The C. elegans ROR receptor tyrosine kinase, CAM-1,non-autonomously inhibits the Wnt pathway

Wnt signaling is necessary for development, but causes cancer when dysregulated. The canonical Wnt pathway is initiated when a secreted Wnt glycoprotein binds to a transmembrane Frizzled (Fz) receptor and ultimately leads to β-catenin-mediated regulation of gene transcription(Logan and Nusse, 2004). The Wnt pathway is actively constrained by secreted antagonists and inhibitors of signal transducers (Kawano and Kypta,2003; Logan and Nusse,2004). The importance of negative regulators, both developmentally and to prevent tumorigenesis, prompted us to investigate the mechanistic activity of ROR proteins, a poorly understood class of Wnt inhibitors(Billiard et al., 2005; Forrester et al., 2004; Mikels and Nusse, 2006a; Mikels and Nusse, 2006b).

ROR proteins are conserved receptor tyrosine kinases (RTKs) characterized by an extracellular Fz domain [also called cysteine-rich-domain (CRD)], an immunoglobulin (Ig) domain, and a kringle domain(Fig. 1A). Mutations in ROR genes cause developmental defects including skeletal abnormalities in mice and humans (reviewed by Forrester,2002). Studies of vertebrate RORs showed that the ROR CRD, like the Fz CRD (Bhanot et al.,1996), can bind to Wnts(Billiard et al., 2005; Hikasa et al., 2002; Kani et al., 2004; Mikels and Nusse, 2006a; Oishi et al., 2003). In cell culture, ROR2 abrogates expression of a canonical Wnt reporter(Billiard et al., 2005; Mikels and Nusse, 2006a);however, whether this antagonistic activity is cell-autonomous is unknown. To study how RORs modulate Wnt signaling in a multicellular environment, we investigated the function of the sole C. elegans ROR family member, cam-1.

Forrester et al. (Forrester,2002; Forrester et al.,2004) studied CAM-1, which is equally similar to ROR1 and to ROR2,for its role in cell migration, where the CRD is required to antagonize EGL-20/WNT activity. During canal-associated neuron (CAN) migration, this CAM-1 function is cell-autonomous(Forrester et al., 1999). Although Forrester and others postulated that CAM-1 sequesters Wnts, reports that ROR2 can bind to Fz receptors (Oishi et al., 2003) raise the question of whether CAM-1/ROR inhibits Wnt signaling by interacting with the receptor or the ligand. We addressed these questions using vulva development as a model, as this process involves every C. elegans Wnt (lin-44, cwn-1, egl-20, cwn-2 and mom-2) and Wnt receptor (mig-1, lin-17, mom-5, cfz-2 and lin-18) (Gleason et al.,2006), and also because the well-characterized cellular phenotypes facilitate identification of signaling defects.

The C. elegans vulva comprises 22 cells generated by well-defined signaling events (reviewed by Sternberg,2005) (Fig. 1B). The vulval cells are descendents of three vulval precursor cells (VPCs)located on the ventral surface of the worm(Sulston and Horvitz, 1977). During larval development, the VPCs are induced to divide by LIN-3 (EGF)secreted by the anchor cell (AC), (Hill and Sternberg, 1992). The VPC most proximal to the AC, P6.p,receives the most LIN-3 inductive signal through the receptor LET-23 (EGFR)(Katz et al., 1995; Yoo et al., 2004), triggering a MAP kinase cascade that induces P6.p to adopt the primary fate (1°) and produce eight vulval progeny. P5.p and P7.p receive lower levels of LIN-3 and a repressive lateral signal from P6.p mediated by LIN-12 (NOTCH)(Simske and Kim, 1995; Sternberg and Horvitz, 1989). These cells adopt the secondary fate (2°) and each produces seven vulval progeny. The remaining VPCs receive sub-threshold LIN-3 signal and adopt either the tertiary fate (3°), dividing once before fusing (P4.p, P8.p and sometimes P3.p), or the fused fate (F), fusing with the epidermis without dividing (P3.p adopts this fate half the time)(Sulston and Horvitz, 1977). A Wnt pathway involving BAR-1 (β-catenin) is required for the VPCs to be induced by LIN-3 and defective Wnt signaling frequently causes P5.p-P7.p to become 3° or F, instead of 1° or 2°, and also causes P3.p, P4.p and P8.p to become F instead of 3°(Eisenmann et al., 1998).

Because wild-type C. elegans development is essentially invariant,even slight deviations from the wild-type induction pattern can be detected and are informative. Worms producing fewer than 22 vulval cells are called`underinduced' (UI) and worms producing greater than 22 vulval cells are called `overinduced' (OI). The UI phenotype(Fig. 1D) is caused by reduced Wnt signaling or reduced Ras/MAPK signaling. The OI phenotype(Fig. 1E) is caused by increased Ras/MAPK signaling (Ferguson et al., 1987), increased lateral signaling(Greenwald et al., 1983) or increased Wnt activity (Gleason et al.,2002; Korswagen et al.,2002).

Here, we show that CAM-1 inhibits Wnt pathway activity during vulval development by limiting the levels of Wnts that interact with the VPCs. We find that expression of the CAM-1 ECD in non-vulval tissue is sufficient to limit Wnt pathway activity in the VPCs, whereas CAM-1 expression in the VPCs failed to rescue the cam-1 mutant phenotype, suggesting a non-autonomous mode of inhibition. We also find that the CAM-1 ECD specifically binds to Wnts, supporting the model that CAM-1 sequesters Wnt ligands. Our results demonstrate how CAM-1/ROR contributes to the complex spatial profile of Wnt signaling by modifying the range of Wnt activity.

