The Mix family homeodomain gene bonnie and clyde functions with other components of the Nodal signaling pathway to regulate neural patterning in zebrafish

The establishment of cell fates along the anteroposterior (AP) axis of the neural plate is modulated by multiple signaling pathways, including the Wnt,Bmp and Nodal pathways (reviewed by Yamaguchi 2001; Thisse et al., 2000; Erter et al., 2001; Kudoh et al., 2002). The Nodal signaling pathway has been most extensively studied for its role in the formation and patterning of the mesoderm and endoderm (reviewed by Schier and Shen, 1999). Studies in amphibians, mice and zebrafish all point to Nodal ligands as potent inducers of mesodermal and endodermal cell fates(Conlon et al., 1994; Feldman et al., 1998; Sampath et al., 1998; Osada and Wright, 1999). In patterning the neuroectoderm, Nodal signaling has been suggested to specify anterior fates, as mouse chimeras with Nodal-mutant cells in the visceral endoderm lack anterior fates(Brennan et al., 2001). In addition, analyses of a hypomorphic nodal allele reveal that reduced levels of Nodal function result in anterior patterning defects in mouse(Lowe et al., 2001). However,it is unclear how a reduction in Nodal signaling leads to neural patterning defects.

Nodals belong to the Tgfβ superfamily of ligands that bind to and activate heteromeric type I and type II Activin-like receptors (reviewed by Whitman, 2001). The founding member of this Tgfβ subgroup, mouse Nodal, was identified from studying a retroviral insertion that affects node formation(Zhou et al., 1993). In zebrafish, two nodal-related genes, cyclops (cyc)and squint (sqt), are required for the induction of the axial and trunk mesoderm, as well as the endoderm(Feldman et al., 1998; Sampath et al., 1998). Nodal signaling also appears to be important for neural patterning, as embryos mutant for both cyc and sqt appear to have expanded anterior neural fates and loss of trunk spinal cord(Feldman et al., 2000). Additionally, in maternal-zygotic one-eyed pinhead(MZoep)-mutant embryos, which lack an EGF-CFC cofactor essential for Nodal signaling, anterior fates appear expanded(Gritsman et al., 1999). However, compound mutant analyses of embryos lacking sqt and bozozok (boz), a homeobox gene required for axis formation,indicate that sqt acts in parallel with boz to specify anterior neuroectoderm, whereas cyc represses anterior neural development (Sirotkin et al.,2000,Sirotkin et al.,2000). These data suggest that Nodal signaling can play both positive and negative roles in neuroectoderm patterning, and that the correct balance needs to be achieved for the process to occur correctly.

Loss- and gain-of-function analyses indicate that Nodal signaling is transduced by Smad2 (Madh2 – Zebrafish Information Network), and to some extent Smad3 (Madh3a – Zebrafish Information Network). These receptor-activated Smads are phosphorylated by ligand binding to the receptor complex (Waldrip et al., 1998;Tremblay et al., 2000; Brennan et al.,2001). Mouse Smad2 mutants, like Nodal mutants,exhibit defects in the formation of the primitive streak, mesoderm and endoderm (Waldrip et al.,1998; Weinstein et al., 1998). Interestingly, Nodal;Smad2transheterozygous embryos exhibit anterior neural truncations, further suggesting that precise levels of Nodal signaling are required for neuroectoderm patterning (Nomura and Li,1998). Upon activation, the receptor-activated Smads form a complex with Smad4 and translocate to the nucleus. Here, the Smad complex is recruited to Nodal target genes by its interaction with other DNA-binding proteins to regulate gene expression(Derynck et al., 1998; Whitman, 1998).

