Non-cell-autonomous role for Cripto in axial midline formation during vertebrate embryogenesis

The emergence of axial mesendoderm from the anterior primitive streak is a crucial event during vertebrate gastrulation that requires precise control of cell fate determination in a limited spatial and temporal context. The recruitment and differentiation of axial mesendoderm progenitors result in formation of the node, prechordal plate, notochordal plate and a population of anterior definitive endoderm that contributes to the foregut(Kinder et al., 2001; Tam and Beddington, 1987). Previous studies have established that several components of the Nodal signaling pathway are essential for axial mesendoderm formation (e.g. Chu et al., 2004; Dunn et al., 2004; Hoodless et al., 2001; Lowe et al., 2001; Norris et al., 2002; Vincent et al., 2003; Yamamoto et al., 2001). Nodal signaling is also believed to play a central role in specifying axial mesoderm subtypes, as differential levels of Nodal activity can specify prechordal mesoderm versus notochord during gastrulation(Agius et al., 2000; Gritsman et al., 2000; Lowe et al., 2001; Norris and Robertson, 1999; Thisse et al., 2000; Vincent et al., 2003). Furthermore, high levels of Nodal pathway activity are also essential for formation of the definitive endoderm (reviewed by Ober et al., 2003; Tam et al., 2003).

Molecular genetic studies in several vertebrate systems have identified EGF-CFC proteins as essential components of the Nodal signaling pathway, and have demonstrated their ability to act in cis as co-receptors for Nodal ligands (Gritsman et al.,1999; Kumar et al.,2001; Reissmann et al.,2001; Yan et al.,2002; Yeo and Whitman,2001). EGF-CFC proteins are small extracellular proteins that contain a divergent epidermal growth factor (EGF) motif and a novel conserved cysteine-rich domain (termed the CFC motif) (reviewed by Shen and Schier, 2000), and they are attached to the cell membrane through a glycosyl-phosphatidylinositol(GPI) linkage (Minchiotti et al.,2000). Two EGF-CFC genes, Cripto(Tdgf1) and Cryptic (Cfc1 - Mouse Genome Informatics), have been identified in mammals; other family members include chick Cripto, frog FRL-1 and zebrafish one-eyed pinhead (oep). In the case of oep, cell transplantation experiments have shown its cell-autonomy in prechordal plate, floor plate, and endoderm (Gritsman et al.,1999; Schier et al.,1997; Strahle et al.,1997). These data, together with the absence of phenotypic consequences following oep overexpression in zebrafish embryos(Zhang et al., 1998), are consistent with a strict role for Oep as a cis-acting co-receptor for Nodal.

Zygotic oep mutants lose axial mesendoderm and display associated ventral midline and forebrain defects(Schier et al., 1997). Despite a relatively low overall level of sequence conservation, all EGF-CFCfamily members appear to have functionally similar activities in assays for phenotypic rescue of oep mutants(Zhang et al., 1998). However,it has been unclear to date whether, like Oep, mammalian EGF-CFC proteins are required in axial midline formation and forebrain patterning, and whether they function as cis-acting co-receptors for Nodal in this context. Intriguingly,by contrast with Oep, studies of Cripto function have raised the possibility that it can function as a secreted trans-acting signaling factor(Bianco et al., 2003; Kannan et al., 1997; Minchiotti et al., 2001; Parisi et al., 2003; Yan et al., 2002). Furthermore, a previous low-resolution chimera analysis has suggested potential non-cell-autonomy for Cripto(Xu et al., 1999), although the absence of cell-marking in this experiment precluded analysis of specific requirements for Cripto function during embryogenesis.

In the studies described below, we show that Cripto, like other Nodal pathway components, is required for axial mesendoderm and definitive endoderm formation in mouse embryogenesis. However, we find that this requirement for Cripto is non-cell-autonomous, and provide evidence that Cripto can act as a soluble trans-acting factor in vivo. Complementary gain-of-function approaches in chick embryos reveal effects of exogenous soluble Cripto protein on differentiation of chick node/head process mesoderm cells, suppressing posterior axial mesoderm fates and promoting anterior mesendodermal fates. Our findings potentially resolve long-standing discrepancies concerning the mechanisms of Cripto function, and suggest that a GPI-linked protein can act as a trans-acting signal.

