WNT5a Signaling through ROR2 Activates the Hippo Pathway to Suppress YAP1 Activity and Tumor Growth

Abstract

Noncanonical Wnt signaling by WNT5a has oncogenic and tumor suppressive activities, but downstream pathways mediating these specific effects remain to be fully established. In a subset of prostate cancer organoid culture and xenograft models, inhibition of Wnt synthesis stimulated growth, while WNT5a or a WNT5a mimetic peptide (Foxy5) markedly suppressed tumor growth. WNT5a caused a ROR2-dependent decrease in YAP1 activity that was associated with increased phosphorylation of MST1/2, LATS1, MOB1, and YAP1, indicating Hippo pathway activation. Deletion of MST1/2 abrogated the WNT5a response. WNT5a similarly activated Hippo in ROR2-expressing melanoma cells, while WNT5a in ROR2-negative cells suppressed Hippo. This suppression was associated with increased inhibitory phosphorylation of NF2/Merlin that was not observed in ROR2-expressing cells. WNT5a also increased mRNA encoding Hippo pathway components including MST1 and MST2 and was positively correlated with these components in prostate cancer clinical datasets. Conversely, ROR2 and WNT5a expression were stimulated by YAP1, and correlated with increased YAP1 activity in clinical datasets, revealing a WNT5a/ROR2 negative feedback loop to modulate YAP1 activity. Together these findings identify Hippo pathway activation as a mechanism that mediates the tumor suppressive effects of WNT5a and indicate that expression of ROR2 may be a predictive biomarker for responsiveness to WNT5a-mimetic drugs.

Introduction

Prostate cancer (PCa) is a leading cause of cancer-related death in men. Androgen deprivation therapy (ADT) remains the standard systemic treatment, but patients invariably progress to metastatic castration-resistant prostate cancer (mCRPC). Many of these tumors will respond to further treatment with androgen signaling inhibitors (ASIs), but resistance inevitably develops. Alterations in androgen receptor (AR) and interacting pathways contribute to persistent AR activity and tumor growth in many cases (1–3), while others become independent of AR, with a subset of these tumors having neuroendocrine features (neuroendocrine prostate cancer, NEPC) (4). Increased Wnt signaling is one of the mechanisms that has been implicated in the progression to mCRPC and ASI resistance (5–7).

The canonical Wnt/β-catenin pathway is initiated when a Wnt ligand binds to a Frizzled (FZD) family receptor and its coreceptor, LRP 5/6, which results in sequestration of the β-catenin destruction complex (8). The stabilized β-catenin protein then translocates to the nucleus and acts as a transcriptional coactivator for TCF transcription factors, and may also coactivate additional factors including AR (9). Notably, genomic alterations that activate this pathway are found in over 20% of mCRPC, with mutations in APC and in CTNNB1 (encoding β-catenin) being most common (10–14). Wnt/β-catenin signaling may also be more generally increased by expression of particular Wnts or decreased expression of multiple negative regulators of this pathway. However, while inhibitors of Wnt synthesis or agents acting at other nodes in this pathway can suppress growth in some PCa models (7,15–17), the precise roles of canonical Wnt/β-catenin and noncanonical Wnt signaling in PCa remain unclear.

While most Wnts can activate the canonical Wnt/β-catenin pathway, a subset including WNT2, 4, 5a, 5b, 6, 7b, and 11 can preferentially activate noncanonical Wnt pathways (18). The co-receptors for this noncanonical Wnt signaling include ROR1, ROR2, RYK, and PTK7, which may act as heterodimers with a subset of FZD receptors (19). The downstream signaling pathways are less well-defined and are context dependent, but include the planar cell polarity (PCP) and Wnt-Ca⁺⁺ pathways. The PCP pathway involves activation of Rac and Rho, with downstream regulation of cell adhesion and migration, while the Wnt-Ca⁺⁺ pathway involves PLC activation and subsequent Ca⁺⁺ release from the endoplasmic reticulum and activation of proteins including PKC and CaMKII. Notably, noncanonical Wnt signaling has also been reported to inhibit the tumor-suppressive Hippo pathway and induce YAP1/TAZ activity, with subsequent stimulation of the TEAD family transcription factors (20–22). Conversely, YAP1/TAZ have been reported to suppress canonical Wnt/β-catenin signaling by binding to components of the β-catenin destruction complex (23).

WNT5a is the predominant noncanonical Wnt whose expression is often dysregulated in cancer, and it has been implicated in both tumor-suppressive and tumor-promoting activities (24). Low WNT5a expression is predictive of unfavorable outcome in some tumors (breast, prostate, and colon cancer), and a WNT5a mimetic peptide drug (Foxy5) (25) is currently being examined in colon cancer (Foxy5 as Neo-Adjuvant Therapy in Colon Cancer, NeoFox, NCT03883802). Conversely, a tumor-promoting function has been observed in melanoma, where elevated WNT5a expression has been associated with a worse prognosis (26,27). In PCa, increased WNT5a expression has been associated with progression to mCRPC (28–30) and with NEPC (31), and WNT5a signaling has been found in a subset of circulating tumor cells (CTCs) with low AR activity in men with mCRPC (32). In contrast, increased WNT5a has been associated with better prognosis in localized PCa (33,34). Moreover, WNT5a produced in the bone microenvironment was found to suppress growth and induce dormancy in PC3 PCa cells (34,35), and treatment with the WNT5a-mimetic peptide Foxy5 suppressed bone metastasis of DU145 PCa cells (36).

While these studies indicate that noncanonical Wnt signaling can have growth suppressive as well as stimulatory effects, the downstream signals mediating these growth suppressive specific effects remain to be established. We show here that WNT5a mediates ROR2-dependent activation of the Hippo pathway to suppress YAP1/TAZ activity and tumor growth. Moreover, we find that YAP1 increases expression of ROR2, revealing a negative feedback mechanism that regulates YAP1 activity. Together these and additional data indicate that the oncogenic versus tumor suppressive effects of WNT5a are modulated by ROR2, and that ROR2 expression may be a biomarker of tumors that will respond to WNT5a and related agents.

Materials and Methods

Cell lines and reagents.

