Vg1 is an essential signaling molecule in Xenopusdevelopment

The Tgfβ family member Vg1 was one of the first examples of a specific mRNA localized in eggs (Melton,1987; Rebagliati et al.,1985). During oogenesis, Vg1 mRNA becomes restricted to the vegetal cytoplasm of the oocyte, and is inherited by the most vegetal cells of the embryo (Weeks and Melton,1987). The vegetal cells of the blastula have two important functions: they differentiate into the endoderm of the embryo, and they signal to the adjacent marginal (equatorial) region to form the mesoderm(Dale et al., 1985). Vg1 therefore became a strong candidate to be the mesoderm-inducing signal. However, the in vivo function of Vg1 has proven difficult to assess. Whereas the overexpression of other Tgfβ family members causes mesoderm formation(Jones et al., 1995; Koster et al., 1991; Smith et al., 1990), Vg1 mRNA does not (Dale et al.,1989; Tannahill and Melton,1989). Furthermore, endogenous mature Vg1 protein cannot be detected, and although Vg1 pro-protein can easily be detected from the translation of exogenous mRNA, it is not secreted and cleaved(Dale et al., 1989; Tannahill and Melton, 1989). Gain- and loss-of-function experiments have been performed using chimeric Vg1 proteins with either Bmp (bVg1) or activin (aVg1) pro-domains, both of which are cleaved to produce mature Vg1 (Dale et al., 1993; Joseph and Melton,1998; Kessler and Melton,1995; Thomsen and Melton,1993). However, the possibility remains that such chimeric proteins may not reflect endogenous function, particularly as recent studies have shown that the pro-domains of Tgfβ family members are important determinants of signaling function (Cui et al., 2001; Jones et al.,1996; Le Good et al.,2005). A loss-of-function experiment was performed in which the Bmp pro-domain was linked to a putative dominant-negative form of the Vg1 mature protein (Joseph and Melton,1998). This construct blocked the action of overexpressed bVg1,and caused a loss of dorsal mesodermal and dorsal axial structures, and the loss of expression of the endodermal marker Xlhbox8. In the absence of processing of the full-length Vg1, or of detectable levels of the mature form, these facts have generated a long-standing paradox over the in vivo function of this maternally localized mRNA.

More recently, further complications have arisen with the discovery of VegT, a maternally encoded T-box transcription factor whose mRNA is also localized in the vegetal cells of the embryo(Zhang and King, 1996). Depletion of the maternal stockpile of VegT mRNA abrogates formation of the endoderm and the mesoderm (Zhang et al., 1998). VegT activates the transcription of at least six zygotic Tgfβ family members (Xanthos et al., 2001). If nodal signaling is blocked, then mesoderm induction is blocked (Agius et al.,2000; Kofron et al.,1999). Any in vivo function of Vg1 must therefore be reconciled with these facts.

Here, we analyse the role of maternal Vg1, and find it is required for Smad2 phosphorylation and for early zygotic gene expression, particularly of anterior mesendodermal genes that encode Bmp and Wnt antagonists, chordin,cerberus, noggin and dickkopf. Embryos depleted of Vg1 develop with delayed gastrulation, and a dose-responsive reduction in anterior and dorsal development. Although the original Vg1 transcript does not rescue Vg1-depleted embryos, we report that a second allele is effective.

Oligonucleotides (oligos) were as follows:

• Vg1A (antisense),5′-G^*C^*C^*ATGTAACCTTG^*A^*G^*G-3′,where phosphorothioate-modified residues are indicated by an asterix; and

• Vg1MO (morpholino), 5′-CCACAGTCTCAGCCACACCATACTG-3′.

Total RNA was prepared using the proteinase K method and RNase-free DNase treatment before cDNA synthesis (Zhang et al., 1998). cDNA was synthesized using oligo dT primers. Random hexamer (R6) primers were used where indicated. Real-time RT-PCR was carried out using the Light Cycler System (Roche), using the primers and cycling conditions described previously (Birsoy et al., 2005; Kofron et al.,1999; Kofron et al.,2004; Zhang et al.,1998).

Full-grown oocytes were manually defolliculated and cultured in oocyte culture medium (OCM), as described previously(Xanthos et al., 2001). Oocytes were injected vegetally with the antisense oligo (Vg1A) or the morpholino (Vg1MO) and cultured for 48 hours at 18°C before maturation. Control uninjected oocytes were cultured in the same way in each experiment. All of the oocytes were matured by addition of 2 μM progesterone in OCM and cultured for another 12 hours. Control uninjected and oligo-injected oocytes were then labeled with vital dyes and fertilized using the host transfer technique, as described (Zuck,1998).

