Cerberus is a feedback inhibitor of Nodal asymmetric signaling in the chick embryo

Chick Cerberus (cCer; also known as Caronte) is a member of the Cerberus-Dan family of cysteine-knot-secreted proteins(Rodriguez Esteban et al.,1999; Yokouchi et al.,1999; Zhu et al.,1999). Cerberus-related proteins have been identified in other vertebrate species: the founding member Xenopus Cerberus (XCer)(Bouwmeester et al., 1996),zebrafish Charon (Hashimoto et al.,2004), mouse Cerberus-like (hereafter denominated Cerl-1)(Belo et al., 1997; Biben et al., 1998; Shawlot et al., 1998) and mouse Cerberus-like-2 (Cerl-2; also known as Dand5 or Dante)(Marques et al., 2004). Xenopus XCer, mouse Cerl-1 and chick Cer genes are syntenic(www.metazome.net)and, at early stages, are expressed in equivalent embryonic structures, such as the anterior endomesoderm, anterior visceral endoderm and hypoblast,respectively (Bouwmeester et al.,1996; Belo et al.,1997; Foley et al.,2000). Mouse Cerl-1 and chick Cer transcripts are also detected in the anterior definitive mesendoderm(Belo et al., 1997; Rodriguez Esteban et al.,1999). However, at later stages, Cerberus-related genes have very distinct patterns: XCer expression is no longer detected,mouse Cerl-1 transcripts are found in nascent somites and presomitic mesoderm, zebrafish charon and mouse Cerl-2 are expressed around the node region (Cerl-2 expression levels are higher on the right side), and chick Cer is expressed in the left paraxial and lateral plate mesoderm (Bouwmeester et al.,1996; Belo et al.,1997; Rodriguez Esteban et al., 1999; Marques et al.,2004; Hashimoto et al.,2004). The understanding of how these different patterns of expression are generated may bring some insights into the evolution of Cerberus-related genes and their functions in the different vertebrate species.

A conserved regulator of vertebrate left-right patterning is Nodal, a member of the transforming growth factor-β (TGF-β)family of signaling molecules that is expressed in the node region and left lateral plate mesoderm (reviewed by Hamada et al., 2002; Schier,2003). In the mouse embryo, Nodal activity is restricted to the left side by Cerl-2 (Marques et al.,2004), by the midline barrier and by Lefty2, a Nodal antagonist also expressed in the left lateral plate mesoderm (reviewed by Juan and Hamada, 2001). The left-side expression of Nodal and Lefty2 is directly regulated by Nodal itself. Our present findings demonstrate that cCerasymmetric expression is also directly activated by Nodal signaling and suggest that the cis-regulatory sequences of Cerberus-related genes have diverged among vertebrates.

Zebrafish charon, mouse Cerl-2 and chick Cerhave all been implicated in the determination of the left-right axis, but their functions seem to differ: zebrafish charon and mouse Cerl-2 have a role in preventing Nodal signals from crossing to the right side (Hashimoto et al.,2004; Marques et al.,2004), whereas chick Cer was reported to have a role in transferring the positional information from the node to the left lateral plate mesoderm (Rodriguez Esteban et al.,1999; Yokouchi et al.,1999). At the molecular level, Cerberus-related proteins behave as antagonists of members of the TGF-β family(Hsu et al., 1998; Rodriguez Esteban et al.,1999; Piccolo et al.,1999; Belo et al.,2000). During left-right patterning, zebrafish Charon and mouse Cerl-2 proteins were shown to act as Nodal antagonists(Hashimoto et al., 2004; Marques et al., 2004), whereas cCer has been proposed to act as a bone morphogenetic protein (BMP) antagonist(Rodriguez Esteban et al.,1999; Yokouchi et al.,1999). Chick Cer would allow the expression of Nodal in the left lateral plate mesoderm by inhibiting the repressive activity of BMPs on Nodal transcription. However, more recent reports have shown that BMP signaling is indeed essential for the activation of Nodal expression in the left lateral plate (Piedra and Ros,2002; Schlange et al.,2002), leaving the role of cCer in left-right patterning unexplained. Our results from overexpression and knockdown experiments demonstrate that cCer acts as a negative regulator of Nodal expression and prevents Nodal signaling from crossing to the right side. In conclusion, we propose that chick Cer, zebrafish Charon and mouse Cerl-2 evolved different regulatory mechanisms but retained a similar role in restricting Nodal activity to the left side.

