Oligomerization of Frizzled and LRP5/6 protein initiates intracellular signaling for the canonical WNT/β-catenin pathway

Upon binding to the canonical WNT glycoproteins, Frizzled family receptors (FZDs) and low-density lipoprotein receptor-related protein 5/6 (LRP5/6) undergo a series of polymerizations on the cell surface that elicit canonical WNT/β-catenin signaling. The hyperactivation of WNT/β-catenin signaling is the major cause of tumorigenesis, but the mechanism in tumors such as hepatoma remains unclear. Here, we observed that WNT3A manifested the hyperactivity in β-catenin–dependent signaling after binding to FZD's competitive inhibitory molecule secreted Frizzled-related protein 2 (SFRP2). To understand the mechanism of FZDs in the presence of SFRP2, we explored how FZDs can bind and activate the LRP5/6 signalosome independently of WNT glycoproteins. Our findings further revealed that oligomerizations of FZDs and LRP5/6 can integrate the cytoplasmic protein Dishevelled into the LRP5/6 signalosome, resulting in a robust activation of ligand-independent β-catenin signaling. We propose that besides WNT-bridged FZD–WNT–LRP5/6 protein complexes, the homo- and hetero-oligomerizations of WNT receptors may contribute to the formation of the LRP5/6 signalosome on the cell surface. Of note, we identified four highly expressed FZDs in the hepatoma cell line HepG2, all of which significantly promoted ligand-independent LRP5/β-catenin signaling. As FZDs are ectopically expressed in numerous tumors, our findings may provide a new perspective on tumor pathologies. Furthermore, the results in our study suggest that the composition and stoichiometry of FZDs and LRP5/6 within the LRP5/6 signalosome may tune the selection of bound WNT glycoproteins and configure downstream WNT/β-catenin signaling.

Signaling elicited by the WNT family of secreted glycoproteins is highly conserved in evolution, innervating cell prolifer-ation, cell specification, and stem cell renewal during embryonic development and tissue homeostasis (1,2). In most mammals, including humans, WNT family proteins consist of 19 members and are divided into canonical and noncanonical subfamilies. Upon binding of a canonical WNT ligand to lowdensity lipoprotein receptor-related protein 5/6 (LRP5/6) 3 and Frizzled family receptors (FZDs), intracellular modulators Dishevelled (DVL) and AXIN1 are recruited to the plasma membrane through the intracellular domains of FZDs and LRP5/6, resulting in repression of a "destruction complex" formed by AXIN1, glycogen synthase kinase 3␤ (GSK3␤), and adenomatous polyposis coli protein (APC); hence, ␤-catenin proteins escape from N-terminal phosphorylation and subsequent proteasome-dependent degradation. Thereafter, stabilized ␤-catenin proteins enter the nucleus and facilitate transcription of WNT target genes via partnering with T-cell factor/lymphoid-enhanced binding factor 1 family transcription factors (3). Because of the pivotal role of WNT/␤-catenin signaling in tumorigenesis and other pathological processes (1,4), the molecular mechanisms underlying this pathway have been extensively investigated over the past two decades.
A prominent feature of WNT/␤-catenin signaling is the requirement for both FZDs and LRP5/6; however, the mechanism underlying the formation of FZDs-LRP5/6 coreceptor complexes remains ambiguous. The paradigm of WNTbridged FZD-WNT-LRP5/6 trimeric protein complex which has long been considered as an ideal model for the initiation of WNT/␤-catenin signaling on the cell surface (1,5) is now insufficient to explain the architecture of large-sized LRP5/6-FZDs signalosome crucial for WNT/␤-catenin activation (6). There have been a large number of studies to illustrate how the LRP5/ 6-FZDs signalosome is formed. For example, the oligomerization of AXIN1 or DVL has been described as the key molecular basis underlying signalosome formation (7,8). At the same time, the structural analyses of WNT glycoproteins as well as their receptors also provide interesting information regarding the establishment of LRP5/6 -FZD signalosome. The 3.25-Å structure of Xenopus Wnt8 in complex with mouse FZD8 cysteine-rich domain (CRD) revealed that there may be double binding sites on Wnt8 to grasp the FZD8 CRD (9), raising the possibility of more than one Frizzled receptor binding to one WNT glycoprotein. In contrast, LRP6 appears to engage different subsets of WNT glycoproteins via different ␤-propeller modules, similarly implying the potency of LRP6 simultaneously binding to two WNT glycoproteins (10 -12). Hence, these observations strongly suggest that the arrangement of the LRP5/6 signalosome may be variable rather than a simple oligomerization of trimeric FZD-WNT-LRP5/6 protein complex. In addition, during signalosome formation on the cell surface, whether or not the simultaneously binding of WNT glycoproteins to LRP5/6 and FZDs is essential is still unknown. In fact, there is still a lack of sufficient biochemical and structural evidence for simultaneous binding of a WNT ligand to a Frizzled receptor and an LRP5/6 receptor. Meanwhile, it is also worth noting that the large-sized LRP6 signalosome can also be induced by APC loss in tumor cells regardless of the presence or absence of WNT glycoproteins (13).
Recently, the function of FZDs in the LRP5/6 signalosome has been challenged as several studies have demonstrated that the secreted Frizzled-related proteins (SFRPs), the conserved inhibitors competing with FZDs to bind to WNT glycoproteins, can promote, but not inhibit, ligand-dependent ␤-catenin signaling. These observations contradict the established mechanism for FZDs in the canonical WNT pathway. Unlike LRP5/6 proteins, which appear to possess intrinsic capability of inducing intracellular ␤-catenin signaling (14,15), FZDs themselves barely activate the canonical WNT pathway. In structure, FZDs resemble G protein-coupled receptors (GPCRs) (16,17); however, the mechanism involved in G protein-and GPCR-dependent intracellular signaling still remains controversial. Therefore, except for the CRD crucial for binding to WNT glycoproteins, the role of other structure modules on FZDs in the formation of the LRP5/6-FZD signalosome and subsequent signaling transduction is not fully understood.
Here, we confirmed that WNT3A could induce ␤-catenin signaling more strongly after binding to SFRP2. Although the binding of FZD5 to WNT3A was inhibited by SFRP2, the regulation of FZD5 in the canonical WNT/␤-catenin pathway appeared to be unaffected by the presence of SFRP2. A previous report has demonstrated that the homo-oligomerization of LRP6 via LDLR repeats is required for the canonical WNT/␤catenin signaling (18). In the current study, we further revealed that the hetero-oligomerization of FZD5 and LRP5 could induce robust activation of ␤-catenin signaling in a ligand-independent manner and that the oligomerization between the FZD5 extracellular N terminus and the LDLR repeats of LRP5 was critical for this ligand-independent mechanism. Moreover, we also observed the homo-oligomerization of FZD5, which may also be involved in initiation of canonical WNT signals on the cell surface. Furthermore, the state of receptor oligomerization within the LRP5/6 signalosome may be involved in the configuration of the extracellular and intracellular signaling axis for canonical WNT/␤-catenin signaling transduction. Therefore, these findings provide novel insight into the architecture of the LRP5/6 signalosome and the underlying mechanisms for fine-tuning the signaling axis according to physiological or pathological changes of cells or tissues.

