The Ubiquitin Ligase RNF220 Enhances Canonical Wnt Signaling through USP7-Mediated Deubiquitination of β-Catenin

Abstract

Wnt/β-catenin signaling plays critical roles in embryonic development and disease. Here, we identify RNF220, a RING domain E3 ubiquitin ligase, as a new regulator of β-catenin. RNF220 physically interacts with β-catenin, but instead of promoting its ubiquitination and proteasomal degradation, it stabilizes β-catenin and promotes canonical Wnt signaling. Our analysis showed that RNF220 interacts with USP7, a ubiquitin-specific peptidase, which is required for RNF220 to stabilize β-catenin. The RNF220/USP7 complex deubiquitinates β-catenin and enhances canonical Wnt signaling. Interestingly, the stability of RNF220 itself is negatively regulated by Gsk3β, which is a key component of the β-catenin destruction complex and is inhibited upon Wnt stimulation. Accordingly, the RNF220/USP7 complex works as a positive feedback regulator of β-catenin signaling. In colon cancer cells with stimulated Wnt signaling, knockdown of RNF220 or USP7 impairs Wnt signaling and expression of Wnt target genes, suggesting a potentially novel role of RNF220 in Wnt-related tumorigenesis.

INTRODUCTION

Research conducted over the past decade has shown that the ubiquitin system regulates a variety of cellular processes, including cell signaling transduction, cell cycle regulation, and gene transcription. Previous reports found that reversible modification of target proteins with ubiquitin regulates an assortment of signaling pathways, either through proteasomal degradation or by altering the activity and/or localization of constituent proteins, though there is an incomplete picture of the precise working of these activities and the constituent proteins. Ubiquitin conjugation itself is mediated via an E1-E2-E3 cascade, of which the ubiquitin ligases (E3) are primarily responsible for substrate selection (1, 2). In humans, there are around 600 known such ubiquitin E3 ligases, grouped into two major classes defined by the presence of either a HECT or RING domain (1, 3). Protein ubiquitination can be reversed by specific deubiquitinating enzymes (DUBs). Among the approximately 90 deubiquitinases encoded in the human genome, the ubiquitin-specific protease (USP) family of DUBs has been shown to play essential roles in numerous cellular processes and signaling pathways (4). For example, USP7 (also known as HAUSP) is a key regulator of the tumor suppressor p53 and plays a critical role in cancer progression (5, 6). USP7 preferentially deubiquitinates the p53 E3 ligase MDM2 and thereby destabilizes p53, although it is also able to directly bind p53 itself with a lower affinity (7).

The canonical Wnt signaling pathway (also known as the Wnt/β-catenin signaling pathway) has been found to play crucial roles in embryogenesis, homeostasis maintenance of adult tissues, and regeneration of various tissues and organs. In humans, dysregulation of this signaling pathway is correlated with tumorigenesis, congenital disorders, and degenerative diseases, among others (8, 9). The key component of this signaling pathway, β-catenin, is regulated by the ubiquitin-proteasome system. In the absence of Wnt ligand, β-catenin in the cytosol is phosphorylated by the destruction complex, which consists of the adenomatous polyposis coli (APC), axin, glycogen synthase kinase 3β (Gsk3β), and casein kinase 1α (Ck1α) (10, 11). Phosphorylated β-catenin is specifically recognized by the E3 ligase β-Trcp, which induces β-catenin polyubiquitination and the subsequent 26S proteasomal degradation (12–14). In the presence of Wnt, the destruction complex dissociates and β-catenin in its hypophosphorylated form escapes degradation, undergoes nuclear translocation, and activates Wnt target genes, including c-myc and cyclin D1 (11).

Alongside β-Trcp, several other E3 ligases that target β-catenin have also been isolated, including Siah-1, Jade-1, and c-Cbl (14–17). Siah-1, a RING-type E3 ligase, recognizes β-catenin in a phosphorylation-independent manner and targets it for degradation. Since it is induced by p53 activation, the regulation of β-catenin levels by Siah-1 is believed to link genotoxic stresses to cell cycle control (15, 16). Jade-1 and c-Cbl are both Wnt-responsive ubiquitin ligases that target β-catenin and inhibit Wnt signaling under different conditions. Unlike β-Trcp, both Jade-1 and c-Cbl are primarily nuclear localized and are thought to regulate nuclear-active β-catenin in the Wnt-on state (15–17). Moreover, nonproteolytic ubiquitination through atypical ubiquitin linkages has also been noted in β-catenin protein stability regulation. For example, Lys29- or Lys11-linked polyubiquitination of β-catenin is considered to stabilize it and promote Wnt signaling (18).

Despite the variety of findings on E3 ligases that target β-catenin, few deubiquitinating enzymes targeting β-catenin have been established. Although the deubiquitinase FAM (USP9X) is able to target and stabilize β-catenin, it has been suggested to be mainly involved in the trafficking of β-catenin and E-cadherin in epithelia (19). USP14 was also found to be overexpressed in lung adenocarcinoma and able to stabilize β-catenin; however, it was shown to work by directly targeting the upstream regulator Dishevelled (Dvl) (20, 21).

RNF220 is a RING finger domain ubiquitin E3 ligase that we identified previously (22). Here, we report that RNF220 specifically interacts with β-catenin. However, instead of ubiquitinating and destabilizing it, RNF220 promotes the deubiquitination and stabilization of β-catenin independently of its E3 ligase activity. Our results show that RNF220 works through USP7-mediated deubiquitination in this process and therefore acts as a canonical Wnt signaling enhancer.

MATERIALS AND METHODS

Cell culture, transfection, and luciferase reporter assay.

