NSC 696085 – Phenanthroline Monohydrate

FRMD6 has tumor suppressor functions in prostate cancer

Abstract

Available tools for prostate cancer (PC) prognosis are suboptimal but may be improved by better knowledge about genes driving tumor aggressiveness. Here, we identiﬁed FRMD6 (FERM domain-containing protein 6) as an aberrantly hypermethylated and signiﬁcantly downregulated gene in PC. Low FRMD6 expression was associated with postoperative biochemical recurrence in two large PC patient cohorts. In overexpression and CRISPR/Cas9 knockout experiments in PC cell lines, FRMD6 inhibited viability, proliferation, cell cycle progression, colony formation, 3D spheroid growth, and tumor xenograft growth in mice. Transcriptomic, proteomic, and phospho-proteomic proﬁling revealed enrichment of Hippo/YAP and c-MYC signaling upon FRMD6 knockout. Connectivity Map analysis and drug repurposing experiments identiﬁed pyroxamide as a new potential therapy for FRMD6 deﬁcient PC cells. Finally, we established orthotropic Frmd6 and Pten, or Pten only (control) knockout in the ROSA26 mouse prostate. After 12 weeks, Frmd6/Pten double knockouts presented high- grade prostatic intraepithelial neoplasia (HG-PIN) and hyperproliferation, while Pten single-knockouts developed only regular PIN lesions and displayed lower proliferation. In conclusion, FRMD6 was identiﬁed as a novel tumor suppressor gene and prognostic biomarker candidate in PC.

Introduction

These authors contributed equally: Jakob Haldrup, Siri. H. Strand Prostate cancer (PC) is the most commonly diagnosed noncutaneous malignancy and the second leading cause of cancer-related death amongst men in the Western world [1]. While many PCs remain indolent throughout the patient’s lifetime, others progress to highly aggressive disease, causing signiﬁcant morbidity and mortality. Routine prog- nostic tools cannot clearly distinguish aggressive from indolent PC, resulting in overtreatment of many clinically insigniﬁcant tumors and suboptimal treatment of aggressive tumors [2].

In the present study, we aimed to identify novel mole- cular drivers of PC aggressiveness. By RNA-sequencing (RNA-Seq) of non-malignant (NM) vs. PC tissue samples as well as less vs. more aggressive isogenic PC cell lines, we identiﬁed FRMD6 (FERM domain-containing protein 6) as a novel tumor suppressor gene (TSG) candidate in PC. FRMD6 belongs to the Ezrin/Radixin/Moesin protein family and has been proposed as the human orthologue of Drosophila Expanded [3]. FRMD6 can bind to actin ﬁla- ments [4] and nectins [5], thus regulating actomyosin con-
tractility in the cytoskeleton and epithelial cell–cell junction complexes to maintain epithelial structure [6]. FRMD6 has previously been identiﬁed as an upstream regulator of the Hippo signaling cascade regulating mammalian cell growth and differentiation, including cell contact inhibition, apop- tosis, proliferation, and tissue regeneration [7]. Furthermore, the deregulation of Hippo pathway components has been reported in various human cancers, including PC [8–10]. In this study, we investigated the expression pattern, biomarker potential, and biological role of FRMD6 in PC development and progression. By analyzing transcription proﬁling data from >800 PC patients, we found the sig- niﬁcant downregulation of FRMD6 in PC vs. NM prostate tissue specimens and in metastatic vs. localized PC. Low FRMD6 expression was associated with aberrant promoter hypermethylation in PC tumors and with higher post- operative biochemical recurrence risk. Furthermore, func- tional studies in 2D, 3D, and xenografts showed that FRMD6 knockout increased PC cell viability and pro- liferation, and stimulated colony formation and cell cycle progression. By integrative analysis of the transcriptome, proteome, and phospho-proteome of PC cells, we identiﬁed Hippo, mTOR, and c-MYC signaling as mediators of FRMD6 and PC aggressiveness. For in vivo validation, we established a murine orthotopic Frmd6/Pten double- knockout model, where prostatic cells with combined Frmd6/Pten knockout displayed higher cell proliferation than Pten single-knockout controls. In summary, our results establish FRMD6 as a novel TSG and prognostic biomarker candidate in PC.

