Research ArticleCancer

The E3 ligase RNF43 inhibits Wnt signaling downstream of mutated β-catenin by sequestering TCF4 to the nuclear membrane

See allHide authors and affiliations

Sci. Signal.  08 Sep 2015:
Vol. 8, Issue 393, pp. ra90
DOI: 10.1126/scisignal.aac6757

RNF43 halts Wnt at the nucleus

Wnt signaling is critical to development and is often reactivated in cancer. The E3 ubiquitin ligase RNF43 inhibits Wnt–β-catenin signaling. Rather than promoting the degradation of cell surface Wnt receptors, Loregger et al. found that RNF43 sequestered TCF4, a β-catenin partner and transcription factor, at the nuclear membrane through a mechanism independent of its E3 ligase function. When expressed in frog embryos, RNF43 bearing mutations like those found in human gastrointestinal tumors increased Wnt signaling. Coexpression of wild-type RNF43 suppressed Wnt signaling even in cells with a constitutively active mutant of β-catenin, indicating tumor-suppressive activity of RNF43 functioning downstream of β-catenin.


Given its fundamental role in development and cancer, the Wnt–β-catenin signaling pathway is tightly controlled at multiple levels. RING finger protein 43 (RNF43) is an E3 ubiquitin ligase originally found in stem cells and proposed to inhibit Wnt signaling by interacting with the Wnt receptors of the Frizzled family. We detected endogenous RNF43 in the nucleus of human intestinal crypt and colon cancer cells. We found that RNF43 physically interacted with T cell factor 4 (TCF4) in cells and tethered TCF4 to the nuclear membrane, thus silencing TCF4 transcriptional activity even in the presence of constitutively active mutants of β-catenin. This inhibitory mechanism was disrupted by the expression of RNF43 bearing mutations found in human gastrointestinal tumors, and transactivation of the Wnt pathway was observed in various cells and in Xenopus embryos when the RING domain of RNF43 was mutated. Our findings indicate that RNF43 inhibits the Wnt pathway downstream of oncogenic mutations that activate the pathway. Mimicking or enhancing this inhibitory activity of RNF43 may be useful to treat cancers arising from aberrant activation of the Wnt pathway.


Aberrant regulation of Wnt signaling is observed in more than 90% of sporadic colon tumors. Such mutations mostly occur in the genes encoding the adenomatous polyposis coli (APC) or β-catenin (1) and lead to constitutive formation of nuclear β-catenin–T cell factor 4 (TCF4) complexes, resulting in activated transcription of TCF target genes (2). A number of endogenous inhibitors of the Wnt signaling pathway have been identified, reflecting the need for tight regulation of this pathway. However, these inhibitors mostly restrict the biogenesis and secretion of Wnt ligands, prevent the interaction between Wnt ligands and the cell surface receptors, or disrupt frizzled (FZD) receptor–mediated activation of Dishevelled proteins (3), thus having minor function in tumors that harbor mutations in APC or β-catenin. In this context, different studies claimed that RING finger protein 43 (RNF43) and its homolog ZNRF3 inhibit Wnt signaling by reducing FZD receptor abundance at the plasma membrane (46), suggesting a tumor suppressor function of RNF43. These results contrasted with previous publications that showed RNF43 to be localized to the nuclear envelope, the endoplasmic reticulum (ER), and the nucleoplasm (7, 8). In line with this, RNF43 has been described to be involved in DNA damage response (DDR) (9), suggesting a possible nuclear function for RNF43. Thus, there are conflicting findings regarding the function of RNF43 depending on its subcellular localization, which may be due to the fact that the localization of the endogenous protein has not yet been clarified.

Early findings described RNF43 as an oncoprotein highly expressed in colorectal tumors (10), whereas more recent studies supported the function as a tumor suppressor reporting mutations in the RNF43 gene in different types of tumors including pancreatic cancer (11), mucinous tumors of the ovary (12), liver fluke–associated cholangiocarcinoma (13), and colon cancer (14). However, these mutations have not been characterized regarding their impact on RNF43 function. Also, it is still unclear whether RNF43 is involved in tumor development or may be associated with prognosis in cancer patients.

Here, we investigated the localization of endogenous RNF43, an alternative mechanism through which it inhibits Wnt signaling, and how mutations in RNF43 impair this newly uncovered tumor-suppressive function in the colon.


Endogenous RNF43 exhibits nuclear localization in colonic crypt and cancer cells and is overexpressed in colon tumors correlating with poor prognosis

Initially, we aimed at characterizing the expression of endogenous RNF43 at the cellular and subcellular level in human samples from normal tissues and colon tumors. By in situ hybridization, we found RNF43 expression in the stem cell compartment in human small intestine and colon, correlating with the expression of OLFM4 (Fig. 1A), a marker of the intestinal stem cell compartment, and found increased amounts of RNF43 mRNA in colon adenomas and adenocarcinomas.

Fig. 1 RNF43 is overexpressed in colon tumors and correlates with poor prognosis.

(A) In situ hybridization of RNF43 and olfactomedin 4 (OLFM4) in human tissues. Scale bars, 50 μm (small intestine and colon); 500 μm (colon adenoma and adenocarcinoma); 20 μm (insets). (B) RNF43 detected by immunohistochemistry in human small intestine [top images; scale bars, 50 μm (left) and 20 μm (right)] and colon adenocarcinoma [bottom images; scale bars, 500 μm (left) and 20 μm (right)]. (C and D) RNF43 mRNA expression in human normal colon mucosa (n = 8 samples), stage II (n = 8) and IV colon tumors (n = 6) (C) and colon cancer cell lines (n = 2 independent experiments performed in duplicate) (D). *P ≤ 0.05, **P ≤ 0.005, Kruskal-Wallis analysis of variance (ANOVA). (E) Subcellular localization of wild-type RNF43 and mutant RNF43H292R in HCT116 and SW480 cells. Scale bars, 10 μm. (F) Subcellular fractionation of HCT116 cells after transfection of wild-type RNF43 or mutant RNF43H292R. Calnexin, lamin A, or β-actin and α-tubulin were used as markers for ER, nuclear fraction, or cytoplasmic fraction, respectively. (G) Subcellular fraction of MCF-7 breast cancer cells expressing endogenous RNF43. Fraction controls as in (F), with CDC5L (cell division cycle 5-like) as an additional nuclear marker. (H) Subcellular localization of wild-type RNF43 and RNF43R437A-K655A lacking the NLS in HEK293 cells. Scale bar, 10 μm. (I) Kaplan-Meier analyses showing the probability of survival, relapse after surgery, and metachronic metastasis of 126 stage II colon cancer patients. Conditional Monte Carlo analysis was performed for consistency testing.

