β-Arrestin is a necessary component of Wnt/β-catenin signaling in vitro and in vivo

Results

Based on previous findings on the interaction between β-arrestin and Dvl (9, 10), we aimed at characterizing that interaction in more detail. We first used as a model SN4741 cells (14), a dopaminergic cell line in which Wnt signaling has been characterized (7, 15, 16). We found that immunoprecipitation of either Myc-Dvl2 or HA-β-arrestin2, in cells ectopically expressing Myc-Dvl2 and FLAG-β-arrestin2 (Fig. 1A) or FLAG-Dvl3 and HA-β-arrestin2 (Fig. 1B), resulted in the pulldown of FLAG-β-arrestin2 and FLAG-Dvl3, respectively. Furthermore, expression of β-arrestin2-GFP and Myc-Dvl2 in SN4741 cells showed a strong colocalization of the two proteins in characteristic punctate Dvl aggregates (17, 18), as detected by immunocytochemistry and confocal laser scanning microscopy (Fig. 1C). Similar results have been obtained with Myc-Dvl2 and HA-β-arrestin2 or FLAG-β-arrestin2 (data not shown), to ensure that the association of β-arrestin with Dvl represents a true colocalization.

To narrow the interaction interface of β-arrestin2 with Dvl, we used a set of deletion mutants of FLAG-Dvl3 (19) (Fig. 2A) and FLAG-β-arrestin2 (Fig. 2B) expressed in COS-7 cells. Coimmunoprecipitation of HA-β-arrestin2 in cells overexpressing also FLAG-Dvl3 mutants revealed that HA-β-arrestin2 can interact with FLAG-Dvl3 mutants 1, 2, 5, 6, and 7 but not with 3 and 4 [supporting information (SI) Fig. 6A], which lack a region N-terminal of the PDZ domain of Dvl (Fig. 2A). Immunoprecipitation of Myc-Dvl2 resulted in the pulldown of full-length FLAG-β-arrestin2 (construct 1) as well as the deletion mutants 2, 3, 5 but not 4 (SI Fig. 6B). To confirm the weak interaction between β-arrestin2 construct 2 and Dvl2-Myc (SI Fig. 6B), we performed coimmunoprecipitation of construct 2 with XDsh-Myc (20) and found a strong interaction between them (SI Fig. 6C). Thus, our results identify the region aa163–300 as the Dvl-interacting domain on β-arrestin2.

The small portion of Dvl that was found to bind β-arrestin, N-terminal to the PDZ domain, was previously shown to contain Ser/Thr residues recognized by CK1ε (21) and PAR-1 (partitioning defective mutant in C. elegans) (20, 22). Moreover, endogenous CK1δ/ε is induced by Wnt ligands (16, 23), phosphorylates Dvl, and results in formation of active PS-Dvl (7, 16). To assess the importance of CK1-mediated phosphorylation in the Dvl–β-arrestin interaction, we treated cells with the CK1-specific inhibitor D4476 (24), which interferes with the activation of Dvl. In cells overexpressing Myc-Dvl2, treatment with D4476 reduced the formation of PS-Myc-Dvl2 (16) and reduced the interaction of Myc-Dvl2 with HA-β-arrestin2 (Fig. 2C), suggesting that phosphorylation of Dvl by CK1 increases the affinity for β-arrestin. We have shown previously (16) that expression of CK1ε in SN4741 cells promotes an even cytosolic distribution of Myc-Dvl2, whereas a kinase-dead mutant, CK1εK→R, promotes the typical punctate Dvl appearance in the cytosol. In triple-expression experiments using HA-β-arrestin2, Myc-Dvl2, and CK1ε or CK1εK→R (Fig. 2D), we found that HA-β-arrestin2 follows the localization of CK1ε–Dvl. In the presence of CK1ε, HA-β-arrestin2, Myc-Dvl2, and CK1ε were distributed evenly throughout the cytoplasm, whereas in cells expressing the kinase-dead mutant of CK1ε, the proteins adopted a punctate distribution. This finding demonstrated that β-arrestin overexpression does not affect the CK1ε-induced translocation of Myc-Dvl2 and suggests that βarrestin probably acts downstream from Dvl, phosphorylated by CK1. This finding also strengthens the biochemical data of β-arrestin interaction with both Dvl and PS-Dvl.