A cCer genomic clone (clone MPMGc125J2191Q3 from RZPD, Germany)was isolated by screening a chicken cosmid library (RZPD no. 125) with a cCer sequence probe (gift from J. C. Izpisúa Belmonte, The Salk Institute, La Jolla, CA). Shorter DNA fragments of this clone were introduced into pBluescriptIIKS (Stratagene), sequenced and identified as containing the 5′, cDNA, intronic and 3′ regions of the cCer gene.

To recognize possible binding sites for known transcription factors, cCer 5′ genomic sequences were analyzed using MatInspector Professional release 7.4 (Quandt et al.,1995).

To identify the transcription initiation site(s), 5′ rapid amplification of cDNA ends was performed using total RNA from HH3-9 chick embryos and the RLM-RACE kit (Ambion). PCR products were size-fractionated by agarose gel electrophoresis, purified using a gel extraction kit (Qiagen),cloned into the pGEM-T Easy vector (Promega) and sequenced.

cCer 5′ genomic sequences were subcloned into an enhanced green fluorescence protein (EGFP) reporter vector containing the EGFPcoding sequence and the SV40 early mRNA polyadenylation signals from pEGFP-N3(Clontech). Deletions or point mutations of FoxH1- and SMAD-binding elements were designed according to the literature(Zhou et al., 1998; Mostert et al., 2001) and introduced into the Cer0.36-EGFP construct by PCR-based site-directed mutagenesis.

For the enhancer assays, cCer genomic sequences were either amplified by PCR (PCR1-5) or synthesized as complementary oligonucleotides,and subcloned into the p1229-EGFP enhancer-less vector. This vector carries the human β-globin minimal promoter and was generated by replacing the lacZ gene in the β-globinlacZ BGZA or p1229 vector(Yee and Rigby, 1993) with the EGFP coding sequence (Clontech).

Chick expression plasmids were based on a modified pCAGGS-MCS vector (gift from D. Henrique, Instituto de Medicina Molecular, Lisbon, Portugal)(Niwa et al., 1991). The coding sequence of Xenopus Cerberus-short (XCerS) was amplified by PCR from a pCS2-XCerS vector (gift from S. Piccolo)(Piccolo et al., 1999). The cCer coding sequence (cCerCDS) was isolated by reverse transcriptase(RT)-PCR according to the published sequence (GenBank accession no. AF179484)(Rodriguez Esteban et al.,1999) and subcloned into the XhoI and NotI sites of pCAGGS-MCS.

The pCAGGS-RFP vector (gift from D. Henrique), carrying the cDNA of monomeric red fluorescent protein (RFP; Clontech)(Campbell et al., 2002) under the control of the CAGGS promoter, was used to control the extent and efficiency of electroporation.

To generate the luciferase (luc) reporter constructs, cCerregulatory sequences were amplified by PCR (using Cer-EGFP plasmids as template) and subcloned into the pGL2-Basic vector (Promega).

Fluorescein-tagged antisense morpholino oligonucleotides (cCer MO:5′-CATGGTCCTGCTGATGCTGTAGATC-3′; cCer CoMO:5′-CATcGTCgTGCTcATGaTGTAcATC-3′, mismatches in lowercase) were designed and produced by Gene Tools. The efficacy of cCer morpholinos to inhibit the translation of Cer-Luc reporter constructs was tested in a cell-free transcription/translation system (see Fig. S3 in the supplementary material) (Summerton et al.,1997).

Fertilized chicken eggs (Quinta da Freiria) were incubated at 37.5°C for the appropriate period. Embryos were staged according to Hamburger and Hamilton (HH) (Hamburger and Hamilton,1951), explanted at HH stage 4-7 (HH4-7) together with the vitelline membrane and anchored to a metacrilate ring according to the protocol of New (New, 1955). Affigel-blue beads (Bio-Rad) were soaked in Shh protein [1 mg/ml in 0.1%bovine serum albumin (BSA)/phosphate-buffered saline (PBS); R&D Systems];heparin acrylic beads (Sigma) were soaked in recombinant Nodal protein (0.5 mg/ml; R&D Systems); and AG1-X2 anion-exchange beads (Bio-Rad) were soaked either in SU5402 [3 mM in dimethylsulfoxide (DMSO); Calbiochem](Mohammadi et al., 1997) or in SB-431542 (10 mM in DMSO; Tocris) (Inman et al., 2002). Treated embryos were cultured at 37.5°C in a humid chamber, fixed in 4% paraformaldehyde and processed for whole-mount in situ hybridization.

