Frizzled receptor dimerization is sufficient to activate the Wnt/β-catenin pathway

The Wnt family of secreted glycoproteins regulates a large number of developmental processes including cell growth, cell polarity and tissue specification. Wnts are crucial cell signaling molecules during development,as well as in adult life, and deregulation of their signaling pathway has major oncogenic effects (Cadigan and Nusse, 1997; Miller et al.,1999). The current data suggest that Wnts signal through seven-transmembrane receptors of the frizzled family(Bhanot et al., 1996). Frizzleds are required for Wnt signaling in vivo, as depletion of both Drosophila frizzled 1 and 2 (Dfz1 and Dfz2) results in wingless-like phenotypes (Bhanot et al.,1999; Bhat, 1998; Chen and Struhl, 1999; Kennerdell and Carthew, 1998). Furthermore, Wnts can bind to soluble forms of the frizzled proteins(Hsieh et al., 1999a; Hsieh et al., 1999b; Wang et al., 1997), which may act as a dominant-negative form of the receptor(Cadigan et al., 1998; Hsieh et al., 1999a; Hsieh et al., 1999b; Wang et al., 1997; Zhang and Carthew, 1998). Frizzled proteins share common structural motifs: a putative N-terminal signal sequence, a conserved cysteine-rich domain (CRD), a hydrophilic and highly divergent linker region, seven transmembrane segments and a variable intracellular C-terminal domain. The CRD is believed to act as the Wnt binding site for the receptor. Indeed, in Drosophila, the CRD of Dfz2 is necessary and sufficient for binding the wingless ligand(Cadigan et al., 1998; Rulifson et al., 2000; Zhang and Carthew, 1998). Several groups have shown that co-receptors for frizzleds are essential for Wnt signaling (Pinson et al.,2000; Tamai et al.,2000; Wehrli et al.,2000). They are related to the low-density-lipoprotein (LDL)receptor family and named LRP5 and LRP6. LRP6 is able to bind Wnt-1 and to associate with the Fz8 CRD in a Wnt-dependent fashion(Tamai et al., 2000).

A crucial function of the Wnt pathway is to activateβ-catenin-dependent transcription. In the absence of Wnt receptor activation, the modular protein Axin provides a scaffold for the binding of glycogen synthase kinase-3β (GSK3), adenomatous polyposis coli protein(APC) and β-catenin. This, in turn, facilitates β-catenin phosphorylation by GSK3 (Ikeda et al.,1998; Itoh et al.,1998) and leads to the rapid degradation of β-catenin via the ubiquitin pathway (Aberle et al.,1997). Upon Wnt's binding of the frizzled receptor, the Axin–GSK3–APC–β-catenin complex is disrupted by a process involving the cytoplasmic protein Dishevelled (Dsh), and dephosphorylation and recruitment of Axin to LRP5, which is associated with Axin destabilization (Mao et al.,2001; Smalley et al.,1999; Yamamoto et al.,1999). Subsequently, the β-catenin is no longer targeted for ubiquitin degradation and has been shown to accumulate in the nuclei(Fagotto et al., 1998; Yost et al., 1996), where it may interact with members of the lymphoid enhancer factor (LEF)/T-cell factor(TCF) classes of transcription factors to regulate the expression of target genes such as siamois (Brannon et al., 1997; Carnac et al.,1996; Lemaire et al.,1995; Molenaar et al.,1996). In Xenopus, Wnt/β-catenin signaling plays a crucial role in dorsoventral axis specification. Ventral overexpression of certain Wnts or frizzleds, Dsh or β-catenin ectopically activates siamois transcription and leads to the generation of a complete secondary axis (Moon and Kimelman,1998; Sokol,1999). In normal embryos, β-catenin protein accumulates in the cytoplasm and nuclei of dorsal blastomeres during early cleavage stages(Larabell et al., 1997). Moreover, overepression of GSK3 and Axin or depletion of maternalβ-catenin RNA causes deficiencies in dorsal structures(He et al., 1995; Heasman et al., 1994; Yost et al., 1998; Zeng et al., 1997).

