Nodal signaling induces the midline barrier by activating Nodalexpression in the lateral plate

The establishment of three axes (anteroposterior, dorsoventral and left-right) is a fundamental aspect of the development of a body plan. Substantial insight has recently been achieved at the genetic and molecular levels into the generation of left-right (LR) asymmetry in vertebrate embryos(Beddington and Robertson,1999; Capdevila et al.,2000; Wright,2001; Hamada et al.,2002). The establishment of LR asymmetry is thought to be achieved in four distinct steps: (1) the breaking of LR symmetry in or near the node,(2) transfer of LR-biased signals from the node to the lateral plate, (3)asymmetric expression of signaling molecules such as Nodal and Lefty on the left side of the lateral plate and (4) the induction by these signaling molecules of LR asymmetric morphogenesis of visceral organs. In addition, the midline barrier must be established to separate the two sides of a developing embryo.

Nodal and Lefty (Ebaf — Mouse Genome Informatics), both of which are members of the transforming growth factor β (TGFβ) family of proteins, play important roles in several embryonic patterning events(Schier and Shen, 2000;Brennan et al., 2001;Juan and Hamada, 2001). Lefty antagonizes Nodal signaling by acting as a feedback inhibitor(Meno et al., 1999;Cheng et al., 2000;Sakuma et al., 2002). In LR patterning, genetic evidence suggests that Nodal expressed on the left side of the lateral plate acts as a left-side determinant and induces left side-specific morphogenesis of visceral organs(Oh and Li, 1997;Yan et al., 1999;Lowe et al., 2001), whereas Lefty2 (Leftb — Mouse Genome Informatics), which is induced by Nodal in the left lateral plate, restricts the timing and the region of Nodal activity(Meno et al., 2001). Nodal and Lefty have similar roles among vertebrates from the zebrafish to mouse(Sampath et al., 1997;Rebagliati et al., 1998;Bisgrove et al., 1999;Thisse and Thisse, 1999).

Despite the recent progress in our understanding of LR patterning, many important questions remain unanswered. One such question concerns the mechanism by which symmetry is broken in the first place. Breaking of symmetry in mammals appears to involve nodal flow, the leftward flow of extra-embryonic fluid in the node generated by the vortical movement of nodal cilia(Nonaka et al., 1998). Thus,nodal flow is impaired in mutant mice in which LR patterning is randomized(Okada et al., 1999). Indeed,many of the genes whose mutation results in LR patterning defects encode proteins required for the formation or motility of cilia. Furthermore,imposition of an artificial flow was able to direct LR patterning in early mouse embryos (Nonaka et al.,2002). The mechanism by which Nodal flow achieves this effect,however, has remained unclear. It is possible that the flow transports a LR determination factor toward the left side, but the identity of such a factor is unknown.

The mechanism of signal transfer from the node to the lateral plate is also unknown. Both the identity of the signal (or signals) transported from the node to the lateral plate mesoderm (LPM) and whether the signal is transferred directly from the node to the lateral plate or is first relayed to an intermediate region such as the paraxial mesoderm remain to be determined.

Another important and related question concerns the mechanism by which the expression of Nodal is initiated in left LPM. Although an autoregulatory mechanism involving signaling by Nodal and the transcription factor Foxh1 (previously known as Fast2) is responsible for amplification ofNodal expression in left LPM(Saijoh et al., 2000;Norris et al., 2002), it is not known how Nodal expression is initiated. Ectopic expression of Nodal in right LPM of chick embryos was not able to induce Nodalexpression (M. Levin, PhD thesis, Harvard University, 1996), suggesting that an unknown factor other than Nodal initiates Nodal expression in left LPM. Bone morphogenetic protein (BMP) signaling has been proposed to regulateNodal expression negatively in the chick, and a BMP antagonist such as Caronte may initiate Nodal expression by inhibiting BMP activity on the left side (Rodriguez Esteban et al., 1999; Yokouchi et al.,1999). However, recent evidence has suggested that BMP signaling positively regulates Nodal expression by inducing an EGF-CFC factor in LPM (Schlange et al., 2001;Schlange et al., 2002;Piedra and Ros, 2002). The factor responsible for the initiation of asymmetric Nodal expression in LPM thus remains elusive. Finally, the midline structures serve as a barrier that prevents the diffusion of asymmetric signals(Danos and Yost, 1996;Lohr et al., 1997;Meno et al., 1998), but it is unclear how the midline barrier is established and precisely how it functions. Analysis of Lefty1 mutant mice has shown that Lefty1, a Nodal antagonist expressed on the left side of the prospective floor plate (PFP),contributes to midline barrier function(Meno et al., 1998). However,it is unknown how Lefty1 expression is induced at the midline.

