pole. At the 16-cell stage, the VCL domain was inherited by the micromeres and, to a lesser extent,the macromeres. The VCL domain became more difficult to detect at later stages, but could be identified in some embryos as late as the 32-cell stage.

Dsh is a multi-domain protein that interacts with many partners(Penton et al., 2002, Wharton, 2003). We generated deletions and point mutations in LvDsh to determine which domains were required for VCL (Figs 5, 6). Although we did not quantify expression levels of the various protein constructs by immunoblotting, all were expressed at levels that were readily detectable by fluorescence microscopy. It was apparent that vegetal targeting did not correlate with relative levels of expression based on GFP fluorescence, i.e. of some constructs that yielded very bright fluorescence, some showed targeting while others did not, and the same was true of constructs that showed faint fluorescence. In addition, all constructs were injected over at least a 10-fold range of concentration and we did not detect changes in the distribution of GFP-tagged proteins over this range.

A putative SH3-binding motif (see Penton et al., 2002) and the C-terminal region of LvDsh were completely dispensible for VCL(Fig. 5; LvDsh.ΔPR1.GFP and LvDsh.ΔC.GFP, respectively). Deletion of the DEP or PDZ domains[LvDsh.ΔDEP.GFP and LvDsh.ΔPDZ.GFP, respectively; see also LvDsh.Δ(DEP+C).GFP)] only partially suppressed VCL. Deletion of the entire N-terminal region (LvDsh.ΔN.GFP), however, or the DIX domain(LvDshΔDIX.GFP) alone, completely abolished targeting.

Motifs within the DIX domain have been identified that regulate binding to actin and vesicles in cultured mammalian cells. Point mutations in the actin-binding motif (K58A) and vesicle (phospholipid)-binding motif(K68A/E69A) have been shown to abolish the corresponding binding activities without compromising the structural integrity of the DIX domain(Capelluto et al., 2002). These two motifs are well conserved within the DIX domain of LvDsh. A point mutation in the actin-binding domain of LvDsh (LvDsh.K47A.GFP) that corresponded to the K58A mutation described by Capelluto et al.(2002) did not affect VCL, but the corresponding double mutation in the phospholipid-binding domain(LvDsh.K57A/E58A.GFP) completely abolished targeting(Fig. 5; Fig. 6E). Two other regions were identified that were required for targeting: a region between the DIX and PDZ domains that includes multiple phosphorylation sites(LvDsh.DIXΔPDZ.GFP) and amino acid sequences other than the proline-rich region that lie between the PDZ and DEP domains (LvDsh.PDZΔDEP.GFP). The smallest portion of the LvDsh protein sufficient for VCL consisted of approximately the N-terminal half of the protein[LvDsh.Δ(DEP+C).GFP].

To test whether Dsh function was required for the vegetal stabilization ofβ-catenin, we overexpressed the DIX domain alone. This region of Dsh is required for canonical signaling (Kishida et al., 1999; Rothbacher et al., 2000, Penton et al.,2002) and overexpression of the DIX domain phenocopies Dsh null mutations, indicating that it acts as a dominant negative(Axelrod et al., 1998). We found that overexpression of DIX produced an animalized phenotype that was indistinguishable from the phenotype observed following overexpression of cadherin (Wikramanayake et al.,1998; Logan et al.,1999), i.e. lack of endoderm and reduction or complete absence of mesoderm (Fig. 7A). LvDsh DIX also blocked the accumulation of β-catenin in the nuclei of vegetal blastomeres, as shown by immunostaining with an antibody against endogenous Lv-β-catenin (Fig. 7D,E). At these levels of expression, DIX did not block VCL of Lv-wt-Dsh-GFP(Fig. 6F), indicating that it acted by a different mechanism. By contrast to the striking phenotype resulting from DIX overexpression, overexpression of the PDZ domain of LvDsh had no effect on embryo morphology (Fig. 7B). Axin, another protein that regulates β-catenin degradation, also contains a DIX domain. To determine whether the effects of DIX domain overexpression were specific to LvDsh, we cloned LvAxin (GenBank Accession number AY624075) and overexpressed the LvAxin DIX domain. Injection of LvAxin DIX mRNA at concentrations equal to or higher than LvDsh DIX mRNA produced no apparent phenotype (Fig. 7C).

Our in-vivo measurements provide the first direct demonstration of differential stability of β-catenin along an axis of an early metazoan embryo. We have found a gradient in β-catenin half-life along the A-V axis of the sea urchin embryo at early cleavage stages. On average, the protein is approximately 8 times more stable in the most vegetal cells, the micromeres, than in animal blastomeres. This difference in protein stability in the micromere and mesomere territories is sufficient to produce a>100-fold difference in protein levels in 2 hours, approximately the time required for the fertilized egg to reach the 8-cell stage. Averaged over all cell lineages, our estimates of β-catenin half-life in the early sea urchin embryo are consistent with those previously reported based on bulk biochemical analyses of whole Xenopus embryos(Yost et al., 1996) (see also Guger and Gumbiner, 2000),whole Drosophila embryos (Pai et al., 1997) and tissue culture cells(Byers et al., 1996).

