The Foxh1-dependent autoregulatory enhancer controls the level of Nodal signals in the mouse embryo

The Nodal signaling pathway plays fundamental roles during early development of all vertebrates (reviewed by Schier and Shen, 2000; Whitman, 2001). In the mouse, Nodal is required for initial specification of the anteroposterior (AP) axis (Brennan et al., 2001), mesoderm formation (Zhou et al., 1993; Conlon et al., 1994) and left-right (LR) patterning (Collignon et al., 1996; Lowe et al., 1996). Similarly, in zebrafish embryos, the Nodal homologs cyclops and squint control the development of head and trunk mesoderm and endodermal tissues (Feldman et al., 2000), and in Xenopus Nodal-related molecules act as mesoderm inducers (Jones et al., 1995). Recent studies in both Xenopus and zebrafish demonstrate the strength of Nodal signals determines cell fate. Thus, low levels of expression are sufficient to induce mesoderm, while high levels of expression are required for the development of endodermal cell lineages (Agius et al., 2000; Gritsman et al., 2000). These important issues have proven difficult to address in mouse embryos because of reciprocal Nodal-dependent interactions between the epiblast and the extra-embryonic tissues (Brennan et al., 2001).

Nodal signals via the Alk4 or Alk7 type I receptor in association with either the ActRIIA or ActRIIB type II receptor (reviewed by Whitman, 2001). The activated receptor complex phosphorylates the intracellular molecules Smad2 and Smad3, which in turn associate with Smad4 (Massague et al., 2000) and translocate to the nucleus. Cooperatively with other DNA-binding proteins, Smad proteins regulate transcription of target genes (reviewed by Wotton and Massague, 2001). Studies of activin responsive promoters in Xenopus have identified two Smad2/4-interacting DNA partners. The forkhead domain protein Foxh1 (formerly known as FAST1/2) binds as part of a complex to the activin responsive element of the Mix2 gene (Chen et al., 1996; Chen et al., 1997). Similarly members of the Mix family of homeodomain proteins, in association with Smad2/4, activate the Gsc promoter (Germain et al., 2000). Loss of Foxh1 function in mouse results in a range of developmental defects broadly consistent with a role in mediating Nodal signals (Hoodless et al., 2001; Yamamoto et al., 2001). Foxh1 is not required for mesoderm formation, but mutant embryos show AP axis patterning defects, as well as abnormal development of the node and definitive endoderm, both of which are derivatives of the anterior primitive streak. Similarly, zebrafish schmalspur (Foxh1 homolog) loss-of-function mutant embryos lack axial mesendoderm (Boggetti et al., 2000; Pogoda et al., 2000; Sirotkin et al., 2000). These defects are considerably less severe than those resulting from loss of Nodal function (Zhou et al., 1993; Conlon et al., 1994), suggesting that Nodal activates target genes required for initial axis specification and mesoderm formation via Foxh1-independent pathways.

Nodal expression is tightly regulated in a highly dynamic fashion at discrete tissue sites during early mouse development (Collignon et al., 1996; Varlet et al., 1997). Nodal mRNA is first detected throughout the epiblast and overlying visceral endoderm (VE). As development proceeds, Nodal is confined to the prospective posterior epiblast, marking the site of primitive streak formation. Nodal is lost from the VE after the onset of gastrulation, and becomes progressively down-regulated in the primitive streak. In the node, Nodal transcripts appear in a discrete population of cells located on the edges of the ventrally located notochordal plate, and become more strongly expressed on the left. At later stages Nodal is induced in a broad stripe of left lateral plate mesoderm (LPM). Mutations that disturb asymmetric Nodal expression inevitably cause defective LR axis patterning (Collignon et al., 1996; Lowe et al., 1996). Similarly, Nodal mis-expression in chick, fish or Xenopus embryos reverses the orientation of the LR body axis (reviewed by Capdevila et al., 2000).