In murine small intestine, Rnf43 mRNA expression was limited to few cells located at the base of crypts, whereas high expression was detected in aberrant foci and intestinal adenomas and tumors of APCmin mice (fig. S1, A and B). Analysis of Rnf43 mRNA expression in different tissues from adult mice and embryos at different embryonic stages revealed that Rnf43 is mainly expressed in endoderm-derived tissues as well as in tissues presenting high Wnt activity (fig. S1, C to E).

To corroborate these findings, we established an immunohistochemical staining using a polyclonal antiserum directed against a peptide comprising amino acids 298 to 427, which shares no homology with other human proteins. This staining confirmed the increased abundance of RNF43 in colon tumors (Fig. 1B) and revealed the localization of endogenous RNF43 in the nucleus of cells in the intestinal crypt and colon cancer cells. We further performed immunocytochemical analysis of endogenous RNF43 in the colon cancer cell line HT-29, which expresses wild-type RNF43 (15). RNF43 was detected in the nucleus of the cells (fig. S1F). The specificity of the antibody was corroborated by staining HT-29 cells in which RNF43 was knocked down (fig. S1F). RNF43 mRNA expression in human tumors differed depending on the stage. Whereas the expression in nonmetastatic UICC (Union for International Cancer Control) stage II tumors was very heterogeneous, all metastatic UICC stage IV tumors showed very high RNF43 mRNA expression (Fig. 1C). A similar heterogeneous expression was observed in colon cancer cell lines (Fig. 1D).

The observed presence of RNF43 in the nucleus of the cells (Fig. 1B) prompted us to further analyze the subcellular localization of RNF43 in colon cancer cells. RNF43 was localized at the nuclear envelope and, to a lesser extent, in the nucleoplasm (Fig. 1E and fig. S2A). The same localization pattern was observed for the RING domain mutant (RNF43H292R), which was generated by introducing two point mutations in the two histidine residues involved in zinc coordination. Colocalization with the lamin B receptor, a marker for the inner nuclear membrane, and the nuclear RNA binding protein PSF (polypyrimidine tract-binding protein-associated splicing factor) (8) was observed (fig. S2, B and C). RNF43 was also found occasionally in the ER, where it colocalized with the ER marker calnexin (fig. S2B). Subcellular fractionation experiments performed after overexpression of RNF43 in HCT116 cells showed RNF43 unambiguously in the nuclear fraction (Fig. 1F). It has to be noted that the ER was also included in this fraction, as indicated by the presence of the ER marker calnexin. This was in line with immunofluorescence showing RNF43 in the nuclear membrane as well as in the ER. To exclude a possible mislocalization of the protein due to overexpression, we also obtained subcellular fractions from the breast cancer cell line MCF-7, which expresses wild-type RNF43. Also in these cells, the localization of endogenous RNF43 was limited to the nuclear fraction including the ER (Fig. 1G). RNF43 contains two putative nuclear localization signals (NLS) at amino acid position 434 (PLRRARP) and 654 (RKRR) as predicted by PSORTII, cNLS mapper, and NucPred software. By site-directed mutagenesis, we introduced two point mutations in each of the NLS resulting in two amino acid exchanges (RNF43R437A-K655A) and analyzed the subcellular localization of RNF43 in human embryonic kidney (HEK) 293 cells. When these NLS were mutated, RNF43 was found in the cytoplasm (Fig. 1H). Our results, showing nuclear localization of RNF43, indicate that RNF43 might have diverse functions depending on its subcellular localization.

Because RNF43 was heterogeneously expressed in tumor stage II samples (Fig. 1C), we next examined a possible correlation with clinicopathological characteristics in a larger cohort by performing quantitative polymerase chain reaction (qPCR) (fig. S3) and Kaplan-Meier analyses (Fig. 1I). Lower survival rates and more frequent metachronic metastases were observed in patients exhibiting higher RNF43 mRNA expression in tumors, even though not attaining significance. However, there was a significant positive correlation between RNF43 mRNA expression and relapse after surgery (P = 0.046). The RNF43 expression showed no obvious association with known clinical parameters such as grading or anatomical localization within the large intestine and was not correlated to microsatellite instability or to oncogenic mutations in KRAS or BRAF. Prognosis was not correlated with the expression of the Wnt target gene encoding osteopontin, indicating that the association observed for RNF43 was not based on differences in Wnt activity. These results suggest that high RNF43 expression is a biomarker for poor prognosis in stage II colorectal cancer.

RNF43 inhibits Wnt signaling downstream of β-catenin

A regulatory feedback loop has been proposed for RNF43 within the Wnt pathway (4, 16). Thus, to analyze whether RNF43 expression was enhanced by Wnt, we knocked down β-catenin in LS174T cells. This led to a significant reduction of RNF43 mRNA expression (Fig. 2A). By chromatin immunoprecipitation (ChIP)–coupled high-throughput sequencing experiments (ChIP-seq), we identified seven TCF4 binding sites within the RNF43 gene sequence that were evolutionarily highly conserved. These are located in the promoter region and within several introns of the RNF43 gene (fig. S4A), indicating that RNF43 is a direct target gene of TCF4. ChIP-qPCR experiments verified the binding of both TCF4 and β-catenin to the proximal promoter and the most prominent intronic TCF4 binding region (fig. S4B). Reporter assays in HCT116 and LS174T cells demonstrated that luciferase-driven expression by the most prominent intronic TCF4 binding region is TCF4-dependent (fig. S4C). Together, these results show that RNF43 is a direct TCF4 target gene.

Fig. 2 RNF43 inhibits Wnt signaling in cells and in vivo downstream of β-catenin.

