Inhibition of Wnt/β-Catenin Signaling by a Soluble Collagen-Derived Frizzled Domain Interacting with Wnt3a and the Receptors Frizzled 1 and 8

cDNA Clones

Human Igκ-FZC18-myc/pSecTag2 (carrying an Igκ signal peptide) and mouse Wnt3a-V5/pCDNA3.1 mammalian expression vectors, Super8•Topflash and Super8•Fopflash CRT reporters, Cyclin D1 promoter reporter D1Δ-944pXP2 and the normalization Renilla luciferase vector pGL4.70[hRluc] were previously described [7]. The episomal expression vector pCEP-PU was from T. Sasaki [11]. Igκ-FZC18-myc was transferred from pSecTag2 to pCEP-PU by PCR synthesis of an 875-bp fragment carrying 5′-NheI and 3′-BamHI restriction sites in the forward and reverse primers, respectively. The PCR fragment was ligated into pCRII-Topo-TA (Invitrogen), excised and subcloned into pCEP-PU. Mouse FZD8_CRD-myc/pcDNA3, mFZD8_CRD-Fc/pRK5 [12] (Addgene plasmid 16689) and empty pRK5 plasmid were from X. He, mFZD8_CRD-myc-GPI was from J. Nathans [13], mFZD8-myc receptor/pEF1A [14] was from R. Nusse, pEF1/myc-his was from Invitrogen and rat FZD1-myc (Addgene plasmid 16798) was from R. Moon. Human FZC18_CRD was PCR cloned into BamHI and KpnI in pIDZ-Fc in frame with an Igκ signal sequence and a C-terminal human IgG Fc tag for affinity purification. A thrombin cleavage site was included to allow removal of the Fc tag. The sequences of primers were: forward, 5′-GGG GGA TCC GCC CTG CTC GGG GCT GAC-3′; reverse #1, 5′-GGG CTC GAG AGA TCC ACG CGG TAC CAG TGC AGC CGG CCC AAT GAG-3′; reverse #2, 5′-GGG CTC GAG TGC AGC CGG CCC AAT GAG-3′. Constructs were stably transfected in DHFR-deficient CHO cells with Effectene transfection reagent (Qiagen), and clones selected in media containing G418 (500 µg/ml, Sigma) and lacking hypoxanthine and thymidine. All cDNAs were checked by automatic sequencing (Sequencing Facility, Rennes University Hospital, France).

Biological Activity of Soluble FZC18

Collection of conditioned media (CM) from parental L cells (control CM) and Wnt3a CM was performed as recommended by ATCC and by R. Nusse lab website. For other CM, HEK293 EBNA cells were seeded at 2.2×10⁶ cells/10 mm dishes and transfected with either FZC18-myc/pCEP-PU, mFZD8_CRD-myc/pcDNA3, mFZD8_CRD-Fc/pRK5 or with the respective empty vectors and, 24 hr later, fresh media were replaced by DMEM (4.5 g/l glucose) without phenol red or FCS (Invitrogen). Conditioned media were collected 48 hr later, centrifuged at 450 g and filtered (0.2 µm). To obtain recombinant FZC18_CRD, conditioned media from hFZC18_CRD-Fc clones were screened for protein expression by ELISA and positive clones were confirmed by Western blot analysis using anti-human IgG-Fc antibody. The positive clones were further adapted to CD OptiCHO medium supplemented with 8 mM L-Glutamine. hFZC18_CRD-Fc producing cells were seeded into spinner flasks at 2×10⁵ cells/ml and incubated at 37°C and 5% CO₂ with agitation at 80 rpm in humidified air for 10 days. The medium was collected, cleared by centrifugation, filtered (0.45 µm) and stored at 4°C until purification. The samples were loaded on a protein A column following equilibration with 20 mM sodium phosphate, 20 mM sodium citrate, pH 7.5. The column was washed with the same buffer until effluent absorbance returned to baseline. The bound proteins were eluted with 20 mM sodium phosphate, 100 mM sodium citrate, pH 2.5 followed by rapid neutralization by adding 0.1 volume of 1 M Tris-hydrochloride, pH 9.0. The yield of the purified proteins was approximately 1∼5 mg/l and purity was over 40%, as estimated by sandwich ELISA using anti-Fc antibody (Abcam AB1927) for capture and peroxidase-conjugated secondary antibody (Sigma A0170) for detection. Purified proteins were stored at −80°C until use.

