An essential role of Xenopus Foxi1a for ventral specification of the cephalic ectoderm during gastrulation

In vertebrate gastrula embryos, the ectoderm is subdivided into various regional tissues by complex inductive processes. Along the dorsoventral (DV)axis, the ectoderm becomes subdivided into the dorsal (CNS or neural plate),intermediate (e.g. presumptive neural crest, placodes and cement gland) and ventral (epidermal) ectoderm. Although it is generally believed that the early DV specification is controlled by a Bmp activity gradient in Xenopus(Wilson et al., 1997; LaBonne and Broner-Fraser, 1998; Marchant et al.,1998; Tríbulo et al.,2003), how the exact subdivision boundaries are determined remains largely elusive.

Along the anteroposterior (AP) axis, the neural plate is finely regionalized into the forebrain, midbrain, hindbrain and spinal cord. Recent molecular studies implicated Wnts, Nodal, Fgfs and RA in the AP regionalization of the CNS (McGrew et al.,1995; Kengaku and Okamoto,1995; Piccolo et al.,1999; Gavalas and Krumlauf,2000; Kiecker and Niehrs,2001; Kudoh et al.,2002; Onai et al.,2004).

In contrast to the CNS, relatively little is known about the AP specification of the non-CNS (intermediate and epidermal) ectoderm. In Xenopus, the non-CNS ectoderm of the head region (referred to as`cephalic non-neural ectoderm' hereafter) has several characteristic features. For example, unlike that of the trunk, the non-neural ectoderm of the cephalic region differentiates into the cranial placodes and special exocrine glands such as the cement gland and hatching gland, in addition to the epidermis and neural crest. The cranial placodes, which develop within the preplacodal field of the head intermediate ectoderm, give rise to a number of sensory tissues(reviewed by Baker and Bronner-Fraser,2001). To date, the molecular mechanism underlying the determination of the cephalic non-neural ectoderm (versus the CNS and the trunk ectoderm) remains largely to be elucidated.

In this study, we have investigated the molecular control of the initial specification of the cephalic non-neural ectoderm by focusing on the roles of a Foxi1 family gene in Xenopus. The winged-helix transcription factor Foxi1 plays an essential role for the formation of placode-derived ectodermal tissues such as the otic vesicle (Hulander et al., 1998; Nissen et al.,2003; Solomon et al.,2003a) in mice and zebrafish. In Xenopus, three Foxi1-related genes have been reported: Xfoxi1a, Xfoxi1b(pseudoalleles generated by the pseudotetraploidy of Xenopus laevis,see alignment of Xfoxi1a and Xfoxi1b in Fig. 1M) and Xfoxi1c(which is not an Xfoxi1a pseudoallele) are expressed in the preplacodal area at the neurula stage (Lef et al., 1994; Pohl et al.,2002).

Interestingly, Xfoxi1a and Xfoxi1b are also expressed even earlier than the establishment of the preplacodal expression at the neurula stage; they are expressed widely in the animal side of the embryo at the late blastula stage and in the anteroventral ectoderm at the late gastrula stage. By contrast, Xfoxi1c is expressed only after the gastrula stage and not during the blastula and gastrula stages(Pohl et al., 2002). The expression of a foxi1 gene in a broad domain of the gastrula ectoderm has been reported also in zebrafish(Nissen et al., 2003; Riley and Phillips, 2003; Solomon et al., 2003a). However, the role of the Foxi1 family genes during the gastrula stage has not yet been elucidated. In addition, although several transcription factors have been implicated in the development of the non-neural ectoderm in Xenopus (e.g. Dlx3, Msx1, Gata1 and Xvent1/2)(Onichtchouk et al., 1996; Suzuki et al., 1997; Ault et al., 1997; Onichtchouk et al., 1998; Feledy et al., 1999; Beanan and Sargent, 2000; Woda et al., 2003), none of them are expressed in a pattern limited to the cephalic non-neural ectoderm during gastrulation. These facts led us to investigate the role of Xfoxi1a (including that of the Xfoxi1b; the term Xfoxi1a/b is used hereafter when the combined functions are considered) in the head ectoderm of the Xenopus gastrula. By focusing on the role at the early stage, we demonstrate that Xfoxi1a/b is essential for the specification of the non-neural ectoderm in the head. We also show that Xfoxi1a/b misexpression promotes epidermal differentiation at the cost of neural tissues. We discuss a possible mode of the Xfoxi1a/b action, focusing on the critical period of Xfoxi1a/b-mediated ectodermal patterning.