Whole-mount in situ hybridization on chicken and mouse embryos was performed as described by Liguori et al.(Liguori et al., 2003). Detailed descriptions of the RNA probes used are available from the authors on request. Embryos were developed with BM purple (Roche) for purple color and with INT/BCIP (Roche) for orange.

Embryos were processed for New culture(New, 1955) at HH3-5 and transferred into a silicon rubber pool containing a 2 mm-square cathode(CY700-1Y electrode; Nepa Gene). The ring was then covered with warmed Hank's buffer (GibcoBRL) and the embryo was injected with a DNA solution (0.5-3 mg/ml; 0.1% Fast Green; Sigma) using a pulled glass capillary and an IM-300 microinjector (Narishige). Electroporation was performed by placing a 2 mm-square anode (CY700-2 electrode; Nepa Gene) over the embryo and applying five pulses (10 V for 50 ms at 350 ms intervals) using a square wave electroporator (ECM830; BTX). The embryo was then placed on a 30 mm Petri dish with albumen (New, 1995), incubated for the appropriate period of time(7-48h), and observed under a fluorescence stereomicroscope (Leica MZ16FA).

Capped sense mouse Nodal mRNA was synthesized using the mMessage mMachine kit (Ambion). Eggs were obtained from Xenopus laevisfemales, cultured and microinjected as previously described(Medina et al., 2000). Embryonic stages were determined according to Nieuwkoop and Faber(Nieuwkoop and Faber, 1967). Xenopus embryos were injected in each animal blastomere of the eight-cell stage with a total of 200 pg of reporter plasmid, with or without Nodal mRNA (50 pg), and 25 pg of pTK-Renilla luciferase. Animal caps were isolated from the blastula stage, cultured until sibling embryos reached stage 12 and lysed in 20 μL of passive lysis buffer per cap. Firefly and Renilla luciferase values were obtained by analyzing 20 μL lysate by the standard protocol provided in the Dual Luciferase Assay kit (Promega) in a luminometer (MicroLumatPlus, Berthold Technologies). Each assay was performed in triplicate and repeated independently at least twice.

The transgenic mouse line Cer2.5-EGFP was generated by microinjection of linearized reporter construct DNA into the pronuclei of fertilized eggs from FVB mice, as described(Nagy et al., 2003). F1 embryos were collected from embryonic day (E)7.5 to E10.5, observed under a fluorescence stereomicroscope (Leica MZ16FA), fixed in 4% paraformaldehyde and processed for whole-mount in situ hybridization. For histological analyses,embryos were embedded in gelatin, cryosectioned and photographed under a fluorescence microscope (Leica DMRA2). In some slides, cell nuclei were labeled with DAPI (Molecular Probes).

In the chick embryo, sonic hedgehog (Shh) signaling positively regulates Nodal asymmetric expression(Pagán-Westphal and Tabin,1998). In turn, Nodal was shown to induce the expression of XCer (Osada et al.,2000) and mouse Cerl-1(Waldrip et al., 1998; Brennan et al., 2001). Therefore, the induction of cCer expression by Shh(Rodriguez Esteban et al.,1999; Yokouchi et al.,1999; Zhu et al.,1999) might be mediated by Nodal. To test this hypothesis, beads soaked in recombinant mouse Nodal protein were implanted on the right side of HH6 chick embryos and cCer expression was examined by whole-mount in situ hybridization. Indeed, ectopic expression of cCer was observed in the right-side mesoderm of these embryos (81%, n=33, Fig. 1A-F). cCer was also induced in embryos electroporated with a Dorsalin-Nodalexpression construct (pCAGGS-DcNodal)(Bertocchini and Stern, 2002),but with less efficiency (35%, n=23, data not shown).

It has been suggested that Shh induces Nodal in the chick lateral plate mesoderm via a secondary signal(Pagán-Westphal and Tabin,1998). In the mouse embryo, Nodal expression in the left lateral plate mesoderm is directly activated by Nodal protein produced in the node (Saijoh et al., 2003; Yamamoto et al., 2003). Accordingly, we observed that exogenous Nodal protein is able to activate Nodal expression in the right lateral plate mesoderm of chick embryos(68%, n=19, Fig. 1C-F). Nodal protein induced ectopic cCer expression in less than 1 hour (73%, n=11, Fig. 1B and data not shown) and Nodal expression in approximately 2 hours (71%, n=7, Fig. 1D), whereas beads soaked in Shh protein started to activate cCer and Nodaltranscription no sooner than 4-5 hours after implantation (66%, n=6, Fig. 1H; 0% 2 hours after implantation, n=5, data not shown). Taken together, these observations suggest that the transcription of cCer and Nodal is directly regulated by Nodal.