Transgenic approaches have been used to map cis-regulatory elements responsible for spatially and temporally restricted Nodal expression patterns (Adachi et al., 1999; Norris and Robertson, 1999; Brennan et al., 2001) (D. P. N. and E. J. R., unpublished). An intronic enhancer, termed the ASE, controls early expression in the epiblast and VE, and also at later stages on the left side of the embryo. The minimal ASE enhancer region contains two Foxh1-binding sites (Saijoh et al., 2000). Similar Foxh1-dependent regulatory elements control asymmetric Lefty2 (Leftb – Mouse Genome Informatics) and Pitx2 expression (Saijoh et al., 1999; Shiratori et al., 2001). Moreover, Foxh1-dependent cis-acting regulatory elements have been mapped within the Nodal locus of all vertebrates examined (reviewed by Whitman, 2001). We have used gene targeting techniques to delete the ASE and test its role in axial patterning. The developmental abnormalities seen in trans-heterozygous embryos carrying the ASE deletion and a null allele (Nodal^Δ600/–), and homozygous mutant embryos (Nodal^Δ600/Δ600) were compared. In Nodal^Δ600/– embryos, Nodal is expressed at low levels in the early epiblast and becomes confined to proximal epiblast cells; however, expression is undetectable in the VE. Reduced Nodal expression levels are sufficient for formation of the anterior visceral endoderm (AVE) and mesoderm induction. Many of these embryos, however, incorrectly position the AP axis, leading to abnormal gastrulation and cell movements. Moreover, Nodal^Δ600/– embryos lack anterior definitive endoderm (ADE) and display rostral CNS patterning defects. By contrast, in homozygous Nodal^Δ600/Δ600 embryos, Nodal is more efficiently expressed in the epiblast but absent from the VE. Unexpectedly, these embryos develop normal AP pattern. Later, at early somite stages, Nodal expression in the LPM is significantly reduced, leading to failure to activate Lefty2, delayed activation of Pitx2 and defective LR axis patterning. Overall, we conclude the Foxh1-dependent autoregulatory enhancer maintains and amplifies Nodal signals that selectively activate target genes responsible for patterning the embryonic body plan.

Nodal^LacZ/+ ES cells were transfected with the Δ600 targeting construct (Norris and Robertson, 1999) and selected in Hygromycin B. DNA prepared from individual drug-resistant colonies was genotyped by Southern blot analysis, as previously described (Norris and Robertson, 1999). After identification of targeted clones, the selection cassette was removed by transient expression of Cre recombinase and colonies isolated by low density plating. Approximately one in 15 of the drug-resistant clones were targeted with equivalent recombination frequencies on both chromosomes. Targeting events and removal of the hygromycin selection cassette were confirmed using 5′ and 3′ probes. Selected ES cell lines were used to generate germline chimeric mice that were subsequently bred to ICR females (Taconic).

Nodal^LacZ, Nodal^Δ100.lacZ, Nodal^Δ600.LacZ and transgenic lines expressing lacZ were PCR genotyped for the presence of the lacZ gene. Δ600 mice were genotyped by PCR using Δ600-5 (5′ GCT AGT GGC GCG ATC GGA ATG GA 3′) and Δ600-6 (5′ AAG GGA AGT GAA CTG GAA AGG TAT GT 3′): a 350 bp fragment shows presence of the deletion; a 950 bp fragment shows the wild-type allele. Foxh1 mutant mice were a kind gift from Jeff Wrana and Pamela Hoodless, and were genotyped according to published protocols (Hoodless et al., 2001). All strains used in this analysis were maintained by breeding to ICR mice (Taconic).

Whole-mount in situ hybridization was performed according to standard procedures. Probes for the following genes were used in this study: α cardiac actin (Roebroek et al., 1998), Bmp4 (Winnier et al., 1995), cripto (Ding et al., 1998), Eomes (Russ et al., 2000), Foxa2 (previously known as Hnf3β) (Sasaki and Hogan, 1996), Gsc (Blum et al., 1992), Hex (Thomas et al., 1998), Lefty (Ebaf – Mouse Genome Informatics) (Meno et al., 1996), Lhx1 (Barnes et al., 1994), Nodal (Conlon et al., 1994), Otx2 (Ang et al., 1994), Pitx2 (Ryan et al., 1998), Shh (Echelard et al., 1993), T (Herrmann, 1991) and Wnt3 (Liu et al., 1999). For histology embryos were fixed in 4% paraformaldehyde (PFA), dehydrated through an ethanol series and embedded in wax before sectioning. Hematoxylin and Eosin staining and X-gal staining were performed according to standard protocols. India ink was injected into the left ventricle of embryos at 14.5 dpc and 18.5 dpc, and allowed to fill the ventricle. Whether or not the ink moved directly into the right ventricle was scored visually.