The biochemical mechanisms by which the binding of the Wnt ligand to its frizzled receptor elicits signal transduction within the cell are poorly characterized. Numerous studies have suggested that several G-protein-coupled receptor (GPCR) families exist as dimers or even higher structures(Devi, 2001; Gouldson et al., 2000; Hebert and Bouvier, 1998; Milligan, 2001). Recently,biophysical methods based on luminescence and fluorescence energy transfer have confirmed the existence of such oligomeric complexes in living cells(Angers et al., 2000; Angers et al., 2001; Kroeger et al., 2001; Overton and Blumer, 2000). However, whether dimerization is a general property of this class of receptors and whether this is functionally relevant for signal transduction remains controversial (Cvejic and Devi,1997; George et al.,1998; Gouldson et al.,1998; Hebert et al.,1998; Marshall et al.,1999). In several cases, receptors appear to fold as constitutive dimers shortly after biosynthesis, whereas ligand-promoted dimerization at the cell surface has been proposed for others(Jones et al., 1998; Kaupmann et al., 1998; Kuner et al., 1999; White et al., 1998). Dimerization is required for normal functioning of β-adrenergic receptors and has been shown to rescue the function of mutant forms of β-adrenergic and angiotensin type I receptors (Hebert et al., 1998).

In this report, we address the question of the potential dimerization property of two Xenopus frizzled receptors and the potential role of this dimerization in Wnt/β-catenin signal transduction. Taking advantage of the differential ability of Xfz3 and Xfz7 to activate the Wnt/β-catenin pathway when overexpressed in the Xenopus embryo(Umbhauer et al., 2000a), we carried out a comparative study of Xfz3 and Xfz7, which revealed an unexpected correlation between the presence of frizzled dimers and activation of the Wnt/β-catenin pathway. We show that Xfz3 but not Xfz7 exists as a homodimer when expressed in the Xenopus embryo and activates transcription of the Wnt/β-catenin target gene siamois. Moreover, forced homodimerization of Xfz7 is sufficient for activation of the Wnt/β-catenin pathway, even in the absence of the CRD ligand-binding domain. We therefore propose that dimerization is a crucial step in the transduction of the Wnt/β-catenin signal by frizzled receptors.

All constructs (Fig. 1) are cloned in pCS2+ vector (Turner and Weintraub, 1994). Myc-tagged version of wild-type Xfz3 (Xfz3-myc)and Xfz3ΔC (Xfz3ΔC-myc) has been previously described(Umbhauer et al., 2000a). Myc-tagged Xfz7 (Xfz7-myc) was made by insertion of an EcoRI/NotI digested PCR fragment of Xfz7 cDNA into pCS2-Xfz3-myc digested by EcoRI and NotI. The PCR fragment was amplified from pCS2-Xfz7 (Djiane et al., 2000) using the upstream primer 5′-TAGGAATTCCCAGCAGTACCACGGG-3′ and the T3 primer.

To obtain Xfz3-HA, the coding sequence of Xfz3 was PCR amplified with an EcoRV site added at the C-terminus and cloned in frame with two HA epitopes. To obtain Xfz7ΔC-flag, a BamHI/ClaI digested fragment corresponding to the 526 first amino acids of Xfz7 was inserted in frame in the pCS2-extraXfz7-flag (Djiane et al., 2000) digested by BamHI and ClaI. To make a myc version of the N-terminal extracellular region of Xfz3(extra3-myc), a nontagged construct corresponding to the first 196 amino acids of Xfz3 was first cloned in pSP64T (Krieg and Melton, 1987). Then, a XhoI/XbaI fragment was inserted in pCS2-Xfz3-myc digested with XhoI and XbaI. To obtain extra3-flag, pCS2-extraXfz7-flag(Djiane et al., 2000) was digested by BamHI and ClaI to replace extraXfz7 with a PCR fragment corresponding to the first 196 amino acids of Xfz3. The antisense primer was 5′-TGAATCGATAGGCAAAGGAGAG-3′ and the sense was the SP6 primer.