Competitively binding of SFRP2 to WNT3A leads to hyperactivation of ␤-catenin signaling
As the SFRPs resemble the ligand-binding CRD of the FZD family and thereby can compete with FZDs for binding to WNT glycoproteins, they have been identified as the largest family of secreted WNT inhibitors. However, a number of recent studies demonstrate that SFRPs, such as SFRP1 and SFRP2, can also act as agonists in canonical WNT/␤-catenindependent physiologies and pathologies (19 -22). To determine the mechanism underlying SFRP2 regulation, purified WNT3A proteins were preincubated with an excess amount of hemagglutinin (HA)tagged SFRP2-conditioned medium (CM) for 6 h and then subjected to coimmunoprecipitation (co-IP) and ␤-cateninresponsive TOPFlash reporter assays, respectively (Fig. 1A). The result of co-IP experiments showed that ϳ90% of input WNT3A proteins was precipitated by SFRP2-HA immobilized on beads (Fig. 1B), indicating that WNT3A proteins were nearly completely embedded by SFRP2-HA molecules. At the same time, WNT3A proteins embedded by SFRP2 exhibited a much stronger agonistic ability than free WNT3A proteins in the ␤-catenin-responsive TOPFlash reporter assays (Fig. 1C). In addition, we also confirmed this agonistic activity in other cells such as U2OS osteosarcoma cells and C2C12 pluripotent precursor cells and observed similar regulation of SFRP1 in canonical WNT/␤-catenin signaling (data not shown). By contrast, when cells were pretreated with vehicle or SFRP2 CM for 6 h and then CM was withdrawn, the subsequent WNT3Astimulated ␤-catenin signaling in both vehicle-and SFRP2-pretreated cells was comparable (Fig. 1D). The immunofluorescence against ␤-catenin proteins validated that the presence of SFRP2 could really enhance WNT3A-mediated stabilization of intracellular ␤-catenin proteins (Fig. 1E). Moreover, the presence of SFRP2 significantly accelerated WNT3A-mediated accumulation of cytosolic ␤-catenin proteins (Fig. 1F) and enhanced the sensitivity of cells responding to WNT3A (Fig.  1G). Taking these results together, we conclude that SFRP2 elicits the enhanced WNT3A-dependent ␤-catenin stabilization by competitively binding to WNT3A rather than other cell surface molecules.

