ortholog of Drosophila Melted impedes SMAD2 PNAS PLUS release from TGF-β receptor I to inhibit TGF-β signaling

Premalatha Shathasivama,b,c, Alexandra Kollaraa,c, Maurice J. Ringuetted, Carl Virtanene, Jeffrey L. Wranaa,f, and Theodore J. Browna,b,c,1

aLunenfeld-Tanenbaum Research Institute at Mount Sinai Hospital, Toronto, ON, Canada M5T 3H7; Departments of bPhysiology, cObstetrics and Gynaecology, dCell and Systems Biology, and fMolecular Genetics, University of Toronto, Toronto, ON, Canada M5S 3G5; and ePrincess Margaret Cancer Centre, University Health Network, Toronto, ON, Canada M5G 1L7

Edited by Igor B. Dawid, The Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, MD, and approved May 5, 2015 (received for review March 11, 2015)

Drosophila melted encodes a pleckstrin homology (PH) domain- The locus encompassing human VEPH1, 3q24-25, lies containing that enables normal tissue growth, metabo- within a region frequently amplified in ovarian cancer (6, 7). Tan lism, and photoreceptor differentiation by modulating Forkhead et al. (8) found that this locus was also amplified in 7 of 12 ep- box O (FOXO), target of rapamycin, and Hippo signaling pathways. ithelial ovarian cancer cell lines. A gene copy number analysis of Ventricular zone expressed PH domain-containing 1 (VEPH1) is the 68 primary tumors by Ramakrishna et al. (9) identified frequent mammalian ortholog of melted, and although it exhibits tissue- (>40%) VEPH1 gene amplification that correlated with tran- restricted expression during mouse development and is poten- script levels. We determined the impact of VEPH1 on gene ex- tially amplified in several human cancers, little is known of its pression in an ovarian cancer cell line using a whole-genome function. Here we explore the impact of VEPH1 expression in ovar- expression array. The results indicate a gene-expression profile ian cancer cells by gene-expression profiling. In cells with elevated that is partially consistent with that reported for Melted and raises VEPH1 expression, transcriptional programs associated with me- the possibility that VEPH1 may modulate TGF-β signaling. tabolism and FOXO and Hippo signaling were affected, analogous TGF-β is a pleiotropic cytokine that regulates tissue develop- to what has been reported for Melted. We also observed altered ment, repair, remodeling, and homeostasis by affecting cell pro-

β β CELL BIOLOGY regulation of multiple transforming growth factor- (TGF- ) target liferation, differentiation, survival, and migration. TGF-β signals . Global profiling revealed that elevated VEPH1 expression by inducing the formation of a heterotetrameric complex of type II suppressed TGF-β–induced transcriptional responses. This inhibitory β β β (T RII) and type I (T RI; ALK5) serine/threonine kinase trans- effect was verified on selected TGF- target genes and by reporter membrane receptors (10). Ligand-bound, constitutively active gene assays in multiple cell lines. We further demonstrated that TβRII phosphorylates TβRI, resulting in TβRI association with VEPH1 interacts with TGF-β receptor I (TβRI) and inhibits nuclear and C-terminal phosphorylation of Sma- and Mad-related protein accumulation of activated Sma- and Mad-related protein 2 (SMAD2). β 2 (SMAD2) and/or SMAD3 (SMAD2/3) (11). In the canonical We identified two T RI-interacting regions (TIRs) with opposing β effects on TGF-β signaling. TIR1, located at the N terminus, inhibits TGF- signaling pathway, phosphorylated SMAD2/3 rapidly β β dissociates from TβRI and oligomerizes with SMAD4. The canonical TGF- signaling and promotes SMAD2 retention at T RI, – similar to full-length VEPH1. In contrast, TIR2, located at the C-ter- SMAD2/3 SMAD4 complex then accumulates in the nucleus minal region encompassing the PH domain, decreases SMAD2 re- tention at TβRI and enhances TGF-β signaling. Our studies indicate Significance that VEPH1 inhibits TGF-β signaling by impeding the release of ac- tivated SMAD2 from TβRI and may modulate TGF-β signaling during Ventricular zone expressed pleckstrin homology domain-con- development and cancer initiation or progression. taining 1 (VEPH1) is among genes on 3q24-26, a region amplified in several cancers. Although little is known of VEPH1 | TGF-β | SMAD2/3 | Melted | ALK5 mammalian VEPH1, its Drosophila ortholog, Melted, is involved in neural and eye development, metabolism, and size deter- entricular zone expressed pleckstrin homology domain-con- mination through effects on Forkhead box O, target of Vtaining 1 (Veph1) was initially identified as a novel gene rapamycin, and Hippo signaling. We show that VEPH1 expression encoding an 833-amino-acid protein expressed in the developing affects similar gene categories as Melted and potently inhibits β β murine central nervous system (1). In the adult mouse, Veph1 transforming growth factor- (TGF- ) signaling. VEPH1 interacts with TGF-β type I receptor (TβRI) and inhibits dissociation of acti- expression is restricted to the eye and kidney. Although the β function of Veph1 in mammals is unknown, the Drosophila vated Sma- and Mad-related protein 2 from T RI, resulting in impaired TGF-β signaling. TGF-β acts initially as a tumor suppres- ortholog melted impacts tissue growth and metabolism (2). Dis- sor through its cytostatic activity, but subsequently promotes tu- ruption of melted in Drosophila results in a 10% reduction in adult melted mor progression. These findings suggest that VEPH1 could affect size, which is rescued by transgenic expression of or human TGF-β activity during cancer development/progression. VEPH1. Mutant flies have reduced fat (triglyceride) content at- tributable to decreased Melted, specifically within the fat body. Author contributions: P.S., A.K., M.J.R., J.L.W., and T.J.B. designed research; P.S. and A.K. Comparison of profiles of fat bodies from wild- performed research; P.S., J.L.W., and T.J.B. contributed new reagents/analytic tools; P.S., type (WT) and Melted-deficient flies identified altered expression A.K., C.V., and T.J.B. analyzed data; and P.S., A.K., M.J.R., J.L.W., and T.J.B. wrote the paper. of genes involved in metabolism, protein degradation, and im- mune response. Melted inhibits Forkhead box O (FOXO) and The authors declare no conflict of interest. promotes target of rapamycin (TOR) signaling—pathways in- This article is a PNAS Direct Submission. volved in regulating metabolism and tissue growth—by localizing Data deposition: The microarray data reported in this paper have been deposited in the Gene Expression Omnibus (GEO) database, www.ncbi.nlm.nih.gov/geo (accession no. FOXO and Tsc1/Tsc2 complexes to the plasma . GSE67765). Melted has also been shown to regulate R8 color photoreceptor 1To whom correspondence should be addressed. Email: [email protected]. warts differentiation by opposing the expression of , a tumor sup- This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. pressor and key component of the Hippo signaling pathway (2–5). 1073/pnas.1504671112/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1504671112 PNAS Early Edition | 1of10 Downloaded by guest on September 28, 2021 A 30 Dataset: B C ^ Cancer Cell, 2010 Amplificaon ‡ Cancer Research, 2012 25 Deleon ≠ Cell, 2012 3.5

& Cell, 2013 c HOC7 OVCAR3 SKOV3 OVCA429 ES2 HEY HepG2 COS7 3.0 20 ø JCO, 2013 VEPH1 ¶ Nature, 2011 * Nature, 2012 2.5 Acn 15 # Nature, 2013 2.0 % Hepatology, 2014 10 = Nature, 2014 1.5 b ¥ Nature Genecs, 2010 b D EGFP-VEPH1 DAPI

† Provisional / TBP mRNA levels 1.0 5 § In revision 0.5 Alteraon frequency (%) frequency Alteraon a a a 0 VEPH1 0.0

ES2 HEY HOC7 † ACC SKOV3 * CCLE † DLBC † GBM † Liver % Liver OVCAR3 & GBM † ccRCC # ccRCC = Breast‡ NCI-60 † Breast * Breast OVCA429 † Uterine # Uterine † Glioma † Ovarian† Cervical * Prostate¶ Ovarian † Bladder= Bladder† Prostate ø Bladder ^ Prostate † Lung* Lung squ squ = Stomach† Stomach ¥ Sarcoma & Prostate † Sarcoma† Pancreas Melanoma † Esophagus † Uterine CS † † Lung adeno = Lung adeno ≠ Lung adeno † Head§ Head & neck & neck

Fig. 1. VEPH1 is differentially expressed in ovarian cancer. (A) Histogram showing frequency of VEPH1 gene amplification or deletion in large-scale DNA copy-number datasets accessed through the cBioPortal for Cancer Genomics. Individual datasets are identified by the tissue. ACC, adrenocortical carcinoma; adeno, adenocarcinoma; CCLE, Cancer Cell Line Encyclopedia; ccRCC, kidney renal clear cell carcinoma; CS, carcinosarcoma; DLBC, lymphoid neoplasm diffuse large B-cell lymphoma; GBM, glioblastoma; NCI-60, NCI-60 cell lines; Squ, squamous. (B) RT-qPCR analysis of VEPH1 mRNA expression in six ovarian cancer cell lines, normalized to TATA-binding protein (TBP) mRNA expression. Bars represent mean ± SEM, n = 3. Bars with different letters are statistically different from one another as determined by ANOVA followed by the Student–Neuman–Keuls (SNK) post hoc test (P < 0.05). (C) Western blot of VEPH1 and actin levels in total lysates from ovarian cancer, HepG2, and COS7 cell lines. (D) GFP immunofluorescence staining showing localization of VEPH1 (green) in HepG2 cells transfected with pEGFP-VEPH1. Chromatin is visualized by DAPI staining (blue). (Magnification, 40×; scale bar, 9 μm.)