We have now studied the role of Nodal-Foxh1 signaling in LR patterning by analyzing Foxh1 conditional mutant mice. We have also examined Nodal function by developing transplantation and electroporation systems for use with mouse embryos and applying these systems to the Foxh1 mutant mice. Unexpectedly, Nodal-Foxh1 signaling was shown to be able to initiateNodal expression in LPM and to induce Lefty1 expression at the midline. Our results indicate that the leftsided expression ofNodal in LPM is initiated by Nodal produced in the node, and thatLefty1 expression at the midline is induced by Nodal produced in left LPM. We propose that Nodal activity travels from the node to left LPM, and from left LPM to the midline.

The Lefty2-3.0 Cre transgene was constructed by ligating the 3.0-kb upstream region of mouse Lefty2 to a Cre cassette derived from pBS-Cre (kindly provided by H. Kondoh). Several transgenic lines harboring this transgene were established. To examine the specificity ofCre expression, we crossed each line with Cre-reporter mice harboring a lacZ gene that is expressed only after Cremediated excision(Sakai and Miyazaki, 1997). Embryos were genotyped and stained with X-gal. One transgenic line (77b) that exhibited bilateral Cre expression in the lateral plate was used in this study. Line 77b animals were crossed with Foxh1^+/- mice to obtain heterozygotes harboring the Cre transgene; theseFoxh1^+/-, Lefty2-3.0 Cre mice were then mated with Foxh1^flox/flox animals(Yamamoto et al., 2001), and the resulting embryos were analyzed at E8.2, E9.5 and E10.5.

Mouse embryos were staged on the basis of their morphology(Downs and Davies, 1993). Whole-mount in situ hybridization was performed according to standard procedures (Wilkinson, 1992). Wild-type and mutant embryos were processed in the same tube. Embryos were genotyped by PCR analysis of yolk sac DNA.

Fragments of tissue (containing ∼20 cells) were isolated for transplantation from the left LPM of mouse embryos at the four-somite stage. For use as recipients, mouse embryos were recovered at the two-somite stage,dissected free of decidual tissues and Reichert's membrane, and maintained in culture until manipulation. Cell clumps from donor embryos were grafted to the right LPM, left paraxial mesoderm or left LPM of the host embryos with the use of tungsten needles. The transplanted embryos were cultured for 3 hours at 37°C in 35 mm disposable dishes containing 4 ml of 50% Dulbecco's modified Eagle's medium supplemented with 50% rat serum(Lawson et al., 1986); this volume of medium was sufficient for culture of eight embryos(Sturm and Tam, 1993). A transplant remained as a mass after the culture, showed distinct density and thickness, and was easily distinguished from the host tissues.

The full-length cDNAs of mouse Nodal, constitutive active human ALK4(caALK4), lacZ and EGFP were subcloned into the eucaryotic expression vectors pEF-BOS (Mizushima and Nagata,1990), pcDNA3, pEF and pCX, respectively. Each plasmid was suspended in phosphate-buffered saline at a concentration of 5 mg/ml. Mouse embryos at the two-somite stage were dissected free of decidual tissues and Reichert's membrane and maintained in culture until electroporation. Electroporation was performed with a CUY21 electroporator and a pulse monitor(BTX, Tokyo). Platinum electrodes were used as an anode and a cathode and were positioned near the posterior and anterior regions of the embryo,respectively. Electric pulses were applied (14 V for 129 ms, five times) while the DNA solution (1 μl) was injected into the anterior region of left or right LPM. The electroporated embryos were cultured for 6 hours in a 50 ml disposable tube containing 4 ml of 50% Dulbecco's modified Eagle's medium supplemented with 50% rat serum (Lawrence and Struhl, 1996), a volume sufficient for the culture of eight embryos (Sturm and Tam, 1993). The tubes were rotated at 30 rpm on a roller apparatus placed in a 37°C incubator containing 5% CO₂. This culture condition is optimized for normal LR development so that left-sided Nodal expression in LPM is preserved in >95% of the cultured embryos. A Nodal or caALK4expression vector was introduced together with an EGFP expression vector by electroporation, and regions that received the vectors were confirmed by the presence of EGFP fluorescence.