HEK293, HCT116, and SW480 cell lines were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (FBS). The HEK293 cells stably transfected with RNF220 shRNAs were screened in complete medium with 10 ng/ml puromycin (Sigma). Plasmids and small interfering RNAs (siRNAs) were transfected using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Luciferase Wnt reporter assays in HEK293, HCT116, and SW480 cells were done in 96-well plates with at least three repeats. The cells were seeded and allowed to reach 50 to 60% confluence prior to transfection. The amount of plasmid transfected per well was as follows: SuperTOPflash, 80 ng; pRL-TK,10 ng; Wnt3a, 30 ng; β-catenin,10 ng; RNF220 and all other constructs, 30 to 100 ng. pCS2 vector DNA was used to adjust the DNA amount to 200 ng per well. Firefly and Renilla luciferase activities were measured between 40 and 48 h posttransfection by using dual-luciferase assay kits (Promega). TOPflash luciferase activities were normalized to the Renilla activities.

Plasmids and reagents.

The pCMV-myc-mRNF220, pCS2-FLAG-mRNF220, and related deletion constructs were constructed as described previously (22). An RNF220-KR mutant construct was created via a site-directed mutation with PCR-driven overlap extension (23).

For knockdown of RNF220 in HEK293 cells, shRNAs against RNF220 were prepared in pSIH-H1 and lentivectors. The two shRNA-targeting sequences were as follows: shRNF220–1#, 5′-GCGUACCACCACACACAUUUA-3′, and shRNF220–2#, 5′-GCGACUUAAGAACGCCAUGA-3′. Two small interfering RNAs targeting RNF220 (siG000055182A/B; RiboBio) were used to knock down RNF220 in both HCT116 and SW480 cells.

Two small interfering RNAs targeting USP7 were synthesized according to the previously established sequences: siUSP7 1#, 5′-ACCCUUGGACAAUAUUCCU-3′, and siUSP7 2#, 5′-UGACGUGUCUCUUGAUAAA-3′ (24, 25).

Coimmunoprecipitation assay and immunoblotting.

HEK293 cells were transfected in 6-well plates, and at 48 h posttransfection the cells were lysed in 300 μl of lysis buffer (50 mM Tris-HCl [pH 7.4], 150 mM NaCl, 5 mM EDTA [pH 8.0], and 1% Triton X-100) that contained a protease inhibitor mixture (Roche Applied Science) for 20 min on ice. Following centrifugation at 14,000 rpm for 15 min at 4°C, 90% of the supernatant was incubated with FLAG-M2 beads (Sigma) at 4°C for 4 h. For β-catenin/USP7 pulldown, lysates of nontransfected HEK293 cells were incubated with β-catenin/USP7 antibodies before being loaded onto protein A/G beads (Santa Cruz Biotechnology), which were then washed three times with the lysis buffer at 4°C for 5 min. The bound proteins were eluted with SDS loading buffer at 95°C for 5 min. Final total lysates and immunoprecipitates were subjected to SDS-PAGE and Western blot analysis. The antibodies used for this process were as follows: anti-FLAG (M2; Sigma), anti-myc (Sigma), antiubiquitin (Santa Cruz Biotechnology), anti-β-actin (Abcam), anti-USP7 (Bethyl Laboratories), anti-RNF220 (Sigma), and anti-β-catenin (Cell Signaling Technology). Horseradish peroxidase-coupled anti-mouse or anti-rabbit IgG (Pierce) was used as the secondary antibody. Chemiluminescence detection (Pierce) was carried out according to the manufacturer's instructions.

In vivo ubiquitination assays.

HEK293 cells were transfected with a hemagglutinin (HA)-ubiquitin construct together with the indicated plasmids. Proteasomal inhibitor MG132 at 25 nM was added 6 h prior to harvesting. At 48 h after transfection, cells were harvested and lysed in an SDS lysis buffer (50 mM Tris-HCl [pH 6.8], 1.5% SDS) at 95°C for 15 min. Following 10-fold dilution of the lysate with extraction buffer C (EBC)–bovine serum albumin (BSA) (50 mM Tris-HCl [pH 6.8], 180 mM NaCl, 0.5% NP-40, and 0.5% BSA) plus protease inhibitors (Roche), the cell lysates were immunoprecipitated with anti-FLAG M2 beads (Sigma). The bound proteins were eluted with a Laemmli sample buffer at 95°C for 5 min. Western blot analysis was performed using an antiubiquitin antibody. The membrane was then stripped and reprobed with the other indicated antibodies.

Embryos, microinjection, and animal cap assays.

Adult Xenopus laevis frogs were obtained from Nasco. In vitro fertilization, embryo culture, staging, and microinjection were carried out as described previously (26). For the secondary axis induction assays, 5 pg β-catenin either with or without 500 to 1,000 pg RNF220 mRNAs was injected into the ventral blastomeres of four-cell-stage embryos. For animal cap assays, 5 pg β-catenin either with or without 500 to 1,000 pg RNF220 mRNAs was injected into the animal poles of the blastomeres of two-cell-stage embryos, and animal caps were excised from late-blastula-stage embryos, cultured to stage 15, and then harvested for further reverse transcription-PCR analysis.

Reverse transcription and real-time PCR.