to modulate gene transcription in association with additional pression in subpopulations of ovarian cancer tumors (8, 9). VEPH1 transcriptional coregulators (10, 11). protein was predominantly localized to the cell membrane, as in- Dysregulated TGF-β signaling is implicated in multiple pathol- dicated by immunofluorescent imaging of enhanced GFP (EGFP)- ogies and plays a dual role in epithelial carcinogenesis (12, 13). tagged VEPH1 in HepG2 cells (Fig. 1D), consistent with the Initially, it acts as a tumor suppressor by inhibiting cell pro- subcellular localization reported for Melted in wing-disk cells (2). liferation but subsequently promotes cancer progression through induction of epithelial-to-mesenchymal transition, migration, in- VEPH1 Expression Affects Multiple Signaling Pathways. To explore vasion, metastasis, and immunosuppression (13, 14). Mutations in the role of VEPH1 in ovarian cancer, we established SKOV3 cell β TGF- receptors or SMADs have been identified in epithelial lines harboring CdCl2-inducible Flag-tagged VEPH1 expression cancers, indicating that dysregulation of TGF-β signaling is an (SKOV3-Ve1) and mock-transfected (SKOV3-M) cells. Western – important oncogenic event (12, 15 17). However, mutations in blot analysis using an anti-VEPH1 antibody indicated that CdCl2- these signaling mediators are less prevalent in ovarian cancer, induced expression of VEPH1 in SKOV3-Ve1 cells is comparable indicating that modulators of the TGF-β signaling pathway may to endogenous VEPH1 levels in OVCA429, ES2, and HEY cells instead be altered to result in this dysregulation. (Fig. 2A). Comparison of whole-genome gene expression profiles In this study, we demonstrate that VEPH1 suppresses TGF-β of SKOV3-Ve1 and -M cells indicated differential levels of 3,941 signaling by impeding the nuclear accumulation of activated of 38,666 expressed probe sets (10.2%), representing 3,630 genes. SMAD2. Our data indicate that this effect is mediated by VEPH1 A similar widespread alteration of gene expression (6% of repre- interaction with TβRI, which suppresses dissociation of phosphor- sented genes) was reported in melted-deficient Drosophila fat body ylated SMAD2 from the TGF-β receptor complex. These findings cells (2). Of 238 Drosophila genes with known human homologs highlight an additional pathway that may be affected by Melted and identified as affected by Melted, expression of 45 (19%) were suggestthatmodulationofTGF-β signaling by VEPH1 may play significantly altered by VEPH1 expression in SKOV3-Ve1 cells. a role in the initiation or progression of a subset of ovarian cancers. To identify cellular processes affected by VEPH1 expression, we used both the DAVID functional annotation tool (david. Results abcc.ncifcrf.gov/summary.jsp) and an independent gene set en- VEPH1 Is Differentially Expressed in Ovarian Cancer. Amplification richment analysis (GSEA) (30, 31) on the list of genes differ- of the VEPH1 locus has been reported in ∼40% of epithelial entially expressed between SKOV3-Ve1 and -M cells. Kyoto ovarian cancers (9). To determine whether this observation ex- Encyclopedia of Genes and Genomes (KEGG) pathways iden- tends to additional ovarian cancer datasets and other cancers, we tified by DAVID included metabolism, cytochrome P450-medi- interrogated large-scale copy number analysis datasets using the ated metabolism, and immune function (Table S1)—consistent cBioPortal for Cancer Genomics (www.cbioportal.org). Putative with what has been reported for Melted in the Drosophila fat amplification of the VEPH1 locus is present in 100 of 579 body (2). GSEA also indicated an impact of VEPH1 on gene sets (17.3%) ovarian serous cystadenocarcinomas (The Cancer Ge- associated with metabolism and immune response, and addi- nome Atlas; Provisional). Amplification of VEPH1 is also pre- tionally identified an impact on mRNA stability, protein syn- sent in other cancers, most notably in cervical, lung squamous thesis, cell adhesion and communication, and phototransduction cell, esophageal, and head and neck squamous cell (18–29) (Fig. (Fig. 2B and Fig. S1). These findings indicate that VEPH1 can 1A). We then examined the expression of VEPH1 in a panel of influence multiple cell processes and signaling pathways in six human epithelial ovarian cancer cell lines. High levels of mammalian cells, consistent with a conservation of function be- VEPH1 transcripts were detected in OVCA429, ES2, and HEY tween human VEPH1 and Drosophila Melted. cells, whereas little or no expression was detected in SKOV3, Previous studies indicate that Melted affects FOXO and OVCAR3, and HOC7 cells (Fig. 1B). Consistent with these data, Hippo signaling pathways in Drosophila. Moreover, putative Western blot analysis revealed VEPH1 protein expression only FOXO and 14-3-3 protein-binding motifs are conserved between in OVCA429, ES2, and HEY cells (Fig. 1C). Expression was not Melted and human VEPH1 [Eukaryotic Linear Motif (ELM) detected in HepG2 human epithelial hepatocarcinoma cells and (32)]. In the present study, DAVID analysis identified the Wnt COS7 African green monkey kidney fibroblasts (Fig. 1C). This signaling pathway as possibly affected by VEPH1 (P = 0.069). variable expression of VEPH1 in ovarian cancer cell lines is VEPH1 altered the expression of multiple genes encoding Wnt consistent with previous reports of VEPH1 amplification and ex- signaling components (Table S2). TGF-β has been implicated in

2of10 | www.pnas.org/cgi/doi/10.1073/pnas.1504671112 Shathasivam et al. Downloaded by guest on September 28, 2021 PNAS PLUS ACVEPH1, up-regulated genes

DNAJC1 ZBED2SCARA3 RBMS2 HAND1 TAGLN PTPRM LHX1 PRDM8 STOX2 2 FRMD6 REPS2 TNFRSF19 STARD13

HOC7 CaOV3 SKOV3 OVCAR3 ES2 HEY OVCA429 (-) SKOV3-M (+) SKOV3-M (-) SKOV3-Ve1 (+) SKOV3-Ve1 PARD6A NPTX1 CD24 C21orf7 ISL1 VEPH1 NT5E TMEM47 ASCL2 LAMC2BTG3 PDLIM3 SERPINE2 MGAT4A IL7R Tubulin HIST1H2BD ST6GAL1 MPZL2 ARHGEF6CHN1 2.0 RNF128KRT34 ELOVL4 DUSP6 RFTN1 NFE2L3 MAP7 DOCK11C14orf37 PHGDH CPVL FOXO COL4A5 KCNK1 YAP KITLG 1.5 MT1F CPE SLC2A12 SGK IRX4 C10orf47 TUBA4A STS-1 CITED2 MARCKS 1.0 COL17A1 GABRG2 NEO1 BCL11B IGFBP3 PNMA2 NPTX2 AKT3 1 3 VCAN C5orf13 RASSF2 1 KHDRBS3 EMP1 CREB5F11R TIAM1

VEPH1/Tubulin LC2A 0.5 S PMEPA1 CXADR COL1A1 HS6ST2 CAMK2NTACSTD H2AFY2 F3 CTGF JUP SOX4 CST6 CCND1 SLC39A8 0.0 PRKD1 ADAMTSL1WNT3A ES2 BAMBI EVI1 C20orf100 HEY PPAP2B ANGPT1 SKOV3- GPC4SEPT3 OVCA429 Ve1 (+) GRIN1 TGFBIRGS2 ADAM19 WNT AKAP12 B CYFIP2NR2E1 carbohydrate BMP7 F2R LDOC1 SPRY2 metabolism PTPRR TGF-β SRPX TSPAN13 BTBD11 GAD1 SLC7A7 DKK1 SYT11 LEF1 CYP24A1 ALOX5A MYO10 DDX43 RAC2 DPYSL3 Metabolism ARHGEF16 CTSH OSBPL6 cellular TGFBR3 MN1 LAMB3 NEDD9 CLDN1 FOXD1 respiraon P LHX6 QPCT BMP6 AOX1 SNAI2 IL11 CSRP2 COL7A1 D VEPH1, down-regulated genes RGS20 MYLK cholesterol AKR1C4 SATB2 NOS3 POPDC3 biosynthesis PRKAG2 PBK B4GALT6 ALDH1A1 MYBL GNE PSRC1 PIM2 CELL BIOLOGY TRIM6 GDF15TYSND1 CAV2 Translaon and 1 MID2 LOC645638PCDH7 DEXI protein synthesis PRSS23 OLFM1 ASPH ELOVL6 AMY2A FBXO4 PLD1 DDR2 FOXO GRIK2 BCAP29 GNG11 EGFLAM YAP SNN ZAK TNFRSF21 Lipid and CXCL2 SDPR PHF15 bile and sterol BDKRB1 SEMA4B carboxylic metabolism FLJ20273 5 DLX1 SLC26A6 UNC84B C8orf5 NXN acid transport 0 EFEMP1 CXXC5 SLC38A2 CCL2 CRYAB LAMA4 ABCC2 XPO7ARL4A ST3GAL3TPM2 AKR1C3 NRCAM INSIG1 lipid CXCL1 AKR1C2 CD36 SLC25A10 biosynthesis CA12 IFIT1 BMP5 C1QTNF1 TMEM166 CRABP2 LPHN2 LRRC20 DKFZp761P0423 COL1A2 HEY1 TACC2 AKR1B1 WNT SDC1 TRIB3 DDIT4 RCOR2 TNFRSF1A GTF2I PLA2G3 GSTK1 CDC42EP4 HMOX1 TGF-β NINJ2 BZW2 APOC1 MGST1 CAV1 YBX2 NIPSNAP1 CPA4 EZH2 SIX5 NISCH NFKBIA CDK5RAP2 GALT 3 VEGFA PTMS SYNM SLC2A6 YWHAG PDGFC INHBE RAB1 CEBPB CBS HMGCR ARL6IP5 APOE S100A4 Cell adhesion and communicaon, immune response, CCNG1 phototransducon, nitric oxide synthesis and acvity IMPA2 CEBPD