SFRP2 regulates the paracrine activity of WNT glycoproteins
The binding of WNT glycoproteins to FZDs is essential to canonical WNT/␤-catenin signaling. Hence, several indirect mechanisms have been proposed to explain the unexpected agonistic activity of SFRP1/2, for example via forming a complex with FZDs, inhibiting noncanonical WNT signaling, or promoting the spread of WNT glycoproteins among tissues (23)(24)(25). Here, we showed that SFRP2 did bind to WNT3A with a higher affinity ( Fig. 2A) and could inhibit the interaction of WNT3A with the CRD derived from the N terminus of Frizzled-5 (FZD5-N) (Fig. 2B). The result that SFRP2 suppresses FZDs activate LRP5/6 signalosome independently of WNT but does not participate in WNT3A-FZD5-N protein complexes seems to rule out the possibility of SFRP2-WNT3A-FZD trimeric complexes. Interestingly, we further detected a high ratio (Ͼ2.0) of WNT3A:SFRP2 in SFRP2-WNT3A coimmunoprecipitation complexes (Fig. 2C), suggesting that the SFRP2-WNT3A complex may not be composed in a one-toone manner. There have been reports that the C-terminal netrin-related motif (NTR) but not the CRD of SFRP1 is also capable of binding to Wingless (26). However, our result clearly showed that the CRD but not the NTR of SFRP2 is responsible for binding to WNT3A (Fig. 2D). Furthermore, deletion of the CRD completely abolished SFRP2-mediated enhancement in WNT3A/␤-catenin signaling (Fig. 2E). In addition, although the NTR could not bind to WNT3A, its absence impaired SFRP2 regulation, but only moderately (Fig. 2E). In conclusion, these results support that CRD-dependent competitive binding to WNT3A is pivotal in SFRP2 regulation. Next, we used the protein-serine O-palmitoleoyltransferase porcupine (PORCN) inhibitor LGK974 (27) to block the production of noncanonical WNT glycoproteins. Fig. 2F shows that transfected WNT11mediated suppression of WNT3A/␤-catenin signaling was well restored by LGK974 treatment; however, the activity of WNT3A or SFRP2-embedded WNT3A could not be counteracted by LGK974 treatment, strongly ruling out potential inhibition of endogenous noncanonical WNT glycoproteins in SFRP2 regulation. Unlike R-spondin 1, which can facilitate both autocrine and paracrine activities of WNT3A glycoproteins, SFRP2 could only act on paracrine WNT3A proteins (Fig. 2G). To further confirm this finding, cells were transfected with WNT1 or WNT3A plasmids and with TOP- Figure 1. Competitive binding of SFRP2 to WNT3A leads to hyperactivation of ␤-catenin signaling. A, schematic diagram showing the experimental procedure for determining the activity of SFRP2-embedded WNT3A proteins. Purified WNT3A proteins were premixed with vehicle (VC) or SFRP2 CM for 6 h and then subjected to coimmunoprecipitation or reporter assays, respectively. B, coimmunoprecipitation experiments determined the percentage of SFRP2embedded WNT3A proteins. Western blot image is a representative of three independent experiments, and the percentage shown in the graph was measured by calculating the ratio of WNT3A proteins in the immunoprecipitations to the input. Error bars represent S.D. C, HEK293T cells transfected with TOPFlash were stimulated with free or SFRP2-embedded WNT3A for 6 h, and then luciferase activity was examined. TOP, ␤-catenin-responsive luciferase reporter TOPFlash. Error bars represent S.D. D, the activity of WNT3A/␤-catenin signaling was determined in vehicle-or SFRP2 CM-pretreated cells. HEK293T cells transfected with TOPFlash were pretreated with vehicle or SFRP2 CM for 6 h, and then CM was withdrawn. The pretreated cells were subsequently stimulated by WNT3A for an additional 6 h. Error bars represent S.D. E, immunofluorescence experiments with a ␤-catenin antibody examined intracellular ␤-catenin accumulation. F, SFRP2 accelerated the accumulation of cytosolic ␤-catenin (␤-cat) stimulated by WNT3A. The cytosolic ␤-catenin proteins were extracted as described under "Experimental procedures." G, SFRP2 enhanced the sensitivity of cells responding to WNT3A. Error bars represent S.D. *, p Ͻ 0.05; **, p Ͻ 0.01; ns, not significant; Student's t test; n ϭ 3. ␤-tub, ␤-tubulin; DAPI, 4Ј,6-diamidino-2-phenylindole.

FZDs activate LRP5/6 signalosome independently of WNT
Flash reporter separately. The reporter assays showed that both paracrine WNT1 and WNT3A could be similarly enhanced by SFRP2 supplement (Fig. 2H). Therefore, these results demonstrate that SFRP2 regulates the paracrine activity of WNT glycoproteins.