Reverse transcription was carried out by using a First-Strand cDNA synthesis kit (Fermantas). The primers used for real-time PCR were as follows: human RNF220 forward, 5′-GATGCCATCCACCAGCAA-3′, and reverse, 5′-CACGAGATAGCTGCCGTTCA-3′; human USP7 forward, 5′-CGGCCTGATGCTTTTTGGAC-3′, and reverse, 5′-TGAGAGCCGGTACATCAGGA-3′; human GAPDH forward, 5′-GAGAAGGCTGGGGCTCAT-3′, and reverse, 5′-TGCTGATGATCTTGAGGCTG-3′; human CCND1 forward, 5′-GCTGCGAAGTGGAAACCATC-3′, and reverse, 5′-CCTCCTTCTGCACACATTTGAA-3′; human axin2 forward, 5′-CAACACCAGGCGGAACGAA-3′, and reverse, 5′-GCCCAATAAGGAGTGTAAGGACT-3′; Xenopus Xnr3 forward, 5′-GTTTATCTCCCCACTGATGGCGATG-3′, and reverse, 5′-GCTTTGGACGGTATCAGATTCCTG-3′; Xenopus Siamois forward, 5′-TACCGCACTGACTCTGCAAG-3′, and reverse, 5′-CTGAGGCTCCTGTGGAATTC-3′; Xenopus ornithine decarboxylase (ODC) forward, 5′-GGGCTGGATCGTATCGTAGA-3′, and reverse, 5′-TGCCAGTGTGGTCTTGACAT-3′.

Statistical analysis.

Statistical analysis was performed using a two-tailed unpaired t test, with P values of <0.05 considered statistically significant, and P values of <0.001 were considered statistically very significant. SPSS 16.0 was used for all statistical calculations (SPSS, Inc., Chicago, IL).

RESULTS

RNF220 interacts with β-catenin.

In a previous study, we found that RNF220 encodes an E3 ubiquitin ligase (22). To screen for its potential target proteins, we used a yeast two-hybrid system (22) and isolated one positive clone that encoded a partial sequence of β-catenin. To confirm their interaction, we used coimmunoprecipitation (co-IP) assays. In HEK293 cells transiently cotransfected with FLAG-tagged RNF220 and myc-tagged β-catenin, β-catenin coimmunoprecipitated with RNF220 (Fig. 1A). In the reverse experiment, RNF220 also coimmunoprecipitated with β-catenin (Fig. 1B). When β-catenin was immunoprecipitated from HEK293 cell lysate using anti-β-catenin antibody, endogenous RNF220 was also detected in the immunoprecipitate (Fig. 1C).

FIG 1.

RNF220 interacts with β-catenin. (A) β-Catenin was coimmunoprecipitated with RNF220. (B) RNF220 was coimmunoprecipitated with β-catenin. HEK293 cells were transiently transfected with different combinations of expression vectors of β-catenin and RNF220, as indicated. Cell lysates were incubated with anti-FLAG beads, washed, and subsequently analyzed by Western blotting. (C) Endogenous RNF220 could be immunoprecipitated by β-catenin. Cell extracts of HEK293 cells were immunoprecipitated with antibody against β-catenin, and endogenous RNF220 was detected by an anti-RNF220 antibody, with IgG as a negative control. (D) Schematic representation of the structures of mouse RNF220 truncates. (E) The ΔNΔR truncate of RNF220 showed a markedly lower affinity for β-catenin. HEK293 cells were transiently transfected with different combinations of expression vectors of myc-tagged β-catenin and FLAG-tagged RNF220 or its truncates, as indicated. Cell lysates were incubated with anti-FLAG beads and subsequently analyzed by Western blotting. (F and G) Interactions of various Xenopus β-catenin truncation constructs with RNF220 in co-IP experiments. A schematic map of the Xenopus β-catenin truncates is shown, with amino acid numbers (F). WCL, whole-cell lysate; IB, immunoblot; IP, immunoprecipitation; WT, wild type; TAD, transactivation domain.

To map the domains of RNF220 and β-catenin that are responsible for the observed interaction, a series of RNF220 and β-catenin truncates were constructed and tested via IP assays in HEK293 cells (Fig. 1D to G). RNF220 contains a characteristic RING finger domain at its C terminus. We constructed a ΔRING (with the RING finger domain deleted), a ΔN (with the N-terminal amino acids 1 to 314 deleted), and a ΔNΔR (with both the RING finger domain and the N-terminal amino acids 1 to 314 deleted) (Fig. 1D). Our results showed that the ΔRING, ΔN, and ΔNΔR constructs of RNF220 could all associate with β-catenin, though the ΔNΔR construct worked less efficiently than the others (Fig. 1D and E). Testing the series of β-catenin deletion constructs via co-IP assays showed that the Armadillo repeats 2 to 8 of β-catenin were responsible for the interactions with RNF220 (Fig. 1F and G). While deletion of the N-terminal 473 amino acids, including this domain, abolished its interaction with RNF220, deletions of either the N-terminal sequences up to amino acid 222 or the C-terminal transactivation domain had no clear effect (Fig. 1F and G).

RNF220 stabilizes β-catenin by attenuating its ubiquitination.

Given that RNF220 is an E3 ligase that often regulates the stability of its substrates (3, 22), we first examined if the β-catenin protein level was regulated by RNF220. Unexpectedly, cotransfection of RNF220 stabilized β-catenin protein in the HEK293 cells (Fig. 2A). Likewise, in β-catenin reporter assays, RNF220 enhanced the β-catenin transcriptional activity (Fig. 2B). We further carried out polyubiquitination assays to examine if RNF220 may affect the ubiquitination status of β-catenin, and we found that coexpression of RNF220 strongly reduced the ubiquitination of β-catenin protein (Fig. 2C).

FIG 2.