Immunological Methods

Coimmunoprecipitations were done by incubating either FZC18-myc or FZD8_CRD-myc pre-cleared CM with recombinant mWnt3a (100 ng/ml; 2.7 nM) and either mouse anti-myc or mouse IgG₁ (Dako) on a rotary wheel at 4°C overnight. Then, protein G magnetic beads (New England Biolabs), saturated overnight in protein extracts from 293EBNA cells in RIPA buffer (TrisHCl 50 mM, pH 7.4; 1% Triton-X-100; 25 mM Hepes; 150 mM NaCl; 0.2% Sodium deoxycholate, 5 mM MgCl₂), were added to immunocomplexes and incubated at 4°C, for 3 hr. After washing in RIPA buffer, complexes were eluted in denaturing sample buffer, resolved by 10% PAGE-SDS and immunoblotted. For reverse coimmunoprecipitation experiments, either rabbit anti-mWnt3a (C64F2, Cell Signaling) or rabbit IgG (Dako) was incubated either with recombinant mWnt3a plus FZC18-myc CM or with recombinant mWnt3a plus FZD8_CRD-myc CM. Coimmunoprecipitation of FZC18-myc with either recombinant mFZD1_CRD-Fc (100 ng/ml) or mFZD8_CRD-Fc CM was done as described above, using protein G magnetic beads binding the Fc tags. Immunoblots were performed with mouse anti-myc (Invitrogen) and with monoclonal rat anti-FZD1_CRD or anti-FZD8_CRD antibodies (R&D). Signal from immunoblots was detected by enhanced chemiluminiscence, as described [7].

The frizzled domain of collagen XVIII inhibits cell proliferation and DNA synthesis

We produced zeocin-resistant mass cultures of 293T cells stably expressing FZC18 or empty vector. To avoid clonal variability, we expanded colonies showing different densities of FZC18 (+) cells (Figure 1B). As FZC18 locates preferentially at the cell surface [7], cell permeabilization followed by immunocytochemistry allowed identification of all cells expressing the protein of interest, regardless of protein maturation. Thus, batch #1 showed a lower density of FZC18 (+) cells (Figure 1B) and lower FZC18 expression by immunoblot than batch #4 and #5 cells (Figure 1C). When passaged routinely, FZC18 (+) cells grew more slowly, formed smaller cell plates than vector cells and secreted soluble FZC18 (Figure S1, B and C). Simultaneous detection of N- and C-terminal epitopes in this fusion protein indicated preservation of FZC18 integrity in cells (Figure 1C) and in the medium (Figure S1C). An 8-day time course cell proliferation assay showed that FZC18-expressing cells grew more slowly than vector cells (Figure 2A). ^3HThymidine incorporation rates into DNA showed that FZC18 reduces cell proliferation and DNA synthesis (Figure 2B). Throughout the 8-day cell proliferation assay, mitochondrial succinate dehydrogenase activity in living cells (MTT assay) confirmed the decrease in cell growth in FZC18-expressing cells (Figure S2). The decrease in proliferation rates was correlated with the expression levels of FZC18 in the stable cell cultures (Figure 1B). No significant difference in spontaneous cell death was observed in these cells compared to vector-expressing cells by flow cytometry search for subG1, hypo-diploid cells (not shown).

FZC18 reduces cell sensitivity to soluble Wnt3a

Incubation of vector and FZC18-expressing 293T cells with conditioned medium (CM) from L cells secreting soluble Wnt3a (Wnt3a CM) confirmed that FZC18 reduces Wnt3a-induced Wnt signaling (Figure 3), β-catenin stabilization, cyclin D1 promoter activity and protein expression (Figure S3). FZC18-expressing cells showed lower amounts of steady-state β-catenin protein (Figure S3A) and cyclin D1 promoter activity than control cells (Figure S3B). In particular, cyclin D1 protein expression in response to soluble Wnt3a was considerably stronger in vector cells than in FZC18 cells (Figure S3C), indicating that FZC18 abrogates the response to Wnt3a. The dose-response curve to different dilutions of Wnt3a CM showed that FZC18-expressing cells could efficiently build up a CRT response to Wnt3a, in such a way that the higher the concentration of Wnt3a CM, the higher the fold-change in CRT. However, the absolute CRT levels in FZC18-expressing cells were 5 to 8 folds lower than those in vector cells (Figure S3D).