Foxh1 null mutants exhibit defects in anteroposterior patterning and in node formation (Hoodless et al.,2001; Yamamoto et al.,2001). The early mortality of the null mutants, however, has impeded characterization of the role of Foxh1 at later stages of development. We have now examined the contribution of Foxh1 to LR patterning by generating conditional mutant mice. Mutant mice harboring a flox allele of Foxh1(Foxh1^flox) have been described(Yamamoto et al., 2001), and,in the present study, we aimed to delete Foxh1 in the lateral plate with the use of transgenic mice expressing the Cre recombinase in this tissue. The Cre-expressing transgene Lefty2-3.0 Cre contains a 3 kb fragment of the mouse Lefty2 promoter that directs gene expression in nascent mesoderm (Saijoh et al.,1999). One of the transgenic lines (77b) harboring this transgene was used in this study.

To confirm the expression pattern of the Cre transgene in 77b mice, we crossed the animals with Cre-reporter mice that express lacZin response to Cre activity (Sakai and Miyazaki, 1997). Embryos harboring both Lefty2-3.0 Creand the Cre-reporter lacZ transgene were recovered at various stages and stained for β-galactosidase activity with the substrate X-gal. Staining was first evident at embryonic day 6.75 (E6.75) in the nascent mesoderm. At E7.5, most of the embryonic region of the mesoderm exhibited staining, whereas the primitive streak, ectoderm and endoderm were negative(Fig. 1A). Mesoderm-specific staining remained evident at E8.2; however, the midline structures, including the axial mesoderm, lacked Cre activity(Fig. 1B). At this stage, the pattern of Cre activity differed between the anterior and posterior regions of the embryo (Fig. 1C). In the anterior region, X-gal staining was complete in the LPM, paraxial mesoderm,definitive endoderm and heart. In the region posterior to the node, however,Cre activity was detected only in the most lateral region of LPM. About 30 to 50% of the extra-embryonic mesoderm cells in the yolk sac and amnion were also positive for X-gal staining. At E9.5, mesoderm-derived cells in the anterior region were positive whereas the mesoderm in the posterior region was negative(data not shown). Thus, at stages later than the early somite stage, Cre activity was mostly restricted to the mesoderm of the anterior region of the embryo.

Foxh1^flox/flox mice were then mated withFoxh1^+/-, Lefty2-3.0 Cre mice to obtainFoxh1^flox/-, Lefty2-3.0 Cre mice (for simplicity, these conditional mutant mice are referred to as Foxh1^c/-). Cre-mediated deletion of Foxh1 in the Foxh1^c/-embryos was confirmed by examining Foxh1 expression. Whereas Foxh1 mRNA is abundant in all three germ layers of gastrulating wild-type embryos(Fig. 1D,E), Foxh1expression was absent in the mesoderm of Foxh1^c/- embryos(Fig. 1F,G).

