Interaction of Frizzled Related Protein (FRP) with Wnt Ligands and the Frizzled Receptor Suggests Alternative Mechanisms for FRP Inhibition of Wnt Signaling*

Frizzled related proteins (FRPs) comprise a family of secreted molecules that contain an N-terminal cysteine-rich domain (CRD) highly similar to the CRDs of the frizzled family of membrane-anchored Wnt receptors. FRPs have been shown to interact with Wnt proteins and antagonize Wnt signaling in a Xenopusdevelopmental model. We demonstrated that FRP antagonizes the Wnt-induced increase in uncomplexed β-catenin in both transient cotransfection and stable transformation models, where Wnt-induced morphological alterations are inhibited as well. We showed further that FRP inhibits Wnt signaling in a paracrine mode using a T-cell factor luciferase reporter to measure Wnt function. Investigation of the mechanisms responsible for FRP inhibition revealed that FRP forms complexes with WNT-1 or WNT-2 through its CRD domain. Transfection analysis with FRPs containing different tags revealed that FRP itself forms complexes and that this ability is conferred by its CRD domain. Finally, we demonstrated by cotransfection that FRP forms complexes with a prototype frizzled. All of these findings are consistent with a model by which FRP inhibits Wnt signaling through interactions with Wnt and/or formation of nonfunctional complexes with the frizzled receptor.

The Wnt proteins comprise a highly conserved, multimember ligand family, which play important roles in patterning and cell fate determination (1,2). A number of downstream components of Wnt signaling have been identified by a combination of genetic and biochemical approaches. Wnts act through the cytoplasmic protein Dishevelled to inhibit the activity of the serine-threonine kinase GSK3. GSK3 appears to bind through a bridging molecule, Axin, to the ␤-catenin-APC 1 complex and phosphorylate ␤-catenin causing its rapid degradation. Wntinduced inhibition of GSK3 leads to ␤-catenin stabilization resulting in an increased level of the uncomplexed soluble form (3)(4)(5). The latter form can interact with TCF/LEF transcription factors and, after translocation to the nucleus, activate target genes (6).
There is evidence that activation of Wnt signaling can contribute to the neoplastic process (7,8). Inappropriate expression of these ligands due to promoter insertion of the mouse mammary tumor virus (1) or targeted expression in transgenic mice causes mammary tumor formation (9). Moreover, in cell culture, several Wnt family members have been shown to induce altered morphology and increased saturation density of certain epithelial (10,11) and fibroblast (12,13) cell lines. Finally, genetic alterations affecting APC or ␤-catenin, associated with increased uncomplexed ␤-catenin levels, have been observed in human colon cancers (14), melanomas (15), and hepatocellular carcinomas (16), indicating that aberrations of Wnt signaling pathways are critical to the development of these and possibly other human cancers.
Recent studies have identified the products of the multimember frizzled family as Wnt receptors (17). These proteins are characterized by a large cysteine-rich extracellular domain (CRD), a seven-transmembrane spanning domain, and a cytoplasmic tail. Exogenous expression of Drosophila frizzled 2 (Dfz2) in a suitable recipient insect cell line conferred to the Drosophila Wnt prototype, Wingless (Wg), the ability to bind to the cell surface and to signal by increasing intracellular levels of ␤-catenin. Furthermore, transfection of Dfz2 or several different mammalian fz cDNAs into human 293T cells conferred the ability to bind Wg. The fz CRD has been shown to be its Wnt binding domain by the ability of Wg to bind to a glycosylphosphatidylinositol-anchored Dfz CRD in the absence of the transmembrane portion (17). More evidence implicating fz as the Wnt receptor has come from experiments performed in the Xenopus embryo system in which Wnt induced axis duplication when coinjected with members of the fz family (18).