The ASE, which is responsible for early Nodal expression in the VE and epiblast and at later stages asymmetric expression in LPM (Adachi et al., 1999; Norris and Robertson, 1999), is thought to be regulated by the forkhead transcription factor Foxh1 (Saijoh et al., 2000). Deletion of the 100 bp core region containing the two Foxh1-binding sites causes striking changes to the expression of a Nodal^LacZ reporter allele (Norris and Robertson, 1999). Thus lacZ activity was restricted to the proximal epiblast, absent from the VE and severely attenuated in the LPM of Nodal^Δ100.lacZ/+ embryos. To demonstrate directly that Foxh1 modulates Nodal activity via associations with the ASE, we examined ASE-lacZ transgene expression in Foxh1 mutant embryos (Hoodless et al., 2001). In wild-type embryos at early post-implantation stages, the transgene is active throughout the epiblast and VE, but expression becomes localized to the prospective posterior prior to the initiation of gastrulation (Fig. 1B) (Norris and Robertson, 1999). In striking contrast, expression is undetectable in Foxh1^–/– embryos (Fig. 1C). To test whether Foxh1 is sufficient for activation of the ASE, we removed the entire 600 bp ASE (Fig. 1A). The resulting Nodal^Δ600.LacZ allele is expressed proximally but not distally in the epiblast (Fig. 1E,G) due to the activity of the 5′ posterior epiblast enhancer (PEE) (Norris and Robertson, 1999; Brennan et al., 2001) (D. P. N. and E. J. R., unpublished). Similar Nodal^Δ600.LacZ expression patterns were observed in a Foxh1^–/– background (Fig. 1H), confirming the Foxh1-independent activity of the PEE. A proportion of Foxh1^–/–Nodal^+/– mutant embryos were previously shown to display more severe patterning defects than Foxh1^–/– embryos (Yamamoto et al., 2001). Similarly, here some (∼25%) of Foxh1^–/–Nodal^Δ600.LacZ and Foxh1^–/–Nodal^LacZ/+ mutant embryos express lacZ throughout the epiblast (Fig. 1F, 1I). Interestingly, this expanded Nodal expression domain and the associated morphological disturbances closely resemble those observed for Smad2 mutant embryos (Waldrip et al., 1998).

To examine ASE function in vivo, we deleted the 600 bp enhancer in the context of the wild-type locus (Fig. 1A). At day 6.5, the majority of Nodal^Δ600/– embryos (60%) are indistinguishable from wild type (Table 1). However, the remaining 40% develop abnormally and display a characteristic morphology (Fig. 2A). In these overtly abnormal Nodal^Δ600/– embryos, gastrulation is initiated and mesoderm induced radially around the proximal epiblast. These embryos also have a thickened patch of visceral endoderm characteristic of the AVE (Kimura et al., 2000) located at or just lateral to the distal tip of the epiblast, suggesting that the AVE forms normally but fails to rotate, as is the case for Otx2 and cripto mutants (Ding et al., 1998; Perea-Gomez et al., 2001). These embryos also display a marked constriction at the embryonic/extra-embryonic boundary (Fig. 2B,C), and as gastrulation proceeds, the pro-amniotic canal often becomes constricted, resulting in the formation of two separate cavities (Fig. 2B,C).

To assess tissue disturbances further, we analyzed a panel of molecular markers. In the overtly abnormal Nodal^Δ600/– embryos, posterior epiblast markers such as cripto (Ding et al., 1998), T (Thomas et al., 1998) and Wnt3 (Liu et al., 1999) are ectopically expressed throughout proximal regions of the epiblast (Fig. 2B-D). Occasionally, T-positive cells corresponding to proximal nascent mesoderm were found in the extra-embryonic region (Fig. 2C; also observed for Nodal^LacZ expression in Fig. 3C). This cell population lacks expression of the epiblast marker Pou5f1 (formerly Oct4), and thus represents mesodermal cells migrating into the yolk sac (data not shown). Collectively these results demonstrate that mesoderm induction occurs throughout the proximal epiblast in Nodal^Δ600/– embryos. Foxa2 (formerly Hnf3β) is initially expressed throughout the VE, but by 6.5dpc localizes to the AVE and the anterior primitive streak (Perea-Gomez et al., 1999) (Fig. 2E). In mutant Nodal^Δ600/– embryos, Foxa2 transcripts are found at the proximal and distal most extent of the embryonic region, suggesting the proximal distal axis is formed correctly but fails to rotate. Additional AVE markers, including Hex (Thomas et al., 1998) and Lhx1 (Perea-Gomez et al., 1999) gave similar results (Fig. 2F,G). Expression of the extra-embryonic markers Eomes and Bmp4 is unperturbed in Nodal^Δ600/– embryos (Fig. 2J,K). We also examined expression of Gsc and Lefty1, putative targets of Nodal/Foxh1 signaling (Labbe et al., 1998; Yamamoto et al., 2001). Gsc normally marks the AVE and the anterior primitive streak; however, in mutant embryos, expression is markedly reduced and in some cases the AVE expression domain is entirely absent (Fig. 2H). Using a cross-reactive probe (Meno et al., 1997) that detects both Lefty1 and Lefty2 (Fig. 2I), we found Lefty2 expressed in the nascent mesoderm, but Lefty1 fails to be induced in the VE of Nodal^Δ600/– embryos. Thus, we observe a distinctive perturbation of AP axis position in a high proportion of Nodal^Δ600/– embryos.