To determine whether endogenous Nodal signaling is necessary for cCer expression, chick embryos were electroporated with a pCAGGS expression vector containing the Nodal-specific antagonist Xenopus Cerberus-short (XCerS)(Piccolo et al., 1999; Bertocchini and Stern, 2002). Embryos were co-electroporated with pCAGGS-red fluorescent protein (RFP) and initially scored for the co-localization of RFP fluorescence and XCerS mRNA (data not shown). As expected, the inhibition of Nodal by XCerS resulted in the downregulation of cCer (89%, n=9, Fig. 1J) and Nodal(85%, n=13, data not shown), whereas control electroporations had no effect (n=10, Fig. 1I). Similarly, cCer expression was also repressed by SB-431542, an inhibitor of Nodal receptors (88%, n=16, see Fig. S1 in the supplementary material). Therefore, we conclude that endogenous Nodal signaling is required for normal activation of cCer and Nodal expression in the left lateral plate mesoderm. In addition,ectopic induction of cCer by Shh protein was inhibited by XCerS (86%, n=7, Fig. 1L), which demonstrates that Nodal signaling is required for the activation of cCer expression by Shh.

To investigate further whether cCer is a direct target of Nodal signaling, we analyzed the regulatory sequences responsible for cCertranscription in the left-side mesoderm. For this, cCer 5′genomic sequences of different lengths were subcloned into an enhanced green fluorescence protein (EGFP) reporter vector (Cer-EGFP constructs) and introduced into chick embryos by microinjection and electroporation in New culture (New, 1955). A representation of these constructs and their electroporation results are summarized in Fig. 2A.

Our initial results showed that a 2.5 kb DNA fragment upstream of the ATG of cCer (Cer2.5) was able to drive the expression of EGFPinto the cell populations that express cCer (i.e. the anterior mesendoderm and left paraxial and lateral plate mesoderm)(Rodriguez Esteban et al.,1999; Yokouchi et al.,1999; Zhu et al.,1999). Fluorescent cells are later detected in the foregut and heart (data not shown), which are derivatives of the chick anterior mesendoderm. The subsequent analysis of EGFP expression driven by shorter fragments (Cer0.9, Cer0.4 and Cer0.36) revealed a similar pattern(Fig. 3A,C, and data not shown). However, left-side expression was specifically disrupted in embryos electroporated with Cer0.34-EGFP and shorter constructs (data not shown). These observations indicate that the cCer left-side enhancer is located in the 360 bp 5′ region and that the -360 to -340 sequence contains an essential regulatory element.

To confirm that cCer left-side expression is directly activated by Nodal, we first analyzed the cCer left-side enhancer sequence and looked for the presence of FoxH1- and SMAD-binding sites. These transcription factors are nuclear effectors of the Nodal signaling pathway (reviewed in Schier and Shen, 2000) and were shown to directly regulate the asymmetric expression of the Nodal,Lefty2 and Pitx2 genes(Saijoh et al., 2000; Osada et al., 2000; Yashiro et al., 2000; Shiratori et al., 2001). Sequence analysis of the cCer 360 bp 5′ region (Cer0.36) using MatInspector software (Professional release 7.4) revealed the presence of a possible TATA box at -60, and consensus binding sites for several putative regulators of cCer transcription, including two FoxH1 and one SMAD element (Fig. 2B).

To determine whether the FoxH1 elements (F1 and F2) or the SMAD element (S)are necessary for the regulation of cCer asymmetric expression, we constructed Cer0.36-EGFP reporter vectors containing deletions (del)or mutations (mut) in each one of those sites that were previously shown to disrupt their activity (F1del, F1mut, F2del, F2mut, Sdel and Smut)(Zhou et al., 1998; Mostert et al., 2001)(indicated in Fig. 2B). Each of the F1del, F1mut, F2del, F2mut, Sdel and Smut constructs was electroporated into chick embryos, and EGFP fluorescence was analyzed both in the anterior mesendoderm and in the left-side mesoderm (results are summarized in Fig. 2C). All constructs were able to drive EGFP expression in the anterior mesendoderm(Fig. 3B and data not shown),but left-side expression was specifically abolished(Fig. 3D and data not shown). These observations demonstrate that the FoxH1 and SMAD sites in Cer0.36 are essential for the induction or maintenance of left-side transcription.