C. elegans was handled as described previously(Brenner, 1974). All strains used are derivatives of C. elegans N2 Bristol strain. LG1: lin-17(n671), lin-17(n677), lin-44(n1792), mom-5(or57), mom-5(zu193).LGII: cwn-1(ok546), cam-1(gm122), cam-1(ks52), cam-1(gm105), cam-1(sa692),cam-1(ak37), rol-6(e187). LGIII: lin-12(n952), unc-119(ed4). LGIV: lin-3(e1417), lin-3(n378), ark-1(sy247), dpy-20(e1282),egl-20(n585), egl-20(hu120), cwn-2(ok895). LGV: him-5(e1490). LGX:lin-18(e620), bar-1(ga80), gap-1(n1691), unc-2(e55), sli-1(sy143),daf-3(mgDf90). For RNAi experiments, gravid hermaphrodites were fed RNAi-expressing bacteria and L4 progeny were scored.

Vulval induction was scored in mid-L4 stage hermaphrodites by counting vulval cell nuclei using Nomarski DIC optics. If both VPC daughters divided,that VPC was counted as induced (1.0). If only one VPC daughter divided, that VPC was counted as half-induced (0.5). Pmyo-3::CAM-1::GFP displayed increased penetrance of the UI phenotype at 25°C. Thereafter, all CAM-1::GFP transgenic worms (except cwEx164) were grown at 25°C. All other strains were grown at 20°C.

Our results are inconsistent with the positive role for LIN-17 in vulva induction reported by Gleason et al.(Gleason et al., 2006). Whereas Gleason et al. report that 12% of lin-17(n671) worms are UI,we did not observe any UI lin-17(n671) worms. To address this discrepancy we obtained lin-17(n671) worms used by Gleason et al. from the Eisenmann laboratory (-DE) and did not detect any UI worms(see Table S2 in the supplementary material). By contrast, we observed one lin-17(n671)-DE worm that was OI and had five VPCs induced (see Fig. S1 in the supplementary material). Our examination of mig-1(e1787);lin-17(n671) and lin-17(n671); cfz-2(ok1201) double mutants did not reveal a UI phenotype. Also, lin-17(lf) did not enhance the UI phenotype of cwn-1(lf) mutant worms. lin-17(n671)-DE;cam-1(lf) double-mutant worms recapitulated the synthetic OI phenotype, as did double mutants containing another lin-17 allele, n677. The elevated Wnt signaling observed in the lin-17(lf);cam-1(lf) background, which cannot be explained by signaling through LIN-17, is likely to be due to increased signaling through another Frizzled receptor, such as MOM-5. Thus, we examined vulval induction in mom-5mutants (see Table S2 in the supplementary material). In contrast to lin-17, we found that mutation of mom-5 caused a dramatic UI phenotype, suggesting that mom-5, but not lin-17, is required for vulval induction.

Extrachromosomal arrays were generated by co-injecting CAM-1b::GFPdriven by various promoters with unc-119 (+) (60 ng/μL) into unc-119(ed4) hermaphrodites as described(Mello et al., 1991). Of the three cam-1 splice variants, the `b' isoform was selected because it appears to have a weak signal sequence, whereas the `a' and `c' variants have no detectable signal sequence. cam-1 tissue-specific constructs were made by shuttling various promoters upstream of CAM-1b::GFP using 5′ BamHI and 3′ NotI sites. All constructs were injected at 50 ng/μL except syEx863, syEx864, and syIs198, which were injected at 75 ng/μL. To facilitate examination of Pcam-1::CAM-1::GFP and Pcam-1::CAM-1ΔIgKriIntra::GFP, dpy-20(e1282) was crossed into strains WF1863 and WF1729, respectively(Forrester et al., 1999; Kim and Forrester, 2003) to suppress the roller phenotype. syIs75(Plin-18::LIN-18::GFP) is an integrated line of syEx363[pTI00.43(60ng/μl) + unc-119(+)(30ng/μl)] (Inoue et al., 2004). syEx1022[LIN-17::GFP(40ng/μl) +unc-119(+)(90ng/μl) + myo-2::DsRed(15ng/μl)] was made with plasmid PSH22 (gift from H. Sawa, RIKEN, Kobe, Japan). syEx1020[Pmyo-3::LIN-17::GFP(50ng/μl) +unc-119(+)(90ng/μl) + Pmyo-2::DsRed(15ng/μl)]contains a Pmyo-3::LIN-17::GFP plasmid that was made by amplifying the N-terminal-encoding portion of lin-17 from PSH22 (forward primer,TCCATCTAGAGGCTCCTTCTCCAAAATGATGCATTCTTTGGGC; reverse primer,GCACAATGCGACTTGGGATCGTGTGG). The lin-17 C-terminal-encoding portion was amplified from cDNA (forward primer, CCAAGCCAACCGGGTGCCCCAG; reverse primer, TCTTCCGGAACG ACCTTACTGGGTCTCCATGAATTCTG). The C-terminal-encoding portion was cleaved by BamHI and BspEI and transferred into Fire vector L4817 (Pmyo-3) that had been cleaved by AgeI and BamHI. The N-terminal-encoding portion was then cleaved by XbaI (cuts twice) and BamHI. The XbaI-BamHI fragment was transferred in first, followed by the XbaI-XbaI fragment.