DU145, CWR22RV1, VCaP, LNCaP, C4–2 and PC3 cell lines were from ATCC, were maintained under conditions recommended by the provider, and aliquots were studied within 3 months of thawing. LAPC4 parental cell line was kindly provided by Dr. CL Sawyers (Memorial Sloan Kettering Cancer Center, New York, NY) (37). The subline used in this study was derived from a castration-resistant xenograft of the parent line and was maintained in RPMI-1640 medium with 10% FBS. STR testing showed identity between this subline and the parental LAPC4. A375P, RPMI7951 and UACC257 cells were kindly provided by Dr. Pere Puigserver (Dana-Farber Cancer Institute, Boston, MA). Cells were confirmed to be Mycoplasma negative by testing with a kit from R&D Systems. ETC-1922159 (ETC-159, which is currently being evaluated in a phase I clinical trial, NCT02521844) was provided by A*STAR, Singapore. LGK974 was purchased from ActiveBiochem (Bonn, Germany). Recombinant human/mouse Wnt5a protein (645-WN), human Wnt3a protein (5036-WN), and Foxy5 peptide (5461) were purchased from R&D Systems. XMU-MP-1 was from SelleckChem (S8334). Verteporfin was from MedChemExpress (Cat. No.: HY-B0146).

3D cultured organoids.

3D/organoid cultures were generated from cell lines and PDXs by culturing in standard DMEM medium with 10% FBS and Matrigel, without further additives. BIDPC4 is a PDX developed by our lab from a biopsy of a soft tissue metastasis and adapted to grow in intact male mice. After euthanizing, the xenografts were harvested and cut into 1~2 mm chunks. The tumor bits were digested with collagenase/dispase (Millipore Sigma, #10269638001 & #11097113001 respectively) at 37⁰C for 30 minutes. The digestion was then stopped by adding medium with 10% FBS and tumor cells collected by centrifugation. Resuspended cell pellets were then seeded with in DMEM with 10% FBS and containing 5% (v/v) Matrigel (CORNING, Cat. # 356230), and plated at 2x10⁵ cells per well in 12 well plates that were precoated with 100 μL Matrigel on the surface of each well. LuCaP70CR is a castration-resistant PDX developed LuCaP70. Generation of LuCaP70CR organoids from LuCaP70CR tumor is similar to BIDPC4 organoids generation. The DU145, LAPC4, CWR22RV1, A375P, UACC257 and RPMI7951 organoids were established from 2D cultured cells using the same conditions, except 0.5 – 1.0x10⁶ cells were seeded in each well of a 12 well plate. The size and morphology of the organoids were followed during a period of 7–14 days. Organoids photos were taken at each time point with a Nikon camera (Coolpix 4300) (40X). 5 pictures were taken randomly for each treatment group. The organoids sizes (diameter) were measured by Image J (https://imagej.nih.gov/ij/) and analyzed with GraphPad Prism 9 Software.

Establishment of CRISPR cell lines.

According to the Zhang Lab general CRISPR cloning protocol (https://www.addgene.org/crispr/zhang/), we established the CRISPR cell lines with ROR2 or WNT5A knockout. We designed sgRNA oligonucleotides for ROR2 and WNT5A. The annealed oligos were cloned into the pX458 vector (Addgene). Plasmids were transfected into DU145 cells using Lipofectamine 3000 (Fisher Scientific, # L3000015). Single cell selection was by flow cytometry sorting of GFP positive cells 72 hours post-transfection. After 4 weeks, each isolated clone was validated by immunoblot analysis for ROR2 or WNT5A knockout. We did not further subclone, so residual protein expression may reflect subclones and we refer to these in the text as cell lines rather than clones. Lentiviral mediated CRISPR knockout lines were generated from LAPC4 cells. Each PsPAX2 plasmid carrying the sgRNA targeting ROR2, Wnt5A, STK3/MST2 or STK4/MST1 was cotransfected with the pMD2.G vector (Addgene) into 293FT cells with Lipofectamine 3000. The virus supernatant was collected at 48 hours post-transfection and filtered through 0.45um PES filters. The cleared virus supernatant was then incubated with LAPC4 (with polybrene) and the cells were selected with 2 μg/ml puromycin for 72 hours. The CRISPR lines were then validated for their intended target gene knockout. For YAP1 and ROR2 overexpression, FLAG-YAP1 (#66853) and ROR2 (#HG16133-CF) plasmids were purchased from Addgene and SinoBiological, respectively.

RNA interference.

NF2/Merlin knockdown with siRNA in DU145, RPMI7951 and A375P cells were performed as described previously (15). Briefly, DU145, RPMI7951 and A375P cells were transfected with 25 nM ON-TARGETplus siRNAs (Dharmacon) with Lipofectamine RNAiMAX (Thermo Fisher Scientific, 13778–075). At 72 hours after transfection, cells were treated with or without WNT5a or WNT5a peptide mimetic, Foxy5 (R&D, # 5461).

RNA isolation and qRT-PCR.

Total RNA was isolated with the RNeasy Mini Plus Kit (Qiagen, # 74134). RNA purification was directly performed from 2D cultured cells. For RNA isolation in 3D cultured organoids, cell recovery solution (Corning, DLW354253) was first used for the removal of Matrigel (1 hour, on ice). qRT-PCR was performed using standard SYBR Green reagents from the StepOnePlus Real-Time PCR system (Thermo Fisher Scientific, # 11780200) or TaqMan One-Step RT-PCR reagents (Thermo Fisher Scientific, #4444434). Target mRNA expression was quantified using the ΔΔCt method and normalized to GAPDH expression.

Immunoblotting.

Immunoblotting was performed as described (38). Briefly, proteins were extracted from cultured cells or homogenized frozen tumor bits using a TissueRuptor (Qiagen, Cat No./ID: 9002755) on ice using RIPA buffer. Protein extraction from 3D/organoid cultures started with digestion of the Matrigel matrix using Cell Recovery Solution (DLW354253, Sigma) for 1 hour on ice. After digestion with Cell Recovery Solution, cells were centrifuged and RIPA buffer was added into the pellets for protein extraction. Antibodies against the following proteins (and clone or catalogue number) were from Cell Signaling: ROR1 (D6T8C), ROR2 (D3B6F), Non-phospho (Active) β-Catenin (Ser45) (D2U8Y), LEF1 (C12A5), Wnt5a/b (C27E8), Dvl2 (30D2), Axl (C89E7), Integrin β2 (D4N5Z), IGFBP3 (D1U9C), YAP/TAZ (D24E4), Phospho-YAP (Ser127) (D9W2I), MST1 (#3682), MST2 (#3952), MOB1 (E1N9D), Phospho-MOB1 (Thr35) (D2F10), LATS1 (C66B5), Phospho-LATS1 (Ser909) (#9157), SAV1 (D6M6X), Phospho-MST1 (Thr183)/MST2 (Thr180) (E7U1D), Merlin (D3S3W), Lamin B2 (D8P3U), Survivin (71G4B7), Phospho-Merlin (Ser518) (D5A4I), Tubulin (#2148). Additional antibodies were against Kif26B (Proteintech, 17422–1-AP), β-Actin (Santa Cruz), GAPDH (Abcam Ab-9485), and Vinculin (Sigma V9131). The primary antibodies were incubated overnight. Gels shown are representative of at least three independent experiments.