The Xfz7-BD-myc construct was obtained by EcoRI/BglII digestion of pCS2-Xfz7-myc and insertion of an EcoRI/BglII-digested PCR fragment amplified from TELmod vector (Lopez et al., 1999). The upstream primer was: 5′-CGGGAATTCGCGCTTGCAGCCAA-TT-3′; the downstream primer was: 5′-GCAAGATCTGCTGAAGG-AGTTCATAGAG-3′.

The fibroblast growth factor receptor-1 (FGFR-1) fusion proteins were obtained as follows: for extra3-R1, the coding sequence for the transmembrane and intracellular domains of FGFR-1(Friesel and Dawid, 1991) was PCR amplified (upstream primer, 5′-CGGAGCT-CCAACTGGAAATT-3′;downstream primer, 5′-ATTATCTAGATC-AGCGTTTTTTAAG-3′), digested by SacI and XbaI and inserted into pCS2-Xfz3-myc digested by the same enzymes. For Xfz3ΔC-R1, the intracellular region of FGFR-1 was PCR amplified (upstream primer, 5′-CGCTTCGAACACCCGTCGAAG-3′;downstream primer, 5′-ATTATCTAGATCAGCGTTTTTTAAG-3′), digested by BstBI and XbaI and inserted into pCS2-Xfz3-myc digested by the same enzymes. Torso-FGFR-1 has been previously described in(Umbhauer et al., 2000b). The wild-type FGFR-1 is as in Amaya et al.(Amaya et al., 1991).

All constructs were checked by sequencing. Plasma membrane subcellular localization in blastula animal cap cells was verified by immunodetection and confocal microscopy for all myc and HA tagged constructs bearing a transmembrane domain.

Xenopus eggs were obtained from females injected with 500 IU of human chorionic gonadotropin (Sigma), and artificially fertilized. Eggs were dejellied with 2% cysteine hydrochloride (pH 7.8) and embryos were staged according to Nieuwkoop and Faber(Nieuwkoop and Faber,1967).

Capped mRNAs were synthesized from linearized plasmids using SP6 RNA polymerase (Boehringer Mannheim) in the presence of 500 μM 5′-mGpppG-3′ cap analog, 500 μM each rUTP, rATP, rCTP and 50μM rGTP. Synthetic mRNA was purified using a Sephadex G-50 column(Pharmacia, Les Ulis, France) and recovered by ethanol precipitation. Microinjection of embryos was performed in 0.1× MBS (Modified Barth's Solution) containing 3% Ficoll 400 using a PLI-100 reproducible pico-injector(Medical Systems, Greenvale, NY). Unless specified, a volume of 10 nl corresponding to 250 pg of synthetic mRNA was injected into each blastomere at the animal pole region of two-cell-stage embryos.

Animal cap explants from control and injected embryos were dissected at mid-blastula stage (stage 8) and cultured to early gastrula stage (stage 10.5). Extraction of total RNA was as previously described(Umbhauer et al., 2000a). About 5 μg of total RNA were reverse-transcribed in the presence of 100 IU of MLV (Murine Leukemia Virus) reverse transcriptase (Life Technologies,Invitrogen, Cergy-Pontoise, France). In each experiment, 5 μg of total RNA from whole embryos treated the same way but without reverse transcriptase were used as a control for PCR specificity. One tenth of the reverse-transcribed cDNA was used for PCR amplification in a 25 μl reaction mixture consisting of 1× PCR buffer (Perkin Elmer Cetus, Life Science, Courtaboeuf,France), dNTP at 0.2 mM each, 25 pmol of sense and antisense primers and 0.3 IU of Taq DNA polymerase (Perkin Elmer Cetus). PCR primers were the following: Siamois (Lemaire et al.,1995) (5′-AAGGAACCCCACCAGGATAA-3′ and 5′-TACTGGTGGCTGGAGAAATA-3′; 30 cycles); Xbra as in Henry et al. (Henry et al., 1996),30 cycles; and ornithine decarboxylase (ODC)(Bassez et al., 1990)(5′-GTCAATGATGGAGTGTATGGATC-3′ and 5′-TCCATTCCGCTCTCCTGAGCAC-3′; 26 cycles). PCR products were resolved on a 2% agarose gel containing 1 μg/ml ethidium bromide.