RNF220 stabilizes β-catenin by attenuating its ubiquitination. (A) β-Catenin protein was stabilized by RNF220 overexpression. FLAG-tagged RNF220, green fluorescent protein (GFP)- and myc-tagged β-catenin plasmids were transfected into HEK293 cells as indicated. After 48 h, cell lysates were analyzed by Western blotting. (B) RNF220 enhanced the transcriptional activity of β-catenin in a dose-dependent manner. TOPflash reporter assays were carried out in HEK293 cells. β-Catenin and increasing amount of RNF220 plasmids (30, 60, and 90 ng) were transfected as indicated. (C) Effects of wild-type RNF220, ΔRING deletion, and the W539R mutant on the polyubiquitination of β-catenin. HEK293 cells were transiently transfected with the indicated plasmids and treated for 6 to 8 h with MG132 before harvest. β-Catenin proteins were immunoprecipitated with anti-FLAG antibody and then detected for polyubiquitin chains with the antibody against ubiquitin. (D and E) Effects of wild-type RNF220, ΔRING deletion, and the W539R mutant on the stability of β-catenin. HEK293 cells were transiently transfected with the indicated plasmids. At 48 h posttransfection, cycloheximide was added to all samples, and the cells were then harvested at the time points indicated. Levels of β-catenin were determined by Western blotting with anti-FLAG antibody. In all cases, β-actin was used as a loading control. The relative levels of β-catenin were quantified densitometrically and normalized against β-actin (E). The data shown in panel E are the averages of three independent experiments. WCL, whole-cell lysate; IB, immunoblot; IP, immunoprecipitation; CHX, cycloheximide.

We then tested whether the ligase activity of RNF220, which is dependent on its RING finger domain, is required for this process. When the ΔRING construct was tested in this assay, it showed a much-reduced activity compared with the wild-type construct (Fig. 2C), suggesting that the RING finger domain of RNF220 is required for its full activity. Next, we constructed a point-mutated form of RNF220 (W539R), which is supposed to be a ligase-dead mutant (3, 27). Indeed, RNF220 (W539R) failed to promote the ubiquitination of Sin3B (data not shown), an established target protein of RNF220 (22). However, RNF220 (W539R) efficiently blocked the ubiquitination of β-catenin like the wild type (Fig. 2C). We also used a series of cycloheximide-based protein chases to test the protein stability of β-catenin coexpressed with different RNF220 constructs. In the presence of both wild-type RNF220 and the ligase-dead mutant, β-catenin was clearly stabilized (Fig. 2D and E). However, the ΔRING construct showed a much weaker stabilization of β-catenin (Fig. 2D and E). The above data imply that RNF220 stabilizes β-catenin by attenuating its ubiquitination, independent of its ubiquitin ligase activity. If so, then the RING finger domain would likely be structurally required for its full activity in this process.

RNF220 bridges USP7 to β-catenin.

Our data indicated that RNF220 promoted deubiquitination of β-catenin, an activity characteristic for the ubiquitin-specific protease (USP). A search of the BioGRID database identified ubiquitin-specific peptidase 7 (USP7) as a potential RNF220 interactor, which was supported by previous global proteomic analyses of the deubiquitinase protein complexes (28, 29). We first confirmed the interaction of RNF220 and USP7 by co-IP of exogenously expressed and also endogenous proteins in HEK293 cells (Fig. 3A to C). USP7 contains a MATH domain, a catalytic domain, and five ubiquitin-like domains (Fig. 3D). To test which domain of USP7 is responsible for its interaction with RNF220, several USP7 truncates were constructed and assayed in a series of co-IP experiments (Fig. 3D and E). Deletion of the ubiquitin-like domains or the ubiquitin-like domains plus the catalytic domain did not affect USP7's association with RNF220, while deletion of the MATH domain abolished its interaction entirely (Fig. 3D and E). These findings imply that the MATH domain of USP7 is responsible for its interaction with RNF220.

FIG 3.

RNF220 bridges USP7 to β-catenin. (A) RNF220 coimmunoprecipitated with USP7 in HEK293 cells. (B) USP7 coimmunoprecipitated with RNF220 in HEK293 cells. (C) Endogenous RNF220 could be immunoprecipitated by USP7. Cell extracts of HEK293 cells were immunoprecipitated with an antibody against USP7, and RNF220 was detected with an anti-RNF220 antibody. IgG was used as a negative control. (D) A schematic map of the USP7 truncates with amino acid numbers. (E) Interactions of the USP7 truncates with RNF220 in co-IP experiments. (F) Interactions of the RNF220 truncates with USP7 in co-IP experiments. (G) Interactions of W539R and ΔRING RNF220 with USP7 in co-IP experiments in HEK293 cells. (H) β-Catenin levels pulled down by USP7 with or without cotransfected RNF220. (I) β-Catenin levels pulled down by USP7 in normal and shRNF220 stably transfected HEK293 cells. For panels A and B and E to I, HEK293 cells or shRNF220 stably transfected HEK293 cells were transiently transfected with different combinations of expression vectors for RNF220, USP7, and β-catenin, as indicated. IB, immunoblot; IP, immunoprecipitation; WCL, whole-cell lysate.

As for RNF220, we found that the RING domain and the N-terminal 315-amino-acid fragment of RNF220 were responsible for its interaction with USP7. Deletion of either domain dramatically reduced its interaction with USP7, while the construct with both regions deleted was unable to bind USP7 (Fig. 1D and 3F). The RNF220 mutant W539R worked equally as well as the wild type in terms of its interaction with USP7 (Fig. 3G).

Since E3 ligases and deubiquitinases often interact and regulate the ubiquitination level and stability of one another (3, 28), we tested whether RNF220 and USP7 could regulate each other as a target. The results showed that neither the protein level nor the ubiquitination level of RNF220 was affected when coexpressed with wild-type USP7 or its deubiquitinase-dead mutant C223S (data not shown). Similarly, the USP7 protein and its ubiquitination level remained unchanged both in cells overexpressing RNF220 and in the two RNF220 shRNA stably transfected cells (data not shown). These findings suggest that RNF220 and USP7 are not effective substrates of each other. Next, we asked if USP7 is recruited to β-catenin by RNF220. As shown in Fig. 3H, β-catenin was pulled down by USP7 weakly in HEK293 cells, but when RNF220 was coexpressed, a greater amount of β-catenin was found in complex with USP7. In contrast, when the RNF220-noninteracting ΔMATH-USP7 construct was used, β-catenin was not pulled down whether or not RNF220 was overexpressed (Fig. 3H). Conversely, when RNF220 was knocked down in HEK293 cells by specific shRNAs, the interaction between USP7 and β-catenin weakened (Fig. 3I). On the other hand, the interaction between RNF220 and β-catenin remained intact whether or not USP7 was present (data not shown). Collectively, these data imply that RNF220 bridges the interaction between USP7 and β-catenin.

