Mixer/Bon and FoxH1/Sur have overlapping and divergent roles in Nodal signaling and mesendoderm induction

Nodal signals induce mesoderm and endoderm and control left-right axis development during vertebrate embryogenesis(Schier, 2003). Nodal signaling is mediated by type I (ALK4, ALK7, TARAM-A) and type II (ActRIIB,ActRIIA) receptor serine/threonine kinases, and requires EGF-CFC proteins as co-receptors. Activation of receptors results in the phosphorylation of regulatory Smad transcription factors such as Smad2, which then associate with Smad4 to translocate into the nucleus. These Smad complexes combine with specific transcription factors to regulate different target genes(Attisano and Wrana, 2000; Hill, 2001; Whitman, 2001; Shi and Massague, 2003). It is generally assumed that the Smad-associated transcription factors determine the specific responses of a cell to a given transforming growth factor β(TGFβ) signal (Hill,2001; Whitman,2001; Shi and Massague,2003). It is unclear, however, how many of these factors are required or sufficient to mediate a particular TGFβ signaling process in vivo. Here, we address this question by analyzing the roles of FoxH1/Sur and Mixer/Bon during Nodal signaling in zebrafish.

Members of the FoxH1 (Fast1, Fast3) and Mix/Bix (Mixer, Milk, Bix3)families are the best characterized partners of phosphorylated Smad2 during embryogenesis (Hill, 2001; Whitman, 2001). FoxH1 proteins are forkhead/winged helix transcription factors that can recruit active Smad complexes to activin responsive elements (AREs) in Xenopus mix.2,xnr1 and other genes (Chen et al.,1996; Watanabe and Whitman,1999; Osada et al.,2000). Mixer and related Mix/Bix proteins are paired-like homeodomain proteins and can recruit active Smad complexes to the distal element (DE) of the Xenopus goosecoid (gsc) promoter(Germain et al., 2000; Randall et al., 2002). The interaction of these transcription factors with the activated Smad complex is mediated through a Smad interaction motif (SID in Fast1, SIM in Mixer)(Chen et al., 1997; Randall et al., 2002).

The in vivo roles of FoxH1 have been analyzed genetically in mouse and zebrafish, and through the use of interference approaches in Xenopus. FoxH1 mutant mice have variable but severe phenotypes, including loss of anterior structures, failure to form the node and its midline derivatives, and defects in definitive endoderm formation(Hoodless et al., 2001; Yamamoto et al., 2001). In contrast to nodal mutants, however, foxH1 mutants develop most mesoderm. Blocking antibodies against Xenopus Fast1 led to defects in mesoderm formation, including the inhibition of the mesodermal marker T/Xbra and the dorsal marker gsc(Watanabe and Whitman, 1999). In addition, Activin-mediated induction of mix2, lim1 and gsc is blocked by anti-Fast1 antiserum(Watanabe and Whitman, 1999). Somewhat conflicting results have been reported when Xenopus Fast1 and Fast3 activity was knocked down using morpholinos: although gastrulation movements were inhibited in these embryos, most marker genes (including gsc, mix2 and lim1) seem to be expressed normally(Howell et al., 2002). Taken together, these results suggest that Xenopus FoxH1 might mediate the activation of gastrulation movements and/or mesoderm induction.

Genetic screens in zebrafish have identified mutations in FoxH1[schmalspur (sur)] (Brand et al., 1996; Schier et al.,1996; Solnica-Krezel et al.,1996; Pogoda et al.,2000; Sirotkin et al.,2000). Mutants that lack zygotic sur activity(Zsur) have variable, relatively mild phenotypes, ranging from randomization of left-right asymmetry but normal early patterning to reduction of prechordal plate and floor plate (Brand et al., 1996; Schier et al.,1996; Solnica-Krezel et al.,1996; Pogoda et al.,2000; Sirotkin et al.,2000). Embryos lacking both maternal and zygotic surfunction (MZsur) can have more severe but variable phenotypes,including reduction of axial midline structures(Pogoda et al., 2000; Sirotkin et al., 2000) (this report). This phenotype is much milder than the one observed upon complete loss of Nodal signaling, which leads to a lack of all endoderm, head and trunk mesoderm, and ventral neuroectoderm(Feldman et al., 1998; Gritsman et al., 1999; Meno et al., 1999; Thisse and Thisse, 1999).