Next, we analyzed Nodal mRNA expression patterns. As shown in Fig. 3A, lacZ expression is undetectable in the distal epiblast of Nodal^Δ600.LacZ/+ embryos. Similarly in Nodal^Δ600/– embryos reduced levels of Nodal mRNA are confined to the proximal epiblast (Fig. 3B). This reduced level, however, is sufficient to activate expression in the VE, as evidenced by the pattern of X-gal staining in the VE of Nodal^Δ600/LacZ embryos (Fig. 3C). By contrast, the VE of Nodal^{Δ600.LacZ/Δ600.LacZ} embryos fails to express lacZ (Fig. 3D). Thus, activation of Nodal in the VE requires Foxh1/Smad2 associations with the autoregulatory ASE.

While only 40% of Nodal^Δ600/– embryos exhibit overt morphological defects at early gastrulation stages, by 7.5 dpc, all Nodal^Δ600/– embryos are distinctly abnormal (Table 1). As mentioned above, approximately 40% exhibit a visible constriction at the embryonic/extra-embryonic boundary at 6.5 dpc, which varies in severity ranging from a slightly noticeable narrowing to almost complete separation of these regions (Fig. 4A,B; summarized in Table 1). Slightly later at 8.5 dpc, this class of most severely disturbed embryos develop external to the visceral yolk sac (VYS) (Fig. 4E). In some cases, the anterior regions are partially externalized, while the primitive streak is retained within the VYS (Fig. 4D). Histological analysis demonstrates the presence of neural tissue (Fig. 4D,E), but midline tissues are largely absent resulting in fused somites (data not shown). This is likely to be due to defective gastrulation movements that secondarily lead to abnormal node morphogenesis and failure to form a notochord. Cells expressing the myocardial marker α cardiac actin are juxtaposed to the VYS in severely affected embryos (data not shown), suggesting that cardiac precursors are specified normally.

The majority of mutant embryos develop within the VYS, exhibit a normal midline, somites and yolk-sac morphology (Fig. 4C), but have defective cardiac looping (data not shown). By 8.5 dpc, Nodal^Δ600/– embryos begin to display a marked reduction in anterior neural tissues. As the AVE is correctly specified, and the node and notochord form normally, these developmental abnormalities probably result from defective formation of the anterior definitive endoderm (ADE), a tissue arising from the anterior streak that is known to be responsible for patterning the CNS (Martinez Barbera et al., 2000). In keeping with this conclusion, Hex expression, a marker of definitive endoderm, is not detected in Nodal^Δ600/– embryos from gastrulation stages onwards (data not shown). Expression of Shh, a marker of ADE, is punctate and highly downregulated even in grossly normal Nodal^Δ600/– embryos (Fig. 4F). The anterior Shh domain normally extends into the ventral forebrain by 8.5 dpc (Fig. 4G,H), but in Nodal^Δ600/– mutant embryos, expression is truncated at the hindbrain (arrowhead in Fig. 4G). The most severely affected Nodal^Δ600/– embryos also display decreased midline tissue and deficiencies in definitive endoderm formation (Fig. 4I). Similar data were obtained analyzing Foxa2 (Fig. 4J,K). Nodal^Δ600/– embryos also display reduced expression of Otx2 (Fig. 4L,M), a marker of forebrain tissue (Ang et al., 1994). Because a normal notochord is present in the majority of Nodal^Δ600/– embryos, specification of the node, a derivative of the anterior streak, appears to be relatively unaffected. Thus, Nodal expression controlled by the 5′ node element (Norris and Robertson, 1999), which remains intact in the Nodal^Δ600 allele, is sufficient to promote node morphogenesis.

As shown in Fig. 5, in contrast to Nodal^Δ600/– embryos, we found that Nodal^Δ600/Δ600 embryos efficiently undergo AVE rotation, display a correctly positioned primitive streak, and gastrulate normally. In comparison with Nodal^Δ600/– embryos, Nodal expression within the epiblast is elevated roughly twofold in Nodal^Δ600/Δ600 homozygous mutants. This increased dose is sufficient to rescue ADE formation and axis patterning defects. Nodal mRNA is expressed at significantly reduced levels in the epiblast, and is undetectable in the VE (Fig. 5A-D). Thus, Nodal expression within the VE is not essential for either specification or positioning of the AVE. Rather, these processes depend on the strength of Nodal signals within the epiblast. At 7.5 and 8.5 dpc, expression within the node closely resembles that in wild type (Fig. 5E,F), except that asymmetry of expression is lost. Later, at early somite stages, Nodal^Δ600/Δ600 embryos display only weak Nodal expression in LPM (Fig. 5J-L), although invariably on the left side of the axis.