The first DNA-binding cofactor identified to interact with the Smad complex is the winged helix transcription factor, Foxh1 (also known as Fast1). Smad2 and Smad4 were shown to form a complex with Foxh1, and to bind to an activin-responsive element in the Xenopus Mix.2 promoter(Chen et al., 1996; Chen et al., 1997). Cloning and mutational analysis of the schmalspur (sur) locus in zebrafish demonstrated that sur encodes Foxh1 and that it is required for the maintenance of Nodal signaling(Pogoda et al., 2000; Sirotkin et al.,2000,Sirotkin et al.,2000). Consistent with this model, embryos lacking both maternal and zygotic sur (MZsur) show defects in axial mesoderm,although they do not exhibit the defects in endoderm and trunk mesoderm formation seen in embryos lacking the Nodal ligands Cyc and Sqt(Feldman et al., 1998). These data have led to the proposal that multiple transcription factors can mediate Nodal signaling in various developmental processes(Pogoda et al., 2000; Stemple, 2000).

Biochemical studies have shown that members of the Mix family of homeodomain proteins also function as transcriptional mediators of Nodal signaling (Germain et al.,2000), for example, by interacting with a Smad2/Smad4 complex upon Tgfβ signaling and binding the goosecoid (gsc)promoter. Mapping of the protein-protein interaction domain identified a common Smad interaction motif within a subgroup of the Mix family members, as well as in winged helix transcription factors, such as Foxh1(Germain et al., 2000).

In zebrafish, the Mix gene bonnie and clyde (bon)functions downstream of Nodal signaling to regulate endoderm formation(Kikuchi et al., 2000). bon expression requires Nodal signaling as it is absent in cyc^–/–;sqt^–/–embryos (Alexander and Stainier,1999). Additionally, misexpression of a constitutively active form of the type I Tgfβ receptor Tarama promotes ectopic bonexpression (Alexander and Stainier,1999). Furthermore, bon overexpression in cyc^–/–;sqt^–/–embryos can induce endodermal gene expression(Kikuchi et al., 2000). Finally, bon^–/– embryos exhibit a severe reduction in the number of endodermal precursors, which indicates that bon plays a crucial role in endoderm formation. Here, we show that Bon also functions in precursors of the axial mesoderm to modulate anterior neural patterning. We further show that Bon functions cooperatively with the Nodal signaling components Sqt and Sur (Foxh1) to regulate this process. Expression analyses in single- and double-mutant embryos show a correlation between the severity of the neural patterning defects and the level of dickkopf 1 (dkk1) expression. The defect in dkk1expression in the mutant embryos is part of an overall defect in dorsal mesendoderm gene expression.

Adult fish and embryos were maintained as described(Westerfield, 1994). Embryos were derived from mating of identified heterozyotes, homozygotes or transheterozygotes. The following mutant alleles were used: bon^m425 (Stainier et al., 1996), sqt^cz35(Feldman et al., 1998) and sur^m768 (Schier et al., 1996). Homozygous sur mutant adults were generated from sur^m768/+ intercrosses.

For restricted morpholino injection experiments, fluorescein-tagged morpholino oligonucleotides for bon(5′-GAT-TCG-CAT-TGT-GCT-GCT-GTC-CTT-C-3′) were dissolved in 5 mM HEPES, pH 7.6, and diluted to 2 ng/nl with 5 mM HEPES/10% Phenol Red. Rhodamine-dextran (10 kDa, 2.5%) was co-injected into some embryos in order to enhance the signal for localizing the morpholino. Antibody staining for the fluorescein-tagged morpholino indicated that the 10 kDa rhodamine-dextran co-localizes with the morpholino (data not shown). Single cells at the 32-cell stage were injected with 1 nl of a 2 ng/nl bon MO stock. Following injections, embryos were fixed for whole-mount in situ hybridization at the tailbud stage, or photographed using a Zeiss Axioplan microscope. Localization of the injected clone was visualized with a rhodamine filter, or an anti-fluorescein antibody following in situ hybridization. Briefly, embryos were treated with 100 mM glycine, pH 2.2, to inactivate alkaline phosphatase and washed with PBS-T (phosphate buffered saline + 0.1% Tween). Anti-fluorescein-alkaline phosphatase conjugated antibody (Boehringer Mannheim; 1:500) was incubated with embryos overnight at 4°C and detected with Fast Red (Sigma).