To generate the Cripto^3loxP targeting construct, a 5.5 kb genomic BamHI fragment containing exon 6 was cloned into the 3′ multiple cloning site of ploxPNT(Shalaby et al., 1995),followed by cloning of a 5′ arm corresponding to a 3.1 kb HindIII-BamHI partial digest fragment containing exons 1-5;a loxP site was inserted into the BamHI site in intron 2. TC1 embryonic stem (ES) cell clones were screened for homologous recombination by Southern blotting with both 5′ and 3′ probes, with positive clones obtained at an efficiency of 25% (n=20).

We used the EIIa-Cre transgenic line to obtain both partial and complete deletions in vivo, taking advantage of the low Cre activity in the germline of EIIa-Cre males (Xu et al., 2001). We obtained all three possible deletion products with approximately equal frequency in the progeny of C57BL/6 wild-type females× EIIa-Cre; Cripto^3loxP/+ males, including the desired Cripto^del and Cripto^floxalleles. (Formal allele names are as follows: Cripto^lacZ=Cripto^tm1Mms; Cripto^3loxP=Cripto^tm2Mms;Cripto^del=Cripto^tm2.1Mms;Cripto^flox=Cripto^tm2.2Mms). PCR primers used for genotyping were: Cripto^3loxP(forward) 5′-CCT CCC AAG TTC ACT ACC AAA TCT-3′, (reverse)5′-TCC CCA CCA TCC ACC ACC AAG TAG-3′; Cripto^del (forward) 5′-AGC CAT CTC ACC AGC CTT CA-3′, (reverse) 5′-CAT CTG GGA CAT GCC CAC TA-3′; Cripto^lacZ (forward) 5′-CCA TCC CCT GCC CGT CTA CAC G-3′, (reverse) 5′-GTC ACG CAA CTC GCC GCA CAT-3′.

Whole-mount in situ hybridization and β-galactosidase staining were performed as described (Ding et al.,1998). Genotypes of 6.5 days-post-coitum (dpc) embryos were determined by PCR using genomic DNA from cultured extraplacental cone(Ang and Rossant, 1994); for 7.5 dpc embryos and older, extra-embryonic tissues following in situ hybridization were lysed for DNA extraction and PCR. Following whole-mount in situ hybridization, embryos were embedded in OCT and 6 μm cryosections were counterstained with Methyl Green, or left unstained. Histological analysis was performed by hematoxylin and eosin staining of paraffin sections of embryos fixed in 10% formalin. Mouse probes for in situ hybridization were: Cer1 (Belo et al.,1997); Cripto (Ding et al., 1998); En1(Davis and Joyner, 1988); Fgf8 (Crossley and Martin,1995); Foxa2 (Ang and Rossant, 1994); Gsc(Belo et al., 1997); Hex (Hhex) (Thomas et al., 1998); Shh(Echelard et al., 1993); and Six3 (Oliver et al.,1995). Chick probes were Gsc(Hume and Dodd, 1993) and Hex (Crompton et al.,1992).

Chimeras were generated by morula aggregation(Nagy et al., 2003), using 4-8 cell embryos from pregnant C57BL/6 wild-type mice and from a Rosa26/Rosa26; Cripto^null/+ × Cripto^lacZ/+ cross (here we refer to the Gt(ROSA)26Sor gene-trap allele as Rosa26). Embryos were recovered at 7.5-9.5 dpc, and genotypes determined by PCR analysis of DNA extracted from extra-embryonic tissues. Whole-mount embryos were stained forβ-galactosidase activity, and cryosections were counterstained with Nuclear Fast Red (Vector Laboratories). In a second method of chimera analysis, marked ES cell lines were established from blastocysts obtained by a Rosa26/Rosa26; Cripto^del/+ × Cripto^lacZ/+ cross, using standard methods(Nagy et al., 2003). The resulting ES colonies were genotyped by PCR, and a Rosa26/+;Cripto^lacZ/del cell line and a Rosa26/+;Cripto^lacZ/+ line were established. Chimeras were generated by injecting approximately 5-10 ES cells into C57BL/6 blastocysts.

Embryos and explants were assayed according to standard techniques(Placzek et al., 1993; Placzek et al., 1990). For explant culture experiments, posterior head process mesoderm or Hensen's node explants were grown for 20 hours in Optimem (Invitrogen) containing 2% fetal calf serum, 1% penicillin/streptomycin, and 1% L-glutamine, either alone or with addition of 10-fold concentrated 293T culture supernatants. 293T cells were either untransfected or transfected with full-length mouse Cripto, which results in production of active Cripto protein in conditioned medium (Minchiotti et al.,2000; Yan et al.,2002). The small molecule inhibitor SB-431542 was used as described previously (Inman et al.,2002). For in-ovo experiments, beads were grafted at Hamburger-Hamilton (HH) stage 3-4 and embryos developed until HH stage 8-9. Explants and embryos were fixed in 4% paraformaldehyde for 2 hours and embedded in OCT for cryosectioning. Sections were processed for immunofluorescence detection using the 3B9 monoclonal antibody [1:10 dilution from culture supernatant (Placzek et al.,1990)] or the Shh monoclonal antibody(Ericson et al., 1996) and a Cy3-labeled anti-mouse secondary antibody (Jackson ImmunoResearch; 1:200 dilution), as described previously(Placzek et al., 1993).

