Oncogene (2014) 33, 279–288 & 2014 Macmillan Publishers Limited All rights reserved 0950-9232/14 www.nature.com/onc

ORIGINAL ARTICLE SOX15 is a candidate tumor suppressor in pancreatic cancer with a potential role in Wnt/b-catenin signaling

KL Thu1, N Radulovich2, DD Becker-Santos1, LA Pikor1, A Pusic1, WW Lockwood1, WL Lam1 and M-S Tsao2

Pancreatic cancer is among the top five deadliest cancers in developed countries. Better knowledge of the molecular mechanisms contributing to its tumorigenesis is imperative to improve patient prognosis. Identification of novel tumor suppressor genes (TSGs) in pancreatic cancer will reveal new mechanisms of pathway deregulation and will ultimately help improve our understanding of this aggressive disease. According to Knudson’s two-hit model, TSGs are classically disrupted by two concerted genetic events. In this study, we combined DNA methylation profiling with copy number and mRNA expression profiling to identify novel TSGs in a set of 20 pancreatic cancer cell lines. These data sets were integrated for each of B12 000 genes in each cell line enabling the elucidation of those genes that undergo DNA hypermethylation, copy-number loss and mRNA downregulation simultaneously in multiple cell lines. Using this integrative genomics strategy, we identified SOX15 (sex determining region Y—box 15) as a candidate TSG in pancreatic cancer. Expression of SOX15 in pancreatic cancer cell lines with undetectable expression resulted in reduced viability of cancer cells both in vitro and in vivo demonstrating its tumor suppressive capability. We also found reduced expression, homozygous deletion and aberrant DNA methylation of SOX15 in clinical pancreatic tumor data sets. Furthermore, we deduced a novel role for SOX15 in suppressing the Wnt/b-catenin signaling pathway, which we hypothesize is a pathway through which SOX15 may exert its tumor suppressive effects in pancreatic cancer.

Oncogene (2014) 33, 279–288; doi:10.1038/onc.2012.595; published online 14 January 2013 Keywords: SOX15; integrative genomics; tumor suppressor; pancreatic cancer; Wnt pathway

INTRODUCTION In this study we performed DNA-methylation profiling in Pancreatic cancer has a very poor prognosis, making it one of the conjunction with copy number and gene expression profiling to top five most lethal malignancies in developed countries.1 Pancreatic identify novel TSGs in PDAC. Our integrative, multi-omics ductal adenocarcinoma (PDAC) is the most prevalent form, approach resulted in the discovery of the candidate TSG, SOX15 comprising up to 90% of cases, and is characterized by mutations (sex determining region Y—box 15), a member of the SOX gene to key oncogenes and tumor suppressors (TSGs) including KRAS, family, which encodes 20 proteins harboring high-mobility group EGFR, TP53, SMAD4 and CDKN2A.2–4 Tumor heterogeneity is a box domains that act as transcriptional activators or repressors, 14 prominent feature of pancreatic malignancies and represents a and function in development and cell differentiation. major challenge in studying PDAC.5 In addition to intratumoral, Additionally, we are the first to show that SOX15 represses Wnt histological and genetic differences, the majority of tumors typically b-catenin signaling. exhibit marked desmoplasia, characterized by infiltration of non- malignant cells such as fibroblasts and inflammatory cells, which RESULTS 2,3 commonly mask the identification of genetic alterations. For this Screen for hypermethylated and underexpressed genes in PDAC reason, cell lines and xenografts enriched for tumor cells have We first screened a panel of 20-cell lines representing the become essential tools in PDAC research. spectrum of PDAC to identify recurrently hypermethylated genes. Despite the recent elucidation of some key driver genes and Clinical and molecular features of these cell lines are summarized core signaling pathways, much remains to be learned about the 6–12 in Table 1. To identify the most robust candidates, we calculated biology driving PDAC tumorigenesis. A better understanding delta b-values (percent difference in methylation) for all CpG of the molecular mechanisms underlying PDAC biology, including sites in each line with reference to normal human pancreatic duct the discovery of new TSGs, is imperative for improving patient epithelial (HPDE) cells, and used a stringent threshold of delta outcome. Inactivation of TSGs can occur through several different b-values in the 95th percentile for defining hypermethylation. mechanisms such as mutation and loss of heterozygosity. This corresponded to a 53% difference in methylation between DNA methylation and deletions are also prominent mechanisms PDAC lines and HPDE. In total, 1752 CpG probes corresponding of TSG silencing in cancer. Thus, consideration of multiple to 1310 unique genes were identified as hypermethylated in 30% genomic dimensions simultaneously is a powerful approach of PDAC lines (Supplementary Table 1). We used a stringent for TSG identification as it facilitates detection of genes with frequency threshold as recurrence suggests biological relevance. A ‘two-hit’ inactivation achieved through different molecular number of these genes have recently been reported as hypermethy- mechanisms.13 lated in PDAC, demonstrating the validity of our approach.15–17