Whole-mount in situ hybridization was performed as described previously(Alexander et al., 1998). dkk1 anti-sense probe was prepared as described by Hashimoto et al.(Hashimoto et al., 2000).

Whole-mount in situ hybridized embryos were genotyped by PCR using restriction polymorphisms for bon^m425 and sur^m768, and agarose polymorphism for sqt^cz35 mutant embryos, as described previously(Feldman et al., 1998; Kikuchi et al., 2000; Sirotkin et al.,2000,Sirotkin et al.,2000). Genotyping was performed after in situ hybridization as follows. After photographing, each embryo was washed with 100% methanol and hydrated with several washes of PBS with 0.1% Tween-20. Genomic DNA was extracted by digestion overnight in 10 mM Tris, 1 mM EDTA, 0.1% NP40, 0.1%Tween-20, 50 μg proteinase K at 55°C.

bon was initially identified as a mutation that causes cardia bifida, a condition in which the precardiac mesoderm fails to migrate to the midline and fuse (Stainier et al.,1996). At 28 hours post-fertilization (hpf), the cardia bifida phenotype is accompanied by pericardial edema(Fig. 1B; arrowhead). Previous characterization of bon^–/– embryos has shown that the primary phenotype is a severe reduction in the number of endodermal precursors, and the likely cause of cardia bifida(Kikuchi et al., 2000). Closer inspection reveals that bon^–/– embryos also exhibit reduced forebrain structures, with a reduction in eye size being most prominent (Fig. 1A,B; arrows). In order to assess whether this reduction reflects defects in neural patterning, we examined the expression of region-specific markers in the neural plate of early somite stage embryos. In bon^–/– embryos, the otx2 expression domain in the presumptive forebrain and midbrain regions(Mori et al., 1994) is approximately 10% smaller than wild type(Fig. 1C,D), suggesting that the anterior neural plate is reduced. Consistent with this result, double staining with emx1, a marker of the anterior boundary of the neural plate, and her5, a marker of the midbrain-hindbrain boundary (MHB),shows a reproducible and consistent reduction in the distance between the anterior edge of emx1 expression and the posterior tip of her5 expression in bon^–/– embryos(Fig. 1E,F). These results suggest that bon functions not only in endoderm formation but also in neural patterning.

bon is expressed in all mesendodermal progenitors prior to the onset of gastrulation (Alexander et al., 1999). The axial mesoderm is thought to promote neuroectodermal fates (reviewed by Harland and Gerhart, 1997),and the nonaxial mesoderm has been implicated in patterning the neuroectoderm(Woo and Fraser, 1997). To determine the mesendodermal derivative in which Bon function is required for neural patterning, we inhibited bon function in a tissue-specific manner by using morpholino antisense oligonucleotides (MO). Restriction of the bon MO was achieved by injecting it into a single cell at the 32-cell stage. The MO was conjugated to fluorescein to track its localization. In control experiments, bon MO injections at the one-cell stage phenocopy the bon mutation very specifically in more than 95% of the embryos (n>1000; data not shown).

To assess the anterior neural plate during the stages of neural patterning,MO-injected embryos were fixed and examined for otx2 expression. Following in situ hybridization, we also performed anti-fluorescein antibody staining to determine the localization of the bon MO. Embryos with axial mesoderm restriction of the bon MO (n=25) showed a reduction in the otx2 expression domain(Fig. 2A), whereas embryos with bon MO restriction in non-axial mesoderm (n=13) exhibited wild-type otx2 expression (Fig. 2B,C).

