The Cerberus/Dan-family protein Charon is a negative regulator of Nodal signaling during left-right patterning in zebrafish

Establishment of left-right (L/R) patterning is one of the central processes of vertebrate embryogenesis. Studies in model animals show that some aspects of the mechanisms for L/R patterning are conserved among vertebrates(reviewed by Burdine and Schier,2000; Hamada et al.,2001; Long et al.,2002; Wright,2001; Wright and Halpern,2002; Yost, 1998). It is believed that the breaking of L/R symmetry, which takes place near organizer derivatives, such as the definitive node in mouse and Hensen's node in the chicken, contributes to the L/R patterning, although there are several reports showing that L/R asymmetry is initiated at the beginning of development (Kramer et al.,2002; Levin and Mercola,1999; Levin et al.,2002). In the mouse, cells on the ventral side of the definitive node have a monocilium that rotates in a counter-clockwise direction and generates a leftward flow of the extra-embryonic fluid (nodal flow)(Nonaka et al., 1998). Accumulated evidence from mouse studies suggest that the nodal flow plays a pivotal role in the L/R patterning (reviewed by Hamada et al., 2002; Tabin and Vogan, 2003). For example, inactivation of mouse left-right dynein (lrd),which is expressed in the node cells, abolishes nodal flow and randomizes L/R asymmetry (McGrath et al.,2003; Okada et al.,1999; Supp et al.,1997). In zebrafish, an lrd homologue (left-right dynein-related; lrdr) appears at the late gastrula stage in dorsal forerunner cells that migrate ahead of the dorsal organizer region(Essner et al., 2002). The dorsal forerunner cells give rise to Kupffer's vesicle during the early segmentation stages (Cooper and D'Amico,1996). Cells within Kupffer's vesicle have a cilium(Essner et al., 2002),suggesting a role for cilia in Kupffer's vesicle for L/R patterning.

The L/R-biased signals are thought to be transferred from the node to the lateral plate mesoderm (LPM), leading to the left side-asymmetric expression of nodal, lefty and the homeobox gene pitx2 in the LPM. Nodal is a member of the transforming growth factor-β (TGF-β) family of cytokines and regulates the expression of pitx2 and lefty, which encodes a feedback regulator for Nodal signaling in the left LPM (Essner et al., 2000; Liang et al., 2000; Shiratori et al., 2001; Yan et al., 1999; Yoshioka et al., 1998). Nodal also activates nodal expression through an auto-regulation mechanism that involves the transcription factor FoxHI(Long et al., 2003; Norris et al., 2002; Osada et al., 2000; Saijoh et al., 2000). In zebrafish, Nodal signaling is also involved in the L/R patterning of the diencephalon (Concha et al.,2000; Gamse et al.,2003; Liang et al.,2000; Long et al.,2003). For the initiation of L/R patterning in the mouse, Nodal is required not only in the LPM, but also near the node(Brennan et al., 2002; Saijoh et al., 2003).

L/R asymmetry is maintained by midline barriers, which block the transfer of the left-side determinants. These midline barriers function either within the organizer itself or within differentiated derivatives of the midline organizer, such as the notochord and floor plate(Bisgrove et al., 2000; Lohr et al., 1997; Schlange et al., 2001). In zebrafish mutants no tail, floating head and bozozok (also known as momo) that perturb midline development, there is an increase in the incidence of the bilateral expression of left side-specific genes in the LPM (Bisgrove et al., 2000; Danos and Yost, 1996). Loss of Lefty1, which is expressed in the left floor plate, in mouse is reported to cause left isomerism (Meno et al.,1998).

In zebrafish, there are three known nodal-related genes, cyclops (cyc), squint (sqt) and southpaw (spaw). cyc and spaw are expressed in the left LPM (Long et al.,2003; Rebagliati et al.,1998a; Sampath et al.,1998). cyc is also expressed in the left diencephalon, in a region that corresponds to the prospective parapineal and pineal bodies(Liang et al., 2000; Rebagliati et al., 1998a; Sampath et al., 1998). spaw is also expressed near Kupffer's vesicle in a similar way to nodal expression near the node in mouse(Long et al., 2003). Mutations in the cyc gene have a minimal effect on visceral organ asymmetry(Bisgrove et al., 2000; Chen et al., 1997; Chin et al., 2000). However,several independent lines of evidence implicate Nodal signaling in the establishment of L/R asymmetry within the zebrafish. Loss of the late zygotic function of One-eyed pinhead (Oep), an EGF-CFC-family protein required for Nodal signaling, leads to defects in left-side gene expression and in visceral organ and diencephalic laterality (Gamse et al., 2003; Liang et al.,2000; Yan et al.,1999). Mutations in schmalspur (sur), which encodes FoxH1 and mediates Nodal signaling, also lead to defects in L/R patterning (Bisgrove et al.,2000; Chen et al.,1997; Pogoda et al.,2000; Sirotkin et al.,2000). Finally, antisense morpholino (MO)-mediated inhibition of Spaw disrupts the left-side expression of cyc, pitx2, lefty1 and lefty2 and leads to defects in L/R patterning in visceral organs and in the diencephalon (Long et al.,2003). sqt is not expressed asymmetrically, but in the absence of sqt, asymmetric expression of spaw is disrupted(Long et al., 2003). In sum,all of these data strongly implicate Nodal signaling in L/R patterning in the zebrafish, with spaw having a major role within the LPM and in at least the initial steps of diencephalic asymmetry.