Ligand-independent oligomerization of FZD5 and LRP5 activates ␤-catenin signaling
Given that SFRP2 competes with FZDs to bind to WNT glycoproteins, we then set out to investigate whether the signaling transduction by LRP5/6 or FZDs can be affected by SFRP2. . SP, signal peptide; TM1-7, transmembrane helices 1-7; C-tail, carboxyl tail. B, SFRP2 suppressed the association of FZD5-N with WNT3A. C, the binding ratio of WNT3A to SFRP2 was determined in SFRP2 immunoprecipitation complexes. Briefly, an excessive dose of WNT3A proteins was preincubated with FLAG-tagged SFRP2 for 6 h, and then SFRP2 was pulled down by FLAG M2 antibodies. The standard protein concentration gradients of FLAG-tagged GST-fused proteins and WNT3A (25, 50, and 100 fmol) were used in Western blotting. D, coimmunoprecipitation experiments showed that the CRD was sufficient to bind to WNT3A proteins. E, the lack of the CRD completely and the lack of the NTR domain weakly inhibited SFRP2 promotion in WNT3A/␤-catenin signaling. Error bars represent S.D. F, inhibition of endogenous noncanonical WNT signaling did not counteract SFRP2 promotion in WNT3A/␤-catenin signaling.

FZDs activate LRP5/6 signalosome independently of WNT
SFRP2 could not regulate LRP5 itself but significantly promoted ligand-dependent LRP5/␤-catenin signaling (Fig. 3A). Moreover, SFRP2 could also augment WNT3A-stimulated LRP6 phosphorylation (Fig. 3B). Interestingly, although the binding of FZD5 to WNT3A could be inhibited by SFRP2 (Fig.  2B), FZD5-mediated WNT3A/␤-catenin signaling was not perturbed by SFRP2, which unexpectedly exhibited a superposition effect on WNT3A/␤-catenin signaling rather than the mutual repulsion in the presence of SFRP2 (Fig. 3C). Therefore, perturbation of the binding of FZDs to WNT3A by SFRP2 seems not to interfere with signaling transduction modulated by both LRP5/6 and FZDs. We therefore speculate that, in addition to binding to the canonical WNT glycoproteins, there may be other mechanism for FZDs to participate in the activation of the LRP5/6 signalosome. Indeed, overexpression of LRP5 along with FZD5 resulted in a much stronger ability to activate ␤-catenin signaling than LRP5 alone, but DKK1 had no effect (Fig. 3D), suggesting that FZD5-regulated LRP5 activation is independent of the binding of WNT glycoproteins to LRP5. To further exclude the potential regulation of endogenous canonical WNT glycoproteins, we treated cells with PORCN inhibitor LGK974. The activities of the transfected WNT1 or FZDs activate LRP5/6 signalosome independently of WNT WNT3A plasmids, but not WNT3A proteins, were completely suppressed, indicating that LGK974 can efficiently block the production of canonical WNT family glycoproteins (Fig. 3E). However, when LGK974 was applied to FZD5-activated LRP5/ ␤-catenin signaling, it did not show inhibition like that in cells transfected with WNT1 or WNT3A (Fig. 3E). In addition to FZD5, other members of Frizzled receptors, such as FZD3 and FZD4, could also activate the LRP5 signalosome (Fig. 3F). Taking these results together, we conclude that FZDs can activate LRP5/␤-catenin signaling independently of the canonical WNT glycoproteins.

The oligomerization of FZD5 and LRP5 is regulated by multiple sites
To further access the molecular basis for the interaction of FZD5 and LRP5 in the LRP5/6 signalosome, we first mapped the critical amino acid sequences responsible for FZD5-LRP5/6 interaction. FZD5 is composed of three parts: the extracellular N terminus containing a CRD and a linker sequence between CRD and the first transmembrane helix, six extracellular and intracellular loop sequences, and a cytosolic tail of ϳ60 amino acid residues. Coimmunoprecipitation experiments revealed that the FZD5 extracellular domain (FZD5-ECD) was sufficient for binding to LRP5 or LRP6 (Fig. 4A). Moreover, the interaction of FZD5-ECD with LRP5 and LRP6 was not regulated by the CRD but instead by the linker sequence between CRD and first transmembrane helix (Fig.  4B). The extracellular domain of LRP5/6 possesses four ␤-propeller motifs that alternate with four epidermal growth factor-like repeats (P1E1-P4E4), which are followed by three LDLR type A repeats. Coimmunoprecipitation experiments demonstrated that P1E1-P4E4 of LRP5 are dispensable for the interaction of LRP5 with FZD5-ECD, but LDLR repeats are required (Fig. 4C). It has been reported that P1E1-P4E4 of LRP6 bind to the extracellular loop sequences of FZD8 by which LRP6 can suppress noncanoni-

FZDs activate LRP5/6 signalosome independently of WNT
cal WNT/c-Jun N-terminal kinase signaling in metastasis cancer cells (29). Here, we also confirmed that the loop and cytosolic tail modules (FZD5-loop-tail) could still bind to LRP5, but the affinity seemed to be weaker than that of FZD5-ECD (Fig. 4D). Taking these results together, we conclude that the FZD5-LRP5 interaction is a multisite-mediated protein interaction.