Real-time quantitative RT-PCR was performed using the ABI Prism 7700 sequence detection system (Applied Biosystems). RNA was standardized usingβ-actin amplification as an internal control. Primers used were: Gsc (forward) 5′-AAA AGA CGG CAC CGG ACT ATC-3′,(reverse) 5′-TCG TTT CCT GGA AGA GGT TTT C-3′, and the probe 5′-CAC TGA CGA GCA GCT CGA AGC GC-3′ (labeled with the reporter dye FAM on the 5′ nucleotide and the quenching dye TAMRA on the 3′nucleotide). Fgf10 (forward) 5′-ATG ACC TCG GCC AGG ACA T-3′, (reverse) 5′-TGA TTG TAG CTC CGC ACG TG-3′, and probe 5′-CTG TCC CCG GAG GCC ACC AA-3′ labeled with the reporter Yakima Yellow on the 5′ nucleotide and the quenching dye Eclipse Dark Quencher on the 3′ nucleotide. β-actin (forward) 5′-GGT CAT CAC CAT TGG CAA TG-3′, (reverse) 5′-CCC AAG AAA GAT GGC TGG AA-3′and the fluorogenic probe 5′-TTC AGG TGC CCC GAG GCC CT-3′ labeled with the reporter dye Yakima Yellow on the 5′ nucleotide and the quenching dye Eclipse Dark Quencher on the 3′ nucleotide. Probes were supplied by Eurogentec. Relative quantification of Gsc and Fgf10 mRNA was calculated using the standard curve method.

We previously described the expression pattern of Cripto in the pre-gastrulation epiblast and in the primitive streak and nascent mesoderm(Ding et al., 1998). At later stages of gastrulation, we observe Cripto expression in the embryonic mesoderm and node, as well as in the axial mesendoderm(Fig. 1A-C). However, this expression of Cripto disappears during neural plate stages (data not shown).

To determine the cell types that derive from this transient expression of Cripto, we took advantage of the Cripto^lacZ null allele, in which the lacZ reporter gene is under the control of the Cripto promoter (Ding et al.,1998). This knock-in allele recapitulates the pattern of endogenous Cripto expression, except for a temporal delay that probably corresponds to the perdurance of β-galactosidase(Ding et al., 1998). In Cripto^lacZ/+ heterozygotes at late neural plate stages, we found β-galactosidase staining throughout the mesoderm and definitive endoderm, with strongest staining in the axial mesendoderm(Fig. 1D,E). Taken together,these data indicate that Cripto-expressing cells give rise to axial mesendoderm and anterior definitive endoderm.

Cripto null mutants display early embryonic lethality due to a failure in anterior-posterior axis positioning before gastrulation(Ding et al., 1998). Therefore, to investigate subsequent functions of Cripto in embryogenesis, we have generated novel hypomorphic (partial loss-of-function),conditional, and null alleles for Cripto(Fig. 1F-H). In our targeting strategy, a `3-loxP' construct containing a `floxed' (flanked by loxP sites) PGK-neo selection cassette was inserted at the Cripto locus. Following germline transmission of this allele, mice carrying this Cripto^3loxP allele were crossed with Ella-Cre transgenic mice to produce all three possible derivatives from Cre-loxP mediated recombination (Xu et al., 2001). This strategy resulted in the generation of a new null allele corresponding to the deletion of exons 3, 4 and 5(Cripto^del) (Fig. 1F), as well as a conditional floxed allele(Cripto^flox) (data not shown). Both homozygous Cripto^del and heteroallelic Cripto^lacZ/del embryos displayed a phenotype identical to that of Cripto^lacZ homozygotes (data not shown).