To make the CAM-1b::GFP backbone, C01G6.8a cDNA was first inserted into Fire vector pPD49.83 using the NheI site. To create hs::CAM-1::GFP, BspEI and ApaI sites were used to switch the 3′ end of cam-1 with the 3′ end of CAM-1::GFPfrom plasmid pMini3 (gift from W. Forrester, Indiana University, Bloomington,IN) which also includes the last two small introns of cam-1. Next,the 5′ end of C01G6.8b was amplified from cDNA using forward primer ATAAGATGCGGCCGCATGGAGGGTACATCAACTGGTCAACG to add a NotI site to the 5′ end (reverse primer TTC CAATGCATTGGCATCTAGCCATCGTTCTGATACAGC). The C01G6.8b 5′ end was then cloned into pBluescript using NotI and BstXI and transferred into hs::CAM-1::GFP using BamHI and BstEII, creating CAM-1b::GFP with a NotI site 5′ of the ATG.

syEx778, syEx781 and syEx814 contain 2.4 kb of Pmyo-3 (myo-3 5′ regulatory region) amplified from Fire vector L4817 with forward primer CGCGGATCCGGTCGGCTATAATAAGTTCTTGAATA and primer ATAGTTTAGCGGCCTCTAGATGGATCTAGTGGTCGTG. syEx798 and syEx799 contain 3.4 kb of Pdpy-8 amplified from genomic DNA using forward primer CGCGGATCCGAACTGAG AATGCTGACGGATG and reverse primer ATAGTTTAGCGGCCGATGGGAAAATAAGAAAAGGAAATGTGG. syEx863 and syEx864 contain 5.5 kb of Psur-2 amplified from cosmid F39B2 using forward primer CGCGGATCCCGAAATTCGGTAGATTTGGGC and reverse primer ATAG TTTAGCGGCCGCTTGTTGCCTGAAAATGTAATAATTTTC. syEx780 and syEx777contain 4.9 kb of Pfos-1a amplified from plasmid pDRS46(Sherwood and Sternberg, 2003)using forward primer CGCGGATCCTGGGCAGCTGTAAAACGTCTTTAC and reverse primer ATAGTTTAGCGGCCTCCACTCTCTTATATAGCAGAGGTG. syEx775 and syEx776contain 3 kb of Psnb-1 amplified from plasmid Psnb-1::slo-1(Davies et al., 2003) using forward primer CGCGGATCCAAGCTTTTTGCTGAAATCTAGGATTAC and reverse primer ATAGTTTAGCGGCCGCTGTTCCCTGAAATGAAGCGA. syIs198 contains 1.6 kb of Plst-1 amplified from plasmid lst-1p-gfp-lacZ (gift from Iva Greenwald c/o Andrew Yoo, Howard Hughes Medical Institute, Columbia University, NY) using forward primer CGCGGATCCCAATTGTTACTACTGACGGCATTCC and reverse primer ATAGTTTAGCGGCCGCGTCAAATAATTCTTTTGAAATGAGAAAGAACTTGGC. To make Pmyo-3::CAM-1ΔIntra::GFP, blunt HpaI and MscI sites were used to switch the C-terminus-encoding part of Pmyo-3::CAM-1b::GFP with a HpaI-HpaI fragment (10.8 kb) from pDM108 (Francis et al.,2005) that contains cam-1 minus the sequence encoding the kinase domain (removal of C-terminal 346 codons), fused to GFP.

Lysates of transfected and untransfected Drosophila S2 cells were run on a 4-12% NuPAGE Bis-Tris gel (Invitrogen) and probed with anti-HA monoclonal antibody G036 (Applied Biological Materials, Vancouver, BC) or polyclonal anti-GAPDH (Sullivan et al.,2003).

The CRD-AP fusion proteins were made in 293T cells as previously described for Drosophila CRD-AP fusions (Wu and Nusse, 2002). The CRD of the sFRP3-AP fusion was replaced with the CRD (or WIF) of C. elegans receptors. Each construct contains sFRP3 signal sequence, C. elegans CRD (or WIF), C-terminal domain of sFRP3 and AP. Sequences across the signal sequence fusion junction are(CRD/WIF underlined): CAM-1, PGAQAAGSNYAPVA; LIN-18,PGAQANVNMFISK; LIN-17, PGAQASIFDQAVKG; MOM-5,PGAQADQRLSSTSI; CFZ-2, PGAQALFGKRQKCE; MIG-1,PGAQAQRCQKVDHE. Downstream fusion junctions are (CRD/WIF domains underlined): CAM-1, STSNCIHALAIVTAD; LIN-18, TDSIDKTRALAIVTAD; LIN-17, PPELCMNALAIVTAD; MOM-5, VTDLCVDALAIVTAD; CFZ-2, TGNICADALAIVTAD; MIG-1, NREKMCMNALAIVTAD. To determine the concentration of CRD-AP fusion protein in the conditioned medium, we immunoprecipitated the CRD-AP fusion proteins with anti-AP antibody (Sigma A-2951), resolved the immunocomplexes by SDS-PAGE and estimated the protein concentration after staining with Coomassie Blue. Activities of the CRD-AP fusion proteins were assayed colorimetrically after incubation with the AP substrate. Each of the CRD-AP fusion proteins was determined to have similar specific activity of 3 pmol/unit activity. The protein was concentrated by ammonium sulfate precipitation (3.2 M) followed by dialysis against Hank's Balanced Salt Solution without calcium and magnesium(HBSS) and the samples were then normalized by AP activity. The Neurotactin(Nrt)-HA-Wnt fusion proteins were made as previously described for Drosophila Nrt-HA-Wnt fusions (Wu and Nusse, 2002) with the exception that we used the pCoBlast selection vector (Invitrogen) and 25 μg/mL blasticidin for selection. The sequences around the regions linking HA and the Wnts are (Wnt sequences underlined): Nrt-CWN-1, WEDEEASLAANRFD; Nrt-CWN-2,WEDEEASLNVQSLL; Nrt-EL-20, WEDEEASPSATYST and WEDEEASGHNVKP; Nrt-MOM-2, WEDEEASKSADAWW; Nrt-LIN-44,WEDEEASAPAGKIV. The binding assay protocol was adapted from those previously published (Cheng and Flanagan,1994; Flanagan and Leder,1990; Wu and Nusse,2002). We observed that Nrt-HA-Wnt expression appeared to decrease with time as cells were passaged. Because of this observation and the non-clonality of the stable lines, we performed the binding assays as soon as sufficient cell numbers had recovered from antibiotic selection and used equal cell numbers for the assay rather than normalizing to levels of Wnt expression. S2 cells stably transfected with the Nrt-HA-Wnt fusion constructs were counted with a hemacytometer and then heat shocked for 45 minutes at 37°C followed by 2 hours incubation at 25°C. At this point, aliquots of 500,000 cells were frozen for western analysis. The remaining cells were then resuspended in HBSS plus 10% BSA and incubated with CRD-AP(7×10^-8 M) in Eppendorf tubes for 90 minutes at 25°C. Three binding reactions of 30,000 cells each were performed for 26 of 30 combinations. For the remaining four combinations (MIG-1, MOM-5 and CFZ-2 CRDs with untransfected S2 cells, and LIN-18 CRD with Nrt-HA-LIN-44-expressing cells), only two reactions of 30,000 cells each were performed. After washing cells three times with HBSS, cells were lysed by adding HBSS plus 1% Triton X-100 with brief vortexing and then heated at 70°C for 10 minutes to kill background phosphatase activity. Supernatant was transferred to a 96-well untreated microtiter plate and incubated with the chromogenic substrate p-nitrophenyl phosphate (Sigma N-7653). After 24 hours the absorbance was read at 405 nm using a microtiter plate spectrophotometer (Bio-Rad) (for raw data see Table S3 in the supplementary material).

