XCR2, one of three Xenopus EGF-CFC genes, has a distinct role in the regulation of left-right patterning

Members of the EGF-CFC gene family encode small extracellular glycosylated proteins, and include Cripto and Cryptic in humans and mouse(Bamford et al., 2000; Ciccodicola et al., 1989; Dono et al., 1991; Dono et al., 1993; Shen et al., 1997), FRL1 in frogs (Kinoshita et al.,1995), one-eyed pinhead, oep, in zebrafish(Zhang et al., 1998) and Cripto/CFC in chick (Colas and Schoenwolf, 2000; Schlange et al., 2001). EGF-CFC family members share two conserved domains: an EGF-like domain and a CFC domain, which are conserved only among EGF-CFC family members (reviewed by Adamson et al.,2002; Minchiotti et al.,2002; Saloman et al.,2000; Shen and Schier,2000). Recent genetic and biochemical studies have demonstrated that EGF-CFC proteins act as essential co-factors for signaling by the nodal/GDF1 subset of TGFβ superfamily ligands (reviewed by Saloman et al., 2000; Schier, 2003). EGF-CFC proteins facilitate the association of nodal and GDF1/VG1 ligands with Type I and Type II transmembrane receptors (Cheng et al., 2003; Reissmann et al., 2001; Yan et al.,2002; Yeo and Whitman,2001). The EGF-like domain mediates association with nodal ligands, whereas the CFC domain interacts with the receptor complex. Loss-of-function studies have established that EGF-CFC proteins function as co-factors for nodal/GDF1 signals in both mesendoderm and left-right axis specification (Bamford et al.,2000; Ding et al.,1998; Gaio et al.,1999; Gritsman et al.,1999; Linask et al.,2003; Schlange et al.,2001; Xu et al.,1999; Yan et al.,1999).

The only reported frog EGF-CFC family member, FRL1, was originally identified in a screen for FGF receptor ligands(Kinoshita et al., 1995). Ectopic expression and antisense oligonucleotide studies have indicated that FRL1 is essential for the activation of ERK and subsequently for neural formation (Yabe et al.,2003), and is also an essential co-factor for the dorsal determinant WNT11 (Tao et al.,2005). Strikingly, however, no defects in either the transduction of nodal/GDF1 signals or left-right patterning were observed following FRL1 depletion, suggesting that the function of FRL1 might differ from that of EGF-CFC factors in other vertebrate model organisms. We report here the identification of two additional EGF-CFC family members in Xenopus. The three frog EGF-CFC factors show distinct spatial and temporal expression patterns during development. We have examined the effects of both loss and gain of function of one of these factors, XCR2, and have found that it is specifically required for normal left-right patterning of XNR1 and XATV expression, and for visceral asymmetry. XCR2 is expressed bilaterally at neurula-tailbud stages, but is required only on the left side for normal patterning. The requirement of XCR2 for left-side-specific expression of the nodal-related gene XNR1, is consistent with prior work demonstrating that the expression of nodal-related ligands are maintained by a FAST1/FOXH1-mediated positive-feedback loop (Adachi et al.,1999; Norris et al.,2002; Osada et al.,2000; Saijoh et al.,2000). In addition, we found that ectopic expression of XCR2 on the right side is sufficient to reverse left-right polarity,and this effect requires XCR2 activity as a co-receptor for nodal/GDF1 ligands. These observations indicate not only that XCR2 is required for nodal signaling during left-right patterning, but also that limitation of nodal signaling across the left-right axis by XCR2 can determine the polarity of the nodal gradient across this axis.

Using the BLAST search algorithm(Altschul et al., 1990), we identified EST clones corresponding to two novel Xenopus laevisEGF-CFC factors: (1) from the NIBB Mochii normalized Xenopus laevisneurula library, cDNA clone XL044d20 (GenBank UniGene Xl. 15503); and (2) from the NICHD_XGC_Emb1 Xenopus laevis library, cDNA clone IMAGE Consortium CloneID 6864330 (GenBank CA987644). We designated these novel EGF-CFC factors XCR2 and XCR3S (short form),respectively.