In addition, the functions of the FoxH1 and SMAD elements in the cCer left-side enhancer were quantified in luciferase reporter assays with Xenopus animal caps. The Cer0.36 reporter construct was clearly activated in the presence of Nodal(Fig. 3I). However, luciferase activity was reduced with the introduction of mutations in one of the FoxH1 or SMAD elements (F1mut, Smut and F2mut constructs; Fig. 3I). Taken together, our results indicate that the cCer left-side enhancer is directly activated by the Nodal-FoxH1/SMAD signaling pathway.

We next investigated whether the FoxH1 and SMAD elements in the cCer left-side enhancer are sufficient to induce left-side expression. For this, potential regulatory sequences were subcloned into enhancerless vectors that contain the human beta-globin minimal promoter upstream of either the EGFP or the luciferase reporter gene. The potential enhancer sequences tested were either shorter fragments of Cer0.36 (PCR1-5) or combinations of individual FoxH1 (F) and SMAD (S)elements (FSF, FF and FS; results are summarized in Fig. 2D). Embryos electroporated with PCR1, PCR2, PCR4, FF or FS did not display EGFPexpression in the left-side mesoderm (Fig. 3F,H, and data not shown). By contrast, asymmetric expression was detected in embryos electroporated with the PCR3, PCR5 or FSF constructs(Fig. 3E,G, and data not shown). Accordingly, luciferase activities of the reporter constructs that lack one of the FoxH1 or SMAD elements (PCR1, PCR2, FS and FF) were severely reduced when compared with those of PCR3 or FSF(Fig. 3J). These observations indicate that the FSF module in the cCer left-side enhancer is sufficient to activate asymmetric expression.

Asymmetric expression of cCer is induced by Shh on the left side and repressed by fibroblast growth factor 8 (FGF8) on the right side(Rodriguez Esteban et al.,1999; Yokouchi et al.,1999; Zhu et al.,1999). To investigate whether the cCer enhancer region is also regulated by these signaling molecules, chick embryos were electroporated with Cer0.4-EGFP and grafted on the right side of the node with beads soaked either in Shh protein or in the FGF receptor-1 inhibitor SU5402. In addition to the expected left-side pattern, EGFP expression was activated on the right side both by the Shh protein (100%, n=10, Fig. 4C) and by SU5402 (62%, n=13, Fig. 4D). These observations demonstrate that the cCer left-side enhancer is regulated by Shh and FGF signaling in the same way as cCerexpression.

To confirm that cCer enhancer activity is regulated by Nodal,embryos were electroporated with the Cer0.4-EGFP reporter construct and grafted with beads soaked in Nodal protein. As expected, Nodal was able to ectopically induce EGFP expression in the right-side mesoderm (100%, n=11, Fig. 4B; compare with control, n=5, Fig. 4A). Conversely, when embryos were co-electroporated with the PCR5-EGFP reporter construct and the Nodal antagonist XCerS(pCAGGS-XCerS), cCer enhancer activity was specifically repressed in the left-side mesoderm (64%, n=11, Fig. 4F; compare with control, n=4, Fig. 4E). XCerS did not have an effect on anterior mesendoderm expression (n=4, Fig. 4G,H). These observations indicate that Nodal signaling is required for the regulation of cCertranscription in the left-side mesoderm.

Chick and mouse Cerberus-related genes have both coincident and distinct domains of expression during embryonic development. At early stages,chick Cer and mouse Cerl-1 are both expressed in equivalent embryonic structures, such as the anterior mesendoderm(Rodriguez Esteban et al.,1999; Belo et al.,1997). However, at later stages, chick Cer is expressed in the left-side mesoderm (Rodriguez Esteban et al., 1999; Yokouchi et al., 1999; Zhu et al.,1999) (see also Fig. 1A), whereas mouse Cerl-1 expression is found in the rostral domain of the nascent somites and presomitic mesoderm(Belo et al., 1997) and mouse Cerl-2 is expressed in the node region(Marques et al., 2004). In order to determine whether the upstream regulators of cCer expression are conserved in mouse, we generated a transgenic line carrying the cCer regulatory region (Cer2.5-EGFP) and analyzed reporter gene expression in mouse embryos at different stages. At E7.5, EGFP fluorescence was detected in the anterior mesendoderm (data not shown), an expression domain common to chick Cer and mouse Cerl-1genes. However, at E8.5, EGFP was expressed in the left lateral plate mesoderm (Fig. 5A,B), which is a cCer-specific pattern. As in Cer-EGFP-electroporated chick embryos, fluorescent cells were also found in the foregut and heart of E8.5 transgenic embryos (Fig. 5Aa,B). These results indicate that the upstream regulators of cCer expression are present not only in tissues that express both cCer and mouse Cerl-1 (i.e. anterior mesendoderm), but also in the mouse left-side mesoderm, a region that expresses cCer but not the mouse Cerl genes.