To study how CAM-1/ROR inhibits Wnt signaling, we investigated the role of CAM-1 in vulval development, a process requiring multiple Wnts. None of the five cam-1 alleles tested (Fig. 1A) caused induction defects(Table 1). However, as vulval development requires several redundant Wnts and receptors(Gleason et al., 2006), we looked for genetic interactions between cam-1 and Wnt receptors lin-17/Frizzled (Sawa et al.,1996) and lin-18/Ryk(Inoue et al., 2004). We found that worms doubly mutant for cam-1 and loss-of-function (lf)mutations in lin-17 or lin-18 displayed an OI phenotype(greater than 22 vulval cells). lin-17(lf) and lin-18(lf)mutants frequently display a polarity defect in the P7.p lineage that is distinct from vulval induction. This polarity defect alters the arrangement,but not the number, of vulval cells. To distinguish between induction defects and polarity defects, we counted the vulval nuclei. Both lin-17(lf)and lin-18(lf) animals had the wild-type pattern of 3-cell induction;P6.p adopted the 1° fate, P5.p and P7.p adopted the 2° fate, and the remaining VPCs were not induced. We found that 17% of lin-17(lf);cam-1(lf) double mutants and 12% of cam-1(lf); lin-18(lf) double mutants were OI (Table 1, Fig. 1E), with either 3.5, 4.0 or 4.5 VPCs induced. Since neither lin-17(lf) nor lin-18(lf)single mutants were OI, these results suggest that CAM-1 negatively regulates vulval induction and that lin-17(lf) and lin-18(lf) provide a sensitized background in which to observe CAM-1 function. To determine whether the OI phenotype is a common phenotype among cam-1; Fz double mutants, we constructed worms doubly mutant for cam-1(lf) and two other Fz receptors, mig-1 and cfz-2. 0/21 mig-1(lf);cam-1(lf) and 1/22 cam-1(lf); cfz-2(lf) double-mutant worms were OI, indicating that sensitization is specific to lin-17.

We analyzed the five available cam-1 alleles(Fig. 1A) in combination with lin-17(lf) and lin-18(lf)(Table 1). The cam-1alleles that caused an OI phenotype are either null (gm122)(Forrester et al., 1999),disrupt the CRD (sa692) (Ailion and Thomas, 2003; Kim and Forrester, 2003), or disrupt the insertion of the ECD into the membrane (ak37) (Francis et al.,2005). By contrast, an allele truncating most of the intracellular domain (gm105) (Forrester et al.,1999), and an allele eliminating the kinase domain (ks52)(Koga et al., 1999), did not cause increased vulval induction. Analysis of these alleles provides structure-function information about CAM-1. Since the sa692 allele eliminates a conserved cysteine in the CRD (Wnt-binding) domain, negative regulation of vulval development by CAM-1 requires membrane-insertion of the ECD containing the CRD, but does not require the intracellular domain. RNAi of cam-1 in lin-17(lf) worms recapitulated the OI phenotype and cam-1 RNAi of cam-1(lf); lin-17(lf) worms did not reduce the OI phenotype, confirming that the OI phenotype is due to reduced CAM-1 activity and not to a neomorphic function of mutant cam-1.

cam-1 expression has been reported in muscle and neurons(Forrester et al., 1999; Koga et al., 1999). We detected additional expression in the VPCs in a previously characterized Pcam-1::CAM-1::GFP strain, WF1863(Forrester et al., 1999) (see Fig. 3A). To test whether cam-1 acts in the VPCs, we tried to rescue the lin-17(lf);cam-1(lf) OI phenotype with an integrated VPC-specific CAM-1::GFP transgene driven by the lst-1 promoter(Yoo et al., 2004). Although Plst-1::CAM-1::GFP was expressed in the relevant VPCs (see Fig. 3G), it failed to rescue the OI phenotype suggesting that CAM-1 is required in other tissues to negatively regulate vulval induction.