USP7 deubiquitination of β-catenin is dependent on RNF220.

We next investigated whether the activity of RNF220 in stabilizing β-catenin was due to RNF220's recruitment of USP7 to β-catenin. We first tested whether USP7 cooperated with RNF220 to promote β-catenin signaling in reporter assays. As expected, coexpression of USP7 with RNF220 enhanced its activity to promote β-catenin signaling dose dependently, while the deubiquitinase-dead mutant C223S could not (Fig. 4A) (30). In HEK293 cells, coexpression of USP7 clearly reduced the ubiquitination level of β-catenin while the C223S mutant and USP4 could not, suggesting that β-catenin may be a direct target of USP7 (Fig. 4B) (31). Further coexpression of wild-type USP7, but not the C223S mutant, with RNF220 promoted the deubiquitination of β-catenin (Fig. 4C). To test whether USP7 is required for RNF220 to stabilize β-catenin in HEK293 cells, two established specific siRNAs were used to knock down the endogenous USP7 (24, 25). Both siRNAs reduced USP7 protein levels in transfected cells (Fig. 4D). In these cells, the ability of RNF220 to reduce the ubiquitination of β-catenin was largely abolished (Fig. 4D). On the other hand, in cells stably transfected with shRNAs against RNF220, the activity of USP7 to deubiquitinate β-catenin was also impaired (Fig. 4E). We conclude that RNF220 and USP7 likely work in complex to deubiquitinate and stabilize β-catenin, as both of these are required for the process.

FIG 4.

USP7 deubiquitination of β-catenin is dependent on RNF220. (A) USP7 cooperated with RNF220 to enhance the transcriptional activity of β-catenin, while its mutant, C223S, could not. RNF220, β-catenin, USP7 wild type, and its C223S mutant constructs together with TOPflash plasmid were transfected into HEK293 cells. At 48 h posttransfection, cells were lysed for luciferase reporter assays. USP7 was transfected with two different doses. (B) Coexpression of USP7 but not its ligase-dead mutant reduced the level of ubiquitinated β-catenin. (C) Coexpression of USP7 and RNF220 strongly reduced the level of ubiquitinated β-catenin. (D) Effects of RNF220 on levels of ubiquitinated β-catenin in control cells and USP7 knockdown cells. (E) Effects of USP7 on the levels of ubiquitinated β-catenin in control cells and RNF220 knockdown cells. (B to E) HEK293 cells or shRNF220 stably transfected HEK293 cells was transfected with USP7, β-catenin, control, and two independent USP7 siRNAs, RNF220 and ubiquitin plasmids, as indicated. At 40 h after transfection, the cells were treated with MG132 for another 6 to 8 h and then harvested for ubiquitination assays. IB, immunoblot; IP, immunoprecipitation; WCL, whole-cell lysate; WT, wild type.

The RNF220 protein level is regulated by Gsk3β.

In the canonical Wnt signaling pathway, phosphorylation of β-catenin at S33/37/41 by Gsk3β is required for its subsequent ubiquitination by β-Trcp and proteasomal degradation (8–11). As RNF220 stabilized β-catenin rather than target it for ubiquitination and degradation, we tested the possibility that RNF220 works through targeting Gsk3β to stabilize β-catenin indirectly. However, Gsk3β protein remained intact when RNF220 was overexpressed in HEK293 cells (data not shown). Unexpectedly, when coexpressed with Gsk3β the level of RNF220 protein reduced dramatically, which could be rescued by LiCl treatment (Fig. 5A). To identify the region in RNF220 that is responsive to Gsk3β regulation, a series of RNF220 truncates were constructed (Fig. 1D). When coexpressed with Gsk3β, the protein levels of most of the constructs (containing amino acids 1 to 512, 315 to 567, 315 to 512, or 1 to 330) were all reduced, similar to the wild-type RNF220 (Fig. 5B). However, the construct containing amino acids 330 to 513 was not responsive to Gsk3β (Fig. 5B). Gsk3β (K85R), the dominant negative form of Gsk3β (32), had no clear effect on the levels of RNF220 or its truncates (Fig. 5B). These data suggest that the region of amino acids 315 to 330 of RNF220 (NARIGKMKRRKQDEG) is responsible for its regulation by Gsk3β, which usually regulates the stability and ubiquitylation status of its substrates by phosphorylating its substrates at specific serines and/or threonines (33, 34). Curiously, there is no serine or threonine in this region that could be phosphorylated by Gsk3β, but there are three lysines that could serve as potential ubiquitination sites responsible for the observed stability regulation (Fig. 5C) (1, 2).

FIG 5.