As the intronic insertion of selection cassettes with strong transcription termination sequences frequently leads to the generation of a hypomorphic allele (Meyers et al., 1998),we performed genetic crosses to determine whether the Cripto^3loxP allele might be hypomorphic. Although homozygous Cripto^3loxP/Cripto^3loxP mice can be recovered as adults at Mendelian ratios(Table 1A), we observed that progeny from crosses of Cripto^3loxP/+ and Cripto^lacZ/+ mice were significantly under-represented for the Cripto^3loxP/lacZ genotype(Table 1B). The surviving mice with this genotype can reach adulthood and breed successfully; gross and histological examination of these mice has not yet revealed any phenotypic abnormality (J.C. and M.M.S., unpublished). Because Cripto^3loxP/del embryos displayed an identical phenotype to Cripto^3loxP/lacZ embryos (data not shown), we refer to both Cripto^3loxP/lacZ and Cripto^3loxP/del embryos as having a Cripto^3loxP/null genotype.

Morphological and histological examination of Cripto^3loxP/null embryos at 8.25-9.5 dpc showed that approximately 56% (n=52) displayed a broad spectrum of phenotypes that recapitulate many of the characteristic features of human holoprosencephaly (Fig. 2). The most severely affected embryos showed anterior truncations as well as fusions of anterior somites across the midline, indicating the absence of notochord(Fig. 2A-C). Less-affected embryos displayed anterior-restricted phenotypes, including cyclopia(Fig. 2D-G) and variable forebrain and midbrain reductions at 9.0-11.5 dpc(Fig. 2D-I). Finally,histological analyses of the brains of severely affected Cripto^3loxP/null embryos showed a single prosencephalic vesicle (Fig. 2J,K), consistent with the absence of ventral patterning signals from the axial mesendoderm.

Next, we examined the expression of specific tissue markers in severely affected Cripto^3loxP/null embryos to ascertain the basis for these phenotypic defects. Sonic hedgehog (Shh) is expressed in ventral midline structures and gut endoderm at 8.25 dpc, but was often greatly reduced or absent in Cripto^3loxP/nullembryos (71%, n=7) (Fig. 3A). Similarly, Hex is a marker of anterior definitive endoderm at 7.5 dpc, but was often undetectable in Cripto^3loxP/null mutants (40%, n=15)(Fig. 3B); by contrast, Hex expression in the anterior visceral endoderm at 6.5 dpc was unaffected (n=5; data not shown). Analyses of neural markers revealed that severely affected Cripto^3loxP/null embryos retain expression of the midbrain-hindbrain marker En1 (n=4)(Fig. 3C). However, Cripto^3loxP/null embryos often displayed truncation of the rostral forebrain, as shown by loss of anterior expression for Six3(67%, n=9) and Fgf8 (63%, n=8)(Fig. 3D-F). The loss of Hex expression, in particular, suggests that the forebrain defects may represent a secondary consequence of the loss of axial mesendoderm derivatives, notably the anterior definitive endoderm and prechordal plate(reviewed by Kiecker and Niehrs,2001; Wilson and Houart,2004).

Consistent with this interpretation, histological and marker analyses of Cripto^3loxP/null embryos during gastrulation stages demonstrated severe defects in the formation of axial mesendoderm and definitive endoderm. In wild-type embryos, the prechordal plate is evident from the apposition of axial mesendoderm to anterior ectoderm at the midline(Fig. 2L), whereas an intact prechordal plate did not form in Cripto^3loxP/null mutants(Fig. 2M,N). At head-fold stages, Shh is expressed in the node, notochordal plate and prechordal plate, but in Cripto^3loxP/null mutants was either absent or occasionally observed in scattered patches of mesodermal cells that were displaced from the midline (53%, n=19)(Fig. 4A-D). Similar results were obtained with the axial midline markers Foxa2 (HNF3β) (36%, n=14) (Fig. 4E-H) and chordin (Chrd, 67%, n=10/15) (data not shown). Expression of goosecoid (Gsc) marks prechordal plate, adjacent midline neuroectoderm and foregut endoderm, but was abolished in approximately half of Cripto^3loxP/null mutants (50%, n=6)(Fig. 4I-L). Finally,expression of cerberus 1 homolog (Cer1) showed the nearly complete loss of definitive endoderm in strongly affected Cripto^3loxP/null embryos (50%, n=4)(Fig. 4M-P), consistent with the results with Hex (Fig. 3B).

