RNF43 inhibits WNT5A-driven noncanonical Wnt signaling pathway

In order to test whether or not RNF43/ZNRF3 controls noncanonical Wnt signaling, we have decided to study T-REx 293 cells. T-REx 293 cells secrete endogenous WNT5A that constitutively activates the noncanonical Wnt pathway – this can be demonstrated by the CRISPR/Cas9-mediated knockout of WNT5A (Kaiser et al., 2020). Removal of endogenous WNT5A in T-REx 293 cells is sufficient to eliminate the activation of readouts of WNT5A signaling such as phosphorylation of ROR1, DVL2, and DVL3 that can be monitored by western blotting as the decrease in the phosphorylation-mediated electrophoretic mobility shifts (Figure 1A; Bryja et al., 2007b; Kotrbová et al., 2020; Radaszkiewicz and Bryja, 2020). Autocrine WNT5A signaling was promoted by the inhibition of endogenous RNF43/ZNRF3 by RSPO1 treatment (Figure 1B, compare lanes 1 and 2) and inhibited by RNF43 overexpression under the tetracycline (Tet)-controlled promoter (TetON) or by block of WNT secretion using Porcupine inhibitor Wnt-C59 (Figure 1B). To confirm that the effects are indeed caused by the block of WNT5A signaling, T-REx 293 cells pretreated with Wnt-C59 and as such unable to produce Wnt ligands were stimulated with increasing doses of recombinant WNT5A. As shown in Figure 1C, overexpression of RNF43 completely blocked signaling induced by recombinant WNT5A. Altogether, this demonstrates that RNF43 has the potential to block WNT5A signaling in mammalian cells.

RNF43 physically interacts with key proteins from the noncanonical Wnt pathway

To address the molecular mechanism of RNF43 action in the noncanonical Wnt pathway, we decided to describe RNF43 interactome by the proximity-dependent biotin identification (BioID; Roux et al., 2012), which was already successfully applied in the challenging identification of E3s substrates (Coyaud et al., 2015; Deshar et al., 2016). We have exploited our recently published dataset (Spit et al., 2020) based on T-REx 293 TetON cells that inducibly expressed RNF43 fused C-terminally (intracellularly) with BirA* biotin ligase. Several core proteins of the noncanonical Wnt signaling pathway – namely, ROR1, ROR2, VANGL1, VANGL2, SEC24B, and all three isoforms of DVL – were strongly and specifically biotinylated by RNF43-BirA* (Figure 1D and D′, Figure 1—source data 1). Furthermore, the noncanonical Wnt pathway was significantly enriched also in the gene ontology (GO) terms (Figure 1—source data 2). Altogether, it suggests that RNF43 can at least transiently interact with multiple proteins involved in the Wnt/PCP pathway, including essential receptor complex components from the ROR, DVL, and VANGL protein families.

To validate the protein-protein interactions identified by BioID, we performed a series of co-immunoprecipitation (co-IP) and co-localization experiments (Figure 2, Figure 2—figure supplement 1). We have focused on the interactions of RNF43 with ROR1/ROR2 and with VANGL1/VANGL2 mainly because these interactions are novel and at the same time highly relevant for the noncanonical Wnt pathway. RNF43 co-immunoprecipitated with both VANGL2 (Figure 2A) and VANGL1 (Figure 2—figure supplement 1A). More detailed analysis of VANGL2 showed co-localization of VANGL2 and RNF43 in the cell membrane (Figure 2B and B′). RNF43 also efficiently pulled down ROR1 (Figure 2C) and ROR2 (Figure 2—figure supplement 1B). Deletion of the cysteine-rich domain (CRD) (ROR2, Figure 2—figure supplement 1B) had no impact on the amount of co-immunoprecipitated RNF43, which suggests that RNF43 primarily interacts with RORs intracellularly. Both ROR1/ROR2 co-localized with RNF43 at the level of the plasma membrane (Figure 2D and D′ and Figure 2—figure supplement 1C and C′). It was described that RORs and VANGLs also bind DVL (Gao et al., 2011; Mentink et al., 2018; Seo et al., 2017; Witte et al., 2010; Yang et al., 2017) and at the same time DVL proteins mediate ubiquitination of FZD receptors by RNF43 in the Wnt/β-catenin pathway (Jiang et al., 2015). To address whether DVL also acts as a physical link between RNF43 and the analyzed PCP proteins, we performed the co-IP experiments with VANGL2 and ROR1 in the T-REx 293 cells lacking all free DVL isoforms (DVL triple knockout cells) (Paclíková et al., 2017). As shown in Figure 2—figure supplement 1D and E, RNF43 was able to bind both VANGL2 and ROR1 as efficiently as in the wild-type cells (compare with Figure 2A and D). In summary, our results indicate that RNF43 interacts, in a DVL-independent way, with PCP proteins from VANGL and ROR families.

RNF43 ubiquitinates VANGL2 and triggers its degradation

Since RNF43 is an E3 ubiquitin ligase, we next tested whether it can ubiquitinate its binding partners from the noncanonical Wnt pathway. Enzymatically inactive RNF43 Mut1 variant (Koo et al., 2012) served here as a negative control. Using His-ubiquitin pulldown assay, we were able to show that VANGL2 (Figure 3A), as well as DVL1 and DVL2 (Figure 3—figure supplement 1A), was ubiquitinated when co-expressed with RNF43 but not with RNF43 Mut1. However, we were unable to detect RNF43-induced ubiquitination of ROR1 or ROR2 (negative data, not shown).

Further analysis showed that overexpression of RNF43, but not its E3 ligase dead variant, decreased VANGL2 protein level (Figure 3B, quantified in Figure 3—figure supplement 1B). Decrease in VANGL2 caused by RNF43 was accompanied by impeded phosphorylation of ROR1 (Figure 3B) and DVL3 (Figure 3B, Figure 3—figure supplement 1B). On the other side, two independent clones of cells deficient in both RNF43 and ZNRF3 (RNF43/ZNRF3 dKO; R/Z dKO) showed higher VANGL2 levels and higher DVL phosphorylation (Figure 3B, Figure 3—figure supplement 1B). These confirm that also endogenous RNF43/ZNRF3 module affects the noncanonical Wnt signaling. Interestingly, treatment with the proteasome inhibitor MG132 but not with autophagosome-lysosome inhibitor chloroquine blocked these effects of RNF43 (Figure 3C). This suggests that RNF43 action in the noncanonical Wnt pathway depends on the proteasomal degradation pathway, which differs from the Wnt/β-catenin pathway, where RNF43 triggers FZD degradation via the lysosomal pathway (Koo et al., 2012). This is in accordance with immunofluorescence analysis, which did not show an increased co-localization of VANGL2 and RNF43 in lysosomes (Figure 3—figure supplement 2B). However, RNF43 overexpression led to the retention of VANGL2 in the Golgin-97-positive area (Figure 3—figure supplement 2C). This suggests that RNF43 can act via similar mechanism that was described for WNT5A-induced regulation of VANGL2 plasma membrane localization during planar cell polarity maintenance (Feng et al., 2021; Guo et al., 2013; Tower-Gilchrist et al., 2019; Yang et al., 2017).