The RNF220 protein level is regulated by Gsk3β. (A) Gsk3β reduced the RNF220 protein level. RNF220 and Gsk3β plasmids were transfected into HEK293 cells as indicated, and 12 h after transfection, cells were treated with LiCl or not, as indicated, for another 36 h and then harvested for Western blotting. IB, immunoblot. (B) The protein levels of the RNF220 truncates transfected alone or with Gsk3β or its dominant negative (K85R) form. Plasmids of EGFP, different RNF220 constructs, and Gsk3β wild type or its dominant negative form were transfected into HEK293 cells. At 48 h posttransfection, cells were lysed for Western blotting. IB, immunoblot; WT, wild type; DN, dominant negative. (C) Schematic representation of the modular structure of RNF220. (D) Ubiquitination assay results for RNF220 and its mutant in transfected HEK293 cells. (E) Effects of Gsk3β on the ubiquitination status of RNF220 and the RNF220 K321/3/6R mutant. Plasmids carrying RNF220 wild type (W.T.) or its K321/3/6A mutant or Gsk3β wild type or its dominant negative (D.N.) form were transfected into HEK293 cells as indicated. At 40 h posttransfection, cells were treated with MG132 for another 6 to 8 h and then harvested for ubiquitination assays. WT, wild type; DN, dominant negative; IB, immunoblot; IP, immunoprecipitation; WCL, whole-cell lysate. (F) Gsk3β was pulled down by RNF220 only in the presence of coexpressed β-catenin. RNF220, Gsk3β, and β-catenin plasmids were transfected into HEK293 cells. At 48 h posttransfection, cells were lysed for IP assays. IB, immunoblot; IP, immunoprecipitation; WCL, whole-cell lysate.

We hypothesized that RNF220 could be regulated by the ubiquitin-proteasome pathway and that this process was sensitive to Gsk3β. To test this possibility, we created a mutated form of RNF220 in which the three lysines (K321/323/326) were mutated to arginine. Indeed, when expressed in HEK293 cells, the ubiquitination of the mutated form was greatly reduced compared with the wild type (Fig. 5D). When coexpressed with Gsk3β but not its dominant negative form, the ubiquitination level of RNF220 increased (Fig. 5E). However, when the KR mutated RNF220 was used, Gsk3β showed no clear effect on either its protein level or ubiquitination level (Fig. 5E). These data suggested that RNF220 is regulated by ubiquitination, which is sensitive to Gsk3β regulation. While direct evidence of this supposition is lacking, Gsk3β most likely phosphorylates RNF220 at multiple sites and primes it for ubiquitination. Indeed, when β-catenin was coexpressed, RNF220 was found in complex with Gsk3β, supporting a direct role of Gsk3β in RNF220 regulation (Fig. 5F).

RNF220 enhances canonical Wnt signaling.

The extant data mentioned above suggest that RNF220 is capable of stabilizing β-catenin by recruiting the deubiquitinase USP7, so we opted to check the behavior of various RNF220 constructs in the Wnt signaling pathway by using the TOPflash reporter system. We found that coexpression of wild-type RNF220 with β-catenin strongly enhanced the reporter activity (Fig. 6A). As implicated from our above data, the ligase-dead mutant W539R worked as well as the wild type in this assay, confirming that the regulation of β-catenin levels by RNF220 is independent of its ubiquitin ligase activity (Fig. 6A). The constructs with deletions of either the N terminus or the RING finger domain, which compromise their ability to bind USP7, both showed a weaker enhancement of Wnt reporter expression (Fig. 1D, 3F, and 6A). Interestingly, in this assay, the construct lacking both the N terminus and the RING finger domain inhibited the Wnt reporter activity (Fig. 6A). Since this form of RNF220 can still weakly bind β-catenin but not USP7 (Fig. 1D and 3F), we reasoned that it may function as a dominant negative form of RNF220. When coexpressed, the ΔNΔR construct did indeed inhibit the activity of wild-type RNF220 in enhancing Wnt signaling by β-catenin (Fig. 6B). We also tested the activity of RNF220 in Xenopus embryos, which is a classical system of monitoring Wnt signaling. In the embryonic axis duplication assay, wild-type RNF220, but not the ΔNΔR form, synergized with β-catenin to induce ectopic axes (Fig. 6C). We also examined the expression of Wnt target genes Siamois and Xnr3 in animal caps injected with β-catenin alone or with RNF220, and we found that RNF220 also enhanced the expression of the Wnt target genes in concert with β-catenin (Fig. 6D and E).

FIG 6.

RNF220 acts as a canonical Wnt signaling amplifier. (A) Effects of RNF220 and its truncates on the stimulation of Wnt reporter expression by β-catenin in 293 cells. Plasmids of RNF220 constructs were transfected with three different doses to evaluate their effects. WT, wild type. (B) The ΔNΔR truncate of RNF220 reversed the stimulation effect of wild-type RNF220 on β-catenin activity in 293 cells. RNF220 plasmids were cotransfected with β-catenin as indicated, and then luciferase reporter assays were performed. (C) RNF220 wild type but not the ΔNΔR truncate cooperated with β-catenin to induce secondary axes in Xenopus embryos. The numbers of embryos in each group are indicated above the bars, and representative embryos are shown in the right panel. (D and E) Real-time PCR analysis of the expression of Siamois and Xnr3 in Xenopus animal caps injected with β-catenin with or without RNF220 mRNA. ns, not significant; **, P < 0.001.

RNF220 and USP7 are involved in Wnt activation in colon cancer cells.