To examine whether endogenous β-arrestin is required for Wnt-3a signaling, we examined whether MEFs lacking (knockout, KO) β-arrestin1, β-arrestin2, or both (β-arr1KO, β-arr2KO, β-arr1/2dKO) respond to Wnt-3a stimulation (Fig. 3A and SI Figs. 7 and 8). The wild-type (WT) MEFs showed time courses and dose responses of Dvl activation similar to those reported previously for SN4741 cells (7). However, the ability of β-arr2KO or β-arr1/2dKO MEFs to induce the formation of PS-Dvl in response to Wnt-3a was delayed and severely reduced (Fig. 3 A and B and SI Fig. 7), suggesting an important role of β-arrestin in Dvl activation and PS-Dvl formation. Interestingly, complete ablation of β-arrestins resulted in a higher basal level of PS-Dvl compared with the WT, β-arr1KO, and β-arr2KO, suggesting up-regulation of yet unknown compensatory mechanisms. Moreover, endogenous β-arrestin2 seemed to be more important compared with β-arrestin1 because the defects on PS-Dvl formation were more pronounced in MEFs lacking β-arrestin2 compared with cells lacking β-arrestin1 (SI Fig. 8). Thus, our results suggest that β-arrestins are crucial components of the molecular machinery mediating Dvl activation and formation of PS-Dvl in response to Wnt-3a. Because the formation of PS-Dvl at the 2-h time point studied here is necessary for the substantial activation of β-catenin in response to Wnt-3a (7), we also investigated dephosphorylation and stabilization of β-catenin. Moreover, we also analyzed the activation and phosphorylation of Ser-1490 in LRP6, using a phosphospecific antibody (25). Notably, LRP6 phosphorylation was strongly induced by Wnt-3a treatment but was not affected by genetic deletion of arrestins (Fig. 3C). In contrast, agonist-induced β-catenin signaling in β-arr2KO and β-arr1/2dKO MEFs was almost completely abolished (Fig. 3C). β-arr1KO MEFs showed a pattern of β-catenin dephosphorylation similar to that of the Wnt-3a-stimulated WT MEFs (SI Fig. 8). Because dephosphorylated β-catenin is stabilized and leads to activation of transcription in a TCF/Lef-dependent manner, we also investigated the role of β-arrestin in Wnt-3a-induced activation of the TOPflash luciferase reporter. Although WT MEFs strongly responded to Wnt-3a stimulation with TCF/Lef reporter activity, the lack of both β-arrestins completely inhibited this response (Fig. 3D). MEFs lacking β-arrestin2 showed a partially reduced response to Wnt-3a stimulation, supporting the idea of redundancy between β-arrestin1 and β-arrestin2 in Wnt signaling. Again, depletion of β-arrestins 1 and 2 increased basal TCF/Lef signaling slightly, indicating the existence of compensatory mechanisms. Thus, our in vitro data argue for an important function of β-arrestin in Wnt signaling, in particular for the Wnt-induced formation of PS-Dvl and subsequent activation of the Wnt/β-catenin pathway.

In addition, we sought to characterize further the potential β-arrestin-dependent mechanisms upstream and downstream from Dvl. First, we investigated the role of β-arrestin in the FZD-induced redistribution of Dvl to the plasma membrane (26). Overexpression of FZD₄-GFP together with Myc-Dvl2 in both WT and β-arr1/2dKO MEFs, resulted in complete redistribution of Myc-Dvl2 from cytoplasmic punctae to the plasma membrane (SI Fig. 9). Thus, β-arrestin is not crucial for the FZD₄-induced Dvl2 translocation. Second, we analyzed the interaction of axin with β-arrestin by overexpression in COS-7 cells. As shown in Fig. 4A, a full-length axin or a deletion mutant 1–477 axin (Δaxin) lacking the DIX domain (27), which is necessary for the interaction with Dvl (28), coimmunoprecipitated with β-arrestin. Thus, β-arrestin can interact directly with axin, and such interaction is most likely not mediated by Dvl. These data suggest that a trimeric Dvl–β-arrestin–axin complex might be central for the recruitment of Dvl to axin and the subsequent dissociation of the destruction complex (27–29). To test this possibility, we transfected in Fig. 4B the indicated combinations of vectors encoding for Dvl (XDsh), β-arrestin, and full-length/Δaxin for coimmunoprecipitation. We found that β-arrestin, XDsh, and axin are precipitated together and are, thus, likely to form a trimeric complex. Interestingly, Δaxin is also present in such complexes, suggesting that it might be recruited to XDsh by β-arrestin. However, β-arrestin is not necessary for the XDsh–axin interaction, at least not in an overexpression system, because overexpressed XDsh interacts to a similar extent with axin in both WT and β-arr1/2dKO MEFs (data not shown). In addition it appears that axin could act as a stabilizing component of XDsh/β-arrestin binding because the association of XDsh and β-arrestin is stronger when axin is coexpressed (compare lanes 3 and 4 with lane 5, WB: XDsh, Fig. 4B).