In the case of Mix/Bix genes, misexpression studies in Xenopus have shown that members of this family can induce endoderm development but their individual requirements in this process have not been resolved (Ecochard et al.,1998; Henry and Melton,1998; Lemaire et al.,1998). Supporting a role for Mix/Bix genes in endoderm formation, zebrafish mutants for the mixer-like gene bonnie and clyde (bon) (Chen et al.,1996; Stainier et al.,1996; Alexander et al.,1999; Kikuchi et al.,2000) or embryos lacking bon activity because of morpholino (MO) injection (Kikuchi et al.,2000) have a dramatic reduction of endoderm. Additional phenotypes include cardia bifida and pericardial edema, but mesoderm induction appears largely normal in these embryos (Stainier et al., 1996; Chen et al.,1996; Kikuchi et al.,2000). In contrast to the sur phenotypes, the bon phenotype is fully penetrant and largely invariable.

The role of Mix/Bix genes in Nodal signaling is complicated by the observation that some of these genes are regulated transcriptionally by Nodal signals. For instance, biochemical studies and sequence analysis indicate that Bon can serve as a binding partner of phosphorylated Smad2(Randall et al., 2002),suggesting that Bon is a component of the Nodal signaling pathway. Other studies have emphasized that bon is a transcriptional target of Nodal signaling (Alexander and Stainier,1999). In particular, bon expression is absent or barely detectable at the onset of gastrulation in the absence of Nodal signaling,suggesting that bon is primarily a target of the Nodal signaling pathway rather than a necessary transducer of Nodal signals(Alexander and Stainier, 1999). This raises the question of whether Bon is a Smad-associated component and/or a downstream gene of the Nodal signaling pathway. An additional level of complexity derives from the observation that some members of the Mix/Bix family, such as mouse Mixl1, appear not to interact with phosphorylated Smad2(Germain et al., 2000; Randall et al., 2002) but are involved in processes that are regulated by Nodal signaling. For instance,mouse Mixl1 mutant embryos display complex gastrulation defects and,in chimeras, Mixl1 mutant cells are largely excluded from endoderm and heart (Hart et al.,2002).

Here, we analyze the regulation of bon and sur by Nodal signaling, determine whether bon and sur have overlapping,additive or antagonistic functions, and test whether Nodal signaling is mediated exclusively by bon and sur. We find that sur expression is independent of Nodal signaling, whereas bon is initially expressed in the absence of Nodal signaling but requires Nodal signaling and sur for full and maintained expression. We find that MZsur;bon double mutants and MZsur;bonMO embryos have severe phenotypes not observed upon loss of either bon or sur. Double mutants lack heart,prechordal plate and ventral neuroectoderm, a subset of the phenotypes seen upon complete loss of Nodal signaling. Analysis of Nodal downstream genes indicates that bon and sur have both divergent and overlapping functions in gene regulation, and reveals that some Nodal-dependent genes do not require bon and sur activity. Overall, our study establishes that sur and bon have both independent and overlapping roles as components of the Nodal signaling pathway but do not account for all effects of Nodal signaling during mesendoderm induction.

Embryos were staged as described(Kimmel et al., 1995). The following mutant alleles were used: oep^tz57(Hammerschmidt et al., 1996; Zhang et al., 1998), bon^s9 (Chen et al.,1996; Kikuchi et al.,2000) and sur^m768(Solnica-Krezel et al., 1996; Schier et al., 1996; Sirotkin et al., 2000; Pogoda et al., 2000). Misexpression studies have indicated that all three alleles are complete loss-of-function mutations (Zhang et al.,1998; Kikuchi et al.,2000; Pogoda et al.,2000; Sirotkin et al.,2000). In addition, a very weak antimorphic phenotype has been described for bon^s9(Kikuchi et al., 2000). Therefore, we have corroborated results obtained with bon^s9 by using MOs that block bon. For simplicity and to distinguish zebrafish genes from mouse and frog genes, we use bon and sur throughout the text.

In situ hybridization and preparation of RNA probes were performed as described (Schier et al.,1997). Phosphorylated-Smad2 detection was as described(Mintzer et al., 2001).