By the eight-somite stage, Nodal^Δ600/Δ600 embryos exhibit abnormal heart looping morphogenesis (Fig. 6). To evaluate the onset of LR patterning defects, we assessed expression of downstream targets, including the homeodomain protein Pitx2 and the Nodal antagonists Lefty2 and Lefty1 (Saijoh et al., 2000; Shiratori et al., 2001). Lefty 2 expression is normally downregulated concomitantly with that of Nodal (Meno et al., 1996). Only a few Nodal^Δ600/Δ600 embryos (30%) express Lefty2 transcripts (Fig. 6B). As shown in Fig. 6A, Lefty1 is asymmetrically expressed in the prospective floorplate of the neural tube, whereas in Nodal^Δ600/Δ600 embryos Lefty1 is missing from the midline and only weakly expressed in a few cells close to the node (Fig. 6B).

Pitx2 mutant embryos exhibit right sided lung isomerisms, abnormal cardiac morphogenesis, reduced spleen development and incomplete turning (Gage et al., 1999; Kitamura et al., 1999; Lin et al., 1999; Lu et al., 1999). Low levels of Pitx2 are required for specification of atrial situs and increased expression is necessary for establishment of lung situs (Gage et al., 1999; Kitamura et al., 1999; Liu et al., 2001). Pitx2 transcription is initiated in left LPM coincident with the onset of asymmetric Nodal expression. We found the onset of asymmetric Pitx2 expression is slightly delayed in Nodal^Δ600/Δ600 embryos and becomes visible only at the five- to six-somite stage (Fig. 6D). Moreover, the anterior extent of the Pitx2 expression domain is markedly reduced by the eight- to 10-somite stage (Fig. 6E,F). At 9.5 dpc, Pitx2 expression normally extends behind the sinoatrial region of the heart, but in Nodal^Δ600/Δ600 embryos the anterior extent of expression is reduced by approximately one to two somite widths (Fig. 6G,H).

To analyze embryonic turning and establishment of organ situs, Nodal^Δ600/Δ600 embryos were allowed to develop beyond 8.5 dpc. At early somite stages, none of a large panel of mutant embryos showed morphological defects (Table 1). However, by 9.0 dpc, heart looping is visibly abnormal in Nodal^Δ600/Δ600 embryos, with the heart positioned more ventrally than in wild type (Fig. 7A,B). The heart tubes appear to fold normally, suggesting they have acquired regional patterning. Moreover, heart contractions and blood circulation develop normally (data not shown). By 11.5 dpc, mutant hearts display an ambiguously positioned apex (Fig. 7C). In addition, the lungs of mutant embryos display lobation defects (Fig. 7C; Table 1). Differences in branching morphogenesis normally leads to the formation of four lobes on the right and a single lobe on the left side of the thoracic cavity. By contrast, Nodal^Δ600/Δ600 embryos develop partial right isomerisms (Fig. 7C).

LR patterning not only affects heart and lung morphogenesis, but is also essential for establishment of the correct connection of the heart to the vasculature (Kathiriya and Srivastava, 2000). Defects in the aortic arches (Fig. 7D-G) were assessed visually and by India ink injections. The aortic arch normally exits the left ventricle, behind the pulmonary artery, and arches to the right; three arteries branch upwards delivering blood to the upper body (Fig. 7D). The pulmonary artery exits the right ventricle arches to the right and splits into two arteries that lead to the lungs. Prenatally, however, the pulmonary artery empties through the ductus ateriosus into the dorsal aorta. In Nodal^Δ600/Δ600 embryos an enlarged aorta rises in front of the pulmonary artery (Fig. 7F,G), giving rise to four ascending arteries. The pulmonary artery empties into the descending aorta through the ductus ateriosus (Fig. 7G). Moreover, ink injected into the right ventricle immediately filled the left ventricle implicating ventriculoseptal defects (data not shown). Approximately 30% of Nodal^Δ600/Δ600 embryos survive to birth (Table 1). However, these are often runted, and none survives beyond 6 weeks of age.