A family of related molecules, which contain N-terminal CRDs highly similar to the fz CRD but then diverge and lack any transmembrane domain, has recently been identified. When coexpressed with Wnt family members in Xenopus embryos, such proteins including human frizzled related protein (FRP) and Frzb-1 were shown to antagonize Wnt-induced duplication of the dorsal axis (19 -21), indicating that these proteins may function as inhibitors of Wnt action during development. Frzb-1 has been further shown to coimmunoprecipitate with Wnt proteins in vitro and after cotransfection of COS7 cells (20). Moreover soluble Frzb-1 was shown to bind to 293 cells expressing a WNT-1 transmembrane chimera (19). In addition, mouse members of this family expressed as glycosylphosphatidylinositol-anchored fusion proteins conferred to Wg the ability to bind 293 cells (22). More recently Lin et al. (23) have provided evidence that the bovine Frzb-1 can prevent WNT-1-induced cytosolic accumulation of ␤-catenin in 293 cells and have localized sites within the CRD, which are critical for WNT-1 binding and inhibition of WNT-1-mediated axis duplication in Xenopus embryos. The present studies were undertaken to investigate the mechanisms by which FRP antagonizes Wnt function. The results have implications with respect to how FRP modulates the complex functions of the Wnt multimember family of developmentally regulated signaling molecules.

MATERIALS AND METHODS
Plasmid Construction-pCEV/WNT-1-HFc and pCEV/WNT-2-HFc, in which the IgGHFc open reading frame (24) was fused in frame downstream of the wnt coding region, were previously described (12). The pcDNA/WNT-2-HFc was constructed by inserting WNT-2-HFc into pcDNA3 (Invitrogen). pBabe/WNT-2-HFc containing the puromycin resistance gene was constructed by inserting the WNT-2-HFc fragment into pBabe, under the control of the cytomegalovirus transcriptional unit. pLNCX/Wnt-1-HA was kindly provided by Dr. J. Kitajewski (Columbia University, New York). In pMMT/FRP, the FRP coding region was introduced into MMTneo (24). To construct pcDNA/FRP-HA or pcDNA/FRP-FLAG, the FRP coding region lacking its stop codon (21) was PCR-amplified using the Expand High Fidelity PCR system (Roche Molecular Biochemicals) as described previously (12), in the presence of a forward BamHI-flanked primer, GGAAGGATCCGCCGGCAT-GGGCATCGGGCGCA, that included a Kozak consensus sequence and a reverse EcoRI-flanked primer, GGCCGAATTCCTTAAACACGGACT-GAAAGGTGGG. The FRP PCR product was inserted in frame into pcDNA3, upstream of an EcoRI-flanked HA (YPYDVPDYA) or M2 (DYKDDDDK) (Eastman Kodak Co.) epitope-encoding sequence. To construct pcDNA/CRD-FRP-HA or pcDNA/CRD-FRP-FLAG, the N-terminal fragment of FRP was PCR-amplified using the above mentioned forward BamHI-flanked primer and a reverse EcoRI-flanked primer, GGAAGAATTCGGCGATGCAGACGTCCCCCTCCGG, and the PCR product was introduced in frame upstream of the HA-or M2-encoding sequence, as detailed above. To construct pcDNA/CRD-Hfz6-FLAG, the N-terminal domain of Hfz6 (GenBank TM accession number AF072873) was PCR-amplified using the forward HindIII-flanked primer, GGAAAAGCTTGCCACCATGGAGATGTTTACATTTTTGTTG, that included a Kozak consensus sequence and a reverse EcoRI-tagged primer, GGAAGAATTCGCATGGAGGCGCACACTGGTCAAT. The PCR product was introduced in frame upstream of M2-encoding sequence as detailed above. Sequence analysis was performed to ascertain the authenticity of the PCR-generated products. pGST-E-cadherin was described previously (12). Two TCF/Luc reporter constructs, pGL3-OT containing TCF-responsive domains and pGL3-OF containing mutantbinding sites, were kindly provided by B. Vogelstein (Johns Hopkins Oncology Center, Baltimore). A schematic diagram of the constructs generated for this study is shown in Fig. 1.