Members of the Cerberus/Dan family have been implicated in L/R patterning by virtue of the asymmetric expression patterns of some of these proteins. In chick, for instance, caronte exhibits left-side expression within the paraxial mesoderm and LPM (Rodriguez Esteban et al., 1999; Yokouchi et al., 1999). Thus, Caronte has been postulated to transmit a signal from the node to the left LPM. Caronte can function as an inhibitor of BMP signaling (Rodriguez Esteban et al.,1999; Yokouchi et al.,1999). However, it was reported that BMP signaling positively regulates nodal expression in the left LPM by inducing an EGF-CFC protein that is required for the LPM's competence to respond to Nodal ligands(Fujiwara et al., 2002; Piedra and Ros, 2002; Schlange et al., 2002; Schlange et al., 2001),raising the possibility that Caronte has functions other than inhibiting BMP signaling. Another member of the Cerberus/Dan family, Dante, is expressed around the node in mouse (Pearce et al.,1999). A role of Dante in L/R patterning has not yet been established.

Here we report the isolation of a novel gene, named charon, that encodes a Cerberus/Dan family secreted protein. charon is expressed from the early segmentation stages in the region embracing Kupffer's vesicle,adjacent and medial to the bilateral (`perinodal') spaw-expression domains. We found that Charon inhibits the activities of the Nodal-related proteins and identified Southpaw as a physiological target. Specifically, our data indicate that the antagonistic interaction between Charon and Nodal(Southpaw) plays an important role in L/R patterning in zebrafish.

Wild-type zebrafish (Danio rerio) embryos were obtained from natural crosses of fish from a pet shop and fish with the AB/India genetic background. bozozok (boz), no tail (ntl), sqt, cyc and oep were obtained after crossing fish heterozygous for the boz^m168, ntl^b195,sqt^cz35, cyc^m294 or oep^tz57mutations. The Fugu and flounder embryos were prepared as described(Hashimoto et al., 2002; Suzuki et al., 2002).

Four distinct Fugu and zebrafish genes homologous to chick caronte were found in the NCBI database and in the Fugu(Fugu rubripes) genome database of the Doe Joint Genome Institute(http://fugu.jgi-psf.org)by a BLAST search. Three of these genes encoded proteins with strong homology to Cerberus/Caronte, PRDC and Gremlin. The other was distantly related to any of the known Cerberus/Dan family proteins. Here, we describe the isolation of Fugu, zebrafish and flounder cDNAs of the gene charon, whose protein displayed the strongest homology to Cerberus/Caronte. The Fugu charon cDNA containing the whole open reading frame (ORF) was amplified by PCR with the following primer set: sense,5′-CGGGATCCCAGACGACAATTTTCCTGTTG-3′, and antisense,5′-CCATCGATGCAGGCGTCCCGAAGCTGCGT-3′. The resulting fragment was cloned into pBluescript II (pBS-fugu-charon). The cDNA fragment of zebrafish charon was isolated from a segmentation stage (15 hours post-fertilization, 15 hpf) cDNA library, which was constructed using a Marathon cDNA library synthesis kit (Clontech), with 5′ and 3′RACE using the primers 5′-GGTTTCACACTTGCACTCCTCAACG-3′ and 5′-GCACTCCTCAACGATCAGTACGCACC-3′ for 5′ RACE, and 5′-CAGCGCATAACGGAGGAGGGCTGTG-3′ and 5′-GGAGGGCTGTGAGACGGTGACCGTT-3′ for 3′ RACE. The resulting fragment was subcloned into pDrive (Qiagen) (pDrive-zcharon for the 5′RACE clone). After determining the 5′- and 3′ ends of the full-length charon cDNA, the zebrafish charon ORF was amplified by PCR with the following primers: sense,5′-CGGGATCCCGAAACCTTGAACCGCAAGATT-3′, and antisense,5′-CCATCGATGTAAATTAAACATATCTGTGTT-3′. The resulting fragment was cloned into pCS2MT or pCS2 (pCS2MT-zcharon or pCS2-zcharon). A part of the flounder charon cDNA was obtained by PCR with primers that corresponded to the conserved amino acids RVTAAGC and ETGREEK: sense,5′-AGCGTGTGACGGCGGCGGGATG-3′, and antisense,5′-CCTTTTCCTCGCGGCCTGTTTC-3′. The obtained fragment was verified by sequencing. Using this fragment as a probe, a putative full-length clone of flounder charon was isolated from a lambda ZipLox cDNA library of 20-somite flounder embryos, and the lambda phage clone was converted to the plasmid (pZL-fl-charon). The nucleotide sequences of zebrafish charon,Fugu charon and flounder charon were deposited in the DDBJ databank under accession numbers AB110416, AB110417 and AB1100418,respectively. Zebrafish PRDC and gremlin will be published elsewhere.