To examine the importance of β-arrestin for Wnt signaling in vivo, during embryonic development, we turned to one of the most classical and robust models to study Wnt signaling, the early X. laevis embryo. The axis duplication assay, as induced by injection of XWnt-8 or DshΔDEP mRNA in the ventral blastomeres, was used to examine the role of β-arrestin overexpression in canonical signaling. The lack of effect of β-arrestin RNA injection on axis duplication under basal conditions or in blastomeres injected with XWnt-8 or DshΔDEP suggested that β-arrestin per se is not sufficient to induce Wnt/β-catenin signaling. In contrast, down-regulation of Xenopus β-arrestins (for sequence and alignment, see SI Fig. 10) with β-arrestin morpholinos (β-arrMO) showed that β-arrestin is required for Wnt/β-catenin pathway (Fig. 5 B and C) because the XWnt-8-induced axis duplication was blocked (percent axis duplication: XWnt-8, 51.0; XWnt-8 + β-arrMO, 8.5). Control injections with β-arrMO alone did not result in axis duplication (data not shown; percent axis duplication: β-arrMO, 0, n = 261). In addition, β-arrMO (Fig. 5 B and C) completely blocked DshΔDEP − (percent axis duplication: DshΔDEP, 21.2; DshΔDEP + β-arrMO, 0) as well as CK1ε-induced secondary axis formation (percent axis duplication: CK1ε, 13.6; CK1ε + β-arrMO, 1.4). However, β-arrMO did not affect secondary axis induced by β-catenin (% axis duplication: β-catenin, 44.4; β-catenin + β-arrMO, 38.6), arguing for β-arrestin acting upstream from β-catenin and downstream from CK1ε and Dvl, which is in good agreement with our in vitro data. As control for the CK1ε injection we also injected a kinase-dead mutant of CK1εK→R, which alone or in combination with β-arrMO failed to induce axis duplication (data not shown; % axis duplication, CK1ε, 0, n = 87; CK1ε + β-arrMO, 0, n = 117). Detection of activated, dephosphorylated β-catenin (ABC, Fig. 5D) and expression of the β-catenin target gene Xnr3 (Fig. 5E) in control and β-arrestin depleted Xenopus embryos further showed that β-catenin activity is reduced in β-arrestin-depleted embryos. In summary, the Xenopus experiments demonstrate that endogenous β-arrestin is required for normal development and for Wnt/β-catenin signaling in vivo.

Discussion

Here, we identify the multipurpose scaffolding protein β-arrestin as a necessary component in the Wnt/β-catenin pathway in vitro and in vivo. We show that β-arrestin binds to and forms a trimeric complex with axin and Dvl. Moreover, endogenous β-arrestin does not affect phosphorylation of LRP6 but is necessary for proper activation of Dvl, signaling along the Wnt/β-catenin pathway, activation of endogenous Wnt target genes, and XWnt-8-, DshΔDEP-, or CK1ϵ-induced axis duplication in Xenopus embryos in vivo.