The coding sequence of Xfoxi1a (GenBank Accession Number X74315)was amplified from Xenopus stage 12.5 cDNA by PCR using KOD Plus DNA polymerase (Toyobo, Osaka, Japan). The resulting cDNA was subcloned into the EcoRI-XhoI site of the pCS2 vector(Turner and Weintraub, 1994). The ligand-binding domain of the glucocorticoid receptor(Hollenberg et al., 1985; Hollenberg et al., 1993; Kolm and Sive, 1995) was amplified by PCR with these primers: forward 5′-GCCGGATCCACCATGACCTCTGAAAATCC-3′ and reverse 5′-GCCATCGATCCTTTTGATGAAACAGAAG-3′. The resulting products were fused in frame at the BamHI-ClaI site in the above plasmid(Xfoxi1a/pCS2). The Xfoxi1a and Xfoxi1b (GenBank Accession Number X74316) with the 5′-UTR sequence and an additional C-terminal flag-tag sequence were constructed by using the following primers:forward 5′-GCCATCGATTCAGTTGGGAAAGAGCAGAAGCCGCTG-3′ and reverse 5′-GCCCTCGAGTTACTTATCGTCGTCATCCTTGTAATCGTACCTTCCCTGGTACAGAGGAGACCTGC-3′;forward 5′-GCCATCGATTCTGCATCAGTTAGAAAAGAGCGATT-3′ and reverse 5′-GCCCTCGAGTTACTTATCGTCGTCATCCTTGTAATCATACCTTCCCTGGTACAAAGGGGG-3′,respectively. The products were inserted in to the ClaI-XhoI site of pCS2.

Eggs were collected from adult Xenopus laevis and fertilized in vitro as described previously (Sasai et al., 2001). Embryos were staged according to Nieuwkoop and Faber(Nieuwkoop and Faber, 1967). After dejellying the embryo by treatment with 2% cysteine (pH 7.8),microinjection was carried out in 1× Barth's solution. Embryos were grown in 0.1× Barth's solution until sibling embryos reached the desired stage. For animal cap assays, ectodermal explants were excised at stage 9 and then cultured in 1× LCMR supplemented with 0.2% BSA until the stages mentioned. For the treatment of the embryo with dexamethasone (Dex), Dex was added to the 0.1× Barth's solution to a 10 μM final concentration at stage 11 or 13, as described by Gammill and Sive(Gammill and Sive, 1997). The embryos were harvested at the neurula stage.

Capped mRNAs for the microinjection were synthesized by using an SP6 mMassage Machine kit (Ambion, Austin, TX) according to the manufacturer's protocol. RNA was injected into all animal blastomeres or into the unilateral blasomeres of eight-cell embryos. Morpholino antisense oligonucleotides (Gene Tools, Philomath, OR) were designed against the 5′ regions (see Fig. 3A)of Xfoxi1a (Xfoxi1a-MO,5′-GATCAGCGGCTTCTGCTCTTTCCCA-3′) and Xfoxi1b(Xfoxi1b-MO, 5′-GGTTCATCTCGCTCACTGGCTAATC-3′). Oligonucleotides with five mismatches (5-mis-Xfoxi1a,5′-GATCAcCGGgTTCTcCTgTTTCgCA-3′; 5-mis-Xfoxi1b,5′-GGTTgATgTCGCTgACTcGCTAtTC-3′) were used as negative controls. For the rescue experiment, wild-type Xfoxi1a mRNA lacking the 5′-UTR sequence was co-injected with Xfoxi1a-MO (containing no complimentary sequence). After fixing the embryo with MEMFA at the appropriate stage, whole-mount in situ hybridization was performed as described previously(Sasai et al., 2001). For double in situ hybridization, fluorescein-labeled probe was stained with BCIP(Roche, Mannheim, Germany) and digoxigenin-labeled probe was stained with BM-purple (Roche, Germany) or Magenta-Phos (Biosynth, Switzerland). All of the injection experiments were carried out at least twice and gave reproducible results.