To generate pCS-XCR2, pCS2-XCR3S, pCS2-hCFC1, pCRII-XNR1 and pCRII-XATV,the ORF region of each EST clone (XCR2 or XCR3S) or cDNA insert of pSP64PA-humanCFC1 (Bamford et al.,2000), pCS2+XNR1 (Lustig et al., 1996) or pCS2+XATV(Cheng et al., 2000) was subcloned into pCS4+, a derivative of pCS2+ described by Yeo and Whitman(Yeo and Whitman, 2001), into the pCS2+ vector or into the pRII vector (Invitrogen). To generate pCS2-XCR3L(long form), the ORF region of XCR3L was amplified by PCR from cDNA of stage 10.5 embryos, and subcloned into the pCS2+ vector. pCS-XCR2 mEGF (mutated EGF)and pCS-XCR2 mCFC (mutated CFC) were generated by PCR-based subcloning: Asn93 and Thr96 to Gly93 and Ala96 in XCR2 mEGF1, and His129 and Trp132 to Gly129 and Gly132 in XCR2 mCFC. All constructs generated by PCR amplification were sequence verified.

Xenopus laevis genomic clones of XCR2 were isolated by PCR using the following primers: up (5′-AAGCAATTTCACATCAAC-3′) and down (5′-GGTGGGCCCCGCTGCCTCTAATG-3′; the linker sequence is underlined). A splice-site-targeted antisense morpholino oligonucleotide,TACACTCACTGTTAGTTCTTACCTC (XCR2 MO; 25-mer, underlined sequence is the region in the intron), was designed against the consensus sequence at the first exon-intron boundary derived from the sequence of two distinct XCR2genomic clones. A standard control morpholino oligo (SC MO) from GeneTools was used as a control.

Plasmids used for transfection were: pCS2+XNR1(Lustig et al., 1996);derrière/CS2+ (Sun et al.,1999); pCS-mouse nodal and pCS-mouse Cripto-3Flag(Yeo and Whitman, 2001);pCS-XCR2-3Flag that was generated by PCR-based subcloning and includes three repeats of the Flag epitope fused after Val191 of XCR2 in pCS-XCR2;pCS2-6Myc-mouse FAST1 (FOXH1) (Weisberg et al., 1998); Mix.2 ARE A3-lux(Chen et al., 1996); and pcDNA3.1/V5-His-TOPO/lacZ (Invitrogen). Human embryonic kidney 293T cells were cultured in 5% CO₂ at 37°C in DMEM supplemented with 10% FBS,100 U/ml of penicillin and 100 μg/ml of streptomycin. Transient transfection assays were performed by the calcium phosphate method in triplicate (Sambrook and Russell,2001). Plasmid mixtures contained 200 ng of each expression construct, 200 ng of the luciferase reporter plasmid, 80 ng of CMV-β-gal plasmid, and various amounts of pCS2+ vector to maintain a constant amount of total DNA. β-gal activity was confirmed to vary linearly with the quantity of plasmid transfected. Luciferase activity was measured 48 hours post-transfection and was normalized to the activity of co-transfectedβ-galactosidase activity. The fold differences in luciferase activity were calculated using the luciferase activity of the pCS2-6Myc-mouse FAST1(FAST1) transfected control as basal level activity. Expression of each epitope-tagged protein was confirmed by western blotting.

Xenopus laevis embryos were obtained by artificial fertilization,and embryos were staged according to Nieuwkoop and Faber(Nieuwkoop and Faber, 1956). Synthetic EGFP mRNA was made using SP6 mMESSAGE mMACHINE (Ambion),and was co-injected with XCR2 MO and/or various plasmids into the marginal region of one blastomere of two-cell-stage embryos. Embryos were sorted into left- or right-injected groups based on EGFP-fluorescence, and then fixed for in situ hybridization analysis, or scored for heart and gut morphology at stage 45-46 according to Branford et al.(Branford et al., 2000).

In situ hybridization analyses were performed according to Sive et al.(Sive et al., 2000), but with the addition of 0.3% CHAPS in 2×SCC and 0.2×SSC for probe washing. In Fig. 6I,K-V, Fig. 7C,D and Fig. 8A-K,M-Q, hybridization and wash steps were performed at 70°C. RNA probes were synthesized from pCRII-XNR1 and pCRII-XATV. After staining, pigmented embryos were bleached and cleared. Immunohistochemistry was carried out according to Faure et al.(Faure et al., 2000) and Lee et al. (Lee et al., 2001). The sarcomere myosin-specific antibody MF 20 was provided by the Developmental Studies Hybridoma Bank at the University of Iowa. The signal was detected with AlexaFluor594 donkey anti-mouse IgG (H+L) antibody (Molecular Probes), or with peroxidase-conjugated donkey anti-mouse IgG (H+L) antibody (Jackson ImmunoResearch), and the DAB Substrate Kit (Vector Laboratories).