Fig. 2. Genome-wide gene-expression profiling indicates that VEPH1 expression affects multiple cell-signaling pathways and processes. (A) Western blot using an anti-VEPH1 antibody showing VEPH1 expression in stably transfected SKOV3-Ve1 cells and ovarian cancer cell lines. Histogram shows VEPH1 levels

normalized to tubulin. + or − indicates presence or absence of CdCl2 pretreatment, respectively. (B) Cytoscape visualization of cellular processes altered by VEPH1 expression as determined by GSEA of genes differentially expressed between SKOV3-M and -Ve1 cells. Three major cluster groups were identified along with several noncluster processes (Fig. S1). Each node represents a set of genes involved in the identified cell process, with the size of the node reflecting the number of genes affected by VEPH1. Red represents enrichment in SKOV3-M cells, and blue represents enrichment in SKOV3-Ve1 cells. Edges indicate sharing of affected gene members between individual nodes, with the thickness of the edge reflecting the number of genes shared. (C and D) Cytoscape visualization of FOXO, Hippo (YAP), Wnt, and TGF-β–regulated target genes altered by VEPH1 expression greater than or equal to twofold. (C) VEPH1, up-regulated genes. (D) VEPH1, down-regulated genes.

regulation of several of the VEPH1-affected KEGG pathways, by VEPH1 expression. Multiple genes are downstream targets of including focal adhesion, neurotrophin signaling, and amyo- two or more of the highlighted signaling pathways. This analysis trophic lateral sclerosis (33–36). A high-throughput screen by demonstrates that VEPH1 likely affects FOXO and Hippo sig- Barrios-Rodiles et al. (37) identified the PH-domain–containing naling as reported for Melted and additionally indicates that Wnt region of murine Veph1 as a potential interactor of TβRI. We and TGF-β signaling may be modulated by VEPH1. therefore determined whether downstream targets of these four signaling pathways are altered in SKOV3-Ve1 cells relative to VEPH1 Has a Global Negative Impact on TGF-β Signaling. To further SKOV3-M cells. Comparison of probe sets altered ≥1.5-fold due determine whether VEPH1 modulates TGF-β signaling, we to VEPH1 expression to compiled lists of known or putative compared gene-expression profiles of SKOV3-Ve1 and -M cells target genes revealed that expression of 173 FOXO, 274 Hippo treated with or without TGF-β. Treatment of SKOV3-M cells [Yes-associated protein (YAP)], 170 Wnt, and 133 TGF-β target with TGF-β for 3 h resulted in altered expression of 454 probe genes was affected by VEPH1 (Tables S3 and S4). Fig. 2 C and D sets. In contrast, only 283 probe sets were modulated by TGF-β in illustrate those target genes up- and down-regulated, respectively, SKOV3-Ve1 cells. Approximately 20% of probe sets (92 of 454)

Shathasivam et al. PNAS Early Edition | 3of10 Downloaded by guest on September 28, 2021 modulated by TGF-β in SKOV3-M cells were also altered by ABSKOV3-M SKOV3-Ve1 SKOV3-M SKOV3-Ve1 F 7 TGF-β in SKOV3-Ve1 cells (Fig. 3A). Applying a TGF-β–induced 1.5-fold change or greater cutoff to these gene lists indicates that 6 c β TGF- altered 65 probe sets in SKOV3-M cells, whereas only 9 362 92 191 57 8 1 5 were affected by TGF-β in SKOV3-Ve1 cells (Fig. 3B and Tables S5 and S6). Of these nine, only one was modulated by TGF-β in 4 SKOV3-Ve1 cells that was not affected in SKOV3-M cells. Fur- 3 b β / TBP mRNA levels thermore, the impact of TGF- is diminished in the presence C 1.2 - TGF-β 1.2 - TGF-β d d + TGF-β + TGF-β 2 of VEPH1 in seven of the eight probe sets common to both cell b b a 1.0 1.0 SMAD7 a a a lines. Overall, these findings indicate a global inhibition of TGF- c 1 β–induced gene responses by VEPH1 expression. 0.8 0.8 β– 0.6 0.6 0 To verify the impact of VEPH1 on TGF- dependent target Time a er 0123 0123 d a a

0.4 / TBP mRNA levels 0.4 TGF-β (h): siLuc siVEPH1 (#2)

gene expression observed in the gene expression profile study, we levels / TBP mRNA a selected three well-established TGF-β target genes that exhibited 0.2 c 0.2 β PAI-1 diminished TGF- stimulation in the presence of VEPH1 rela- 0.0 SMAD7 0.0 tive to control transfected cells: plasminogen activator inhibitor 1 SKOV3-M SKOV3-Ve1 SKOV3-M SKOV3-Ve1 (PAI-1), SMAD7, and inhibitor of differentiation 1 (ID1). Whereas TGF-β treatment in SKOV3-M cells resulted in a 4.6- and 2.4-fold D SKOV3-M SKOV3-Ve1 G PAI-1 SMAD7 Time (h): 0 1.5 3 4 0 1.5 3 4 - TGF-β up-regulation of and transcripts, respectively, this up- ID1 1.8 + TGF-β e regulation was reduced to 2.8- and 1.9-fold, respectively, in SKOV3- 1.6 d Tubulin Ve1 cells (Fig. 3C).TheimpactofVEPH1onTGF-β–induced ID1 fold- 1.4 expression was examined by Western blot analysis. The highest change: 1.0 18.1 14 10.7 1.4 8.2 5.8 4.1 β 1.2 levels of ID1 protein were observed 1.5 h after TGF- treatment in siVEPH1 siVEPH1 b E mRNA levels 1.0 both SKOV3-Ve1 and -M cells (Fig. 3D). However, VEPH1 ex- siLuc (#1) (#2) Time (h): 24 48 72 96 24 48 72 96 24 48 72 96 β– / TBP 0.8 pression decreased TGF- induced ID1 protein levels at all time VEPH1 0.6 c points examined (SKOV3-Ve1 vs. SKOV3-M). Tubulin c VEPH1 inhibition of TGF-β–induced SMAD7 expression was SMAD7 0.4 a siVEPH1 siVEPH1 siVEPH1 further demonstrated by knockdown studies in HEY cells, which (#3) (#4) (pool) siLuc 0.2 express high levels of endogenous VEPH1 (Fig. 1 B and C). Four Time (h): 24 48 72 96 24 48 72 96 24 48 72 96 24 0.0 VEPH1 siLuc siVEPH1 siVEPH1 individual VEPH1-targeting siRNAs were screened, with siVEPH1 (#1) (#2) nos. 1 and 2 providing the highest levels of knockdown (Fig. 3E). Tubulin β SMAD7 TGF- increased mRNA, with the highest transcript levels Fig. 3. VEPH1 negatively regulates activation of TGF-β–induced target observed 1 h after treatment (Fig. 3F). Silencing VEPH1 expression SMAD7 genes. (A) Venn diagram showing total number of probe sets significantly with siVEPH1 no. 1 or 2 increased transcripts relative to altered by TGF-β in SKOV3-Ve1 and -M cells. (B) Venn diagram showing TGF-β–treated controls (Fig. 3 F and G). Together, these findings probe sets significantly altered ≥1.5-fold by TGF-β in SKOV3-Ve1 and -M cells. demonstrate that VEPH1 negatively modulates TGF-β signaling. (C) VEPH1 inhibits TGF-β activation of PAI-1 and SMAD7 in SKOV3-Ve1 cells,