We mapped the domain of physical interaction between β-arrestin and Dvl (9, 10) to a region N-terminal of the PDZ domain of Dvl and aa163–300 of β-arrestin. This stretch of Dvl contains important CK1ε (21), PAR-1 (20, 22), and possibly also casein kinase 2 (CK2) (30) phosphorylation sites, which may work as potential regulators of β-arrestin–Dvl interaction. Importantly, CK1, CK2, and PAR-1 are all necessary for Wnt/β-catenin signaling (20, 31, 32), and at least CK1ε and CK2 kinase activity is induced directly by Wnts (16, 23, 33). It has been reported that the association of Dvl with β-arrestin is potentiated by CK2 phosphorylation (9), and we report herein that this interaction is reduced by CK1 inhibition, a finding that also supports this hypothesis. We also found that cells lacking β-arrestin resemble cells with CK1δ/ε knockdown (7, 16) in their inability to form PS-Dvl properly in response to Wnt. We thus suggest that the formation of the activated form of Dvl, PS-Dvl, is initiated by and depends on Dvl phosphorylation by Wnt-induced kinases, as we demonstrated previously for CK1ε (7, 16). In this model, β-arrestin recognizes phosphorylated (primed) Dvl and subsequently recruits other kinases to complete PS-Dvl formation and Dvl activation. A second function of β-arrestin that is likely to be crucial for Wnt signaling is its capacity to bind axin directly. This binding did not require the DIX domain of axin, which is necessary for the interaction of Dvl (7). This finding suggested the possibility of a trimeric complex consisting of Dvl, β-arrestin, and axin, which was confirmed by coimmunoprecipitation experiments. We hypothesize that such a trimeric complex could contribute to destabilize the degradation complex, resulting in the stabilization of β-catenin. Thus, our results suggest that the formation of a trimeric complex is an essential step in Wnt signaling that provides a molecular link between the LRP6/axin and FZD/Dvl branches of Wnt signaling after activation.

Experiments in β-arrestin-deficient MEFs suggest that β-arrestins 1 and 2 might be at least partially redundant with respect to mediating Wnt signaling and activation. Interestingly, unstimulated β-arr1/2KO MEFs contain a higher amount of PS-Dvl and higher background activity in the TOPflash reporter assay compared with WT and β-arr1KO or β-arr2KO MEFs, indicating that mechanisms exist to accomplish the formation of PS-Dvl in the absence of β-arrestin. These compensatory mechanisms become evident also regarding the obvious delay and decrease in efficacy of Wnt-3a-induced PS-Dvl formation rather than a complete block. It is possible that high concentrations of Wnt or long-term exposure may to some extent lead to PS-Dvl formation in MEFs deficient in β-arrestins, although not to the same extent as in WT MEFs. These findings show that β-arrestin is not indispensable for PS-Dvl formation and suggest that redundant mechanism(s) exist, which is also evident from the quantification of PS-Dvl/Dvl data (SI Fig. 7), indicating that absence of β-arrestin2 alone has a more pronounced effect on Dvl activation compared with cells depleted of both β-arrestins. However, β-arrestin seems to be uniquely required for proper dynamics of the PS-Dvl formation and for correct Wnt signaling in vivo, as demonstrated by the dramatic defects in β-arrestin-depleted Xenopus embryos. It should be noted, however, that although β-arrestin is a necessary component for Wnt signaling, it is not sufficient to induce PS-Dvl formation (Fig. 1A) or axis duplication in Xenopus embryos (Fig. 5). Thus, our data suggest that β-arrestin is required, but not sufficient, for Wnt signaling and that β-arrestin is recruited to Wnt signaling upon Wnt activation.

Our findings regarding the necessity of β-arrestin for Wnt signaling, together with the recognized function of β-arrestins as multifunctional adapter proteins, also open other interesting possibilities such as that β-arrestin might be necessary for other signaling pathways or that β-arrestin might mediate an interaction between Wnt and other signaling pathways (12). This assumption is further supported by structural data indicating that the domain of β-arrestin2 that binds Dvl (aa 163–300) lies in a region of the protein containing large parts of the C-terminal β-fold (β sheet X-XVI; see alignments of nonvisual vertebrate β-arrestins in SI Fig. 10). Thus, both the N and C termini of β-arrestin could remain open for interaction with other partners after Dvl binding. β-Arrestins are also known to play a very important role in clathrin-mediated endocytosis (34). In addition, blockade of endocytosis by, for example, hyperosmolar sucrose blocks Wnt-induced β-catenin signaling (35) and down-regulates Dvl (36). Combined, these findings together with the recently appreciated role of endocytosis in Wnt/Frizzled signaling (10, 35, 37–40) suggest the provocative possibility that β-arrestin-mediated endocytosis might be required for some aspects of Wnt signal transduction, an issue that remains to be explored.

In summary, we hereby identify β-arrestin as an integral component of the Wnt signaling pathway. This finding opens up the possibility that Wnts may signal through novel additional pathways, allowing extensive cross-talk with other, e.g., G protein-coupled receptor-dependent, pathways. Future studies will explore such interactions and will focus on further elucidating the role of β-arrestin in Dvl activation and the function of the trimeric complex formed by Dvl–β-arrestin–axin in Wnt signaling.

Methods

Supplementary Material