Extraction of total RNA and reverse transcription were performed according to Watanabe and Whitman (Watanabe and Whitman, 1999). Ornithine decarboxylase (ODC)was used as an internal control. Reverse transcriptase negative (RT-)reactions were carried against all samples using ODC primers to confirm the absence of genomic DNA contamination. Primers used for PCR were: FRL1/XCR1 (up, 5′-CTGGTTTTTGCTAAGGACAC- 3′;down, 5′-TTGCAATGCTTGATAAAATG-3′); for temporal expression analysis of XCR2 (up, 5′-GCTGCGCATATGGGGTTCTT-3′; down,5′-CGATAATGCAGCCTTGTTTTCTCT-3′); for RNA splicing analysis of XCR2 (up, 5′-GCCCTTGGGATCCTTACATT-3′; down,5′-GAGTCAA- TGTTATAAATATGAAT-3′); XCR3 (up,5′-GCTGTAATTCGCTTGG-GAAC-3′; down,5′-TTTTGGACATGCACAGAAGC-3′); VG1 (up,5′-GACCGCTAACGATGAGTG-3′; down,5′-AGGAATGTCTTCTGGCTC-3′); XNR1(Lustig et al., 1996); ODC(http://www.xenbase.org/). PCR products were separated by agarose gel electrophoresiss and visualized by ethidium bromide staining. PCR products of XCR1 (FRL1), XCR2, XCR3S and XCR3L were recovered from gels and confirmed by sequencing.

Because studies of the only reported Xenopus EGF-CFC factor, FRL1, suggested a functional difference from the proposed role of EGF-CFC factors in other vertebrate embryo model systems(Shen and Schier, 2000; Tao et al., 2005; Yabe et al., 2003), we searched the X. laevis EST database for additional EGF-CFC family members. Full sequencing of two EST clones of potential novel EGF-CFC family members identified these as two novel Xenopus EGF-CFC genes, XCR2 and XCR3 (Fig. 1A). The XCR2 and XCR3 proteins are 34.6% and 30.6% identical to FRL1, respectively, and 30.6% identical to each other. Examination of the X. tropicalis EST and genomic databases did reveal the existence of clear X. tropicalis orthologs for each of the X. laevisEGF-CFC genes (data not shown). While their overall sequences are quite divergent from one another, the three X. laevis EGF-CFC proteins share substantial conservation in the EGF-like and CFC domains, and all contain a putative O-linked fucosylation site, which has been identified in mammalian cripto as being crucial for function as a nodal co-receptor (Schiffer et al.,2001; Yan et al.,2002). The three Xenopus EGF-CFC proteins are nearly identical in size, with the notable exception being that the XCR3transcript occurs in two alternatively spliced forms; the longer form encodes 72 additional amino acids near the N terminus of the protein(Fig. 1A). This additional stretch is cysteine-rich but shows no homology to other known proteins or domains. The full-length sequences for XCR2, XCR3S (short form) and XCR3L (long form) have been submitted to GenBank under Accession Numbers AY796186, AY796188 and AY796189, respectively. XCR2 and XCR3 have also been identified independently by Dorey and Hill (K. Dorey and C. Hill, personal communication). To render the nomenclature for Xenopus EGF-CFC factors consistent we, Dorey and Hill, and the investigators who originally identified FRL1(Kinoshita et al., 1995) (M. Kirschner, personal communication) propose to rename FRL1 as XCR1.