Results

FRMD6 is downregulated in PC

To identify novel genes involved in PC development and progression, we analyzed total RNA-Seq data from 18 NM and 55 PC tissue specimens (Set#1; Supplementary Table S1), one normal primary prostate epithelial cell line (PrEC), and two pairs of less/more aggressive isogenic PC cell lines (PC3/PC-3M and DU145/DU145-MN1; Supplementary Table S2). For candidate selection, we speciﬁcally searched for transcripts that were differentially expressed in NM vs. PC tissue specimens as well as in less vs. more aggressive PC cell lines. We found that FRMD6 was signiﬁcantly downregulated in PC vs. NM tissue specimens (Set#1, false discovery rate (FDR) = 9.64E − 08, Fig. 1A) and in the more aggressive PC-3M and DU145-MN1 cell lines vs. their isogenic counterparts PC3 and DU145 (Fig. 1B). These results were conﬁrmed by RT-qPCR analysis of FRMD6 expression in 28 independent patient samples (19 PC and 9 NM tissue specimens) (Supplementary Fig. S1).For external validation, we used FRMD6 transcriptional expression data from four public PC patient sets [11–14].

FRMD6 was signiﬁcantly downregulated in PC vs. NM prostate tissue samples in The Cancer Genome Atlas (TCGA/PRAD) cohort [11, 15, 16] (479 PC vs. 29 NM; p < 0.0001; Fig. 1C), in the Taylor cohort [12] (150 PC vs. 29 NM; p < 0.0001; Fig. 1D), and in the Grasso cohort [13] (57 PC vs. 22 NM; p < 0.001; Fig. 1E). FRMD6 was further downregulated in advanced mCRPC samples (n = 29) analyzed in the Grasso cohort (p ≤ 0.0001; Fig. 1E). Moreover, in the Long cohort [14], FRMD6 expression was signiﬁcantly lower in PC patients with vs. without BCR (p = 0.0027; Fig. 1F). These results indicate that PC development and progression are associated with the downregulation of FRMD6 and suggest a possible tumor suppressor role for FRMD6 in PC. FRMD6 downregulation in PC is associated with aberrant promoter hypermethylation To explore the mechanism(s) behind FRMD6 down- regulation in PC, we investigated whether the FRMD6 gene is subject to mutations or copy number alterations (CNAs) in PC tumors. Based on whole exome/genome sequencing and SNP/comparative genomic hybridization array data from 13 PC studies in cBioPortal (www.cbioportal.org), genomic alterations affecting the FRMD6 gene were detected in only 18 out of 3441 (0.5%) primary PC samples. In contrast, DNA methylation array data available in -house (33 NM and 43 PC; Set#2 + 3 [15–17]) and from TCGA (29 NM and 497 PC samples [18, 19]) showed very common and highly PC-speciﬁc promoter hypermethylation of FRMD6 in both patient sets with AUCs ≥0.88 in receiver operating characteristics (ROC) curve analyses (Fig. 1G–J; Supplementary Fig. S2). There was a signiﬁcant inverse correlation between expression and FRMD6 promoter methylation levels in both patient sets (Set#3: Rho = −0.814, p < 0.0001. TCGA: Rho = −0.613, p < 0.0001, Fig. 1K, L), consistent with epigenetic silencing. These results indicate that FRMD6 is a common target for aberrant promoter hypermethylation in PC, whereas somatic muta- tions and CNAs are rare. Prognostic potential of FRMD6 To investigate the prognostic potential of FRMD6 tran- scriptional expression, we performed Kaplan–Meier ana- lyses of post-operative BCR in two large PC patient cohorts with sufﬁcient clinical follow-up (Taylor [12, 20] and Long [14] cohorts; Supplementary Table S3). Patients with high FRMD6 expression had signiﬁcantly reduced risk of BCR in the Taylor cohort (p = 0.036; Fig. 1M), while this was borderline signiﬁcant in the Long cohort (p = 0.056; Fig. 1N). Similarly, in univariate Cox regression analysis, high FRMD6 expression was associated FRMD6 to a multivariate model of clinicopathological factors improved predictive accuracy (C-index) from 0.788 to 0.798 and 0.706 to 0.747 in the Taylor and Long cohort, respectively. These results indicate that reduced FRMD6 expression is an adverse predictor of PC recurrence, inde- pendent of clinicopathological parameters. FRMD6 inhibits PC cell viability, proliferation, cell cycle progression, and anchorage-independent growth in vitro To explore the biological function of FRMD6 in PC, we transfected PC-3M and DU145-MN1 cells (both have low endogenous FRMD6 expression levels; Fig. 1B) with a plasmid encoding the most predominant isoform of FRMD6. Overexpression of FRMD6 was veriﬁed by RT- qPCR and Western blotting, and lasted for at least 96 h (Supplementary Fig. S3A-D). Ectopic expression of FRMD6 signiﬁcantly attenuated PC-3M and DU145-MN1 cell viability and proliferation at 72 h compared to mock-transfected controls (p < 0.05; Fig. 2A–C).Next, we used CRISPR/Cas9 to establish PC3 and DU145 cell lines with genomic knockout (KO) of FRMD6 (PC3_KO_FRMD6 and DU145_KO_FRMD6) and scram- bled controls (PC3_SCR and DU145_SCR). We chose PC3 and DU145 due to their high endogenous FRMD6 expres- sion level (Fig. 1B). The knockout was validated by TIDE analysis, RT-qPCR, and Western blotting (Supplementary Fig. S4A–C). Both cell viability and proliferation were signiﬁcantly increased for both PC3_KO_FRMD6 and DU145_KO_FRMD6 as compared to scrambled controls (p < 0.05; Fig. 2D–F). Transient FRMD6 rescue (validated by RT-qPCR; Supplementary Fig. S5A, B) attenuated the increased cell viability in both PC3_KO_FRMD6 and DU145_KO_FRMD6 to parental levels at 72 h (p < 0.005; Fig. 2G). Cell cycle analysis revealed a sig- niﬁcantly higher proportion of PC3_KO_FRMD6 and DU145_KO_FRMD6 cells in the S phase as compared to