RNF43 induces ROR1 endocytosis by a clathrin-dependent pathway

ROR1 and ROR2 are the key receptors for WNT5A that we found to interact with RNF43 (Figures 1 and 2). We thus speculated that RNF43 can regulate ROR1/ROR2 surface levels. T-REx cells express dominantly ROR1, and indeed flow cytometric analysis demonstrated that cell lacking endogenous RNF43 and ZNRF3 have more ROR1 receptor on the surface than parental T-REx cells (Figure 3D). The staining is specific as demonstrated by the validation of the ROR1-APC antibody in ROR1 KO T-REx 293 cells (Figure 3—figure supplement 1C and D). When we introduced inducible RNF43 into RNF43/ZNRF3 dKO T-REx cell line, we were able to rescue this phenotype, and after 3 hr of tetracycline treatment, we detected decreased surface ROR1 and the overnight exposition to tetracycline had no significant effect (Figure 3E and E′).

In our analysis of RNF43 interactors (Figure 1D), we identified also multiple proteins involved in endosomal transport. It included proteins involved in the clathrin endocytic pathway – STAM1, HRS, ZFYVE16, PICALM, NUMB, RAB11-FIP2, and subunits of the associated adaptor protein complexes AP-3 and AP-4 (Supplementary file 1; Bache et al., 2003; Cullis et al., 2002; Hirst et al., 2013; Raiborg et al., 2001; Santolini et al., 2000; Seet and Hong, 2005; Tebar et al., 1999). Based on the BioID results analysis, we speculated that RNF43 may promote clathrin-mediated endocytosis of ROR1. Thus, we applied dansylcadaverine to block this pathway (Blitzer and Nusse, 2006). In agreement with our hypothesis, treatment with this inhibitor prevented RNF43-mediated effect on the ROR1 surface expression in T-REx 293 RNF43/ZNRF3 dKO RNF43 TetON cells (Figure 3F).

To get a better insight into the mechanism of RNF43-induced internalization of ROR1, we analyzed the co-localization of ROR1 and RAB5 (marker of early endosomes) and RAB11 (marker of recycling endosomes) in T-REx 293 R/Z dKO RNF43 TetON (Figure 3G and Figure 3—figure supplement 3A) and T-REx 293 RNF43 TetON cells (Figure 3—figure supplement 1E and Figure 3—figure supplement 3B). Hyperactivation of Rab5 by overexpression of wild-type Rab5 leads to the formation of giant early endosomes (Bucci et al., 1992) where we observed ROR1/RAB5 co-localization after 3 hr of tetracycline treatment. The co-localization decreased after overnight exposition to tetracycline. RAB11⁺ endosomes were recruited to the ROR1 as well as after RNF43 induction and RAB11 co-localized strongly with ROR1 even after ON treatment. We conclude that surface ROR1 is controlled by RNF43 via interference with RAB5- and RAB11-mediated endocytosis and RNF43-mediated effect is not persistent due to the activity of recycling RAB11 recycling pathway (Ullrich et al., 1996).

RNF43 expression is decreased in human melanoma

Our data shown in Figures 1—3 demonstrate that RNF43 can inhibit WNT5A-induced noncanonical signaling via downregulation of the receptor complexes. But is RNF43 capable of blocking WNT5A-induced biological processes? WNT5A signaling plays a crucial role in melanoma, one of the most malignant tumor types. High expression of WNT5A in this cancer is a negative overall survival and positive metastasis formation factor (Da Forno et al., 2008; Luo et al., 2020; Weeraratna et al., 2002). Signaling cascade activated by WNT5A in melanoma drives epithelial–mesenchymal transition-like program (EMT), resulting in increased metastatic properties of melanoma cells in vitro and in vivo (Dissanayake et al., 2008; Dissanayake et al., 2007; Sadeghi et al., 2018). In melanoma, WNT5A acts through FZD and ROR1/ROR2 (O’Connell et al., 2010; Tiwary and Xu, 2016; Weeraratna et al., 2002). The importance of WNT5A-driven signaling in melanoma is thus well recognized, and melanoma represents probably the most characterized (and most clinically relevant) pathophysiological condition where noncanonical WNT5A signaling drives cell invasion and disease progression (Arozarena and Wellbrock, 2017b; Da Forno et al., 2008; Dissanayake et al., 2007; Lai et al., 2012; Liu et al., 2018; O’Connell et al., 2010; O’Connell et al., 2008; Weeraratna et al., 2002).

Interestingly, the in silico analysis of gene expression (Talantov et al., 2005; Xu et al., 2008) showed that RNF43 expression dramatically decreases between benign melanocytic skin nevus and cutaneous melanoma (Figure 4A; Talantov et al., 2005) and further between primary site and metastasis (Xu et al., 2008; Figure 4B). Importantly, analysis of other datasets (Anaya, 2016) showed that RNF43 low melanoma patients have shorter overall survival (OS; Figure 4C). ZNRF3 expression had no prognostic value (Figure 4—figure supplement 1A). Interestingly, the expression of two genes encoding direct targets ubiquitinated by RNF43, namely, DVL3 and VANGL1, increased during melanoma progression (Figure 4—figure supplement 1B and C) and high expression in both cases correlates with bad prognosis and shorter overall survival (Figure 4D, Figure 4—figure supplement 1D). All these findings are in line with the hypothesis that RNF43 acts in melanoma as a tumor suppressor that restricts WNT5A-induced biological processes and gets silenced during melanoma progression.