To determine whether the requirement for Cripto in axial mesendoderm and definitive endoderm is cell-autonomous or non-cell-autonomous,we performed genetic mosaic analysis using two distinct and complementary methods to generate marked chimeras of wild-type and Cripto null mutant cells. In the first approach, we generated chimeric embryos using morula aggregation between wild-type embryos and morulas generated from a cross between mice carrying two different Cripto null mutant alleles(Fig. 5A)(Rivera-Perez et al., 1999). In this scheme, chimeric embryos are genotyped retrospectively using visceral yolk sac DNA, while the use of two Cripto alleles that can be distinguished by PCR allows for the unambiguous detection of chimeras containing null mutant cells (Fig. 5B). Cells derived from progeny of the Criptoheterozygote cross are marked by the Rosa26 gene-trap allele, which ubiquitously expresses β-galactosidase activity(Friedrich and Soriano, 1991; Zambrowicz et al., 1997). One principal advantage of this approach is that control chimeras are generated together with the desired mutant chimeras, thereby reducing any potential experimental bias in their analysis.

Using morula aggregation, we generated 31 total chimeras, including 9 Cripto^lacZ/del↔ +/+ chimeras, which were examined at head-fold stages or at 9.5 dpc. These embryos were sectioned and assessed for their extent of chimerism as measured by the percentage ofβ-galactosidase stained cells, corresponding to expression of the Rosa26 marker (Table 2). In control chimeras (Cripto^+/+↔ +/+, Cripto^lacZ/+ +/+, or Cripto^del/+↔+/+), the β-galactosidase-stained cells were equally distributed among all germ layers and tissue types (data not shown). Unexpectedly, we found that low-grade (<50% contribution of marked cells) and medium-grade (∼50%contribution) Cripto^lacZ/del↔ +/+ chimeras were phenotypically wild type (Fig. 6A-J; data not shown). In medium-grade Cripto^lacZ/del↔ +/+ chimeras at head-fold stages,there was no discernible bias in contribution of Cripto null mutant(blue) cells to all three germ layers (Fig. 6B,D). In particular, β-galactosidase-staining cells could be found in the notochordal plate and prechordal plate, which appeared morphologically normal (Fig. 6C,E). At 9.5 dpc, Cripto^lacZ/del↔ +/+chimeras were also phenotypically wild type, with abundant contribution of marked cells to notochord, floor plate and foregut endoderm(Fig. 6F-J). Surprisingly, only a proportion of higher-grade (>50% contribution) Cripto^lacZ/del↔ +/+ chimeras displayed abnormal phenotypes, which resembled those of hypomorphic Cripto^3loxP/null mutants(Fig. 6K,L); interestingly,there was abundant mesodermal contribution of marked cells even in the most severely affected chimeras (Fig. 6L).

In a second approach for chimera analysis, we used marked Cripto^lacZ/del ES cells, which were derived from crosses of Rosa26/+; Cripto^del/+ mice with Cripto^lacZ/+ heterozygotes (Materials and methods) for microinjection into unmarked wild-type host blastocysts. In analysis of seven sectioned embryos (40-70% chimerism) at head-fold stages, we found an identical unbiased contribution of Cripto null mutant cells to derivatives of all three germ layers, including the prechordal plate, midline neuroectoderm and foregut endoderm (Fig. 6M-O). Thus, both methods of chimera analysis demonstrate that Cripto acts non-cell-autonomously in the axial mesendoderm and anterior definitive endoderm.

The non-autonomous function of Cripto suggests that Cripto might act as a secreted trans-acting factor in the differentiation of anterior mesendoderm cell types, a process known to be dosage-sensitive for Nodal pathway activity (Agius et al.,2000; Dunn et al.,2004; Gritsman et al.,2000; Vincent et al.,2003). Consistent with this possibility, chick Cripto(also known as CFC) is expressed in the anterior primitive streak/Hensen's node at HH stages 3 and 4, as well as in notochord and prechordal plate at stages 5 and 6 (Colas and Schoenwolf, 2000; Schlange et al., 2001). To determine whether soluble Cripto protein might affect the differentiation of anterior mesendodermal cell types, we used chick embryos to assay the activity of Cripto-containing conditioned medium(Minchiotti et al., 2000; Yan et al., 2002).

Initially, we examined cell differentiation in node or head process mesoderm explants cultured in control medium, or medium conditioned by untransfected 293T cells. When HH stage 4^--4 Hensen's node explants were cultured under these conditions (Fig. 7A), Gsc-expressing prechordal mesoderm cells, Hex-expressing anterior endoderm cells and 3B9-expressing notochord cells differentiated from the explanted node (n=20; Fig. 7B-D, data not shown). By contrast, when posterior head process mesoderm from HH stage 5 embryos was cultured (Fig. 7E), neither Gsc-expressing prechordal mesoderm cells nor Hex-expressing anterior endoderm cells was detected; however, differentiating notochord cells marked by the cell surface marker 3B9 were detected (n=30)(Fig. 7F-H, data not shown).