We first examined LR patterning in the Foxh1^c/- embryos by analyzing the transcription of asymmetrically expressed genes such asNodal, Lefty1, Lefty2 and Pitx2. In wild-type embryos,Nodal is expressed in two domains at the early somite stage: the node and left LPM (Fig. 2A,B). In most Foxh1^c/- embryos examined (31/37, 84%), however,left-sided expression of Nodal in left LPM was absent(Fig. 2C); in the remaining embryos (6/37, 16%), a low level of Nodal expression was detected in a small region of left LPM adjacent to the node(Fig. 2D,E), probably because a Nodal-positive loop was operative in this region before deletion ofFoxh1 was complete. Nodal expression in the node was maintained in all (37/37) Foxh1^c/- embryos (insets inFig. 2C,D). The expression ofLefty2 apparent in left LPM of wild-type embryos(Fig. 2F) was abolished in all(16/16) Foxh1^c/- embryos examined(Fig. 2G). The expression ofLefty1 observed in the PFP of wild-type embryos(Fig. 2F) was also lost in all(16/16) Foxh1^c/- embryos(Fig. 2G) (in someFoxh1^c/- embryos, Lefty1 expression was detected in a small region of PFP adjacent to the node at the two-somite stage but this expression disappeared at four-somite stage). This latter effect was unexpected given that Foxh1 is preserved in midline structures including the PFP (Fig. 1B). The expression of Lefty1 apparent in the node of wild-type embryos was maintained in the mutant embryos (inset ofFig. 2G). The asymmetric expression of Pitx2 apparent in wild-type embryos(Fig. 2H,I,N-P) was abolished in two-thirds (29/43) of Foxh1^c/- embryos at E8.2(Fig. 2J) and in ∼70%(18/26) of the mutant embryos between E9.0 and E10.5(Fig. 2K,Q-S), whereas bilateral Pitx2 expression in the branchial arch was maintained in all Foxh1^c/- embryos. In the remaining mutant embryos, a reduced level of left-sided Pitx2 expression was detected both in LPM at E8.2 (Fig. 2L) and in several organs and other structures, including the common atrial chamber, lung bud, sinus venosus, vitelline vein, common cardinal vein and gut, at later stages (Fig. 2M,T-V).

Foxh1^c/- mice were carried to term but died within several days after birth. Examination of the visceral organs ofFoxh1^c/- neonates revealed various LR defects with right isomerism as the major phenotype (Fig. 3). Although the lungs of wild-type mice contain one and four lobes on the left and right sides, respectively(Fig. 3A), the lungs of most(40/45, 89%) of the Foxh1^c/- mice examined manifested right isomerism, having four lobes on both sides(Fig. 3B). A small proportion(3/45, 7%) of Foxh1^c/- mice exhibited partial right isomerism, having two or three lobes on the left side and four lobes on the right side. Abnormal positioning of the great arteries, either arterial transposition (9/41, 22%) (Fig. 3P,S) or double-outlet right ventricle (31/41, 76%)(Fig. 3Q,T), was also frequently observed in Foxh1^c/- mice. The position of the heart apex was reversed (toward the right; 15/41, 37%)(Fig. 3K), ambiguous (in the middle; 7/41, 17%) (Fig. 3J) or normal (toward the left; 19/41, 46%) in Foxh1^c/- mice. Additional heart malformations, including ventricular septal defect without valve defects (VSD; 25/41, 61%) (Fig. 3T), atrial septal defect without valve defects (ASD; 30/41, 73%)(Fig. 3Q) and endocardial cushion defect (common atrioventricular valve plus ASD and VSD; 10/41, 24%)(data not shown), were also observed. The external morphology of the atrium was also affected, showing right isomerism (11/43, 26%) or LR inversion(13/43, 30%). The azygos vein, which is normally located on the left side(Fig. 3C), was reversed (13/43,30%) (Fig. 3D), bilateral(14/43, 33%) or normal (16/43, 37%) in Foxh1^c/- mice. Additional defects included hypoplasia of the spleen (21/40, 53%)(Fig. 3F) and the presence of the stomach on the right side (8/43, 19%)(Fig. 3H). The relative positions of left and right renal veins were reversed (14/44, 32%) or the two veins were located at the same level (12/44, 27%)(Fig. 3M). The portal vein,which normally passes dorsally to the duodenum, passed ventrally to the duodenum in Foxh1^c/- mice (12/38, 32%)(Fig. 3N). These morphological defects resemble those observed with cryptic mutant mice(Yan et al., 1999), and are,in general, consistent with a lack of Nodal, the left-side determinant, in left LPM. The phenotype of Foxh1^c/- mice also resemble that of the zebrafish mutant lacking Foxh1(Chen et al., 1997; Bisgrove,2000).