Several observations indicate that the turnover of Xenopuswild-type β-catenin-GFP mimics that of the endogenous sea urchin protein. Most significantly, the pattern of nuclear accumulation of Xl-wt-β-catenin-GFP we observed (Fig. 1) closely matches the pattern of nuclear localization of endogenous, sea urchin β-catenin detected by immunostaining(Logan et al., 1999). In addition, several studies have shown that homologous Wnt pathway proteins from vertebrates and sea urchins are functionally interchangeable and therefore likely to be regulated in similar ways. For example, Xenopus and sea urchin forms of cadherins, GSK3β and LEF/TCF have been tested in overexpression studies, and in each case the homologs have similar effects(Emily-Fenouil et al., 1998; Wikramanayake et al., 1998: Logan et al., 1999; Vonica et al., 2000). In cnidarians, much more distant relatives of vertebrates than sea urchins, the Xenopus wild-type β-catenin construct used in this study has been compared to a GFP-tagged form of the endogenous protein and the two have identical dynamics (Wikramanayake et al.,2003).

The findings reported here point to differential proteolysis as the predominant mechanism controlling the polarized nuclear localization ofβ-catenin during cleavage. Nevertheless, other mechanisms might operate in parallel. β-Catenin mRNA is uniformly distributed in the egg and early embryo (Miller and McClay,1997), and differential localization of maternal β-catenin mRNA (or localized transcription of the β-catenin gene) is therefore unlikely to be a contributing factor. We cannot rule out differential translation as a contributing mechanism, however, as our mRNA constructs might lack important translational regulatory elements, and the rate of translation of endogenous β-catenin mRNA in animal and vegetal blastomeres has not been measured. If differential translation contributes to the polarized nuclear accumulation of β-catenin, this effect can be overridden by experimental manipulations that alter post-translational processing of the protein. Thus, endogenous β-catenin can be driven into the nuclei of more animal blastomeres by treating embryos with LiCl, an inhibitor of GSK3β(Logan et al., 1999). Finally,another plausible mechanism is regulated nuclear import and/or export,possibly involving interactions with cytoplasmic or nuclear anchoring proteins(Henderson and Fagotto, 2002; Tolwinski et al., 2003). Our observation that wild-type and hyperstable β-catenin-GFP accumulate in the nuclei of animal blastomeres (the former transiently and the latter persistently) argues strongly against the possibility that animal blastomeres lack factors required for nuclear import or retention. A potential limitation of overexpression studies, however, is that mechanisms that normally regulate nuclear import/export of β-catenin might be overwhelmed. For example,artificially high levels of β-catenin might saturate a cytoplasmic anchoring protein that normally sequesters β-catenin in the cytoplasm of animal blastomeres. Although the differential translation and cytoplasmic sequestration models warrant further study, our findings strongly support the view that differential degradation is a key mechanism regulating the nuclear accumulation of β-catenin during cleavage.

The methods used in the present study do not allow us to determine precisely when the polarity in β-catenin degradation first arises. It is therefore unclear whether differential stability along the A-V axis involves a local activation of degradation in the animal hemisphere, an inhibition of degradation in the vegetal region, or a combination of mechanisms. A difference in β-catenin stability along the A-V axis is clearly established by the 8-cell stage, when animal and vegetal blastomeres are first separated from one another by a horizontal cleavage. The finding that LvDsh-GFP becomes localized at the vegetal pole even prior to first cleavage,however, suggests that the molecular machinery underlying differential degradation is polarized maternally or immediately after fertilization. This observation also indicates that VCL is not dependent on Wnt signals transmitted between blastomeres, consistent with other evidence that cell-cell signaling is not required for vegetal nuclear accumulation of β-catenin in the sea urchin embryo (Logan et al.,1999).

Our finding that endomesoderm specification and nuclear accumulation ofβ-catenin are suppressed by overexpression of the DIX domain of Dsh, a putative dominant negative form of the protein with respect to canonical Wnt signaling (Axelrod et al.,1998), provides the strongest evidence to date that Dsh normally plays an essential role in early deuterostome embryo polarity. It has been proposed that Dsh plays a role in axis specification in Xenopus,based on the finding that Dsh protein becomes concentrated on the dorsal side of the early embryo (Miller et al.,1999) (see below). There are also differences in levels of Dsh phosphorylation along the dorsal-ventral axis(Rothbacher et al., 2000),although their functional significance (and the role of Dsh phosphorylation in general) remains unclear (Wharton,2003). The function of Dsh in early patterning in Xenopushas been controversial because dominant negative approaches have not yet revealed a role for the protein in endogenous axis formation. Dominant negative forms of Dsh that interfere with Wnt-induced secondary axis formation do not suppress the formation of the endogenous dorsal axis(Sokol, 1996). Others have noted possible technical reasons for the failure of these constructs to suppress normal axis formation, however. Levels of expression of dominant negative constructs might be insufficient for competing with maternal pools of protein, particularly if the maternal proteins are already complexed with other molecules (Miller et al.,1999; Rothbacher et al.,2000).