1BC Cancer Research Centre, Vancouver, BC, Canada and 2Ontario Cancer Institute, Princess Margaret Hospital, University Health Network and Department of Laboratory Medicine and Pathobiology, University of Toronto, Toronto, ON, Canada. Correspondence: Dr WL Lam, BC Cancer Research Centre, Vancouver, BC, Canada or Dr M-S Tsao, Ontario Cancer Institute, Princess Margaret Hospital, University Health Network and Department of Laboratory Medicine and Pathobiology, University of Toronto, Toronto, ON, Canada. E-mail: [email protected] or [email protected] Received 21 June 2012; revised 23 October 2012; accepted 2 November 2012; published online 14 January 2013 SOX15 inactivation in pancreatic cancer KL Thu et al 280 Table 1. Clinical and molecular characteristics of 20 PDAC cell lines analyzed in this study

Name Gender Ethnicity Karyotype Tissue Mutations (source)

AsPC1 Female Caucasian NA Pancreatic adenocarcinoma CDKN2A, FBXW7, KRAS, MAP2K4, TP53 (COSMIC) BxPC3 Female NA NA Pancreatic adenocarcinoma CDKN2A, MAP2K4, SMAD4, TP53 (COSMIC) Capan1 Male Caucasian Hypotriploid Pancreatic adenocarcinoma BRCA2, CDKN2A, KRAS, SMAD4, TP53 (COSMIC) Capan2 Male Caucasian Pseudotriploid Pancreatic adenocarcinoma Unknown CFPAC1 Male Caucasian Hyperdiploid Pancreatic ductal adenocarcinoma KRAS, SMAD4, TP53 (COSMIC) HPAC Female Caucasian Hypotriploid Pancreatic ductal adenocarcinoma Unknown HPAFII Male Caucasian NA Pancreatic adenocarcinoma CDKN2A, KRAS, TP53 (COSMIC) Hs766T Male Caucasian Hyperdiploid Pancreatic carcinoma Unknown MIAPaca2 Male Caucasian Hypotriploid Pancreatic carcinoma CDKN2A, KDM6A, KRAS, TP53 (COSMIC) Panc0203 Female Caucasian NA Pancreatic adenocarcinoma KRAS (ATCC) Panc0213 Female Caucasian NA Pancreatic adenocarcinoma Unknown Panc0327 Female Caucasian NA Pancreatic adenocarcinoma CDKN2A, KRAS, SMAD4, TP53 (COSMIC) Panc0403 Male Caucasian NA Pancreatic adenocarcinoma KRAS (ATCC) Panc0504 Female Caucasian NA Pancreatic adenocarcinoma KRAS (ATCC) Panc0813 Male Caucasian NA Pancreatic adenocarcinoma CDKN2A, KRAS, SMAD4 (COSMIC) Panc1 Male Caucasian Hypertriploid Pancreatic ductal epithelioid carcinoma Unknown Panc1005 Male Caucasian NA Pancreatic ductal adenocarcinoma KRAS, TP53 (COSMIC) PL45 Male Caucasian NA Pancreatic ductal adenocarcinoma KRAS (ATCC) SU8686 Female Caucasian NA Pancreatic ductal carcinoma Unknown SW1990 Male Caucasian Near triploid Pancreatic adenocarcinoma CDKN2A, KRAS (COSMIC) Abbreviation: COSMIC, Catalog of Somatic Mutations in Cancer; NA, information unavailable on ATCC website; PDAC, pancreatic ductal adenocarcinoma.