RNF43 inhibits invasive properties of melanoma cells in vitro

A375 and A2058 are human melanoma cell lines carrying BRAF V600E mutations that are broadly used to study WNT5A role in melanoma (Anastas et al., 2014; Connacher et al., 2017; Da Forno et al., 2008; Ekström et al., 2014; Linnskog et al., 2016; Liu et al., 2018). For the purpose of our study, we chose A375 wild-type (A375) cells and their derivate with the increased metastatic potential referred to as A375 IV (Kucerova et al., 2014). Both A375 variants and A2058 express WNT5A, RNF43, and ZNRF3 (Figure 4—figure supplement 2A–C, I–K). A375 and A375 IV secrete WNT5A and WNT11 to the culture medium, whereas A2058 produces only WNT11 (Figure 4E). WNT5A protein levels are higher and WNT11 levels lower in case of A375 IV metastatic derivate in comparison to parental A375 cell line (Figure 4E′ and E′′). Interestingly, RNF43 expression in the A375 IV cells was significantly lower than that in the A375 parental cells (Figure 4—figure supplement 2B). Expression of ZNRF3 did not differ and it was not affected by RNF43 overexpression (Figure 4—figure supplement 2C). Further, A375 line has lower expression of ROR1 than IV derivate (Figure 4—figure supplement 2D). A2058 cells are ROR1 and ROR2 positive with ROR1 level being higher (Figure 4—figure supplement 2L and M).

To study the RNF43 function, we generated A375 cells lacking RNF43/ZNRF3 (R/Z dKO) by CRISPR/Cas9 method (sequencing results are presented in Supplementary file 1), cells stably overexpressing RNF43 (RNF43 OE) as well as A375, A2058, and additionally NRAS^Q61L/WT HRAS^G13D/G13D MelJuso -based doxycycline-inducible RNF43 TetON lines (Figure 4F). The initial characterization of A375 derivatives essentially confirmed the findings from T-REx 293 (see Figure 1), where RNF43 loss- and gain of function correlated strongly with the level of Wnt pathway activation assessed as DVL phosphorylation (Figure 4G, quantified in Figure 4—figure supplement 1E and F). Total protein levels of DVL2, DVL3, as well as expression of their genes remained unaffected by the RNF43 manipulation (Figure 4—figure supplement 2E–H). Similarly to T-REx 293 cells, in A375 (Figure 4H), A375 IV (Figure 4I), and A2058 (Figure 4J and quantified effect of RNF43 OE on the endogenous WNT pathway activity in Figure 4—figure supplement 1G and H) melanoma cells RNF43 overexpression efficiently blocked WNT5A-induced signaling. To extend the importance of our studies further, we tested the RNF43-mediated effect also in the RAS-mutant MelJuso cells. Here we also confirmed the suppression of ROR1, DVL2, and DVL3 shifts by forced RNF43 expression (Figure 4—figure supplement 3A). Inducible model of A375 RNF43 TetON cells drew a similar picture (Figure 4—figure supplement 3B), compared to the control cells that underwent the same procedures (Figure 4—figure supplement 3C).

WNT5A signaling has been related to numerous biological features that support the invasive properties of melanoma (Arozarena and Wellbrock, 2017b; O’Connell and Weeraratna, 2009; Prasad et al., 2015; Weeraratna et al., 2002). To address if RNF43 affects any of these WNT5A-controlled properties, we have compared parental and RNF43-derivative melanoma cells in a panel of functional assays that included (1) wound healing assay, (2) collagen I hydrogel 3D chemotaxis assay, (3) Matrigel invasion assay, (4) invadopodia formation assay, and (5) gelatin degradation assay. Firstly, all A375, A375 IV, and A2078 cells overexpressing RNF43 showed suppressed 2D collective migration in the wound healing assay (Figure 5A–E). Next, we analyzed the impact of RNF43 expression on directional invasion through the collagen I hydrogel in response to chemokines. This assay mimics the taxis of melanoma cells in body: CCL21 drives lymph nodes metastasis and CXCL12 promotes lung invasion (Figure 5F; Jacquelot et al., 2018; McArdle et al., 2016). A375 IV and A2058 cell lines showed significant response to these treatments and RNF43 blocked completely these invasion events (Figure 5G and H). Importantly, the response of A375 cells in this assay was negligible (Figure 5—figure supplement 1B), confirming their decreased metastatic capacity compared to the A375 IV. In line, invasion of individual cells through the extracellular matrix (ECM) mimicking Matrigel was higher in A375 IV in comparison to A375 but in both cases reduced by RNF43 OE (Figure 5—figure supplement 1D). The same is true for the number of invadopodia-specialized structures mediating adhesion and remodeling of the surrounding ECM (Eddy et al., 2017; Masi et al., 2020). A375 IV cells formed more invadopodia than A375 parental cells (Figure 5I) and cells overexpressing RNF43 formed less of them (Figure 5I). In agreement, we also observed reduced gelatin degradation activity in A375 and A375 IV cells overexpressing RNF43 (Figure 5J). Furthermore, treatment with WNT5A enhanced the gelatin degradation capacity of A375 cells, but not their RNF43-overexpressing derivate (Figure 5J). Representative images from the conducted assays are shown in Figure 5—figure supplement 1, Figure 5—figure supplement 2, Figure 5—figure supplement 3, Figure 5—figure supplement 4. All these assays strongly support the conclusion that RNF43 acts as a strong molecular inhibitor of WNT5A-triggered proinvasive features of melanoma. Interestingly, RNF43/ZNFR3 removal did not lead to significant potentiation in these functional assays, which suggests that the noncanonical Wnt pathway-controlled pro-metastatic features of these cells are already close to its maximum. In agreement, treatment with WNT5A showed the effect only in case A375 cells in the gelatin degradation assay (Figure 5J).

RNF43 prevents acquisition of resistance to BRAF V600E targeted therapy

The mitogen-activated protein kinase (MAPK) pathway is hyperactivated in melanoma (Davies et al., 2002) as a result of UV-induced mutations triggering constitutive activation of this signaling axis. The most common genetic aberration – BRAF V600E is a target of anti-melanoma therapy (Akbani et al., 2015; Birkeland et al., 2018; Chapman et al., 2011; Flaherty et al., 2010; Hodis et al., 2012; Shain et al., 2015). Drugs targeting mutated BRAF (e.g., vemurafenib/PLX4032) in melanoma have improved patients’ survival (Chapman et al., 2011; Flaherty et al., 2010; Joseph et al., 2010). Unfortunately, patients receiving BRAF inhibitors (BRAFi) relapse after several months of monotherapy because of the acquired resistance (Nazarian et al., 2010). WNT5A was shown to play a crucial role in the process leading to the vemurafenib resistance (Anastas et al., 2014; Arozarena and Wellbrock, 2019; Mohapatra et al., 2019; O’Connell et al., 2013; Prasad et al., 2015; Webster et al., 2015). Therefore, we were interested in checking whether RNF43 inhibits, via its effects on WNT5A signaling, cellular plasticity in response to vemurafenib (PLX4032), a clinically used BRAF V600E inhibitor.