In the mouse embryo, the asymmetric expression of both Nodal and Lefty2 is directly regulated by Nodal signaling(Saijoh et al., 2000; Saijoh et al., 2003; Yamamoto et al., 2003). In Cer2.5-EGFP mouse embryos, EGFP mRNA expression is exclusively detected in the left lateral plate mesoderm at E8.25, and coincides with the expression patterns of Nodal(Fig. 5C,C′) and Lefty2 (Fig. 5D,D′), which reinforces the hypothesis that cCerregulatory sequences are directly regulated by Nodal.

In the chick embryo, Lefty is expressed in the midline (as is mouse Lefty1) and in a small posterior domain of the left lateral plate mesoderm at late stages, whereas the cCer expression pattern is much more similar to that of mouse Lefty2 in the left-side mesoderm(Rodriguez Esteban et al.,1999; Ishimaru et al.,2000). Like Lefty proteins, Cerberus-related molecules were shown to act as Nodal antagonists in zebrafish(Hashimoto et al., 2004), Xenopus (Hsu et al.,1998; Piccolo et al.,1999), chick (Bertocchini and Stern, 2002) and mouse (Belo et al., 2000; Marques et al.,2004) embryos. Therefore, we proposed that cCer has taken the role of mouse Lefty2 in the left-side mesoderm, and acts to restrict the range of Nodal signaling. To test this hypothesis, we have performed cCeroverexpression and knockdown experiments in chick embryos. Because Nodal transcription is autoregulated, Nodal expression was analyzed as a readout of Nodal signaling.

Chick embryos electroporated on the left side with a pCAGGS expression vector containing the cCer coding sequence (pCAGGS-cCerCDS)showed a dramatic reduction or absence of Nodal expression in the left lateral plate mesoderm, but not in the node region (95%, n=20, Fig. 6B; compare with control, n=4, Fig. 6A). On the other hand, when cCer was misexpressed on the right side, Nodal was never ectopically induced (n=18, Fig. 6C) and was downregulated only in one embryo (6%, n=18, data not shown). In this embryo, it is possible that the cCer protein had traveled from the right to the left side,where it inhibited the Nodal signal. In addition, the expression of the Nodal target gene Pitx2 was downregulated by cCer on the left side (56%, n=16, data not shown), whereas it was never induced on the right side(n=11, data not shown). At older stages, chick embryos showed reversed heart looping when cCer was overexpressed on the left side(47%, n=15, data not shown), but not when cCer was misexpressed on the right side (n=7) nor in control electroporations(n=4). Taken together, these observations suggest that cCer may act as a negative regulator of Nodal signaling.

To investigate the effect of cCer downregulation on Nodalexpression, fluorescein-tagged morpholino oligonucleotides against cCer (MO) or against a related sequence (five mismatches; CoMO; see Fig. 2B) were electroporated into the future left-side mesoderm of HH4-6 chick embryos. At HH8-11, Nodal transcription was ectopically induced on the right side of cCer morphant embryos (MO: 51%, n=37, Fig. 6E,F versus CoMO: 7%, n=27, Fig. 6D). In nine HH10-11 morphant embryos, Nodal expression on the right was higher than on the left side (Fig. 6F). This observation can be explained by a right-side-biased amplification of Nodal signaling, as predicted by the self-enhancement and lateral inhibition (SELI) model in the absence of Nodal inhibitors(Nakaguchi et al., 2006). The Lefty midline expression domain was normal in cCer MO-treated embryos(n=26, see Fig. S2 in the supplementary material), suggesting that the midline barrier was not affected. At older stages, cCer knockdown resulted in the inversion of heart looping (43%, n=7 versus CoMO: 0%, n=5, data not shown). These results indicate that the main function of cCer in the left-side mesoderm is to prevent Nodal signaling from crossing to the right side.