Asymmetric Nodal expression in left LPM begins in a region adjacent to the node and expands along the anteroposterior axis. It is currently unknown how Nodal expression is initiated in left LPM; it may involve a positive autoregulatory mechanism, but direct evidence is lacking. To study the mechanism of Nodal induction, we developed a system for cell transplantation into LPM. A piece of tissue was dissected from the left LPM of a donor embryo at the four-somite stage (Nodal is already expressed in the entire left LPM at this stage) and was transplanted into the right LPM of a recipient embryo at the two-somite stage(Nodal expression is initiated but not expanded in left LPM at this stage) (Fig. 4A). When theFoxh1^c/- embryo was used as a recipient, a donor LPM was transplanted into the left LPM. We then examined whether the transplanted left LPM was able to induce Nodal expression in the right LPM of the recipient embryo. Indeed, the transplanted left LPM induced Nodalexpression in the recipient right LPM (3/3). Thus, Nodal expression in right LPM was not only apparent in the region adjacent to the transplanted left LPM but expanded along the anteroposterior axis(Fig. 4B). In most instances,the induced Nodal expression did not extend over the entire area of right LPM, possibly as a result of its premature termination by Lefty2 present in the transplanted left LPM (exogenous Nodal expression alone in the right LPM induced Nodal expression throughout the entire region of right LPM, as shown below). By contrast, transplantation of right LPM into the right LPM of a host embryo failed to induce Nodal expression in the recipient right LPM (3/3, data not shown).

We performed similar transplantation experiments withFoxh1^c/- embryos. Left LPM obtained from a wild-type embryo was thus transplanted to the anterior region of left LPM ofFoxh1^c/- embryos. Such manipulation failed to induceNodal expression in the left LPM of the host embryo (10/10)(Fig. 4F,G), indicating that the induction of Nodal expression in LPM requires Foxh1.

Our results suggested that an unknown factor derived from left LPM is able to initiate Nodal expression in right LPM. An obvious candidate for this factor was Nodal itself present in left LPM. To test this possibility, we introduced a Nodal expression vector and an EGFP expression vector into embryos at the early somite stage by electroporation. The embryos were first examined for fluorescence to locate the cells that received the vectors(Fig. 5A), and were subjected to in situ hybridization. Introduction of the Nodal vector into right LPM of wild-type embryos resulted in the induction of endogenousNodal expression (Fig. 5B,C). Nodal expression induced in right LPM expanded along the anteroposterior axis and extended throughout the entire region of the right LPM (25/25) (arrowheads in Fig. 5C). Electroporation of an EGFP expression vector alone into right LPM of wild-type embryos did not give rise to such expanded Nodalexpression except in one (1/20) case. By contrast, introduction of theNodal expression vector into the left or right LPM ofFoxh1^c/- embryos failed to induce Nodal(Fig. 5F). Thus, ectopic Nodal is able to induce Nodal expression in right LPM, but this induction requires the presence of Foxh1 in LPM.

Lefty1 plays an essential role as the midline barrier(Meno et al., 1998). InFoxh1^c/- embryos, Lefty1 expression in the floor plate was lost (Fig. 2G)despite the fact that Foxh1 was not deleted in the PFP(Fig. 1B). Furthermore, bothNodal (Fig. 2C) andLefty1 (Fig. 2G)expression was preserved in the node. These observations suggested thatLefty1 expression in the floor plate might be induced by Nodal produced in left LPM. Consistent with this notion, previous studies(Chen and Schier, 2001;Meno et al., 2001) have suggested that Nodal is able to act over a long distance. We therefore tested this possibility with the use of our transplantation and electroporation systems.

Lefty1 is expressed in the PFP on the left side of wild-type embryos (Meno et al., 1997). A piece of left LPM transplanted to the right LPM of wild-type embryos was able to induce not only Nodal in the right LPM but also Lefty1 in the right PFP (3/3) (Fig. 4C). Thus, Lefty1 expression in the PFP became bilateral only at the levels where Nodal was ectopically expressed in right LPM(Fig. 4D,E), supporting the idea that Nodal produced in LPM induces Lefty1 expression in the PFP. Similar experiments were performed with Foxh1^c/- embryos,which retain Foxh1 in the PFP but lack Lefty1 expression in this region. Transplantation of left LPM from wild-type embryos to the left LPM of Foxh1^c/- embryos did not result in the induction ofLefty1 expression in the PFP (0/10)(Fig. 4F,G). However,transplantation of the left LPM to the paraxial mesoderm, a site closer to the PFP, resulted in the induction of Lefty1 (3/4)(Fig. 4H,J). Furthermore, left LPM obtained from Lefty2^ΔASE/ΔASE embryos, in which Nodal activity is increased as a result of the lack of Lefty2(Meno et al., 2001), inducedLefty1 expression in the PFP even when transplanted to the left LPM of Foxh1^c/- embryos (4/5)(Fig. 4I,K).