RT-PCR was performed as described previously(Mizuseki et al., 1998; Kuo et al., 1998; Tsuda et al., 2002). The other primers used first in this study were as follows: Dlx3(Papalopulu and Kintner, 1993; Dirksen et al., 1994) (forward primer, ATGAGTGGCCCCTATGAGAAGAAG; reverse primer, GGTTCTCTGTAATGGACAAACGG); Sox2 (Mizuseki et al.,1998) (forward primer, GAGGATGGACACTTATGCCCAC; reverse primer,GGACATGCTGTAGGTAGGCGA), Bmp4 (Dale et al., 1992) (forward primer, GCATGTACGGATAAGTCGATC; reverse primer, GATCTCAGACTCAACGGCAC), Xfoxi1a(Lef et al., 1994) (forward primer, CCAGAACTGAAATCTTAGCAA; reverse primer, TAACAAAGATAAAGCCAGAGGT), MyoD (Hopwood et al.,1989) (forward primer, AGGTCCAACTGCTCCGACGGCATGAA; reverse primer,AGGAGAGAATCCAGTTGATGGAAACA), H4(Perry et al., 1985) (forward primer, CGGGATAACATTCAGGGTATCACT; reverse primer,ATCCATGGCGGTAACTGTCTTCCT).

Animal caps were lysed in the extraction buffer [20 mM HEPES (pH 7.9), 420 mM NaCl, 1.5 mM MgCl₂, 0.2 mM EDTA, 0.5% NP-40, 1:100 dilution of protease inhibitor cocktail; Cytoskeleton, Denver, CO] and cleared by micro-centrifugation at 20,000 g for 10 minutes. Aliquots of 10-30 μg proteins were resolved by 10% SDS-PAGE and then blotted on to a PVDF membrane filter (Millipore, MA). For the primary antibody, anti-FLAG M2 mouse monoclonal antibody (1:1000, Sigma) was used. For the secondary antibody, an anti-mouse IgG horseradish peroxidase linked F(ab′)2 fragment (1:5000, Amersham) was used. Signals were detected with ECL reagents(Amersham).

To understand the role of Xfoxi1a during the early steps of embryogenesis (blastula to early neurula), we first performed whole-mount in situ hybridization experiments to analyze the precise pattern of Xfoxi1a expression (Fig. 1). Consistent with previous reports(Lef et al., 1994; Pohl et al., 2002), no maternal expression of Xfoxi1a was observed(Fig. 1A). Xfoxi1a was widely expressed in the animal cap region at the blastula stage(Fig. 1B). Interestingly, Xfoxi1a expression was gradually shut off in the dorsal and posterior ectoderm during early gastrulation, and by the mid-gastrula stage it was localized to the anteroventral ectoderm(Fig. 1C). This expression pattern is consistent with a zebrafish study reporting that a foxi1gene is expressed in the anteroventral quadrant of the early gastrula(Nissen et al., 2003; Riley and Phillips, 2003; Solomon et al., 2003a). Double in situ hybridization showed that the expression domains of Xfoxi1aand that of Sox2 (neural plate)(Mizuseki et al., 1998) were complementary to each other and did not overlap(Fig. 1D), indicating that Xfoxi1a expression is confined to the cephalic non-neural ectoderm. At the early neurula stage, Xfoxi1a expression gradually became limited to the most anterior part of the non-neural ectoderm(Fig. 1E). By the mid-neurula stage, Xfoxi1a expression was found only in a horseshoe-shaped domain within the anterior intermediate ectodermal (or preplacodal) region(Fig. 1F). At this stage, an obvious gap was seen between the anterior neural plate and the Xfoxi1a⁺ domain (Fig. 1G). This gap area expressed another early preplacodal marker, Six1 (Pandur and Moody,2000; Ghanbari et al.,2001) (Fig. 1H),suggesting that the preplacodal region is already divided at the marker level into different DV subdomains by mid-neurulation (Six1 and Xfoxi1a expressions partially overlap in the lateral region but do not coincide in the medial region). Consistent with this idea, double in situ hybridization with a probe for Xag1(Sive et al., 1989) indicated that the Xfoxi1a⁺ domain (but not the Six1⁺ domain) partly overlap with the cement gland anlage(data not shown). At the tailbud stage, Xfoxi1a was expressed in restricted branchial arch regions of the head ectoderm (the profundal placodes and the head lateral line system)(Schlosser and Northcutt,2000) (Fig. 1I) but not in the otic placodes, consistent with a previous study(Pohl et al., 2002).