as determined by RT-qPCR. RNA was extracted from CdCl2-induced SKOV3- VEPH1 Inhibits SMAD2- and SMAD3-Dependent TGF-β Signaling. To Ve1 and -M cells treated with or without 25 pM TGF-β for 3 h. mRNA levels determine whether VEPH1 affects canonical SMAD2/3-dependent were normalized to TBP expression. (D) Western blot showing VEPH1 in- TGF-β signaling, we examined the impact of ectopic human hibition of TGF-β–induced ID1 protein expression in SKOV3-Ve1 cells. Total VEPH1 expression on TGF-β–dependent transcriptional responses protein was extracted from cells treated with 25 pM TGF-β for the indicated in SKOV3-M, -Ve, COS7, and HepG2 cells using SMAD2- and times. Tubulin expression is shown as a loading control. Fold changes shown SMAD3-dependent reporters [pARE- and pSBE4-luciferase (38, are relative to SKOV3-M cells in the absence of TGF-β.(E) VEPH1 knockdown efficiency was determined for a pool of four VEPH1 targeting siRNAs or each 39), respectively]. In the absence of CdCl2 treatment, SKOV3-Ve1 clones express detectable VEPH1 levels, whereas SKOV3-Ve2 individually in HEY cells. HEY cells were transfected with luciferase targeting clones express little or no VEPH1 (Fig. 4 A and B, Westerns). In siRNA (siLuc) as a control. Protein lysates were collected at 24, 48, 72, and SKOV3-M cells, TGF-β increased pARE- and pSBE4-luciferase 96 h after siRNA transfection and subjected to immunoblotting for VEPH1. β (F and G) VEPH1 knockdown enhances TGF-β–induced SMAD7 gene ex- 4.4- and 3.7-fold, respectively, whereas in SKOV3-Ve1 cells, TGF- pression in HEY cells, as determined by RT-qPCR. HEY cells were transfected induction of pARE-luciferase was blocked, and induction of A with siLuc or siVEPH1 (no. 1 or 2) 96 h (F)or48h(G) before treatment with pSBE4-luciferase was decreased by 50% (Fig. 4 ). A slight decrease or without 100 pM TGF-β. RNA was extracted after TGF-β treatment for the in pARE-luciferase activity in the absence of TGF-β was observed indicated times (F) or 1 h after treatment (G). SMAD7 mRNA levels were in SKOV3-Ve1 cells, consistent with an impact on autocrine TGF-β normalized to TBP expression. Bars represent mean ± SEM, n = 3. Bars with signaling in SKOV3 cells (Fig. S2). Induction of VEPH1 expression different letters are statistically different from one another as determined in SKOV3-Ve2 cells with increasing amounts of CdCl2 resulted in a by ANOVA followed by SNK post hoc test (P < 0.05). dose-dependent decrease in TGF-β–induced pARE- and pSBE4- luciferase activity (Fig. 4B). To determine whether this negative impact of VEPH1 extends to other cell types, we performed parallel expression on SMAD2/3 nuclear accumulation. Cytoplasmic and experiments in transiently transfected COS7 and HepG2 cells. nuclear protein fractions were isolated from SKOV3-Ve1 and -M TGF-β stimulationresultedin8.5-and145-foldactivationof cells treated with or without TGF-β. As expected, TGF-β treat- SMAD2-dependent signaling in COS7 and HepG2 cells, re- ment decreased cytoplasmic and increased nuclear SMAD2 levels spectively, and ectopic VEPH1 expression inhibited this activation in the absence of VEPH1 (Fig. 5A). VEPH1 expression reduced in a dose-dependent manner in both cell lines (Fig. 4 C and D). At TGF-β–induced nuclear SMAD2 accumulation at all time points the highest levels of VEPH1, SMAD2-dependent signaling was examined. This reduction was also evident by immunofluorescence decreased by >70%. A marked reduction of up to 50% in SMAD3- (Fig. S3). Knockdown of endogenous VEPH1 expression in ES2 dependent signaling by VEPH1 was also observed in HepG2 cells cells resulted in increased SMAD2 and SMAD3 nuclear accu- (Fig. 4D). Overall, these data demonstrate that VEPH1 attenuates mulation following TGF-β treatment (Fig. 5B). Time course both SMAD2- and SMAD3-dependent TGF-β signaling. analysis of TGF-β–induced phospho-SMAD2 levels indicated that VEPH1 does not impact C-terminal phosphorylation of SMAD2 VEPH1 Inhibits SMAD2 Nuclear Accumulation and Promotes SMAD2 (Fig. 5C and Fig. S4A). Thus, the decreased nuclear accumulation Retention at TβRI. To determine the mechanism by which VEPH1 of SMAD2 in the presence of VEPH1 is not a result of reduced inhibits TGF-β signaling, we first examined the impact of VEPH1 ability of TβRI to phosphorylate SMAD2.

4of10 | www.pnas.org/cgi/doi/10.1073/pnas.1504671112 Shathasivam et al. Downloaded by guest on September 28, 2021 PNAS PLUS A B 4.0 3.0 4.0 - TGF-β 6 b - TGF-β b + TGF-β + TGF-β 3.5 b 3.5 5 b 2.5 c 3.0 3.0 4 2.0 SKOV3-M SKOV3-Ve1 2.5 SKOV3-Ve2 2.5 c c CdCl2 (μM): 0 5.0 0 5.0 2.0 3 CdCl2 (μM): 02.55.0 d a, d 2.0 1.5 Flag-VEPH1 1.5 Flag-VEPH1 d 2 a a Tubulin Tubulin 1.5 1.0 a a a 1.0 1 a 0.5 c c 1.0 a a 0.5 Relave pARE-luciferase acvity 0.0 0 SKOV3-M SKOV3-Ve1 Relave pSBE4-luciferase acvity SKOV3-M SKOV3-Ve1 0.5 Relave pARE-luciferase acvity Relave pSBE4-luciferase acvity 0.0 0.0 CdCl2: 02.55.0 CdCl2: 02.55.0 C D (μM) (μM) 10 b - TGF-β + TGF-β 8 c 120 18 c b - TGF-β b - TGF-β + TGF-β 16 + TGF-β 6 100 14 4 80 12 Relave c d 10 d 2 a a 60 e a a c 8 Relave Relave pARE-luciferase acvity pARE-luciferase 0 40 d 6 Flag-VEPH1 (ng) 0 200 400 800 e 4 20

pARE-luciferase acvity pARE-luciferase 2 pSBE4-luciferase acvity pSBE4-luciferase a a a aa a a a Flag-VEPH1 0 0 Tubulin Flag-VEPH1 (ng) 0 300 500 700 Flag-VEPH1 (ng) 0 300 500 700

Fig. 4. VEPH1 inhibits SMAD2- and SMAD3-dependent TGF-β signaling. (A)CdCl2-induced SKOV3-Ve1 cells have diminished SMAD2-dependent (pARE- luciferase) and SMAD3-dependent (pSBE4-luciferase) TGF-β signaling relative to SKOV3-M cells. (B) VEPH1 expression decreases TGF-β–dependent activation of both SMAD2- and SMAD3-dependent reporters in a dose-dependent manner in SKOV3-Ve2 cells. Cells in A and B were treated for 24 h with or without 25 pM

TGF-β 24 h after reporter transfection. CdCl2 induction of VEPH1 expression in SKOV3-Ve1 (A, Left)andSKOV3-Ve2(B, Left) cells was confirmed by immuno- CELL BIOLOGY blotting with Flag antibody. (C and D) Increased expression of VEPH1 inhibits SMAD2- and SMAD3-dependent TGF-β signaling in COS7 or HepG2 cells. COS7 (C) and HepG2 (D) cells were transiently transfected with empty vector or increasing amounts of Flag-VEPH1 and either a SMAD2-dependent (C and D) or SMAD3- dependent (D) reporter. Transfected cells were treated with or without 100 pM TGF-β for 24 h. Immunoblotting with anti-Flag antibody in C shows VEPH1 protein levels in COS7 cells transfected with increasing amounts of Flag-VEPH1. For all reporter assays, firefly luciferase activity was normalized to either β-galactosidase or Renilla luciferase (RL) activity. For pARE-luciferase assays, cells were cotransfected with a FoxH1 expression construct. Bars represent mean ± SEM, n = 3. Bars with different letters are statistically different from one another as determined by ANOVA followed by SNK post hoc test (P < 0.05).

To identify how VEPH1 antagonizes SMAD2 nuclear accu- was also observed in the presence of VEPH1, consistent with our mulation, we analyzed the interaction of VEPH1 with TβRI and analysis of endogenous interactions in SKOV3- and HepG2-Ve1 the impact of VEPH1 on SMAD2–TβRI association. Endoge- cells (Fig. 6 B and C). The SMAD2 retained at TβRI in the pres- nous TβRI coprecipitated with both endogenous VEPH1 in ence of VEPH1 is phosphorylated at the C terminus as shown in HEY and Flag-tagged VEPH1 in SKOV3-Ve1 cells independent HEK293T cells (Fig. 6E). In the absence of VEPH1, association of of TGF-β stimulation (Fig. 6 A and B). VEPH1 impact on phospho-SMAD2 with TβRI could not be detected, consistent with SMAD2–TβRI association was examined in both SKOV3 and phosphorylation-induced release of SMAD2 from the receptor. The HepG2 cells stably expressing CdCl2-inducible Flag-VEPH1 combined data demonstrate that VEPH1 impedes release of acti- (SKOV3- and HepG2-Ve1) or mock transfected (SKOV3- and vated SMAD2 from TβRI. HepG2-M) cells. In the absence of VEPH1, an interaction be- tween TβRI and SMAD2 was detected in TGF-β–treated Modulation of TGF-β Signaling by VEPH1 Is Dependent upon Its HepG2-M cells (Fig. 6C), which is consistent with TGF-β– Interaction with TβRI. To further investigate the relationship be- induced transient recruitment of SMAD2 to TβRI. This in- tween VEPH1-induced retention of SMAD2 at TβRI with its teraction was not detected in SKOV3-M cells (Fig. 6B), likely inhibitory effects on TGF-β signaling, we generated Flag-tagged due to the lower level of TGF-β sensitivity of these cells. In the VEPH1 deletion constructs (Fig. 7A) and examined their in- presence of VEPH1, increased SMAD2 coprecipitation with teraction with TβRI. VEPH1 deletion constructs lacking amino TβRI was observed independent of TGF-β treatment in both acids 661–833 (Δ661) or 320–833 (Δ320) both interacted with SKOV3- and HepG2-Ve1 cells (Fig. 6 B and C), indicating that TβRI, as did a construct consisting of only the C-terminal PH- VEPH1 recruits and retains SMAD2 at TβRI independent of domain–containing region (662Δ) (Fig. 7B). These findings thus TβRI activation by TGF-β. indicate two TβRI-interacting regions (TIRs) on VEPH1: TIR-1, Semiquantitative luminescence-based immunoprecipitation assay located within the first 319 amino acids of the N-terminal region; (LUMIER) (37) was used to quantify the impact of VEPH1 on and TIR-2, located within the 171 C-terminal amino acids con- the transient SMAD2–TβRI interaction. HEK293T human em- taining the PH domain. Interestingly, expression of VEPH1 bryonic kidney cells were cotransfected with WT or constitutively fragments possessing TIR-1 inhibited TGF-β induction of pARE- active (TD) TβRI and Renilla luciferase (RL)-tagged SMAD2 luciferase similar to full-length VEPH1, whereas TIR-2 (662Δ) expression constructs, with or without VEPH1. Total protein enhanced TGF-β responsiveness (Fig. 7C). The enhancing effect extracts were immunoprecipitated with an anti-TβRI antibody, of TIR-2 was also observed in HEY cells, in which expression of and immunoprecipitates were assessed for RL activity. In the full-length VEPH1 only slightly inhibited TGF-β signaling (Fig. absence of VEPH1, an interaction between RL-SMAD2 and 7D), likely due to high levels of endogenous VEPH1. TβRI(TD) was observed (Fig. 6D), consistent with previous To delineate the sequences within TIR-1 responsible for the reports (37). As expected, little or no interaction was found interaction with TβRI and inhibition of TGF-β signaling, we gen- between RL-SMAD2 and TβRI(WT). A VEPH1-induced dose- erated deletion constructs consisting of amino acids 1–90 (Δ91), dependent increase in RL-SMAD2 interaction with TβRI(TD) 91–200 (90Δ–Δ201), or 201–319 (200Δ–Δ320) (Fig. 7A). The was observed (Fig. 6D). A RL-SMAD2 interaction with TβRI(WT) strongest interaction with TβRI was observed with 200Δ–Δ320,