For whole-mount in situ hybridization with cerberus and chordin probes, gastrula-stage embryos were fixed in MEMFA for two hours. In situ hybridization was performed as described(Harland, 1991).

For histology, tailbud-stage embryos were fixed in Bouin's fixative for three hours, dehydrated and embedded in paraplast, serially sectioned and stained with Hematoxylin and Eosin.

Wild-type animal caps dissected at mid-blastula stage were incubated with control or Vg1-depleted vegetal masses, from oocytes injected with 6 ng oligo,dissected at mid-blastula stage. After 1.5 hours of co-culture, caps were separated from the vegetal masses and frozen down when sibling wild-type embryos reached stage 11.

Western blots were carried out under reducing conditions, as described(Birsoy et al., 2005). Antibodies used were anti-Vg1 antibody (D5; 1:1000)(Tannahill and Melton, 1989),anti-phospho-Smad1 (Cell Signaling Technology; 1:1000), anti-phospho-Smad2(Cell Signaling Technology; 1:1000) and total Smad2 (BD Transduction Laboratories; 1:500). α-Tubulin (DM1A Neomarkers 1:20,000) was used as a loading control. IPLab Gel software (V. 1.5) was used to quantify protein levels.

To study Vg1 function, we targeted Vg1 mRNA using an antisense oligo (Zuck, 1998), which depletes Vg1 mRNA in a dose-responsive fashion(Fig. 1A), and results in a dose-dependent reduction in Vg1 protein levels(Fig. 1B). After fertilization,the amount of Vg1 mRNA and protein continues to decline through the gastrula stages, as there is no transcription of zygotic Vg1 at this time (Fig. 1A,B) (see also Rebagliati et al., 1985). Reduction of Vg1 protein below 50% arrests development at the gastrula stage(data not shown). A similar delayed gastrulation phenotype results when Vg1 protein is depleted (Fig. 1B)using a morpholino oligo, Vg1MO (Table 1B).

Vg1-depleted embryos develop normally until the gastrula stage, but then have a severe, dose-dependent abnormality in gastrulation when compared with controls, where the timing of blastopore formation is delayed, and the blastopore remains enlarged at the late gastrula stage(Fig. 1C; Table 1A). Whole-mount in situ hybridization of Vg1-depleted embryos shows that cerberus and chordin expression is reduced during gastrulation in a dose-dependent manner (Fig. 1C).

At the neurula and tailbud stages, Vg1-depleted embryos have different degrees of anteroposterior and dorsoventral axis abnormality(Fig. 1C; Table 1C). Typically, 6 ng of antisense oligo causes embryos to develop with a loss of head structures,whereas lower doses result in stunted embryos. Histological sections of tailbud-stage embryos show that all three germ layers are present in Vg1-depleted embryos; however, an absence of the notochord and fusion of somites in the midline are observed, with much reduced and abnormal neural structures.

Development of the early embryo involves the interplay of at least three signaling pathways; the Wnt signaling pathway(Heasman et al., 1994)activates the expression of the target genes siamois and Xnr3; the VegT pathway activates the expression of endodermal genes and nodal-related proteins (Xanthos et al., 2001), and initiates Smad2 phosphorylation(Lee et al., 2001), and the Bmp pathway activates Smad1 phosphorylation(Graff et al., 1996; Lee et al., 2001). In Vg1-depleted embryos, the expression of Xnr3 and siamois is not significantly altered (Fig. 1D), suggesting that Vg1 is not required for maternal Wnt pathway activation. However, Vg1 is required for Smad2 phosphorylation, as Smad2 phosphorylation is reduced in Vg1-depleted embryos at the late-blastula stage(Fig. 1E; repeated in three experiments). At the early and mid-gastrula stage, the amount of Smad2 phosphorylation in Vg1-depleted embryos increases, but does not reach wild-type levels. This correlates with the timing of the onset of expression of the VegT target gene derrière, which occurs normally in both control and Vg1-depleted embryos (data not shown). By contrast, Bmp signaling via phospho-Smad1 is initially normal in Vg1-depleted embryos, but is reproducibly elevated (in three experiments) as gastrulation proceeds(Fig. 1E; arrowhead). We conclude that Vg1 signaling is required for Smad2 phosphorylation at the late-blastula stage, and to prevent excess Bmp signaling at the gastrula stages.

Because Smad2 phosphorylation is essential for mesoderm induction, we tested whether Vg1 depletion reduces the mesoderm-inducing signals released by vegetal cells at the blastula stage by performing Nieuwkoop assays. Vg1-depleted vegetal masses co-cultured with wild-type animal caps have a reduced mesoderm-inducing ability compared with wild-type vegetal masses, as measured by the induction of the general mesodermal markers Xbra and Fgf8, and the anterior endo-mesodermal marker chordin, in wild type animal caps (Fig. 1F).