The process of vemurafenib resistance acquisition can be modeled in vitro. We applied an experimental scheme optimized for A375 (Anastas et al., 2014). This model (Figure 6A) allows to study both acute responses to vemurafenib (24 hr treatment) as well as the gradual adaptation of long-term cell culture to increasing vemurafenib doses. Vemurafenib-resistant (VR) cells can be obtained after approximately 2 months. As shown in Figure 6B, treatment with vemurafenib resulted in rapid and complete inhibition of ERK1/2 phosphorylation, the readout of MAPK activation (compare lanes 1 and 2). In contrast, A375 VR cells showed constitutive ERK1/2 phosphorylation in the presence of 2 μM vemurafenib (compare lanes 2 and 3). Interestingly, transient exposition to vemurafenib resulted in the impeded phosphorylation of ROR1, DVL2, and DVL3 and increased ROR2 protein level without change in its mRNA (Figure 6B–D, Figure 6—figure supplement 1A and B). On the other side, VR cells displayed elevated ROR1 levels (both protein and mRNA) and increased phosphorylation of DVL2 and DVL3 (Figure 6B–D, Figure 6—figure supplement 1A and B). Strikingly, the expression of WNT5A was also higher in the resistant VR cells (Figure 6E). This suggests that activation of noncanonical WNT5A-induced signaling and ROR1 and ROR2 changes is indeed a part of the melanoma adaptation mechanism to vemurafenib. Therefore, we challenged A375 and its RNF43-expressing derivatives with vemurafenib. As shown in Figure 6F, exogenous RNF43 decreased colony formation and proliferation of cells seeded in low density and vemurafenib further enhanced this effect. Importantly, both A375 and A375 IV-overexpressing RNF43 completely failed to develop resistance to vemurafenib and died off during selection at 1 μM vemurafenib concentration (Figure 6G). Both A375 and A375 IV RNF43/ZNRF3 double KO cells survived selection with BRAFi but showed decreased proliferation rate. Lower proliferation of RNF43/ZNRF3 dKO cells can be caused by senescence because these cells have elevated WNT5A pathway activity (i.e., Figures 3B and 4I, Figure 4—figure supplement 1E and F) and WNT5A at the same time causes a senescence-like phenotype of melanoma after exposition to vemurafenib (Webster et al., 2015). Alternatively, RNF43/ZNRF3 KO could affect also canonical Wnt/β-catenin pathway that was shown to drive distinct features of melanoma cells than the ones related to the WNT5A-specific events (Arozarena et al., 2011; Arozarena and Wellbrock, 2019; Uka et al., 2020; Webster and Weeraratna, 2013).

To strengthen our results, we exposed A375 and A2058 parental and RNF43-expressing cells temporarily and chronically to vemurafenib, MEK inhibitor – cobimetinib and to the FDA-approved combination of these drugs (Ascierto et al., 2016; Signorelli and Shah Gandhi, 2017; experimental scheme in Figure 6—figure supplement 1C’). We failed to obtain A375-resistant line, but A2058 acquired resistance to cobimetinib alone and its combination with vemurafenib (Figure 6H and L). Treatments and resistance were validated by pERK1/2 levels (Figure 6H). A2058 enriched with ectopically expressed RNF43 did not survive cobimetinib monotherapy and cobimetinib with vemurafenib combination, whereas vemurafenib-only treatment did not fully eliminate those cells (Figure 6L). Thus, we aimed to detect the differences in mechanisms behind targeted therapies resistance acquisition to characterize better the function of RNF43 in melanoma. Firstly, we noticed the shift of VANGL1 in all delivered resistance models, while VANGL2 was shifted only in cells being in the long-term culture with cobimetinib and with cobimetinib and vemurafenib (Figure 6H). This corresponds with decreased RNF43 and ZNRF3 expression (Figure 6J, Figure 6—figure supplement 1D) and increased WNT5A mRNA level (Figure 6K). Interestingly, both MITF protein level and MITF gene expression increased in VR cells and decreased in all cobimetinib-insensitive lines (Figure 6H and I). Moreover, MITF upregulation was accompanied by clear pigmentation of resistant cells (Figure 6I′), suggesting cell phenotype change (Arozarena and Wellbrock, 2019; Ji et al., 2015; Müller et al., 2014). Expression of ROR1 decreased in case of MEKi resistance, while ROR2 remained unaffected (Figure 6—figure supplement 1E and F). Contrary to A375, ROR1 did not significantly increase in VR-resistant cells and we did not detect the increased DVL2 and DVL3 phosphorylation (Figure 6—figure supplement 1C).

Altogether, these data confirm earlier findings on the importance of WNT5A signaling in the acquisition of resistance to targeted therapies and demonstrate that RNF43 can block this process. Moreover, we show that RNF43 could be a negative regulator of melanoma phenotype plasticity by targeting MITF^-low/WNT5A^-high melanoma cells (Hoek et al., 2006; Kim et al., 2017; Sensi et al., 2011; Tirosh et al., 2016).

RNF43 as onco-suppressor in vivo: impact on tumors and resistance to vemurafenib

Next, we aimed to confirm our results also in the in vivo model. We decided to use cell lines with the same origin but varying in the RNF43 expression. For this purpose, A375 RNF43 TetON cells and A375 control lines (Figure 4—figure supplement 3B and C) were tested by qPCR. A375 control line had the lowest expression (referred to as ‘RNF43 low’) of RNF43, and doxycycline-untreated A375 RNF43 TetON showed intermediate expression due to TetON leakage (‘RNF43 mid’) that could be further enhanced by the doxycycline treatment (‘RNF43 high’) (Figure 7A). Next, we prepared and tested a novel vemurafenib formulation in 25% Kolliphor-water solution ensuring the solubility of this BRAFi and which could be administrated by oral gavage in the previously published dose 25 mg/kg (Figure 7—figure supplement 1A; Wang et al., 2019). Cells were injected intradermally into the NOD-Rag1^null IL2rg^null mice. Animals were divided into treatment cohorts once tumors became prominent in size (Figure 7B, Figure 7—figure supplement 1B). Cohort I (RNF43 low) had significantly shorter survival and formed macroscopic tumors earlier than RNF43 mid and high cohorts (Figure 7C, Figure 7—figure supplement 1C). Moreover, tumors originating from cells with the lowest RNF43 expression were significantly bigger than RNF43 mid ones (Figure 7D). To validate RNF43 expression induction and impact on its targets, protein lysates from tumors were analyzed by western blotting (Figure 7—figure supplement 1D). We found out that the level of RNF43-specific signal in cohorts I–III was significantly and inversely correlated with ROR1 and VANGL2 protein levels and similar trend was also noticed for DVL2 phosphorylation (Figure 7E, Figure 7—figure supplement 1D and E), recapitulating the described above in vitro effect also in the in vivo experiment. Importantly, we observed the RNF43 leakage (HA signal in Figure 7—figure supplement 1D), meaning that TetON system did not efficiently suppress the ectopic expression of RNF43 even in the absence of doxycycline, which also provides explanation for different RNF43 levels in studied here cell lines (Figure 7A). In conclusion, the dose-dependent effect mediated by RNF43 showed the features of the onco-suppression in melanoma in vivo model.