pcDNA3.1-Charon-Myc was constructed by insertion of the Charon-Myc fragment(a Myc tag in the carboxy terminal) from pCS2MT-zcharon into the BamHI and EcoRI sites of pcDNA3.1. pcDNA3.1-PRDC-Myc was constructed in a similar manner to that for pcDNA3.1-Charon-Myc. pcDNA3.1-HA-Spaw was constructed by insertion of the EcoRI-NotI fragment from pCS2+ActHASpaw(Long et al., 2003) into pcDNA3.1. pBS-fugu-charon was used for in situ hybridization of the Fugu embryo. A DIG-labeled riboprobe was made with T3 RNA polymerase(Promega) after NotI digestion. pZL-fl-charon was used for the flounder embryo. A DIG-labeled riboprobe was made with SP6 RNA polymerase(Promega) after SalI digestion. The zebrafish riboprobe was made with BamHI-digested pDrive-zcharon using SP6 RNA polymerase (Promega). Synthetic zebrafish charon RNA was produced with pCS2MT-zcharon or pCS2-zcharon. After NotI digestion, the RNA for Myc-tagged Charon or untagged Charon was transcribed in a solution containing an RNA cap structure analog (New England BioLabs) and SP6 RNA polymerase (Promega). Synthesis of RNAs for cyc, sqt and spaw was performed as described previously (Long et al., 2003; Rebagliati et al., 1998a; Rebagliati et al., 1998b). The method for detecting spaw, cyc, ntl, goosecoid, six3.2, pax2.1,nkx2.5 and cardiac myosin light chain (cmlc2)expression was also previously published(Chen and Fishman, 1996; Hashimoto et al., 2000; Long et al., 2003; Yelon et al., 1999). Whole-mount in situ hybridization and two-color staining were performed as described (Hashimoto et al.,2000; Long et al.,2003; Long and Rebagliati,2002).

COS7 cells in a 10 cm-diameter dish (approximately 10⁶ cells)were transfected with 10 μg of pcDNA3.1-Charon-Myc, pcDNA3.1-PRDC-Myc or pcDNA3.1-HA-Spaw by a standard calcium phosphate precipitation method. After 20 hours, the medium was changed from 10% fetal calf serum-contained DMEM to serum-free Opti-MEM1 (Invitrogen). After 48 hours, the supernatants were harvested. The supernatants containing Charon-Myc (25 μl), PRDC-Myc (25μl) or HA-Spaw (250 μl) were mixed in a combination described in Fig. 5 and incubated with anti-Myc (9E10, Invitrogen) or anti-HA (3F10, Roche) antibodies, and protein G sepharose, at 4°C for 15 hours. The precipitates were washed five times with washing buffer: 20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.1% Triton X-100,proteinase inhibitor Complete Mini EDTA-free (Roche), eluted with Laemmli's sodium dodecyl sulfate (SDS) loading buffer, and separated on an SDS-4 to 20%gradient polyacrylamide gel. The immune complexes were visualized by a chemiluminescence system (Western Lightning; PerkinElmer Life Sciences).

The antisense MOs were generated by Gene Tools (LLC, Corvallis, OR, USA). For the charon-MOs, the sequences were: charon-MO,5′-CAAAAAAGCCGACCTGAAAAGTCAT-3′, and 4mis-MO,5′-CAtAAAtGCCGACCTGAtAAGaCAT-3′ (lower case letters indicate mis-paired bases). The spaw-MO was previously published(spaw-MO1) (Long et al.,2003). The MOs were dissolved in and diluted with 1×Danieau's buffer (Nasevicius and Ekker,2000).

We are interested in elucidating how the node/organizer initiates L/R asymmetry within the LPM in the teleost embryo. To this end, we attempted to isolate fish homologues of chick caronte, which is reported to mediate a signal for left-side formation that is sent by the node to the left LPM (Rodriguez Esteban et al.,1999; Yokouchi et al.,1999). We searched a Fugu (Fugu rubripes) genome database and found one transcription unit deduced from the genome sequence that displayed a relatively strong homology to chick caronte. We named the gene charon and isolated the full-length charoncDNA by PCR. We found a partial coding fragment of zebrafish charonin the NCBI database (Z35724-a1466a08.p1c), and isolated the full-length cDNA of zebrafish charon by 5′ and 3′ RACE using the sequence information. We also isolated a flounder charon cDNA by combining degenerate PCR and a hybridization screen. The ORFs of the charoncDNAs consisted of 729 bp (zebrafish), 765 bp (Fugu) and 786 bp(flounder); they encoded 243, 255 and 262 amino acid residues, respectively. The zebrafish Charon protein exhibited a 42% amino acid identity to both the Fugu and flounder Charon proteins(Fig. 1A). In addition to the sequence similarities, the expression profiles of the Fugu, zebrafish and flounder charon genes were relatively similar to each other(Fig. 2), suggesting that they are orthologues.