By contrast, addition of Cripto-conditioned medium evoked a very distinct pattern of differentiated cell types within node and head process mesoderm explants. When Cripto-conditioned medium was added to HH stage 4^--4 Hensen's node explants, the differentiation of 3B9-expressing notochord cells was prevented (80%, n=20) (Fig. 7L), but robust differentiation of Gsc- and Hex-expressing cells was observed (100%, n=16)(Fig. 7J,K). This suggests that soluble Cripto protein can promote the differentiation of more anterior mesendodermal cell types (prechordal mesoderm and anterior endoderm), at the expense of posterior notochord cells. In support of this interpretation, when HH stage 5 posterior head process mesoderm was exposed to Cripto, Gsc-expressing prechordal mesoderm cells were detected in the central component of the explant, while Hex-expressing anterior endodermal cells were detected in peripheral regions of the explant (80%, n=10)(Fig. 7N,O). Concomitantly, the differentiation of 3B9-expressing notochord cells was prevented (80%, n=10) (Fig. 7P).

To examine whether Cripto can rapidly elicit a change in cellular fate, we used a sensitive real-time quantitative RT-PCR assay to examine the loss of notochord and induction of prechordal mesoderm properties in HH stage 5 posterior head process mesoderm. Explants were cultured in the absence or presence of Cripto-conditioned medium, and simultaneously monitored for expression of the prechordal mesoderm marker Gsc and the early notochord marker Fgf10. After a 4-hour culture period, Cripto addition caused an increase in Gsc expression and a concomitant decrease in Fgf10 expression (n=7)(Fig. 7I,M).

We next established whether Cripto could similarly elicit a change in fate of axial mesodermal cells in vivo, promoting a prechordal mesoderm fate at the expense of notochord cell fate. Beads soaked in medium containing Cripto protein, or control medium, were implanted adjacent to Hensen's node of HH stage 4^--4 chick embryos (n=6 each; Fig. 7Q), and the embryos were allowed to develop in ovo to HH stage 8-9(Fig. 7U). In control-operated embryos, axial mesoderm displayed the normal phenotypic properties of notochord cells, as no expression of Gsc was detected(Fig. 7R), but the cells co-expressed Shh and 3B9 (Fig. 7S,T). By contrast, in embryos exposed to Cripto-conditioned medium, axial mesoderm at an equivalent rostrocaudal level(Fig. 7U) appeared to be anteriorized. Although situated adjacent to somites, axial mesoderm cells expressed the prechordal mesoderm marker, Gsc(Fig. 7V), as well as Shh(Fig. 7W), but expression of 3B9 was eliminated (Fig. 7X).

Finally, we examined whether the ability of exogenous Cripto to effect fate changes in axial mesoderm is dependent upon Nodal signaling. To do so, we utilized SB-43152, a small molecule inhibitor of the type I receptors ALK4/5/7(Inman et al., 2002), to antagonize the Nodal receptors ALK4 and ALK7(Reissmann et al., 2001; Yan et al., 2002; Yeo and Whitman, 2001). Culture of HH stage 5 posterior head process mesoderm with SB-43152 alone did not result in the loss of the notochord marker 3B9, appearing similar to control explants (n=9) (Fig. 7Y,AA). However, SB-43152 eliminated the ability of Cripto to anteriorize axial mesoderm, as culture of HH stage 5 head process mesoderm with Cripto resulted in loss of 3B9 and induction of Gsc and Hex (Fig. 7I,M-P,Z),while culture with Cripto and SB-43152 resulted in the sustained expression of 3B9 (80%, n=20) and no induction of Gsc or Hex(n=6 each) (Fig. 7BB,data not shown).

In summary, our findings in these gain-of-function assays show the ability of Cripto to act as a soluble transacting factor, eliciting effects consistent with increased Nodal signaling activity. These data strongly suggest that the non-autonomous requirement for Cripto in axial mesendoderm formation reflects the ability of Cripto protein to act intercellularly in vivo.

Our analyses demonstrate that Cripto is transiently expressed in precursors of the axial mesendoderm and definitive endoderm, and is essential for formation and differentiation of these tissues. This requirement for Cripto is non-cell-autonomous, as null mutant cells readily contribute to the axial mesodendoderm in chimeric mouse embryos, and soluble Cripto protein elicits Nodal-dependent effects on axial mesoderm in chick embryos. In conjunction with analyses of other Nodal pathway components, our findings provide evidence that Cripto protein can act as an intercellular transacting factor in vivo.