Further, we also investigated the synergistic effect of RNF43 and vemurafenib in vivo. RNF43 mid and RNF43 high cohorts received vemurafenib, and this treatment significantly prolonged the survival (Figure 7F), validating our formulation and dosing efficacy. Importantly, doxycycline supplementation (RNF43 high) in addition to vemurafenib resulted in the more efficient suppression of tumor growth, suggesting the delayed the BRAFi resistance acquisition (Figure 7G). Nevertheless, there was no difference at the experimental end point (Figure 7F and H), possibly due to the loss of transgene expression observed in cells during this long-term experiment (HA signal in Figure 7—figure supplement 1D). Moreover, the protein level of WNT5A was significantly increased in tumors treated with vemurafenib, which could help to bypass the RNF43-related effect as well (Figure 7—figure supplement 1E′′′). Our main findings are summarized as a graphical abstract (Figure 8).

1. Cell lines and treatments

T-REx-293 (R71007, Thermo Fisher Scientific), GFP labeled human melanoma A375 wild-type (A375) and its metastatic derivates A375 IV (Kucerova et al., 2014), A2058 (ECACC 91100402), and MelJuso (Štětková et al., 2020) (kindly gifted by Stjepan Uldrijan) cell lines were propagated in the Dulbecco’s modified Eagle’s medium (DMEM, 41966-029, Gibco, Life Technologies) supplemented with the 10% fetal bovine serum (FBS, 10270-106, Gibco, Life Technologies), 2 mM L-glutamine (25030024, Life Technologies), 1% penicillin-streptomycin (XC-A4122/100, Biosera) under 5% (vol/vol) CO₂ controlled atmosphere at 37°C. Routine checks for mycoplasma contamination were performed. For inhibition of endogenous Wnt ligands, cells were treated with the 0.5 μM Porcupine inhibitors C-59 (ab142216, Abcam) or 1 μM LGK-974 (1241454, PeproTech). The time points and doses have been chosen based on the purpose of the experiment. Changes in the phosphorylation of the WNT5A-downstream proteins in the noncanonical Wnt pathway transduction have been analyzed after 3 hr with the recombinant human WNT5A (645-WN, R&D Systems) in doses 40–100 ng/ml as given in the figure legends. Longer stimulation (overnight and longer) has been used in the functional experiments. For canonical Wnt signaling activation, the recombinant human WNT3A (5036-WN, R&D Systems) was used overnight in 40 ng/ml, 60 ng/ml, or 80 ng/ml concentrations. Co-treatment with the recombinant human R-Sponidin-1 (120-38, PeproTech) in 50 ng/ml dose was applied where indicated. Dansylcadaverine (D4008, Sigma-Aldrich) 50 µM treatment along with 3 hr tetracycline was applied to block clathrin-dependent endocytosis pathway (Blitzer and Nusse, 2006).

For preparation of stable cell lines, antibiotic selection after plasmid DNA transfection was performed using 5 μg/ml blasticidin S (3513-03-9, Santa Cruz Biotechnology) or 200 μg/ml of hygromycin B (31282-04-9, Santa Cruz Biotechnology) for T-REx-293 cells and accordingly 400 μg/ml and 5 μg/ml in case of A375 melanoma cell line. As a result, tetracycline-inducible T-REx-293 RNF43 and RNF43 Mut1 TetON, T-REx-293 RNF43/ZNRF3 dKO RNF43 TetON, A375+ RNF43, and A375 IV + RNF43 were obtained. A2058 RNF43 TetON, MelJuso RNF43 TetON A375 RNF43 TetON, and A375 TetON ctrl (not expressing exogenous RNF43) cells were obtained by lentiviral transduction of doxycycline-inducible pCW57-RNF43 (blast) plasmid encoding HA- and FLAG- tagged RNF43, using published protocol (Barta et al., 2016), followed by 5 μg/ml blasticidin S selection and limiting dilution.

T-REx-293 DVL1/2/3 tKO cells were described previously (Paclíková et al., 2017). For transgene expression induction (TetON), T-REx-293 cells were treated with 1 μg/ml of tetracycline (60-54-8, Santa Cruz Biotechnology) for the indicated time (3 hr to overnight) and melanoma cells were induced overnight by 1 μg/ml doxycycline (HY-N0565B, MedChem Express). Lysosomal degradation pathway was blocked by the 10 μM chloroquine (C662, Sigma) treatment, whereas 10 μM MG-132 (C2211, Sigma) was used for the proteasome inhibition. Generation of the melanoma cells resistant to vemurafenib (HY-12057, MedChem Express) and cobimetinib (HY-13064, MedChem Express) was performed according to the published protocols (Anastas et al., 2014). Resistant cells were cultured in the presence of 2 μM (A375) and 5 μM (A2058) of vemurafenib and 0.5 μM cobimetinib or their combination (A2058). For transient treatments (24 hr) of melanoma cell lines, 0.5 μM vemurafenib has been used (A375) or 2.5 μM of vemurafenib and 0.5 μM of cobimetinib (A2058).

2. Plasmids/cloning

Backbone of the plasmid pcDNA4-TO-RNF43-2xHA-2xFLAG (kindly gifted by Bon-Kyoung Koo together with pcDNA4-TO-RNF43Mut1-2xHA-2xFLAG; Koo et al., 2012) was used for further cloning. Briefly, for generation of the BioID-inducible pcDNA4-TO-RNF43-BirA*-HA plasmid, cDNA encoding RNF43 without stop codon was amplified by the PCR and cloned into the pcDNA3.1 MCS-BirA(R118G)-HA (Addgene plasmid #36047; RRID:Addgene_36047) using HpaI (ER1031, Thermo Fisher Scientific) and EcoRI (ER0271, Thermo Fisher Scientific) restriction enzymes to fuse it in frame with the BirA*-HA sequence. Then, RNF43-BirA*-HA cDNA was amplified and cloned by the In-Fusion cloning method (639690, Takara Bio) into linearized by HindIII (ER0501, Thermo Fisher Scientific) and XbaI (ER0681, Thermo Fisher Scientific) pcDNA4-TO plasmid. To eliminate BirA* enzyme-mediated potential false-positive results, pcDNA3-RNF43-HA was prepared by subcloning RNF43 PCR product containing HA encoding sequence in the reverse primer to the pcDNA3 backbone (Invitrogen). pCW57-RNF43 (blast) plasmid was obtained by RNF43-HA-FLAG cDNA cloning into the EcoRI site of the pCW57-MCS1-P2A-MCS2 (Blast) backbone (Addgene #80921; RRID:Addgene_80921) by the In-Fusion cloning method. All obtained plasmids were verified by the Sanger sequencing method.