To begin to characterize the developmental roles of XenopusEGF-CFC factors, we examined the temporal expression patterns of XCR1,XCR2 and XCR3 by RT-PCR (Fig. 1B). Transcripts for all of the XCR genes were detected in unfertilized eggs, and XCR1 and XCR3 were expressed through early embryogenesis until the early neurula stage (stage 15). XCR2was expressed at very low levels maternally, expression increased markedly at the end of gastrulation (stage 12), and was maintained at high levels through tadpole stages (stage 45). XCR3 was detected as two bands corresponding to the two splicing variants, XCR3S and XCR3L,which showed similar temporal expression patterns. XCR1 and XCR3 were expressed ubiquitously before gastrulation(Fig. 2A), whereas XCR2 expression was too low to compare it in the different regions of the pre-gastrula embryo (not shown). We next examined the spatial expression patterns of the three XCR genes by in situ hybridization(Fig. 2). XCR1expression was strong in the entire prospective ectoderm at the beginning of gastrulation, and became restricted to the prospective neural plate by stage 12 (Fig. 2B,C), consistent with previous reports (Wessely et al.,2004; Yabe et al.,2003). XCR3 was strongly expressed in the Organizer and the prospective endoderm at the onset of gastrulation(Fig. 2D). This expression became enriched anteriorly during gastrulation(Fig. 2E). XCR2 was expressed at very low levels until stage 12, when its expression was detectable on the dorsal side (Fig. 2F). This dorsal expression was further restricted to the midline during neurulation (Fig. 2G),and in the notochord through neurula stages(Fig. 2H,I). XCR2 was also bilaterally expressed in the lateral plate region at stage 20(Fig. 2H,J). XCR2 was expressed in the prospective heart region at stage 25, and in the floor plate and dorsal endoderm in the dorsal midline(Fig. 2K,M). Dorsal midline expression decreased by stage 30, whereas the heart region expression was maintained (Fig. 2L,N). The markedly different temporal and spatial expression patterns of the XCR genes suggest that they have distinct developmental roles.

The distinctive expression pattern of XCR2 suggested a role for XCR2 that was distinct from those of XCR1 and XCR3. We therefore examined whether XCR2, like other EGF-CFC factors, acts to facilitate nodal-like ligand signaling. The activity of XCR2 on XNR1 or derrière signaling was tested in 293T cells transiently transfected with FAST1/FOXH1 and a nodal-responsive luciferase reporter, the Mix.2 ARE A3-luc reporter (Fig. 3). XNR1 or derrière increase luciferase activity in a FOXH1-dependent manner, but only in the presence of co-transfected XCR2. Similar results were seen with co-transfection of the related mouse genes Nodal and Cripto(Fig. 3). These results indicate that XCR2 can act as a co-factor for nodal-related ligands,as has been previously reported for other EGF-CFC factors. The fact that derrière signaling, as well as XNR1 signaling, is supported by XCR expression has not been reported, but is consistent with previous work showing that the structurally similar ligand VG1 requires EGF-CFC factors to signal,and that ectopic derrière shows only limited activity in zebrafish MZoep mutants (Cheng et al.,2003).

We next investigated the role of XCR2 in early frog development using antisense morpholino oligonucleotide (MO)-mediated depletion. Because preliminary experiments indicated that two antisense MOs targeting XCR2 translation were ineffective (data not shown), we designed an antisense MO targeting the first exon-intron junction to block XCR2pre-mRNA splicing. We cloned genomic DNA containing XCR2 by PCR, and found the position of the first intron to be conserved with known EGF-CFC genes (data not shown, GenBank Accession Number AY796187)(Colas and Schoenwolf, 2000). A 25 nucleotide antisense MO was designed against the first exon-intron boundary; three bases targeted the first exon and 22 bases targeted the first intron (Fig. 4A,B). We first examined whether the XCR2 MO blocks endogenous RNA splicing by RT-PCR, using primers that span the first intron. In XCR2 MO-injected embryos, the amount of mature XCR2 mRNA was decreased and an abnormal splicing variant was generated from a cryptic splice donor site(Fig. 4C). Correct splicing of XCR2 was inhibited through stage 35. Splicing of XCR2 mRNA was not affected by control MO injection. The abnormal splicing variant detected in XCR2 MO-injected embryos has an in-frame termination codon immediately following the end of exon I, resulting in an eight amino acid long predicted coding sequence (Fig. 4B). These results indicate that this splice-site-targeted antisense MO is an effective tool for the inhibition of endogenous XCR2 function.