DVL family proteins are required for ligand-independent FZD-LRP5/6 axis
To investigate the regulatory role of FZD5-LRP5 interaction, we introduced different truncated mutants of FZD5 or LRP5 into HEK293T cells and examined their abilities in ligand-independent ␤-catenin signaling. Compared with WT LRP5, both LRP5-⌬PE (P1E1-P4E4 repeat deletion) and LRP5-⌬PE-LDLR (extracellular domain deletion) possess strong constitutive activity (Fig. 5A); these mutants have been previously reported to be desensitized to the canonical WNT signals (14,15). However, FZD5 could activate LRP5-⌬PE but not LRP5-⌬PE-LDLR (Fig. 5A), demonstrating that the interaction of LDLR repeats with FZD5, but not the interaction of P1E1-P4E4 repeats with FZD5, is required for the ligand-independent FZD-LRP5/6 signaling axis. Meanwhile, deletion of the entire FZDs activate LRP5/6 signalosome independently of WNT N terminus, but not just the CRD, abolished FZD5 regulation in LRP5/␤-catenin signaling (Fig. 5B), further confirming the important role of the linker sequence between the CRD and the first transmembrane helix in FZD5-mediated LRP5 signalosome activation. The interaction between the P1E1-P4E4 repeats of LRP5 and the extracellular loops of FZD5 appears to be dispensable for FZD5 regulation in LRP5/␤-catenin signaling as LRP5 lacking the P1E1-P4E4 repeats could still be activated by FZD5 (Fig. 5A). However, the loop and carboxyl tail of FZD5 could still contribute to the FZD5-LRP5 signaling axis (Fig. 5B), demonstrating that the intracellular signaling axis triggered by these modules of FZD5 may be required for LRP5 signalosome activation.
DVL family proteins act as the key transducer of the canonical WNT/␤-catenin pathway, and they can be recruited to the plasma membrane via binding to the intracellular loop and carboxyl tail of FZDs (5,30). Immunofluorescence images showed that overexpression of FZD5 as well as FZD5-⌬CRD and FZD5-⌬ECD did recruit DVL3 to the plasma membrane, but FZD5-⌬loop-tail could not (Fig. 5C). Although LRP5 and DVL3 themselves could constitutively activate ␤-catenin signaling independently of WNT glycoproteins, they cooperated poorly until coexpressed with FZD5 (Fig. 5D). Importantly, FZD5 lacking the entire N terminus (Fig. 5D), not just the CRD (Fig. 5E), failed to regulate LRP5 and DVL3 collaboration. In addition, FZD5 lacking its loop and carboxyl tail could not regulate such collaboration (Fig. 5F). Of note, the collaboration of myristoylated DVL3, which can be constitutively translocated to the plasma membrane, and LRP5 could not be regulated by FZD5-⌬loop-tail mutant (Fig. 5, G and H). Taking these results together, we concluded that FZD5 facilitates ligand-independent LRP5/6 signalosome activation probably via two key steps. First, the intracellular loop and carboxyl tail of FZD5 binds and recruits DVL family proteins to the plasma membrane. Second, the extracellular N terminus of FZD5 binds to the LDLR repeats of LRP5 and then integrates DVLs and LRP5/6 together. Consistent with this hypothesis, CRISPR/Cas9-mediated triple knockout of DVL family proteins led to a complete suppression of FZD5-activated LRP5/␤-catenin signaling (Fig. 5I). By contrast, the constitutive activity of LRP5, including WT and with P1E1-P4E4 deletion, was not dependent on DVL family proteins. Hence, FZD5-facilitated ligand-independent LRP5/6 signaling may still adopt the common signaling axis of the canonical WNT pathway.