To investigate the signaling involved in CAM-1 inhibition of vulval induction, we first tested whether the synthetic OI phenotype is dependent on the inductive LIN-3 signal. Removing the source of LIN-3 (the AC) by laser ablation of the gonadal primordium eliminates inductive Ras/MAPK signaling. In gonad-ablated wild-type worms, no VPCs are induced(Kimble, 1981; Sulston and White, 1980). Mutations that strongly activate Ras/MAPK signaling can rescue the UI phenotype caused by gonad ablation (Han and Sternberg, 1990). We ablated the gonad in wild-type and lin-17(lf); cam-1(lf) worms and found that vulval induction in lin-17(lf); cam-1(lf) worms was gonad-dependent: all 16 ablated animals had no VPCs induced. Because only strong activation of the Ras/MAPK pathway can rescue vulval induction in gonad-ablated worms, we next tested whether cam-1(lf) affects induction in worms with mildly reduced LIN-3 activity. cam-1(lf) suppressed the UI phenotype of two reduction-of-function (rf) lin-3 alleles (see Table S1 in the supplementary material), suggesting that cam-1 acts downstream of, or parallel to, lin-3. We then tested for a genetic interaction between cam-1 and inhibitors of Ras/MAPK signaling, ark-1, sli-1 and gap-1 (Sternberg,2005; Sundaram,2006), which are each silent when mutated singly, but are OI(30-90%) when combined with loss of another negative regulator(Hopper et al., 2000; Yoon et al., 2000). We found no interaction of cam-1(lf) with mutations in ark-1, sli-1or gap-1, indicating that CAM-1 is probably not a negative regulator of the Ras/MAPK pathway. lin-17(lf); gap-1(n1691) worms were not OI,thus providing further support that loss of CAM-1 does not cause increased Ras/MAPK signaling. Besides Ras/MAPK signaling, Wnt signaling is also required for vulval induction and can cause OI phenotypes when hyperactivated(Gleason et al., 2002). Mutations in bar-1/β-catenin cause a UI phenotype(Eisenmann, 2005; Eisenmann et al., 1998). In contrast to the suppression we observed upon reduced activity of the Ras/MAPK pathway, cam-1(lf) did not suppress the UI phenotype of bar-1(lf), consistent with cam-1 and bar-1functioning in the same pathway.

cam-1 mutants have a withered tail (Wit) phenotype that might position some VPCs closer to the AC and thus increase the local concentration of inductive LIN-3 signal. To investigate whether the OI phenotype is a consequence of increased VPC proximity to the AC, we tested the ability of cam-1 to affect vulval induction independently of the AC. To do this,we used a gain-of-function (gf) allele of lin-12/Notch. When heterozygous, the lin-12(n952gf) allele causes gonad-independent specification of 2° lineages in P3.p-P8.p. As lin-12(gf)/+ also causes loss of the AC, this phenotype is due to increased lateral signaling rather than increased Ras/MAPK signaling. We found that cam-1(lf)increased induction in lin-12(gf)/+ worms (see Table S2 in the supplementary material). Thus, the effect of cam-1(lf) on vulval induction cannot be attributed to mispositioning of the VPCs closer to the AC,which is absent in these worms.

Starvation and passage through dauer, an alternate third larval stage usually entered under conditions of starvation or high temperature(Savage-Dunn, 2005), can affect vulval induction (Ferguson and Horvitz, 1985) and cam-1 mutants are dauer constitutive(Daf-c) (Forrester et al.,1998; Koga et al.,1999). To test whether the OI phenotype we observe is due to passage through dauer, we constructed lin-17(lf); cam-1(lf);daf-3(lf) triple mutants. Although daf-3(lf) suppresses the Daf-c phenotype of cam-1(lf)(Koga et al., 1999), it did not suppress the OI phenotype of lin-17(lf); cam-1(lf) double mutants(see Table S1 in the supplementary material), indicating that the OI phenotype is not due to passage through dauer.

Previous studies of CAN migration demonstrated that CAM-1 inhibits EGL-20/WNT function (Forrester et al.,2004). To determine if this is also the role of CAM-1 in vulval induction, we tested whether a strong rf mutation in egl-20(Harris et al., 1996) could suppress the OI phenotype of lin-17(lf); cam-1(lf) or cam-1(lf);lin-18(lf) double mutants (Table 1). egl-20(rf) fully suppressed the OI phenotype of cam-1(lf); lin-18(lf) worms indicating that the OI phenotype of these worms depends on EGL-20. However, we found that lin-17(lf); cam-1(lf);egl-20(rf) triple mutants were still OI(Table 1), indicating that the OI phenotype of these worms is not dependent on EGL-20. The role of CAM-1 in vulval induction is thus only partly attributed to inhibition of EGL-20 activity.

Of the five Wnts, EGL-20, CWN-1 and CWN-2 strongly promote vulval induction(Gleason et al., 2006)(Table 1). To investigate whether cam-1(lf) causes increased CWN-1 or CWN-2 activity, we tested the ability of mutations in these Wnt genes to suppress the OI phenotype of lin-17(lf); cam-1(lf) or cam-1(lf); lin-18(lf) double mutants. We found that cwn-1(lf) suppressed the OI phenotype of cam-1(lf); lin-18(lf) mutant worms and that cwn-2(lf) weakly suppressed the OI phenotype of lin-17(lf); cam-1(lf) mutant worms. These results indicate that cam-1(lf) increases the activity of CWN-1, EGL-20 and possibly CWN-2 (Fig. 1F). The inability of cwn-1(lf), egl-20(rf), or cwn-2(lf) to fully suppress the OI phenotype of lin-17(lf);cam-1(lf) worms suggests that the OI phenotype in this strain is caused either by one of the remaining Wnts or by multiple Wnts. In some cases,mutation of a Wnt reduced the level of induction in lin-17(lf);cam-1(lf) or cam-1(lf); lin-18(lf) double mutants to below that of wild type, consistent with the role of these Wnts in vulval induction.