Xenografts and drug treatment.

All animal studies were conducted at the Beth Israel Deaconess Medical Center according to protocol approved by the institutional animal care and use committee (IACUC). BIDPC4 PDX were subcutaneously injected into 5–6 weeks old intact ICR SCID male mice. Castrated 5–6 weeks old ICR SCID mice were used for BIDPC5 PDX. LAPC4 and DU145 subcutaneous xenograft tumor models were established in nude mice. When a tumor reached 0.5 cm in largest dimension, the host was randomly assigned to the control, ETC-159 or Foxy5 treatment group. Briefly, the ETC-159 treatment group received 30 mg/kg of ETC-159 (in 0.5% carboxyl methyl-cellulose and by oral gavage) every other day. Foxy5 (# 5461) was intraperitoneal (ip) injected every other day at 2 mg/kg. The control group similarly received the carrier solution only. The tumor volumes were followed by directly palpation and caliper measurement twice a week. Tumor size was calculated by the formula: Volume= Length *Width² *0.5. The fold change of tumor volume was determined by normalizing each tumor measurement to its volume on day 0 at the beginning of drug administration.

Bioinformatic analysis.

The mRNA expressions of genes in PCa patients were obtained from the database of TCGA-Prostate Cancer (PRAD). Top 20% patients with highest (n=50) or lowest (n=50) Wnt5a mRNA expressions were grouped into Wnt5A high and low groups, respectively. YAP1 pathway genes (YAP1, CTGF, IGFBP3, etc.) and Hippo pathway genes (STK3, STK4, LATS1, LATS2, etc.) expressions were analyzed in Wnt5A high vs Wnt5A low groups, using 2-tailed Student’s t test. The mRNA co-expression analysis between Hippo/YAP pathway genes and Wnt5a/ROR2 in PCa patients and melanoma patients were performed through cBioPortal (http://www.cbioportal.org/). Gene Set Enrichment Analysis (GSEA) was performed with whole genes mRNA expressions in PCa patients (TCGA-PRAD dataset), based on ROR2 expression level. Disease free survival and overall survival analysis were done with TCGA-PRAD data.

Statistics.

GraphPad Prism 9 Software (GraphPad Software Inc.) was used for all statistical analysis unless otherwise specified. Significance of difference between 2 groups was determined by 2-tailed Student’s t test. Pearson Correlation was used for all correlation analysis. Log rank test was used for survival analysis. Statistical significance was accepted at P < 0.05.

Study Approval.

Animal studies were approved by the BIDMC IACUC.

Data Availability Statement.

The data generated in this study are available within the article and its supplementary data files. The data analyzed in this study were obtained from the National Cancer Institute Genomic Data Commons at https://gdc.cancer.gov/about-data/publications/prad_2015. Expression vectors generated in this study have been deposited in Addgene.

Results

Subset of PCa grown under 3D/organoid conditions are suppressed by noncanonical Wnt signaling

We initially assessed effects of WNT5a on DU145 and PC3 cells under standard 2D culture conditions, but did not see clear stimulatory or repressive effects. We then cultured DU145 cells under organoid culture conditions in medium that did not contain added Wnts. We refer here to these as organoid cultures as they can support organoid generation, but these tumor cells generally grow as aggregates without clear development of glandular structure. We then treated the cells with the PORCN inhibitor ETC-159 to globally suppress Wnt production, and found that the organoid size was significantly increased (Fig. 1A). Addition of a canonical Wnt (WNT3a) did not have further effects, but addition of WNT5a decreased organoid size. The same results were obtained with an alternate PORCN inhibitor, LGK974 (Fig. S1A). We similarly examined another PCa cell line derived from LAPC4 cells. Organoid cultures from these cells were also stimulated by treatment with ETC-159 (Fig. 1B) or LGK974 (Fig. S1B), and this stimulation was abrogated by addition of WNT5a. Notably, WNT5a alone did not consistently suppress basal growth, which may reflect effects of endogenous WNT5a or other noncanonical Wnts (see below).

Figure 1. Noncanonical WNT5a decreases PCa organoid growth.

Representative images of DU145 (A) and LAPC4 (B) organoids after a 7-day (DU145) or 14-day (LAPC4) treatment with ETC-159 (1 μM) alone or plus WNT3a (0.3 μg/ml) or WNT5a (0.6 μg/ml). Mann-Whitney tests were performed to measure the differences in organoids size. **** = p<0.0001, *** = p<0.0001, ns = not significant. Immunoblotting of active β-catenin was performed in DU145 (C) and LAPC4 (D) organoids with above treatments. (E) Immunoblotting of KIF26b in DU145 and LAPC4 organoids after the 7- or 14-day treatments described above. (F) Immunoblotting of ROR1 and ROR2 in various PCa cell lines. Vinculin was internal control for protein loading.

The effects of ETC-159 and WNT5a did not appear to be related to canonical Wnt/β-catenin signaling as levels of active β-catenin were not altered or were only modestly suppressed by these agents, but as expected were increased by WNT3a (Fig. 1C, D). Together these results indicated that the growth stimulatory effect of ETC-159 may be due to decreased autocrine noncanonical Wnt signaling. In support of this mechanism, we found that expression of KIF26b, which has been shown previously to be degraded downstream of noncanonical WNT signaling (39), was increased by ETC-159 and decreased by WNT5a (Fig. 1E).

In contrast to these results, organoids from the CWR22Rv1 PCa cell line were suppressed by ETC-159, and were not further suppressed by WNT5a at concentrations that suppressed LAPC4 (Fig. S1C, D). Notably, amongst described receptors for noncanonical Wnt signaling (ROR1, ROR2, RYK, and PTK7), we found that mRNA for ROR2 was preferentially expressed by DU145 and LAPC4 cells (Fig. S1E). This was also reflected at the protein level, with only DU145 and LAPC4 cells having readily detectable levels of ROR2 protein (Fig. 1F). Taken together these results indicated that DU145 and LAPC4 organoids were being repressed by autocrine production of WNT5a, or possibly another noncanonical Wnt.