The requirement of Cripto in axial mesendoderm and definitive endoderm strongly indicates that Cripto is required for Nodal signaling in these tissues and/or their progenitors. Indeed, null and/or hypomorphic mouse mutants for several Nodal pathway components display similar phenotypes to those of Cripto hypomorphs, including mutants for Nodal(Lowe et al., 2001; Norris et al., 2002), the type II activin receptors ActRIIA and ActRIIB(Song et al., 1999), the intracellular signal transducers Smad2 and Smad3(Dunn et al., 2004; Heyer et al., 1999; Vincent et al., 2003), Smad4(Chu et al., 2004), and the winged-helix transcription factor FoxH1(Hoodless et al., 2001; Pogoda et al., 2000; Sirotkin et al., 2000; Yamamoto et al., 2001). Nonetheless, we note that Cripto may also have Nodal-independent functions. For example, the EGF-CFC protein Oep has been proposed to act independently of Nodal in regulation of cell motility in the zebrafish blastoderm(Warga and Kane, 2003). In addition, cell culture studies have shown that Cripto can block Activin signaling in the absence of Nodal activity(Adkins et al., 2003; Gray et al., 2003). Finally, a recent study has found that the Xenopus EGF-CFC protein FRL-1 can mediate signaling by Wnt11 in the canonical Wnt/β-catenin pathway during dorsal axis formation (Tao et al.,2005).

Given the transient expression of Cripto during gastrulation, and the early appearance of axial midline defects in Cripto hypomorphs,the requirement for Cripto function (and hence Nodal signaling) is likely to occur in axial mesendoderm progenitors within or shortly after their emergence from the anterior primitive streak. This temporal requirement may also coincide with the timing of Nodal signaling from the axial mesoderm to the overlying neuroectoderm in floor plate induction(Patten et al., 2003)(reviewed by Placzek and Briscoe,2005). Although our investigation of the Cripto hypomorph phenotype does not directly address whether Cripto function is also required in the floor plate, the chimera analysis suggests that such a role for Cripto would also be non-cell-autonomous(Fig. 6H,J; data not shown).

Our analyses of mutant mice show that reductions of Criptoactivity affect rostral midline structures (prechordal plate) in preference to caudal structures (notochord), as indicated by the spectrum of holoprosencephaly phenotypes in Fig. 2, while exposure of chick cells to exogenous Cripto both in vitro and in vivo increases prechordal mesoderm at the expense of notochord. These results complement previous observations that levels of Nodalactivity specify the rostrocaudal identity of axial mesoderm(Agius et al., 2000; Gritsman et al., 2000; Lowe et al., 2001; Norris and Robertson, 1999; Thisse et al., 2000; Vincent et al., 2003). More generally, our findings demonstrate that reduction of Cripto activity can lead to a `pinhead' phenotype that resembles that of zygotic oepmutants (Schier et al., 1997). This observation, together with the identification of a Criptoloss-of-function mutation in a human patient with midline forebrain defects(de la Cruz et al., 2002),provides additional support for the evolutionary conservation of EGF-CFC functions.

In principle, the experimental observation of non-cell-autonomy implies that the gene product, or one of its major regulatory targets, mediates a cell-cell signaling event. In some cases, genes encoding proteins with intracellular functions, such as transcription factors, can nonetheless display non-cell-autonomy (e.g. Rhinn et al., 1999). However, analysis of the Nodal pathway component FoxH1 has shown that this transcription factor is required cell-autonomously for axial mesendoderm and definitive endoderm formation(Hoodless et al., 2001). Thus,the upstream extracellular component Cripto acts non-cell-autonomously in the Nodal pathway, but the downstream transcription factor FoxH1 is cell-autonomous (Fig. 8A). Moreover, chimera analysis of Smad2 has demonstrated its cell-autonomous requirement in definitive endoderm(Tremblay et al., 2000). Based on this pathway relationship (Fig. 8A), we conclude that it is Cripto itself, not one of its regulatory targets, that can act in trans as an intercellular mediator of Nodal signaling.