We next examined the phenotype of XCR2 MO-injected embryos. The XCR2 MO did not cause any detectable defect in dorsoventral or anteroposterior axis patterning through stage 45, either in comparison to control MO-injected or uninjected embryos (Fig. 5A,B). There was also no effect of the XCR2 MO on movement or responsiveness of swimming tadpoles (data not shown). XCR2 MO-injected embryos were, however,extensively randomized with respect to left-right axis formation(Fig. 5C-H). The left-right patterning defects were scored according to Branford et al.(Branford et al., 2000); heart morphology is scored as normal or reversed, and gut morphology is scored for left or right origin of the gut, and for clockwise or counterclockwise coil direction. A low dose of XCR2 MO (20 ng) caused 23% reversed and 55% normal heart formation, 37% reversed and 41% normal gut origin, and 28% reversed and 50% normal gut coil direction (Table 1). A higher dose (50 ng) of XCR2 MO caused complete randomization of heart asymmetry (37% reversed and 39% normal). By contrast, 50 ng of the control MO had no significant effect on left-right patterning (90% and 89%normal heart and gut formation, respectively; Table 1).

XCR2 is expressed symmetrically in the lateral region at neurula stages (Fig. 2). This symmetric expression pattern is shared with other vertebrate EGF-CFC factors. EGF-CFC proteins are thought to be an essential component of a positive-feedback loop in which nodal-like ligands maintain and expand their own expression in the left lateral plate mesoderm (LPM) during the establishment of left-right axis patterning (Burdine and Schier,2000; Hamada et al.,2002). The function of EGF-CFC proteins on the right side,however, has not been specifically addressed. To examine the role of XCR2 in the left versus the right side of the embryo, embryos were injected XCR2 MO into one of two blastomeres after the first cleavage, which divides the embryo along the midline of the future left-right axis. Injection of 10 ng of XCR2 MO caused complete randomization of heart formation (48%reversed and 49% normal), and extensive, but incomplete, randomization of gut formation (23% mirror and 46% normal), when injected on the left side of the embryo (Table 2). By contrast,right-side injection did not affect left-right patterning, even at a dose of 20 ng of XCR2 MO (95% and 91% of normal heart and gut formation,respectively). These results indicate that left-side, but not right-side,expression of XCR2 is necessary for the correct establishment of left-right asymmetry. We also examined left-side-specific expression of XNR1 and XATV in the left LPM to establish whether XCR2 is required in the left LPM for expression of these genes. Left-side injection of 10 ng of XCR2 MO inhibited XNR1 expression in the LPM (84%), whereas right-side injection of XCR2 MO or left-side injection of control MO had no effect on XNR1 expression(Fig. 6A-J, Table 3). Similarly, XATV expression in the left LPM was inhibited by left-side injection of 10 ng of XCR2 MO (90%), but the dorsal midline expression of XATVwas not inhibited by XCR2 depletion(Fig. 6O-R, Table 4).

The symmetric expression of XNR1 has been reported to start before the left-right asymmetric expression of XNR1 is detectable(Lohr et al., 1997; Lowe et al., 1996). We could first detect bilateral XNR1 expression at stage 15 in two patches beside the presumptive notochord region, at the posterior of the archenteron roof (Fig. 7A-C), and left-side-specific XNR1 expression began in the LPM at stage 20(Fig. 7G,H, Table 8). Bilateral injection of XCR2 MO did not affect bilateral expression of XNR1 at stage 15,before left-side-specific XNR1 expression begins(Fig. 7E-F, Table 5). At stage 20, the initiation of expression of XNR1 in the left-side LPM was efficiently inhibited by XCR2 MO, but the bilateral expression of XNR1 was not affected (Fig. 7G-J, Table 5). Left-side-specific expression of XNR1 was not detectable at stage 18 and 19 (data not shown). These observations indicate that at post-gastrula stages, bilateral posterior expression of XNR1 is independent of XCR2, whereas left-side LPM specific expression is dependent on XCR2 at the earliest stage at which this expression can be detected.