Noncanonical Wnt signaling suppresses PCa xenograft growth

We previously reported that PCa xenografts from the VCaP and CWR22Rv1 cells lines, as well as several PCa PDXs (LuCaP70CR, LuCaP77CR, and BIDPC5) were suppressed by treatment with PORCN inhibitors (7,15). Therefore, we next generated xenografts from DU145 and LAPC4 cells to determine whether their growth in vivo was suppressed by PORCN inhibition or noncanonical Wnt signaling. For these in vivo studies we used a WNT5a-mimicking hexapeptide drug, Foxy5 (25). Consistent with the in vitro results, xenografts generated from DU145 (Fig. 2A) and LAPC4 (Fig. 2B) were stimulated by ETC-159 and repressed by Foxy5. Notably, this latter Foxy5 result indicated that endogenous levels of Wnt5a (or other noncanonical Wnts) were not fully activating the growth suppressive mechanisms driven by Foxy5. We then examined a small panel of PCa patient derived xenograft (PDX) models for expression of ROR2 and WNT5a. ROR2 was expressed at varying levels by each, but BIDPC4 had particularly high levels of WNT5a, suggesting substantial autocrine noncanonical Wnt signaling (Fig. 2C). Therefore, we generated organoids from BIDPC4 and found that these were stimulated by ETC-159 and LGK974, and were suppressed by WNT5a (Fig. 2D).

Figure 2. WNT5a suppression of PCa growth in vivo is ROR2 dependent.

Growth curves for subcutaneous xenografts of DU145 (A) or LAPC4 (B) in three treatment groups, including a control (black, n=5), an ETC-159 (blue, n=5) and a Foxy5 arm treatment group (red, n=5). ETC-159 was given by oral gavage at 30 mg/kg/qod. Foxy5 was given by intraperitoneal (ip) injection at 2 mg/kg/qod. Mann-Whitney tests was applied to measure the variations in tumor sizes. ** = p<0.01, * = p<0.05. (C) Immunoblotting of WNT5a in the BID-PDX series, and in LuCaP35CR and 70CR PDXs. (D) Representative images of BIDPC4 organoids after a 7-day treatment with DMSO, LGK974 (0.6 μM), ETC-159 (1.0 μM) or WNT5a (0.6 μg/ml) alone or in combinations as indicated. Mann-Whitney tests of organoids sizes were performed. **** = p<0.0001, ns = not significant. (E) Growth curve of BIDPC4 PDX, with negative control (black, n=5), ETC-159 (blue, n=5) or Foxy5 (red, n=5) treatments (with the same doses and schedule as for DU145 and LAPC4 xenografts). Mann-Whitney tests was applied to measure the variations in tumor sizes. ** = p<0.01, * = p<0.05. (F) Representative images of DU145 organoids with control (sg-NTC) or targeted CRISPR knockout of Wnt5a (sg-Wnt5A) or ROR2 (sg-ROR2). These organoids were treated with DMSO, ETC-159 (1 μM) alone or plus WNT5a (0.6 μg/ml) for 7 days. Mann-Whitney tests was applied to measure the variations in organoids sizes. **** = p<0.0001, *** = p<0.001, or ns = not significant. (G) Representative images of LNCaP organoids with non-targeting control (NTC) or human ROR2 (H-ROR2) overexpression after a 7-day treatment with DMSO, WNT5a (0.6 μg/ml) or Foxy5 (2 μM). Mann-Whitney tests was applied to measure the variations in organoids sizes. **** = p<0.0001, ** = p<0.01.

In contrast, growth of organoids generated from LuCaP70CR, which expresses low levels of WNT5a (Fig. 2C) and ROR2 (Fig S2A) relative to BIDPC4, was suppressed by ETC-159 and not further suppressed by WNT5a (Fig. S2B). Notably, LuCaP70CR organoids also had high levels of KIF26b relative to BIDPC4, indicating low basal noncanonical Wnt signaling (Fig. S2A). Consistent with the in vitro results, BIDPC4 PDXs in vivo were stimulated by ETC-159 and repressed by Foxy5 (Fig. 2E). Notably DU145 cells and our LAPC4 cell line express low or undetectable levels of AR, suggesting the repressive effects of WNT5a may be related to loss of AR dependence (Fig. S2C). However, the BIDPC4 PDX expresses high levels of AR, indicating that the response to WNT5a is not dictated by AR status.

Growth suppressive effect of WNT5a is mediated by ROR2

The relatively high expression of ROR2 in DU145 and LAPC4 cells suggested that it may be mediating growth suppressive signaling from WNT5a. To further assess the role of ROR2 and autocrine WNT5a, we used CRISPR to generate DU145 cell lines with depleted ROR2 or WNT5a (Fig. S2D). Significantly, depletion of either ROR2 or WNT5a enhanced growth of DU145 organoids (Fig. 2F). Moreover, the WNT5a depleted organoids, but not the ROR2 depleted organoids, were suppressed by exogenous WNT5a. We also deleted ROR2 in LAPC4 cells (Fig. S2E). In contrast to DU145, this did not enhance organoid growth or prevent the stimulatory effects of ETC-159, possibly reflecting residual ROR2 or compensatory effects mediated through other Wnt receptors in these cells (Fig. S2F). Nonetheless, the ROR2 decrease in the LAPC4 cells prevented the growth-suppressive effects of exogenous WNT5a. Finally, we examined organoids generated from LNCaP PCa cells stably transfected with ROR2. WNT5a and Foxy5 did not suppress nontransfected control LNCaP cells, and instead had a modest stimulatory effect (Fig. 2G). In contrast, WNT5a and Foxy5 both markedly decreased the size of LNCaP organoids overexpressing ROR2. Notably, the size of the untreated ROR2 expressing organoids was also decreased, which may reflect increased basal signaling. Together these data indicate that the growth suppressive effects of WNT5a are mediated through ROR2.

WNT5a signaling through ROR2 suppresses YAP1 activity

We next sought to determine the mechanism by which WNT5a/ROR2 signaling mediates growth suppression. Immunoblotting for PARP1 cleavage and cleaved caspase 3 indicated that WNT5a was not increasing apoptosis (Fig. S3A). Another possible mechanism was increased TGFβ signaling (40). However, treatment with a TGFβ receptor inhibitor did not increase the growth of LAPC4 organoids, and treatment with TGFβ did not impair the growth promoting effects of ECT-159 (Fig. S3B). Notably, a recent study found that WNT5a could induce dormancy of PCa cells in bone, which was mechanistically linked to increased expression of SIAH2 and subsequent increased degradation of β-catenin (35). Moreover, noncanonical Wnt signaling has also been reported to suppress canonical Wnt signaling by further mechanisms in other contexts. However, we did not observe a consistent decrease in active β-catenin in response to WNT5a (see Fig. 1C).

Previous studies have linked noncanonical Wnt signaling to activation of YAP1 and TAZ (encoded by WWTR1) and suppression of the Hippo pathway (20–22). In contrast, in DU145 organoid cultures we found that ETC-159 increased expression of several previously identified YAP1/TAZ target genes including AXL, LMNB2 (encoding Lamin B2), and BIRC5 (encoding Survivin), and that these were suppressed by WNT5a (Fig. 3A). YAP1/TAZ activities are negatively regulated by phosphorylation, which decreases their nuclear expression and enhances their degradation (41,42). Consistent with the effects on YAP1/TAZ regulated genes, ETC-159 in DU145 organoids decreased levels of phosphorylated YAP1 (pYAP1) and increased total YAP1, with WNT5a having the opposite effects (Fig. 3B). The same result was seen in LAPC4 organoids (Fig. 3C). Moreover, in LAPC4 xenografts we also observed an ETC-159-mediated decrease in YAP1 phosphorylation, an increase in total YAP1, and an increase in YAP1 transcriptional targets (integrin β2, IGFBP3, and Survivin), which were all reversed by Foxy5 (Fig. 3D).