The nine cysteine residues in the cysteine knot domain that are conserved among the Cerberus/Dan family were also conserved in Charon, indicating that Charon is a member of the Cerberus/Dan family. Within the cysteine knot domain, the zebrafish Charon protein had stronger similarities to Xenopus Cerberus (18%), chick Caronte (25%) and human Cerberus-related (20%) than to Dante or the other members of the Cerberus/Dan family (Fig. 1B). Outside of the cysteine knot domain, Charon did not have an apparent similarity to other Cerberus/Dan family proteins.

charon transcripts were first detected at the beginning of somitogenesis (2-3 somite stages) in the hypoblast of the tailbud region in zebrafish (Fig. 2A,B). At the 10-somite stage (14 hpf), the expression domain was observed as a horseshoe-shaped zone (with the anterior side open) in the tail region(Fig. 2C,D). Sagittal sectioning revealed that the charon-expressing cells were either within the epithelial lining of Kupffer's vesicle or very closely apposed to Kupffer's vesicle (Fig. 5M). The expression was strongest at the 10-somite through to the 14-somite stage(16 hpf) (Fig. 2G,H), then gradually disappeared, and it was not detected at 24 hpf. charon was not detected anywhere besides the region adjacent to Kupffer's vesicle at any developmental stage. charon was expressed in the region adjacent to Kupffer's vesicle in Fugu and flounder embryos, as in zebrafish(Fig. 2I-L).

We next examined the regulation of charon expression using the zebrafish mutants boz, ntl, cyc, sqt and oep. boz mutant embryos display variable defects in dorso-axial structures including the dorsal forerunner cells, which give rise to Kupffer's vesicle(Fekany et al., 1999). In boz mutant embryos, charon expression was reduced or not detectable, with a variable penetrance(Fig. 3B,C). The charon expression was not detected in the ntl mutant embryos(Fig. 3D), which display defective development of the dorsal forerunner cells, notochord and tail(Melby et al., 1996). The charon expression was not affected in the cyc mutant embryos(Fig. 3G), but it was reduced or absent in the sqt mutant embryos, which exhibit defective development of Kupffer's vesicle (Dougan et al., 2003) (Fig. 3E,F). In the oep mutant embryos, the charonexpression was comparable to that in the wild-type embryos, but the expression domain was smaller than that in the wild-type embryos, in proportion to the size of Kupffer's vesicle (Fig. 3H). The charon expression was not detected in embryos injected with a large amount of RNA for the Nodal/Activin inhibitor Atv/Lefty1(Thisse and Thisse, 1999)(Fig. 3I). The atv/lefty1 RNA-injected embryos did not have Kupffer's vesicle (data not shown). The data suggest that charon expression depends on the formation of the dorsal forerunner cells/Kupffer's vesicle, which is dependent on the Nodal signaling.

Cerberus/Dan-family proteins are inhibitors for TGF-β family proteins such as BMPs and a Nodal, and for Wnt proteins(Bell et al., 2003; Pearce et al., 1999; Piccolo et al., 1999; Rodriguez Esteban et al.,1999; Yokouchi et al.,1999). To investigate the function of Charon, we misexpressed charon RNA in zebrafish embryos. The RNA injection led to variable levels of defects in the formation of mesendoderm. The phenotypes were classified into three categories (Class I-III) with increasing severity(Fig. 4, Table 1). Class I embryos displayed a slightly shortened axial structure at 10 hpf (the time of yolk plug closure, YPC) (Fig. 4E),lacked the prechordal plate, had fused eyes and reduced notochord structure at the pharyngula stage (24 hpf, Fig. 4F-I). The phenotypes were similar to those observed in mutants with mildly affected Nodal signaling, such as zygotic oep and maternal-zygotic (MZ) sur (lacking maternal and zygotic Sur protein)(Pogoda et al., 2000; Sirotkin et al., 2000; Solnica-Krezel et al., 1996),or in embryos injected with a small amount of atv/lefty1 RNA. Class II embryos exhibited a severe defect in dorsal axis extension at YPC(Fig. 4J), and displayed a cyclopic eye, reduced trunk somites and loss of the notochord at the pharyngula stage (Fig. 4K-M). Class III embryos displayed a lack of dorsal axis extension and had a dorsal vegetal mass at YPC (Fig. 4N). They had a cyclopic eye, but lacked most mesoderm and endoderm at the phayngula stage (Fig. 4O-Q), as observed in MZoep embryos, cyc;sqt double mutant embryos, or embryos injected with a large amount of atv/lefty1 RNA(Feldman et al., 1998; Gritsman et al., 1999; Meno et al., 1999; Pogoda et al., 2000; Sirotkin et al., 2000; Solnica-Krezel et al., 1996),which are more completely deficient in Nodal signaling. The effects of the charon RNA injection depended on the dose of the injected RNA(Table 1). The phenotypes of charon-misexpressing embryos were different from those of embryos misexpressing the BMP inhibitors Noggin1 and Chordin (dorsalized phenotypes),or the Wnt inhibitor Dkk1 (anteriorized phenotypes)(Furthauer et al., 1999; Hashimoto et al., 2000; Miller-Bertoglio et al., 1997; Shinya et al., 2000). These data suggest that Charon functions as an inhibitor for Nodal signaling.