Stable Transfection-NIH3T3 cells were plated at 1.5 ϫ 10 5 cells per 100-mm dish. After 24 h cells were transfected with 1 g of each plasmid DNA, by the calcium phosphate coprecipitation method (25) as described (26). Cultures were incubated with the DNA precipitates for 18 -20 h and then washed in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% calf serum (Life Technologies, Inc.). To ensure selection of cells coexpressing WNT-2 and FRP, different selectable markers were utilized in each construct. Cotransfectants were doubly selected by addition of 750 g/ml geneticin (Life Technologies, Inc.) and 2 g/ml puromycin (Calbiochem) to the medium. Cells expressing WNT-2 or FRP individually were cotransfected with control vectors containing the other selectable marker.
Transient Transfection-For transient transfection 70 -80% confluent cultures of 293T cells were transfected by the calcium phosphate method using 3 g of each plasmid DNA. When necessary the amounts of DNAs used for cotransfection were adjusted to yield similar levels of protein expression. Cells were exposed for 5-6 h to the DNA precipitates and then washed in growth medium. After 48 h, cells were harvested and cell lysates prepared for analysis as indicated below.
GST-E-cadherin Binding Assay-GST-E-cadherin was expressed in bacteria and purified from the bacterial lysates by binding to glutathione-Sepharose beads as described previously (12). Cells were lysed in immunoprecipitation buffer (10 mM sodium phosphate, pH 7, 0.15 M NaCl, 1% Nonidet P-40, 10 g/ml aprotinin, 10 g/ml leupeptin, 2 mM sodium vanadate, 2 mM phenylmethylsulfonyl fluoride), and cell extracts were clarified by centrifugation. In order to perform a quantitative analysis, two different amounts of total protein (0.1 and 1 mg) were analyzed for each sample by incubation with Sepharose beads (Amer-sham Pharmacia Biotech) bound to the GST-E-cadherin. After 1 h of incubation with rotation at 4°C, the beads were collected by centrifugation, washed, and dissolved in Laemmli buffer. Samples were subjected to SDS-PAGE followed by immunoblotting with 0.5 g/ml anti-␤-catenin antibody (Transduction Laboratories). Detection was perperformed with 125 I-labeled protein A (Amersham Pharmacia Biotech).
Paracrine Assay for Wnt Signaling-The paracrine assay was performed as described. 2 Briefly, 293T cells transiently transfected with 1 g of pLNC/Wnt-1-HA or with an empty vector were cocultured with 293T cells transiently transfected with 1 g of pcDNA/FRP-HA, pGL3-OT, or pGL3-OF and 0.1 g of ␤-galactosidase expression vector. After 48 h, the relative luciferase units (RLU) were measured using the Promega luciferase assay system according to the manufacturer's protocol. The activities were normalized for transfection efficiency using ␤-galactosidase activity. TCF/LEF-dependent luciferase activity was calculated by subtracting RLU levels obtained with the pGL3-OF reporter from those obtained by pGL3-OT (27).
Western Blot and Coimmunoprecipitation-Cultures were solubilized in lysing buffer (0.01 M phosphate buffer, 1% Triton, 0.5% sodium deoxycholate, 0.1% SDS, 0.1 M NaCl, 5 M EGTA, 10 g/ml aprotinin, 10 g/ml leupeptin, 2 mM sodium vanadate, 2 mM phenylmethylsulfonyl fluoride). Around 100 g of total cell lysates were resolved by SDS-PAGE, and proteins were transferred to Immobilon-P membranes. After blocking with 5% bovine serum albumin in PBS, filters were incubated with the specific primary antibody (2.6 g/ml horseradish peroxidase-conjugated rabbit anti-mouse HFc, Dako; 10 g/ml anti-FLAG M2, Kodak; 1.5 g/ml anti-HA obtained from the Hybridoma Center, Mount Sinai School of Medicine, New York). When the anti-FLAG or the anti-HA primary antibodies were used, membranes were washed in PBS/Tween 20 and incubated for 1 h with a secondary horseradish peroxidase-conjugated rabbit anti-mouse antibody. After washing 6 times in PBS/Tween 20, membranes were subjected to ECL analysis (Amersham Pharmacia Biotech). For coimmunoprecipitation analysis, cells were solubilized in lysing buffer (see above), and extracts were clarified by centrifugation at 10,000 ϫ g for 20 min at 4°C. Around 1.5-3 mg of total cell lysates were incubated with the specific antibody (24 g of goat anti-mouse HFc, Pierce; 10 g of anti-FLAG M2; 15 g of anti-HA) for 1 h on ice and then incubated with rotation with protein A beads (Pierce) for 1 h at 4°C. The beads were collected by centrifugation, washed three times in immunoprecipitation buffer, dissolved in Laemmli buffer, and analyzed by SDS-PAGE, followed by detection with each specific antibody. The extracellular matrix fraction was prepared as follows. Cultures were washed twice with PBS containing 1 mg/ml aminocaproic acid and 1 mM phenylmethylsulfonyl fluoride and then incubated with the same solution containing 5 mM EGTA for 15 min at 37°C. Cells were removed, and the plates were washed 2-3 times with the PBS solution and twice with distilled H 2 O. The extracellular matrix was then solubilized in 200 l of Laemmli buffer.