(A) RNF43 and LGR5 mRNA expression after knocking down β-catenin expression in LS174T colon cancer cells (n = 3 experiments). (B) TCF transcriptional activity in the presence of wild-type RNF43 or RING domain point mutant RNF43H292R after WNT3A stimulation in HEK293 cells (n = 3 experiments). (C) TCF transcriptional activity after knocking down wild-type RNF43 in HT-29 colon cancer cells (n = 4 experiments). (D) X. laevis axis duplication assay. WNT1 mRNA was injected into ventral blastomeres of a four-cell X. laevis embryo alone (n = 198 embryos) or in combination with wild-type RNF43 (n = 142 embryos) or RNF43H292R mutant (n = 106 embryos). Representative pictures of two different embryos are shown. Data are presented as percentage of embryos showing ectopic axis formation. (E) TCF transcriptional activity induced by constitutive active Dishevelled, Axin2, β-TrCP, and stabilized β-catenin (S33-β-catenin) after RNF43 overexpression in HEK293 cells. (F) TCF transcriptional activity in colon cancer cell lines harboring mutations in β-catenin (HCT116) or APC (DLD-1) after overexpression of wild-type or RNF43H292R RING domain mutant. (G) Effect of ectopic expression of wild-type RNF43 and mutant RNF43H292R on β-catenin protein abundance in HCT116 cells. *P < 0.05, **P ≤ 0.01, ***P < 0.0005, Student’s t test or ANOVA with Dunnett’s multiple comparsion posttest. Non-p-β-catenin, nonphosphorylated β-catenin.

The Wnt regulatory function of RNF43 was initially assessed in HEK293 cells, which were stimulated with WNT3A. Here, RNF43 overexpression significantly decreased WNT3A-induced TCF transcriptional activity (Fig. 2B). Inactivation of the RING domain in RNF43 (RNF43H292R) not only abolished the inhibitory effect on Wnt signaling but also enhanced Wnt signaling activity in a dose-dependent manner (Fig. 2B). Thus, the functional RING domain in wild-type RNF43 is required for its inhibitory effect on the Wnt–β-catenin signaling pathway. To confirm these observations under endogenous conditions, RNF43 was knocked down in the colon cancer cell line HT-29 (fig. S4, D and E). Knocking down RNF43 significantly increased TCF transcriptional activity (Fig. 2C and fig. S4F), confirming that RNF43 is an endogenous repressor of Wnt activity.

To further validate these results in vivo, we performed Xenopus laevis axis duplication assays. Microinjection of WNT1 mRNA into X. laevis embryos induced the formation of a secondary body axis in 48.7% of the embryos, whereas co-injection of RNF43 mRNA reversed the effect of WNT1, given that double axis formation was only observed in 4.9% of all embryos analyzed (Fig. 2D). The presence of an active RING domain was also essential to inhibit Wnt signaling in vivo because co-injection with the mutant RNF43H292R-encoding mRNA did not reduce WNT1-induced double axis formation. These results confirmed that RNF43 can inhibit Wnt signaling in vivo. Notably, the inhibitory activity of RNF43 was already apparent at an early developmental stage just after gastrulation (fig. S5A). In addition, we observed that embryos injected with RNF43 mRNA alone exhibited a loss of posterior structures, whereas embryos injected with the mutant RNF43H292R-encoding mRNA were phenotypically normal at later stages of development (fig. S5B). Moreover, injection of RNF43 mRNA affected the expression of morphogens induced by Wnt, such as MyoD (myogenic differentiation 1) and chordin (fig. S5C). It is important to note that the mRNA generated from the human RNF43 complementary DNA (cDNA) sequence was used for injection into Xenopus embryos, pointing to a high conservation of RNF43 function across species. Indeed, when comparing RNF43 sequences from different species, a very high conservation (greater than 80%) is noticeable over the entire sequence of the protein (fig. S5D), whereas 41.17% identity was detected between human and Xenopus tropicalis RNF43. Together, these results indicate that RNF43 is not only an inhibitor of Wnt signaling in vitro but also a negative regulator of canonical Wnt signaling in vivo.

Because RNF43 was detected in the nucleus of the cells, we analyzed in detail whether RNF43 could be interacting with and thus inhibiting Wnt at an alternative level of the pathway. For this purpose, HEK293 cells, which are devoid of constitutive Wnt signaling, were transfected with constitutive active forms of different molecules in the Wnt signaling cascade [Dishevelled, dominant-negative (dn) Axin2, dn-β-TrCP, and β-catenin], and TCF transcriptional activity was measured. Notably, RNF43 inhibited TCF activity induced by all activators analyzed (Fig. 2E). Furthermore, RNF43 was able to suppress Wnt activity induced by Ser33-phosphorylated (activated) β-catenin, indicating that RNF43 acts downstream or at the level of β-catenin. These findings and the fact that RNF43 has been described as a putative tumor suppressor prompted us to assess the effect of RNF43 in HCT116 and DLD-1 colon cancer cell lines harboring activating mutations in β-catenin (CTNNB1) and APC, respectively (1). Inhibition of Wnt signaling was observed after overexpression of wild-type RNF43, whereas the RING mutant RNF43H292R acted in a dominant-negative manner (Fig. 2F). It has to be noted that the basal Wnt activity detected in these cells is much higher than the Wnt activity induced by WNT3A, and therefore, the observed inactivation was lower. To determine whether the dominant-negative effect induced by the RNF43H292R mutant resulted from increased autocrine Wnt signaling, experiments were performed in the presence of the porcupine inhibitor LGK974. Under these conditions, RNF43H292R still increased Wnt signaling (fig. S6A), excluding an autocrine loop on Wnt signaling. These results further confirmed that RNF43 acts at the level or downstream of β-catenin in the Wnt signaling cascade. This effect was specific for Wnt because overexpression of RNF43 did not alter other signaling pathways such as nuclear factor κB (fig. S6B), and was not induced by toxicity upon overexpression of RNF43 because no cellular phenotypical changes or alterations in cell viability were detected in response to RNF43 transfection (fig. S8, A and B).

Because RNF43 is an E3 ubiquitin ligase, we tested whether RNF43 was able to ubiquitinate β-catenin and mark it for degradation. RNF43 or RNF43H292R was overexpressed in HCT116 cells, and the protein abundance of endogenous β-catenin was analyzed, but found to be unchanged (Fig. 2G). Moreover, coimmunoprecipitation experiments did not reveal a direct interaction between RNF43 and β-catenin (fig. S6C), although the interaction of RNF43 with the nuclear protein PSF, which was suggested earlier (8), was confirmed (fig. S6D). Together, these results indicate that RNF43 can inhibit Wnt signaling downstream of β-catenin.