Shathasivam et al. PNAS Early Edition | 5of10 Downloaded by guest on September 28, 2021 A Time (h): 0.5 1.0 4.0 B siLuc siVEPH1 Fracon: C N C N CNCNCNCN Time (h): 0 0.5 2.0 00.52.0 TGF-β: -++- -++- - - ++ Fracon: C N C N C N C N C N C N SMAD2 TGF-β: -++-+--+++++ EEA1 SKOV3-M SMAD2 LAMIN A/C SMAD3

SMAD2 EEA1 EEA1 SKOV3-Ve1 FIBRILLARIN LAMIN A/C siLuc 1.0 b 0.6 siLuc siVEPH1 0.9 siVEPH1 VEPH1 SKOV3-M 0.8 0.9 open = without TGF-β 0.5 Tubulin SKOV3-Ve1 0.8 filled = with TGF-β 0.7 0.4 0.7 0.6 0.6 0.3 0.5 c 0.5 0.2 0.4 0.4 Rao of nuclear Rao 0.3 SKOV3-M β -treated a 0.1

0.3 SMAD2 SMAD2/total Rao of nuclear Rao a

SMAD2/total SMAD2 SMAD2/total 0.2 0.2 0 0.1 to TGF- 00.5 2 Nuclear/total SMAD2 relave Nuclear/total 0.1 0.0 0.0 2.0 siLuc 0 0.5 1.0 4.0 SKOV3-M SKOV3-Ve1 siVEPH1 Time in TGF-β (h) 1.6 Time (h): 0.75 1.0 1.5 2.0 4.0 6.0 7.5 C TGF-β: - + -++ - -++ - -+ +- 1.2 pSMAD2 SKOV3-M 0.8 SMAD2 Rao of nuclear Rao

SMAD3/Fibrillarin 0.4 pSMAD2 SKOV3-Ve1 0 SMAD2 00.5 2 Time in TGF-β (h)

Fig. 5. VEPH1 impedes SMAD2 nuclear accumulation without altering SMAD2 phosphorylation. (A) Western blot analysis for SMAD2, EEA1 (cytoplasmic

marker), and LAMIN A/C (nuclear marker) in cytoplasmic (C) and nuclear (N) fractions extracted from CdCl2-induced SKOV3-Ve1 and -M cells treated with or without 25 pM TGF-β. Lysates were collected at 0.5, 1, and 4 h after TGF-β treatment. The line graph shows quantification of the immunoblot expressed as the ratio of nuclear to total (C+N) SMAD2 for TGF-β–treated cells. The histogram summarizes three independent experiments (mean ± SEM) examining SMAD2 subcellular localization after treatment with TGF-β for 1 h. Data are expressed as the ratio of nuclear to total SMAD2 relative to that measured in TGF- β–treated SKOV3-M cells (set to 1.0). Bars with different letters are statistically different from one another as determined by ANOVA followed by SNK post hoc test (P < 0.05). (B) VEPH1 knockdown enhances TGF-β–induced SMAD2 and SMAD3 nuclear accumulation in ES2 cells. A representative of three independent Western blots for SMAD2, SMAD3, EEA1, and Fibrillarin (nuclear marker) in C and N fractions extracted from siLuc- or siVEPH1-transfected ES2 cells treated with or without 100 pM TGF-β. ES2 cells were transfected 72 h before treatment with TGF-β, and lysates were collected after 0.5 and 2 h. The line graph shows quantification of the immunoblots expressed as the ratio of nuclear to total (C + N) SMAD2 or nuclear SMAD3 to Fibrillarin. The Western blot Inset in the line graph shows VEPH1 expression levels in siLuc- or siVEPH1-transfected ES2 cells. (C) A representative of three independent Western blots showing equivalent

TGF-β–induced C-terminal phosphorylated SMAD2 levels in SKOV3-M and -Ve1 cells. CdCl2-induced SKOV3-Ve1 and -M cells were treated with or without 25 pM TGF-β for the times indicated.

whereas a weak interaction was detected with Δ91 (Fig. 7E). An with TD TβRI and either full-length VEPH1, Δ661, or 662Δ,and interaction with TβRI was not detected with 90Δ–Δ201. Further- their impact on SMAD2-dependent pARE-luciferase activation more, both 200Δ–Δ320 and Δ91 inhibited TGF-β signaling, whereas was determined in the presence or absence of TGF-β.Asexpected, 90Δ–Δ201hadnoimpact(Fig.7F). Collectively, these findings TGF-β treatment resulted in increased luciferase activity due to demonstrate that the ability of VEPH1 fragments to modulate its activation of endogenous WT TβRI. Both full-length VEPH1 TGF-β signaling associates with their ability to interact with TβRI. and Δ661 inhibited, whereas 662Δ enhanced, pARE-luciferase activity in nontreated, as well as TGF-β–treated, cells relative to Increased SMAD2–TβRI Interaction by Truncated VEPH1 controls (Fig. 8B). These findings indicate that both the inhibitory Correlates with Their Inhibition of TGF-β Signaling. Our results in- effects of VEPH1 and the enhancing effects of TIR-2 are down- dicate that TIR-1 activity is essential for an inhibitory impact of stream of TβRI kinase activation and are consistent with VEPH1 VEPH1 on TGF-β signaling, whereas TIR-2 facilitates TGF-β acting to modulate SMAD2 dissociation from TβRI. signaling in the absence of TIR-1. We therefore compared the To further investigate TIR-2 activities, its impact on SMAD2 impact of TIR-1–containing Δ661 and 662Δ (TIR-2) on SMAD2 nuclear localization and phosphorylation were examined in cells association with TβRI using semiquantitative LUMIER. Full- stably expressing EGFP-tagged 662Δ or EGFP alone as a con- length VEPH1 stabilized SMAD2 interaction with both WT and trol. Stable expression of EGFP-662Δ was verified by Western TD TβRI (Fig. 8A), similar to our analysis above (Fig. 6D). Whereas blot analysis (Fig. 8C). Treatment of cells with TGF-β for 1 h Δ661 further enhanced these interactions, 662Δ destabilized increased nuclear SMAD2, which returned to near baseline by SMAD2 association with TβRI, compared with empty-vector- 2 h after treatment (Fig. 8D). Expression of 662Δ enhanced TGF- transfected cells (Fig. 8A). Thus, the ability of full-length and β–induced SMAD2 nuclear accumulation and prolonged its re- truncated VEPH1 proteins to modulate the stability of the SMAD2– tention. To determine whether these changes were associated with TβRI interaction correlates with their inhibitory or stimulatory altered C-terminal phosphorylation of SMAD2, we compared effect on TGF-β signaling. levels of phospho-SMAD2 in 662Δ or control cells. Time-course To further verify that VEPH1 inhibition of TGF-β signaling is analysis indicated elevated TGF-β–induced phospho-SMAD2 independent of TβRI kinase activity, COS7 cells were transfected levels at all time points in 662Δ-expressing cells compared with

6of10 | www.pnas.org/cgi/doi/10.1073/pnas.1504671112 Shathasivam et al. Downloaded by guest on September 28, 2021 PNAS PLUS A HEY C HepG2-M HepG2-Ve1 D 12 IP: IgG TβRI IP: IgG TβRI IgG TβRI TβRI TGF-β: --+ TGF-β: - -+ -+- + IB: VEPH1 SMAD2 RL IB: SMAD2 10 IB: VEPH1 Input IB: SMAD2

IB: TβRI Input 8 SMAD2 RL IB: TβRI TβRI

B SKOV3-M SKOV3-Ve1 IB: Flag Light (VEPH1) 6 IP: IgG TβRI IgG TβRI TGF-β: --++ -- IP: TβRI SMAD2SMAD2 RLRL HEK293 IB: SMAD2 E 4 TβRITβRI IP: IgG Flag TβRI Ab