Dynamic patterns of Nodal expression are tightly regulated during early mouse development (Collignon et al., 1996; Varlet et al., 1997). An intronic enhancer termed the ASE, which controls expression in the epiblast, VE and left lateral plate mesoderm contains two Foxh1-binding sites (Saijoh et al., 2000). Recent in vitro studies suggest that the range and intensity of Nodal signals are in part controlled by an autoregulatory feedback loop (reviewed by Whitman, 2001). The present experiments demonstrate ASE activity in mouse embryos is strictly Foxh1 dependent. Acting in an autocrine manner, the ASE maintains and amplifies Nodal transcription, and thereby increases the amount of ligand produced locally. The ASE also functions in a paracrine fashion to expand expression domains. Specifically, the ASE controls Nodal-dependent target gene expression in the VE. The Foxh1-dependent autoregulatory enhancer thus establishes a gradient of Nodal signals that is crucial for mesoderm induction, endoderm specification and AP and LR axis patterning during early mouse development (summarized in Fig. 8).

Shortly after implantation, at the egg cylinder stage, the mouse embryo exhibits a distinct molecular pattern along its proximodistal (PD) axis. The final AP axis is positioned by coordinated cell movements that rotate this pre-existing PD axis, and thus restrict mesoderm induction to the proximal posterior epiblast (reviewed by Beddington and Robertson, 1999; Lu et al., 2001). Nodal signals from the epiblast establish this PD pattern (Brennan et al., 2001). The ASE has an important role controlling Nodal activation in the VE and in amplifying Nodal signaling in the epiblast. At day 5.5, Nodal expression is initiated throughout the epiblast of Nodal^Δ600/– embryos, but in contrast to wild type, expression rapidly becomes confined only to proximal epiblast cells. Thus, ASE activity maintains widespread expression in a broader domain of the epiblast, and activates Nodal transcription in the VE. These results demonstrate that Nodal mRNA expression, as assessed by whole-mount in situ hybridization, is greatly reduced in the absence of the feedback loop. Nonetheless, the low levels of Nodal expression in the epiblast of Nodal^Δ600/– embryos are sufficient for establishment of correct PD pattern, and promote expression of genes such as Wnt3 and Eomes, which are required for mesoderm formation. Nodal^Δ600/– embryos also maintain expression of Bmp4 in the adjacent extra-embryonic ectoderm, and markers of the AVE such as Hex are induced in the overlying VE. In a high percentage of Nodal^Δ600/– embryos, the PD axis is imprecisely converted to AP pattern, a phenotype also seen in embryos homozygous for a hypomorphic Nodal allele (Lowe et al., 2001). Because the AP axis is rescued in Nodal^Δ600/Δ600 embryos, we conclude that this defect reflects decreased levels of Nodal signaling in the epiblast. Thus, Nodal transcription is initiated via upstream regulatory elements, probably the PEE. In Nodal^Δ600/– embryos, Nodal activity is sufficient to induce mesoderm formation, but insufficient to promote cell movements that position the AP axis. Loss of the Nodal co-factor cripto causes similar patterning defects (Ding et al., 1998), suggesting that the strength of Nodal ligand/receptor interactions in the epiblast controls cell growth and/or morphogenetic movements.

Embryos that lack the ASE fail to express Nodal mRNA in the VE. Thus, activation of the Nodal locus in the VE is strictly controlled by the Smad2/Foxh1-dependent pathway. However, markers of the AVE are activated normally in Nodal^Δ600/– embryos. Thus, Nodal signals within the VE are nonessential for early anterior patterning. Consistent with this, embryos deficient in Foxh1, and therefore lacking Nodal activity in the VE, also establish a normal AVE (Hoodless et al., 2001; Yamamoto et al., 2001). An interesting feature of Nodal^Δ600/– embryos is that they develop a severe physical constriction at the extra-embryonic/embryonic boundary that subsequently leads to externalization and growth of the embryo outside of the yolk sac. Similar defects have been described for Foxa2 and Lhx1 mutants (Ang and Rossant, 1994; Shawlot and Behringer, 1995). In the case of Foxa2, this probably reflects a requirement for Foxa2 within the VE, as the defect is not seen in chimeric embryos in which the VE is wild type (Dufort et al., 1998). Loss of Foxh1 also results in VE constriction, and this defect is also alleviated in mutant embryos in which the VE is wild type (Yamamoto et al., 2001). By contrast, Nodal^Δ600/Δ600 embryos that lack Nodal activity in the VE develop normally. The morphological defects in Nodal^Δ600/– embryos are therefore most likely due to decreased Nodal signaling from the epiblast that in turn attenuates expression of downstream target genes acting in the VE. In keeping with this idea, we find that expression of genes such as Gsc and Lefty1, which represent downstream targets of the Nodal pathway (Labbe et al., 1998; Yamamoto et al., 2001), are induced poorly, if at all, in Nodal^Δ600/– embryos. We conclude that the strength of Nodal signaling from the epiblast is crucial in controlling activation of discrete Nodal targets in the VE that in turn are responsible for normal growth and morphogenesis of the VE.