their wild-type counterparts (32% vs. 19%, and 31% vs. 13%, respectively, p < 0.005; Fig. 2H). Moreover, in soft agar colony formation assays, both PC3_KO_FRMD6 and DU145_KO_FRMD6 cells formed signiﬁcantly more colonies than parental PC3 and DU145 (p < 0.005; Fig. 2I). In summary, FRMD6 inhibited PC cell viability, proliferation, cell cycle progression, and anchorage- independent growth. Fig. 2 Functional studies in isogenic PC3, PC-3M, DU145, and DU145-MN1 cell lines. A Viability of PC3 and DU145 cells with/ without FRMD6 overexpression/mock. Results are presented relative to the corresponding WT and SCR. B Proliferation of PC-3M and (C) DU145-MN1 cells with/without transient FRMD6 overexpression. Cell index was recorded every 15 min for 96 h by xCELLigence. D Via- bility of PC3 and DU145 cells with/without FRMD6 KO or SCR. Results are presented relative to the corresponding WT. E Proliferation of PC3 and (F) DU145 cells with/without FRMD6 KO. Cell index was recorded every 15 min for 96 h by xCELLigence. G Viability of PC3_KO_FRMD6 and DU145_KO_FRMD6 cells with/without transient FRMD6 rescue/mock. Results are presented relative to the corresponding KO or mock. H Cell cycle analysis of PC3 and DU145 with/without FRMD6 KO. The percentage of cells on G1, S, or G2 phase are presented. I Anchorage-independent colony formation for PC3 and DU145 cells with/without FRMD6 KO/SCR. Colonies were counter under light microscope after 21 days. All viability readouts were performed 72 h post seeding using AlamarBlue. Signiﬁcant P values (two-sided Student’s t-test) are marked by *p < 0.05, **p < 0.005, ***p < 0.0005, or ****p < 0.0001 for all ﬁgures except H (Chi- square test). Results are presented as the ±SD of a minimum of three replicates. WT wild type, SCR scrambled control, KO knockout. FRMD6 inhibits 3D spheroid proliferation Next, we investigated the effect of FRMD6 knockout in 3D cell culture. PC3_KO_FRMD6 and DU145_KO_FRMD6 spheroids grew signiﬁcantly faster than their corresponding controls (p < 0.05; Fig. 3A, B and Supplementary Fig. S6A). We observed no differences in the number of spheroids formed for PC3 nor DU145 cells upon knockout of FRMD6 (28/28 wells; 100% for each cell line). After day 7, the faster growing PC3_KO_FRMD6 spheroids, but not DU145_KO_FRMD6, displayed an increasingly irregular morphology characterized by multiple cellular protrusions extending beyond the roun- ded spheroid core (Fig. 3C and Supplementary Fig. S6A). In addition, we found a signiﬁcant increase in cell death and apoptosis for PC3_KO_FRMD6 spheroids (p < 0.0001; Fig. 3C; Supplementary Fig. S6B, C), whereas cell death, but not apoptosis, was increased in DU145_KO_FRMD6 spheroids (p < 0.0001; Supplementary Fig. S6D, E). This is likely attributable to hypoxia and nutrition depletion in the center of hyperproliferation spheroids. Furthermore, the altered mor- phology may reﬂect a more aggressive phenotype with dis- ruption of cell-cell junctions and epithelial architecture consistent with previous reports [6]. Fig. 3 FRMD6 inhibits 3D spheroid and xenograft tumor growth. Spheroid size (mm2) at indicated time points for (A) PC3 and (B) DU145 cells with/without FRMD6 KO/SCR. Results are the mean ±SD of a minimum of 15 replicates. C Representative pictures of 3D spheroid size/morphology, necrosis, and apoptosis at indicated time points. An untreated (UT), propidium iodide (PI), and caspase (Casp) readout were performed to measure the level of necrosis and apoptosis, respectively, in the indicated PC3 cell line. Scale bars: 500 μm. D Indicated cells were embedded in matrigel and subcutaneously injected into the ﬂank of six immunocompromised (nude) NMRI mice. E Subcutaneous tumor growth curve of PC3 (black), PC3_SCR (gray), and PC3_KO_FRMD6 (red) cells in NMRI male nude mice (n = 6 per group). Mice were sacriﬁced at the time points marked by dotted vertical lines. F Dot plot of the maximum tumor volume for each xenograft. Results are the mean ±SD of six replicates. Signiﬁcant P values (two-sided Student’s t-test) are marked by *p < 0.05, **p < 0.005, ***p < 0.0005, or ****p < 0.0001. WT wild type, SCR scrambled control, KO knockout. FRMD6 inhibits tumor xenograft growth in vivo For in vivo analysis, PC3_KO_FRMD6 cells and their par- ental and scrambled controls were subcutaneously injected into the ﬂank of NMRI male nude mice (n = 6 in each group; Fig. 3D). The tumor volume of PC3_KO_FRMD6 xeno- grafts was signiﬁcantly larger and tumors grew signiﬁcantly faster than parental PC3 and PC3_SCR xenografts (p < 0.05; Fig. 3E, F). Due to substantial tumor burden, three mice with PC3_KO_FRMD6 xenografts were sacriﬁced after 24 days and another three after 32 days, whereas all PC3 and PC3_SCR xenografts were sacriﬁced at the predeﬁned end- point of 45 days. Taken together, the 2D, 3D, and xenograft results highlight FRMD6 as a putative TSG in PC. FRMD6 loss activates c-MYC signaling To elucidate how FRMD6 exerts its proposed tumor sup- pressor function, we conducted RNA-Seq and proteomic proﬁling (Supplementary Data 1 and 2, respectively) of parental PC3 and PC3_KO_FRMD6, which showed pro- minent phenotypic changes in vitro and in vivo. Gene Set Enrichment Analysis found that the targets of the oncogenic c-MYC transcription factor (TF) were particularly enriched in the transcriptome of PC3_KO_FRMD6 (FDR < 0.005; Fig. 4A, B). Concordant with this, gene ontology (GO) analysis of the top 500 differentially expressed proteins revealed