Synthetic capped squint RNA was synthesized and injected as described (Chen and Schier,2001). Approximately 3 ng of bon MO(Kikuchi et al., 2000)dissolved in phenol red buffer was injected into the yolk of one- to two-cell-stage wild-type or MZsur embryos.

Fish were genotyped as described (Chen and Schier, 2001). Primers for bon are described in Kikuchi et al. (Kikuchi et al.,2000). Primers for sur are 5′-TCACCTTGACTGCAGAATCGG-3′ [fast 330 f2(Sirotkin et al., 2000)] and 5′-GCCAGGTAAGAGTACGGTGGTTTGGGATAT-3′ (SurDCWTR2). SurDCWTR2, a dCAP (derived cleaved amplified polymorphic sequence) primer(Neff et al., 1998),introduces an EcoRV restriction site into the wild type to give a 205 base pair band but does not introduce this site into sur^m768, resulting in a 235 base pair band. The conditions for genotyping of both bon and sur were: 94°C for 3 minutes (1 cycle), followed by 45 cycles of 94°C for 30 seconds, 55°C for 30 seconds and 72°C for 30 seconds, and finally 72°C for 5 minutes. The amplified sur product was digested with EcoRV and resolved on 2% agarose gels.

Members of the Mixer family have been implicated as both transcriptional downstream genes of Nodal signaling (i.e. transcribed in response to Smad/transcription factor complexes)(Rosa, 1989; Chen et al., 1996; Vize, 1996; Ecochard et al., 1998; Henry and Melton, 1998; Alexander and Stainier, 1999)and components of the Nodal signaling pathway (i.e. as partners of Smads)(Germain et al., 2000; Randall et al., 2002). In the case of bon, previous studies have implied that bontranscription is almost fully dependent on Nodal signaling(Alexander and Stainier, 1999; Kikuchi et al., 2000). In this scenario, bon would initially not be a component of the Nodal signaling pathway but would instead be a downstream target gene of Nodal signaling. By contrast, the ubiquitous maternal and zygotic expression of sur has suggested that its expression is not regulated by Nodal signaling (Pogoda et al.,2000; Sirotkin et al.,2000), similarly to Xenopus FoxH1(Chen et al., 1996; Watanabe and Whitman, 1999). To explore the regulation of bon and sur further, we investigated their expression in MZoep mutants, which lack all Nodal signaling (Gritsman et al.,1999). Expression of sur is not affected by loss of Nodal signaling (Fig. 1B). By contrast, and in support of previous analyses(Alexander and Stainier, 1999),we found that bon expression is abolished or barely detectable at 50%epiboly in MZoep mutants (Fig. 1D). However, we detected weak bon expression at earlier stages (dome and 30% epiboly) in MZoep mutants(Fig. 1I,K). These results indicate that sur expression is independent of Nodal signaling,whereas maintained and full, but not initial and weak, bon expression requires Nodal signaling.

To determine whether sur and/or bon mediate Nodal signaling to enhance bon expression, we analyzed bon and MZsur mutants and MZsur;bon double mutants (see below). We found that bon expression is unaffected in bonmutants (Fig. 1E) and strongly,but not completely, reduced in MZsur and MZsur; bonmutants (Fig. 1F,G). Expression of sur was not affected in any of these mutants (data not shown). Together with previous biochemical studies(Chen et al., 1996; Germain et al., 2000; Randall et al., 2002), these results indicate that Bon and Sur can act as components of the Nodal signaling pathway and that Sur enhances bon expression during early embryogenesis.

The findings that both bon and sur are expressed in the absence of Nodal signaling, that both Mixer and FoxH1 can interact with phosphorylated Smad2, and that bon and sur have different phenotypes suggested that bon and sur might have either additive or overlapping roles during mesendoderm induction and Nodal signaling. To distinguish between these possibilities, we generated embryos that lack both bon and sur activity using two different approaches (Fig. 2). In a genetic approach (Fig. 2A), we generated embryos that lack both maternal and zygotic sur(sur is expressed maternally and zygotically) and also lacked zygotic bon [bon is only expressed zygotically(Kikuchi et al., 2000)]. We crossed bon/+ and sur/+ heterozygous fish to generate bon/+; sur/+ double heterozygotes. These were crossed to sur/sur fish and resulting embryos were injected with wild-type sur mRNA to rescue sur/sur mutants. This allowed us to raise sur/sur;bon/+ fish to adulthood. Intercrosses of these fish result in embryos that lack both maternal and zygotic sur activity (MZsur), and one-quarter of embryos are also homozygous for bon. These MZsur;bon mutants lack all sur and bon activity. In a second approach(Fig. 2B), we used a MO antisense oligonucleotide against bon(Summerton and Weller, 1997; Heasman et al., 2000; Nasevicius and Ekker, 2000). Previous studies have shown that this MO efficiently phenocopies the bon phenotype (Kikuchi et al.,2000). Injection of bonMO into MZsur mutants at the one or two cell stage results in MZsur;bonMO embryos.