In addition to in situ hybridization with otx2, individual MO-injected embryos were followed for morphological observations. At the effective MO concentration, the fluorescein-tag proved to be an ineffective lineage tracer in live embryos. Thus, as an additional lineage tracer, 10 kDa rhodamine-dextran was co-injected with the bon MO. Following antibody staining for fluorescein, we observed that the 10 kDa rhodamine-dextran co-localized with the bon MO in the co-injected embryos (data not shown), thus providing a reliable method to determine the localization of cells with reduced Bon function. Examples of tissue restriction are shown in Fig. 2D-F. As expected, embryos with bon-MO restriction in the neuroectoderm (n=2; Fig. 2F,F′), where bon is not expressed, were normal(Fig. 2F″). Consistent with the otx2 expression data mentioned earlier, embryos with bon-MO restriction in the ventral mesoderm (n=5; Fig. 2E,E′) were also normal (Fig. 2E″). However, all embryos with bon-MO restriction in the axial mesoderm(n=27; Fig. 2D),derivatives of which populate the notochord (white arrowhead) and head mesenchyme (white arrow; Fig. 2D′), exhibited reduced forebrain structures, with a reduction in eye size being most prominent(Fig. 2D″; arrow),similar to the neural defects seen in bon^–/–embryos. In addition, we excluded the endoderm as a tissue in which bon functions to modulate neural patterning because in embryos lacking all endoderm, such as casanova mutants, neural patterning is unaffected (data not shown). Together, these results indicate that Bon function is required in the axial mesoderm for neural patterning.

To further analyze the requirement of the axial mesoderm during neural patterning, we examined the expression of the anterior axial mesoderm marker gsc (Stachel et al.,1993) at several stages during gastrulation. At the shield stage, bon^–/– embryos show gsc expression that is indistinguishable from that seen in wild-type embryos(Fig. 3A). At 90% epiboly, the gsc expression domain is reduced in bon^–/– embryos(Fig. 3C), indicating a differentiation defect in the anterior axial mesoderm. The same progressive reduction in anterior axial mesoderm gene expression was also observed with bmp4. During gastrulation stages, bmp4 is expressed ventrolaterally, as well as in a discrete domain of the anterior axial mesoderm (Hwang et al., 1997; Martinez-Barbera et al.,1997). This expression pattern allowed us to assess dorsoventral patterning as well as axial mesoderm formation. At 50% epiboly, wild-type and bon^–/– embryos show indistinguishable bmp4 expression ventrolaterally(Fig. 3D,G), indicating that dorsoventral patterning is not affected in bon^–/– embryos. Dorsal bmp4expression also appears unaffected at this stage(Fig. 3D,G; arrowhead). At 90%epiboly, wild-type and bon^–/– embryos show a wild-type pattern of ventrolateral bmp4 expression(Fig. 3H,I), but the anterior axial mesoderm bmp4 expression domain is dramatically reduced in bon^–/– embryos(Fig. 3E,F,H,I; arrows). These data indicate that although the early induction of axial mesoderm occurs properly in bon^–/– embryos, its subsequent differentiation is defective.

In order to better understand the role of bon in neural patterning, we crossed bon^+/– fish with fish heterozygous at other loci regulating axial mesoderm formation, and found a functional interaction between bon and the nodal-related gene sqt. Although bon^+/– and sqt^+/– embryos appear to have a wild-type phenotype,approximately 20% of bon^+/–;sqt^+/– embryos exhibit a cyclopic phenotype similar to that seen in sqt^–/– embryos(Fig. 4C; arrow). In addition,whereas bon^–/– embryos exhibit a slight reduction in forebrain structures (Fig. 2B; arrow), bon^–/–;sqt^–/–embryos exhibit a complete absence of forebrain, lacking telencephalic and diencephalic structures as well as eyes (arrow; Fig. 4D). Interestingly, this interaction was not found with the nodal-related gene cyc,further indicating that sqt and cyc play distinct roles in neural patterning. In addition, MZsqt^–/–embryos do not exhibit as severe a defect as that seen in bon^–/–;sqt^–/–embryos (data not shown). Together, these data suggest that Bon and Sqt function in parallel to regulate neural patterning.