signiﬁcant enrichment of c-MYC targets in PC3_KO_FRMD6 (Fig. 4C). Integrative analysis showed enrichment of multiple ribosomal, peptide biogenesis and mitochondrial gene sets in PC3_KO_FRMD6 (Fig. 4D), which is consistent with both c-MYC activation [21] as well as the oncogenic phenotypes associated with FRMD6 loss (Figs. 2D–F; 3A–C; Supplementary Fig. S6A). Moreover, we found that extracellular matrix (ECM) and multiple ECM binding components, such as collagen, ﬁbronectin, integrin, and laminin gene sets, were sig- niﬁcantly depleted in PC3_KO_FRMD6 (Fig. 4B, D). This is consistent with the irregular spheroid morphology observed in PC3_KO_FRMD6 spheroids (Fig. 3C) and regulation of cell–cell junctions critical for epithelial structure [6]. Integrative analyses showed no evidence for enrichment of gene sets related to cell death or apoptosis, supporting the notion that the increased cell death and apoptosis rates observed for the 3D spheroids were likely attributed to nutrition/oxygen depletion. Pyroxamide inhibits cell proliferation To identify drugs that may inhibit the oncogenic phenotypes associated with FRMD6 loss, we employed the Con- nectivity Map (CMap) [22]. Querying CMap with a sig- nature containing the top 150 most upregulated transcripts in PC3_KO_FRMD6 (Supplementary Data 3), we obtained an inversely correlated connectivity score (CS) to c-MYC knockdown (CS: −99.95; Fig. 5A). To identify potential drugs, we ranked the ~3000 drugs in CMap according to CS and recognized several histone deacetylase inhibitors, including pyroxamide, givinostat, panobinostat, and vor- inostat among top candidates (Fig. 5A). We selected the HDAC1 inhibitor pyroxamide, based on its inversely cor- related CS in PC3 and VCaP (−99.43 and −97.13, respectively), while minimizing the CS in non-PC cell lines (Fig. 5A). Treatment with pyroxamide resulted in a potent reduction of cell viability in a dose-dependent manner in PC3_KO_FRMD6 (IC-50 ≈ 6 µM; Supplementary Fig. S7). Concordantly, pyroxamide signiﬁcantly inhibited PC3_KO_FRMD6 cell proliferation (p < 0.005; absolute reduction in cell index at 120 h = 1.43; Fig. 5B), while only a small effect was observed in parental PC3 cells (p < 0.05; absolute reduction in cell index at 120 h = 0.19; Fig. 5B). These results suggest that FRMD6 deﬁcient PC cells may be vulnerable to treatment with pyroxamide and provides pre- clinical proof-of-principle for drug repurposing experiments. FRMD6 activates Hippo signaling in PC cells As FRMD6 has been proposed as an upstream activator of the Hippo signaling pathway [7], we proﬁled the phospho- proteome in PC3_KO_FRMD6 and parental PC3. Hier- archical clustering separated both the proteome and phospho- proteome into two clusters according to FRMD6 status (Sup- plementary Fig. S8). Querying the top 250 differentially expressed phospho-peptides (Supplementary Data 4), we observed enrichment of GO terms related to mTORC1, het- erochromatin, c-MYC/MAX, and Hippo/YAP signaling (Fig. 5C). Concordant with the literature, we focused on the Hippo pathway and investigated the abundance of protein and phospho-peptides for canonical Hippo members (Supplemen- tary Fig. S9) [23]. The bulk of Hippo core proteins (MST1/2, MOB1A, SAV1 and YAP) and most Hippo-associated members (e.g., KIBRA, MERLIN, PPP2CA, RASSF1, and AMOT) were signiﬁcantly depleted upon FRMD6 knockout (FDR < 0.05; Fig. 5D, E). Likewise, the phosphorylation level of the canonical Hippo phospho-peptides (MOB1A, LATS1, and YAP) were signiﬁcantly depleted at the indicated pSTY residues (FDR < 0.05; Fig. 5F) [7, 8]. Consistent with Hippo- YAP dependent c-MYC activation, we observed a signiﬁcant upregulation of several protein members of the AP-1 TF complex, including JUN, JUNB, and FOSL2 (Supplementary Fig. S10) [24]. These results suggest that FRMD6 is as a potent activator of Hippo signaling, and that FRMD6 loss leads to disruption of the Hippo phosphorylation cascade and oncogenic transformation. To the best of our knowledge, this is the ﬁrst report to associate FRMD6 with Hippo and c-MYC signaling in PC cell lines. Fig. 4 Pathway analysis of PC3_KO_FRMD6 vs. parental PC3 based on RNA-Seq data. Gene Set Enrichment Analysis (GSEA) using (A) ´Hallmark´ and ´C6´ collection for PC3_KO_FRMD6 (tri- angles) vs. PC3 (squares). The normalized enrichment score (NES) of signiﬁcantly depleted or enriched gene sets (FDR < 0.005) are colored according to FDR q value. B Log2 transformed FPKM values of selected differentially expressed genes. C Pathway analysis (Enrich) of top 500 most abundant proteins in PC3_KO_FRMD6 vs. parental PC3 cells. Enriched transcription factors (TFs) and GO/Pathways are pre- sented according to combined Z-score. D GSEA using the ´C5_Gene Ontology´ collection (FDR < 0.002) for PC3_KO_FRMD6 (triangles) vs. PC3 (squares). The normalized enrichment score (NES) of sig- niﬁcantly depleted or enriched gene sets (FDR < 0.002) is colored according to FDR q value. All analyses were corrected for multiple testing. GO gene ontology. In vivo knockout of Frmd6 in the murine prostate To investigate the function of FRMD6 in an androgen receptor (AR) positive cell line, we established LNCaP_- KO_FRMD6 and observed a signiﬁcant increase in cell viability compared to scrambled or parental controls (Fig. 6A; Supplementary Fig. S11A–C). To further char- acterize the endogenous role of Frmd6 in a non-tumorigenic and AR positive in vivo environment, we injected Adeno- Associated Virus (AAV) particles containing sgRNAs for Frmd6 and Pten (combined knockout) into the anterior prostatic lobe of Cas9 transgenic male mice (n = 6). As controls (n = 6), we used AAV particles with a sgRNA targeting only Pten, as orthotopic knockout of Pten is known to induce prostatic intraepithelial neoplasia (PIN) in mouse models and accelerate tumor formation [25, 26]. Anterior prostate tissue samples were analyzed 3 months post transduction for indels generated by CRISPR/Cas9. Fig. 5 Drug repurposing experiments and phospho-proteomics. A Connectivity scores of c-MYC knockdown and several HDAC inhi- bitors, including pyroxamide, identiﬁed by querying top 150 upregu- lated genes in PC3_KO_FRMD6 vs. parental PC3 to Connectivity Map. B Cell proliferation of PC3 and PC3_KO_FRMD6 cells treated with pyroxamide (6 μM). Pyroxamide was added after 48 h and readouts were performed every 12 h using xCELLigence for 72 h. Results are presented as ±SD of a minimum of three replicates. C Pathway analysis (Enrichr) of the top 250 most abundant phospho-peptides in PC3_KO_FRMD6 vs. parental PC3 cells. Enriched tran- scription factors (TFs) and GO/Pathways are presented according to combined Z-score and adjusted for multiple testing. D Proteomics analysis of core Hippo kinase enzymes and (E) other Hippo regulatory proteins in PC3_KO_FRMD6 vs. parental PC3. F The phosphoryla- tion level of indicated serine (S), threonine (T), or tyrosine (Y) residues for signiﬁcantly deregulated Hippo proteins are shown. Signiﬁcant P values (two-sided Student’s t-test) are marked by *p < 0.05 or **p < 0.005. TIDE analysis revealed indels in Pten in the controls, and in both Frmd6 and Pten in the double knockouts, conﬁrming effective in vivo genomic editing in prostatic tissue (Fig. 6B). Furthermore, histopathological analysis revealed that Frmd6/Pten double-knockout promoted hyperproli- feration and premalignant high-grade PIN (HG-PIN) lesions (Fig. 6C, D; lower), while Pten deﬁcient specimens dis- played a less aggressive phenotype with PIN lesions (Fig. 6C, D; upper). IHC staining for p-Akt showed increased activation of Akt in both controls (Fig. 6E; upper) and in double knockouts (Fig. 6E; lower), consistent with Pten loss in PrEC. IHC staining for Frmd6 conﬁrmed luminal Frmd6 expression in the wild-type mouse prostate (Fig. 6F, G) as well as loss of protein expression in areas with HG-PIN lesions in the double knockouts (Fig. 6H; lower). Using Ki-67 staining, we quantiﬁed the number of proliferating cells in PIN areas of Pten single-knockouts and in HG-PIN areas of Frmd6/Pten double knockouts. There was a signiﬁcantly higher number of proliferating (Ki-67+) cells in Frmd6/Pten double-knockout as compared to Pten single- knockout samples (p < 0.05; Fig. 6I, J). These results indicate that Frmd6 loss promotes premalignant prostate carcinogen- esis and prostatic cell proliferation in prostatic tissue in vivo, further highlighting Frmd6 as a novel TSG in PC. Discussion This study is the ﬁrst to demonstrate FRMD6 tumor sup- pressor functions in PC. Speciﬁcally, functional studies in 2D, 3D, and xenografts showed that FRMD6 loss increased PC cell viability, proliferation, cell cycle progression, anchorage- independent colony formation, spheroid growth, and sub- cutaneous xenograft tumor growth. FRMD6 also promoted the development of premalignant HG-PIN in the murine prostate with concomitant Pten loss. Comprehensive transcriptome, proteome, and phospho-proteome proﬁling revealed activation of oncogenic c-MYC signaling and inactivation of the Hippo phosphorylation cascade. The pre- sent report is also the ﬁrst to demonstrate signiﬁcant down- regulation and aberrant promoter hypermethylation of FRMD6 in PC compared to NM prostate tissue samples from clinical patients, indicating an epigenetic silencing mechanism. Moreover, in two large PC patient cohorts, FRMD6 down- regulation predicted postoperative BCR independently of routine clinicopathological parameters. This highlights the potential of FRMD6 as a novel prognostic biomarker candi- date in PC and warrants further large-scale validation. Fig. 6 Orthotropic gene editing in transgenic Rosa26-LSL-Cas9- eGFP mice. A Viability of LNCaP cells with/without FRMD6 KO or SCR control. B TIDE analysis of Pten and Pten/Frmd6 using DNA isolated from bulk prostate tissue samples. C H&E staining of Pten (n = 6) and Pten/Frmd6 (n = 6) deﬁcient histological sections (4 μm) of the anterior prostate lobe at 3 months after AAV delivery. Scale bars: 100um. D Same as C but with scale bars: 20 um. E IHC staining of p- AKT (brown) for representative areas with Pten KO or Pten/Frmd6 KO. Scale bars: 20um. F Representative IHC of Frmd6 in the anterior lobe of wild-type ROSA26 mice. Scale bars: 100um. G Close-up of the dashed area in F with scale bars: 20um. A strong luminal Frmd6 straining is observed and conﬁrms endogenous Frmd6 expression in the murine prostate. H IHC staining Frmd6 (red) for representative areas with Pten KO or Pten/Frmd6 KO. Arrows high- light representative positive cells. Scale bars: 20um. I Ki-67 staining of proliferating cells in representative PIN areas. Scale bars: 20um. J Quantiﬁcation of Ki-67 positive cells from I. Data are presented as ±SD of 30 representative ﬁelds. Signiﬁcant P values (two-sided stu- dent’s t-test) are marked by *p > 0.05, **p > 0.005, ***p > 0.0005. N/ A not available, H&E Hematoxylin and eosin, SCR Scrambled, KO knockout, PX Pten + none, PF Pten + Frmd6.