MZsur;bon and MZsur;bonMO embryos have indistinguishable phenotypes at 30 hours postfertilization (hpf; Fig. 3). Although bonmutant or bonMO embryos have a severe reduction in endoderm(Fig. 3B,C) and MZsurembryos have mild midline defects (Fig. 3D), MZsur;bon (Fig. 3E) and MZsur;bonMO(Fig. 3F) embryos display additional defects not observed in single mutants: the hatching gland and heart are absent, and the ventral forebrain, eyes and floor plate are dramatically reduced. Importantly, this phenotype is highly penetrant and expressive (Fig. 3). Notably,other mesoderm derivatives such as notochord, somites and blood, which are severely affected in the absence of Nodal signaling, appear largely normal and the central nervous system (CNS) is patterned along the anterior-posterior axis. Morphological abnormalities were further confirmed by in situ hybridization using tissue-specific markers (Figs 4, 5). Lack of the prechordal plate-derived hatching gland is evidenced by absence of anterior-most islet-1 expression domain (Fig. 5D). Lack of myocardial cells is apparent from the absence of the markers cmlc1 and cmlc2(Fig. 4M,P, Fig. 5H). Absence of the ventral CNS is highlighted by loss of shh and nk2.2expression (Fig. 4O,R, Fig. 5L). The anterior-most region of the CNS in double mutants expresses emx1(Fig. 4N,Q), defining this territory as telencephalon.

To determine whether these phenotypes are already apparent during early embryogenesis, we analyzed the expression of different markers during late gastrulation and early somitogenesis (Fig. 6). At 80-90% epiboly, endodermal markers (axial,casanova and sox17) are absent or severely reduced in double mutant embryos (Fig. 6H,L,P). The average number of sox17 positive cells was 10 (n=4), in contrast to 24 in bon mutant embryos (n=5). In addition, the prechordal plate marker gsc is not expressed in double mutants at the end of gastrulation (Fig. 6D). Similarly, the anterior expression domain of axial, corresponding to the prechordal plate, is absent in double mutants(Fig. 6H). Remaining midline expression of axial and expression of the notochord marker ntl is compressed along the anterior-posterior axis and broader along the dorsal-ventral axis (Fig. 6H,T), suggesting a defect or delay in dorsal convergence. These results reveal essential, overlapping roles for bon and surin prechordal plate, heart and ventral CNS formation.

The MZsur;bon phenotype only affects a subset of structures dependent on Nodal signaling. This suggested the hypothesis that Sur and Bon regulate only a specific subset of Nodal-dependent genes (e.g. genes involved in prechordal plate or endoderm formation) but that other genes are not affected by Bon and Sur (e.g. genes involved in mesoderm formation, notochord formation or cell internalization). Alternatively, Bon and Sur might affect the expression of many Nodal-dependent genes but to different extents depending on the level of Nodal signaling required for expression. This would result in the loss of structures that require high levels of Nodal signaling. To test these possibilities, we investigated the regulation of different Nodal target genes in bon, MZsur and MZsur;bon embryos at 50% epiboly, the onset of gastrulation(Fig. 7). We analyzed the expression of goosecoid (gsc) (a marker of prechordal plate progenitors; Fig. 7A-D), no tail (ntl) (a pan-mesodermal marker; data not shown), bhikhari (bik) (a retroelement containing FoxH1 binding sites; Fig. 7E-H), mezzo (a mix-like gene implicated in mesendoderm formation; Fig. 7I-L)(Poulain and Lepage, 2002), bon (Fig. 1), casanova (cas; a marker of endoderm progenitors)(Dickmeis et al., 2001; Kikuchi et al., 2001; Sakaguchi et al., 2001)(Fig. 7M-P), snail1 (a marker of internalizing mesendoderm; Fig. 7Q-T) and floating head (flh; a marker of notochord progenitors; Fig. 7U-X).