IB: Flag acvity SMAD2 luciferase (VEPH1) SMAD2:++ + + + Y Flag-TβRI: +++++ Sepharose IB: SMAD2 Input 2 TβRII: ++-+- bead IB: Flag (VEPH1) VEPH1: ++--+ IB: pSMAD2 SKOV3-M SKOV3-Ve1 0 TGF-β: --++ IB: pSMAD2 Flag- 0 0 0.6 0.9 1.2 0 0.6 0.9 1.2 IB: TβRI VEPH1 (μg) TβRI(WT) TβRI(TD) IB: Tubulin IB: SMAD2 Input

IB: SMAD2 Input IB: TβRI IB: TβRI IB: VEPH1

Fig. 6. VEPH1 interacts with TβRI and promotes retention of SMAD2 at TβRI. (A) Coimmunoprecipitation of endogenous TβRI with VEPH1 in HEY cells treated with or without 200 pM TGF-β for 1 h. (B) Coimmunoprecipitation of TβRI with VEPH1 and with SMAD2 in SKOV3-Ve1 cells. Induced SKOV3-Ve1 and -M cells were treated with or without 200 pM TGF-β for 1 h. Lysates immunoprecipitated with nonimmune IgG were used as a control. TβRI levels shown were de- β

termined in the cell lines in a separate experiment. (C) Coimmunoprecipitation of T RI with SMAD2 in HepG2-Ve1 and -M cells. CdCl2-induced HepG2 stable CELL BIOLOGY cells were treated with or without 200 pM TGF-β for 0.5 h. Lysates immunoprecipitated with nonimmune IgG were used as a control. (D) Semiquantitative LUMIER assay showing increased SMAD2–TβRI interaction in the presence of VEPH1. HEK293T cells were transiently transfected with expression constructs for WT or TD TβRI and RL-tagged SMAD2 in the presence or absence of increasing amounts of VEPH1. Cell lysates were immunoprecipitated with an anti-TβRI antibody, and the resulting precipitates were assessed for RL activity. The dashed line represents the minimum value reflecting a significant interaction (37). RL-SMAD2 and TβRI levels in total lysates before immunoprecipitation (input) are shown by Western blot. (E) VEPH1 promotes TβRI retention of phos- phorylated SMAD2. HEK293T cells were transiently transfected with expression constructs for WT TβRI and SMAD2, in the presence or absence of TβRII and VEPH1. Lysates immunoprecipitated with nonimmune IgG were used as a control. IB, immunoblotting antibody; input, nonimmunoprecipitated lysates; IP, immunoprecipitating antibody.

control cells (Fig. 8E). Maximal phospho-SMAD2 was detected at ciation with SMAD2 in the absence of TGFβ, we did not detect 1 h after treatment with TGF-β and began to decrease by 2 h. phospho-SMAD2 in the absence of TGFβ treatment. Further Thus, 662Δ functions to enhance C-terminal SMAD2 phosphor- studies are required to determine whether other proteins might ylation, leading to its increased nuclear accumulation. be involved with the inhibitory role of VEPH1. Knockdown of endogenous VEPH1 resulted in increased Discussion TGF-β–induced nuclear accumulation of SMAD2 and SMAD3. Studies with melted mutant flies indicate important develop- In contrast to SMAD2, the increase in SMAD3 may reflect in- mental effects, mediated at least in part by modulation of the creased SMAD3 stability induced by TGF-β (Fig. S4B). The FOXO and Hippo signaling pathways. Our gene-expression data Axin/GSK3-β complex promotes proteasome-dependent degra- generated with SKOV3 ovarian cancer cells indicate human dation of SMAD3, but not SMAD2, in the absence of TGF-β VEPH1 also impacts these signaling systems and additionally the signaling (42). Although VEPH1 could impair SMAD3 nuclear Wnt and TGF-β signaling pathways. We demonstrate a unique accumulation through a mechanism similar to SMAD2, it is also inhibitory role for VEPH1 on SMAD2/3-dependent TGF-β sig- possible that VEPH1 could affect the Axin/GSK-3β complex to naling. Our findings indicate that VEPH1 is a membrane-local- enhance SMAD3 degradation. ized TβRI-interacting protein that retains SMAD2 at TβRI, The VEPH1 N-terminal and PH-domain–containing regions without affecting SMAD2 C-terminal phosphorylation. This that interact with TβRI have opposing impacts on TGF-β sig- hindered release correlates with diminished SMAD2 nuclear naling that is reflected in their ability to sequester SMAD2 at accumulation and negative modulation of SMAD2/3-dependent TβRI. Whereas the N-terminal region (TIR-1) inhibits signaling TGF-β target genes. similar to the full-length protein and promotes SMAD2 retention The binding, phosphorylation, and release of SMAD2/3 from at TβRI, the isolated PH-domain–containing region (TIR-2) en- TβRI is highly coordinated, involving multiple partnering pro- hances signaling and promotes dissociation of SMAD2 from teins, such as FKBP12 (FK506-binding protein 1A, 12 kDa) and TβRI, resulting in elevated nuclear SMAD2 accumulation. These SARA (SMAD anchor for receptor activation) (reviewed in ref. findings indicate that the TIR-1 region of VEPH1 is crucial for the 40). Inactive TβRI is bound to inhibitory protein FKBP12. In inhibitory effect of the full-length protein. Our data further in- response to TGF-β binding, TβRII forms a stable complex with, dicate that there are two subregions within TIR-1 that interact and transphosphorylates, TβRI. The resulting conformational with TβRI and independently inhibit TGF-β signaling; however, a change dissociates FKBP12, enhancing TβRI binding to SMAD2/3 more pronounced inhibition is obtained with an intact TIR-1. (41). SARA interacts with and controls SMAD2/3 subcellular lo- The enhancing effect of TIR-2 was also observed in HEY cells, calization and recruitment to TβRI (40). VEPH1 expression re- suggesting that this truncated protein opposes endogenous VEPH1 sulted in SMAD2 interaction with TβRI, even in the absence of activity. Of interest, murine Veph1 isoform B (Veph-B; GenBank TGF-β stimulation. Although this result raises the possibility that accession no. BAC02922) is predicted to encode a truncated 253- VEPH1 could promote activation of TβRI to result in its asso- amino-acid protein corresponding to the PH-domain–containing

Shathasivam et al. PNAS Early Edition | 7of10 Downloaded by guest on September 28, 2021 Amino acid posion A D 20 100 200 300 400 500 600 700 800 18 f

PH WT 16 Flag - TGF-β + TGF-β ∆661 14 Flag 12 ∆320 Flag 10 ∆91

Flag 8 b 90∆-∆201 6 d Flag

200∆-∆320 acvity pARE-luciferase Relave 4 Flag e 2 PH 662∆ a c Flag 0 B IP: IgG TβRI Empty WT 662∆ VEPH1 mutant: WT - WT ∆661 ∆320 662∆ E TGF-β: -+ +- -+ +- -+ -+ (99.3 kDa) 90∆- 200∆- Flag-VEPH1 VEPH1 mutant: ∆320 ∆91 ∆201 ∆320 Flag-∆661 IP (IgG or TβRI Ab): IgG Ab IgG Ab IgG Ab IgG Ab (79.0 kDa) Flag-∆320 IB: Flag (39.40 kDa) (40.6 kDa) Flag-∆320 IB: Flag Flag-662∆ (16.95 kDa) (18.2 kDa) Flag-200∆-∆320 Flag-∆91 (13.47 kDa) ∆320

Input 90∆- ∆91 ∆320 ∆201 200∆- IB: Flag Flag-∆320

Input IB: Flag Flag-90∆-∆201 C 20 Flag-200∆-∆320 emptyWT ∆661 ∆320 662∆ e Flag-∆91 18 F 14 - TGF-β IB: Flag + TGF-β b 16 12 b

14 Tubulin 10 d empty∆320 ∆91 90∆-∆201200∆-∆320 12 c,d β 8 10 b - TGF- + TGF-β IB: Flag 8 6 c

6 4 Tubulin Relave pARE-luciferase acvity pARE-luciferase Relave 4 d d acvity pARE-luciferase Relave d d 2 a a a 2 a a a c c c 0 0 Empty ∆320 ∆91 90∆- 200∆- Empty WT ∆661 ∆320 662∆ ∆201 ∆320

Fig. 7. VEPH1 inhibition of TGF-β signaling is dependent on its interaction with TβRI. (A) Schematic showing Flag-tagged VEPH1 fragments generated by site- directed mutagenesis and cloning. WT, full-length VEPH1; Δ661, p.M661X; Δ320, p.E320X; 662Δ, p.M1_M662del., amino acids 1–90 (Δ91); 91–200 (90Δ–Δ201); and 201–319 (200Δ–Δ320). PH, pleckstrin homology domain. (B) VEPH1 N- and C-terminal regions interact with TβRI independent of TGF-β. Lysates were collected after 1-h treatment with TGF-β and immunoprecipitated with anti-TβRI antibody or control IgG. (C) N- and C-terminal regions of VEPH1 inhibit and enhance TGF-β signaling, respectively. COS7 cells transiently transfected with empty expression vector or truncated Flag-VEPH1 expression constructs and pARE-luciferase were treated with or without 100 pM TGF-β for 24 h. Inset shows levels of Flag immunoreactivity indicative of full-length or truncated VEPH1 expression. (D) The C-terminal PH domain-containing region of VEPH1 (662Δ) enhances TGF-β signaling in HEY cells. Cells were transiently transfected with VEPH1 expression constructs or empty vector and pARE-luciferase reporter and treated with or without 100 pM TGF-β for 24 h. (E) Determination of TβRI- binding regions within the VEPH1 N-terminal region. Lysates from COS7 cells transfected with VEPH1 expression constructs were immunoprecipitated with an anti-TβRI antibody or control IgG. (F) Inhibition of TGF-β signaling by VEPH1 N-terminal fragments shown to interact with TβRI. COS7 cells transfected with expression constructs for N-terminal region VEPH1 and pARE-luciferase were treated with or without 100 pM TGF-β for 24 h. Right shows levels of Flag immunoreactivity indicative of truncated VEPH1 expression. For all reporter assays, firefly luciferase activity was normalized to RL activity. For pARE-luciferase assays, cells were cotransfected with a FoxH1 expression construct. Bars represent mean ± SEM, n = 3. Bars with different letters are statistically different from one another as determined by ANOVA followed by SNK post hoc test (P < 0.05). IB, immunoblotting antibody; input, nonimmunoprecipitated lysates; IP, immunoprecipitating antibody.

region. Thus, expression of the isolated TIR-2 region of Veph1, the VEPH1 N-terminal and PH domain regions are well conserved, as well as full-length Veph1 (Veph-A; GenBank accession no. and analysis using the database of ELM (32) BAC02920), may have opposing effects on TGF-β signaling. Both predicts multiple protein interaction motifs within these regions.