Metabolic Labeling-Subconfluent 293T cultures were transiently transfected with 3 g of each plasmid DNA. Forty hours later, cells were washed and incubated for 30 min in methionine-or methionine-and cysteine-free DMEM in the absence of serum. Cells were then labeled for 3 h with 3.5 ml of the same medium containing 200 Ci of [ 35  . Labeled cells were rinsed with cold PBS, lysed in 0.5 ml of lysing buffer, and clarified by centrifugation. Cell extracts were incubated with 24 g of goat antimouse HFc (Pierce), 15 g of anti-HA, 10 g of anti-PDGF, or 10 g of anti-FLAG. Immunoprecipitates were recovered with protein A (Pierce) beads, dissolved in Laemmli buffer, and analyzed by SDS-PAGE. After electrophoresis, the gels were treated with En 3 hance (DuPont), dried, and autoradiographed.

FRP Inhibits Wnt Function in Both Autocrine and Paracrine
Modes-Studies to date showing that FRPs antagonize Wnt action have generally utilized in vivo models, with the exception of one recent report in which cotransfection of 293 cells with Frzb-1 and WNT-1 resulted in the inhibition of Wntinduced accumulation of cytosolic ␤-catenin (23). To confirm and extend these findings with respect to FRP, we cotransfected 293T cells with WNT-1 and FRP. This transient expression system has the advantage of inducing high levels of pro-tein expression. However, we observed that cotransfection led to variable levels of expression of the exogenous proteins. To overcome this difficulty, WNT-1-HFc (12) was cotransfected with the vector control or HA-tagged FRP (Fig. 1) at different DNA ratios in order to ensure that similar WNT-1 protein levels could be obtained in the presence or absence of FRP expression.
Under conditions in which 293T cells transfected with WNT-1, in the presence or absence of FRP, expressed similar WNT-1 levels ( Fig. 2A), FRP was also specifically detected as a 36-kDa species in WNT-1/FRP-cotransfected cells (Fig. 2B). Total and uncomplexed ␤-catenin levels (12) were quantitated in the same cells using a GST-E-cadherin binding assay as described under "Materials and Methods." Although total ␤-catenin levels were not appreciably altered (data not shown), Fig. 2C shows that WNT-1 induced an increase of around 10-fold in the amount of uncomplexed ␤-catenin over that of the vector control. Of note, expression of FRP dramatically inhibited this increase (Fig. 2C). Similar results were obtained with WNT-2 (data not shown). These findings established, under carefully controlled conditions, that FRP acts to inhibit Wnt signaling functions responsible for increased free ␤-catenin levels.
We next investigated the ability of FRP to inhibit Wnt functions in a stable transformation model. To do so, WNT-2 and FRP genes were cotransfected using different selectable markers into NIH3T3 cells, in which both WNT-1 and -2 induce morphologic alterations that include increased saturation density as well as higher steady state levels of uncomplexed ␤-catenin (12). Whereas WNT-2 transfectants exhibited abnormal growth and achieved higher cell density, double marker selected cultures expressing both FRP and WNT-2 showed little if any morphological differences from vector control or FRP transfectants (Fig. 3A). Similar results were obtained in multiple experiments. Moreover, inhibition of the transformed phenotype by FRP was correlated with inhibition of Wnt function as indicated by the substantially lower levels of WNT-2-induced free ␤-catenin in the stable cotranfectants (Fig. 3B).