To further confirm an antagonistic role for Charon in Nodal signaling, we expressed charon RNA alone, or charon RNA with cyc,sqt or spaw RNA, and examined the expression of various genetic markers (Figs 4, 5). Misexpression of charon abolished goosecoid (gsc) and ntlexpression in the dorsal axial mesendoderm as well as a posterior expansion of the forebrain marker six3.2 (Fig. 4S,U,W,Y). All of these expression profiles are similar to those observed in MZoep and cyc;sqt mutant embryos and in atv/lefty1 RNA-injected embryos(Feldman et al., 1998; Gritsman et al., 1999; Meno et al., 1999; Thisse and Thisse, 1999). Injection of cyc (10 pg), sqt (10 pg) or spaw (100 pg) RNA led to dorsalization (data not shown), expansion and/or ectopic expression of gsc (Fig. 5A,C,E). Co-injection of 25 pg of charon RNA with these nodal RNAs suppressed the dorsalization, expansion and ectopic expression of gsc that were elicited by the nodal RNA injection (Fig. 5B,D,F). We also found that Charon but not PRDC (another member of the Cerberus/Dan family) interacted with Spaw (Fig. 5G). Taking these observations together with the phenotypes of the charon-misexpressing embryos, we conclude that Charon functions as an inhibitor for Nodal signaling.

Charon inhibited all the Nodal-related proteins in zebrafish in the misexpression studies. To address which Nodal-related ligand(s) is the physiological target for Charon, we first compared the expression profiles of charon and the nodal-related genes. sqt expression did not overlap with charon expression at any developmental stage(data not shown). cyc is expressed in the tailbud region at the bud stage, but its expression disappears by the 2-3 somite stage(Rebagliati et al., 1998a; Sampath et al., 1998),indicating that the expression of cyc coincides with that of charon spatially but not temporally. spaw displays a similar expression to charon in the tailbud region(Long et al., 2003); spaw expression is first detectable at the 4-6-somite stage (14 hpf),slightly later than the initial charon expression. The spaw-expressing cells were located bilaterally in two domains flanking (or possibly in cells lining) Kupffer's vesicle(Long et al., 2003)(Fig. 5H-J). During somitogenesis, spaw and charon continued to be expressed in adjacent regions close to Kupffer's vesicle(Fig. 5K-M). Two-color staining and examinations of cross-sectioned embryos revealed that the spaw-expressing cells were located dorsal and lateral to the charon-expressing cells (Fig. 5N-Q). The expression domains of spaw and charondid not overlap. However, both genes encoded secreted proteins, so their domains could overlap at the level of protein. The expression of both genes was maintained until the end of somitogenesis (data not shown). These data indicate that Spaw is the best candidate for a functional target for Charon.

We performed loss-of-function experiments by injecting antisense MOs against the translational initiation site of charon(charon-MO, Fig. 6). To examine the specificity of charon-MO, we co-injected charon-MO with charon RNA into embryos. The phenotypes caused by the misexpression of charon were suppressed by the co-injection of charon-MO (Fig. 6A,B, Table 1). Embryos injected with charon-MO alone (charon morphant embryos) did not show any gross morphological abnormalities during early development and survived for at least 1 week after hatching(Fig. 6L). However, we found that the charon morphant embryos displayed abnormal positioning of the heart (Fig. 6C-K, Table 2). The process of heart positioning can be divided into two steps, jogging and looping(Chen et al., 1997); jogging and looping are two aspects of heart L/R asymmetry. Wild-type embryos display a leftward shift (jog) of the heart tube from approximately 26-30 hpf and rightward (D-) looping from 30-60 hpf (Chen et al., 1997) (Table 2). The expression of the homeobox gene nkx2.5 and a cardiac-specific myosin light chain (cmlc2) gene can be used to visualize the jogging and looping of the heart(Chen and Fishman, 1996; Yelon et al., 1999). Using these markers, we found that the laterality of the heart jogging was severely perturbed in the charon morphant embryos, but not in the control morphant embryos (Fig. 6C-H, Table 2). Likewise, heart looping was disrupted in charon morphant embryos, as shown by significant increases in the fraction of embryos with reversed heart loops(L-loops) or unlooped hearts (Fig. 6I-K, Table 2). These data indicate that Charon is involved in L/R-biased heart positioning during heart formation.