RNF43 interacts with TCF4 and prevents TCF4-mediated transcription

We detected a robust interaction between wild-type RNF43 and TCF4 by coimmunoprecipitation (Fig. 3A), whereas the mutant RNF43H292R only weakly coimmunoprecipitated with TCF4. Moreover, we observed that different N-terminally truncated RNF43 mutants interacted with TCF4 (Fig. 3B), suggesting that a region between the RING domain and the C terminus is responsible for RNF43 binding to TCF4. Thus, we generated a C-terminal deletion mutant (RNF43 1–447). Notably, no interaction between RNF43G447X and TCF4 was detected (Fig. 3B), confirming that the C-terminal part of RNF43 is involved in binding to TCF4.

Fig. 3 RNF43 binds TCF4 at the nuclear membrane.

(A) Immunoprecipitation (IP) of TCF4 from HCT116 cells transfected with HA-tagged wild-type or mutant RNF43H292R, followed by Western blotting. (B) Schematic representation of different human RNF43 N-terminal deletion constructs and a C-terminal deletion construct and Western blot after transfection in HCT116 cells. Bottom: Immunoprecipitation of TCF4 from HCT116 cells transfected with the indicated RNF43 constructs, followed by Western blotting. aa, amino acids. (C) Western blot analysis of immunoprecipitates from HCT116 cells transfected with the indicated constructs. RL, Renilla luciferase. (D and E) Western blot analysis (D) and confocal immunofluorescence (E) of TCF4 in HCT116 cells transfected with wild-type or mutant RNF43. Scale bars, 10 μm. Graphs below (E) show signal intensities across the transverse axis of the cell. (F) Immunoprecipitation of TCF4 in HCT116 cells transfected with HA-tagged wild-type or mutant RNF43, followed by Western blot analysis. (G) Expression of Wnt target genes Axin2 and Twist detected by qPCR after RNF43 overexpression in HCT116 cells (n = 3 experiments). *P < 0.05, Student’s t test.

TCF4 has been shown to be ubiquitinated by NARF, a Nemo-like kinase–associated RING finger protein (17) and deubiquitinated by USP4 (ubiquitin-specific protease 4). We therefore tested if RNF43 might be another ubiquitin E3 ligase for TCF4. However, although RNF43 exhibited autoubiquitination activity as described earlier (7), ubiquitination of TCF4 by RNF43 was not detectable (Fig. 3C). In addition, the protein abundance of TCF4 did not change upon overexpression of RNF43 (Fig. 3D). Together, these data suggest that Wnt signaling is not impaired as a result of RNF43-mediated ubiquitination and proteasomal degradation of TCF4.

To analyze more closely the interaction between RNF43 and TCF4, we performed coimmunofluorescence staining of wild-type RNF43 or mutant RNF43H292R and TCF4. TCF4 localizes to the chromatin as punctate structures within the nucleoplasm. Notably, overexpression of RNF43 resulted in an obvious relocalization of TCF4 from the nucleoplasm toward the nuclear rim, where it colocalized with RNF43 (Fig. 3E). In contrast, mutant RNF43H292R did not change the localization pattern of TCF4. Although the N-terminal deletion mutant RNF43(245–783), which is devoid of the putative transmembrane domain in the N terminus, still interacted with TCF4 (Fig. 3B), overexpression of this truncated protein did not change the localization pattern of TCF4 (fig. S6E) but rather colocalized with TCF4 in the nucleoplasm. In addition, the mutant RNF43(245–783) was not able to inhibit Wnt signaling (fig. S6F). When the NLS were mutated, the inhibitory capacity of RNF43 was abolished (fig. S6G), confirming that subcellular localization of RNF43 highly affects its function as a repressor of TCF transcriptional activity. To corroborate this mechanism under endogenous conditions, we analyzed whether depletion of RNF43 in HT-29 cells led to changes in TCF4 localization. In control cells expressing RNF43, TCF4 expression could be observed in a perinuclear pattern (fig. S6H), albeit less pronounced than when overexpressing RNF43. This perinuclear pattern was lost when RNF43 was knocked down, supporting our proposed mechanism of RNF43-mediated TCF repression.

The relocalization of TCF4 to the nuclear membrane induced by RNF43 did not disrupt the transcription complex because the binding of CTBP1 (C-terminal binding protein 1) to TCF4 was still detected (Fig. 3F). However, the mislocalization of TCF4 to the nuclear membrane resulted in a reduced mRNA expression of its reported target genes encoding Axin2, Twist, LGR5 (leucine-rich repeat–containing G protein–coupled receptor 5), and MMP7 (matrix metalloproteinase-7) (Fig. 3G and fig. S6I), suggesting that RNF43 sequesters TCF4 transcription complexes to sites of silent chromatin at the nuclear periphery.

Mutations in RNF43 disrupt the interaction with TCF4, abolishing its Wnt inhibitory capacity

The RNF43 gene has been shown to exhibit mutations in several forms of neoplastic cysts in the pancreas (11, 18), in ovarian cancer (12), in cholangiocarcinoma (13), and in colorectal adenocarcinomas (14). These mutations are distributed over the entire open reading frame, as depicted in the graphical overview showing all mutations annotated so far (fig. S7A).

To determine the impact on Wnt signaling, we tested some of the mutations in functional assays. Specifically, we cloned two stop mutations (R113X and S216X) and three mutations leading to amino acid substitutions (R127P, A169T, and R343H) observed in pancreatic tumor samples (11) or annotated in the Cosmic database. Mutation R113X was recently reported as a germline mutation in subjects diagnosed with multiple sessile serrated adenomas (9). When measuring TCF transcriptional activity in HCT116 cells after transfection, four of the mutated RNF43 variants (R113X, S216X, R127P, and A169T) could not inhibit Wnt signaling anymore (Fig. 4A), whereas the R343H variant was still active. Mutations R113X and R127P induced relocation of RNF43 into the cytoplasm, whereas A169T, S216X, and R343H did not influence the localization of the protein at the nuclear rim (Fig. 4B). Moreover, we did not observe colocalization between TCF4 and the different RNF43 mutants, with the exception of R343H (Fig. 4C). Finally, when performing coimmunoprecipitation for the mutants R127P or A169T and TCF4, no interaction was detectable (Fig. 4D). Thus, our findings indicate that mutations in RNF43 abolish its Wnt inhibitory capacity and confirm that inhibition of TCF4 transcriptional activity involves mislocalization of TCF4 to the nuclear membrane, which can be disturbed by certain RNF43 mutations.