These genes show four distinct classes of response: (1) ntl,snail1 and flh are expressed normally in bon,MZsur and MZsur;bon embryos; (2) bik, mezzo and bon are expressed in bon but strongly downregulated in MZsur (Poulain and Lepage,2002) and MZsur;bon embryos; (3) cas expression in endoderm precursors is slightly reduced in MZsur and strongly impaired in bon (Kikuchi et al.,2001) and MZsur;bon embryos; (4) gsc is weakly reduced in bon mutants, strongly reduced in MZsur mutants(Pogoda et al., 2000; Sirotkin et al., 2000) and undetectable in MZsur;bon mutants. As described above, gscexpression in MZsur mutants recovers at later stages in a bon-dependent manner (Fig. 6C). These results establish a differential dependence of different Nodal target genes on Bon and/or Sur and reveal a correlation between the lack of a given cell type and the regulation of genes marking specific progenitors at the onset of gastrulation.

As an additional test of the requirement for Bon and Sur in mediating Nodal signaling, we determined the response of ntl, cas, gsc and bik to the ectopic activation of the Nodal signaling pathway(Fig. 8). RNA for the Nodal signal Squint was injected at the one- to two-cell stage and gene response was assayed at 50% epiboly. The expression of ntl was induced in bon, MZsur and MZsur;bon embryos (data not shown). The expression of cas was induced in MZsur but not in bon or MZsur;bon embryos(Fig. 8E-H). The expression of bik was induced in bon but not in MZsur embryos(Fig. 8I-L). Surprisingly, the expression of bik was weakly induced in MZsur;bon and MZsur;bon/+ (data not shown) embryos, suggesting that bonmight act as a repressor of bik at high levels of Nodal ligand and in the absence of sur. Finally, the expression of gsc was not induced in MZsur;bon mutants but was activated in bon and MZsur mutant embryos (Fig. 8A-D). These results provide further evidence for independent and overlapping functions of Bon and Sur in the regulation of Nodal downstream genes.

In the analysis of regulatory networks, it is often difficult to distinguish between direct and indirect regulatory effects. For instance, in the case of FoxH1, many genes have been suggested to be directly regulated,based on overexpression assays in the presence of cycloheximide(Watanabe and Whitman, 1999). However, it has also been proposed that the major role of FoxH1 in zebrafish might be in the autoregulation of Nodal signaling and not necessarily in the regulation of downstream genes (Pogoda et al., 2000). This latter scenario postulates that Bon and Sur regulate Nodal signals or other components of the Nodal signaling pathway to allow full Nodal signaling. The activation of other downstream genes would then be mediated by other components in the Nodal signaling pathway. Therefore, we directly tested the extent of Nodal signaling activity by assaying Smad2 phosphorylation (Fig. 9). In the absence of Nodal signaling (MZoep mutants or lefty overexpression), no phosphorylated Smad2 is detectable (data not shown). Although MZsur mutants initially have less phosphorylated Smad2 than wild-type embryos, phosphorylated Smad2 levels have recovered by the shield stage (Fig. 9). Loss of bon does not influence phosphorylated Smad2 levels. Importantly,despite the much more severe phenotype of MZsur;bonMO embryos, no difference in phosphorylated Smad2 levels is observed compared with MZsur. These results indicate that the stronger phenotype of MZsur;bonMO embryos compared with MZsur mutants is not caused by a decrease in overall Nodal signaling activity.