The molecular mechanism of the dominant negative effect of the DIX domain is unknown. This is largely because the mechanism by which Dsh stabilizesβ-catenin has not yet been elucidated. It has been proposed that Dsh acts by binding to Axin, via the DIX domains of the two proteins. This interaction might prevent Axin multimerization (Hsu et al., 1999; Kishida et al.,1999; Sakanaka and Williams,1999) and/or recruit GBP/FRAT-1, an inhibitor of GSK3β, to the degradation complex (Li et al.,1999; Ferkey and Kimelman,2002; Hino et al.,2003). Dsh also forms multimers via the DIX domain, and this might be important for signaling (Kishida et al., 1999; Rothbacher et al.,2000). These observations are consistent with a number of scenarios by which stray DIX domains might disrupt endogenous Dsh-Dsh,Dsh-Axin and/or Axin-Axin interactions. For example, DIX might compete with endogenous Dsh for binding to Axin but be unable to recruit GBP/FRAT-1 to the degradation complex.

Our observations and those of Miller et al.(Miller et al., 1999) point to striking similarities in Dsh localization in early sea urchin and Xenopus embryos. Miller and co-workers reported that in Xenopus, endogenous Dsh and a GFP-tagged form of the protein associate with vesicle-like organelles that translocate from the vegetal pole of the fertilized egg to the future dorsal side during cortical rotation. In L. variegatus, LvDsh-GFP is associated with granular or vesicular structures in the vegetal region and this association is dependent on a vesicle-binding motif within the DIX domain. The major difference appears to be that Dsh is not redistributed after fertilization in sea urchin eggs, which lack a cortical rotation. Although the mutational analysis of Miller et al.(Miller et al., 1999) was not as detailed as that presented here, both found that association of Dsh-GFP with vesicle-like structures was dependent on the DIX domain, but not the C-terminal region of the protein. One difference between the two studies,however, is that Miller et al. reported that deletion of the DEP or PDZ domains eliminated association with vesicles, while we found that deletion of these domains in LvDsh only partially blocked targeting to the vegetal cortex and association with vesicles in that region. Dsh has been found associated with punctate cytoplasmic structures in a variety of cell types(Axelrod et al., 1998; Itoh et al., 2000), although it is not known whether these structures are the same as the granular or vesicular structures observed in eggs and early embryos.

Significantly, in our experiments, we observed no phenotype associated with overexpression of full length LvDsh, either untagged or GFP-tagged forms, even at mRNA concentrations sufficiently high to compromise embryo viability. Others have shown that animal blastomeres can be converted to more vegetal fates by overexpression of kinase-dead GSK3β(Emily-Fenouil et al., 1998), a mechanism that bypasses Dsh. Moreover, in the present study, we showed thatβ-catenin can be driven into the nuclei of animal blastomeres by overexpression of kinase-dead GSK3β or mutation of N-terminal phosphorylation sites in β-catenin. These findings show that disruption of β-catenin degradation downstream of Dsh leads to nuclear accumulation of β-catenin in animal blastomeres and changes in cell fate. Therefore,our finding that overexpression of wild-type LvDsh does not have the same effect strongly suggests that the protein is not active in animal cells. We cannot exclude the possibility that, in our experiments, insufficient levels of LvDsh were expressed in animal blastomeres to produce effects, although GFP-tagging confirmed that the protein was expressed persistently in all cells, including animal blastomeres. An alternative hypothesis, and one that we currently favor, is that Dsh is activated specifically in the vegetal region (Fig. 8). This local activation might involve phosphorylation by a vegetally localized activating kinase or interactions with other vegetal-specific proteins. In light of such a model, the targeting of LvDsh-GFP to the vegetal cortex might be interpreted in one of two ways. First, targeting might be an essential step in Dsh activation. For example, vegetal targeting might bring Dsh into close association with a localized, activating kinase, such as Par-1(Sun et al., 2001). Alternatively, VCL might be a consequence of the vegetal activation of Dsh. In that event, VCL could serve to concentrate the protein to effective levels in vegetal blastomeres, or it might play no functional role. Further studies will be required to identify the mechanisms and function of the vegetal cortical localization of Dsh and the putative, local activation of this protein in the vegetal region of the embryo.

We thank J. Miller and R. Moon for plasmids encoding Xl-wt-β-catenin-GFP and Xl-pt-β-catenin-GFP, D. Kimelman for plasmid encoding Xl-dnGSK3β, and D. McClay for antibody against Lv-β-catenin. C.A.E. was supported by grants from the NSF and NIH and A.W. was supported by grants from the NSF and the Hawaii Community Foundation.