This is the ﬁrst study to investigate the biological role of FRMD6 in PC development and progression. Our ﬁndings that FRMD6 knockout increased PC cell viability, pro- liferation, cell cycle progression, colony formation, and subcutaneous xenograft tumor growth are concordant with previous reports of tumor suppressor functions for FRMD6 in glioblastoma [27] and hepatocellular carcinoma [28] cell lines. Our results are also consistent with a study in breast cancer cell lines, where FRMD6 overexpression inhibited cell proliferation, colony formation, and anchorage- independent growth in vitro as well as xenograft tumor growth [29]. The irregular morphology observed for PC3 spheroids with FRMD6 knockout is in agreement with a previous study, which found FRMD6 to be involved in maintaining epithelial structure through cell–cell junction complexes [6]. The observation of increased epithelial markers and decreased mesenchymal markers in PC3_KO_FRMD6 may seem to conﬂict with the more aggressive phenotypes observed in these cells.

However, the biology behind aggressiveness is complex, and cells may not adhere to strict epithelial/mesenchymal pheno- types. Indeed, recent reports show that cancer cells can metastasize in ways alternative to the traditional EMT paradigm, e.g., by detaching in clusters or existing in stable hybrid epithelial/mesenchymal [30–32].

While previous studies [27, 29] demonstrated tumor suppressor function for FRMD6 in multiple cancer types, neither offered direct evidence for the effects being medi- ated through Hippo signaling. Our phospho-proteomics analyses showed that multiple Hippo kinases were depho- sphorylated in response to FRMD6 knockout. It has pre- viously been reported that this causes deactivation of the Hippo phosphorylation cascade and thus activation of YAP/ TAZ [7, 33]. We observed no difference in MST1/2 phos- phorylation, which is consistent with previous studies demonstrating MST-independent, but LATS-dependent, activation of YAP/TAZ [34].