Interestingly Foxh1^–/– (Hoodless et al., 2001; Yamamoto et al., 2001) and Nodal^Δ600/– embryos both develop anterior CNS truncations, a phenotype similar to that which develops in chimeric embryos lacking Nodal function in the VE (Varlet et al., 1997). Here, by contrast, we observe normal development of Nodal^Δ600/Δ600 embryos, arguing that Nodal is not required in the VE. Similarly Foxh1, while essential for Nodal activation in the VE, is not required for formation of the AVE (Hoodless et al., 2001; Yamamoto et al., 2001). The current findings suggest an alternative explanation for the development of these anterior defects, namely that lowering the dose of Nodal signaling in the epiblast, either genetically or via cell mixing in chimeras, leads to a failure to correctly specify definitive endoderm tissue. The definitive endoderm (DE) initially appears as a discrete cell layer at the most distal end of the primitive streak (Lawson et al., 1986; Lawson and Pedersen, 1987; Kinder et al., 2001), displacing primitive visceral endoderm into the extra-embryonic regions (Lawson and Pedersen, 1987; Tam and Beddington, 1992; Thomas and Beddington, 1996). Recent evidence has shown that the signals provided by the anterior gut endoderm are essential for reinforcing and correctly elaborating pattern in the developing neural plate after displacement of the AVE. Thus, embryos lacking the transcription factor Hex fail to form anterior definitive endoderm and develop anterior CNS truncations (Martinez Barbera et al., 2000). Similarly, the Wnt inhibitor Dkk-1 acting in the anterior mesendoderm (Mukhopadhyay et al., 2001), and the pro-protein convertase SPC4 expressed in the foregut (Constam and Robertson, 2000) are both required for patterning of the rostral CNS. Our current findings indicate in mouse, as for Xenopus (Agius et al., 2000), that Nodal signaling is essential for specification of the definitive endoderm lineage.

Neither Foxh1 (Hoodless et al., 2001) nor Smad2 (Tremblay et al., 2000) -deficient ES cells are competent to form definitive endoderm, suggesting Foxh1/Smad2 complexes govern endodermal fate in the anterior streak. Chimeras composed of Nodal-deficient VE, with Nodal expression only partially restored in the epiblast, phenocopy Nodal^Δ600/– embryos (Varlet et al., 1997). Thus, decreasing Smad2 signals in the anterior streak is sufficient to disrupt formation of DE, in turn leading to cardiac and anterior CNS defects. By contrast, in Nodal^Δ600/Δ600 embryos, increased levels of Nodal activity allow normal specification of the definitive endoderm lineage. Collectively, these findings establish a dose-dependent requirement for the Nodal signaling pathway in eliciting endodermal fates in the mouse.

Conserved Nodal activities are known to be essential for specification of the LR body axis (reviewed by Capdevila et al., 2000). Asymmetric activation of Nodal expression in the left lateral plate mesoderm at early somite stages in turn activates the downstream targets Lefty2 and Pitx2 (reviewed by Capdevila et al., 2000) and enhances Nodal transcription via the Foxh1-dependent autoregulatory enhancer. Deletion of the Nodal ASE enhancer severely compromises this pathway, and only low levels of asymmetric Nodal expression are transiently detected in Nodal^Δ600/Δ600 embryos.

Asymmetric Lefty2 transcription is also mediated by an enhancer containing two Foxh1-binding motifs (Saijoh et al., 1999; Saijoh et al., 2000). Deletion of these sequences in the mouse germline results in failure to activate asymmetric Lefty2 expression and causes ectopic Nodal expression leading to situs defects (Meno et al., 2001). Nodal^Δ600/Δ600 embryos either fail or only weakly activate Lefty2 expression. Lefty2 target gene expression thus requires continuous Nodal signaling. In addition, we found that Lefty1, which is normally expressed in the midline prospective floorplate tissue, is only rarely induced in Nodal^Δ600/Δ600embryos. These results implicate Lefty1 as a downstream target. However, as yet, Nodal responsive cis-acting elements such as Foxh1-binding sites have not been identified at the Lefty1 locus (Saijoh et al., 1999).