Figure 3. WNT5a suppresses YAP1 activity through ROR2.

(A) Immunoblotting of YAP1/TAZ regulated gene products in DU145 organoids treated with DMSO, ETC-159 (1 μM) alone or in combination with WNT5a (0.6 µg/ml) for 7 days. Immunoblotting of phosphorylated (Ser127) or total YAP1 after a 7-day treatment with ETC-159 alone (at the indicated doses, µM) or in combination with WNT5a (ng/ml) in DU145 (B) and LAPC4 (C) organoids. (D) Immunoblotting of phosphorylated (Ser127) and total YAP1, and several YAP1/TAZ regulated gene products in LAPC4 xenografts after a 2-week treatment with ETC-159 (oral gavage at 30 mg/kg/qod), Foxy5 (ip injection at 2 mg/kg/qod), or no treatment control (NC). Three independent tumors of each treatment group were blotted. (E) Immunoblotting of phospho- or total YAP1, active (non-phosphorylated) β-catenin, or LEF1 in LuCaP70CR organoids after 7-day treatments with DMSO, ETC-159 (1 μM) alone or in combination with WNT5a (0.6 μg/ml). Two biological replicates are shown for each treatment. (F) Immunoblotting of WNT5a, ROR2 or phosphorylated (Ser127) and total YAP1 in control (sgNTC) or ROR2 (sgROR2) CRISPR knockout DU145 organoids after a 7-day treatment with DMSO (-) or ETC-159 (1 µM), alone or in combination with WNT5a (0.6 µg/ml). The prominent WNT5a bands in the WNT5a treated cell lysates reflect the exogenously added WNT5a, which is decreased in the ROR2 depleted cells. (G) Immunoblotting of phosphorylated (Ser127) and total YAP1 in control (sgNTC) or ROR2 (sgROR2) knockout LAPC4 organoids after a 7-day treatment with ETC-159 (1 µM) alone or in combination with WNT5a (0.6 µg/ml). Two biological replicates are shown for each treatment.

In contrast, in LuCaP70CR organoids (whose growth is not suppressed by WNT5a, see Fig. S2B), treatment with ETC-159, or with ETC-159 combined with WNT5a, did not have effects on YAP1 (Fig. 3E). Finally, to determine whether these effects were mediated through ROR2, we examined cells with CRISPR depletion of ROR2. Significantly, WNT5a did not increase YAP1 phosphorylation after ROR2 depletion in DU145 cells (Fig. 3F) or in LAPC4 cells (Fig. 3G).

WNT5a signaling through ROR2 activates the Hippo pathway

YAP1 and TAZ are primarily regulated by the Hippo pathway, which is initiated by phosphorylation of MST1 and/or MST2 (encoded by STK4 and STK3, respectively), subsequent downstream phosphorylation of LATS1, LATS2, and MOB1, and then LATS-mediated phosphorylation of YAP1 and TAZ (41–43). Consistent with Hippo pathway activation, ETC-159 treatment of DU145 organoids decreased phosphorylation of LATS1/2 and of MOB1, and these were increased by WNT5a (Fig. 4A). Similar effects were observed in LAPC4 organoids (Fig. 4B), and in organoids derived from the BIDPC4 PDX (Fig. 4C).

Figure 4. WNT5a/ROR2 signaling activates the Hippo pathway.

Immunoblotting of phosphorylated LATS1 (pSer909), MOB1 (pThr35) or total LATS1 and MOB1 in DU145 organoids (A) and LAPC4 organoids (B) after a 7-day treatment with ETC-159 (0.5 – 1 µM) alone or in combination with WNT5a (0.6 – 1.2 mg/ml). (C) Immunoblotting of phosphorylated or total LATS1 or MOB1 in BIDPC4 organoids treated with of ETC-159 (1 µM) alone or in combination with WNT5a (0.6 µg/ml) for 7 days. Two biological replicates are shown for each treatment. (D) Immunoblotting of phosphorylated MST1 (pThr183)/MST2 (pThr180) or total MST1; and phosphorylated LATS1 (pSer909) protein in DU145 organoids after ETC-159 (1 µM), WNT5a (0.6 µg/ml), or combination treatments for 7 days. Two biological replicates are shown for each treatment. (E) Immunoblotting of MST1/2, phosphorylated and total LATS1 in control (sg-NTC) or ROR2 (sg-ROR2) CRISPR knockout LAPC4 organoids after a 7-day single or combination treatment with ETC-159 (1 µM) and WNT5a (0.6 µg/ml). Two biological replicates are shown for each treatment. (F) Immunoblotting of phosphorylated LATS1 (pSer909), phosphorylated (Ser127) YAP1, YAP1, phosphorylated MST1 (pThr183)/MST2 (pThr180) or total MST1, phosphorylated MOB1 (pThr35), or MOB1 in control (NTC) and ROR2 overexpressing LNCaP organoids with 7 days treatment of DMSO, WNT5a (0.6 μg/ml) or Foxy5 (2 μM). Two biological replicates are shown for each treatment.

We then examined MST1/2, and found that MST1/2 phosphorylation also was decreased by ETC-159 and increased by WNT5a in DU145 organoids (Fig. 4D). Notably, total MST1 was similarly altered, suggesting that MST1 synthesis and/or stability may be enhanced downstream of WNT5a (see below). The effects of ETC-159 and WNT5a on total MST1/2 protein levels were even more substantial in LAPC4 organoids, where MST1 and MST2 protein expression were markedly decreased by ETC-159 and restored by WNT5a (Fig. 4E). MST phosphorylation was not clearly increased in the ETC-159/WNT5a treated cells relative to the untreated cells in this experiment (not shown), but LATS1/2 phosphorylation was increased, and this was again ROR2 dependent. Notably, ETC-159 also markedly decreased MST1/2 expression in the ROR2 deficient LAPC4 organoids (Fig. 4E). This Wnt-dependent/ROR2-independent basal regulation of MST1/2 expression may be mediated through an alternative noncanonical Wnt receptor, or even possibly by canonical Wnt signaling. (Fig. 4E). Finally, we examined organoids generated from the LNCaP cells overexpressing ROR2, whose growth was suppressed by WNT5a (see Fig. 2G). WNT5a and Foxy5 treatment in these cells (versus control LNCaP cells) increased phosphorylation of MST1/2, LATS1, MOB1, and YAP1, further supporting Hippo activation (Fig. 4F).