We have searched the literature for a possible known link between Hippo and c-MYC signaling. We focused on the major signaling hub in the Hippo pathway, YAP/TAZ, which has no DNA-binding domains and thus relies on TF partners, e.g., TEAD1-4 and KLF4, to initiate transcription of target genes [33, 35]. Previous work by Zanconato et al. [24] showed that YAP/TAZ/TEAD-bound enhancers inter- act with the c–MYC promoter via chromatin looping in MDA-MB-231 cells. Our results are consistent with a model in which FRMD6 loss entails YAP/TAZ activation, and hence the direct activation of c-MYC transcriptional expression. In addition, AP-1 promotes YAP/TAZ/TEAD- dependent gene expression and enhances YAP-mediated oncogenic growth [24]. Thus, our results are also consistent with a model in which FRMD6 loss leads to increased c- MYC expression via activation of YAP/TAZ/TEAD/AP-1 interaction. In further agreement with this, we identiﬁed the HDAC1 inhibitor pyroxamide as a potent inhibitor of cell proliferation in FRMD6 deﬁcient PC cells. HDAC1 inhibitors are known to inhibit c-MYC by lysine 323 acetylation and interfere with chromatin remodeling in conjugation with HDAC1 [36–38], thus suggesting a possible mechanism for the vulnerability to pyroxamide in FRMD6 deﬁcient PC cells.

To investigate the biological role of FRMD6 in a non- tumorigenic environment, we utilized the Rosa26 transgenic mouse to co-target Pten and/or Frmd6 [39]. Loss of Pten is known to induce prostate gland enlargement and PIN lesions within a relatively short period of 12 weeks in Rosa26 mice [25]. This model is clinically relevant since the loss of PTEN is a recurrent genomic alteration in early stage PC patients [40]. Our ﬁndings indicate that loss of Frmd6 may play a driver role in the early steps of PC development, by accelerating the formation of HG-PIN lesions, which are considered to be precursors of malignant PC in mice [41] and humans [42]. We observe no devel- opment of either primary PC lesions or metastases in the Rosa26 model within 12 weeks. Further studies are war- ranted to determine whether Frmd6/Pten loss can drive the full transformation to malignant PC and/or metastatic pro- gression after an extended period of time. To the best of our knowledge, this is the ﬁrst study to show tumor suppressor function of endogenous Frmd6 in the murine prostate.

In conclusion, we identiﬁed FRMD6 as a novel TSG and prognostic biomarker candidate in PC. This was supported by clinical data from multiple PC patient sets, which showed gradual downregulation of FRMD6 expression from NM prostatic tissue, to primary PC tissue, and to advanced mCRPC samples as well as increased recurrence risk for patients with low FRMD6 expression. Furthermore, results from multiple in vitro and in vivo functional assays as well as multi-omics analyses showed that FRMD6 has tumor suppressor functions in PC.

Materials and methods

Cell line and animal experiments are described in Supple- mentary Methods. Primers and oligos are listed in Supple- mentary Table S4, S5 and S6.