To assess the neural defects resulting from the loss of bon and sqt function, we analyzed the expression of region-specific markers in the neural plate of wild-type, bon^–/–, sqt^–/– and bon^–/–;sqt^–/–embryos. At the 1-somite stage, emx1 marks the anterior boundary of the neural plate, whereas krox20 (egr2b – Zebrafish Information Network) marks rhombomeres 3 and 5 of the hindbrain (r3 and r5; Fig. 4E,I)(Oxtoby and Jowett, 1993). In bon^–/– embryos, the distance between the anterior neural ridge (emx1) and the r5/r6 boundary, as well as the distance between r3 and r5 is reduced (Fig. 4F,J; brackets). In addition, the lateral borders of the neural plate (emx1) appear to be shifted medially(Fig. 4J; asterisks), further indicating a reduction in the neural plate. In sqt^–/– and bon^–/–;sqt^–/–embryos, the reduction in the distance between the anterior edge of the neural plate and the r5/r6 boundary appears to be more dramatic(Fig. 4G,H). Additionally, r3 and r5, as marked by krox20 staining, appear to be closer together in bon^–/–;sqt^–/–embryos (Fig. 4H,L). This apparent merging of r3 and r5, and the reduced distance between the anterior neural ridge and the r5/r6 boundary, indicates a reduction of neural tissue along the AP axis.

Anteriorly, the emx1 expression domain spreads medially to cover the entire anterior ventral neural plate in sqt^–/– and bon^–/–;sqt^–/–embryos (Fig. 4K,L). This expansion appears to be restricted to emx1 expression, as otx2 expression is reduced in sqt^–/–and bon^–/–;sqt^–/–embryos (Fig. 4O,P). Consistent with this result, and with the morphological absence of eyes in bon^–/–;sqt^–/–embryos, the expression of opl (zic1 – Zebrafish Information Network) and rxb, markers of the eye field, is dramatically reduced or absent in bon^–/–;sqt^–/–embryos (data not shown). Together, these data indicate that loss of bon and sqt function leads to synergistic defects in neural patterning.

AP patterning of the neuroectoderm is regulated by posteriorizing signals and their antagonists (reviewed by Yamaguchi, 2001). Recent evidence points to the Wnt signaling pathway as a key regulator of AP patterning, with Wnt8 as a posteriorizing signal and the Wnt antagonist Dkk1 as promoting anterior neural fates (Glinka et al., 1998; Erter et al.,2001). The neural patterning defects in bon^–/–, sqt^–/–and bon^–/–;sqt^–/–embryos were reminiscent of defects caused by an excess of Wnt signaling(Kim et al., 2000; Erter et al., 2001). Therefore,we examined the expression of dkk1 in bon^–/–, sqt^–/–and bon^–/–;sqt^–/–embryos and found that defects in dkk1 expression correlated with the severity of the neural patterning defects observed in these mutant embryos. At 50% epiboly, dkk1 expression is observed in all marginal blastomeres(Fig. 5A)(Hashimoto et al., 2000; Shinya et al., 2000). In bon^–/– embryos, there is a dorsal gap in dkk1 expression (Fig. 5B). This dorsal gap appears more extensive in sqt^–/– and bon^+/–;sqt^+/–embryos(Fig. 5C). In bon^–/–;sqt^–/–embryos, dkk1 expression is seen only in the ventral half of the margin (Fig. 5D). At 70%epiboly, dkk1 is expressed in cells of the prechordal plate (PCP; Fig. 5E)(Hashimoto et al., 2000; Shinya et al., 2000). Consistent with bon^–/– embryos exhibiting defects in anterior axial mesoderm gene expression, the dkk1-expressing cells appear to coalesce aberrantly in these mutants(Fig. 5F). In sqt^–/– and bon^+/–;sqt^+/– embryos, dkk1 expression in the PCP is dramatically reduced(Fig. 5G), reflecting a defect in anterior axial mesoderm formation. This reduction is enhanced in bon^–/–;sqt^–/–embryos, where dkk1 expression appears to be completely absent in the PCP region (Fig. 5H). These data suggest that the defects in dkk1 expression may be responsible,at least in part, for the neural patterning defects. In order to test this hypothesis, we overexpressed dkk1 in bon^–/– embryos and observed an enlargement of the forebrain and eyes, suppressing the anterior neural deficiency (data not shown). However, the cardia bifida phenotype was not rescued, suggesting that dkk1 functions in neural patterning but not in endoderm development.