Introduction of the Nodal expression vector into the right LPM of wild-type embryos also induced Lefty1 expression in the PFP (18/25)(Fig. 5D,E). The spatial level of ectopic Lefty1 expression along the anteroposterior axis of the PFP corresponded to that of ectopic Nodal expression in the right LPM. In most instances, Lefty1 expression was bilateral throughout the entire PFP, while ectopically induced Nodal expression extended throughout the entire region of right LPM(Fig. 5E). Furthermore,introduction of the Nodal expression vector into the left LPM ofFoxh1^c/- embryos also induced Lefty1 expression in the PFP (6/8) (Fig. 5G,H),making it unlikely that Lefty1 was induced by secondary signals produced by the Nodal-Foxh1 pathway. In these various experiments,Nodal expression was never induced in the PFP either by the transplanted left LPM or by introduction of the Nodal expression vector into right LPM (Fig. 4B,Fig. 5B).

These results suggest that Lefty1 expression in PFP is directly induced by Nodal produced in LPM but do not exclude an alternative possibility that it is induced by secondary factor(s) produced in LPM by a Nodal-dependent yet Foxh1-independent pathway. To test the latter possibility, we examined the effects of constitutive active ALK4 (caALK4). As expected, caALK4 was able to induce Nodal in the right LPM of the wild-type embryos (5/5)(Fig. 5I). However,introduction of the caALK4 expression vector into the left LPM ofFoxh1^c/- embryos failed to induce Lefty1expression in the PFP (11/11) (Fig. 5J). These results now demonstrate that Nodal activity produced in LPM directly induces Lefty1 expression in PFP.

Our results suggest that Nodal ectopically produced in LPM may diffuse over the relatively long distance to the PFP and there induce Lefty1expression. We next examined whether the PFP indeed receives Nodal signals from left LPM with the use of a lacZ transgene whose expression is strictly dependent on Nodal signaling. This transgene, (n2)7-lacZ,contains seven tandem repeats of a Foxh1 binding site and its expression is induced by the Nodal-Foxh1 pathway (Saijoh et al., 2000; Sakuma et al.,2002). X-gal staining of transgenic embryos harboring(n2)7-lacZ revealed lacZ expression in the PFP predominantly on the left side as well as in left LPM(Fig. 6A,B). Given thatNodal is not expressed in the PFP, Nodal produced elsewhere (either in left LPM or the node) must have traveled to the PFP. By contrast, X-gal staining of Foxh1^c/- embryos harboring the(n2)7-lacZ transgene revealed that lacZ expression was abolished in the left LPM and PFP, although staining in the allantois and at the base of the allantois remained (10/10)(Fig. 6C,D). Similarly, another Nodal-responsive lacZ reporter gene, Lefty2 ASE-lacZ, that contains the asymmetric enhancer (ASE) of Lefty2 also gave rise to X-gal staining in the PFP in addition to the left LPM of wild-type embryos(Saijoh et al., 1999);however, this transgene was inactive in the PFP ofFoxh1^c/- embryos (11/11) (data not shown).Foxh1^c/- mice lack Nodal expression in left LPM but that in the node is unaffected (Fig. 2C-E). Furthermore, they retain Foxh1 in the PFP(Fig. 1B,C). Finally,introduction of the Nodal expression vector into left LPM ofFoxh1^c/- embryos harboring the (n2)7-lacZ (7/7)(Fig. 6E,F) or Lefty2 ASE-lacZ (2/2) (data not shown) transgene resulted in expression oflacZ in the PFP. Together, these results suggest that Nodal synthesized in left LPM, not Nodal produced in the node, travels to the PFP and activates the Nodal-responsive lacZ transgenes.