To investigate how early the loss of Charon affects the L/R patterning, we analyzed the expression of the left side-specific genes, atv/lefty1,lefty2, pitx2, cyc and spaw. The nodal-related gene spaw is the earliest marker of embryonic L/R asymmetry in the zebrafish embryo and is required globally for correct L/R asymmetry(Long et al., 2003). Thus,changes in spaw expression provide an indicator of the general degree of disruption of overall L/R patterning. The charon morphant embryos exhibited bilateral expression of all of these genes in the lateral plate during somitogenesis (Fig. 7B,E,H, Fig. 8B,E,H, Table 3). The effects of the charon-MO depended on the dose of the injected morpholino. We also observed bilateral expression of atv/lefty1, cycand pitx2 in the diencephalon in these embryos(Fig. 7B, Fig. 8B,E,H, Table 3); these are normally expressed only on the left side (Fig. 7A, Fig. 8A,D,G)(Concha et al., 2000; Essner et al., 2000; Liang et al., 2000; Rebagliati et al., 1998a; Sampath et al., 1998). The data indicate that Charon is required for the asymmetric expression of these genes in the LPM and the diencephalon.

Dorsal midline tissues are required to maintain L/R asymmetry (`midline barrier'). It is thought that atv/lefty1 expression in the dorsal midline contributes to its barrier function(Meno et al., 1998). Defective midline development in zebrafish increases the incidence of bilateral expression of the left side-specific genes(Bisgrove et al., 2000; Danos and Yost, 1996). We examined the development of the midline tissues in the charonmorphant embryos. Expression of ntl in the notochord and sonic hedgehog (shh) in the notochord (13 hpf) and floor plate (24 hpf) was not affected in the charon morphant embryos(Fig. 7L,M,O,S,U). There was also no discernible difference in atv/lefty1 expression in the notochord and prechordal mesoderm at all somite stages examined (6-9 somite stage, Fig. 7Q; 12-13 and 17-20 somite stage, data not shown). We did not detect any morphological abnormality in the notochord and floor plate in these embryos(Fig. 7S,U). The data suggest that Charon is dispensable for the formation of the midline tissue and atv/lefty1 expression in the midline. Because charon is expressed only in the regions adjacent to Kupffer's vesicle but not in the LPM or diencephalon, Charon probably functions in the early process of the L/R patterning, which takes place in or near Kupffer's vesicle.

Left-side expression of spaw, pitx2, atv/lefty1 and lefty2 is lost in embryos with reduced Spaw activity(Long et al., 2003). The same effects are seen in mouse mutant embryos that lack nodal expression in the node (Brennan et al.,2002; Saijoh et al.,2003), suggesting a role for Spaw and Nodal in the initial left-side determination. Because charon was co-expressed with spaw in the tail region and Charon inhibited Spaw's function in the misexpression study, we thought it probable that the antagonistic interaction between Charon and Spaw in the tail region plays a role in L/R patterning. To address this, we conducted an epistatic analysis using the MOs for spaw and charon. We examined the expression of pitx2,atv/lefty1, lefty2 and cyc in embryos injected with both spaw-MO and charon-MO, or charon-MO alone. Injection of 0.8 ng of charon-MO occasionally increased the frequency of embryos expressing pitx2, cyc, atv/lefty1 or lefty2 on the right side, but most often led to bilateral expression of these genes in the LPM and diencephalon (for pitx2, cyc and atv/lefty1)(Fig. 8B,E,H, Table 3). Coinjection of 8 ng of spaw-MO with the charon-MO reduced or abolished the expression of these genes on both the left and right sides(Fig. 8C,F,I, Table 3). Because the same effect is observed in the spaw morphant embryos(Long et al., 2003), we conclude that spaw is epistatic to charon in the expression of the left-specific genes. These data provide additional evidence that Spaw functions as a left-side determinant and, more importantly, indicate a functional interaction between Charon and Spaw in the L/R patterning.

Many Cerberus/Dan family proteins, including Cerberus and Caronte, inhibit BMPs (reviewed by Balemans and Van Hul,2002); only a subset of this family, including Cerberus and Coco,can inhibit all BMP, Nodal and Wnt signaling(Bell et al., 2003; Piccolo et al., 1999). In this study, we found that Charon inhibits Nodal signaling (Figs 4, 5). The misexpression of charon did not elicit dorsalization of the embryos(Fig. 4). Because BMP inhibition is sufficient to dorsalize the zebrafish embryos(Myers et al., 2002), our results suggest that Charon is not a strong BMP inhibitor. Caronte has been shown to function as a BMP inhibitor and transmit a left-side signal from the node to LPM (Rodriguez Esteban et al.,1999; Yokouchi et al.,1999). The net effect of Caronte is induction of nodalexpression in the left LPM. Misexpression of caronte on the right side is sufficient to induce ectopic nodal expression within the right LPM. We identified a novel role of Cerberus/Dan family proteins in L/R patterning that is very different from the role of Caronte. Our data show that Charon acts formally as a negative regulator of Nodal expression within the right LPM, because loss of Charon function results in southpaw(nodal) expression within both the right and left LPM(Fig. 7, Table 3). The Charon-deficient embryos had a normal notochord and floor plate and maintained the normal atv/lefty1-dependent midline barrier. This suggests that Charon acts within or near Kupffer's vesicle to help restrict or counteract left-biased signals generated in or near Kupffer's vesicle. This would be opposite to the proposed effect of Caronte, which is to facilitate or mediate the transmission of a left-side signal from the node to LPM. The different modes of action of Charon and Caronte are consistent with the low absolute level of sequence similarity between these two proteins (Fig. 1). In addition, charon displayed a very different expression profile from cerberus and caronte, which are expressed in the deep endomesoderm and head mesenchyme in Xenopusembryos and in the left paraxial mesoderm and LPM in the segmentation stage in chick embryos, respectively, supporting the idea that Charon is a novel member of the Cerberus/Dan family proteins.