In addition to defects in dkk1 expression, we found that bon^–/–;sqt^–/–embryos have defects in dorsal mesendoderm gene expression. In wild-type and bon^–/– embryos at 50% epiboly, ntl is expressed around the margin of the embryo(Fig. 5I,J). In sqt^–/– or bon^+/–;sqt^+/– embryos, ntl expression appears reduced(Fig. 5K), and, in bon^–/–;sqt^–/–embryos, it is absent from the dorsal half of the margin(Fig. 5L), suggesting that the formation of dorsal mesoderm is defective in bon^–/–;sqt^–/–embryos. This reduction in dorsal mesendoderm gene expression in bon^–/–;sqt^–/–embryos was also observed with other markers, such as wnt8 (data not shown). Thus, in bon^–/–;sqt^–/–embryos, the lack of dkk1 expression from dorsal mesendoderm may reflect an overall deficit in dorsal mesendoderm gene expression, which suggests that bon and sqt function in parallel to regulate dorsal mesendoderm formation as well as neural patterning.

The genetic interaction between bon and sqt suggested that these two genes function in parallel to regulate neural patterning. However, molecular epistasis analyses have indicated that bonexpression is dependent on Nodal signaling, which places bondownstream of sqt (Alexander et al., 1999). Thus, additional signal(s) must function upstream of bon, and additional Nodal transcriptional mediator(s) must function downstream of sqt. The foxh1 gene mutant locus sur was a good candidate to be an additional Nodal transcriptional mediator in neural patterning due to its role in axis formation (Pogoda et al.,2000; Sirotkin et al.,2000,Sirotkin et al.,2000). Therefore, we asked whether bon^–/–;sur^–/–embryos exhibit neural patterning defects. Although bon^–/– embryos exhibit a slight reduction in anterior neural structures(Fig.1B and Fig. 4B; arrow) and sur^–/– embryos exhibit mild cyclopia(Pogoda et al., 2000; Sirotkin et al.,2000,Sirotkin et al.,2000), bon^–/–;sur^–/–embryos exhibit a dramatic reduction of forebrain structures, with the most severally affected embryos exhibiting an absence of telencephalic and diencephalic structures, as well as eyes(Fig. 6B; arrow). Interestingly, bon^–/–;sur^+/–embryos also exhibited anterior truncations at a low percentage (1.8%, n=340) when they originated from bon^+/–;sur^+/– females but not from bon^+/–;sur^+/–males, indicating that a reduction in maternal Sur (Foxh1) can enhance the bon neural phenotype.

The loss of anterior structures in bon^–/–;sur^–/–embryos was reminiscent of the bon^–/–;sqt^–/–phenotype (Fig. 4D); thus, we used the same region-specific neural markers that were employed in the bon^–/–;sqt^–/–analyses to assess neural patterning in bon^–/–;sur^–/–embryos. At the tailbud stage, bon^–/–;sur^–/–embryos exhibit a dramatic reduction in the distance between the anterior edge of emx1 expression and the r5/r6 boundary(Fig. 6F; bracket). In addition, the rhombomeres r3 and r5 appear closer together(Fig. 6J). The similarity in neural patterning defects between bon^–/–;sur^–/–and bon^–/–;sqt^–/–embryos indicates that Sur (Foxh1) may be the additional Nodal transcriptional mediator functioning downstream of Sqt and in parallel to Bon in neural patterning (Fig. 7).

To further analyze the similarity in neural patterning defects between bon^–/–;sur^–/–and bon^–/–;sqt^–/–embryos, we examined dkk1 expression in bon^–/–;sur^–/–embryos. We found that at 50% and 70% epiboly, bon^–/–;sur^–/–embryos exhibit a loss of dkk1 expression(Fig. 6C,D,G,H) similar to that seen in bon^–/–;sqt^–/–embryos (Fig. 5C). Further, we found that expression of ntl is also absent from the dorsal side of bon^–/–;sur^–/–embryos (Fig. 6K,L), which suggests that the formation of dorsal mesendoderm is defective in bon^–/–;sur^–/–embryos. Altogether, these data indicate that Bon and Sur (Foxh1) function in parallel to regulate dorsal mesendoderm gene expression and neural patterning.