Fig. 4 Mutations in RNF43 abolish its Wnt inhibitory capacity.

(A) TOP/FOP luciferase reporter assays in HCT116 cells transfected with wild-type or mutant (H292R, R113X, S216X, R127P, A169T, or R343H) RNF43. Results were normalized to Renilla luciferase values (n = 3 experiments). *P < 0.05, ANOVA with Dunnett’s multiple comparsion posttest. (B) Confocal immunofluorescence analysis of wild-type or mutant RNF43 (red) and β-catenin (green) in HCT116 cells. Scale bars, 10 μm. (C) Confocal immunofluorescence analysis of TCF4 (green) and RNF43 (red) in HCT116 cells cotransfected with TCF4 and wild-type or mutant RNF43. Scale bars, 10 μm. (D) Immunoprecipitation of TCF4 from HCT116 cells transfected with Flag-tagged wild-type or mutant RNF43, followed by Western blot analysis.


The function and localization of the RING finger ubiquitin ligase RNF43 are debated. It has been described as an oncogene (8, 10, 19) as well as a tumor suppressor (4, 12, 20), being localized either in the nucleus (7, 8, 21, 22) or at the cell membrane (4, 6, 23). Given its increasingly recognized role in human gastrointestinal cancer, we conducted a thorough analysis of RNF43 expression, localization, and function under endogenous and overexpression conditions in Xenopus, mouse, and human tissue as well as colorectal cancer tissue.

Our immunohistochemical staining of normal human intestine and colon cancer tissue demonstrated nuclear localization of endogenous RNF43. This result was corroborated by subcellular fractionation experiments and immunofluorescence, which unequivocally showed endogenous and overexpressed RNF43 to be located in the nucleus. Here, RNF43 was localized at the nuclear envelope and, to a lesser extent, in the nucleoplasm. Mutagenesis of the two NLS in the RNF43 gene abrogated nuclear localization. These findings are consistent with previous reports showing that RNF43 is located in the ER and in the nuclear membrane with occasional staining in the nucleoplasm (7, 8), as well as with studies linking RNF ubiquitin ligases, including RNF43, to DDR signaling and chromatin remodeling (9, 24). Recently, RNF43 was reported to interact with the nucleoprotein of influenza A virus and modulate p53-dependent signaling and apoptosis (22). Together, we postulate that RNF43 might have diverse functions depending on its subcellular localization, but seems to be primarily located in the nucleus.

RNF43 was described to inhibit Wnt signaling in colon cancer cells by selectively ubiquitinating FZD receptors and targeting them for degradation (4). This mechanism could not explain our findings that RNF43 was located in the nucleus and was able to inhibit Wnt signaling downstream of β-catenin. We found that RNF43 physically interacts with TCF4 and leads to translocation of TCF4 to the nuclear membrane, resulting in inhibition of TCF4 transcriptional activity. Increasing evidence points toward subnuclear localization as an important regulator of transcriptional activity. Inner nuclear membrane proteins have been shown to sequester transcription factors, limiting their transactivation or transrepression abilities. For example, c-fos sequestering to the nuclear envelope by lamin A or C negatively regulates AP-1 (activating protein 1), favoring cellular quiescence (25), whereas interaction of the inner nuclear membrane protein emerin prevents nuclear accumulation of β-catenin and inhibits its transcriptional activity (26). Moreover, different nuclear pore complex proteins were shown to SUMOylate or deSUMOylate TCF4, thereby enhancing or inhibiting its transcriptional activity and thus regulating Wnt signaling in colon cancer cells (27). We therefore propose a model (fig. S7B) in which RNF43 prevents TCF4 transcriptional activity by sequestering the TCF4 transcription complex from the nucleoplasm, and this mechanism is dependent on direct interaction with, but not ubiquitination of, TCF4 by RNF43.

The novel mechanism of RNF43-mediated inhibition of Wnt signaling at the level of TCF4 described here explains how RNF43 can inhibit this pathway in the presence of activating mutations in APC or β-catenin. This is also compatible with the findings that in the absence of constitutive activating mutations in the Wnt pathway, increased Wnt signaling activity resulting from deletion of RNF43 can be reduced by porcupine inhibitors (28, 29). In contrast, transactivating mutations in RNF43, such as the H292R mutation we identified, would not respond to porcupine inhibitors, supporting that such enhancement of Wnt signaling is not induced by increased autocrine Wnt signaling.

When studying the functional impact of some mutations found in different human cancers on Wnt signaling, we observed that somatic mutations in RNF43 as well as the reported germline mutation R113X (9) reverted the Wnt inhibitory capacity of RNF43. Thus, these mutations may represent a selection advantage for colon cancer cells. Most interestingly, we observed that relocalization of TCF4 to the nuclear membrane was not detected when analyzing RNF43 mutants, indicating that inhibition of TCF4 transcriptional activity by RNF43 indeed involves its tethering to the nuclear envelope. Thus, our data not only reveal a novel functional mechanism of RNF43-mediated inhibition of the Wnt pathway but also provide an explanation for the loss of this Wnt inhibitory capacity observed for RNF43 mutants found in human cancers.

The expression of RNF43 mRNA was highly heterogeneous in tumor stage II colorectal cancer samples, reflecting the well-described heterogeneity of this patient cohort (30). We observed a significant positive correlation between RNF43 mRNA expression and relapse after surgery, suggesting that high RNF43 expression is a biomarker for poor prognosis in stage II colorectal cancer. This finding seemed counterintuitive, given the fact that on one hand, canonical Wnt activity is correlated with disease aggressiveness in colorectal cancer (31), and on the other hand, RNF43 inhibits Wnt signaling and is considered a tumor suppressor (4, 6). One of the main causes of colon cancer recurrence is the presence of colon cancer stem cells resistant to chemotherapy (32). In light of our results, it is tempting to speculate that RNF43 mutation might be an early event predominantly occurring in cancer stem cells, which prevents the inhibitory effect on Wnt signaling, and might confer cells an early proliferative or survival advantage. In turn, abnormal Wnt signaling would enhance the expression of mutant RNF43, creating a feedback loop promoting tumor development. Thus, patients presenting high RNF43 mRNA expression and low survival might harbor RNF43 mutations.