Both Wnt and FRP proteins are processed through the secretory pathway but remain associated with cells that have been analyzed to date (21,28,29). The sites of interactions and mechanisms by which FRP exerts its inhibitory functions on Wnt signaling remain to be elucidated. To address this question, we first analyzed the localization of FRP in transfected 293T cells. As shown in Fig. 4, FRP was found to be present in cell lysates and the extracellular matrix but was not detected in the medium using conditions known to optimize FRP release into the medium of cultured fibroblasts (21). When the results were normalized for the amount of cell lysate, culture fluid, or extracellular matrix in the same Petri dish, FRP was found to be predominantly cell-associated (90%) with the remainder present in the extracellular matrix (Fig. 4). These results indicate that FRP remains tightly cell-associated when expressed by 293T cells.
Experiments were next performed to analyze the ability of FRP to inhibit Wnt actions in a paracrine mode. To do so we utilized a recently developed transient assay involving a luciferase reporter for TCF transcriptional activation (27)  been shown to respond to autocrine Wnt-1 by stimulating TCFdependent transcription (13). We established a cocultivation assay in which cells transiently transfected with Wnt-1 are able to stimulate the activation of the TCF reporter in target cells in a paracrine mode. 2 293T cells were transfected with Wnt-1 or vector control and then cocultivated with 293T cells, which were cotransfected with the TCF reporter and either the vector control or FRP. As shown in Fig. 5, TCF reporter activity was increased over 20-fold in 293T cells in response to cocultivation with Wnt-1-expressing cells. However, the responsiveness of FRP expressing 293T cells to paracrine-acting Wnt-1 was not significantly different from those expressing vector alone. These results indicate that Wnt-1 function can be inhibited by FRP in a paracrine mode.
FRP Interacts with Wnts-Previous studies have shown that Frzb-1 coimmunoprecipitates several members of the Wnt family (20,23). To confirm and extend these findings with respect to FRP, we performed immunoprecipitation experiments with lysates of 293T cells cotransfected with WNT-1-HFc or WNT-2-HFc in combination with either FLAG-tagged FRP or FLAGtagged C-terminally truncated FRP encoding only its CRD (Fig.  1). Antiserum against the HFc tag was used for immunoprecipitation followed by immunoblotting with anti-FLAG to detect FRP or CRD-FRP. As shown in Fig. 6A, FRP and CRD-FRP could be immunoprecipitated in complexes with either WNT-1 or -2. The specificity of these interactions was demonstrated by the fact that anti-HFc failed to immunoprecipitate FRP or CRD-FRP in the absence of Wnt-1 coexpression and by the fact that Wnt with a different tag, HA, was similarly able to immunoprecipitate FRP-FLAG (data not shown).
To investigate further the nature of Wnt-FRP interactions as well as to examine the stoichiometry of the complex, we performed biosynthetic experiments. Thus, 293T cells were transfected with WNT-2-HFc in the presence or absence of FRP. Around 40 h later cultures were labeled for 3 h with [ 35 S]methionine followed by immunoprecipitation analysis of cell lysates with anti-HFc. Fig. 6B shows that anti-HFc specifically immunoprecipitated the radiolabeled 66-kDa WNT-2-HFc from lysates of WNT-2-transfected 293T cells. In contrast, both were collected. Aliquots reflecting 1/50 of the cell lysates and culture fluids, respectively, as well as 1/7 of the extracellular matrix from the same culture dish were analyzed by SDS-PAGE followed by immunoblotting with anti-HA as described under "Materials and Methods." Similar results were obtained with heparin at higher concentrations (50 g/ml).

FIG. 5. FRP inhibits Wnt-1-induced TCF/LEF-dependent transcription in a paracrine mode.