Our results indicate that Nodal signaling induces the initiation and expansion of asymmetric Nodal expression in LPM as well as initiatesLefty1 expression at the midline. They also provide insight into the mechanism by which asymmetric signals are transferred between structures during LR patterning. On the basis of the present and previous data, we propose the following scenario (Fig. 7). First, Nodal produced in the node travels from the node to left LPM, where it initiates asymmetric Nodal expression(alternatively, Nodal may act on left LPM indirectly). Second, Nodal protein produced in the small region of left LPM adjacent to the node diffuses along the anteroposterior axis, resulting in the expansion of Nodalexpression within left LPM. Third, Nodal produced in left LPM also travels toward the midline, where it induces the expression of Lefty1, the product of which is crucial for midline barrier function. Thus, according to this scenario, two critical events of LR patterning, asymmetric expression ofNodal in LPM and Lefty1 expression at the midline, are established by diffusion of Nodal followed by Nodal signaling. Foxh1 is required at least for the induction of Nodal in left LPM.

How is asymmetric Nodal expression initiated in LPM? Our transplantation and electroporation experiments with mouse embryos revealed that ectopically expressed Nodal induced endogenous Nodal expression in right LPM. Nodal is thus able to initiate Nodal expression in LPM,suggesting the possibility that Nodal produced in the node travels to left LPM and there initiates Nodal expression. Previous observations are also consistent with this idea. First, Nodal expression in the node (the perinodal region) begins earlier than that in left LPM(Collignon et al., 1996). Second, asymmetric Nodal expression in left LPM begins in a small region adjacent to the node. Third, asymmetric Nodal expression in left LPM is controlled by a left side-specific enhancer designated ASE(Adachi et al., 1999;Norris et al., 2002;Norris and Robertson, 1999),the most critical elements of which are two Foxh1-binding sites(Adachi et al., 1999;Saijoh et al., 2000). These Foxh1-binding sites act as a Nodal-responsive element, suggesting thatNodal is regulated by a positive autoregulatory mechanism. Fourth,components required to mediate Nodal signaling (such as ALK4, ActRII, Cryptic,Smads and Foxh1) are all expressed in LPM on both sides. Fifth, in various mutant mice with LR defects, asymmetric Nodal expression in LPM is always absent when Nodal expression in the node is abolished(Lowe et al., 2001;Saijoh et al., 2003). Finally,the role of Nodal produced in the node has been more convincingly demonstrated by Brennan et al. (Brennan et al.,2002). Thus, mutant mice specifically lacking Nodal in the node fail to initiate asymmetric Nodal expression in LPM, revealing thatNodal expression in the node is indeed essential for asymmetric gene expression in left LPM (Brennan et al.,2002). Nonetheless, it remains to be seen whether Nodal coming from the node directly acts on left LPM to initiate Nodal expression. Other factors such as GDF1 may also be involved in signal transfer from the node to LPM, as GDF1 is expressed in the node and the lack of GDF1 results in the loss asymmetric Nodal expression in LPM(Rankin et al., 2000).

Foxh1 is expressed bilaterally in LPM at the early somite stage whenNodal is expressed in left LPM(Saijoh et al., 2000), and may function in both the initiation and amplification of Nodal expression in left LPM. Although it is difficult to distinguish these two processes experimentally, Foxh1 is implicated in both by our observation that the transplantation of left LPM to Foxh1^c/- embryos failed to induce Nodal expression even in the cells adjacent to the transplant site.

If Nodal synthesized in the node acts on left LPM, what might prevent Nodal activity from traveling toward the right side? Nodal flow, the leftward flow of extra-embryonic fluid in the node generated by vortical movement of the cilia (Nonaka et al., 1998),may transport Nodal preferentially toward the left side. Indeed, the role of nodal flow in LR patterning was recently demonstrated by testing the effects of artificial fluid flow in embryos(Nonaka et al., 2002). Although ciliated cells can be found in the organizer region of non-mammals,including the chick (Essner et al.,2002), fluid flow may not be generated there. Coincidentally,ectopic introduction of Nodal into the right LPM failed to induce endogenousNodal expression (M. Levin, PhD thesis, Harvard University, 1996). Thus, a different mechanism may operate in the chick for the transfer of asymmetric signals from the node to left LPM.