One of the Cerberus/Dan family proteins, mouse dante, is expressed in the definitive node from the bud stage(Pearce et al., 1999). dante expression in the mouse node is very similar to charonexpression in the Kupffer's vesicle region of the zebrafish, suggesting that Dante is an orthologue of Charon. However, only a partial sequence of dante has been reported and its biochemical function is not yet known, although it has been suggested that Dante functions as a Nodal inhibitor (Brennan et al.,2002). Thus, the relationship between Charon and Dante remains to be elucidated. Future comparison between the zebrafish charonmorphant embryos and dante-deficient mouse embryos, and biochemical analysis of Dante activity against Nodal will clarify this point.

In the mouse, dante and nodal are initially expressed symmetrically in the node region, and the dante expression becomes stronger on the right side of the node by early somitogenesis, in contrast to the stronger expression of nodal on the left side(Collignon et al., 1996; Pearce et al., 1999). We observed that the majority of embryos showed symmetric expression of charon (Fig. 2),implicating that the regulation of charon is different from that of dante in mouse. However, it could not be excluded that charon shows right-biased expression transiently.

Atv/Lefty1 and Lefty2, members of TGF-β family, function as feedback inhibitors for Nodal signaling (Cheng et al., 2000; Meno et al.,1999; Sakuma et al.,2002). atv/lefty1 is expressed in the Kupffer's vesicle region during early somitogenesis, and its expression is abrogated in oep mutant embryos (Bisgrove et al., 1999), suggesting that the atv/lefty1 expression in the Kupffer's vesicle region depends on Nodal signaling. The charonexpression in the Kupffer's vesicle region was dependent on the Nodal signaling. The expression of charon was abolished or strongly reduced in the sqt and atv/lefty1-injected embryos(Fig. 3), which display defective development of Kupffer's vesicle(Dougan et al., 2003). The charon expression was not affected in the cyc embryos, and was reduced in the oep embryos, with the severity of the decrease correlating with the reduction in the size of Kupffer's vesicle. These data suggest that the charon expression depends on the formation of Kupffer's vesicle and that Nodal signaling indirectly regulates the charon expression. Consistent with this, the charonexpression was abrogated or strongly reduced in the boz and ntl mutant embryos (Fig. 3), which have defects in the formation of Kupffer's vesicle(Fekany et al., 1999; Melby et al., 1996). However,because the aforementioned mutants also affect other aspects of mesendoderm development, we cannot rule out the possibility that the charonexpression is regulated by a signal distinct from those that induce morphological development of Kupffer's vesicle. Our data indicate that Charon functions to restrict laterality near Kupffer's vesicle, probably in cooperation with Atv/Lefty1.

Three lines of evidence argue that Southpaw is a physiological target for Charon. First, Charon interacted with Spaw biochemically and the dorsalizing activity of Spaw was inhibited by the overexpression of Charon(Fig. 5). Second, the expression domains of charon and spaw were in close proximity to each other in the tail region(Fig. 5), and both genes encode secreted proteins that will probably be secreted into overlapping regions. Third, the loss-of-function experiments showed that Spaw is epistatic to Charon in the expression of the left side-specific genes(Fig. 8, Table 3) in both the LPM and diencephalon. All of these data suggest that the inhibition of Spaw by Charon is involved in the generation of embryonic L/R asymmetry. The inhibition of Spaw's function leads to a reduction or loss of left side-specific gene expression (Long et al.,2003), suggesting that Spaw functions as a left-side determinant,as proposed for Nodal in mouse (Hamada et al., 2002). However, spaw is expressed not only in the Kupffer's vesicle region but also in the left LPM. The functional relevance of Spaw in the Kupffer's vesicle region has not been tested yet, as the spaw-MOs would block Spaw activity in all the spawexpression domains. In this study, Charon, which is expressed in Kupffer's vesicle, interacted functionally with Spaw, suggesting that Spaw may function in the Kupffer's vesicle region to initiate left-side determination. However,there is one significant argument against this idea: in ntl mutants, spaw expression domains in Kupffer's vesicle are lost, but spaw is expressed bilaterally in the LPM(Long et al., 2003),suggesting that Spaw in the Kupffer's vesicle region is not required for the expression of Spaw in the LPM. The situation is different from nodalin the mouse; elimination of nodal in the node region disrupts nodal expression in the left LPM in the mouse(Brennan et al., 2002; Saijoh et al., 2003).