8of10 | www.pnas.org/cgi/doi/10.1073/pnas.1504671112 Shathasivam et al. Downloaded by guest on September 28, 2021 PNAS PLUS A B D EGFP EGFP-662∆ 35 f Time (h): 0 1.0 2.0 01.02.0 β 18 - TGF- + TGF-β Fracon: CN CNCNCNCNCN 30 TGF-β: -- ++++--++++ 16 f SMAD2 14 25 EEA1 12 e FIBRILLARIN 20 10 0.7 d d 0.6 EGFP 8 15 EGFP-662∆ IP: TβRI 0.5 c 6 c c 0.4 10

SMAD2 luciferase acvity SMAD2 luciferase 4 b b b 0.3 b acvity pARE-luciferase Relave b g 2 e 0.2 a 5 d d of nuclear Rao

SMAD2/total SMAD2 SMAD2/total 0.1 0 a RL-SMAD2:-+ + + + ++++ 0 0 012 - VEPH1: - - - - VEPH1: WT β ∆661 662∆ Time in TGF- (h) WT WT ∆661 ∆661 662∆ 662∆ TβRI: - TβRI(TD) TβRI: - TβRI(WT) TβRI(TD) E EGFP EGFP-662∆ RL-SMAD2 IB: SMAD2 Time (h): 0 0.25 1.0 2.0 00.251.0 2.0 C TGF-β: --++++++ β TβRI IB: T RI pSMAD2 EGFP EGFP-662∆ Flag-VEPH1 SMAD2

Input EGFP-662∆ Flag-∆661 1.2 EGFP EGFP-662∆ IB: Flag IB: GFP 1.0 EGFP 0.8 0.6 Flag-662∆ Tubulin 0.4 CELL BIOLOGY total SMAD2 total 0.2 Rao of pSMAD2/ Rao 0.0 00.25 1 2 Time in TGF-β (h)

Fig. 8. Increased SMAD2–TβRI interaction by truncated VEPH1 proteins correlates with their negative impact on TGF-β signaling. (A) Semiquantitative LUMIER assay showing enhanced SMAD2–TβRI interaction in the presence of TGF-β inhibiting VEPH1 truncated proteins. HEK293T cells were transfected with WT or TD TβRI and RL-tagged SMAD2 expression constructs in the presence or absence of WT, Δ661, or 662Δ VEPH1. Cell lysates were immunoprecipitated with an anti-TβRI antibody, and the resulting precipitates were assessed for RL activity. The dashed line represents the minimum value reflecting a significant interaction (37). Immunoblotting for RL-SMAD2 was performed on immunoprecipitates. TβRI and Flag-VEPH1 protein levels in total lysates before immu- noprecipitation (input) are shown by Western blot. IB, immunoblotting antibody; IP, immunoprecipitating antibody. (B) VEPH1 inhibits TGF-β signaling downstream of TβRI activation. COS7 cells transfected with TβRI(TD), pARE-luciferase, and expression constructs for full-length or truncated VEPH1 were treated with or without 100 pM TGF-β for 24 h. Bars represent the mean ± SEM, n = 3. Bars with different letters are statistically different from one another as determined by ANOVA followed by SNK post hoc test (P < 0.05). (C) GFP immunoblot showing EGFP or EGFP-662Δ expression in stably transfected COS7 cells. (D and E)662Δ enhances TGF-β–dependent SMAD2 nuclear accumulation and C-terminal phosphorylation. (D) Western blot analysis for SMAD2, EEA1 (cy- toplasmic marker), and Fibrillarin (nuclear marker) in cytoplasmic (C) and nuclear (N) fractions extracted from EGFP-662Δ or EGFP stable COS7 cells treated with or without 100 pM TGF-β. Lysates were collected at 1 and 2 h after TGF-β treatment. The line graph shows quantification of the immunoblot expressed as the ratio of nuclear to total (C+N) SMAD2. (E) Western blot of C-terminal phosphorylated and total SMAD2 levels in stable EGFP or EGFP-662Δ COS7 cells treated with or without 100 pM TGF-β for 0.25, 1, or 2 h. The line graph shows quantification of the immunoblot expressed as the ratio of phosphorylated to total SMAD2.

Future studies will define whether these motifs are critical for the tations or reduced SMAD4 expression (45–48), other studies observed divergent effects of TIR-1 and -2 on TGF-β signaling. have reported that alterations in TGF-β signaling are independent Teleman et al. (2) reported that deletion of melted in Dro- of TGF-β receptors or downstream SMADs. In recent studies, sophila led to a “reprogramming” of cells within the fat body, altered expression of known TGF-β signaling regulators (49) and affecting the expression of genes involved in fatty acid metabo- epigenetic silencing of TGF-β target genes in ovarian cancer (50– lism and immune function. Pathway analysis of our expression 52) have been reported. The present study identifies VEPH1 as a data also indicates altered expression of genes involved in me- unique negative modulator of canonical TGF-β signaling. tabolism, phototransduction, and immune response, suggesting The gene encoding VEPH1 localizes to a region frequently re- conservation of function between VEPH1 and Melted. Although ported as amplified in ovarian cancer (8, 9). Microarray compara- Melted has been shown to affect Hippo, FOXO, and TOR sig- tive genomic hybridization analysis indicated amplification of the naling pathways, TGF-β is also involved with vertebrate adipo- 3q24-26 region in 7 of 12 ovarian clear cell carcinoma cell lines cyte differentiation (43) and eye development (44). VEPH1 examined, including ES2 cells (8). In addition to ES2 cells, we found regions containing TIR-1 and -2 have 44% and 42% sequence high levels of VEPH1 expression in OVCA429 and HEY cells, identity and 65% and 62% positivity, respectively, with Melted. which were derived from serous adenocarcinomas. Whether the Thus, future studies should explore the impact of VEPH1 on VEPH1 locus is amplified in these cells is unknown. These three cell cell-signaling pathways affected by Melted. Our findings also lines exhibit a more invasive phenotype than OVCAR3, SKOV3, raise the possibility that some effects of Melted may be mediated and HOC7 cells (53), which lack VEPH1 expression, raising the through an impact on signaling by Drosophila TGF-β superfamily possibility that VEPH1 may contribute to a more aggressive cell members. Collectively, the evidence support that Melted and phenotype or to the initiation of a subset of ovarian cancers. VEPH1 may function as membrane scaffolding proteins to in- In summary, studies with Drosophila indicate that Melted im- tegrate and modulate these signaling pathways. pacts several signal-transduction pathways during development that Although some studies suggest that the dysregulation of TGF- overlap with those we show are affected by VEPH1 expression. β signaling in ovarian cancer is due to TβRI or TβRII gene mu- We now demonstrate that VEPH1 is a previously unidentified

Shathasivam et al. PNAS Early Edition | 9of10 Downloaded by guest on September 28, 2021 modulator of the canonical TGF-β signaling pathway and acts and statistical analysis are provided in SI Materials and Methods.Gene- to restrict SMAD2 nuclear accumulation through retention of expression profiles were generated by using the Illumina Human HT-12 Bead- SMAD2 at TβRI. Because VEPH1 is up-regulated in many ovarian Chip platform (Version 4.0). A corrected (Benjamini–Hochberg; false discovery cancers and is highly expressed in invasive ovarian cancer cell rate < 0.05) ANOVA followed by a Tukey post hoc test was performed to lines, these findings suggest that VEPH1 may contribute to car- identify probes whose mean expression was significantly different between cinogenesis by blocking canonical TGF-β signaling. treatment groups. Microarray data have been deposited into the Gene Ex- pression Omnibus database (accession no. GSE67765). Materials and Methods ACKNOWLEDGMENTS. We thank Drs. Miriam Barrios-Rodiles and Xaralabos A detailed outline of the procedures and specific materials used for cell culture; Varelas for technical advice; Drs. Veronique Voisin and Gary Bader for assistance generation of expression constructs and stably transfected SKOV3, HepG2, and with GSEA and Cytoscape analysis; and Soyeon Park for technical assistance. This COS7 cells; gene expression profiling and analysis; quantitative RT-PCR work was supported by Canadian Institutes of Health Research Grant MOP74726. (RT-qPCR); Western blot analysis; VEPH1-targeted siRNA studies; immunoflu- P.S. was supported in part by a Kristi Piia Callum Memorial Fellowship, a Frank orescence microscopy; reporter gene assays; coimmunoprecipitation assays; Fletcher Memorial Fund, and a Teresina Florio Graduate Scholarship.