Embryos at the early gastrula stage were extracted in lysis buffer [10 mM Tris-HCl, pH 7.5, 100 mM NaCl, 5 mM EDTA, 0.5% NP-40, 1% aprotinin (Sigma), 2 mM phenylmethylsulfonyl fluoride (PMSF) and 0.1% leupeptin (Sigma)]. After 10 minutes centrifugation at 13,200 g at 4°C, supernatants were diluted in 2× Laemmli buffer without β-mercaptoethanol for direct western blotting analysis. For these analyses, samples of protein extracts from the equivalent of 0.5 embryos were fractionated by electrophoresis on SDS-polyacrylamide gels according to Laemmli(Laemmli, 1970). For immunoprecipitation experiments, extracted proteins from an equivalent of ten embryos were separated in two equal parts and processed for immunoprecipitation using an anti-c-myc (9E10, Santa Cruz, Santa Cruz, CA), an anti-influenza-HA (12CA5, Boehringer Mannheim) or an anti-flag (M2, Sigma)monoclonal antibody. Immunoblotting analyses were performed after gel electrophoresis in reducing conditions and electrotransfer of separated proteins on nitrocellulose sheets (HYBON-C, Amersham, Les Ulis, France). The membranes were incubated in the presence of a 1/1000 dilution of monoclonal antibodies 9E10 or 12CA5 or in the presence of a 1/500 dilution of the M2 antibody. After incubation with horseradish peroxidase-conjugated secondary antibody (Immunotech Beckman Coulter Company, Maseille, London, 1/5000 dilution), bands were visualized using the ECL chemiluminescence kit(Amersham).

To analyze the biochemical properties of the Xfz3 receptor, we have constructed tagged-versions of the wild-type Xfz3 (Xfz3-HA, Xfz3-myc) and of a C-terminal truncated form, which retains five amino acids after the seventh transmembrane domain (Xfz3ΔC-myc)(Fig. 1)(Shi et al., 1998; Umbhauer et al., 2000a). These molecules were overexpressed in Xenopus embryos and analyzed either by western blot or immunoprecipitation. Embryos at the two-cell stage were injected with mRNA encoding the tagged receptors and cultured until the early gastrula stage when protein extracts were prepared. Under nonreducing conditions, direct western blot analysis of Xfz3-myc allows for the consistent detection of not only a band at approximately 90-95 kDa, which corresponds to the anticipated weight of the monomeric receptor, but also a band of a potentially dimeric species at 190 kDa(Fig. 2A). Immunoblotting of the truncated form Xfz3ΔC-myc reveals bands of approximately 65-75 kDa as expected for the monomer and an additional band at 130-150 kDa. These data are consistent with the idea that the form with higher molecular weight represents a homodimer. In some cases, discrete bands, which could represent trimeric or even higher order structures, were observed(Fig. 2A, star). Similar results were obtained using an HA rather than a myc epitope tag(Fig. 2B). We performed a competition experiment in which 200 pg of Xfz3ΔC-myc mRNA was co-injected with 800 pg of Xfz3-HA mRNA. Protein extracts were analyzed with the anti-myc antibody by western blotting. Expression of Xfz3-HA led to the disappearance of the 130-150 kDa immunoreactive species representing the dimeric form of Xfz3ΔC-myc. A heavier form, also revealed with the HA antibody (not shown) and which probably corresponds to a Xfz3ΔC-myc/Xfz3-HA heterodimer, appeared(Fig. 2C).