The expression pattern of the pseudoallele Xfoxi1b showed an expression pattern indistinguishable from that of Xfoxi1a (tissue distribution at representative stages shown in Fig. 1J-L).

The in situ hybridization analysis above shows that Xfoxi1a is expressed specifically in the anteroventral (or cephalic non-neural) ectoderm during the mid-gastrula and early neurula stages. We next investigated patterning signals that controlled the spatial expression of Xfoxi1aduring these stages, by focusing on the roles of Bmp and Wnt signals. When Bmp4 (2.5 pg of the expression plasmid DNA per cell)(Dale et al., 1992) was injected into all animal blastomeres at the eight-cell stage, Xfoxi1aexpression significantly expanded into the dorsal ectoderm at stage 12 (77%, n=13; Fig. 2B),whereas overexpression of the Bmp antagonist Chd (50 pg RNA/cell)(Sasai et al., 1995)suppressed Xfoxi1a expression (83%, n=12; Fig. 2C). We then performed experiments using the ectodermal explants (animal cap assay) to distinguish direct effects on the ectoderm from secondary effects via the mesoderm. Consistent with the in vivo observation, Xfoxi1a expression was diminished by Chd injection in the animal cap assay(Fig. 2K-M,O,P), but its expression was rescued by co-injection of Bmp4(Fig. 2K, lane 4), indicating that Bmp signaling positively regulates Xfoxi1a expression by directly acting in the ectoderm.

We next studied the role of Wnt signaling in Xfoxi1a expression. Microinjection of a Wnt1-expression plasmid (2.5 pg DNA/cell) into the animal blastomeres markedly reduced Xfoxi1a expression (100%, n=14; Fig. 2D). By contrast, overexpression of the Wnt inhibitor gene Dkk1 (125 pg/cell)(Glinka et al., 1998) resulted in the expansion of Xfoxi1a expression into the posteroventral ectoderm (38%, n=16; Fig. 2E). Consistent with this, the animal cap assay showed that Wnt1 suppressed Xfoxi1a expression in ectodermal explants(100%, n=30; Fig. 2L,N), without inducing Sox2(Fig. 2O,Q).

These findings suggest that Xfoxi1a expression in the gastrula ectoderm is regulated positively by Bmp signals and negatively by Wnt signals;this regulation presumably occurs as a consequence of the modifications that specify the DV (ventralization by Bmp4) and AP (posteriorization by Wnt) ectodermal identities. Consistent with this idea, Xfoxi1aexpression in the neurula ectoderm (Fig. 2F) was suppressed by Chd (93.8%, n=16; Fig. 2H) and by pCS2Wnt1 (93.8%, n=16; Fig. 2I), and upregulated by Dkk1 (100%, n=15; Fig. 2J). In contrast to the effect on the gastrula ectoderm(Fig. 2B), injection of pCS2Bmp4 did not cause expansion of Xfoxi1a in the neurula head ectoderm (n=16; Fig. 2G), suggesting that late Xfoxi1a expression requires finer local regulation in the neurula ectoderm.