Other plasmids used were described previously and included myc-Vangl1, GFP-Vangl2, GFP-Vangl2ΔN, GFP-Vangl2ΔC, GFP-Vangl2ΔNΔC (Belotti et al., 2012), pLAMP1-mCherry (Addgene #45147), pEGFP-C1-Rab5a (Chen et al., 2009), GFP-rab11 WT (Addgene #12674), His-ubiquitin (Tauriello et al., 2010), pcDNA3-Flag-mDvl1 (Tauriello et al., 2010), pCMV5-3xFlag Dvl2 (Addgene #24802), pCDNA3.1-Flag-hDvl3 (Angers et al., 2006), pcDNA3.1-hROR1-V5-His (gifted by Kateřina Tmějová), pcDNA3-Ror2-Flag and pcDNA3-Ror2-dCRD-FLAG (Sammar et al., 2004), pRRL2_ROR1ΔCYTO and pRRL2_ROR1ΔTail (Gentile et al., 2011), hCas9 (Addgene #41815), gRNA_GFP-T1 (Addgene #41819), and PiggyBack-Hygro and Transposase coding plasmids (gifted by Bon-Kyoung Koo). Sequences of primers used for cloning are presented in Table 1.

3. CRISPR/Cas9

For targeting RNF43 and ZNRF3 in the T-Rex-293, gRNAs TGAGTTCCATCGTAACTGTGTGG (PAM) and AGACCCGCTCAAGAGGCCGGTGG were cloned into gRNA_GFP-T1 backbone and transfected together with PiggyBack-Hygro and Transposase coding plasmids using polyethylenimine (PEI) in a way described below. For ROR1 and WNT5A knockout cell lines generation, gRNA CCATCTATGGCTCTCGGCTGCGG (ROR1) and AGTATCAATTCCGACATCGAAGG (WNT5A) were used. Transfected cells were hygromycine B selected and seeded as single cells. Genomic DNA isolation was performed using DirectPCR Lysis Reagent (Cell) (Viagen Biotech), Proteinase K (EO0491, Thermo Fisher Scientific), and DreamTaq DNA Polymerase (EP0701, Thermo Fisher Scientific) according to the manufacturer’s instructions. PCR products were analyzed by restriction digestion using Taal (ER1361, Thermo Fisher Scientific) in case of RNF43, HpaII (ER0511, Thermo Fisher Scientific) – ZNRF3, TaqI (ER0671, Thermo Fisher Scientific) – WNT5A and TseI (R0591S, New England BioLabs) – ROR1 for detection of Cas9-mediated disruptions in the recognition sites.

For targeting RNF43/ZNRF3 in the A375 and in the A375 IV melanoma lines, gRNAs AGTTACGATGGAACTCATGG (RNF43) and CTCCAGACAGATGGCACAGTCGG (ZNRF3) were accordingly cloned by described protocol into the pU6-(BbsI)CBh-Cas9-T2A-mCherry (Addgene #64324) and pSpCas9(BB)-2A-GFP (PX458) (Addgene #48138) backbones, transfected and sorted as single, GFP, and mCherry double-positive cells. These were then analyzed by restriction enzymes Hin1II (ER1831, Thermo Fisher Scientific) and Taal as described above. Finally, PCR products were sequenced using the Illumina platform and compared with the reference sequence (Malcikova et al., 2015). Sequencing results are presented in Supplementary file 1.

4. RNF43 BioID analysis

Data are available via ProteomeXchange (Deutsch et al., 2020) with identifier PXD020478 in the PRIDE database (Perez-Riverol et al., 2019). The analysis of the mass spectrometric RAW data files was carried out using the MaxQuant software (version 1.6.2.10) using default settings unless otherwise noted. MS/MS ion searches were done against modified cRAP database (based on http://www.thegpm.org/crap) containing protein contaminants like keratin, trypsin, etc., and UniProtKB protein database for Homo sapiens (ftp://ftp.uniprot.org/pub/databases/uniprot/current_release/knowledgebase/reference_proteomes/Eukaryota/UP000005640_9606.fasta.gz; downloaded 19.8.2018, version 2018/08, number of protein sequences 21,053). Oxidation of methionine and proline, deamidation (N, Q) and acetylation (protein N-terminus) as optional modification, carbamidomethylation (C) as fixed modification, and trypsin/P enzyme with two allowed miss cleavages was set. Peptides and proteins with FDR threshold <0.01 and proteins having at least one unique or razor peptide were considered only. Match between runs was set among all analyzed samples. Protein abundance was assessed using protein intensities calculated by MaxQuant. Protein intensities reported in proteinGroups.txt file (output of MaxQuant) were further processed using the software container environment (https://github.com/OmicsWorkflows, Kristina, 2021), version 3.7.2 a. Processing workflow is available upon request. Briefly, it covered (1) removal of decoy hits and contaminant protein groups, (2) protein group intensities log2 transformation, (3) LoessF normalization, (4) imputation by the global minimum, and (5) differential expression using LIMMA statistical test. Prior to volcano plot plotting, suspected BirA* binders were filtered out (proteins identified by at least two peptides in both technical replicates of particular BirA* sample, and present in more than three samples). Volcano plot was created in R using ggplot2 and ggrepel R packages by R version 3.6.1. Proteins with an adjusted p-value < 0.05 and log fold change >1 were further subjected to gene ontology tools, considering only the first ID of majority protein IDs: g:Profiler online tool (https://biit.cs.ut.ee/gprofiler/gost, version e98_eg45_p14_ce5b097; Raudvere et al., 2019) was used and selected GO terms were highlighted. RNF43 interactors from BioID assay are listed in Figure 1—source data 1, and the results obtained by g:Profiler are presented in Figure 1—source data 2.

5. Transfection

T-REx-293 cells were transected using 1 μg/ml, pH 7.4 PEI, and plasmid DNA in a 4:1 ratio (Paclíková et al., 2017). Plasmid DNA were in amount of 3 µg for 6 cm culture dish (ubiquitination assay) and 6 µg for 10 cm dish (co-immunoprecipitation or stable cell lines preparation). Approximately 1 × 10⁶ of A375 and A375 IV cells were electroporated with 6 μg of plasmid DNA utilizing Neon Transfection System (Thermo Fisher Scientific) 1200 V, 40 ms, 1 pulse. Culture media were changed 6 hr post-transfection.