We reasoned that increased Smad1 phosphorylation in Vg1-depleted embryos may indicate that Vg1 controls the expression of Bmp antagonists. In support of this, the levels of expression of chordin, noggin and cerberus are reproducibly reduced to 20% or less of control levels in Vg1-depleted embryos(in three experiments) at the gastrula stage(Fig. 1G). Also, because these proteins normally restrict the range of activity of Bmp to the posteroventral quadrant (Khokha et al.,2005), their reduction may cause a concomitant increase in the expression of Bmp target genes. We find that the ventrolaterally expressed Bmp-target gene sizzled (Salic et al., 1997) is upregulated in Vg1-depleted embryos at the mid-gastrula stage, whereas the level of Bmp4 mRNA expression is unchanged (Fig. 1G).

To confirm that this phenotype is specifically caused by Vg1 depletion, we attempted to rescue Vg1-depleted embryos by reintroducing Vg1 mRNA prior to fertilization. Previous studies in which Vg1 mRNA was overexpressed used an allele, Vg1(Pro), that contains a proline at position 20 from the N terminus of the sequence (asterisk in Fig. 2A)(Dale et al., 1993; Dohrmann et al., 1996; Tannahill and Melton, 1989). This protein is translated in Xenopus embryos but is very inefficiently processed (Fig. 2B) (Dale et al.,1993; Tannahill and Melton,1989); in addition, it does not have mesoderm-inducing activity,and is unable to rescue Vg1-depleted embryos (data not shown). However, we recently identified a second allele of Vg1, Vg1(Ser)(Birsoy et al., 2005), that is equally represented with Vg1(Pro) in oocyte and gastrulae libraries,and is more efficiently processed than Vg1(Pro)(Fig. 2B). The Xenopus tropicalis and Xenopus borealis Vg1 homologs, also have a serine rather than proline residue at the equivalent position(Fig. 2A). As the antisense oligo Vg1A is complementary to both serine and proline alleles(Fig. 2A), it was expected that it would bind and deplete them both with the same efficiency. We were unable to confirm this, as neither the PCR primers nor the Vg1 antibody can differentiate between the two allelic forms. Unlike Vg1(Pro), Vg1(Ser) mRNA partially rescues the phenotype of Vg1-depleted embryos(Fig. 2C,D; Table 1D,E), Smad2 phosphorylation (Fig. 2C) and the expression of molecular markers (Fig. 2E,F) at the gastrula stage.

Because Vg1(Ser) is able to rescue the phenotype of Vg1-depleted embryos,we reasoned that unlike Vg1(Pro), it should also have an inducing activity when overexpressed in embryos. We find that Vg1(Ser) has some mesoderm-inducing activity in animal caps, whereas Vg1(Pro) does not in the same experiment (data not shown). Fig. 2G shows that when Vg1(Ser) mRNA is overexpressed in the vegetal area, it causes the upregulation of the endodermal and mesodermal zygotic genes chordin, Xnr1, Fgf8 and Xsox17α. We compared the inducing activity of Vg1 with another Tgfβ family member,Xnr5. Xnr5 is a zygotic gene regulated by VegT, and, like Vg1, is restricted in its expression to vegetal cells(Takahashi et al., 2000). In comparison to Xnr5, 200 pg of Vg1 mRNA induces Fgf8and Xnr1 to a similar extent as does 40 pg of Xnr5 mRNA at the early gastrula stage. Interestingly, Xnr5 induces Xsox17α and chordin in mid-gastrula stage embryos more effectively than Vg1.

In the current model of axis formation in Xenopus, Wnt target genes are activated on the dorsal side as a result of the cortical rotation and dorsal concentration of localized maternal Wnt11 mRNA at the early cleavage stage (Tao et al.,2005). Because the targets of Vg1 activity, chordin,cerberus and dickkopf, are all first expressed in the dorsal vegetal quadrant of the early gastrula(Bouwmeester et al., 1996; Glinka et al., 1998; Sasai et al., 1994), we tested whether Vg1 is also concentrated in dorsal compared with ventral cells by hemisecting 32-cell stage embryos into dorsal and ventral halves. Fig. 2H shows that both Vg1 mRNA and protein are more concentrated in the dorsal halves than in the ventral halves at the 32-cell stage. Here, protein and mRNA levels are examined in dorsal and ventral halves taken from the same batch of embryos,and the experiment was repeated three times with the same result. The comparison of oligo dT (dT)- versus random hexamer (R6)-primed cDNA shows that the enrichment of Vg1 mRNA on the dorsal side is not due to differential polyadenylation (data not shown).