293T cells in 6-well plates were cotransfected with FRP or empty vector, as well as either pGL3-OT or pGL3-OF TCF luciferase reporters and a ␤-galactosidase expression vector. The transfected cultures were then cocultivated with 293T cells transfected with either Wnt-1 or vector as described under "Materials and Methods." After 48 h luciferase activity was measured and levels of Wnt-1-induced RLU presented relative to the RLU obtained by coculturing with the empty vector transfectants. Results of one representative experiment of three independently performed assays are shown. Determinations were made in duplicate for each experimental point, and the standard deviation was less than 4%.
WNT-2-HFc and a 36-kDa species corresponding to FRP were coimmunoprecipitated from cotransfected cultures. No other major bands were observed, arguing against the possibility that other protein(s) participated in FRP-WNT-2 binding interactions. Of note, the intensity of the band corresponding to FRP was significantly greater than the signal associated with WNT-2-HFc in these complexes (Fig. 6B). Similar results were obtained in several independent experiments. The ratio between the intensity of the FRP and the WNT bands was around 1.7 as calculated by densitometry. As a similar number of methionines are present in both proteins, these results raised the possibility that Wnt-2 might be present in a complex containing more than one FRP molecule.
In order to establish the specificity of the Wnt-FRP interactions, we performed biosynthetic experiments using PDGFB as a control for binding interactions. PDGFB, like Wnt, contains multiple cysteine residues and is also processed through the secretory pathway (30,31). FRP-HA and WNT-2-HA were cotransfected with PDGFB, and cells were labeled with [ 35 S]methionine and [ 35 S]cysteine. Cell lysates were subjected to immunoprecipitation with either a PDGFB monoclonal antibody (32) or anti-HA. Fig. 6C shows that neither WNT-2 nor FRP interacted with PDGFB. These results strongly support the specificity of the Wnt-FRP complexes observed (Fig. 6, A and  B).
Analysis of FRP Structure-We took advantage of the availability of both HA-and FLAG-tagged FRP and CRD-FRP constructs ( Fig. 1) to investigate FRP structure. Thus, 293T cells were transfected individually or cotransfected with the FLAGand HA-tagged versions of FRP or CRD-FRP, respectively. Fig.  7A shows the analysis of transfected cell lysates by immunoblotting with anti-FLAG (top) or following immunoprecipitation with the same antibody (bottom). The results revealed that each of the FLAG-tagged proteins was expressed well and was immunoprecipitable by anti-FLAG. Fig. 7B further shows that anti-HA detected FRP-HA and CRD-FRP-HA in cells transfected with these expression vectors (Fig. 7B, top). Of note, FRP-HA was specifically immunoprecipitated by anti-FLAG from cells coexpressing FRP-FLAG. Similarly, CRD-FRP-HA was specifically immunoprecipitated from lysates of cells cotransfected with CRD-FRP-FLAG (Fig. 7B, bottom). Furthermore, antibody specificity controls indicated that FRP-HA and CRD-FRP-HA were immunoprecipitated by anti-FLAG only in the presence of their FLAG-tagged versions. These findings indicate that FRP, itself, can form complexes and that its CRD is sufficient to mediate these interactions.
We next investigated whether FRP was capable of forming such complexes following secretion. Because FRP remains tightly cell-associated (Fig. 4), experiments were performed by independently transfecting HA or FLAG-tagged FRPs in 293T   FIG. 6. A, coimmunoprecipitation of FRP or CRD-FRP with WNT-1 or WNT-2. 293T cells were cotransfected with FLAGtagged pcDNA/FRP or pcDNA/CRD-FRP together with HFc-tagged pCEV/WNT-1 or pCEV/WNT-2. The lysates were subjected to immunoprecipitation (IP) with anti-HFc, followed by immunoblotting with anti-FLAG to detect FRP and CRD-FRP. B, analysis of FRP-WNT-2 interactions in metabolically labeled cells. 293T cells were transfected with either vector control HFc-tagged pcDNA/WNT-2 alone or in the presence of pcDNA/FRP. After labeling, cell lysates were immunoprecipitated with anti-HFc, followed by SDS-PAGE analysis. C, PDGF does not form complexes with Wnt or FRP. Lysates from labeled 293T cells transfected with vector, pcDNA/PDGFB, or pcDNA/PDGFB cotransfected with HA-tagged WNT-2 or FRP were subjected to immunoprecipitation with either anti-PDGF or anti-HA and analyzed by SDS-PAGE. cells, followed by cocultivation of the transfectants for 48 h. Immunoprecipitation of cell lysates revealed FRP-HA-FRP-FLAG complexes indicating that these interactions can occur following secretion from cells (Fig. 8).