Therapeutic inhibition of the Wnt signaling pathway has been hampered by the fact that most substances act upstream of the oncogenic mutations found within the destruction complex or β-catenin itself. To our knowledge, our study is the first to identify a tumor suppressor acting directly by preventing TCF4 transcriptional activity that, more importantly, acts in the presence of constitutive activation of Wnt signaling found in many tumors. Our findings may pave the way for the development of novel therapeutic strategies in cancers arising from mutation of the Wnt pathway.



RNF43-Flag, RNF43-hemagglutinin (HA), and RNF43-2×HA-3×Flag plasmids were generated by PCR subcloning. The mutant RNF43H292R construct was generated by site-directed mutagenesis using internal primers to generate mutations in amino acids 292 and 295 of the RNF43 gene. Plasmid derivatives encoding N-terminal truncation mutants were also generated by PCR subcloning.

TOP and FOP plasmids were provided by M. van de Wetering; expression plasmids for Dishevelled, Axin2, β-TrCP, and S33-β-catenin, provided by W. de Lau, express the dominant-active forms of the respective proteins. TCF4 binding sites within the RNF43 gene identified by ChIP were PCR-amplified and cloned into pGL3b enhancer vector.

The pCMV-Renilla plasmid, which was used to evaluate the transfection efficiency, was purchased from Promega. pCR3.1-his-ubiquitin was provided by K. Sabapathy (Laboratory of Molecular Carcinogenesis, National Cancer Centre, Singapore). TCF4-HA was provided by C.W. Wu (Graduate Institute of Life Sciences, National Defense Medical Center, Taiwan), and the plasmid pCR3.1-PSF(HA) was provided by P. Tucker (Department of Molecular Genetics and Microbiology, University of Texas, Austin, TX). A genomic sequence encompassing 824 base pairs (bp) of DNA around the most intense intronic TCF4 binding site in the RNF43 gene [located at chr17:53,823,655–53,824,478 (hg18 assembly on UCSF genome browser)] was cloned in front of a minimal fragment encompassing the TATA box of the adenovirus major late promoter cloned in front of the firefly luciferase gene in the pGL4.10 reporter vector.

The RNF43 NLS mutant RNF43(R437A-K655A) as well as the RNF43 mutant forms RNF43R113X, RNF43S216X, RNF43R127P, RNF43A169T, RNF43R219H, and RNF43R343H were generated by site-directed mutagenesis using the QuikChange Lightning Site-Directed Mutagenesis Kit (Agilent Technologies). NLS were predicted by PSORTII (, cNLS mapper (, and NucPred software (

Cell culture, transfections, and expression studies

All cell lines used in this study were purchased from American Type Culture Collection (ATCC) and regularly tested for the presence of mycoplasma. Cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (Invitrogen) supplemented with 10% fetal calf serum (Sigma-Aldrich) and maintained in a humidified atmosphere of 5% CO2 at 37°C. Plasmid transfection was performed using Lipofectamine 2000 (Invitrogen) following the manufacturer’s instructions.

LS174T-pTER-β-catenin cell line carrying a doxycycline-inducible short hairpin RNA (shRNA) against β-catenin (33) was used for the expression studies. The cells were grown in the presence or absence of doxycycline (1 μg/ml) for 72 hours before RNA isolation.

shRNA-transduced cells

Vectors were constructed using standard cloning procedures. RNF43 shRNA sequence: TTCTTGGTAAGATCGAGAG and RNF43 shcontrol sequence: GTACAGCCGCCTCAATTCT were used. All shRNA oligonucleotides were purchased from Sigma-Aldrich. The oligonucleotides were annealed and cloned into lentiviral pLVTHM plasmid using the Mlu 1–Cla 1 sites (Addgene, plasmid 12247) (34). Viral particles were produced in HEK293T cells (ATCC CRL-3216). Briefly, 293T cells were seeded in DMEM containing 10% fetal bovine serum. psPAX2, pMD2.g, and the lentiviral vector pLVTHM containing both the green fluorescent protein reporter gene and the shRNA sequence were transfected in the packaging cell line by calcium phosphate precipitation. Transductions were carried out in the presence of polybrene (8 μg/ml) (Sigma).

RNF43 small interfering RNA knockdown

A human RNF43–targeted small interfering RNA (siRNA) kit (SR310324, OriGene) was used to knock down RNF43 abundance. The siRNAs were introduced into HT-29 (ATCC HTB-38) cells using a standard protocol (4D-Nucleofector Protocol for HT-29 cells, Lonza).

Luciferase assays

TOP/FOP luciferase assays were performed using the Dual-Luciferase Reporter Assay Kit (Promega) according to the manufacturer’s instructions.

Quantitative PCR

Total RNA was isolated from cultured cells using the GenElute Mammalian Total RNA Miniprep Kit (Sigma-Aldrich) and was reverse-transcribed with Moloney Murine Leukemia Virus Reverse Transcriptase RNase H− Point Mutant (Promega) according to the manufacturer’s instructions. The KAPA SYBR FAST qPCR Universal Master Mix (Peqlab) was used, and transcript abundance was assessed using the LightCycler 480 (Roche Applied Science) and Bio-Rad CFX384 systems. The sequences of the primers used are as follows: RNF43, sense: 5′-CCTGTGTGTGCCATCTGTCT-3′ and antisense: 5′-GCAAGTCCGATGCTGATGTA-3′; GAPDH, sense: 5′-GAAGGTGAAGGTCGGAGT-3′ and antisense: 5′-GAAGATGGTGATGGGATTTC-3′; LGR5, sense: 5′- TGATGACCATTGCCTACAC-3′ and antisense: 5′- GTAAGGTTTATTAAAGAGGAGAAG-3′; AXIN2, sense: 5′-ATGGGATGATCTGTTGCAGAGGGA-3′ and antisense: 5′-TGTCATTTCCACGAAAGCACAGCG-3′; TWIST, sense: 5′-CATGTCCGCGTCCCACTAG-3′ and antisense: 5′-TGTCCATTTTCTCCTTCTCTGG-3′; MMP7, sense: 5′-AAACTCCCGCGTCATAGAAAT-3′ and antisense: 5′-TCCCTAGACTGCTACCATCCG-3′. The PCR conditions used were 40 cycles of amplification with 15-s denaturation at 95°C and 1-min annealing and amplification at 60°C. Gene expression analysis was performed using the comparative ΔΔCT method with the housekeeping gene GAPDH for normalization. Template-free and reverse transcriptase–free negative controls were included in each experiment.