Despite lower pre-Wnt3a and post-Wnt3a β-catenin stabilization and downstream signaling events in FZC18 cells (Figures 3 and S3), a given strength of Wnt3a stimulus induced the same fold-change in CRT both in vector and FZC18 cells (Figure S4). Likewise, it has been recently shown that in a normal cell context, different cell systems respond to Wnt stimulation with similar fold-change despite their different starting and output levels in Wnt/β-catenin signaling [15]. Thus, FZC18 may not impair downstream processing of Wnt stimuli, but it seems to decrease cell sensitivity to Wnt3a, probably by blocking Wnt access to frizzled receptors.

Soluble FZC18 binds Wnt3a

Co-expression of FZC18 and Wnt3a in non permeabilized AT3F1S315 mouse hepatoma cells [16] followed by confocal microscopy analysis revealed that both proteins colocalized at the cell surface, highlighting cell contacts (Figure 4A). We further confirmed that HEK293T cell clones stably expressed FZC18 at the cell surface by immunostaining with antibodies detecting the N- and C-termini of the FZC18-myc fusion protein (Figure S5A). Moreover, subcellular fractionation confirmed that FZC18 was exclusively detected in the crude membrane fraction (Figure S5B), indicating that the protein is indeed addressed to the secretory pathway.

Using protein extracts from cells cotransfected with both FZC18 and Wnt3a, we previously showed that these ectopically expressed molecules interact [7]. However, cysteine-rich proteins like Wnts and the Frizzled CRDs may get clogged within the secretory pathway [2], leading to high intracellular concentration and spurious interactions. Here, we wished to study these interactions in a cell-free system, using soluble FZC18 and Wnt3a, at low concentrations. Despite easy detection of FZC18-myc by immunocytochemistry, immunoblot detection showed no signal in non-concentrated CM, in contrast to FZD8_CRD-myc, which was detected at high levels (Figure S5C). Concentration of FZC18 CM by 13 folds was required to observe a detectable signal by immunoblot (Figure S5C). For coimmunoprecipitation, we added purified recombinant Wnt3a to non-concentrated FZC18-myc or FZD8_CRD-myc CM. Wnt3a concentration (2.7 nM) was within the physiological range [13], [17], [18], [19], [20]. Under these conditions, both FZC18-myc and FZD8_CRD-myc pulled down Wnt3a (Figure 4B). Accordingly, reverse co-immunoprecipitation revealed that Wnt3a pulled down both FZC18-myc and FZD8_CRD-myc (Figure S5D). Additionally, both precipitation and immunoblot with anti-myc antibody confirmed the presence of soluble FZC18 and FZD8_CRD at the expected amounts in these CM (Figure S5E).

Extracellular FZC18 inhibits Wnt3a-induced Wnt/β-catenin signaling

We tested whether the CRD of FZC18 (FZC18_CRD) could effectively inhibit Wnt signaling. FZC18_CRD was cloned in frame with a human Igκ signal sequence and a human Fc tag from IgG. Human FZC18_CRD-Fc preparations with 40% purity were tested for biological activity. As expected, hFZC18_CRD-Fc dose-dependently inhibited Wnt3a-induced CRT (Figure 5). Inclusion of a thrombin cleavage site did not significantly affect Wnt inhibitory activity of hFZC18_CRD-Fc (Figure S6). Taken together, these findings support the concept that FZC18 exerts its biological effects in the extracellular compartment and that the CRD of FZC18 has Wnt inhibitory activity.

FZC18 seems to provide short-range signals, thus working as an SFRP-like molecule [8]. We thus confirmed whether FZC18-expressing cells impact on the microenvironment of adjacent cells and modulate their response to Wnt stimuli. FZC18-expressing cells were co-cultured with parental 293 cells expressing the CRT reporter. Co-cultures were established at different ratios of FZC18 (+) cells to a constant number of reporter cells in the presence of 50% Wnt3a CM (Figure S7). Under these conditions, the response of the reporter cells to soluble Wnt3a was inversely proportional to the number of FZC18 (+) cells in the co-culture system.

Frizzled 1 and 8 receptors partially rescue the inhibition of Wnt3a-induced β-catenin signaling by FZC18