1. Muto E, et al. (2004) Identification and characterization of Veph, a novel gene en- 29. Hammerman PS, et al.; Cancer Genome Atlas Research Network (2012) Comprehen- coding a PH domain-containing protein expressed in the developing central nervous sive genomic characterization of squamous cell lung cancers. Nature 489(7417): system of vertebrates. Biochimie 86(8):523–531. 519–525. 2. Teleman AA, Chen YW, Cohen SM (2005) Drosophila Melted modulates FOXO and 30. Subramanian A, et al. (2005) Gene set enrichment analysis: A knowledge-based ap- TOR activity. Dev Cell 9(2):271–281. proach for interpreting genome-wide expression profiles. Proc Natl Acad Sci USA 3. Mikeladze-Dvali T, et al. (2005) The growth regulators warts/lats and melted interact 102(43):15545–15550. in a bistable loop to specify opposite fates in Drosophila R8 photoreceptors. Cell 31. Mootha VK, et al. (2003) PGC-1alpha-responsive genes involved in oxidative phos- 122(5):775–787. phorylation are coordinately downregulated in human diabetes. Nat Genet 34(3): – 4. Jukam D, Desplan C (2011) Binary regulation of Hippo pathway by Merlin/NF2, Kibra, Lgl, 267 273. 32. Dinkel H, et al. (2012) ELM—the database of eukaryotic linear motifs. Nucleic Acids and Melted specifies and maintains postmitotic neuronal fate. Dev Cell 21(5):874–887. Res 40(Database issue):D242–D251. 5. Jukam D, et al. (2013) Opposite feedbacks in the Hippo pathway for growth control 33. Iłzecka J, Stelmasiak Z, Dobosz B (2002) Transforming growth factor-Beta 1 (tgf-Beta and neural fate. Science 342(6155):1238016. 1) in patients with amyotrophic lateral sclerosis. Cytokine 20(5):239–243. 6. Iwabuchi H, et al. (1995) Genetic analysis of benign, low-grade, and high-grade 34. Krieglstein K, Unsicker K (1996) Distinct modulatory actions of TGF-beta and LIF on ovarian tumors. Cancer Res 55(24):6172–6180. neurotrophin-mediated survival of developing sensory neurons. Neurochem Res 7. Haverty PM, Hon LS, Kaminker JS, Chant J, Zhang Z (2009) High-resolution analysis of 21(7):843–850. copy number alterations and associated expression changes in ovarian tumors. BMC 35. Walsh JE, Young MR (2011) TGF-beta regulation of focal adhesion proteins and Med Genomics 2:21. motility of premalignant oral lesions via protein phosphatase 1. Anticancer Res 8. Tan DS, et al. (2009) PPM1D is a potential therapeutic target in ovarian clear cell 31(10):3159–3164. – carcinomas. Clin Cancer Res 15(7):2269 2280. 36. Cicchini C, et al. (2008) TGFbeta-induced EMT requires focal adhesion kinase (FAK) 9. Ramakrishna M, et al. (2010) Identification of candidate growth promoting genes in signaling. Exp Cell Res 314(1):143–152. ovarian cancer through integrated copy number and expression analysis. PLoS ONE 37. Barrios-Rodiles M, et al. (2005) High-throughput mapping of a dynamic signaling 5(4):e9983. network in mammalian cells. Science 307(5715):1621–1625. 10. Schmierer B, Hill CS (2007) TGFbeta-SMAD signal transduction: molecular specificity 38. Zawel L, et al. (1998) Human Smad3 and Smad4 are sequence-specific transcription and functional flexibility. Nat Rev Mol Cell Biol 8(12):970–982. activators. Mol Cell 1(4):611–617. 11. Shi Y, Massagué J (2003) Mechanisms of TGF-beta signaling from cell membrane to 39. Labbé E, Silvestri C, Hoodless PA, Wrana JL, Attisano L (1998) Smad2 and Smad3 the nucleus. Cell 113(6):685–700. positively and negatively regulate TGF beta-dependent transcription through the 12. Akhurst RJ, Derynck R (2001) TGF-beta signaling in cancer—a double-edged sword. forkhead DNA-binding protein FAST2. Mol Cell 2(1):109–120. Trends Cell Biol 11(11):S44–S51. 40. Massagué J, Chen YG (2000) Controlling TGF-beta signaling. Genes Dev 14(6):627–644. 13. Massagué J (2008) TGFbeta in Cancer. Cell 134(2):215–230. 41. Huse M, et al. (2001) The TGF beta receptor activation process: An inhibitor- to 14. Siegel PM, Massagué J (2003) Cytostatic and apoptotic actions of TGF-beta in ho- substrate-binding switch. Mol Cell 8(3):671–682. meostasis and cancer. Nat Rev Cancer 3(11):807–821. 42. Guo X, et al. (2008) Axin and GSK3- control Smad3 protein stability and modulate 15. Massagué J, Gomis RR (2006) The logic of TGFbeta signaling. FEBS Lett 580(12): TGF- signaling. Genes Dev 22(1):106–120. 2811–2820. 43. Lu H, Ward MG, Adeola O, Ajuwon KM (2013) Regulation of adipocyte differentiation 16. Pardali K, Moustakas A (2007) Actions of TGF-beta as tumor suppressor and pro- and gene expression-crosstalk between TGFβ and wnt signaling pathways. Mol Biol metastatic factor in human cancer. Biochim Biophys Acta 1775(1):21–62. Rep 40(9):5237–5245. 17. Wang YQ, Qi XW, Wang F, Jiang J, Guo QN (2012) Association between TGFBR1 44. Grocott T, Johnson S, Bailey AP, Streit A (2011) Neural crest cells organize the eye via β polymorphisms and cancer risk: A meta-analysis of 35 case-control studies. PLoS ONE TGF- and canonical Wnt signalling. Nat Commun 2:265. 7(8):e42899. 45. Antony ML, Nair R, Sebastian P, Karunagaran D (2010) Changes in expression, and/or 18. Taylor BS, et al. (2010) Integrative genomic profiling of human prostate cancer. mutations in TGF-beta receptors (TGF-beta RI and TGF-beta RII) and Smad 4 in human ovarian tumors. J Cancer Res Clin Oncol 136(3):351–361. Cancer Cell 18(1):11–22. 19. Imielinski M, et al. (2012) Mapping the hallmarks of lung adenocarcinoma with 46. Lynch MA, et al. (1998) Mutational analysis of the transforming growth factor beta receptor type II gene in human ovarian carcinoma. Cancer Res 58(19):4227–4232. massively parallel sequencing. Cell 150(6):1107–1120. 47. Wang D, et al. (2000) Analysis of specific gene mutations in the transforming growth 20. Brennan CW, et al.; TCGA Research Network (2013) The somatic genomic landscape of factor-beta signal transduction pathway in human ovarian cancer. Cancer Res 60(16): glioblastoma. Cell 155(2):462–477. 4507–4512. 21. Koboldt DC, et al.; Cancer Genome Atlas Network (2012) Comprehensive molecular 48. Chen T, et al. (2001) Transforming growth factor-beta receptor type I gene is fre- portraits of human breast tumours. Nature 490(7418):61–70. quently mutated in ovarian carcinomas. Cancer Res 61(12):4679–4682. 22. Baca SC, et al. (2013) Punctuated evolution of prostate cancer genomes. Cell 153(3): 49. Sunde JS, et al. (2006) Expression profiling identifies altered expression of genes that – 666 677. contribute to the inhibition of transforming growth factor-beta signaling in ovarian 23. Barretina J, et al. (2010) Subtype-specific genomic alterations define new targets for cancer. Cancer Res 66(17):8404–8412. – soft-tissue sarcoma therapy. Nat Genet 42(8):715 721. 50. Chou JL, Chen LY, Lai HC, Chan MW (2010) TGF-β: Friend or foe? The role of TGF- 24. Cancer Genome Atlas Research Network (2014) Comprehensive molecular charac- β/SMAD signaling in epigenetic silencing of ovarian cancer and its implication in – terization of urothelial bladder carcinoma. Nature 507(7492):315 322. epigenetic therapy. Expert Opin Ther Targets 14(11):1213–1223. 25. Kandoth C, et al.; Cancer Genome Atlas Research Network (2013) Integrated genomic 51. Chou JL, et al. (2010) Promoter hypermethylation of FBXO32, a novel TGF-beta/ characterization of endometrial carcinoma. Nature 497(7447):67–73. SMAD4 target gene and tumor suppressor, is associated with poor prognosis in hu- 26. Bell D, et al.; Cancer Genome Atlas Research Network (2011) Integrated genomic man ovarian cancer. Lab Invest 90(3):414–425. analyses of ovarian carcinoma. Nature 474(7353):609–615. 52. Chan MW, et al. (2008) Aberrant transforming growth factor beta1 signaling and 27. Barretina J, et al. (2012) The Cancer Cell Line Encyclopedia enables predictive mod- SMAD4 nuclear translocation confer epigenetic repression of ADAM19 in ovarian elling of anticancer drug sensitivity. Nature 483(7391):603–607. cancer. Neoplasia 10(9):908–919. 28. Grasso CS, et al. (2012) The mutational landscape of lethal castration-resistant pros- 53. Sodek KL, Ringuette MJ, Brown TJ (2007) MT1-MMP is the critical determinant of tate cancer. Nature 487(7406):239–243. matrix degradation and invasion by ovarian cancer cells. Br J Cancer 97(3):358–367.

10 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1504671112 Shathasivam et al. Downloaded by guest on September 28, 2021