Ligand-independent FZD-LRP5/6 signaling may be involved in tumorigenesis
Hyperactivation of canonical WNT/␤-catenin signaling caused by hereditary or incidental mutations contributes to the occurrence of most tumors (1,3,31). However, mutations giving rise to the loss of function of APC and AXIN1 crucial for degrading ␤-catenin or the gain of function of CTNNB1 occur only in a few tumor cells. In the present study, we demonstrate that the interaction of FZDs with LRP5/6 can induce robust activation of ligand-independent ␤-catenin signaling, thus providing an alternative mechanism for activating WNT/␤catenin signaling in tumor cells. Using quantitative RT-PCR and protein Western blotting, we observed that the total expression levels of LRP5 and LRP6 were roughly equivalent in various tumor cells (Fig. 6, A-C). However, the expression of Frizzled receptors showed evident tumor-context specificity. Quantitative RT-PCR analyses indicated that FZD1, FZD2, and FZD3 were ubiquitously expressed in most tumor cells examined here, whereas the expressions of other Frizzled receptors appeared to be more distinct but random (Fig. 6D). For example, two kinds of cancer cells with different metastatic abilities, Saos2 and U2OS osteosarcoma cells and MCF7 and MDA-MB-231 mammary cancer cells, exhibited no significant difference in expression levels and members of FZD family; however, PC3 prostate cancer cells harbored prominent expression of FZD8 (32) compared with LNCAP cells, which have low metastatic ability (Fig. 6D). Interestingly, HepG2 hepatoma cells expresses four Frizzled receptors, FZD3/4/5/9 (Fig. 6D), all of which could strongly regulate the collaboration of DVL3 and LRP5 to activate ␤-catenin signaling (Fig. 6E) and recruit DVL family proteins to the plasma membrane (Fig. 6F). One of these FZDs especially, FZD9, could dramatically activate the LRP5 signalosome dependent on DVL family proteins (Fig. 6, E and G). Hyperactivation of WNT/␤-catenin signaling has been considered as the major cause for hepatoma genesis; however, the mechanisms still remain unclear (31,33). In fact, the destruction complex formed by AXIN1, GSK3␤, and APC is still functional in HepG2 cells as cytosolic ␤-catenin proteins were dramatically stabilized when treated with the GSK3␤ inhibitor CHIR99021 (Fig. 6H). However, cytosolic ␤-catenin stabilization could not be suppressed by DKK1 (Fig. 6H), implying that the endogenous canonical WNT glycoproteins were not involved in suppressing GSK3␤-AXIN-APC complexes. Therefore, we concluded that ectopic expression of Frizzled receptors (34,35), such as FZD3/4/5/9, which cause ligand-independent LRP5/6 signalosome activation, might be the putative pathological factor for the constitutive activation of WNT/␤-catenin signaling in HepG2 cells.
In fact, in addition to the proposed oligomerization between FZDs and LRP5/6, LRP6 is capable of dimerizing, which has been reported to be required for the initiation of canonical WNT/␤-catenin signaling on the cell surface (18). Therefore, these findings prompted us to hypothesize that the oligomerization of FZDs and LRP5/6 may establish the fundamental large-sized LRP5/6 signalosome. The increase in FZD5 protein content, which can alter the stoichiometry of FZDs and LRP5/6 on the cell surface and thus appears to lead to more WNT-
Interestingly, the receptor complex with LRP5-predominant stoichiometry appeared to favor the paracrine activity of the WNT1 subtype glycoproteins consisting of WNT1, WNT2, WNT2B, and WNT6 (Fig. 7H), which have been suggested to bind to the P1E1 and P2E2 repeats of LRP5/6 (10 -12). By contrast, WNT3A and WNT3, which bind to the P3E3 and P4E4 repeats of LRP5/6 (12), could be used to form both FZD5-and LRP5-dominant receptor complexes, but they may prefer the FZD5 preponderance (Fig. 7H). Taking into account the above results, we conclude that the state of receptor oligomerizations may affect the recognition and binding of the LRP5/6 signalosome to different subtypes of canonical WNT glycoproteins. A recent study demonstrated ligand-independent LRP5/6 signaling in APC-deficient tumor cells where clathrin-instead of caveolin-mediated endocytosis is required (13). It has been reported that clathrin-and caveolin 1-mediated endocytosis

FZDs activate LRP5/6 signalosome independently of WNT
are both required for the canonical WNT/␤-catenin pathway (36 -38). To illustrate whether the state of receptor oligomerization affects the preference of the canonical WNT pathway for endocytic pathways, targeted disruption of clathrin heavy chain 1 (CLTC) was established in HEK293T cells (Fig. 8A). Consistently, the ␤-catenin signaling elicited by APC knockdown but not AXIN1/2 double knockdown was impaired in CLTC-deficient cells (Fig. 8, A and B). Although the basal and constitutive activity of LRP5 was not regulated by CLTC (Fig. 8C), FZD5-activated LRP5/␤-catenin signaling could be significantly suppressed by CLTC deficiency (Fig. 8D). Moreover, WNT1-or WNT3A-stimulated activity of overexpressed LRP5 was unaffected by CLTC knockout (Fig. 8, E and F) but could be inhibited by caveolin 1 (CAV1) knockdown (Fig. 8G). Of note, both caveolin 1-and clathrin-mediated endocytosis appeared to be dispensable for ␤-catenin signaling stimulated by WNT3A and overexpressed FZD5 (Fig. 8, G and H). Taking these results together, we conclude that receptor oligomerizations can con-