For quantitative determination of RNF43 expression in tumor samples, tumor tissue from patients with histopathologically confirmed colon cancer (stage II and III), who underwent complete tumor resection (R0) at our surgical department between 1987 and 2005, was obtained after approval from the ethics committee of the Klinikum rechts der Isar (35). RNA was isolated using the Qiagen AllPrep DNA/RNA Mini Kit according to the manufacturer’s protocol. RNA integrity was evaluated by denaturing gel electrophoresis and spectrometric measurement (ND-1000, Thermo Fisher Scientific). All samples presented high RNA integrity corresponding to an RIN (RNA integrity number) value ≥5. Retrotranscription was performed as previously described (36). Maxima SYBR Green/ROX qPCR 2× Master Mix (Fermentas) was used for PCR, and data collection was performed in an Applied Biosystems 7900HT cycler. For each sample, 20 to 30 mg of frozen tumor tissue were used for histology-guided sample selection (37) to ensure a sufficient amount of tumor cells (more than 30%).

Immunoprecipitation and Western blot analysis

For coimmunoprecipitation experiments, cells were lysed in lysis buffer containing 20 mM tris (pH 7.5), 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, and phosphatase and protease inhibitors. Cleared cell lysates were incubated overnight at 4°C with antibodies to TCF4 (Cell Signaling), HA (Sigma-Aldrich), and mouse or rabbit immunoglobulin G (IgG) (Santa Cruz Biotechnology) as a negative control, according to the manufacturer’s instructions. Protein A/G agarose beads were added and incubated for 4 hours at 4°C. The beads were collected by centrifugation and washed five times with lysis buffer, and the bound proteins were eluted in SDS sample buffer for Western blot analysis.

Protein samples were resolved by SDS–polyacrylamide gel electrophoresis, transferred onto nitrocellulose membranes, and incubated with primary antibodies overnight at 4°C after blocking with 5% low-fat milk. Secondary antibodies conjugated with horseradish peroxidase (Promega) were used for signal visualization with Pierce ECL Western Blotting Substrate (Thermo Fisher Scientific). The antibodies used were TCF4 (Cell Signaling), Flag M2 (Sigma-Aldrich), HA (Sigma-Aldrich), β-catenin (BD Transduction Laboratories), nonphosphorylated β-catenin (Cell Signaling), and β-actin (Sigma-Aldrich).

ChIP assays

ChIP was performed as previously described (38). In brief, LS174T cells were cross-linked with 1% formaldehyde for 30 min at room temperature. The reaction was terminated with glycine at a final concentration of 125 mM. The cells were washed for 10 min each with phosphate-buffered saline (PBS), buffer B [0.25% Triton X-100, 10 mM EDTA, 0.5 mM EGTA, 20 mM Hepes (pH 7.6)], and buffer C [0.15 M NaCl, 1 mM EDTA, 0.5 mM EGTA, 20 mM Hepes (pH 7.6)] at 4°C. The cells were then resuspended in ChIP incubation buffer [0.3% SDS, 1% Triton X-100, 0.15 M NaCl, 1 mM EDTA, 0.5 mM EGTA, 20 mM Hepes (pH 7.6)] and sheared using Covaris S2 (Covaris) for 10 min with the following settings: duty cycle, max; intensity, max; cycles/burst, max; mode, power tracking. The sonicated chromatin was incubated for 12 hours at 4°C with the appropriate antibody [polyclonal antibody against TCF4 (sc-8631) or polyclonal antibody against β-catenin (H-102) (sc-7199, Santa Cruz Biotechnology)] at 1 μg of antibody per 106 cells with 70 μl of protein G beads (Millipore). The beads were successively washed two times with buffer 1 [0.1% SDS, 0.1% deoxycholate, 1% Triton X-100, 0.15 M NaCl, 1 mM EDTA, 0.5 mM EGTA, 20 mM Hepes (pH 7.6)], one time with buffer 2 [0.1% SDS, 0.1% sodium deoxycholate, 1% Triton X-100, 0.5 M NaCl, 1 mM EDTA, 0.5 mM EGTA, 20 mM Hepes (pH 7.6)], one time with buffer 3 [0.25 M LiCl, 0.5% sodium deoxycholate, 0.5% NP-40, 1 mM EDTA, 0.5 mM EGTA, 20 mM Hepes (pH 7.6)], and two times with buffer 4 [1 mM EDTA, 0.5 mM EGTA, 20 mM Hepes (pH 7.6)] for 5 min each at 4°C. The precipitated chromatin was eluted by incubation of the beads with elution buffer (1% SDS, 0.1 M NaHCO3) at room temperature for 30 min. The immunoprecipitated chromatin was de–cross-linked by incubation at 65°C for 5 hours in the presence of 200 mM NaCl, extracted with phenol-chloroform, and precipitated with ethanol. The following primers were used: huRNF43ChIPfor1, TAACCAGACCCGTTCTTTCGCTCT (promoter); huRNF43ChIPrev1, TTGAGTGGCCCTGATTTGCATGTG; huRNF43ChIPfor2, TGGAGTCCTTCCAAAGCACCTCAT (intronic enhancer); huRNF43ChIPrev2, TCCTCTGACTTTCTTGCTTGGCCT; huAXIN2ChIPfor1, GCCAGAGTCAAGCCAGTAGTC (negative control region); huAXIN2ChIPrev1, TAGCCTAATGTGGAGTGGATGTG.

Subcellular fractionation

Cells were lysed in lysis buffer (250 mM sucrose, 20 mM Hepes, 10 mM KCl, 1.5 mM MgCl2, 1 mM EDTA, 1 mM EGTA, 1 mM dithiothreitol) containing protease inhibitors. Lysates were passed through a 25-gauge needle and incubated on ice for 20 min. After centrifuging at 3000 rpm for 5 min at 4°C, the nuclear pellet was washed in lysis buffer, whereas the supernatant fraction was centrifuged at 15,000 rpm for 30 min at 4°C to obtain the crude cytosolic and membrane fractions. The nuclear pellet was again passed through a 25-gauge needle and finally centrifuged at 3000 rpm for 10 min. The pellet was resuspended in radioimmunoprecipitation assay (RIPA) buffer containing 10% glycerol and 0.1% SDS and sonicated briefly on ice before Western blot analysis.