To further establish the role of MST1/2 in the LAPC4 cells, we used CRISPR to deplete MST1 and MST2 (Fig. 5A). Consistent with Hippo pathway mediation of WNT5a/ROR2 signaling, organoids generated from MST1 or MST2 depleted LAPC4 cells had increased growth, and became unresponsive to ETC-159 or WNT5a (Fig. 5B). Moreover, the WNT5a-mediated increased phosphorylation of LATS1, MOB1, and YAP1 was abrogated in the MST1 and MST2 knockout cells (Fig. 5C). Notably, the MST1 knockout cells also had lower MST2, suggesting that MST2 may be the predominant mediator of Hippo signaling in these cells (Fig. 5C). Alternatively, there may be a requirement for MST1/2 heterodimers, or possibly a requirement for a threshold level of MST1 and/or MST2.

Figure 5. WNT5a/ROR2 mediated growth suppression is dependent on MST1/2 and Hippo activation.

(A) Immunoblotting of MST1 or MST2 in LAPC4 cells with or without CRISPR-mediated knockout of respective MST. Two independent lines for each CRISPR target were examined. (B) Representative images of control (sgNTC), MST1 (sgMST1) and MST2 (sgMST2) knockout LAPC4 organoids after 7-day treatment with ETC-159 (1 µM) alone or in combination with WNT5a (0.6 µg/ml). Mann-Whitney tests were used to measure the variations in organoids sizes (bottom panel). **** = p<0.0001, *** = p<0.001, ns = no significance. (C) Immunoblotting of Hippo pathway components in control and MST1 or MST2 knockout LAPC4 organoids after a 7-day treatment with ETC-159 (1 µM) alone or in combination with WNT5a (0.6 µg/ml). (D) Representative images of DU145 organoids treated with ETC-159 (1 μM), WNT5a (0.6 μg/ml), or verteporfin (5 μM), alone or in combination for 5 days. Mann-Whitney tests were used to measure the variations in organoids sizes (bottom panel). **** = p<0.0001, ** = p<0.01, ns = no significance. (E) Immunoblotting of AXL, YAP1, IGFBP3 and CTGF in DU145 organoids after 5-day treatment with ETC-159 (1 μM), WNT5a (0.6 μg/ml), or verteporfin (5 μM), alone or in combination.

We next used treatment with verteporfin, which blocks YAP1/TAZ coactivation of TEAD transcription factors, to address whether the suppressive effects of WNT5a were due to YAP1/TAZ inactivation. Verteporfin suppressed basal and ETC-159 stimulated growth of DU145 organoids (Fig. 5D). Moreover, WNT5a did not further suppress the vertoporfin treated cells, consistent with WNT5a suppressing through YAP1/TAZ. WNT5a and verteporfin also had comparable suppressive effects on the ETC-159 stimulated expression of the YAP1/TAZ regulated genes AXL and IGFBP3 (Fig. 5E). Notably, although the CCN2 gene (encoding CTGF protein) is regulated by YAP1/TAZ in many contexts, it appears not to be YAP1/TAZ regulated in these cells.

ROR2 modulates response to WNT5a in melanoma cells

In contrast to observations in PCa, a tumor-promoting function for WNT5a has been observed in melanoma (26,27), and has been linked to YAP1 stabilization (22). Therefore, we next examined a series of melanoma cell lines for expression of ROR2 and WNT5a, and found low or undetectable expression of ROR2 in most (Fig. 6A). As UACC-257 expressed substantial ROR2 and WNT5a, we assessed its response to WNT5a. Similarly to the PCa cells, UACC-257 organoids where stimulated by ETC-159, and this stimulation was abrogated by addition of WNT5a (Fig. 6B). Moreover, WNT5a increased LATS1 and MOB1 phosphorylation, consistent with Hippo activation (Fig. 6C).

Figure 6. WNT5a activates Hippo pathway in ROR2 expressing melanoma cells.

(A) Immunoblotting of ROR2 and WNT5A in melanoma cell lines. (B) Representative images of UACC257 organoids after 7-day treatment with ETC-159 (1 µM) and/or WNT5a (0.6 µg/ml) (image for Wnt5a alone is not shown). Mann-Whitney tests were used to measure the variations in organoids sizes (bottom panel). **** = p<0.0001. (C) Immunoblotting of KIF26B, pLATS1 (pSer909), pMOB1 (pThr35), or total LATS1 and MOB1 in UACC257 organoids treated with DMSO, ETC-159 (1 µM) alone or in combination with WNT5a (0.6 µg/ml). Two biological replicates are shown for each treatment. (D) Immunoblotting of pLATS1 (pSer909), active (nonphosphorylated) YAP1, phosphorylated (Ser127) YAP1, phosphorylated MOB1 (pThr35) and MOB1 in A375P organoids treated with DMSO, ETC-159 (1 µM) alone or in combination with WNT5a (0.2 – 0.4 µg/ml) or Foxy5 (2 – 4 µM) for 7 days. Two biological replicates are shown for each treatment. (E) Immunoblotting of Merlin, phospho-Merlin (S518) and active (nonphosphorylated) YAP1 in control siRNA (NTC) or Merlin siRNA treated (si-Merlin) 2D-cultured A375P cells in response to WNT5a (0.1 – 0.2 µg/ml) treatment for 4 hours. Two biological replicates are shown for each treatment. The lower pMerlin band that is depleted by siMerlin corresponds to the molecular weight of Merlin. (F) Immunoblotting of phosphorylated (Ser127) YAP1, total YAP1, phospho-Merlin (S518), total Merlin, and GAPDH (loading control) in LAPC4 organoids treated with DMSO, ETC-159 (1 µM) alone (5 days) or in combination with WNT5a (0.4 µg/ml). Two biological replicates are shown for each condition.

We then examined A375P cells, which expressed low or undetectable levels of ROR2 and WNT5a, and in which WNT5a has been reported to increase YAP1 (22). Consistent with this previous report, WNT5a, and to a greater extent Foxy5, decreased YAP1 phosphorylation in ETC-159 treated cells (Fig. 6D). Moreover, this was associated with a decrease in phosphorylation of LATS1 and MOB1, indicating downregulation of Hippo.