LIN-17 and LIN-18 function as typical Wnt receptors in P7.p polarity. We speculate that in addition, loss of LIN-17 and LIN-18 increases levels of extracellular Wnt and that loss of CAM-1 further increases these levels,crossing the threshold to induce the VPCs(Fig. 1F). This hypothesis is consistent with observations in the Drosophila wing where clones mutant for Frizzleds fz and fz2 have increased extracellular levels of Wingless (Wnt) (Han et al.,2005). This increase might be caused by reduced endocytosis of ligand-bound receptor. We thus tested whether worms lacking both lin-17 and lin-18 display an OI phenotype. Of 51 lin-17(lf); lin-18(lf) double-mutant worms observed, only one displayed an OI phenotype (see Table S2 and Fig. S1 in the supplementary material). However, it is possible that the class of Wnts elevated by removal of CAM-1 complements those elevated by removal of LIN-17 and LIN-18, but the Wnts elevated by removal of LIN-17 and LIN-18 do not complement each other. Another possibility is that removal of lin-17 or lin-18 only mildly increases extracellular Wnt levels and that these levels do not cross the threshold unless cam-1, a more important regulator of Wnt levels,is also removed. We next tested whether overexpression of LIN-17 and LIN-18 might reduce extracellular Wnt levels and cause a UI phenotype. Plin-18::LIN-18::GFP (Inoue et al., 2004) caused a weak UI phenotype (see Table S2 in the supplementary material) and significantly increased the fraction of cwn-1(lf) worms with a more severe UI phenotype (<2 VPCs induced),consistent with the hypothesis that lin-18 expression affects extracellular Wnt levels. Although transgenes can sometimes decrease gene expression by titrating out transcriptional activators(Gill and Ptashne, 1988), it is unlikely that the phenotype we see here is caused by reduced lin-18 expression because Plin-18::LIN-18::GFP is an overexpression construct (not a promoter::GFP array) and rescues the lin-18(lf) phenotype (Inoue et al., 2004). However, we cannot rule out the possibility that the phenotype is due to promoter effects on a different gene. In contrast to Plin-18::LIN-18::GFP, Plin-17::LIN-17::GFP did not affect vulval induction; however, Plin-17::LIN-17::GFP caused a mild, though not statistically significant, increase in the fraction of UI cwn-1(lf)worms. Again, it is unlikely that this phenotype is caused by promoter effects on lin-17 expression because Plin-17::LIN-17::GFP rescued the P7.p polarity defect of lin-17(lf) worms (data not shown). Also,loss of lin-17 did not increase the fraction of UI cwn-1(lf)worms (see Materials and methods and Table S2 in the supplementary material). As with lin-18 overexpression, we cannot rule out the possibility that the phenotype is due to promoter effects on a different gene. Because Plin-17::LIN-17::GFP displays a more restricted expression pattern than Plin-18::LIN-18::GFP, we expressed LIN-17 in body wall muscle using the myo-3 promoter (Okkema et al., 1993). Pmyo-3::LIN-17::GFP did not significantly affect vulval induction, nor did it enhance the UI phenotype of cwn-1(lf) worms. Although the mechanism by which lin-17 and lin-18 mutations provide a sensitized background for cam-1effects on vulval induction is unclear, the role of cam-1 as an inhibitor of vulval induction is confirmed by other experiments not dependent on lin-17 or lin-18 mutants (e.g. cam-1(lf); lin-3(rf),cam-1(lf); lin-12(gf/+) see above).

Our data suggest that non-vulval CAM-1 normally antagonizes Wnt signaling by a mechanism dependent on the CAM-1 ECD, possibly by directly binding to and impeding Wnts. Detecting association of the CAM-1 ECD with Wnts by co-immunoprecipitation experiments was impractical owing to the characteristic insolubility of Wnt proteins and the lack of available recombinant C. elegans Wnts. To circumvent these obstacles, we employed a reverse binding assay (Rulifson et al.,2000; Wu and Nusse,2002) in which C. elegans Wnts are expressed in stably transfected insect cells and tethered to the membrane by N-terminal fusion to Neurotactin (Nrt) (Fig. 2A). Binding is determined by measuring the alkaline phosphatase (AP) activity retained by the cells after incubation with secreted CAM-1 CRD-AP fusion proteins. As an internal control we assayed all combinations of C. elegans Wnts and Wnt receptors. This set included five Wnts (LIN-44,CWN-1, EGL-20, CWN-2, MOM-2), four Fz receptors (MIG-1, LIN-17, MOM-5, CFZ-2),and two RTKs (CAM-1/ROR, LIN-18/RYK) and confirmed that no Wnt bound indiscriminately to all receptors (see Table S3 in the supplementary material). Consistent with our genetic data, we found that the CAM-1 CRD bound to CWN-1 and EGL-20 to a significantly greater extent than to control cells(Fig. 2B). The CAM-1 CRD also bound significantly to cells expressing Nrt-MOM-2.