To confirm that the species of higher molecular weight observed in direct western blot corresponded to an Xfz3 homodimer, we devised a differential co-immunoprecipitation strategy. Xfz3-HA, Xfz3ΔC-myc and Xfz3-myc were expressed either separately or in combination in Xenopus embryos. The receptors were then immunoprecipitated with an anti-HA or an anti-myc antibody, subjected to SDS-PAGE and blotted with one or the other antibody. The anti-myc antibody specifically precipitated Xfz3ΔC-myc, whereas the anti-HA antibody did not (Fig. 3, lanes 1 and 2). Immunoprecipitation of protein extracts from embryos co-expressing Xfz3ΔC-myc and Xfz3-HA with the anti-HA antibody resulted in the co-precipitation of Xfz3ΔC-myc(Fig. 3, lane 3). Similar results were obtained for the wild-type Xfz3-HA and Xfz3-myc(Fig. 3, lanes 5 and 6). In both cases, two bands corresponding in size to the monomeric and the dimeric forms of the receptors were co-immunoprecipitated, suggesting that Xfz3 was also present as an oligomer. Overexpression of the same tagged proteins in COS cells yielded identical results (not shown). Taken together, these results show that Xfz3 protein is able to form homodimers when expressed in Xenopus embryos.

To determine whether disulfide bonds are important for Xfz3 dimer formation, treatment with 2% or 20% β-mercaptoethanol was performed before SDS-PAGE on protein extracts from embryos expressing Xfz3ΔC-myc. In parallel, to prevent artificial disulfide bonds, protein extracts were also treated with 10 mM iodoacetamide (IAA). As shown in Fig. 4, β-mercaptoethanol treatment led to a significant diminution of the dimer/monomer ratio. In the IAA-treated extracts, the dimeric form remained, excluding the possibility that the band with high molecular mass was an artefact due to disulfide bond exchange reactions during the preparation of samples. Thus, Xfz3 is able to form a SDS resistant dimer in Xenopus embryo and its dimerization relies, at least partly, on disulfide linkages.

The N-terminal extracellular region of frizzled receptors contains a conserved CRD interacting with Wnt ligand. We have tested whether this region alone is capable of dimerization when expressed in Xenopus embryos. Embryos were injected with mRNA encoding the first 196 N-terminal amino acids of Xfz3 (corresponding to the N-terminal extracellular region including the CRD) bearing either a flag (extra3-flag) or a myc (extra3-myc) tag. Co-immunoprecipitation experiments were performed using an anti-myc or an anti-flag antibody. In protein extracts derived from embryos expressing both constructs, anti-flag co-immunoprecipitated the extra3-myc protein(Fig. 5A, lane 5), revealing the presence of extra3-myc/extra3-flag dimers. In the embryo, the N-terminal extracellular region of Xfz3 has therefore an intrinsic capacity to form a homodimer. We asked next whether this region was sufficient to induce dimerization when added to a membrane-bound heterologous context. Two chimeric constructs between Xfz3 and the tyrosine kinase FGFR-1 receptor were devised. For one construct (extra3-R1), the extracellular domain of FGFR-1, which contains the FGF ligand binding site, has been replaced by the N-terminal extracellular region of Xfz3 (including the CRD)(Fig. 1). The other construct,Xfz3ΔC-R1, was made by replacing the C-terminal cytoplasmic region of Xfz3 by the intracellular portion of the FGFR-1 receptor. The experimental concept is based on the fact that tyrosine kinase receptors are known to transduce intracellular signaling via dimerization on ligand binding. If dimerization of the chimeric receptors occurs, the Xfz3 part of the chimeric protein should be responsible for this dimerization, as no FGF ligand-binding site is present in these receptors.