To understand the role of Xfoxi1a/b in early head ectodermal patterning, we performed loss-of-function experiments by injecting morpholino antisense oligonucleotides (MOs; Fig. 3A for the design; Fig. 3B-D for the efficiency and specificity tests using flag-tagged Xfoxi1a/b constructs). The unilateral injection of Xfoxi1a-MO into two right animal blastomeres at the eight-cell stage(12.5 ng/cell) expanded the expression of the neural plate marker Sox2 in the anterior ectodermal region (48.3%, n=31; Fig. 3E), consistent with the in vivo expression pattern of Xfoxi1a. Conversely, the expression of the epidermal markers XK81 (embryonic type I keratin)(Jonas et al., 1985) and Dlx3 (Papalopulu and Kintner,1993; Dirksen et al.,1994; Feledy et al.,1999) decreased on the injected side (40%, n=45 and 36.4%, n=44; Fig. 3F,G, respectively). Xfoxi1a-MO injection also inhibited the expression of markers for the cephalic intermediate ectoderm, including for the neural crest (FoxD3)(Sasai et al., 2001, 72.7%, n=11; Fig. 3H), cement gland (Xag1) (Sive et al.,1989) (45.7%, n=35; Fig. 3I) and pre-placodal region (Six1, 90%, n=22 in Fig. 3J) [Eya1(David et al., 2001) 63.6%, n=11; data not shown]. In the tailbud-stage embryo, Six1expression in the nasal placode disappeared on the injected side (100%, n=11; Fig. 3K). Xfoxi1a-MO injection caused no change in the expression of the axial mesodermal marker Chd (n=18; Fig. 3L) or the paraxial mesodermal marker MyoD (n=10; Fig. 3M), suggesting that the effects of Xfoxi1a-MO on ectodermal patterning are not secondary to the defect in mesodermal development. The control MO with a five-base mismatch(see Materials and methods) showed no effects on the expression of the marker genes (Fig. 3N-P). The expansion of Sox2 expression and the repression of FoxD3 and Six1 caused by Xfoxi1a-MO injection were reversed by co-injecting wild-type Xfoxi1a mRNA lacking the Xfoxi1a-MObinding site (no expansion in 87.5%, n=16; and no repression in 43.5%, n=23, 86.9%, n=23, respectively; Fig. 3Q,R and data not shown). Similar observations were obtained in the knockdown experiments using Xfoxi1b-MO (Fig. 3S-U,and data not shown). The co-injection of Xfoxi1a-MO and Xfoxi1b-MO produced qualitatively indistinguishable effects from single injections (data not shown). These results demonstrate that a sufficient expression level of Xfoxi1a/b (higher than a certain threshold) is essential for the development of the cephalic non-neural ectoderm, and that Xfoxi1a/b has a pivotal role in the `non-neural versus neural'specification of the head ectoderm.

We next performed the animal cap assay to further study the requirement of the Xfoxi1a/b function for the non-neural specification of the ectoderm. In RT-PCR analysis (Fig. 4), control animal caps (prepared at stage 9 and cultured until stage 14) strongly expressed the non-neural ectodermal markers XK81,Msx1 and Dlx3 (lane 2). Consistent with the in vivo study,injection of the MOs for both Xfoxi1a and Xfoxi1b (but not their corresponding five-base mismatched MOs) significantly suppressed XK81, Msx1 and Dlx3 (lanes 5 and 6), suggesting that the Xfoxi1a/b function is essential for naïve ectodermal cells to differentiate into the non-neural ectodermal fate in vitro. In contrast to the in vivo situation (Fig. 3),injection of a single MO, i.e. against pseudoallele Xfoxi1a or Xfoxi1b (lanes 3 and 4; at a dose sufficient to evoke the in vivo phenotypes), caused only moderate suppression, suggesting that the extent of the sensitivity to the gene dose is somehow context-dependent in these cases. Interestingly, the neural marker Sox2 was not substantially induced in the animal caps by MO injections in any cases (lanes 3-5).

To further understand the role of Xfoxi1a/b, we performed gain-of-function studies. The unilateral injection of Xfoxi1a mRNA(25 pg/cell) into the right animal blastomeres of eight-cell stage embryos caused a significant reduction of Sox2 expression (50%, n=52; Fig. 5A) in the anterior neural plate. By contrast, ectopic expression of the epidermal markers XK81 (33.3%, n=54; Fig. 5B) and Dlx3(39.5%, n=43; Fig. 5C)in the neural plate region was observed on the injected side. These gain-of-function phenotypes are consistent with the observations in the loss-of-function analysis (Fig. 3), supporting the idea that Xfoxi1a/b plays a decisive role in the non-neural specification of the head ectoderm.

In the animal cap assay (Fig. 5F), the co-injection of Xfoxi1a suppressed the Chd-induced Sox2 expression (lanes 3 and 4), while the expressions of the epidermal/non-neural ectodermal markers (XK81, Dlx3,Msx1 and Xfoxi1a), which were suppressed by Chd, were rescued. The mesodermal marker MyoD was not induced regardless of the mRNA injection. Next, we further analyzed the relationship between Xfoxi1a and Bmp signaling by co-injecting with the dominant-negative Bmp receptor (dnBMPR) (Suzuki et al., 1994). Neural differentiation caused by dnBMPRinjection in the animal cap was suppressed by co-injecting Xfoxi1a(Fig. 5G). Although Bmp signaling was blocked at the receptor level, Sox2 was suppressed by Xfoxi1a, while the non-neural ectodermal marker XK81 was induced (lanes 3 and 4). These suggest that Xfoxi1a does not act upstream of BMPR, but rather functions downstream and/or in a parallel fashion. Taken together, these findings indicate that Xfoxi1apromotes epidermal differentiation at the cost of neural differentiation both in vivo and in vitro.