We tested whether increasing the availability of cell surface frizzled receptors in FZC18-expressing cells could compete with FZC18, thereby enhancing Wnt signaling. To this end, FZC18-expressing and vector cells were transfected with increasing amounts of full-length FZD1 or FZD8 receptor cDNAs and incubated in the presence of Wnt3a CM (Figure 6, A and B). FZD1 and FZD8 receptor expression led to up-regulation of CRT in both vector and FZC18-expressing cells. Vector and FZC18 cells expressing FZD1 exhibited similar CRT (Figure 6A). By contrast, 100 ng of FZD8 cDNA was required to reach 50% of vector cells' maximal CRT in FZC18-expressing cells, whereas vector cells' maximal activity was obtained with 5 ng of FZD8 cDNA. Interestingly, although lower amounts of FZD1 (0.5 and 0.75 ng) or FZD8 (0.5 to 10 ng) cDNAs activated CRT in a dose-dependent manner, higher amounts of FZD1 (1 to 250 ng) or FZD8 (50 to 250 ng) cDNAs gradually reduced CRT (Figure 6, A and B). Although high doses of FZ receptor cDNA appeared non-specifically inhibitory (Figures 6 and S8), mouse FZD1 and Drosophila FZ3 may behave as antagonists of canonical Wnt/β-catenin signaling [21], [22]. Excessive ectopic stimulation of β-catenin signaling via frizzled receptors may result in saturation of the signal transduction capacity of pathway components downstream to the frizzled receptors. In keeping with this hypothesis, mouse wild-type FZD1 inhibits Wnt signaling less efficiently than does C-terminally deleted mouse FZD1 [21]. These data led us to test a hypothetical synergy between FZC18 and FZ CRDs, capable of binding Wnts but unable to transduce downstream signal.

Using FZC18-expressing cells, we compared the effects of the full-length FZD8 receptor, FZD8_CRD and a FZD8_CRD carrying a C-terminal glycosylphosphatidylinositol (GPI) anchor attaching the CRD at the cell surface (FZD8_CRD-GPI). Within the 10 pg-100 ng range of transfected cDNA, FZD8_CRD-GPI and FZD8_CRD further reduced Wnt signaling in FZC18-expressing cells (Figure 6C). Under the same conditions, full-length FZD8 antagonized the effects of FZC18, increasing CRT (Figure 6C), supporting the data shown in Figure 6B. Effects of full-length FZD8 receptor, FZD8_CRD and FZD8_CRD-GPI on cells expressing the empty vector are provided on Figure S9.

These results show an additive effect of FZC18 and either FZD8_CRD or FZD8_CRD-GPI, the dose-response curve outlining a higher efficacy of FZD8_CRD-GPI. Thus, Wnt/β-catenin signaling can be concomitantly downregulated by different Wnt-binding proteins. As FZD8_CRD is soluble, its cell surface bioavailability may thus be lower than that of FZD8_CRD-GPI, probably resulting in lower Wnt inhibitory activity.

Taken together, these data underline the specificity of the Wnt inhibitory activity of FZC18. As FZD receptors rescued Wnt/β-catenin signaling, but FZD CRDs further enhanced the inhibitory effects of FZC18, these findings indicate that the effects of FZC18 may not result from endoplasmic reticulum toxicity through clogging of the secretion pathway with cysteine-rich proteins.

FZC18 forms homodimers and binds the CRDs of FZD1 and FZD8 receptors

One of the well-known features of frizzled CRDs is their capacity to form homo and heterodimers [4], conferring to SFRPs the ability to bind to frizzled receptor CRDs [5]. To investigate whether these features applied to FZC18, we cotransfected HEK293T cells with both V3Nter-V5 and FZC18-myc or with V2Nter-V5 and FZC18-myc (Figure 7, A and B). V3Nter is a precursor of FZC18 originated by endogenous proteolysis in human tissues [7]. V2Nter-V5 contains the same aminoterminal sequences of C18 as V3Nter, but lacks the FZC18 domain (Figure 7A). As both V2Nter and V3Nter share the DUF-959 domain, V2Nter was used as a negative control. Immunoprecipitation with anti-myc, followed by immunoblotting with anti-DUF-959 or with anti-V5 tag antibodies showed that FZC18 bound V3Nter but not V2Nter. This also excludes the possibility that FZC18 could bind other portions of the V3Nter molecule, such as the DUF-959 or the Tsp-1 domains (Figure 7A).

Next, we tested whether FZC18 could bind the CRDs of FZ receptors. To this end, we used soluble FZD1 or FZD8 CRDs fused to the Fc portion of human IgG (Figure 7C). Adding recombinant FZD1_CRD-Fc or FZD8_CRD-Fc CM to non-concentrated FZC18 CM and immunoprecipitating with protein G coated beads, followed by immunobloting with anti-myc antibody showed that FZC18 could bind FZD1 and FZD8 CRDs. Taken together, the results suggest a model whereby FZC18 could bind both Wnts and FZD CRDs, hampering access of Wnts to the FZD receptors, thus blocking Wnt/β-catenin pathway activation in an SFRP-like mode (Figure 8).