If CAM-1 negatively regulates Wnt signaling by binding to and impeding Wnts, then overexpression of CAM-1 in non-vulval tissue might cause a UI phenotype. To test this, we made full-length CAM-1::GFP translational fusions driven by the tissue-specific promoters Psnb-1 (pan-neuronal)(Nonet et al., 1998), Pmyo-3 (muscle) (Okkema et al.,1993), Pdpy-8 (epidermis), Plin-31(VPCs)(Tan et al., 1998), Psur-2 (VPCs) (Singh and Han,1995), Plst-1 (VPCs)(Yoo et al., 2004) and Pfos-1a (somatic gonad) (Sherwood and Sternberg, 2003) (Fig. 3). We observed membrane-localized GFP in the expected tissues for all lines except Plin-31, in which we were unable to detect fluorescence. We found that expression of CAM-1::GFP in body wall muscle (myo-3 promoter) and in neurons (snb-1 promoter)caused a UI phenotype (Table 2)similar to loss of bar-1/β-catenin and Wnt genes:specifically, P3.p adopted the F fate at an increased frequency, P4.p was often F instead of 3°, and P5.p occasionally adopted the F or 3° fate instead of the normal 2° fate.

Also similar to the consequences of mutations in Wnt pathway components,both Pmyo-3::CAM-1::GFP and Psnb-1::CAM-1::GFP had a greater effect on the anterior VPCs than on the posterior VPCs. To test whether this activity of cam-1 requires the intracellular domain, we expressed a version of CAM-1::GFP lacking the intracellular domain(CAM-1ΔIntra::GFP) in muscle. Pmyo-3::CAM-1ΔIntra::GFP caused a UI phenotype,indicating that the intracellular domain is not required. This observation is consistent with our analysis of cam-1 mutant alleles. Although expressed at levels similar to the other transgenes, based on GFP expression,neither Psur-2::CAM-1::GFP, Plst-1::CAM-1::GFP, Pdpy-8::CAM-1::GFPnor Pfos-1a::CAM-1::GFP caused a UI phenotype. These CAM-1 overexpression experiments indicate that CAM-1 can non-autonomously inhibit vulval induction. Because our analysis of cam-1 mutant alleles suggested that the CAM-1 CRD is necessary to inhibit vulval induction, we tested whether overexpression of the membrane-tethered CAM-1 CRD is sufficient to inhibit vulval induction. The cwEx164 transgene expresses CAM-1::GFP lacking the intracellular domain and the extracellular immunoglobulin and kringle domains (CAM-1Δ IgKriIntra::GFP) (Kim and Forrester, 2003). Pcam-1::CAM-1ΔIgKriIntra::GFP was sufficient to cause frequent fusion of P3.p and P4.p and to cause occasional F or 3° fates in P5.p. The mild effects on P5.p fate caused by Pcam-1::CAM-1ΔIgKriIntra::GFP compared with other transgenes could be due to less robust expression under Pcam-1 or to instability of the severely truncated protein.

Loss of any single Wnt causes only minor induction defects(Gleason et al., 2006)(Table 1); therefore, Pmyo-3::CAM-1::GFP and Psnb-1::CAM-1::GFP are likely to interfere with multiple Wnts. To determine with which Wnts CAM-1::GFPinterferes, we analyzed Pmyo-3::CAM-1::GFP in worms mutant for cwn-1, egl-20 and cwn-2, the three Wnts contributing most to VPC induction (Table 2). Loss of a Wnt that retains inductive activity in a Pmyo-3::CAM-1::GFPbackground should display enhancement of the UI phenotype, whereas loss of a Wnt that is already fully antagonized by Pmyo-3::CAM-1::GFP should not enhance the phenotype. Both egl-20(rf) and cwn-2(lf)significantly enhanced the UI phenotype of Pmyo-3::CAM-1::GFP(Table 2), indicating that these Wnts retain some or all of their inductive activity. By contrast, we found that mutation of cwn-1 did not significantly enhance the UI phenotype, indicating that the inductive activity of CWN-1 is largely abrogated by Pmyo-3::CAM-1::GFP.

Despite studies in several different organisms, the mechanism of ROR action remains elusive. In this work, we characterized the role of CAM-1/ROR as a regulator of Wnt distribution and determined that one function of ROR proteins is to sequester Wnts (Fig. 4).

Previously, it was hypothesized that CAM-1/ROR could sequester Wnts. Kim and Forrester (Kim and Forrester,2003) found that expression of the membrane-anchored CAM-1 ECD was sufficient to rescue the cell migration defects of cam-1(lf) worms and that overexpression of the membrane-anchored CAM-1 CRD caused defects in HSN and Q cell migration similar to those caused by mutation of egl-20/Wnt, leading these authors to propose that the CAM-1 CRD might sequester EGL-20/WNT. Indeed, CAM-1 was later shown to inhibit EGL-20 signaling in cell migration independently of the CAM-1 cytoplasmic domain(Forrester et al., 2004). However, the mechanism of this inhibition was not demonstrated. In particular,as the ROR2 CRD is capable of dimerizing with Fz(Oishi et al., 2003), the CAM-1 ECD could potentially function cell-autonomously by inhibiting the Wnt receptor.

The genetic data presented here indicate that CAM-1 antagonizes Wnt signaling during vulval development. We found that in lin-17 and lin-18 mutant backgrounds, cam-1 mutations cause an OI phenotype owing to elevated levels of Wnt activity. Loss of lin-17 or lin-18 might provide a sensitized background if LIN-17 and LIN-18,like CAM-1, also affect the extracellular distribution of Wnts. According to this hypothesis, mutation of lin-17 or lin-18 would similarly result in elevated extracellular Wnt levels; however, our data do not conclusively support this hypothesis.