Previous studies noted that the activation of the Wnt pathway is linked to various cancers, in particular to colon cancer (9, 10). Accordingly, we checked the potential involvement of RNF220 in Wnt stimulation in two human colon cancer cell lines, HCT116 and SW480 (35). In these cells, wild-type RNF220, the ΔRING deletion, and the ligase-dead W539R mutant all enhanced canonical Wnt signaling in reporter assays (Fig. 7A and B). Cycloheximide-based protein chase assays showed that RNF220 stabilized endogenous β-catenin in both HCT116 and SW480 cells (Fig. 7C and D). To test the function of RNF220/USP7 in vivo, we performed knockdown experiments using RNF220/USP7 siRNAs in the two cell lines. The two independent RNF220 siRNAs reduced the endogenous RNF220 protein level by more than 50% in HCT116 and SW480 cells (Fig. 7E). Likewise, a significant reduction of β-catenin protein was observed in both HCT116 and SW480 cells when RNF220 was knocked down (Fig. 7E). In addition, the expression of the transfected Wnt reporter was also reduced when RNF220 was knocked down in HCT116 and SW80 cells (Fig. 7F and G). Cycloheximide-based protein chase assays showed that both RNF220 or USP7 knockdown could destabilize endogenous β-catenin in the two cells lines (Fig. 7H to K). Also in both cell lines, knockdown of RNF220 or USP7 similarly reduced the expression of endogenous axin2 and CCND1, two established target genes of canonical Wnt signaling (Fig. 7L to S). Taken together, these data support the idea that RNF220/USP7, the β-catenin-destabilizing complex, is involved in canonical Wnt signaling activation in colon cancer cells, bolstering the potential role of RNF220 in tumorigenesis.

FIG 7.

RNF220 and USP7 are involved in Wnt activation in colon cancer cells. (A and B) Effects of wild-type, ΔRING, and the W539R mutant RNF220 on expression of Wnt reporter genes in HCT116 and SW480 cells. RNF220 plasmids together with the reporter plasmids were transfected into HCT116 and SW480 cells, and luciferase reporter assays were performed 48 h after transfection. WT, wild type; Mut, W539R mutant; RLU, relative luciferase units. **, P < 0.001. (C and D) Effects of RNF220 overexpression on the stability of endogenous β-catenin in HCT116 and SW480 cells. HCT116 and SW480 cells were transiently transfected with RNF220 plasmids. At 48 h posttransfection, cycloheximide was added to all samples, and the cells were then harvested at the time points indicated. Levels of endogenous β-catenin were determined by Western blotting with an anti-β-catenin antibody. In all cases, α-tubulin was used as a loading control. IB, immunoblot; CHX, cycloheximide. (E) Effects of RNF220 knockdown on β-catenin levels in HCT116 and SW480 cells. The indicated siRNAs were transfected into the indicated cells, and 48 h after transfection cells were harvested for Western blotting to examine the endogenous protein level. IB, immunoblot. (F and G) Effects of RNF220 knockdown on Wnt signaling activity in HCT116 and SW480 cells. The reporter plasmids were first transfected into HCT116 and SW480 cells, and after 6 h the indicated siRNAs were transfected. After 48 h, the cells were harvested for luciferase reporter assays. ns, not significant; *, P < 0.05; **, P < 0.001. RLU, relative luciferase units. (H to K) Effects of RNF220 (H and I) or USP7 (J and K) knockdown on the stability of endogenous β-catenin in HCT116 and SW480 cells. RNF220 or USP7 siRNA was transfected into HCT116 and SW480 cells, and 48 h after transfection cells were treated with cycloheximide for the indicated times and then harvested for Western blotting to examine the endogenous protein level. Note that the exposure times for β-catenin in the left and right panels were different, in order to better illustrate the degradation effect, since knockdown of RNF220 or USP7 reduced total β-catenin levels. IB, immunoblot; CHX, cycloheximide. (L to S) The expression of endogenous axin2 and CCND1 in HCT116 and SW480 cells when RNF220 (L to O) or USP7 (P to S) was knocked down. Control luciferase siRNA and two independent RNF220 or USP7 siRNAs were transfected into HCT116 and SW480 cells. After 48 h, the cells were harvested for real-time PCR. The expression levels of axin2 and CCND1 were normalized with that of glyceraldehyde 3-phosphate dehydrogenase (GAPDH). NC, negative control. ns, not significant; *, P < 0.05; **, P < 0.001.

DISCUSSION

Controlling the stability of crucial signaling components via ubiquitination has emerged as an important aspect of Wnt pathway regulation (36). β-Catenin, the key effector of the pathway, has been found to be targeted by multiple E3 ligases that are regulated by distinct stimuli in different contexts (12–17). The roles of deubiquitinating enzymes in controlling the Wnt/β-catenin pathway have also been exemplified in several studies, wherein they target either β-catenin, APC, or axin (20, 21, 37). Our present study extends these findings by defining for the first time the RING finger ubiquitin E3 ligase RNF220 as a β-catenin stabilizer through USP7-mediated deubiquitination, thereby suggesting a previously unknown broader role of ubiquitin E3 ligases.

Our data demonstrate that the RNF220/USP7 complex works as a β-catenin-deubiquitinating enzyme complex that is able to stabilize β-catenin and promote canonical Wnt signaling. In this complex, RNF220 serves as an adaptor protein that bridges USP7 and β-catenin. This process seems to be independent of the E3 ligase activity of RNF220, since the ligase-dead form of RNF220 works as efficiently to stabilize β-catenin as the wild type (Fig. 2C). However, the RING finger domain is a prerequisite for efficient binding of USP7 to achieve full activity in stabilizing β-catenin (Fig. 2C and 3F and G). Though RNF220 is an active E3 ligase able to ubiquitinate itself and its substrate Sin3B (22), it is unable to ubiquitinate β-catenin. As a consequence, the deubiquitinase activity of USP7 is necessary for this process, suggesting that the complex does not simply block β-catenin ubiquitination via structural masking or competing with active E3 ligases of β-catenin.

Ubiquitin E3 ligases have been frequently found to be in complex with DUBs, wherein the E3 ligase may destabilize the DUB through ubiquitination and, reciprocally, the DUB may stabilize the E3 through deubiquitination (4). For example, the interaction of USP7 and the p53 E3 ligase MDM2 has previously been suggested to stabilize MDM2 and reduce p53 protein levels (24, 25, 38–40). In the case of RNF220 and USP7, however, neither overexpression nor knockdown of either protein had any clear measurable effect on the protein levels of the other, implying that RNF220 and USP7 are not effective substrates of each other (unpublished data). RNF220 binds to the MATH domain of USP7, which is also responsible for the binding of USP7 to MDM2 and p53 (Fig. 3D and E) (7). Whether the binding of RNF220 has any substantial effects on the function of USP7 in p53 regulation remains to be seen.