Xfoxi1a overexpression in the embryo suppressed the intermediate ectodermal markers FoxD3 and Six1 (67%, n=43, 59%, n=59, respectively; Fig. 5D,E). These phenotypes were similar to those with the loss-of Xfoxi1a function (Fig. 3H,J), suggesting the possibility that the inhibition by Xfoxi1a overexpression involves certain indirect effects on the specification of the intermediate ectoderm.

As shown in Fig. 1, the in vivo expression of Xfoxi1a dynamically changes during gastrulation and neurulation, suggesting the possibility that Xfoxi1a plays distinct roles in a stage-dependent manner. To examine the crucial period of the Xfoxi1a function in the `non-neural' specification of the head ectoderm, we performed a temporally controlled misexpression experiment by using an inducible fusion protein construct comprising Xfoxi1a and the ligand-binding domain of the human glucocorticoid receptor(GR-Xfoxi1a) (Kolm and Sive,1995). GR-Xfoxi1a mRNA (50 pg/cell) was injected unilaterally into the animal blastomeres of eight-cell embryos. In the absence of dexamethasone (Dex), the expressions of Sox2, XK81, FoxD3 and Six1 appeared normal in the injected embryos at the mid-neurula stage(Fig. 6A,D,G,J). When 10 μM Dex was added to the medium at stage 11, a significant reduction of Sox2 was reproducibly observed on the GR-Xfoxi1a-injected side (53.4%, n=43; Fig. 6B), whereas embryos treated with Dex from stage 13 onwards exhibited no change in Sox2 expression (n=13; Fig. 6C). Consistent with this finding, expansion of the epidermal maker XK81 was seen in the neural plate of GR-Xfoxi1a-injected embryos treated with Dex from stage 11 onwards (35.7%, n=42; Fig. 6E) but not in those treated from stage 13 onwards (n=12; Fig. 6F). Similarly, the reduction of the neural crest and preplacodal markers FoxD3 and Six1 was observed in GR-Xfoxi1a-injected embryos treated with Dex from stage 11 onwards (55%, n=20; 74%, n=20,respectively; Fig. 6H,K), but not in those treated with Dex from stage 13 onwards (n=20, n=25, respectively; Fig. 6I,L). To eliminate the possibility that the GR-Xfoxi1a protein became degraded and ineffective by stage 13, we assessed the protein expression levels by western blot (flag-tagged GR-Xfoxi1a, which causes phenotypes indistinguishable from those caused by GR-Xfoxi1a,was used). As shown in Fig. 6M,the flag-tagged GR-Xfoxi1a products were detected at similar levels in the gastrula and neurula stages. Taken together, these observations suggest that Xfoxi1a promotes the epidermal specification by acting during the gastrula stages in vivo.

In mouse and zebrafish genetic studies, Foxi1-related genes have been shown to play essential roles in the formation of the head ectodermal derivatives (Hulander et al.,1998; Lee et al.,2003; Nissen et al.,2003; Riley and Phillips,2003; Solomon et al.,2003a). Mouse Foxi1 is required for normal development of the inner ear (Hulander et al.,1998). In zebrafish, foxi1 is the responsible gene for the hearsay mutant, in which the otic placode formation and jaw development are impaired (Riley and Phillips, 2003; Solomon et al., 2003a). Multiple Foxi1-related genes exist in each vertebrate species and can be classified into three subgroups according to their structures. Interestingly, mouse Foxi1, zebrafish foxi1 and Xfoxi1a/b/c belong to distinct subgroups: B, A and C, respectively (Solomon et al.,2003b). At present, it is not clear whether these subgroup factors function for ectodermal patterning in a distinct or redundant manner(Ohyama and Groves, 2004).