To confirm the specificity of the effects of XCR2 MO on left-right asymmetric gene expression and visceral patterning, we examined whether ectopic expression of XCR2 can rescue the defects caused by XCR2 MO,using a CMV promoter-driven XCR2 expression plasmid (pCS-XCR2; Fig. 6D-R; Tables 3, 4). XNR1 and XATV expression in the LPM were severely reduced in 84% or 90% of embryos, respectively, injected with 10 ng of XCR2 MO on the left side. XNR1 and XATV expression in the LPM was restored in 86% or 45% of embryos co-injected with XCR2 MO and pCS-XCR2 plasmid on the left side. However, the spatial patterns of pCS-XCR2-rescued expression of XNR1were variable and often did not coincide well with the normal expression patterns of XNR1 (Fig. 6E), which was consistent with the inability of plasmid injection to accurately recapitulate the spatial pattern, timing and dose of endogenous XCR2 expression, confirmed by in situ hybridization of XCR2(data not shown). In addition, we examined the abilities of other Xenopus EGF-CFC members and human CFC1 (Cryptic) to rescue the effect of the XCR2 MO. Co-injection with XCR2 MO and a CMV promoter-driven XCR1, XCR3S, XCR3L or human CFC1 expression plasmid rescued XNR1 expression in the LPM in 50%, 83%, 82% or 59% of embryos, respectively (Fig. 6K-N, Table 3). We also examined left-right asymmetric visceral patterning in tadpoles co-injected with pCS-XCR2 and the XCR2 MO. Co-injection of 10 pg of pCS-XCR2 rescued the left-right asymmetry of the heart and gut (73% and 80% of normal heart and gut formation, respectively; see Table 6). The efficiency of the rescue of morphological asymmetry was not improved by increasing the amount of injected XCR2 plasmid (10, 20 and 40 pg). As in the case of the rescue of XNR1 expression, this probably reflects the variable distribution of the rescue plasmid. These observations demonstrate, however,the specificity of action of the antisense MO, and also suggest a significant functional overlap among different members of the EGF-CFC family in their ability to participate in signaling during left-right patterning.

To determine whether XCR2 may be limiting as well as necessary for nodal-family signaling during left-right patterning, we evaluated the effects of ectopic expression of XCR2 by injection of pCS-XCR2. Right-side-specific injection of 10-40 pg of pCS-XCR2 with control MO causes significant perturbation of the asymmetric expression of XNR1relative to the left-side injection(Table 3). The right-side injection of pCS-XCR2 with control MO also caused abnormal heart and gut asymmetry (Table 6). Interestingly, right-side injection of 40 pg of pCS-XCR2 not only caused ectopic right-side expression of XNR1 (38%), but also a significant number of reversals in the asymmetry of XNR1 expression (52%)(Table 3). To investigate whether ectopic expression of XCR2 on the right side is due to its function as a nodal/VG1 ligand co-receptor, we tested two mutants of XCR2 in this assay: XCR2 mEGF, which is mutated in the EGF-like domain required for binding to nodal family ligands, and XCR2 mCFC, which is mutated in the CFC domain and no longer interacts with Type I nodal/VG1 receptors (Yeo and Whitman, 2001). In contrast to injection of wild-type XCR2 on the right side, which caused ectopic XNR1 expression in the right LPM and inhibited endogenous XNR1 expression on the left side in 40% of embryos (Fig. 8B, Table 7),neither pCS-XCR2 mEGF nor pCS-XCR2 mCFC reversed the asymmetry in XNR1 expression (Fig. 8A,C,D; Table 7). These results indicate that the ability of ectopic XCR2 to change the left-side-specific XNR1 expression depends on the function of XCR2 as a co-receptor for nodal/VG1 family ligands.

To clarify how right-side misexpression of XCR2 causes left-right axis reversal, we examined the dynamics of XNR1 and XATVexpression in pCS-XCR2 injected embryos(Fig. 8, Tables 8, 9). Expression of XNR1or XATV in the right LPM following right-side injection of pCS-XCR2 was detectable by stage 15 in 17% (XNR1) or stage 17 in 45%(XATV) of injected embryos. This is in marked contrast to the normal expression of XNR1 or XATV in the LPM, which is first detectable at stage 20 (XNR1) or at stage 23 (XATV)(Fig. 8G-K,M-Q; Tables 8, 9). In embryos injected on the right side with pCS-XCR2, endogenous left-side XNR1 and XATVexpression was absent at stage 20 in 67% (XNR1) and at stage 23 in 73% (XATV) of embryos, respectively. The endogenous expression of XCR2 itself on the left side, however, was not affected by pCS-XCR2 injection on the right side (Fig. 8L,R). We also investigated whether the left-right asymmetry defect was overcome by XNR1 misexpression on the left side, by using an EF1α promoter-driven XNR1 expression plasmid(pXEX/XNR1) (Sampath et al.,1997). The heart and gut morphology was inverted by pCS-XCR2 injection into the right side in 57% and 46% of embryos, respectively(Table 10). Concomitant injection of 20 pg of pXEX/XNR1 on the left side rescued the orientation of heart and gut morphology in 79% and 64% of injected embryos,whereas 20 pg of the pXEX control vector did not. This observation indicates that ectopic expression of XCR2 on the right is sufficient to initiate premature, ectopic expression of XNR1 and XATV on the right side, leading to suppression of the normal expression of these factors on the left side.