FRP Forms Complexes with the Hfz6 Receptor-Evidence that the CRD of FRP has a predicted structure similar to that of fz (21) suggested the possibility that fz itself might interact with FRP. We have cloned several members of the fz family including Hfz6 from a human cDNA library. For the present studies, we generated an expression vector encoding a FLAGtagged C-terminal truncated version of Hfz6 containing its CRD region but lacking the entire transmembrane and cytoplasmic domains (Fig. 1). This construct, designated CRD-Hfz6-FLAG, was cotransfected into 293T cells with either vector control, FRP-HA, or CRD-FRP-HA. Immunoblotting analysis with anti-FLAG revealed expression (Fig. 9A, top) and specific immunoprecipitation (Fig. 9A, bottom) of FLAG-tagged CRD-Hfz6. The expression of HA-tagged FRP and CRD-FRP in transfectants can be seen in Fig. 9B (top). When anti-HA immunoblotting analysis of anti-FLAG immunoprecipitates was performed, the results showed the specific interaction of CRD-Hfz6 with both FRP and CRD-FRP (Fig. 9B, bottom). Immunoprecipitation experiments performed following cocultivation of cells independently transfected with CRD-Hfz6 or FRP confirmed these interactions (data not shown). These findings establish the ability of FRP to form complexes with a prototype fz and provide evidence that these interactions are mediated by their homologous CRD regions.
To assess the stoichiometry of the FRP-CRD-HFz6 complexes, we performed biosynthetic experiments. 293T transfectants expressing FRP-HA and CRD-Hfz6-FLAG individually or together were labeled with [ 35 S]methionine and [ 35 S]cysteine. Cell lysates were then immunoprecipitated with anti-HA or anti-FLAG. As shown in Fig. 10, each protein was readily detected in singly transfected cultures, with FRP expressed at higher levels than CRD-Hfz6. When cotransfectants were subjected to coimmunoprecipitation with anti-HA or anti-FLAG, FRP-CRD-Hfz6 complexes were detectable in both cases. When the lysates were immunoprecipitated with anti-HA, the band corresponding to FRP was stronger than that corresponding to CRD-Hfz6, consistent with the higher level of FRP expression. However, when anti-FLAG was used to immunoprecipitate CRD-Hfz6, present in limiting amounts, the coprecipitated FRP band showed a very similar intensity. The ratio between the intensity of the signal for FRP and the CRD-Hfz6 band was found to be 1.0 when analyzed by densitometry. Since the number of cysteines and methionines in the two molecules is similar, these observations support the heterodimeric interaction between the two proteins. No other species was consistently detectable in both anti-HA and anti-FLAG coimmunoprecipitates. However, we cannot rigorously exclude the presence of other molecules in the complex. To confirm the specificity of the FRP-CRD-Hfz6 interactions, cotransfection of CRD-Hfz6 with PDGFB revealed the absence of any detectable complex.

DISCUSSION
Our present studies provide evidence concerning mechanisms involved in antagonism of Wnt signaling by members of the FRP family. Several previous studies have reported that Wnts can bind to fz receptors or FRPs presented at the cell surface (17,22). There are also reports of Wnt binding to Frzb as measured by in vitro or in vivo coexpression and coimmunoprecipitation (20,23,33). We confirmed these observations with respect to FRP, further demonstrating complexes containing WNT-1 or WNT-2 and FRP in coexpressing cells. Metabolic labeling analysis suggested a direct interaction between FRP and Wnt, which does not require the participation of any additional major protein. Moreover, this binding appeared to involve more than one FRP interacting with each Wnt molecule.