The present work has introduced a new role for a Foxi1 family member, Xfoxi1a/b, in the ventral specification of the early head ectoderm during gastrulation. During the mid- and late gastrula stages, Xfoxi1a/b is expressed in the anteroventral ectoderm. This gastrula expression is complementary to that of Sox2, indicating that all of the head ectoderm except for the neural plate tissues expresses Xfoxi1a (Fig. 1). Consistently, the loss-of-function study has demonstrated that Xfoxi1a/b is essential for the proper development of the non-neural domain of the head ectoderm (epidermis, cement gland, neural crest and placodes) and for suppression of the ectopic expansion of the neural plate(Fig. 3). Conversely,misexpression of Xfoxi1a induces ectopic keratin expression and suppresses Sox2 expression in the neural plate region(Fig. 5). This activity of Xfoxi1a is limited to the gastrula stage(Fig. 6). These results indicate that Xfoxi1a/b plays a pivotal role for the `neural versus non-neural' decision of the head ectoderm during gastrulation.

In the animal cap study, overexpression of Xfoxi1a inhibits neural differentiation caused by the injection of Chd(Fig. 5D) or dnBMPR(Fig. 5G), demonstrating that Xfoxi1a can exert an anti-neuralizing activity in the isolated ectodermal tissue. In addition, as the effect of dnBMPR is reversed by Xfoxi1a, it is likely that Xfoxi1a does not act upstream of Bmpr (although Xfoxi1a weakly induces Bmp4 in the animal cap; Fig. 5G, lane 4), but rather acts downstream of Bmpr or in parallel.

Interestingly, Xfoxi1a/b-MO injection (at the amount sufficient for keratin suppression and Sox2 expansion in vivo) suppresses the epidermal markers (XK81, Msx1 and Dlx3) but does not induce the neural marker Sox2 in the animal cap explant(Fig. 4). This suggests the possibility that the expansion of Sox2 expression by Xfoxi1a/b-MO in vivo (Fig. 3) depends on some additional factors, although Xfoxi1a/bregulates the epidermal fate determination in a tissue-autonomous manner. This idea is supported by our preliminary observation that the ectopic Sox2 expression in the embryo is always limited to the lateral region of the head ectoderm and not found in the more ventral region. One candidate factor may be Fgf signals, as a recent report(Delaune et al., 2005) has shown that Fgf signaling is required for anti-Bmp factors to induce ectopic Sox2 expression in the ventral-most part of the ectoderm.

The molecular mechanism underlying the regulation of ventral specification of the head ectoderm by Xfoxi1a/b remains elusive. Dlx3 and Msx1, which are required for non-neural ectodermal development(Suzuki et al., 1997; Feledy et al., 1999; Beanan and Sargent, 2000; Woda et al., 2003), may be among candidate mediators of Xfoxi1a/b activities as their expression is positively regulated by Xfoxi1a (Figs 3, 4, 5 and data not shown). The exact relationship between these factors and Xfoxi1a should be carefully analyzed along the temporal axis by using the combination of MOs and inducible constructs in future investigation. Our preliminary study has shown that Xfoxi1a-MO injection (which causes the expansion of Sox2 expression) does not significantly suppress Bmp4expression in the head region (data not shown). This suggests that the effect of Xfoxi1a-MO is not primarily mediated by the inhibition of Bmp4 expression, consistent with the dnBMPR study. In future, it will be important to systematically identify downstream target genes (and possible co-factors) of Xfoxi1a in the ventral specification.

This study has mainly focused on the role of the early Xfoxi1a/bfunction in the ventral specification of the head ectoderm during gastrulation. Later, by the mid-neurula stage, Xfoxi1a expression fades in the ventralmost area of the head ectoderm and becomes limited to the preplacodal region (Fig. 1). Although this late expression pattern of Xfoxi1a/b seems relevant to the requirement of the Foxi1 family genes for proper development of the head placodes of other species(Hulander et al., 1998; Lee et al., 2003; Nissen et al., 2003; Solomon et al., 2003a), the exact role of Xfoxi1a/b in late ectodermal patterning requires more careful interpretation. The intermediate head ectoderm (which gives rise to the neural crest, cement gland and preplacodal region) is complex and contains considerable heterogeneity even within the preplacodal region(Schlosser and Ahrens,2004).