By contrast, decreased Nodal activity in Nodal^Δ600/Δ600 embryos is sufficient to activate robust Pitx2 expression. Recent experiments have described a Pitx2 transcriptional enhancer containing three Foxh1-binding motifs. In contrast to Nodal, Pitx2 transcription is maintained via the activity of an Nkx2.1-dependent enhancer (Shiratori et al., 2001). The increased number or distinct configuration of the Foxh1-binding sites may allow Pitx2 occupancy by activated Smad2/3/4 complexes transiently formed in Nodal^Δ600/Δ600 embryos. Nonetheless delayed activation of Pitx2 transcription in the LPM results from the markedly reduced Nodal expression, that correlates with the development of right pulmonary isomerisms and heart defects in Nodal^Δ600/Δ600 embryos. As Pitx2-deficient embryos show similar thoracic defects (Gage et al., 1999; Kitamura et al., 1999; Lin et al., 1999; Lu et al., 1999), these abnormalities probably reflect altered Pitx2 expression patterns.

Cryptic, a Nodal co-factor (Yan et al., 1999), and Gdf1, a TGFβ family member (Rankin et al., 2000), are both essential for asymmetric activation of Nodal, Lefty2 and Pitx2. Loss-of-function mutants display right pulmonary isomerism and randomization of cardiac and abdominal situs, associated with randomization in the direction of embryonic turning and heart looping. Surprisingly, we found that dramatically decreasing Nodal activity in the lateral plate by elimination of the positive feedback loop had a relatively minor impact on LR patterning. Thus, the direction of embryonic turning or initial cardiac looping is unaffected in Nodal^Δ600/Δ600 embryos. This cannot be due to asymmetric activation of Pitx2 in these embryos, as loss-of-function of Pitx2 does not affect the direction of turning or prevent initial establishment of abdominal situs (Gage et al., 1999; Kitamura et al., 1999; Lin et al., 1999; Lu et al., 1999). Embryonic turning and abdominal situs may be therefore be controlled by distinct, Nodal-independent, molecular pathways. However, in chick (Logan et al., 1998; Piedra et al., 1998; Ryan et al., 1998) and Xenopus (Sampath et al., 1997; Ryan et al., 1998; Campione et al., 1999), ectopic Nodal expression can reverse body situs, suggesting that as yet undescribed Nodal targets may be activated to control this process. In this case, based on the phenotype of Nodal^Δ600/Δ600 embryos, only low levels of asymmetric Nodal activity are predicted to be required to activate these pathways.

It has been suggested that LR asymmetry is initiated in the mouse by a net leftward flow of extracellular fluid generated by the cilia located on the ventral surface of the node (reviewed by Capdevila et al., 2000). This asymmetric morphogen gradient is thought to act at a distance to induce asymmetric Nodal expression in the left lateral plate. Mutations affecting the formation or motility of the cilia disturb establishment of the asymmetric Nodal expression domain. Nodal ligand produced at the edge of the notochordal plate potentially represents this ‘morphogen’ and acts to induce its own transcription in the LPM. Consistent with this, recent work in zebrafish shows that the Nodal homolog squint can act at a distance (Chen and Schier, 2001). Similarly removal of the asymmetric expression domain of the Nodal antagonist Lefty2, results in bilateral Nodal expression, consistent with long range diffusion and autoactivation by secreted Nodal ligand (Meno et al., 2001).

However, the present findings challenge this simple model and allow us to draw two new conclusions about Nodal activities involved in establishing the LR axis. First, we demonstrate that asymmetric Nodal expression in the developing node is not essential to induce asymmetric Nodal transcription in the lateral plate, as Nodal^Δ600/Δ600 embryos express wild-type levels of Nodal transcripts in the node in a symmetric fashion, but Nodal is correctly induced exclusively on the left side of the axis. Thus, the functional significance of Nodal asymmetry in the mouse node remains unclear. Second, deletion of the autoregulatory enhancer fails to eliminate Nodal expression in the left LPM. Thus, tissue-specific activation of the locus occurs selectively on the left via an ASE- and probably Foxh1-independent mechanism(s). While key molecules controlling LR patterning have been identified in recent years, clearly additional work will be required to dissect the components of the functional pathways they control, and how they interact with each other to regulate this process.

We thank Jeff Wrana and Pamela Hoodless for providing the Foxh1 mutant mice, Patti Lewko and Joe Rocca for animal care, Debbie Pelusi for genotyping assistance, Gerry Chu for advice about histology, and Cindy Lu and Ray Dunn for comments on the manuscript. D. P. N. was supported by a Charles A King Trust Fellowship and J. B. by a Fellowship from the Wellcome Trust. This work was supported by a grant from the NIH to E. J. R.