The intercellular activity of Cripto may reflect its ability to be released from the cell surface in an active form, as we have previously shown(Yan et al., 2002). As Cripto is a GPI-linked protein, several possible mechanisms may explain its release from the cell surface. These include protein shedding due to cleavage of the GPI anchor (e.g. Kondoh et al.,2005; Metz et al.,1994), or formation of membrane vesicles (argosomes) that can traverse neighboring cells (Greco et al.,2001). Moreover, there are several precedents for shedding of GPI-linked proteins under physiological circumstances, such as the regulated proteolytic cleavage of ephrin-A2 during neuronal axon repulsion(Hattori et al., 2000), or the release of the glypican Dally-like, which modulates Wingless signaling in Drosophila (Kreuger et al.,2004). However, we note that our findings show only that Cripto is capable of acting intercellularly under physiological circumstances in vivo,but do not necessarily imply that such an activity is utilized during wild-type development. For example, GFRα is a GPI-linked co-receptor that has been proposed to act in trans to provide a response to GDNF family ligands in cells that express the Ret receptor but not GFRα (e.g. Mikaels-Edman et al., 2003; Paratcha et al., 2001), but this model has been challenged by a recent study showing the absence of Ret-independent GFRα function in vivo(Enomoto et al., 2004).

Our data may provide a unifying basis for two bodies of literature on Cripto activity and function that have previously seemed difficult to reconcile. While work in zebrafish, frog and mouse systems has supported a role for EGF-CFC proteins as cis-acting co-receptors for Nodal ligands, there is also ample evidence for trans-acting soluble activities of Cripto, particularly in cell culture. Notably, work from the Salomon laboratory has shown that recombinant Cripto protein or peptide fragments can elicit specific responses in cell culture, for example resulting in activation of the MAPK pathway (Bianco et al., 2002; Bianco et al.,1999; Bianco et al.,2003; De Santis et al.,1997; Ebert et al.,1999; Kannan et al.,1997). In some cases, these cellular responses are due to activation of the Nodal pathway, whereas other Cripto activities appear to be Nodal-independent (Bianco et al.,2002; Bianco et al.,2003). Other investigators have found that truncated Cripto protein lacking the C-terminal GPI anchor can rescue the cardiac differentiation defects of Cripto-deficient ES cells(Parisi et al., 2003), and that injection of truncated Oep protein into the syncytial yolk layer of zebrafish embryos will rescue the oep mutant phenotype(Minchiotti et al., 2001). Finally, our studies using a reconstituted Nodal signaling assay in mammalian cells have shown that the ability of soluble Cripto to signal in this system is Nodal- and FoxH1-dependent (Yan et al.,2002). All of these data support a model in which Cripto possesses intrinsic activity as a non-membrane-associated protein that can mediate Nodal signaling in trans (Fig. 8B,C).

Consistent with possible trans-acting functions, several studies have reported phenotypes associated with overexpression/ectopic expression of EGF-CFC genes in vivo. For example, overexpression of wild-type human Cryptic - but not mouse Cryptic - in zebrafish leads to developmental delays and abnormal gastrulation movements(Bamford et al., 2000), while overexpression of secreted forms of oep can promote dorsoanterior development (Kiecker et al.,2000). In chick embryos, implantation of cells expressing human Cryptic, mouse Cryptic or chick Cripto on the right side of Hensen's node leads to randomization of cardiac looping(Schlange et al., 2001). Finally, overexpression of frog FRL-1 in Xenopus animal caps results in neural induction in the absence of mesoderm formation(Kinoshita et al., 1995; Yabe et al., 2003; Yokota et al., 2003).

By contrast, other studies have shown the absence of a gain-of-function phenotype following overexpression of wild-type oep in fish(Bamford et al., 2000; Gritsman et al., 1999; Zhang et al., 1998). Moreover,rigorous cell transplantation assays have established the strict cell-autonomy of oep function in several tissues(Gritsman et al., 2000; Schier et al., 1997; Strahle et al., 1997; Warga and Kane, 2003). These opposite findings in zebrafish versus mice (and chick) strongly suggest an underlying divergence in the mechanism of EGF-CFC function. Therefore, despite the evolutionary conservation of EGF-CFC function in axial midline formation, EGF-CFC proteins may utilize disparate non-conserved molecular mechanisms for this essential developmental process.

We thank Tony Wynshaw-Boris and Heiner Westphal for the EIIa-Cremice and advice on the partial deletion strategy, Jose Belo, Alex Joyner, Gail Martin, Andy McMahon, Guillermo Oliver, Janet Rossant and the late Rosa Beddington for generously providing probes, and Cory Abate-Shen, Rebecca Burdine, Ken Irvine and Ruth Steward for helpful discussions and comments on the manuscript. This work was supported by grants from the American Heart Association (J.D.), the Leukemia and Lymphoma Society (J.D.), the Medical Research Council of Great Britain (M.P.), the Wellcome Trust (M.P.) and the NIH (M.M.S.).