Using vulval development as a model, we showed conclusively that CAM-1/ROR can act non-autonomously. The source of the Wnts required for vulval induction is unknown and a sequestration model would require that Pmyo-3::CAM-1::GFP (muscle expression) and Psnb-1::CAM-1::GFP (neuronal expression) are expressed in positions that enable them to restrict diffusion or transport of the Wnts to the VPCs(Fig. 3H). EGL-20/WNT forms a gradient of decreasing concentration from its site of expression in the tail extending anteriorly past the VPCs(Coudreuse et al., 2006). The distance between the source of EGL-20 and the VPCs provides ample opportunity for CAM-1 expressed in nervous or muscle tissue to prevent EGL-20 from reaching the VPCs. CWN-1/WNT is expressed in ventral cord neurons (VCNs) and posterior body wall muscle (Gleason et al., 2006; Hilliard and Bargmann, 2006). Endogenous CAM-1 expression in body wall muscle and VCNs, which are in close proximity to the VPCs(Fig. 3H), could place CAM-1 between the source of cwn-1 expression and the VPCs, allowing CAM-1 to act as a barrier and limit the amount of Wnt signal received by the VPCs(Fig. 4). CAM-1 could also function at the Wnt source to limit secretion. Consistent with inhibition by sequestration, CAM-1 overexpression antagonizes Wnt signaling independently of the cytoplasmic domain. Also, phenotypes of cam-1 mutants indicate that the membrane-anchored ECD is sufficient to inhibit Wnt signaling.

A sequestration model also predicts that CAM-1 specifically binds to those Wnts that it antagonizes. In agreement with our genetic data, we found that the CAM-1 CRD can bind to Wnts CWN-1, EGL-20 and MOM-2 in vitro. Our initial experimental design included measuring binding at various concentrations of CRD-AP that would allow us to calculate the binding affinity of each receptor-ligand pair. However, our preliminary results showed high background binding to untransfected S2 cells. We thus chose the concentration of CRD-AP at which we saw the greatest difference between binding to Nrt-Wnt-expressing and to untransfected cells and tested all of the combinations at this concentration in triplicate. Wu and Nusse(Wu and Nusse, 2002) reported that the binding of DFz2CRD-AP to Nrt-Wg-expressing cells was 10-fold higher than to untransfected cells. In our experiments, we never observed a difference greater than 2-fold. Weaker binding could be caused by a species barrier, whereby the Drosophila cells do not express a necessary cofactor or do not process Wnts in a manner conducive to high-affinity binding to C. elegans receptors. Although the binding we detected is not as robust as that observed for Drosophila Wnts and Fzs, we feel that it might still be informative and have included these values in a supplementary table (see Table S3 in the supplementary material).

Although sequestration through Wnt-CRD binding can account for many functions of CAM-1/ROR, there are examples in which CAM-1 might function by a different mechanism. The membrane-anchored ECD, but not the membrane-anchored CRD alone, was sufficient to rescue all cell migration defects of cam-1(lf) worms (Kim and Forrester, 2003). In cases where the CRD was not sufficient,ligand binding might require additional CAM-1 ECD(s) - e.g. the kringle or Ig domain - or these might be cases in which CAM-1 functions by a non-sequestration mechanism. Other examples of CAM-1 function that are probably not due to sequestration include cell-autonomous roles in CAN migration (Forrester et al.,1999) and development of the ASI sensory neuron(Koga et al., 1999). Also,CAM-1 function in Pn.aap division orientation in males requires CAM-1 kinase activity (Forrester et al.,1999; Kim and Forrester,2003). Although our study has furthered our understanding of ROR function, the role of the cytoplasmic domains remains elusive. CAM-1 shares 44% identity in the kinase domain to human ROR1 and ROR2 and none of the 21 invariant amino acids is altered(Forrester, 2002). Although ROR proteins have demonstrated kinase activity(Masiakowski and Carroll,1992; Oishi et al.,1999), the precise function of this activity has not been identified.

Our genetic and biochemical observations that CAM-1 interacts not only with EGL-20, but also with other Wnts, suggest that CAM-1 is an important general regulator of Wnt activity, rather than a specific EGL-20 antagonist. As a system in which neighboring cells reproducibly adopt distinct fates, vulva induction has enabled us to study how CAM-1 affects the precision of Wnt distribution. The subtle effects we observed upon cam-1 manipulation suggest that CAM-1 serves to buffer Wnt levels rather than to dramatically affect Wnt localization. Such buffering mechanisms might provide robustness to the Wnt morphogen gradient. The high degree of similarity between CAM-1 and vertebrate ROR proteins (Forrester,2002), in addition to the ability of ROR proteins to inhibit Wnt signaling in a kinase-independent manner, suggest a conserved function of ROR proteins to fine-tune the spatial profile of Wnt activity and to help create regions of distinct cell fate in complex multicellular organisms.

We thank Gladys Medina and Barbara Perry for technical assistance and members of the Sternberg laboratory for helpful discussions. For reagents and worms we thank W. Forrester, C. Wu, A. Fire, A. V. Maricq, M. Francis, D. Sherwood, I. Greenwald and H. Sawa. Some Nematode strains were from the Caenorhabditis Genetics Center, funded by the NIH National Center for Research Resources. We thank Cheryl Van Buskirk, Ryan Baugh, Jagan Srinivasan and Mihoko Kato for critically reading the manuscript; the Benzer laboratory(especially Gil Carvallo) for use of their microplate spectrophotometer; the Hay laboratory for use of their tissue culture hood; Jost Vielmetter and Inderjit Nangiana of the Caltech Protein Expression Facility for production of CRD-AP conditioned media; and Julie Gleason for kindly exchanging lin-17 and mom-5 mutant alleles. P.W.S. is an investigator with the Howard Hughes Medical Institute and J.L.G. was supported by the Thomas Hunt Morgan Fellowship for graduate study toward the doctor of philosophy degree in biology at the California Institute of Technology.