We have demonstrated that Nodal is sufficient and necessary for the induction of cCer expression (Fig. 1). However, previous studies have reported that Nodal is unable to activate cCer expression in the right lateral plate mesoderm(Rodriguez Esteban et al.,1999; Yokouchi et al.,1999; Zhu et al.,1999). The inconsistency between these and our results may be a consequence of the usage of different Nodal overexpression methods: we grafted beads soaked in active Nodal protein (mature form; R&D Systems), whereas others used retroviral vectors carrying a Bmp4-Nodal fusion protein that were introduced either by direct injection(Rodriguez Esteban et al.,1999) or by implantation of expressing cell pellets(Yokouchi et al., 1999; Zhu et al., 1999). In fact,induction of ectopic cCer expression was much less efficient when we used a Dorsalin-Nodal fusion construct (34%, versus 81% with Nodal protein beads) and was never detected when we overexpressed chick Nodal complete cDNA(n=10, data not shown). Therefore, it is possible that the proprotein convertase required for Nodal maturation is present at low levels in the right side of the chick embryo (Constam and Robertson, 2000), or that the Nodal protein encoded by these constructs is less stable than the recombinant protein(Le Good et al., 2005).

In zebrafish, Xenopus and mouse embryos, Nodal expression in the left lateral plate mesoderm is directly induced by Nodal protein released by the node (Long et al.,2003; Osada et al.,2000; Saijoh et al.,2003; Yamamoto et al.,2003). Accordingly, our results indicate that Nodal signaling positively regulates Nodal expression in the left lateral plate in the chick embryo (Fig. 1). This observation supports the hypothesis that Nodal itself is the intermediary signal that transfers the asymmetric information from the node to the lateral plate (Pagán-Westphal and Tabin,1998).

Here we have provided evidence that the FoxH1 and SMAD elements present in the cCer left-side enhancer are essential and sufficient for the activation of asymmetric expression (Figs 2, 3). FoxH1 and SMAD transcription factors are nuclear mediators of Nodal signaling (reviewed by Schier and Shen, 2000). As expected, Nodal is necessary and sufficient to activate the cCerleft-side enhancer (Fig. 4). These observations, together with the evidence that cCer asymmetric expression is activated by Nodal (Fig. 1), indicate that Nodal signaling directly regulates cCertranscription in the left-side mesoderm via the activity of FoxH1 and SMAD factors. In the future, the isolation of chick FoxH1 and the investigation of its direct binding and activation of the FoxH1 elements present in the cCer left-side enhancer may bring additional support to these results.

Previous reports have proposed that cCer is able to induce ectopic Nodal expression on the right side by antagonizing the repressive activity of BMPs on Nodal transcription(Rodriguez Esteban et al.,1999; Yokouchi et al.,1999). However, BMPs can have opposite effects on Nodalexpression: Bmp4 is a negative regulator of Nodal in the right side of the chick node at early stages (HH5-6)(Monsoro-Burq and Le Douarin,2001), whereas Bmp2 positively regulates Nodal signaling in the lateral plate mesoderm at later stages (HH7-8)(Piedra and Ros, 2002; Schlange et al., 2002). Because cCer was introduced at early stages (HH6) on the right side of the node (Rodriguez Esteban et al.,1999; Yokouchi et al.,1999), it might be inducing Nodal expression by blocking the inhibitory function of Bmp4 in the node region. In fact, when cCer-expressing cells were implanted at later stages (HH7-8) in the lateral plate, Nodal expression was not affected(Zhu et al., 1999).

In our hands, cCer misexpression in the node region (n=7,data not shown) or right lateral plate(Fig. 6C) was never able to induce Nodal, whereas cCer overexpression on the left side actually repressed Nodal (Fig. 6B). Conversely, Nodal was ectopically expressed on the right side of cCer-knockdown embryos (Fig. 6E,F). These findings revealed that cCer acts as a negative regulator of Nodal signaling. Similar results have been obtained with the Nodal antagonist Lefty: Nodal activity was repressed by the ectopic expression of chick Lefty (Rodriguez Esteban et al., 1999) or mouse Lefty1 or Lefty2(Yoshioka et al., 1998) in the chick embryo, whereas it was upregulated on the right side of Lefty2 mutant mice (Meno et al., 2001). Given the similarities between the expression patterns and functions of chick Cer and mouse Lefty2 (reviewed by Juan and Hamada, 2001), we propose that cCer has taken the role of mouse Lefty2 in left-right patterning,and acts in addition to the midline barrier to confine Nodal signaling to the left side.