Total RNA-Seq: Set#1 (discovery cohort) and isogenic PC cell lines RNA-Seq

All samples used for RNA-Seq had an RNA integrity number score >7. Whole transcriptome, strand-speciﬁc RNA-Seq libraries were prepared using Ribo-Zero Gold and ScriptSeq v2 kit (Epicentre) and paired-end sequencing on Illumina HiSeq 2000 (12–46 million reads/sample). Paired de- multiplexed FastQ ﬁles were generated using CASAVA software (Illumina) and reads were mapped to the human genome (hg19) using TopHat [43] with the Bowtie aligner [44]. HTSeq [45] was used to summarize reads per gene of interest with the “union” overlap resolution mode. Fragments per kilobase of exon per million fragments mapped (FPKM) values were calculated using the Tuxedo suite [46].

Set#1

A total of 55 PC and 18 NM prostate tissue samples (Sup- plementary Table S1) were collected at Department of Urol- ogy, Aarhus University Hospital, Denmark (2004–2016). Total RNA was isolated from fresh-frozen (tissue tek) and macro-dissected prostate (cancer) tissue samples from radical prosta- tectomy (RP) specimens using the RNeasy Mini Kit (Qiagen) as previously described [17, 47].

Set#2

Genome-wide DNA methylome data were available from 21 PC and 21 NM macro-dissected RP specimens analyzed on the Inﬁnium HumanMethylation450 (450K) BeadChip (Illumina) [17]. Raw 450K data were processed using the ChAMP package (v1.2.0) [48, 49] in R [50]. Each CpG site was
assigned a β‐value ranging from 0 (unmethylated) to 1 (fully methylated).

Set#3

Genome-wide DNA methylome data (450K) and matched total RNA-Seq data were available from a distinct set of 22 PC and 12 NM laser capture micro‐dissected RP prostate tissue samples, as previously described [15, 16].

External PC patient datasets

TCGA/PRAD [19]: RNA-Seq data (FPKM values) and DNA methylation data (Beta values, Illumina 450K array) were downloaded from the TCGA data portal [18, 19], as previously [51]. FRMD6 expression and DNA methylation data were available for 479 PC and 29 adjacent normal (AN) specimens.

Long et al. [14]: RNA-Seq data (FPKM values) for 106 RP patients were downloaded from the Gene Expression Omnibus (GEO) database (accession number GSE54460). For biochemical recurrence (BCR) free survival analysis, we excluded 10 patients who suffered BCR within 1 month from RP (Supplementary Table S3) and eight patients with missing clinical data.Taylor et al. [12]: Affymetrix human exon 1.0 ST array data and clinical data [20] were downloaded for 150 PC and 29 AN tissue specimens from GEO (accession number GSE21036). Clinical follow-up and FRMD6 expression data were available for 126 PC patients (Supplementary Table S3).Grasso et al. [13]: Whole Human Genome Microarray and Whole Human Genome Oligo Microarray data from 28 NM, 59 PC, and 35 mCRPC tissue specimens were downloaded from the GEO database (accession number GSE35988). FRMD6 expression data were available for 22 NM, 57 PC, and 29 mCRPC tissue specimens.

Bioinformatics and statistical analyses

Statistical analyses were performed using R [50] or STATA v. 15 (StataCorp). Differential expression ana- lysis was performed using EdgeR [52] with maximum complex dispersion for each gene. Differential expression in external datasets was assessed by the Wilcoxon rank-
sum test. Correlations between expression and methyla- tion levels were analyzed using Spearman’s test. Differ- ences in FRMD6 immunoreactivity scores and cell cycle distribution were assessed using the chi2 test. For func- tional studies in cell culture, Student’s t-test (two sided) was used to evaluate differences between treated and controls, unless stated otherwise.BCR-free survival was assessed by Kaplan–Meier and uni-/multivariate Cox regression analyses. For multivariate testing, all clinicopathological parameters signiﬁcant in univariate analysis were included. Predictive accuracy was estimated using Harrell’s C-index [53]. For Kaplan–Meier analysis, FRMD6 expression was dichotomized based on ROC curve analysis (BCR status at 36 months). Where applicable, correction for multiple testing (FDR) was per- formed according to Benjamini-Hochberg [54]. FDR and P values <0.05 were considered signiﬁcant. Study approval This study was approved by The Central Denmark Region Committee on Health Research Ethics (2000–0299) and notiﬁcation was given to the Danish Data Protection Agency (No. 2013-41-2041). Written informed consent was obtained from all patients. All animal experiments were approved by the Danish Animal Experiments Inspectorate (permission 2013-15-2934-00901, C2 (Xenografts) and 2016-15-0201-01083 (Orthotopic knockout of Frmd6 in Rosa26 mice) and housed according to Danish legislation NSC 696085 and the Directive 2010/63/EU on the protection of animals used for scientiﬁc purposes.