One major regulator of Hippo is NF2/Merlin, which serves as a scaffold at the plasma membrane to facilitate Hippo signaling (44–46). This membrane binding and subsequent Hippo activation can be disrupted by S518 phosphorylation of Merlin, which is mediated by PAK1 and displaces Merlin from the plasma membrane (47,48). Consistent with this mechanism for Hippo suppression in A375P cells, WNT5a caused an increase in Merlin S518 phosphorylation (Fig. 6E). Moreover, Merlin depletion by siRNA caused an increase in active (nonphosphorylated) YAP1, and abrogated the increase in active YAP1 in response to WNT5A. In contrast, WNT5a did not decrease Merlin S518 phosphorylation in LAPC4 organoids, indicating that ROR2 is not acting through Merlin to driving Hippo (Fig. 6F).

WNT5a regulates expression of Hippo pathway proteins

As noted above, in addition to stimulating increased phosphorylation of Hippo proteins, we observed that total MST1/2 expression was increased by WNT5a. To further assess whether WNT5a regulates basal expression of Hippo pathway proteins, we examined cells with CRISPR-mediated WNT5a deletion. The loss of WNT5a in DU145 cells caused a decrease in MST1 and MST2, and marked decrease in MOB1 (Fig. 7A). Downstream of MST1/2, there was also a decrease in LATS1 phosphorylation, an increase in YAP1, and an increase in YAP1/TAZ regulated genes (AXL and IGFBP3). Similar effects were observed after loss of WNT5a in LAPC4 cells (Fig. 7B).

Figure 7. WNT5a regulates expression of Hippo pathway proteins.

Immunoblotting of Hippo pathway components and target genes (AXL, IGFBP3) in sg-control (NTC) and WNT5a knockout (sg-Wnt5a) DU145 cells (A) and LAPC4 cells (B). In DU145 (C) and LAPC4 (D) organoids, MST2/STK3 and MST1/STK4 mRNA expression in sg-control cells (sgNTC) or WNT5a knockout cells (sgWnt5a) without or with WNT5a (0.6 µg/ml) (sgWnt5a+Wnt5a) for 7days was measured by qRT-PCR. Mann-Whitney tests were used to measure the mRNA level differences between groups. ** = p<0.01, * = p<0.05, ns = not significant. STK3 and STK4 mRNA expression levels in DU145 (E) and LAPC4 (F) organoids after 7-day treatment with DMSO (control) or ETC-159 (1 µM). Mann-Whitney tests were used to measure the mRNA level differences between groups. ** = p<0.01, * = p<0.05, ns = not significant. (G, H) Co-expression analysis between WNT5A and Hippo pathway components (STK3, STK4, SAV1, MOB1A, MOB1B and LATS1) was performed with TCGA-Prostate Cancer (G) or Melanoma (H) datasets by cBioPortal. R values (Spearman) and p-values were shown.

WNT5a loss in DU145 organoids (Fig. 7C) and in LAPC4 organoids (Fig. 7D) also caused a decrease in MST1/STK4 and MST2/STK3 mRNA, and these levels were then increased by treatment with WNT5a. We also examined the effects of ETC-159 in wild-type DU145 and LAPC4 organoids. In both, ETC-159 caused a significant decrease in MST1/STK4 and MST2/STK3 mRNA (Fig. 7E, F). Together these results indicate that WNT5a-mediated increases in Hippo pathway components occur at the mRNA level, although additional effects on protein translation and stability are also possible.

Based on these observations, we addressed whether there was a correlation between WNT5a and expression of Hippo pathway components in clinical samples. Notably, in the TCGA dataset of primary PCa, WNT5a was significantly correlated with the expression of MST1 and MST2, as well as additional Hippo pathway components (MOB1A, MOB1B, LATS1, SAV1, and NF2) (Fig. 7G). We also examined the association of WNT5a with Hippo components in melanoma and found a similar correlation (Fig. 7H).

In contrast to WNT5a, Hippo pathway components were not significantly correlated with ROR2 expression in the TCGA PCa (Fig. S4A) or TCGA melanoma cases (Fig. S4B). However, ROR2 mRNA in TCGA PCa cases was positively correlated with increased expression of YAP1 and of YAP1 regulated genes including IGFBP3, AXL, CCN1 and CCN2 (encoding CTGF), although not BIRC5 (Fig. 8A). These associations also were found in additional metastatic PCa and melanoma datasets (Fig. S5A-C). These findings suggested that YAP1/TEAD may be driving the expression of ROR2, possibly as a negative feedback pathway. Consistent with this hypothesis, treatment of DU145 organoids with verteporfin decreased expression of ROR2, and also decreased WNT5a (Fig. 8B). Analysis of mRNA by RT-qPCR confirmed that YAP1 inhibition was decreasing ROR2 and WNT5a mRNA (Fig. 8C). Notably, a previous study showed that YAP1 overexpression in MCF10A breast cancer cells increased expression of WNT5a, which was found to antagonize canonical Wnt/β-catenin signaling (20). Taken together, these data suggest that increased WNT5a and ROR2 in response to YAP1 activation reflects a negative feedback loop that stimulates Hippo to suppress YAP1 activity (Fig. 8D).

Figure 8. Inhibition of YAP1 activity decreases WNT5A and ROR2 expression.

(A) Co-expression analysis between ROR2 and YAP pathway components (YAP1, IGFBP3, AXL, CCN1/ CYR61, CCN2/CTGF, BIRC5/Survivin) was performed with TCGA-Prostate Cancer datasets by cBioPortal. R values and p-values are shown. (B) Immunoblotting of ROR2 and WNT5A in DU145 organoids treated with DMSO or verteporfin (1–10 µM) for 7 days. (C) ROR2, WNT5a, IGFBP3 and CTGF mRNA expression in DU145 2D-cultured cells without or with verteporfin (VP, 5 µM for 24 hours) treatment was measured by qRT-PCR. Mann-Whitney tests were used to measure the mRNA level differences between groups. ** = p<0.01, * = p<0.05, ns = not significant. (D) Model depicting WNT5a/FZD-mediated suppression of Hippo versus WNT5a/ROR2-mediated activation of Hippo. The figure suggests a role for p21 activated kinases in FZD signaling to Hippo, but this remains to be established. Similarly the direct or indirect mechanisms connecting ROR2 to Hippo are a current focus of investigation.

Interestingly, there was a progressive decrease in ROR2 mRNA in going from benign prostate to localized PCa and to mCRPC (Fig. S6A, B). Moreover, tumors with lower ROR2 had increased expression of multiple gene sets associated with more aggressive behavior (Fig. S6C, D). Consistent with this correlation, lower ROR2 was associated with decreased disease-free and overall survival in TCGA PCa cases (Fig. S6E, F). Lower ROR2 was similarly associated with shorter time to biochemical recurrence in another primary PCa dataset (Fig. S6G), and was associated with decreased overall survival in men with mCRPC (Fig. S6H). In conjunction with the data above, we suggest that persistent ROR2 expression may be indicative of tumors that have increased YAP1 activity and dependence, but have retained a physiological negative feedback loop that can modulate YAP1 signaling, and that these tumors may be responsive to treatment with WNT5a.