Homologs of Xenopus Vg1 are expressed maternally in zebrafish(DVR1) (Helde and Grunwald,1993), and at early embryonic stages in chick (Vg1) and mouse(Gdf1) (Lee, 1990; Shah et al., 1997). Although the role of zebrafish DVR1 has not been established, overexpression experiments suggest that chick Vg1 is important in primitive streak formation in chick (Shah et al., 1997; Skromne and Stern, 2002). By contrast, a loss-of-function mutant of mouse Gdf1 has normal germ layer specification, is viable to E14.5 and has left-right axis abnormalities(Rankin et al., 2000; Wall et al., 2000). In Xenopus, previous loss-of-function studies used a dominant-negative approach (Joseph and Melton,1998). The Vg1 depletion phenotype caused using the antisense oligo approach is less severe than the ventralized phenotype caused by using mutant bVg1 ligands to disrupt Vg1 function(Joseph and Melton, 1998). There are also significant differences in the expression of zygotic genes caused by the two approaches. Here, we see no effect on Xnr3expression and a severe decrease in chordin expression as a result of Vg1 depletion, whereas Joseph and Melton showed increased Xnr3 and chordin expression (Joseph and Melton, 1998). It is unlikely that the differences are due to different degrees of inactivation of Vg1, as, when Vg1 is depleted to below 50% of wild-type levels, embryos arrest at gastrulation. Also, the antisense depletion reduces chordin expression, whereas the dominant-negative approach increases it. It is possible that the mutant ligands used in the dominant-negative study were not specific for Vg1.

The effect of Vg1 depletion is also less severe than that caused by depletion of the localized maternal transcription factor VegT(Zhang et al., 1998). In Vg1-depleted embryos, derrière and Xnr1, and the endodermal genes Xsox17α and Gata5, continue to be expressed, and Smad2 phosphorylation occurs, albeit at a reduced level; these activities are completely abrogated by VegT depletion(Xanthos et al., 2001; Lee et al., 2001). Why does the presence of maternal Vg1 mRNA not alleviate the VegT-depletion phenotype? The likely explanation is that the VegT phenotype is in fact a compound phenotype, as VegT mRNA depletion causes the mis-localization of Vg1 mRNA, and leads to a reduction in the levels of Vg1 protein (Heasman et al.,2001). In support of this, the injection of a VegT morpholino oligo, which acts not by degrading VegT mRNA, but by blocking translation, does not affect Vg1 mRNA localization, and it causes a less severe phenotype than the regular VegT oligo does(Heasman et al., 2001).

Because Vg1 is the only dorsally localized maternally inherited Tgfβprotein, it is likely that it initiates the first activation of Smad2 in the dorsal vegetal quadrant after the mid-blastula transition (MBT)(Lee et al., 2001). Activated Smad2 is known to bind to the maternal transcription factor Foxh1(Chen et al., 1996), whose depletion also causes the loss of anterior structures(Kofron et al., 2004). In this model, the VegT-target Tgfβs, which are synthesized after MBT, constitute the second wave of signaling activity.

This study resolves the long-standing paradox over the in vivo function of Vg1. The original clone, with proline at position 20 of the prodomain, is less efficiently processed than the version of the protein with serine at this position. Overexpressed Vg1(Ser) is secreted(Birsoy et al., 2005), whereas Vg1(Pro) is not secreted (Dale et al.,1989; Tannahill and Melton,1989). The Vg1(Ser) clone is likely to be a second pseudoallele,resulting from the pseudo-tetraploidy of the Xenopus laevis genome,as equal numbers of serine (13 ESTs) and proline (14 ESTs) forms exist in the EST databases, and both X. borealis and X. tropicalissequences have a serine at this position. Other Vg1-related genes in zebrafish, chick and mouse genomes are divergent in this region of the prodomain when compared with Xenopus Vg1. Although Vg1(Ser) is not an efficiently secreted protein when compared with Xnr5(Birsoy et al., 2005), we find that, unlike Vg1(Pro), it does induce mesodermal and endodermal gene expression when the mRNA is injected into the vegetal area of the Xenopus embryo (Fig. 2G). In addition, it has much less inducing activity when overexpressed in animal caps (data not shown), indicating the importance of correct localization in development.

Many thanks to Dr Dan Kessler for supplying the Vg1 antibody and the Vg1(P)plasmid, and for much advice; to Aaron Zorn and Scott Rankin for the cerberus in situ probe; and to Helbert Puck and Mansoor Haque for technical assistance. This work was supported by NICHD RO1 HD33002.