Discussion
WNT-bridged FZD-WNT-LRP5/6 trimeric complexes have long been considered as an ideal model to describe the initiation of canonical WNT/␤-catenin signaling. However, the discovery of large-sized LRP5/6 signalosomes consisting of FZDs, AXIN1, and DVL (6) excludes the monomeric mode of a FZD-LRP5/6 coreceptor complex on the cell surface. There have been a large number of models illustrating how the LRP5/6 signalosome is orchestrated, for example those involving oligomerization of AXIN1 or DVL (6 -8). Nonetheless, the details of LRP5/6 signalosome arrangement are still unclear. Notably, large-sized LRP5/6 signalosomes can also be induced by APC loss in a ligand-independent manner. Therefore, a comprehensive understanding of the mechanism underlying LRP5/6 signalosome assembly will help to elaborate cell context-specific WNT/␤-catenin signaling. Here, we surprisingly observed that Frizzled receptors could activate the LRP5/6 signalosome in a ligand-independent manner via receptor hetero-oligomerizations and further propose that the receptor oligomerizations may facilitate LRP5/6 signalosome formation and configure the downstream signaling axis for ␤-catenin stabilization.
Although FZDs are essential for WNT glycoprotein binding to the LRP5/6 signalosome, the traditional role of FZDs in canonical WNT/␤-catenin signaling is now being challenged as SFRP1/2, which competes with FZDs to bind to WNT ligands, has been recently observed to activate ligand-dependent ␤-catenin signaling (19 -22). In the present study, the cooperation of SFRP2 and overexpressed FZD5 in transducing WNT3A/␤-catenin signaling prompted us to explore the unexpected role of oligomerization between FZD5 and LRP5/6 in ligand-independent ␤-catenin signaling pathway. Moreover, the oligomerization of Frizzled family receptors with LRP5/6 may be a universal mode on the cell surface. With the exception of FZD5, other FZDs, including FZD3, FZD4, FZD8, and especially FZD9, can also significantly promote LRP5/␤-catenin signaling in a ligand-independent manner. Because ectopic expression of FZDs in tumor cells is a common phenomenon (34,35), these findings may provide an alternative mechanism to comprehensively understand the molecular basis underlying tumorigenesis.
Importantly, FZDs activate ligand-independent LRP5/6 signalosomes, which still require the intracellular components of canonical WNT/␤-catenin signaling. For example, AXIN1 and GSK3␤ may promote the oligomerization of FZD5 and LRP5, and the presence of DVL family proteins can dramatically enhance FZD5-mediated LRP5/␤-catenin signaling. Interestingly, the recruitment of DVL family proteins to the LRP5/6 signalosome requires not only FZD intracellular module transport of DVL proteins to the plasma membrane but also interaction of the FZD extracellular N terminus with the LDLR repeats of LRP5/6. Therefore, the arrangement of the LRP5/6 signalosome requires the coordination of both extracellular and intracellular regulations. Taking into account the homo-and hetero-oligomerization of FZDs and LRP5/6, it is worth noting that the state of receptor oligomerization, such as the composition and stoichiometry of FZDs in the LRP5/6 signalosome, may vary in different types of cells or tissues (39) as the expression levels of FZDs and LRP5/6 are cell context-specific. Even in the same cell, the asymmetrical distribution of FZDs on the cell surface also results in mixed LRP5/6 signalosomes with different FZD composition and stoichiometry. More interestingly, we further demonstrate that the state of receptor oligomerization may determine the extracellular and intracellular signaling mechanisms. First, different subtypes of WNT glycoproteins appear to recognize and bind to the LRP5/6 signalosome with a distinct stoichiometry of FZDs and LRP5/6. Second, ligand-independent signaling facilitates clathrin-mediated endocytosis to activate intracellular ␤-catenin signaling, whereas canonical WNT signals via LRP5-predominant receptor complexes may favor caveolin-mediated endocytosis. In addition, canonical WNT signals via FZDpredominant receptor complexes do not require any regulation originating from endocytosis.
Due to cell context specificity, pathological ectopic expression of FZDs and LRP5/6, or dysfunctions of molecules like APC, there may be multiple modes of signaling initiation and transduction in the canonical WNT/␤-catenin pathway depending on the oligomerization of FZDs and LRP5/6 ( Fig. 8I). In addition, these multiple modes may also coexist in the same cells. Overall, our presented model provides a new perspective for understanding the complexity and diversity of canonical WNT/␤-catenin signaling transduction and an alternative mechanism for tumor pathologies.

Reporter assays
HEK293T cells were transiently transfected with TOPFlash reporter and GFP using Lipofectamine LTX from Life Technologies according to the manufacturer's instructions. LRP5 (15 ng/well), FZDs (10 ng/well), DVL3 (25 ng/well), and WNTS (25 ng/well) were transfected as indicated; the LacZ plasmid was added to make the total amount of DNA equal (0.25 g of total plasmids/well in a 48-well plate). Luciferase reporter assays were performed, and the luciferase activities presented were normalized against the levels of GFP expression as described previously (40).