The chimeric receptors extra3-R1 and Xfz3ΔC-R1 were overexpressed in Xenopus blastula animal cap cells and transcriptional activation of Xbra was tested by RT-PCR. Xbra is a target gene of the transduction pathway activated by the FGFR-1 in response to FGF ligand in ectoderm animal cap cells. Activation of its transcription will indicate the presence of receptor dimers. As a positive control, we overexpressed a constitutive form of FGFR-1 receptor, torso-R1(Umbhauer et al., 2000b) and the wild-type FGFR-1 as a negative control. The same quantity of mRNA (50 pg)was injected for each construct. As shown in Fig. 5B, extra3-R1 and Xfz3ΔC-R1 overexpression leads to the transcriptional activation of Xbra, as does the torso-R1 positive control. This ability is correlated with phosphorylation on tyrosine residues of these hybrid proteins(Fig. 5C). In the same experiment, Xbra was not expressed in response to overexpressing the wild-type FGFR-1 (Fig. 5C). These experiments show that the N-terminal extracellular portion of Xfz3 is sufficient to induce dimerization either in a frizzled receptor or in an unrelated context.

In previous work, we have shown that overexpression of Xfz3 in Xenopus blastula animal cap cells is sufficient to activate transcription of the Wnt/β-catenin target gene siamois. Overexpression of the receptor Xfz7, however, does not lead to siamois expression (Umbhauer et al., 2000a). Using the differential ability of these two receptors to activate the Wnt/β-catenin pathway after overexpression in Xenopus embryo, we have addressed the functional significance of frizzled dimerization.

We first analyzed the ability of the receptor Xfz7 to form dimers when overexpressed in Xenopus embryos. Western blot analysis of protein extracts derived from Xfz7-myc-injected embryos revealed only one group of three bands corresponding to the expected molecular weight of the monomeric Xfz7-myc (Fig. 6A). In most cases, no other detectable signal was observed, although in some cases (two of ten), a faint band corresponding to the size of a putative Xfz7-myc dimer was visible. Co-immunoprecipitation experiments confirmed these western blot analyses. The anti-flag antibody did immunoprecipitate Xfz7ΔC-flag(Fig. 6B, lane 4) but it did not co-precipitate Xfz7-myc in protein extracts derived from embryos injected with Xfz7-myc and Xfz7ΔC-flag, (Fig. 6B, lane 3). These results show that, unlike Xfz3, Xfz7 protein is almost solely present in a monomeric form when expressed in the embryo.

Activation of the Wnt/β-catenin pathway in blastula animal cap cells in response to Xfz3 but not Xfz7 may be due to the endogenous expression of a specific ligand for Xfz3. If this is true, co-expression of the N-terminal extracellular region of Xfz3 should block activation of the Wnt/β-catenin pathway in response to Xfz3 by sequestering the ligand in the extracellular compartment. Embryos were injected at the two-cell stage with Xfz7-myc RNA,Xfz3-HA RNA or a combination of Xfz3-HA and extra3-myc RNA. Animal caps were dissected at the blastula stage, cultured to early gastrula and analysed by RT-PCR for siamois expression(Fig. 7A). As expected, Xfz3-HA but not Xfz7-myc activated siamois expression. Notably, the level of siamois expression in caps co-expressing at the same time Xfz3-HA and extra3-myc was very similar to the level obtained in response to Xfz3-HA alone(Fig. 7A). Moreover,co-injection of extra3-myc did not inhibit the formation of Xfz3 dimers, as shown by immunoprecipitation and western blot analysis(Fig. 7B). These results suggest that Xfz3 dimerizes and activates the Wnt/β-catenin pathway in a ligand-independent manner, at least in ectoderm animal cap cells.

These results described above show a correlation between the presence of dimers and the activation of the Wnt/β-catenin pathway. To determine whether activation of the Wnt/β-catenin pathway was indeed directly related to frizzled receptor dimerization, we designed a construct to artificially force the dimerization of the Xfz7 receptor. We interchanged the CRD of Xfz7 with an unrelated oligomerization domain (Xfz7-BD), the B domain of ETS transcription factor TEL (translocated Ets leukemia)(Lopez et al., 1999). Western blot and immunodetection analyses revealed that this protein indeed migrates as monomeric and dimeric forms, and even higher molecular forms(Fig. 8A), and is correctly expressed at the plasma membrane (Fig. 8B). Expression of Xfz7-BD in animal cap cells induces siamois expression (Fig. 8C) at a similar level to that induced by Xfz3. Induced dimerization of Xfz7 is therefore sufficient to activate the Wnt/β-catenin pathway.