Ubiquitination analysis

Cells were treated with 10 μM MG132 (N-carbobenzyloxy-l-leucyl-l-leucyl-l-leucinal; Sigma-Aldrich) for 12 hours before cell lysis in RIPA buffer [50 mM tris-HCl (pH 7.4), 150 mM NaCl, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS]. SDS was added to cell lysates to a final concentration of 1%. The lysates were boiled for 5 min (95°C) before NP-40 buffer was added to each tube.

In situ hybridization

Paraffin-embedded human tissue samples were obtained from the tissue bank of the Institut für Pathologie, Klinikum Bayreuth, Germany. Digoxigenin-labeled (Roche) antisense and appropriate sense control RNA probes were generated from cDNA-containing vectors by in vitro transcription using T3/T7 and T7/SP6 RNA polymerase (Promega). The following probes were used: OLFM4 [(GenBank accession no. NM_006418.4, nucleotides 19 to 1886; IMAGE ID #40125880, provided by H. Clevers (Utrecht)] and RNF43 (NM_017763.4, nucleotides 3824 to 4402). For the detection of murine Rnf43, a 1629-bp fragment of mouse Rnf43 cDNA (NM_172448) was cloned into pCMV-SPORT6 vector (Invitrogen). In situ hybridization was performed as previously described (39). Whole embryos were obtained at different embryonic stages, fixed in 3.7% paraformaldehyde at 4°C overnight, and subsequently embedded in paraffin. A mouse Rnf43 gene fragment (1567 bp) was amplified from mouse cDNA (NM_172448, nucleotides 2701 to 4267) and cloned into pGEM-T vector (Promega).

Whole-mount in situ hybridizations were performed as previously described (40). All embryos were stained using BM Purple AP Substrate (Roche) according to the manufacturer’s instructions. The incubation time until development of strong staining was about 1 to 2 days. Afterward, the embryos were postfixed with 4% paraformaldehyde in PBS. In all cases, sense probes were run in parallel, and all of them were negative.

Immunofluorescence analysis

For immunofluorescence stainings, cells were grown on coverslips. After transfection, the cells were incubated for 24 to 48 hours at a humidified atmosphere of 5% CO2 at 37°C before being fixed with ice-cold 1:1 acetone/methanol for 15 min. After being washed with PBS, the samples were incubated with blocking and permeabilization solution (3% bovine serum albumin, 1% saponin, and 0.5% Triton X-100 in PBS) for 10 min at room temperature and with primary antibodies overnight at 4°C. Goat anti-rabbit–Alexa Fluor 594 and chicken anti-rabbit–Alexa Fluor 488 (Sigma-Aldrich) were used as secondary antibodies. The cells were mounted in Vectashield mounting medium containing 4′,6-diamidino-2-phenylindole (DAPI) (Vector Laboratories) and visualized using a Leica SP5 confocal microscope. Signal intensities across the transverse axis of the cell were calculated using Fiji software and the Profile plot program.


Paraffin-embedded human tissue samples were obtained from the tissue bank of the Institut für Pathologie, Klinikum Bayreuth, Germany. Heat-induced antigen retrieval was performed using 10 mM sodium citrate (pH 6), and endogenous RNF43 was detected with a specific antibody to RNF43 (1:1000; Atlas Antibodies AB).

Preparation of synthetic RNA and microinjection of X. laevis embryos

Because the X. laevis RNF43 sequence is not available and high conservation among species has been observed, human RNF43 cDNA sequence was used as a template for cloning RNF43 into pCS2myc. Capped mRNAs were transcribed in vitro from linearized DNA templates using the mMESSAGE mMACHINE kit (Ambion). WNT1 mRNA (5 pg) (41) was co-injected with 500 pg of RNF43 or RNF43H292R mRNA in a total volume of 4 nl into the marginal zone of both ventral blastomeres of four-cell stage embryos. Embryos were staged according to Nieuwkoop and Faber (42) and scored for the appearance of secondary body axes.

Statistical analysis

Results are presented as mean ± SD of three independent experiments. Statistical analysis of normally distributed data was performed using t test or ANOVA with Dunnett’s multiple comparsion posttest, whereas non-normally distributed data were compared using Kruskal-Wallis ANOVA. Statistical significance was set at P ≤ 0.05.

Statistical evaluation of RNF43 expression in tumor samples was performed using the IBM SPSS Statistics version 19 (SPSS Inc.). Recurrence-free survival was considered as the primary end point. To derive an optimal cutoff value of RNF43 gene expression level, maximally selected log-rank statistics with R software version 2.13.0 (R Foundation for Statistical Computing) was used. To consider multiple test issue within these analyses, the R function maxstat.test was used (43). Time-dependent survival probabilities were estimated with the Kaplan-Meier method, and the log-rank test was used to compare independent subgroups. The Monte Carlo analysis for consistency testing was used as previously described (35).


Fig. S1. RNF43 expression.

Fig. S2. RNF43 subcellular localization.

Fig. S3. RNF43 is heterogeneously expressed in stage II colon tumors.

Fig. S4. RNF43 is a Wnt target gene.

Fig. S5. RNF43 inhibits Wnt signaling in vivo.

Fig. S6. Mode of action of RNF43-mediated Wnt inhibition.

Fig. S7. Model of RNF43 mode of action and RNF43 mutations.

Fig. S8. RNF43 overexpression does not affect cell viability.


Acknowledgments: We thank H. Clevers for the critical discussion; M. van de Wetering and W. de Lau for providing plasmids; M. Hrabe de Angelis for help with mouse embryo in situ hybridization. Funding: This work was supported by a grant from Deutsche Forschungsgemeinschaft (GE 2042/2-1) to M. Gerhard. Author contributions: A.L. and M. Grandl performed experiments, analyzed the data, and contributed to the writing of the manuscript. M.A. performed in situ hybridization. K.D. cloned and analyzed RNF43 mutants. V.H., M.N., R.R., and I.V. performed exome sequencing and bioinformatics analysis. C. O. and P.H. performed ChIP and luciferase reporter assays with RNF43 promoter. K.-P.J. and U.N. provided tumor samples and performed statistical analysis of clinical data. D.G., O.v.d.B., and O.D. performed Xenopus experiments. B.K. generated plasmids. A.J. provided tumor samples. R.M.S. and D.H.B reviewed the manuscript. K.U. performed statistical analysis. R.M.-L. and M. Gerhard conceived and designed the experiments, analyzed the data, and wrote the manuscript. Competing interests: The authors declare that they have no competing interests.
View Abstract

Navigate This Article