Gene transcription analysis
Total RNAs were extracted from cultured cells by directly adding the TRIzol reagent to the cells. cDNAs were synthesized using the ProtoScript II First Strand cDNA Synthesis kit (New England Biolabs, Ipswich, MA). Quantitative real-time PCR analysis was performed using Power SYBR Green PCR Master Mix (Life Technologies). The following primer pairs were used: human GAPDH: forward, ctt tgg tat cgt gga agg act ca; reverse, ttc ccg ttc agc tca ggg atg a; CAV1: forward, tta ctt cgc cat tct ctc tt; reverse, agt tga tgc gga cat tgc t; LRP5: forward, aac gtg gtc atc tcc ggc ctg gtc tct; reverse, ggg gtc caa ggc gat ggc cct cgg ct; LRP6, forward, gca tgt gat tgg ctt gga gaa aaa tt; reverse, cac ttc tcc cca gtc tgt cca gta ca; FZD1: forward, ggc tgc acc atc ctc ttc atg at; reverse, gat ggt ctt gat ggc cgg cac a; FZD2: forward, gct tcg tgt cgc tct tcc gca t; reverse, acg agc gct ccc agt gct cgc gg; FZD3: forward, cat ccc tgc aca ata taa ggc tt; reverse, ttc ttc tca ata gct tca cta c; FZD4: forward, gga tgt gca ata att ttc ttg ct; reverse, aat ggt ttt cac tgc ggg gat g; FZD5: forward, tcg cca cct tct gga tag gcc tg; reverse, cat ggc cca cga cca gac gca c; FZD6: forward, ttc tgg ggg aca agg ata taa gt; reverse, aat ctt cta aca tca att aaa a; FZD7: forward, gac gct ctt tac cgt tct cac ct; reverse, acc gtg cgg tag cca tcg tcc g; FZD8: forward, tcg ccg gct act cgc agt act t; reverse, aag agg tag atg acc agc ggc g; FZD9: forward, tga cgc tca cct ggt tcc tgg ct; reverse, ccg tgc tgg cca cgt agc aaa g; FZD10: forward, caa cat gga tta ctg gaa gat c; reverse, gtc ttg gag gtc caa atc cac at.

Protein complex assays
Co-IP experiments were carried out in HEK293T cells. Briefly, cells were transfected with the indicated constructs for 24 h and then lysed in a cell lysis buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Triton X-100, 0.5 mM EDTA, protease inhibitor mixture (Roche Applied Science), and phosphatase inhibitor mixture (Roche Applied Science)). After centrifugation, the supernatants were mixed with the indicated antibodies for 1 h at 4°C, and the Protein A/G Plus-agarose was added for an additional 1 h. The beads were washed three times, resuspended in 30 l of SDS loading buffer, and then analyzed by Western blotting.

Immunofluorescence
Cells on coverslips were washed once with Dulbecco's phosphate-buffered saline (DPBS) and then fixed for 20 min in DPBS containing 4% paraformaldehyde at room temperature. Fixed cells were permeabilized by 0.1% Triton X-100 for 5 min and then blocked by 2% BSA for 30 min. Finally, cells were stained with primary antibodies followed by FITC-conjugated secondary antibodies. Immunofluorescence images were captured using Leica SP5.

Assays of ␤-catenin stabilization
HEK293T cells were stimulated with WNT3A, SFRP2, or in combination for indicated hours. Then the cells were harvest in cold DPBS, and centrifuged. The cell pellet was re-suspended in low salt buffer (10 mM Hepes, pH 7.9, 20 mM NaCl, protease inhibitor mixture and phosphatase inhibitor mixture) swelling for 10 min on ice. The swelled cells were homogenized by a needle for 10 times, and then were centrifuged at 25000g for 30 min at 4°C. The supernatants were harvested and analyzed by western-blotting.

Knockout cell lines
To generate DVL triple knockout cell line (41) and CLTC knockout cell line, HEK293T cells were transfected with plasmids encoding Cas9 and guide RNA for 24 h, and then transfected cells were treated with puromycin (2 g/ml) for additional 24 h. Clones were selected and verified by immunoblotting. The guide RNA sequence for DVL1/2, 5Ј-cta cat tgg ctc cat cat ga-3Ј; DVL3, 5Ј-acc atg ctt caa tgg ccg gg-3Ј; CLTC, 5Ј-gat cgc cat tct agc ctt gc-3Ј.

siCAV1 design and assays
Human CAV1 siRNA targeted the following sequence: 5Ј-cac ctt cac tgt gac gaa a-3Ј. HEK293T cells were transfected with control or CAV1 siRNA using Lipofectamine RNAiMAX from Life Technologies according to the manufacturer's instructions. After 48 h of transfection, cells were reseeded for the reporter assay.

FZDs activate LRP5/6 signalosome independently of WNT Statistical analysis
We used two-tailed Student's t tests to evaluate statistical significance. p Ͻ 0.05 was considered statistically significant. All values are presented as means Ϯ S.D.