Previous studies have shown that members of both the Mixer and the FoxH1 families can associate with phosphorylated Smad2 to confer recognition of specific cis elements (Chen et al.,1996; Germain et al.,2000; Hill, 2001; Whitman, 2001; Randall et al., 2002). Additional studies have indicated that some Mixer-like genes are transcriptionally regulated by Activin/Nodal signaling(Rosa, 1989; Vize, 1996; Chen et al., 1996; Ecochard et al., 1998; Henry and Melton, 1998; Alexander and Stainier, 1999). In particular, genetic studies in zebrafish have led to the view that bon is predominantly a transcriptional target of Nodal signaling, not a component of the pathway (Alexander and Stainier, 1999; Kikuchi et al., 2000; Poulain and Lepage,2002). Our studies clarify the regulatory interactions between Nodal signaling, Sur and Bon. Our results, together with biochemical studies(Chen et al., 1996; Watanabe and Whitman, 1999; Osada et al., 2000; Germain et al., 2000; Randall et al., 2002), suggest that Sur is a component of the Nodal signaling pathway, whereas Bon is both a component and a transcriptional target of Nodal signaling(Fig. 1). In particular, we find that there is an early Nodal-signaling-independent, albeit reduced,expression of bon at the blastula margin. It is not clear how this activation is achieved. Studies in Xenopus have demonstrated that the transcription factor VegT can act as a maternal vegetal determinant and activate mix-like genes (Yasuo and Lemaire, 1999). However, functional homologs of VegT have not been identified in zebrafish. Embryological experiments by Chen and Kimelman(Chen and Kimelman, 2000) have led to the suggestion that a secreted factor derived from the extraembryonic yolk syncytial layer induces gene expression at the margin, independently of Nodal signaling. It is thus conceivable that this unknown signal also induces bon expression in the absence of Nodal signaling.

Although not required for the initiation of bon expression, Nodal signaling is essential for normal bon expression(Alexander and Stainier, 1999)(Fig. 1). At the onset of gastrulation, bon expression is lost or barely detectable in the absence of Nodal signaling. The enhancement and maintenance of bonexpression is in part mediated by sur, because MZsur mutants display reduced bon expression. The downregulation of bonmight explain the reduced expression of the endodermal markers axialand sox17 in MZsur mutants(Pogoda et al., 2000; Sirotkin et al., 2000)(Fig. 6), because bonis required for axial and sox17 expression(Alexander and Stainier, 1999; Kikuchi et al., 2000). The low levels of bon in MZsur embryos are apparently sufficient for many processes that are disrupted upon complete loss of both sur and bon in MZsur;bon embryos. For instance, cardiac mesoderm,endoderm, prechordal plate and ventral neuroectoderm form in MZsurembryos despite the lower levels of bon. It is conceivable that the phenotypic variability observed in MZsur mutants(Pogoda et al., 2000; Sirotkin et al., 2000)(Fig. 3) is in part caused by slightly varying levels of bon in these mutants. Reduction of bon expression is not as severe in MZsur or MZsur;bon mutants as in MZoep mutants(Fig. 1), indicating that factors other than Sur and Bon are also involved in Nodal signaling to enhance bon expression (see below). In contrast to bon, surexpression is not affected by loss of Nodal signaling. Taken together, these results indicate that both Sur and Bon are initially expressed in responsive cells independently of Nodal signaling and can thus serve as components of the Nodal signaling pathway. Nodal signaling, in part mediated by Sur but not Bon,then further enhances and maintains bon expression, allowing efficient activation of Bon target genes.

The finding that Bon can associate with phosphorylated Smad2(Randall et al., 2002) and is initially expressed independently of Nodal signaling also offers an explanation for the finding that Bon is not able to activate Nodal target genes such as cas in the animal region of the blastula(Kikuchi et al., 2000). We suggest that Bon is only active upon association with phosphorylated Smad2,and this association is Nodal dependent. In turn, Bon might restrict the expression domain of some targets of Nodal signaling, because bon is expressed only in cells at the margin. For example, cas and sox17 are only expressed in the domain where high levels of Nodal signaling overlap with and induce bon expression. Ectopic expression of bon extends the territory of cas and sox17expression, but this domain is still within the normal range of Nodal signals(Kikuchi et al., 2000; Chen and Schier, 2001). Interestingly, these observations are reminiscent of the Dorsal-dependent regulation of a subset of target genes in the Drosophila embryo. High levels of Dorsal induce expression of the transcription factor Twist in the ventral-most region of the embryo. Dorsal and Twist then act together to activate a group of ventrally expressed target genes such as snail(Ip et al., 1992; Stathopoulos and Levine,2002). Analogously, phosphorylated Smad2 might activate bon in the margin region of the zebrafish embryo. Phosphorylated Smad2 and Bon would then associate in marginal-most cells where phosphorylated Smad2 levels are high and specifically regulate vegetally expressed target genes. Hence, both Dorsal and phosphorylated Smad2 appear to induce transcriptional activators to regulate a specific set of target genes. It is tempting to speculate that this strategy is a general mechanism to translate the graded activity of a transcription factor into discrete downstream responses.