In this study, we show that the Mix homeodomain gene bon is required in the axial mesoderm to regulate neural patterning. Our results indicate that the severity of the neural patterning defects in bon^–/– embryos correlates with the degree of reduction in dkk1 expression in the dorsal mesendoderm and,subsequently, the anterior axial mesoderm. Genetic interactions between bon and the components of the Nodal signaling pathway, sqtand sur (foxh1), reveal a complex network that mediates Nodal signaling in neural patterning. First, the genetic interaction between bon and sqt suggests that the relationship between bon and sqt is not strictly linear as previously suggested by molecular epistasis studies (Alexander et al., 1999; Kikuchi et al., 2000). Second,the bon;sur interaction demonstrates that these two transcriptional factor genes play overlapping functions in neural patterning. Finally,expression studies indicate that Bon, Sqt and Sur (Foxh1) function to regulate dorsal mesendoderm genes, such as ntl and dkk1, the latter playing an important role in neural patterning(Glinka et al., 1998; Hashimoto et al., 2000; Mukhopadhyay et al., 2001; Shinya et al., 2000).

Genetic and embryological analyses indicate that Mix genes are potent inducers of mesodermal and endodermal gene expression. Ectopic expression of Mix.1, Milk, Mixer, Bix1, mezzo and bon leads to the expression of mesodermal and endodermal genes(Henry and Melton, 1998; Lemaire et al., 1998;Alexander et al., 1999; Latinkic and Smith, 1999; Poulain and Lepage; 2002). Additionally, a genetic lesion in the zebrafish Mix gene bon leads to a reduction in endodermal precursors(Kikuchi et al., 2000). Our data point to an essential role for Bon in the axial mesoderm for neural patterning. We found that a reduction in Bon function in the axial mesoderm caused by restricted MO injection is associated with anterior neural defects. In addition, bon^–/– embryos display defects in axial mesoderm gene expression. Furthermore, based on the expression pattern of bon in mesendodermal progenitors before involution, we favor a model in which Bon regulates the transcription of neural patterning genes that are expressed in mesendodermal precursors. The finding that dkk1expression is absent from the dorsal side of bon^–/– embryos is consistent with this model. It is interesting to note that studies in Xenopus had hinted at a role for Mixer in head formation and Dkk1 expression(Henry and Melton, 1998).

Recent findings have revealed that the spatial variation in the level of Wnt signal plays a crucial role in the AP patterning of the neuroectoderm(reviewed by Yamaguchi, 2001; Erter et al., 2001; Kudoh et al., 2002). Extensive evidence from genetic and overexpression studies points to the importance of Wnt antagonism for anterior neural patterning. Specifically, Dkk1mouse mutant embryos lack head structures anterior to the midbrain, whereas overexpression of dkk1 in amphibians and zebrafish embryos leads to enlarged heads (Glinka et al.,1998; Hashimoto et al.,2000; Mukhopadhyay et al.,2001; Shinya et al.,2000). Conversely, ectopic expression of wnt8 suppresses anterior fates, whereas a deficiency in the wnt8 locus or a reduction of Wnt8 caused by MO injection in zebrafish embryos leads to a loss of posterior neural fates (Erter et al.,2001; Lekven et al.,2001). Our data indicate that the precise level of Wnt signaling required for neural patterning is transcriptionally controlled by Nodal signaling as well as by Bon and Sur (Foxh1).