The midline structures, including the floor plate and notochord, are required to separate the two sides of the embryo(Danos and Yost, 1996), with Lefty1 being critical for midline barrier function(Meno et al., 1998). Our observations now suggest that Lefty1 expression in the PFP is induced by Nodal produced in left LPM. First, Fixhl^c/-embryos, which lack Nodal expression in left LPM but retain it in the node, fail to express Lefty1 in the PFP, suggesting that Nodal produced in the node is unable to induce Lefty1 expression in the PFP. Second, and more importantly, transplanted left LPM or a Nodalexpression vector introduced into right LPM induced Lefty1 expression in the PFP of wild-type embryos but not in that ofFoxh1^c/- embryos. Third, introduction of constitutively active ALK4 into the left LPM ofFoxhl^c/- embryos was unable to induceLefty1 expression in PFP, excluding a possibility that an unknown factor produced by a Nodal-dependent yet Foxh1-independent pathway inducesLefty1. The idea that Lefty1 expression is induced by LPM-derived Nodal is also consistent with previous observations. Comparison of the kinetics of Lefty1 and Nodal expression thus revealed that Lefty1 expression in the PFP is preceded by Nodalexpression in left LPM (C. M. et al., unpublished data). Furthermore,Nodal is not expressed in the PFP. Finally, mutant mice lacking a component of the Nodal signaling pathway, such as the coreceptor Cryptic, fail to express Lefty1 in the PFP as well as Nodal in left LPM(Yan et al., 1999).

After Nodal produced in left LPM travels to the midline and inducesLefty1 expression, Lefty1, which is also able to travel over long distances (Sakuma et al.,2002), might then be expected to diffuse toward the LPM. Lefty1 that reaches the right LPM would render it incompetent for Nodal signaling and prevent Nodal expression there. Lefty1 that reaches left LPM,together with Lefty2 produced in the left LPM, may contribute to rapid repression of Nodal expression in this region. Midline barrier function is abolished in mutant mice that lack Lefty1, resulting in bilateral expression of Nodal and Lefty2(Meno et al., 1998). We previously suggested that, in the absence of Lefty1, an unknown left-side determinant travels across the midline and reaches the right LPM, where it induces the expression of Nodal and Lefty2(Meno et al., 1998). Our data now suggest that this left-side determinant is most likely Nodal.

Although our results indicate that Lefty1 expression at the midline is induced by Nodal produced in left LPM, it is not clear whether this expression depends on Foxh1. Lefty1 expression is lost even in the least severe type of Foxh1-null mutant(Yamamoto et al., 2001). However, this effect may be secondary to misspecification of the midline cells in the absence of Foxh1 (Hoodless et al.,2001; Yamamoto et al.,2001). Our previous analysis of the transcriptional regulatory elements of Lefty1 by transgenic approaches(Saijoh et al., 1999)suggested that the 1.2 kb region immediately upstream of Lefty1 is sufficient for its asymmetric expression in the PFP. Although this 1.2 kb region contains three potential binding sites for Foxh1, mutation of these sequences did not impair the PFP-specific expression of Lefty1 (Y.S. and H.H., unpublished). Nodal signaling that induces Lefty1 in the PFP therefore may not involve Foxh1.

Overall, our results obtained with Foxh1^c/-mice suggest that Nodal activity travels from the node to left LPM, and from left LPM to the midline. A direct test of this conclusion will require visualization of the behavior of Nodal in mouse embryos.

We thank anonymous reviewers for many constructive comments; Yayoi Ikawa and Sachiko Ohishi for technical assistance; Jun Takeuchi and Toshihiko Ogura for introducing us to electroporation; Kohei Miyazono for a caALK4 expression vector; and Masaru Okabe for pCX-EGFP. This work was supported by a grant from CREST (Core Research for Evolutional Science and Technology) of the Japan Science and Technology Corporation (to H.H.), and by a grant from the Ministry of Education, Science, Sports, and Culture of Japan (to C.M.). M.Y. is a recipient of a fellowship from the Japan Society for the Promotion of Science for Japanese Junior Scientists.