6. His-ubiquitin pulldown assay

Cells were transfected with the plasmid encoding polyhistidine-tagged ubiquitin, RNF43-HA, or enzymatically inactive RNF43, protein of interest, and cultured overnight. Next, cells were treated with 0.2 µM epoxomicin (E3652, Sigma) for 4 hr and lysed in the buffer containing 6 M guanidine hydrochloride (G3272, Sigma), 0.1 M Na_xH_xPO₄ pH 8.0, and 10 mM imidazole (I5513, Sigma), sonicated, and boiled. Insoluble fraction was removed by the centrifugation (16,000 g, room temperature [RT], 10 min). For the pull down of tagged proteins, 10 µl of equilibrated in lysis buffer His Mag Sepharose beads Ni (GE28-9799-17, GE Healthcare) was added to each sample and kept on a roller overnight. Then, the beads were washed three times in the buffer containing 8 M urea (U5378, Sigma), 0.1 M Na_xH_xPO₄ pH 6.3, 0.01 M Tris, and 15 mM imidazole, resuspended in 100 μl of western blot sample buffer, boiled for 5 min, and loaded onto SDS-PAGE gel. Approximately 10% of cellular lysate was used as a transfection control after ethanol precipitation and resuspension in the western blot sample buffer.

7. Western blotting and antibodies

Western blot analysis was performed as described before using samples with the same protein amount, measured by the DC Protein Assay (5000111, Bio-Rad), or lysed directly in the sample buffer (2% SDS, 10% glycerol, 5% β-mercaptoethanol, 0.002% bromophenol blue, and 0.06 M Tris HCl, pH 6.8) and protease inhibitor cocktail (11836145001, Roche) after PBS wash (Mentink et al., 2018). Protein extraction from mouse tissues was done by homogenizing in the 1% SDS, 100 mM NaCl, 100 mM Tris, pH 7.4 buffer, sonication, clarification by centrifugation (16,000 g, 4°C, 15 min), and protein concentration measurement. Next, volumes of samples containing the same protein amounts were mixed with western blot sampling buffer and loaded onto SDS-PAGE gels. Briefly, after electrophoretic separation, proteins were transferred onto Immobilon-P PVDF Membrane (IPVH00010, Millipore) and detected using primary and corresponding HRP-conjugated secondary antibodies on Fusion SL imaging system (Vibler) using Immobilon Western Chemiluminescent HRP Substrate (Merck, WBKLS0500). Molecular size of the bands is marked in each panel (kDa). A list of used antibodies is presented in Appendix 1—key resources table. Shifts of ROR1, ROR2, DVL2, DVL3, VANGL1, and VANGL2 are marked by arrowheads. Empty arrowhead marks phosphorylation-dependent shift. Densitometric analysis of western blot signals was performed using ImageJ software. Activation level of DVL2 and DVL3 (Bryja et al., 2007a) is presented as the ratio of intensities of upper band; representing active – phosphorylated protein fraction and lower band (black arrowhead; unphosphorylated). Total DVL2 and DVL3 levels were quantified as the sum of two bands intensities.

8. Immunofluorescence and confocal microscopy

Cells growing on the glass were fixed in 4% paraformaldehyde (PFA) in PBS. Fixed cells were permeabilized by 0.1% Triton X-100 in PBS and blocked in 1% solution of bovine serum albumin (BSA) in PBS. Then, samples were incubated overnight at 4°C with primary antibodies diluted in 1% BSA in PBS and washed. Corresponding Alexa Fluor secondary antibodies (Invitrogen) were incubated with samples for 1 hr at RT, along with 1 µg/ml Hoechst 33342 (H1399, Thermo Fisher Scientific) for nuclei staining. After PBS washes, the samples were mounted in the DAKO mounting medium (S3023, DAKO). Images were taken on the confocal laser scanning microscopy platform Leica TCS SP8 (Leica). For co-localization analysis, histograms for each channel were prepared in LAS X Life Science (Leica) software and plotted in GraphPad Prism 8. Co-localization is marked by arrowheads.

9. Immunoprecipitation

T-REx-293 cells were transfected with the proper plasmid DNA and cultured for 24 hr. Then, cells were washed two times with PBS and lysed for 15 min in the buffer containing 50 mM Tris pH7.6, 200 mM NaCl, 1 mM EDTA, 0.5% NP40, fresh 0.1 mM DTT (E3876, Sigma) and protease inhibitor cocktail (04693159001, Roche). Insoluble fraction was removed by centrifugation (16,000 g, RT, 15 min), 10% of total cell lysate was kept as western blot control. Lysates were incubated with 1 μg of antibody for 16 hr at 4°C on the head-over-tail rotator. Next, 20 μl of protein G-Sepharose beads (17-0618-05; GE Healthcare) equilibrated in complete lysis buffer were added to each sample and incubated for 4 hr at 4°C, following six washes using lysis buffer and resuspension in 100 μl of western blot sample buffer. Immunoprecipitation experiments were analyzed by the western blot.

10. Flow cytometric determination of ROR1 surface expression

Determination of the ROR1 surface expression of T-REx-293 and its derivates was done using anti-ROR1-APC (#130-119-860, Miltenyi Biotec) and Accuri C6 (BD Biosciences) (RNF43/ZNRF3 dKO cells) or using BD FACSVerse Flow Cytometer (BD Biosciences) (TetON cells). Cells were harvested in 0.5 mM EDTA/PBS, washed in PBS, and incubated in 2% FBS in PBS with anti-ROR1-APC antibody (1:25, #130-119-860, Miltenyi Biotec) on ice for 30 min. The cells were washed and resuspended in PBS, and incubated with propidium iodide (10 ng/ml, #81845, Sigma-Aldrich) for 5 min to exclude dead cells from analysis. For the detection of ROR1 surface expression in HA-positive cells, ROR1-APC-stained cells were washed in PBS, fixed in 4% PFA at RT for 15 min, permeabilized in 0.02% Triton X-100 at RT for 15 min, and incubated with anti-HA antibody (1:1000, #9110, Abcam) in staining buffer at RT for 30 min. After two washes, cells were incubated with secondary antibody ALEXA Fluor 488 Donkey anti-Rabbit (#A21206, Invitrogen) at RT for 20 min, washed, and measured using FACS Verse (BD Biosciences). Data were analyzed using NovoExpress Software (ACEA Biosciences).