Although a plethora of factors has been identified that interact with regulatory Smads, an in vivo requirement for these factors during vertebrate development has been established in only a few cases(Brand et al., 1996; Chen et al., 1996; Schier et al., 1996; Solnica-Krezel et al., 1996; Stainier et al., 1996; Kikuchi et al., 2000; Pogoda et al., 2000; Sirotkin et al., 2000; Hoodless et al., 2001; Yamamoto et al., 2001). Our double mutant analysis now provides evidence that partners of regulatory Smads have overlapping roles in vivo (Figs 3, 4, 5, 6, 7, 8). Formation of heart,prechordal plate and ventral neuroectoderm are only mildly affected in bon or MZsur embryos but are severely disrupted in embryos lacking both sur and bon activity. Moreover, although the penetrance and expressivity of MZsur mutants are variable, loss of sur and bon leads to fully penetrant and expressive phenotypes.

The wider roles for bon revealed in MZsur;bon mutants are also supported by the phenotypes of embryos lacking both bon and mezzo (Poulain and Lepage,2002) or bon and spadetail(Griffin and Kimelman, 2002)activity. Loss of the T-box transcription factor spadetail and bon results in loss of myocardium, indicating overlapping roles of these genes during cardiac development(Griffin and Kimelman, 2002). Mezzo is another member of the Mix family but, in contrast to Bon, does not contain phosphorylated Smad interaction motifs and thus appears to act exclusively downstream of Nodal signaling(Poulain and Lepage, 2002). Loss of mezzo and bon results in heart and prechordal plate defects, suggesting overlapping roles of these two genes in the formation of these structures (Poulain and Lepage,2002). Although removal of Mezzo enhances the bonphenotype, we have found no enhancement of MZsur;bon embryos upon depletion of Mezzo (S.Z. and A.F.S., unpublished).

The requirements for sur and bon are already reflected before gastrulation in the regulation of downstream genes (Figs 7, 8). We found Nodal-regulated genes whose expression requires bon but not sur(cas), sur but not bon (bhikhari, bon,mezzo), bon or sur (gsc), or neither bon nor sur (flh, ntl, snail1). It is as yet unclear whether all these genes are directly regulated by Nodal signaling, but studies in Xenopus indicate that at least some of these genes might be direct targets. Experiments involving cycloheximide and/or VP16 fusion constructs suggest that Mixer-like proteins can directly regulate gsc(Germain et al., 2000) and FoxH1 can directly activate mix.2, Xbra, lim-1 and gsc(Chen et al., 1996; Watanabe and Whitman, 1999; Osada et al., 2000). Similarly, zebrafish cas, mezzo and ntl appear to be directly regulated by Nodal signaling(Poulain and Lepage, 2002). Moreover, bik elements contain binding sites for FoxH1(Vogel and Gerster, 1999) and the zebrafish gsc promoter contains sequences resembling Mixer binding sites (McKendry et al.,1998). These observations suggest that Nodal signaling leads to the activation of genes regulated by Bon or Sur, Bon and Sur, or neither Bon nor Sur.

The use of different transcription factors, such as Sur and Bon,associating with phosphorylated Smad2 allows Nodal signaling to diverge downstream of receptor activation. For instance, and as outlined above, the restricted expression of bon might contribute to the restricted expression of Nodal-regulated genes implicated in endoderm formation. Indeed,we might speculate that, during evolution, specific genes have come under the control of Nodal signaling by the phosphorylated-Smad2-mediated recruitment of different transcription factors. In this scenario, subsets of genes were initially regulated by transcription factors independently of phosphorylated Smad2. Interaction with and eventual dependence on phosphorylated Smad2 would then usurp these factors into the Nodal signaling pathway. Intriguingly, some members of the mix family are independent of phosphorylated Smad2,whereas others interact with phosphorylated Smad2(Rosa, 1989; Vize, 1996; Chen et al., 1996; Ecochard et al., 1998; Henry and Melton, 1998; Alexander and Stainier, 1999; Germain et al., 2000; Hill, 2001; Whitman, 2001; Randall et al., 2002). Moreover, FoxH1-VP16 fusion proteins can regulate Nodal targets in the absence of Nodal signaling (Watanabe and Whitman,1999; Pogoda et al.,2000). It is thus conceivable that ancestral Mixer- and FoxH1-like proteins were active independently of phosphorylated Smad2 and have only recently been recruited into the Nodal signaling pathway. Support for this model is also provided by the observation that Forkhead transcription factors,but not Activin/Nodal signals, are involved in endoderm formation in Caenorhabditis elegans and Drosophila(Gaudet and Mango, 2002; Stainier, 2002).