By use of different immunologic tags, FRP complexes containing both tags were specifically detectable. Such complexes were further shown to involve the CRD region of the molecule. FRP and fz contain analogous CRD domains (21), and it was possible to demonstrate that FRP also possessed the ability to form complexes with the CRD of a prototype fz, Hfz6. Biosynthetic experiments further revealed that these complexes were consistent with FRP-CRD-Hfz6 heterodimers. Thus, FRP has the ability to independently interact with Wnts and with their receptors.
It might be argued that cysteine-rich proteins such as Wnt, FRP, and Hfz6, which are processed through the same secre-tory pathway, could form illegitimate complexes by improper disulfide bond formation under conditions of exogenous coexpression. However, cotransfection with another cysteine-rich, secreted molecule, PDGFB (30,31), yielded no evidence of complexes with any of the same molecules. Moreover, FRP complexes were detected when cells were independently transfected with FRPs containing different tags and cocultivated. All of these findings strongly support the specificity of the complexes involving FRP and either Wnts or Hfz6. The strength of the binding interactions for each of these complexes was sufficient to survive exposure to 0.1% SDS. Such conditions do not impair antigen-antibody binding but do disrupt interactions between heterodimers of tyrosine kinase growth factor receptors (34,35). Continued investigation of the physical and biochemical characteristics of the complexes involving FRP with Wnts and the fz receptors should provide further insights into the nature of these interactions.
Our present studies are consistent with at least two models by which FRP may function as a naturally occurring antagonist of Wnt signaling. According to the first, FRP binds Wnt and exerts its inhibitory function by competing for the ability of the Wnt ligand to interact with the fz receptor. However, a second model emerges from our results showing that FRP forms complexes with the fz receptor and that these interactions are mediated by their homologous CRDs. Such findings support the possibility that fz, itself, may form complexes which function as the signaling and/or binding receptor for Wnt. According to this model, FRP could act by a dominant negative mechanism, interacting with the fz receptor and forming nonfunctional complexes that are incapable of transmitting the Wnt signal.
In several assay systems, FRP acted to inhibit Wnt-induced alterations of ␤-catenin regulation and TCF transcription. Moreover, we showed that FRP inhibited Wnt function in both autocrine or paracrine modes. Similar conclusions were derived from microinjection of Xenopus embryos (20). Although as shown here and in previous studies (21), secreted FRP remains predominantly associated with the cell or extracellular matrix, it could be postulated to inhibit paracrine-acting Wnts by either of the above models. Lin et al. (23) reported that certain Frzb deletion mutants that retained the ability to bind WNT-1 failed to block WNT-1-induced axis duplication. This lack of a strict correlation between the ability of Frzb to interact physically with Wnts and its ability to inhibit Wnt signaling suggests that the mechanism of antagonism is more complex than can be explained solely by Wnt binding. Genetic analysis aimed at localizing FRP domains required for Wnt binding, as well as for formation of complexes with the fz receptor, should be helpful in assessing the proposed models for FRP inhibition of Wnt signaling. These models are not mutually exclusive, and cooperation of the two mechanisms might be necessary to ensure more complete functional inhibition of Wnt signaling.
The FRP family is not the only example of naturally occurring Wnt antagonists. Recently, dickkopf-1 (dkk-1), which encodes Dkk-1, a secreted inducer of Spemman's organizer in Xenopus was isolated by an expression cloning strategy (36). This 40-kDa secreted protein contains two cysteine-rich domains and represents a new multigene family. Dkk-1 and Frzb have overlapping patterns of expression during development. Genetic analysis indicates further that Dkk-1 inhibits Wnt signaling upstream of dishevelled (36). It will be of interest to determine whether Dkk-1 acts like FRP/Frzb to bind Wnt ligands and/or their fz receptors or, indirectly, by activating an independent Wnt inhibitory pathway.