11. Quantitative polymerase chain reaction (qPCR)

Messenger RNA was isolated using the RNeasy Mini Kit (74106; Qiagen) according to the manufacturer’s instructions. 1 μg of mRNA was transcribed to cDNA by the RevertAid Reverse Transcriptase (EP0442, Thermo Fisher Scientific) and analyzed by use of LightCycler 480 SYBR Green I Master (04887352001, Roche) and LightCycler LC480 (Roche). Results are presented as 2^−ΔΔCT and compared by unpaired Student’s t-test. Mean expression of B2M and GAPDH or HSPCB and RPS13 (A375 VR cells) was used as reference. Primers are listed in Appendix 1—key resources table.

12. Databases

RNF43, VANGL1, and DVL3 gene expression in different melanoma stages was analyzed through Oncomine (RRID:SCR_007834; Rhodes et al., 2004) database in the different datasets (Talantov et al., 2005, Xu et al., 2008, Haqq et al., 2005). OncoLnc (Anaya, 2016) database was employed to elucidate whether the expression of the RNF43, ZNRF3, VANGL1, and DVL3 gene expression has a significant impact on the melanoma patients’ overall survival. RNF43 BioID data are available via ProteomeXchange (RRID:SCR_004055; Deutsch et al., 2020) in the PRIDE database (PXD020478) (RRID:SCR_003411; Perez-Riverol et al., 2019).

13. Wound healing assay, Matrigel invasion assay, fluorescent gelatin degradation assay, invadopodia formation assay, and collagen I hydrogel 3D invasion assay

For the determination of cellular motility and invasive properties in vitro wound healing (O’Connell et al., 2008), Matrigel invasion towards 20% FBS as chemoattractant followed by crystal violet staining of invaded cells, fluorescent gelatin degradation in the presence of 5% FBS after overnight starvation, and invadopodia formation assays were prepared according to the established protocols (Makowiecka et al., 2016). The wound gap was photographed using the Olympus ix51 inverted fluorescence microscope after 48 hr from scratch. Percentage of the cell-free surface was measured by ImageJ software. For the fluorescent gelatin degradation assay purpose, 80 ng/ml of rhWNT5A was used during 16 hr of cells’ incubation on the coverslips coated with gelatin-Oregon Green conjugate (G13186, Thermo Fisher Scientific). Alexa Fluor 594 phalloidin (A12381, Thermo Fisher Scientific) and TO-PRO-3 Iodide (642/661) were employed for the cells’ visualization on confocal microscopy platform Leica TCS SP8. For the invadopodia formation assay, an immunofluorescence imaging protocol employing phalloidin and anti-cortactin antibody was performed. Invadopodia – as structures double positive for F-actin and cortactin staining – was quantified for tested cell lines and conditions and presented as the number of invadopodia per one cell. Two independent repetitions were performed.

Collagen I hydrogel 3D invasion assay is a modification of the inverted vertical invasion assay (McArdle et al., 2016). Cells were plated on the µ-Slide 8 Well glass bottom coverslips (80827, Ibidi). At 80% confluence, the full medium was replaced with one containing 0.5% FBS for proliferation suppression. Doxycycline for RNF43 induction was applied at this step when needed. Next day, a solution of rat tail collagen type I in final concentration 1.5 mg/ml prepared accordingly to the manufacturer’s protocol was overlaid over the cells and left for polymerization for 30 min at 37°C, 5% CO₂. Then medium with final FBS concentration 10% and 100 ng/ml CXCL12 (350-NS, R&D Systems) or 100 ng/ml CCL21 (366-6C, R&D Systems) was added to the wells. After 24 hr, cells were PFA fixed, permeabilized, and stained with Hoechst 33342 (H1399, Thermo Fisher Scientific) and Alexa Fluor 594 phalloidin. Corresponding photos at the coverslip level and at 50 μm (A375 and A2058) or 70 μm (A375 IV) were taken using a confocal laser scanning microscopy platform Leica TCS SP8 (Leica). Invasion index was calculated as the ratio of invaded cells at a specified height to the number of noninvasive ones.

14. Colony formation assay

To assess the ability of colony formation in the presence of 0.3 μM vemurafenib, 300 of the melanoma cells were plated onto 24-well plate and were subsequently cultured for 7 days. After that time, the medium was removed and colonies were washed in PBS, fixed in the ice-cold methanol for 30 min, and stained with 0.5% crystal violet in 25% methanol. After washing and drying, bound crystal violate was eluted with 10% acetic acid and absorbance at 590 nm was measured on Tecan Sunrise plate reader. Results were normalized to the nontreated A375 wild-type results.

15. Animal studies

Animal experiments were approved by the Academy of Sciences of the Czech Republic (AVCR 85/2018), supervised by the local ethical committee, and performed by certified individuals (Karel Souček, Markéta Pícková, Ráchel Víchová).

A375 ctrl and A375 RNF43 TetON cells were implanted as a suspension of 50,000 cells in 50 μl saline intradermally into 9-week-old males of NOD-Rag1^null IL2rg^null strain, obtained from Jackson Laboratory. Animals were checked daily, and weight and tumors sizes were measured weekly along perpendicular axes using an external caliper. Tumor volumes were calculated using the equation volume = ½ (length × width²). Upon tumor establishment, animals were divided into five cohorts based on administered cells and treatment: (I) A375 ctrl + vemurafenib (N = 5); (II) A375 RNF43 TetON Dox - (N = 3); (III) A375 RNF43 TetON Dox + (N = 4); (IV) A375 RNF43 TetON VEMURAFENIB Dox - (N = 5); and (V) A375 RNF43 TetON VEMURAFENIB Dox + (N = 4). Doxycycline supplemented for RNF43 expression induction was administrated in drinking water, 0.2 mg/ml in 0.1% weight:volume sucrose. Control cohort obtained drinking water with 0.1% sucrose. Vemurafenib was supplemented daily by oral gavage in concentration of 25 mg/kg/day as freshly prepared formulation in 25% Kolliphor ELP (61791-12-6, Sigma-Aldrich) with 2.5% DMSO. Animals not treated with vemurafenib received the same formulation without inhibitor. Mice were sacrificed when tumors reached approximately 1000 mm³. Tumor samples were analyzed by western blot, and the times to reach the experimental end point was compared.

16. Software and statistics

Statistical significance was confirmed by two-tailed paired or unpaired Student’s t-tests. Survival was analyzed by Mantel–Cox test. Correlation between RNF43(HA) protein level and its targets in the in vivo experiments was tested by one-tailed Spearman correlation test. Statistical significance levels were defined as *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. All statistical details including the number of biological or technical replicates can be found in each figure legend. Statistical analysis and data visualization were performed in GraphPad Prism 8.0 software. Graphs are presented with error bars as ± SD if not stated differently in the figure legends.