Discussion

Activation of canonical Wnt/β-catenin signaling is generally oncogenic, but β-catenin independent noncanonical signaling can activate several pathways and have context dependent oncogenic or tumor suppressive effects. We initially observed in 3D/organoid PCa models that PORCN inhibitors stimulated growth, and that WNT5a treatment could prevent this stimulation, indicating growth of these cells was being suppressed by autocrine noncanonical Wnt signaling. These effects were also seen in vivo with PORCN inhibition, and with a WNT5a peptide mimetic, Foxy5. The suppressive effects of WNT5a were abrogated after ROR2 depletion, supporting the conclusion that this Wnt response was mediated through ROR2. Although noncanonical Wnt signaling had been linked previously to suppression rather than activation of Hippo (20–22), we next examined effects of WNT5a on YAP1. Notably, WNT5a treatment resulted in an increase in YAP1 phosphorylation, consistent with Hippo pathway activation (41,42). Hippo pathway activation was further indicated by the finding that WNT5a increased phosphorylation of the upstream LATS and MST kinases. Moreover, we confirmed that the WNT5a-stimulated phosphorylation of LATS and YAP1 could be prevented by CRISPR-mediated depletion of MST1 or 2. Together these findings identify Hippo pathway activation as a mechanism that mediates the growth suppressive effects of WNT5a.

We confirmed the WNT5a mediated activation of Hippo in ROR2 expressing melanoma cells. Conversely, consistent with previous reports (20–22), WNT5a suppressed Hippo signaling in melanoma cells with low or absent ROR2 expression. Our data suggest that this latter Hippo suppression may be mediated through inactivation of NF2/Merlin, which serves as a scaffold at the plasma membrane to facilitate Hippo signaling (44–46). Mechanistically, previous studies have shown that YAP1/TAZ could be stabilized by G-protein coupled receptors (GPCRs) through increases in Gα proteins and subsequent activation of Rho family small GTPases (49–51), which is mediated by Gα binding to RhoGEFs (52). Similarly, Hippo pathway suppression by WNT5a appears to be mediated by a FZD family receptor acting as a GPCR to increase Gα and activate Rho (20). Notably, increased Gβγ downstream of GPCRs also activates RAC and CDC42 (53), and Gαs was shown recently to activate CDC42 through PDZ-RhoGEF (54). This activation of these p21 GTPases (RAC and CDC42) indicates that PAK1 (p21 activated kinase 1), which is a negative regulator of NF2/Merlin through S518 phosphorylation (which displaces it from the plasma membrane) (47,48), may also be activated downstream of WNT5a/FZD signaling.

Consistent with PAK1 contributing to the suppression of Hippo signaling through Merlin, we found that WNT5a-mediated activation of YAP1 in melanoma cells was associated with increased S518 phosphorylation of Merlin, and that this YAP1 activation was abrogated by Merlin depletion. In contrast, WNT5a/ROR2-mediated Hippo activation was not associated with altered (and specifically decreased) Merlin phosphorylation. We hypothesize this reflects a switch from WNT5a signaling through FZD to ROR2, although the precise basis for this switch and its effectors directly downstream of ROR2 remain to be determined (Fig. 8D). As noted in the Results, we have not observed clear Hippo pathway activation in response to WNT5a under standard 2D culture conditions. This does not appear to reflect altered levels of ROR2 expression, and we presume it does reflect a difference in ROR2 interacting proteins or other intermediaries between ROR2 and Hippo. Importantly, responses in the 3D/organoid cultures are reflective of the response in vivo to Foxy5. However, as the WNT5a stimulation studies outlined here have been carried out over several days in 3D/organoid culture conditions (or in vivo), we do not yet know if ROR2 is directly or indirectly activating Hippo. Addressing precisely how ROR2 is linked to Hippo is a focus of current studies.

Several studies have found an association between increased YAP1 and more aggressive PCa, although it was also found to be reduced in NEPC (55–59). Furthermore, a recent study found that a major stem-cell like PCa subtype was associated with YAP1/TAZ activation (60). There are conflicting reports regarding the role of WNT5a and noncanonical Wnt signaling in PCa, with several studies linking this pathway to more aggressive behavior and progression to mCRPC and NEPC (28–32) and others finding it associated with better prognosis in localized PCa (33,34). Intriguingly, WNT5a in the bone microenvironment was found to induce dormancy in PCa cells (34,35), but this was associated with decreased active β-catenin, and the possible role of Hippo was not explored. Notably, we found that WNT5a could increase the expression of multiple Hippo signaling elements, and WNT5a expression also was correlated with expression of these Hippo elements in clinical datasets. This induction appears to be driven independently of Hippo signaling, but could serve to prime cells for Hippo pathway activation by diverse signals.

Conversely, we found that YAP1 induces expression of ROR2 and WNT5a. The YAP1 induction of WNT5a has been reported previously in breast cancer cells (20), and its function was suggested to be suppression of canonical Wnt signaling. We suggest that it (along with increased ROR2) acts as a negative feedback loop, wherein YAP1/TAZ activation feeds back through increased WNT5a/ROR2 signaling to activate Hippo (Fig. 8D). Notably, increased ROR2 expression, although associated with increased YAP1 activity, also is associated with earlier stage PCa and with improved disease free and overall survival. We hypothesize the increased ROR2 in these cases is indicative of cells that are less aggressive as they maintain the negative feedback loop regulating YAP1 activity, and may also be more responsive to other growth regulatory signals.

From a therapeutic perspective, this suggests that increased ROR2 may be a biomarker of tumors that are more likely to respond to WNT5a and related agents. This may be particularly the case for PCa (or other cancers) that both express ROR2, and are being substantially driven by YAP1/TAZ. Previous studies indicate that ROR2 is frequently expressed in many solid tumors, and may be particularly high in breast cancer, glioblastoma, and neuroblastoma, but may be decreased in mCRPC (61,62). However, further studies with validated antibodies are needed to firmly establish its distribution and correlation with YAP1/TAZ (63). Notably, a phase 1 clinical trial of a ROR2 antibody drug conjugate in solid tumors is underway, and may provide further insight into ROR2 expression (NCT03504488). Finally, while our data indicate that ROR2 expression will enrich for patients who would respond to a WNT5a mimetic drug, it is not unlikely that only a subset will respond, and that additional biomarkers indicative of tumors where ROR2 is linked to Hippo will further improve patient selection.

Significance:

WNT5a signaling through ROR2 activates the Hippo pathway to downregulate YAP1/TAZ activity and suppress tumor growth, identifying ROR2 as a potential biomarker to identify patients that could benefit from WNT5a-related agents.