Taken together, our findings suggest a feedback mechanism by which Nodal signaling is restricted in the left lateral plate mesoderm of the chick embryo(Fig. 7). During the establishment of the left-right axis, Nodal expression is first activated in the left perinodal region by the Notch and Shh signaling pathways(reviewed by Raya and Izpisua-Belmonte,2004). This initial Nodal signal directly induces cCerexpression via the activation of the cCer left-side enhancer by FoxH1 and SMAD transcription factors. We hypothesize that SMAD2 and/or SMAD3 regulate cCer transcription, because they are thought to transduce Nodal signal in the left-side mesoderm (reviewed by Schier, 2003) and phospho-SMAD2 has been detected in the chick lateral plate mesoderm(Faure et al., 2002). Additionally, we propose that, as in the mouse embryo (reviewed in Hamada et al., 2002), Nodal also activates its own transcription, leading to the amplification of Nodal and cCer expression throughout the left lateral plate. The partial overlap of the Nodal and cCer expression domains is possibly determined by functional differences in their regulatory regions and/or in Nodal and cCer diffusion rates, as proposed for the mouse Nodal and Lefty2 proteins (Nakaguchi et al.,2006). Together with the midline barrier, cCer has a key role in preventing the Nodal signal from crossing to the right side. Ultimately, the negative-feedback regulation of Nodal signaling by cCer results in the downregulation of Nodal and cCer expression in the left lateral plate mesoderm. Further support for this model may come from the analysis of chick Nodal transcriptional regulation as well as from the investigation of the diffusion rates and stability of Nodal and cCer proteins.

Unlike other known Cerberus-related genes, cCer is expressed on the left side of the paraxial and lateral plate mesoderm. Variations in the expression patterns of orthologous genes may arise either from the presence of particular cis-regulatory elements in their genomic sequence, or from the existence of differences in the localization and activation status of their upstream regulators, or both. Cross-species studies of cis-regulatory sequences are likely to help distinguish between these two hypotheses. In our study, the analysis of Cer-EGFP transgenic mouse embryos revealed that the upstream regulators of the cCer left-side enhancer (i.e. Nodal) are present in the mouse left lateral plate mesoderm,and suggested that the regulatory regions of Cerberus-related genes have diverged in chick and mouse. In agreement with this, the comparison between the cCer left-side enhancer and non-coding sequences of human, mouse, Xenopus and Fugu Cerberus-related genes using ConSite and VISTA programs was unable to detect any conserved FoxH1-binding sites or other common regulatory elements. FoxH1- and SMAD-binding sites are indeed present in the asymmetric enhancers of several left-side-specific genes, such as the ascidian, Xenopus, mouse and human Nodalgenes, mouse and human Lefty2 genes, and mouse and Xenopus Pitx2 genes (Saijoh et al.,2000; Osada et al.,2000; Yashiro et al.,2000; Shiratori et al.,2001). Our findings add cCer to this list, and underscore the essential role of evolutionarily conserved FoxH1-SMAD modules in the transcriptional regulation of asymmetric gene expression(Osada et al., 2000).

Although chick Cer, zebrafish charon and mouse Cerl-2 have different expression patterns, the Cerberus-related proteins encoded by these genes seem to have a conserved function in left-right development, which is to restrict Nodal signaling to the left side of the embryo. Nodal expression in the node region also differs among vertebrate species: it is bilateral in zebrafish and in early mouse embryos,whereas it is restricted to the left side in the chick embryo (reviewed in Raya and Izpisua-Belmonte,2004). This difference may justify the need for a Nodal antagonist in the node of zebrafish and mouse embryos, which is not required in the chick node.

The divergence in gene regulation between chick, Xenopus and mouse Cerberus homologues, here demonstrated by the presence of a FoxH1-SMAD module in the cCer regulatory region, has conveyed a novel scenery for the activity of cCer and enabled it to take the role of Lefty2 as a negative-feedback regulator of Nodal signaling in the left lateral plate. In the future, the analysis of novel Cerberus-related molecules involved in the left-right development of other vertebrate species (such as Xenopusand rabbit) should provide further insight into the evolution of Cerberus gene regulation and function.

We are grateful to J. C. Izpisúa Belmonte, H. Hamada, D. Henrique,S. Piccolo and C. D. Stern for probes and plasmids; S. Marques, N. Moreno, I. Marques and B. Lenhart for technical assistance; and V. Teixeira, A. Jacinto and L. Saúde for critically reading the manuscript. This work was supported by Fundação para a Ciência e a Tecnologia(POCTI/CBO/46691/2002, POCTI/BME/46257/2002 and POCI/SAU-MMO/59725/2004),Centro de Biologia do Desenvolvimento and IGC/Fundação Calouste Gulbenkian, where J.A.B. is a Principal Investigator.