It has been speculated that Sur might exclusively enhance the expression of Nodal signals (Pogoda et al.,2000). This suggestion was based on the observation that the expression of the Nodal genes cyclops and squint appears to be downregulated in MZsur mutants. In apparent contradiction to this,however, a Sur-VP16 fusion can rescue aspects of the MZoep mutant phenotype (Pogoda et al.,2000). These mutants are unable to transmit Nodal signals(Gritsman et al., 1999) and so Sur-VP16-mediated activation of cyclops and squint would not have any effect, indicating that, in this context, Sur must regulate other genes to rescue MZoep mutants. Our results also indicate that a purely autoregulatory role of Sur is unlikely. In particular, we find that phosphorylated-Smad2 levels are reduced at dome but not shield stage in MZsur embryos (Fig. 9). These observations are consistent with results in Xenopus, in which a FoxH1-Engrailed fusion construct has no effect on overall phosphorylated Smad2 levels during gastrulation(Lee et al., 2001). It is possible that the effects on phosphorylated Smad2 levels are relatively minor,because FoxH1 does not only regulate the expression of Nodal ligands but also feedback inhibitors of Nodal signaling, such as Lefty and Cerberus(Whitman, 2001; Hamada et al., 2002; Schier, 2003). Therefore, the net effect of elimination of FoxH1 on phosphorylated Smad2 levels might be quite limited (Fig. 9). In addition, no change in phosphorylated Smad2 levels is seen upon blocking bon in wild-type or MZsur mutants. This indicates that the much more severe phenotype of MZsur;bon mutants is unlikely to be due to the reduced activity of cyclops, squint or other components of the Nodal signaling pathway upstream of Smad2 phosphorylation.

Previous studies have identified several transcription factors that interact with Smad proteins to regulate the expression of specific genes(Massague and Wotton, 2000; Whitman, 2001; Hill, 2001). It is unclear how many of these factors are required or sufficient to mediate a particular TGFβ signaling process in vivo. We find that the defects in morphology and gene regulation observed in MZsur;bon double mutants represent only a subset of the phenotypes observed upon complete block of Nodal signaling. In particular, Nodal mutants lack all trunk mesoderm, including blood, pronephros, somites and notochord, and display disrupted expression of genes such as snail1, flh and ntl(Feldman et al., 1998; Gritsman et al., 1999). These defects are not observed in MZsur;bon double mutants, establishing that Bon and Sur cannot account for all Nodal signaling during mesendoderm induction. The p53 tumor suppressor has recently been implicated in the regulation of a subset of Nodal target genes(Cordenonsi et al., 2003). However, blocking p53 in wild type does not lead to mesendoderm defects in zebrafish (Langheinrich et al.,2002) and depletion of p53 in MZsur;bon embryos does not enhance the phenotype (J.T.B. and A.F.S., unpublished). Our results thus indicate that at least one additional Smad-associated transcription factor remains to be identified as a component of the Nodal signaling pathway.

We thank members of the Schier and Yelon laboratories for discussions, D. Yelon, W. Talbot, K. Joubin, B. Ciruna and H. Knaut for comments on the manuscript, and T. Bruno and N. Dillon for fish care. P.S.K. thanks R. Lehmann for generous support. Y.C. was the Rebecca Ridley Kry Fellow of the Cancer Research Fund of the Damon Runyon-Walter Winchell Foundation. M.W. is supported by grants from the NICHD. A.F.S. is a Scholar of the McKnight Endowment Fund for Neuroscience, a Irma T. Hirschl Trust Career Scientist and an Established Investigator of the American Heart Association, and is supported by grants from the NIH.