We report here the identification of three different EGF-CFC family members and examine the role of one of them, XCR2, in Xenopusembryogenesis. Our results, and those of K. Dorey and C. Hill (personal communication), suggest that all three are likely to be involved in mediating signaling by nodal-related ligands during embryogenesis. The spatially restricted expression patterns of frog EGF-CFC genes and their necessary role in nodal-related ligand signaling events, strongly suggest that the expression of these factors is locally limiting for signaling by nodal-related ligands in the post-gastrula embryo. Antisense morpholino-mediated loss-of-function experiments show that XCR1 and XCR3 are important for nodal-related signaling during pre-gastrula germ layer specification and patterning, whereas XCR2 is specifically required for left-right patterning during neurula-tailbud stages, indicating that all three are likely to function as co-receptors (K. Dorey and C. Hill, personal communication). Although XCR2 is expressed bilaterally on both sides of late neurulae, it is required only on the left side for correct patterning,consistent with a model in which nodal-related signaling on the left side of the LPM is required for left-side specification.

Maintenance of localized expression of nodal ligands has been shown to depend on positive-feedback regulation mediated through a FAST1/FOXH1-responsive enhancer in the first intron of nodal orthologs in a broad range of chordate species (Whitman,2001). Depletion of XCR2 suppresses XNR1expression in the left-side LPM, consistent with a role for XCR2 as a co-factor in XNR1 signaling. The signals that initiate left-side-specific expression of XNR1 in the frog embryo are not known. One possibility is that the posterior bilateral expression of XNR1 is functionally comparable with nodal expression in the peri-nodal region of the mouse (Brennan et al., 2002; Saijoh et al., 2003), and that the XNR1 signal from this posterior region is responsible for the initiation of XNR1 expression in the left-side LPM through an XCR2-dependent pathway. A cilia-based mechanism for the left-right asymmetric distribution of nodal protein has been proposed(McGrath et al., 2003; Nonaka et al., 2002; Nonaka et al., 1998); whether a comparable mechanism might distribute the posterior-bilateral XNR1 asymmetrically is an interesting possibility, but it requires additional investigation. Alternatively, an asymmetric distribution of VG1 activity has been proposed as a mechanism for the initiation of left-right signaling(Chen et al., 2004; Hyatt et al., 1996; Kramer and Yost, 2002). This signal could also be dependent on XCR2 (alternatively it could depend on XCR1 and XCR3 (K. Dorey and C. Hill, personal communication), and therefore our data do not distinguish between these possibilities. We found that inhibition of XCR2 blocks left-side-specific expression of XNR1 as early as this expression is detectable. Although we cannot rule out the possibility that a non-XCR2-dependent signal initiates left-side XNR1 expression at a level not detectable by in situ hybridization, our data are consistent with the possibility that the initiation of asymmetric XNR1 expression is XCR2 dependent. Once asymmetric XNR1 expression is initiated, XCR2 is also likely to be required for the activity of the FAST1/FOXH1-mediated positive-feedback loop maintaining XNR1expression in the left LPM.

The bilateral expression of XNR1 in the posterior region of neurula stage embryos is not affected by depletion of XCR2. While it is possible that bilateral posterior expression is supported by residual XCR2, or by the perdurance of XCR1 or XCR3 from early embryogenesis, it seems likely that this expression is maintained by a mechanism distinct from the positive-feedback loop of XNR1. The pattern of XCR2- independent expression of XNR1 is similar to the expression of mouse Nodal in the peri-nodal region, which is independent of Cryptic (Gaio et al., 1999; Yan et al.,1999), indicating a broad conservation of this posterior pattern of neurula-stage gene expression of nodal ligands among vertebrates. It is possible that this posterior bilateral expression of XNR1 is transferred anteriorly in a left-right asymmetric manner to establish the asymmetric expression of XNR1 in the LPM(Wright, 2001), but the basis for this transition to asymmetry remains unclear.