In this study we have shown that Xfz3, but not Xfz7, is observed on western blot as dimers and/or oligomers when overexpressed in Xenopusembryos. Co-immunoprecipitation experiments using differentially epitope-tagged receptors further show the existence of Xfz3 dimers. Although the structural requirements for GPCR dimerization are as yet unknown, several dimerization interfaces have been proposed. These are the extracellular N-terminal domain for the glutamate and calcium-sensing receptors(Bai et al., 1998; Romano et al., 1996; Romano et al., 2001), the intracellular third loop and the sixth transmembrane domain for theβ-adrenergic and dopamine receptors(George et al., 1998; Hebert et al., 1996), and the C-tail for the GABA-B and δ-opioid receptors(Cvejic and Devi, 1997; Jones et al., 1998; Kaupmann et al., 1998; Kuner et al., 1999; White et al., 1998). Given the ability of the tagged truncated mutant Xfz3ΔC protein to form homodimers and oligomers, the C-tail is clearly not required in the case of Xfz3. Rather,our results suggest that the N-terminal extracellular domain of Xfz3 may play a role in receptor dimerization. In co-immunoprecipitation experiments, two truncated soluble forms of Xfz3 (extra3-flag and extra3-myc) bearing different epitope tags interacted. Using the chimeric protein extra3-R1, composed of the extracelluar domain of Xfz3 fused to the transmembrane and cytoplasmic domains of FGFR-1, we showed that a membrane-bound form of the extracellular N-terminal part of Xfz3 is capable of forming dimeric complexes. Our results are consistent with the fact that crystal structural analyses of CRD, both from secreted frizzled-related protein 3 (sFRP-3) and from mouse frizzled 8(mFz8), have revealed that CRDs exhibit a putative conserved dimer interface(Dann et al., 2001).

The sensitivity of the dimeric Xfz3 receptor to reducing agents indicates that disulfide bonds are important for Xfz3 dimer formation and/or stabilization. The β-mercaptoethanol treatment can either disrupt intramolecular disulfide bonds and consequently affect conformation of the receptor in a manner that impedes dimerization, or it can disrupt intermolecular disulfide bonds directly responsible for dimer formation. If intermolecular disulfide linkages are implicated in Xfz3 dimerization, none of the ten cysteines located in the CRD is likely to be involved in these interactions as these conserved cysteines have been shown to be engaged in pairwise intramolecular bonds in the secreted frizzled-related proteins sFRP-1 and sFRP-3, as well as in the frizzled module recently determined in rat Ror1 receptor tyrosine kinase (Chong et al.,2002; Roszmusz et al.,2001). However, cysteines outside the CRD could be involved in intermolecular disulfide bonds implicated in Xfz3 dimerization. Several cysteines located in the first and second extracellular loops are conserved among the frizzled proteins and, interestingly, two conserved cysteines,located on these same loops in the m3 muscarinic receptors have been identified as key residues for covalent dimer formation by site-directed mutagenesis (Zeng and Wess,1999).