There are several possible explanations for this discrepancy. First, a low level of Spaw could be transiently expressed in the Kupffer's vesicle region in the ntl mutant embryos and induce spaw expression in the LPM in the absence of the midline barriers and the Nodal inhibitor Charon. Second, if Nodal-class proteins turn over slowly, then pre-existing Nodal proteins, such as Cyc and Sqt, might compensate for the loss of Spaw in the Kupffer's vesicle region in the ntl mutant embryos. Consistent with this possibility, in ntl mutants, cyc expression is lost in the tailbud but is concomitantly upregulated or shifted into posterior-lateral territories at the bud stage (Rebagliati et al., 1998a). Charon could inhibit the dorsalizing activity of Cyc (Fig. 5). We could not exclude the possibility that there is a non-Nodal factor, possibly another TGF-β, which is expressed in the Kupffer's vesicle region and activates Nodal signaling in the LPM in the ntl mutant. In mouse, GDF1 expression in the node is required for the expression of nodal,lefty2 and pitx2 in the left LPM(Rankin et al., 2000). It is also possible that, in the absence of a midline barrier and Charon, spaw could be expressed in the LPM by a constitutive or default mechanism.

Loss of the Charon function by the morpholino-mediated inhibition led to bilateral expression of the left side-specific genes and randomization of heart jogging and looping (Figs 6, 7, Table 2). In this sense, the phenotypes of the charon morphant embryos are similar to those observed in zebrafish mutants boz (momo), ntl and flh, which have defects in the formation of midline structures(Bisgrove et al., 2000; Long et al., 2003), and in mutant mice deficient in lefty1, which is expressed in the left floor plate (Meno et al., 1998),raising the question as to whether Charon is involved in the formation of `the midline barrier'. However, the development of midline tissues, such as notochord and floor plate, and the expression of ntl, shh and atv/lefty1 were not affected in the charon morphant embryos(Fig. 7), suggesting that Charon is dispensable for the formation of the midline barrier. We cannot completely rule out the possibility that Charon is involved in expression or function of an unidentified component(s) of the midline barrier. Because charon is expressed only in the Kupffer's vesicle region, we favor the hypothesis that Charon functions near Kupffer's vesicle rather than regulating midline barrier functions of the notochord or floor plate.

lr dynein-related is expressed in the dorsal forerunner cells, and cells within Kupffer's vesicle have a cilium, in zebrafish(Essner et al., 2002),suggesting that Kupffer's vesicle is equivalent to the node cavity in mouse. Likewise, the fact that asymmetric expression of the nodal-related gene spaw begins near Kupffer's vesicle, and that this pattern is disrupted by mutations that affect forerunner cells, also supports this equivalence (Long et al.,2003). It has not yet been reported whether the cilia in the Kupffer's vesicle rotate and generate the leftward flow (nodal flow) in zebrafish. It has also not yet been shown whether the L/R signal is initiated in Kupffer's vesicle in zebrafish. Here, we have demonstrated that charon is expressed in or next to Kupffer's vesicle and that Charon functions in the L/R patterning without affecting the development of the midline tissue and atv/lefty1-dependent midline barrier, suggesting that L/R patterning is initiated in the Kupffer's vesicle region. Given that loss of molecular components of the mouse node and zebrafish Kupffer's vesicle can result in similar L/R defects, our data provide support for the proposal that these structures are functionally equivalent and that they have an evolutionarily conserved role in L/R patterning(Essner et al., 2002). It is tempting to speculate that Nodal-related molecule(s) such as Spaw (or possibly Cyc) function in the Kupffer's vesicle region in zebrafish in the same way as had been proposed for Nodal protein in the perinodal region in mice. In this scenario, the Nodal signaling is biased to the left side of the Kupffer's vesicle region in zebrafish. Charon might be involved in creating an `all or nothing' condition of Nodal signaling on each side of the Kupffer's vesicle region by reducing the Nodal signaling baseline.

In this study, we isolated Charon, a novel player in L/R patterning, and found antagonistic interactions between Charon and Nodal (Southpaw) to be involved in L/R patterning. This finding sheds new light on the role of Cerberus/Dan-related proteins in the process by which the L/R-biased signal is created during early vertebrate development.

We thank D. Yelon, C. Thisse, H. Yost, M. Kobayashi and H. Takeda for various useful plasmids; H. Takeda, W. S. Talbot and L. Solnica-Krezel for mutant fish; T. Shimizu and M. Nikaido for useful comments; and C. Fukae and Y. Kuga for excellent technical assistance. H.H. was funded by the Postdoctoral Fellowship from the Japan Society for the Promotion of Science(JSPS). This work was supported by Grants-in-Aid for Scientific Research from the Ministry of Education, Science, Sports, and Technology (KAKENHI 13138204 to M.H., KAKENHI 14034275 to T.S.) and from JSPS (KAKENHI 13680805 to M.H.), a grant from RIKEN (to M.H.), a Beginning Grant-in-Aid from the American Heart Association Heartland Affiliate (AHA 0360057Z) (to M.R.) and grants from the Bio-Design Project sponsored by the Ministry of Agriculture, Forestry and Fisheries, Japan (to T.S.).