In contrast to effects on the left side LPM, depletion of XCR2 on the right side of the embryo does not detectably alter left-right patterning. Although expression of both nodal family ligands and the nodal antagonists lefty/antivin is excluded from the right side(Branford et al., 2000; Cheng et al., 2000; Joseph and Melton, 1997; Lowe et al., 1996; Tanegashima et al., 2000), the extent or significance of their potential diffusion across and/or from the midline is not known. On the one hand, our results indicate that XCR2-dependent activity of neither the nodal family ligands nor their antagonists is essential on the right side for correct patterning. On the other hand, ectopic expression of XCR2 is sufficient to reverse the left-right polarity, as can ectopic expression of TGFβ family ligands or activated receptors (Chen et al.,2004; Hanafusa et al.,2000; Hyatt et al.,1996; Sampath et al.,1997; Toyoizumi et al.,2000). Data from several laboratories, including our own, have shown that EGF-CFC proteins function as co-factors for nodal/GDF1 signaling,but are not sufficient to activate SMAD2 signaling in absence of the ligands(Fig. 3) (Saijo et al., 2000; Yan et al., 2002; Yeo and Whitman, 2001; Reissmann et al., 2001). That ectopic XCR2 expression on the right side is enough to reverse left-right polarity indicates, therefore, that there are sufficient levels of nodal/GDF1 family ligands in the right-side LPM to initiate the left-side program when an exogenous co-receptor is provided. Ectopic XCR2 is also sufficient to activate ectopic XNR1 expression in the somites(Fig. 6), indicating that ectopic XCR2 can sensitize somites to endogenous nodal-related signals. This also suggests that the diffusion of XNR1 or related ligands is not restricted to the LPM at somite stages, although the possibility that nodal ligands are expressed at undetectable, but functionally significant, levels in the somite itself cannot be ruled out. Because left-side or right-side specific injections at the two-cell stage lead to variable expression in midline structures, our data do not distinguish a function for XCR2 expressed in midline structures such as the notochord in left-right patterning.

Why ectopic XCR2 expression on the right is sufficient to flip the left-right orientation of XNR1 expression remains an interesting theoretical question. This effect is eliminated by point mutations in either of the two domains required for XCR2 function as a nodal ligand co-receptor,strongly indicating that it is this function that mediates the observed effect on axis orientation. That ectopic XCR2 expression is sufficient to reverse the polarity of XNR1 expression suggests: (1) that XCR2 is limiting for activity of endogenous nodal/GDF1 family ligands on the right side of the embryo; (2) that elevated XCR2 enhances the activity of these ligands more than it enhances the activity of any lefty/antivin antagonists(Chen and Shen, 2004; Cheng et al., 2004; Tanegashima et al., 2004); (3)that the enhancement of activity of nodal/GDF1 ligands present on the right side is sufficient to establish a positive-feedback loop for XNR1expression in the right-side LPM; and (4) that this early XNR1expression on the right induces XATV, which in turn diffuses to suppress the normal activation of XNR1 signaling and expression on the left side.

The establishment of the left-right axis is a fascinating example of how an initial symmetry breaking event establishes a system of feedback loops that maintain a sharply divided asymmetry in the activity of a diffusible signaling molecule, in this case XNR1. Our observations suggest that XCR2 may be a critical limiting component of both the amplitude and the spatial extent of the left-side signal during left-right patterning. Consideration of this role for EGF-CFC factors will therefore be important for theoretical modeling of the dynamics of left-right patterning.

We thank Drs Marc Kirschner, Maximilian Muenke, Hazel Sive and Christopher V. E. Wright for gifts of plasmids; and the NIBB Xenopus laevis EST project and the American Type Culture Collection for EST clone resources. We thank Drs Caroline Hill and Karel Dorey for sharing results before publication. We thank Dr Michael Levin for helpful advice on photography. This work was supported by grants from the NICHD. C.-Y.Y. was supported by a grant(R08-2003-000-10943-0) from the Basic Research Program of the Korea Science& Engineering Foundation.