Although dimerization has been shown for several GPCRs, the functional roles of such a process are currently unclear. The GABA-B receptors associate in the endoplasmic reticulum as heterodimers and are targeted to the plasma membrane as preformed dimers, independent of agonist regulation(Jones et al., 1998; White et al., 1998). Dimerization of these receptors provides a mechanism to control the efficient delivery of active GPCRs to the cell surface. The β-adrenergic receptor,however, undergoes ligand-dependent dimerization and activation, suggesting that dimerization favors receptor/G-protein coupling efficiency(Angers et al., 2000; Hebert and Bouvier, 1998; Hebert et al., 1996). In the case of frizzled receptors, our results strongly suggest that dimerization plays a role in transducing the signal through the Wnt/β-catenin pathway. When overexpressed in blastula animal cap cells, Xfz3 exists as a dimer and leads to activation of the Wnt/β-catenin pathway, whereas Xfz7 remains monomeric and does not activate the Wnt/β-catenin pathway. Moreover, the addition of a heterologous dimerization domain to Xfz7 is sufficient to activate the Wnt/β-catenin pathway in the animal cap assay. Activation of this pathway can also be obtained when Xfz7 is co-expressed with an appropriate Wnt ligand, suggesting that the lack of activity of Xfz7 in Wnt/β-catenin signaling in the animal cap cells may be due to the absence of a specific ligand (Umbhauer et al.,2000a).

Although it is tempting to speculate that frizzled dimerization is induced by the Wnt ligand, we have failed to show any reproducible effect of Wnt expression on the frizzled dimer/monomer ratio using western blot or co-immunoprecipitation studies (data not shown). In blastula animal cap cells,the N-terminal extracellular domain of Xfz3 was not sufficient to inhibit Xfz3 dimerization and activation of the Wnt/β-catenin pathway in response to the full-length receptor Xfz3. These results suggest that activation of the Wnt/β-catenin pathway and formation of Xfz3 dimers are both ligand independent, but they do not exclude a different mechanism for the receptor Xfz7. The question of whether ligands affect GPCR oligomerization is currently largely debated. Agonist treatment of several GPCRs has variably been reported to increase, decrease or have no effect on the oligomeric complexes. This may reflect the specific behavior of each receptor or it may be a consequence of using different experimental approaches. For example, immunoprecipitation experiments have suggested that agonists favor monomer formation of theδ-opioid receptor (Cvejic and Devi,1997), whereas the use of both bioluminescence resonance energy transfer (BRET) and time-resolved fluorescence resonance energy transfer(FRET) approaches does not reveal any effect of agonist or antagonist on this receptor at the cell surface (McVey et al., 2001).

In addition to the canonical Wnt/β-catenin pathway, Wnt and frizzled have been shown to transduce signals through at least two other pathways(Kuhl et al., 2000b; Mlodzik, 1999). Wnt5A and rat frizzled 2 trigger intracellular calcium release(Slusarski et al., 1997) and activate protein kinase C and calmodulin kinase II (CamKII)(Kuhl et al., 2000a; Sheldahl et al., 1999) in a heterotrimeric G-protein-dependent manner. In Drosophila, the planar cell polarity pathway (PCP) identified downstream of DFz1 involves the cytoplasmic protein Dsh, the small GTPase RhoA and a Jun N-terminal kinase cascade. A similar PCP pathway might be implicated in the control of gastrulation movements in Xenopus(Wallingford et al., 2000; Yamanaka et al., 2002). Dorsal overexpression of Xfz7 or Xwnt11 perturbs convergence extension movements,which can be rescued by co-injection of a dominant-negative form of the small GTPase (Djiane et al., 2000; Medina et al., 2000; Tada and Smith, 2000). Our results suggest that dimerization of frizzled receptors might not be required for the activation of the PCP pathway, as overexpression of Xfz7 as a monomer is sufficient to alter gastrulation movements. Because several frizzleds seem to be bifunctional receptors capable of transducing signals via two biochemically distinct pathways, dimerization could enable the discrimination between the alternate pathways. Therefore, dimerization of frizzled is emerging as a possible mechanism for transduction specificity.

We wish to thank I. Buisson for excellent technical assistance, L. Neves and S. Mine for assistance at different phases of this work, R. Schwartzmann for confocal microscopy, J. Ghysdael for the TELmod plasmid and E. Amaya for the FGFR-1 plasmid. We also thank P. Denoulet and J. F. Riou for critical reading of the manuscript. This work was supported by grants from the Centre National de la Recherche Scientifique (CNRS), the Association Française contre les Myopathies (AFM) and the Association pour la Recherche contre le Cancer (ARC).