For β-catenin ubiquitination, β-Trcp is the dominant E3 ligase and β-Trcp can only ubiquitinate phosphorylated β-catenin at S33/37/41 via Gsk3β (8, 41). Recent studies demonstrated that phosphorylation-independent ubiquitination of β-catenin is also important for canonical Wnt signaling regulation. For example, the ubiquitin E3 ligase c-Cbl specifically targets nuclear-active β-catenin, and Siah1/2 can target β-catenin for degradation in a Gsk3β- and β-Trcp-independent manner (15–17). Both RNF220 and USP7 predominantly localize to the nucleus, but they are also weakly present in the cytoplasm. We suggest that RNF220/USP7 targets both the cytoplasm and nuclear β-catenin for deubiquitination, with RNF220 likely being recruited to the Gsk3β complex via β-catenin (Fig. 5F). In the present study, we found that RNF220 and USP7 could also stabilize the S33/37/41A mutant β-catenin and enhance its transcriptional activity (data not shown), supporting the contention that RNF220/USP7 is also active in the nucleus in stabilizing β-catenin.

In both Wnt reporter assays in cultured cells and the secondary axis induction assay in Xenopus embryos, RNF220 was found to cooperate with β-catenin to stimulate Wnt signaling. This result was not surprising, given RNF220's ability to stabilize β-catenin with USP7. In two human colon cancer cell lines with stimulated Wnt signaling—HCT116 and SW480 (35)—knockdown of RNF220 reduced the expression of both the Wnt reporter and endogenous Wnt target genes (Fig. 7E to L), suggesting that RNF220 may contribute to Wnt stimulation and tumorigenesis in colon cancer. Curiously, in HEK293 cells, which exhibit weak background Wnt signaling, overexpression or knockdown of RNF220 had no clear effect on Wnt reporter activity (data not shown), though a clear cooperative effect was observed when it was coexpressed alongside β-catenin (Fig. 6A). These findings suggest that RNF220 is not absolutely required for canonical Wnt signaling but instead likely acts as an amplifier or enhancer.

The cooperative enhancing or amplifying effect of RNF220 fits well with our other findings indicating that the stability of RNF220 itself is regulated by Gsk3β, and thus Wnt signaling activity. Gsk3β is a key serine/threonine kinase that phosphorylates β-catenin and promotes its ubiquitination and consequent proteasomal degradation (8, 33, 34, 41). In the presence of the Wnt ligand, Gsk3β dissociates with β-catenin, allowing its stabilization and activation. Our analysis showed here that overexpression of Gsk3β reduced RNF220 protein by promoting its ubiquitination (Fig. 5). Given the observation that RNF220 is recruited to Gsk3β through β-catenin (Fig. 5F), we assume that Gsk3β works via phosphorylation of RNF220 and promotes ubiquitination, as is the case for β-catenin. We further suggest that Gsk3β targets multiple serines and threonines in RNF220, as most of our RNF220 deletion constructs remained responsive to Gsk3β overexpression. Quantified analysis showed that mutation of one potential Gsk3β phosphorylation site (S282/6) partially reduced its responsiveness to Gsk3β (data not shown). We did find one motif in RNF220 (amino acids 315 to 330) that is apparently required for Gsk3β induced reduction, which we confirmed to be the ubiquitination site responsive to Gsk3β regulation (Fig. 5). Indeed, Gsk3β promotes the ubiquitination of wild-type RNF220 but not that of the K321/3/6R mutant (Fig. 5E).

Based on the results of our analyses, we propose the following model for the role of RNF220/USP7 in Wnt signaling (Fig. 8). In the absence of the Wnt ligand, both β-catenin and RNF220 are recruited to the Gsk3β/axin/APC destruction complex, phosphorylated, and degraded through the ubiquitin-proteasome pathway. In the presence of the Wnt ligand, the destruction complex dissociates, β-catenin and RNF220 are both stabilized, and RNF220 works with USP7 to deubiquitinate β-catenin, further increasing its protein level and signaling activity. We suggest that the stability of RNF220 may be regulated by factors other than Gsk3β; if so, then RNF220 should be able to activate Wnt signaling independent of upstream Wnt activators. Further targeted studies should be able to more conclusively verify this possibility.

FIG 8.

Proposed model for the control of β-catenin stability by RNF220/USP7. In the absence of the Wnt ligand (left panel), both β-catenin and RNF220 are recruited to the Gsk3β/axin/APC destruction complex, phosphorylated, and degraded through the ubiquitin-proteasome pathway. Whether Gsk3β directly phosphorylates RNF220 remains to be determined, and the ubiquitin ligase for RNF220 in this process is unknown. In the presence of the Wnt ligand (right panel), the destruction complex dissociates, β-catenin and RNF220 are both stabilized, and RNF220 works with USP7 to deubiquitinate β-catenin, further increasing its protein level and signaling activity.

Wnt signaling is an important morphogen during embryonic development that is capable of inducing different cell fates at different dosages. The enhancement of Wnt signaling via RNF220/USP7 may potentially serve as a mechanism to sharpen the Wnt gradient, implying a potential role in cell fate determination that should be assessed via further in vivo analysis. Nevertheless, as our data already suggest, RNF220-mediated signaling amplification might have a role in Wnt stimulation in cancer cells. In conclusion, our findings provide a novel mechanism for Wnt signaling enhancement via deubiquitination of β-catenin by USP7, which, intriguingly, is mediated by the ubiquitin E3 ligase RNF220.