An intriguing but slightly puzzling observation regarding the role in the regulation of intermediate ectodermal genes is that the phenotypes caused by Xfoxi1a overexpression are basically the same as those with the loss-of Xfoxi1a function; both result in suppression of FoxD3 and Six1 (Figs 3 and 5). This is in contrast to the situation of the regulation of Sox2 and XK81 by Xfoxi1a/b, in which gain- and loss-of-function experiments show the opposite phenotypes (Figs 3 and 5). One interpretation of this discrepancy is that Xfoxi1a affects the development of the intermediate head ectoderm in a non-cell-autonomous fashion; both augmentation and attenuation of Xfoxi1a may interfere with the interactions between the neural plate and epidermis, which are required for the proper differentiation of the intermediate ectoderm(Dickinson et al., 1995; Selleck and Bronner-Fraser,1995; Mancilla and Mayor,1996; LaBonne and Bronner-Fraser, 1999; Glavic et al., 2004). This idea is in agreement with the largely non-overlapping expression patterns of Xfoxi1a and Six1 or FoxD3 in the mid-neurula (Fig. 1H and data not shown). Alternatively, the role of Xfoxi1a could be cell-autonomous, given that Xfoxi1a is expressed throughput the Sox2-negative head ectoderm (which should include the intermediate ectoderm) at the mid-gastrula stage(Fig. 1D), unlike at the mid-neurula stage (Fig. 1G). In this case, the gain- and loss-of-function phenotypes in the intermediate head ectoderm should be caused by distinct mechanisms.

The study with GR-Xfoxi1a suggests that the inhibitory effects of Xfoxi1a on the intermediate ectodermal markers are related to the Xfoxi1a activity before the late gastrula stage(Fig. 6). FoxD3 and Six1 expressions at the neurula stage are clearly suppressed when GR-Xfoxi1a-injected embryos are treated with Dex from stage 11 but not from stage 13 (Fig. 6G-L). However, as the neural plate marker Sox2 is affected in a similar manner (Fig. 6A-C), it remains to be clarified whether the suppression of FoxD3 and Six1 is directly or indirectly caused by Xfoxi1a.

Early Xfoxi1a expression in the anteroventral ectoderm (stage 12)is strongly influenced by Bmp and Wnt signals(Fig. 2). Working as upstream regulators, Bmp signaling positively controls Xfoxi1a expression in the ectoderm whereas Wnt signaling has a negative effect. The role of Bmp in the DV patterning of the cephalic non-neural ectoderm described here is in agreement with a previous report (Wilson et al., 1997). Although Wnt signals are known to be crucial for the AP patterning of the CNS (and of the mesoderm), experimental knowledge about their roles in the AP patterning of the non-neural ectoderm has been limited. Both our in vivo and in vitro analyses(Fig. 2) have shown that Wnt signaling suppresses Xfoxi1a, indicating a direct regulatory role of Wnts in the determination of the cephalic non-neural ectoderm. A consistent effect of Wnt signals on Xfoxi1a expression is also found in the neurula embryo (Fig. 2I,J).

By contrast, the late Xfoxi1a expression at the neurula stage responds to Bmp4 in a slightly different manner. Although Xfoxi1aexpression is also suppressed by Chd, injection of the Bmp-expression plasmid does not upregulate Xfoxi1a expression at this stage(Fig. 2G,H). This may be explained by the stage-dependent difference of the Xfoxi1a expression domains. In contrast to the wide expression domain in the anteroventral ectoderm at the late gastrula stage, Xfoxi1a expression at the mid-neurula stage is limited to a band in the head ectoderm, which is narrow in the dorsoventral direction (Fig. 1F). Therefore, it is likely that the late Xfoxi1aexpression requires some additional positional information other than the ventralizing signal of Bmp4.

The present study suggests a role of Xfoxi1a/b as an important player that mediates early patterning signals (such as Bmp and Wnt) in the ventral specification of the head ectoderm. Further studies of the regulation and function of Xfoxi1a/b should improve our understanding of the molecular mechanisms that underlie the complex multiple-step patterning of the vertebrate head ectoderm.

We are grateful to Dr R. Ladher and members of the Sasai laboratory for comments on this work, and to Dr T. D. Sargent for the Dlx3, XK81 and Xag marker plasmids. This work was supported by grants-in-aid (Y.S.) from MEXT,the Kobe Cluster Project and the Leading Project.