Overexpression and mutant analyses have indicated that Bon functions exclusively downstream of Nodal signaling in endoderm formation (Alexander et al., 1999; Kikuchi et al.,2000). However the synergistic neural patterning defects seen in bon^–/–;sqt^–/–embryos indicate that Bon also functions in parallel to Sqt signaling. Biochemical analyses indicate that a subset of Mix homeodomain proteins, as well as winged-helix transcription factors, physically interact with the Smad2/Smad4 complex through a conserved motif in their C terminus(Germain et al., 2000). This Smad interaction motif is present in Bon and Sur (Foxh1)(Pogoda et al., 2000; Randall et al., 2002), raising the possibility that Bon and/or Sur (Foxh1) can interact with the Smad2/Smad4 complex, upon Sqt activation of the Nodal pathway, to activate downstream targets. The loss of dkk1 expression in bon^–/–;sqt^–/–and bon^–/–;sur^–/–embryos indicates that dkk1 is one of the genes regulated in this manner. Whether Bon and Sur (Foxh1) bind directly to the dkk1promoter needs to be investigated.

In addition, we also found defects in wnt8 expression at the margin of bon^–/–;sqt^–/–embryos suggesting that the neural patterning defect in these double-mutant embryos may not be solely due to an expansion of Wnt signaling. We do observe a shortening of the body axis in bon^–/–;sqt^–/–and bon^–/–;sur^–/–embryos, which may lead to a misplacement of neural organizing centers, such as the anterior neural boundary cells and the MHB (reviewed by Liu and Joyner, 2001; Houart et al., 1998), which would further affect AP patterning of the neural plate (see Fig. 4E-H, Fig. 6E-F).

By combining our results with biochemical(Germain et al., 2000) and molecular epistasis data (Alexander et al., 1999), a model emerges in which the Nodal signal provided by Sqt is transduced by a complex of Smad2/Smad4 that is recruited to specific target genes by either Bon or Sur (Foxh1; Fig. 7). These two transcriptional mediators of Nodal signaling have unique functions during the formation of endoderm and axial mesoderm but have overlapping activities in neural patterning. The genetic interactions between bon;sqt and bon;sur indicate that Bon functions in parallel to Sqt and Sur(Foxh1) to regulate the expression of mesendodermal genes, such as dkk1, which in turn is required for neural patterning.

In endoderm formation, Bon functions downstream of Nodal signaling in an Oep-dependent fashion (Alexander et al., 1999; Kikuchi et al., 2000). bon expression is unaffected in MZsur^–/– embryos (data not shown), suggesting that an additional Smad-binding transcription factor is involved in regulating bon expression (Fig. 7; factor Y). A possible candidate for this activity could be the Mix-like transcription factor, Mezzo that was shown to function downstream of Nodal signaling. However, bon expression is probably not regulated by Mezzo as Mezzo lacks a Smad interaction motif and mezzo MO-injected embryos do not exhibit endoderm defects(Poulain and Lepage, 2002). Thus, we propose that an additional, as yet unidentified, Smad-binding transcription factor (Y) is involved in the initiation of bonexpression.

Once bon expression is initiated, our model places Bon and Sur(Foxh1) as the two transcriptional mediators of Sqt signaling in neural patterning. However, it should be re-emphasized that MZsqt^–/– embryos exhibit a less severe neural patterning defect than that seen in either bon^–/–;sqt^–/–or bon^–/–;sur^–/–embryos, which indicates that Sqt is not the sole signal regulating Bon transcriptional activity. Thus, an additional factor (X) may function upstream of Bon and in parallel to Sqt in neural patterning. In this model, factor X could correspond to Cyc, as it has been suggested that the ventrolateral mesoderm, which requires Nodal signaling for its formation, can provide a secondary posteriorizing signal to the neural plate(Erter et al., 2001; Feldman et al., 2000; Woo and Fraser, 1997). The neural defect seen in bon^–/–;sqt^–/–and bon^–/–;sur^–/–embryos, but not in cyc^–/–;sqt^–/–embryos, may be caused by the presence of ventrolateral mesoderm and its posteriorizing effect on the neural plate. Further studies should reveal how the various Nodal ligands, as well as other signals, regulate neural patterning, either directly, or through their regulation of mesendodermal gene expression.

We thank Pia Aanstad, Elke A. Ober and Nick Osborne for comments on the manuscript and helpful discussions. L.A.T. was supported by a National Science Foundation Predoctoral fellowship. This work was supported in part by grants from the NIH (DK 58181) and the Packard Foundation to D.Y.R.S.