If biologically relevant, gene hypermethylation should result in by each of these alterations across the 20 PDAC lines. Given that downregulation of gene expression. Thus, to prioritize our list of our definition of two-hits required all three of these events to hypermethylated genes to those most likely to have an effect on occur simultaneously, we multiplied the average proportions for PDAC biology, we looked for those with a minimum two-fold each event to calculate the proportion of genes expected to decrease in expression relative to HPDE within individual cell lines. undergo two-hits by chance. The average proportions for each Using this criterion, we identified 247 genes with concurrent event were: hypermethylation 0.0547, copy-number loss 0.2064 hypermethylation and underexpression in at least 30% of PDAC and underexpression 0.2601. Thus, the average probability of a lines (Figure 1a, Supplementary Table 2). Among these genes were gene being affected by two-hits in a single cell line was 0.0029. a number of known and candidate TSGs including SFRP1, SOX17, Furthermore, the probability of observing two-hits in a single gene TUSC3, CDH13 and TIMP3.18–24 in three of 20 lines (15% of samples) drops to 2.4  10 À 8,a probability so low that we would not expect to identify any genes Two-hit genes: concurrent hypermethylation, DNA copy-number by chance (11 984 genes  2.4  10 À 8 ¼ 0.0003 genes). loss and underexpression Integration of DNA methylation, copy number and mRNA expres- Clinical relevance of gene expression changes sion profiles for individual genes within individual cancer lines allowed us to screen our list of hypermethylated genes for those As a final means to refine our list of two-hit genes, we examined which also harbored copy-number loss. As our strategy is based on expression in clinical pancreatic tumors compared to non- the two-hit model, we hypothesized DNA deletions would malignant pancreatic tissues (Figure 1a). As cell lines have complement hypermethylation, providing a second hit for gene adapted to artificial in vitro conditions, assessing candidate gene inactivation. Array CGH data were segmented to determine the expression in actual tumor samples is essential to rule out copy number status (gain, loss or neutral) for each gene in each artefacts of tissue culture. Expression data for the 91 candidates PDAC genome. Copy number alterations are summarized in was extracted from two independent clinical data sets from the Supplementary Tables 3 and 4. Of our 247 hypermethylated genes, Gene Expression Omnibus data repository (GSE11838: 28 tumors 185 exhibited concurrent copy-number loss in at least one sample. and 4 normals, and GSE15471: 36-matched tumors and normals, Inactivation of genes by two-hit events is a rare occurrence that see Methods). A total of 24 genes were differentially expressed implies selection against important genes, and likely leads to gene (U test P-valueo0.05), having significantly lower expression silencing and a growth advantage for cancer cells. As recurrence of in pancreatic tumors versus non-malignant pancreas tissues, the two-hit phenomenon in multiple samples would provide even consistent with gene downregulation (Table 2). further evidence of selection against a gene in PDAC, we implemented an arbitrary but stringent two-hit frequency threshold Over-representation of two-hit genes in the Wnt/b-catenin of 15% for candidate genes to move forward in our analysis pathway (Figure 1a). This recurrence threshold revealed 91 two-hit genes Upon refining our list of two-hit genes, we sought to understand based on our integrative genomic analysis (Supplementary Table 5). how they might contribute to PDAC biology. We used pathway analysis to uncover biological relationships between these genes Two-hits at specific gene candidates occur remarkably more often in order to discern the cellular pathways deregulated as a than expected by chance consequence of their disruption. This analysis revealed that the To determine if the two-hit genes were identified by chance, we Wnt/b-catenin pathway was significantly enriched for two-hit performed a calculation based on our observed frequencies of inactivation of pathway inhibitors, including SFRP1 and DKK2.In methylation, copy number and expression changes. For each cell fact, the Wnt/b-catenin pathway was the only significantly line, we calculated the number of genes with hypermethylation, enriched pathway we observed (Benjamini Hochberg corrected copy-number loss and underexpression, and calculated the Fisher’s exact test P-value ¼ 0.005, Figure 1b, Supplementary average proportion of genes (of 11 984 genes assessed) affected Figure 1). Two-hit inactivation of known pathway inhibitors is

Oncogene (2014) 279 – 288 & 2014 Macmillan Publishers Limited SOX15 inactivation in pancreatic cancer KL Thu et al 281 -log(BH p-value) Hypermethylation 0.0 0.5 1.0 1.5 2.0 2.5 and underexpression in ≥ 30% of cell lines Wnt/β-catenin Signaling Second hit (ie. concurrent Role of Osteoblasts, DNA copy number loss) Osteoclasts and Chondrocytes in ≥ 15% of cell lines in Rheumatoid Arthritis

Differential expression in two Role of Macrophages, clinical tumor datasets Fibroblasts and Endothelial Cells in Rheumatoid Arthritis

24 genes Granzyme A Signaling

gene Ascorbate and Aldarate expression Metabolism

0.000 0.025 0.050 0.075 0.100 0.125 0.150 Ratio

Gene Alteration AsPC1 BxPC3 Capan1 Capan2 CFPAC1 HPAC HPAFII Hs766T MIAPaca2 Panc0203 Panc0213 Panc0327 Panc0403 Panc0504 Panc0813 Panc1005 Panc1 PL45 SU8686 SW1990 Hyper Methylation Copy Number Loss SOX15 Under Expression Two-Hit Hyper Methylation Copy Number Loss SFRP1 Under Expression Two-Hit Hyper Methylation Copy Number Loss DKK2 Under Expression Two-Hit Hyper Methylation Copy Number Loss WNT7A Under Expression Two-Hit Figure 1. Genes frequently disrupted by two-hits are significantly enriched for involvement in the Wnt/b-catenin signaling pathway. (a) Schematic illustrating the process used to identify potential TSGs inactivated by ‘two-hits’ in PDAC. A total of 24 genes exhibited concurrent hypermethylation, DNA copy-number loss and underexpression in at least 15% of PDAC lines, as well as significantly reduced expression in two independent clinical pancreatic cancer cohorts (U test, Po0.05). Black lollipops on one allele indicate DNA hypermethylation, whereas the gap in sequence on the second allele indicates loss of DNA, contributing to two-hit inactivation of candidate TSGs. The occurrence of these events within a single cancer cell line results in decreased gene expression. (b) Ingenuity Pathway Analysis results for two-hit genes identified in PDAC cell lines. The orange vertical line indicates the corrected P-value significance threshold of 0.05. The dotted orange line indicates the ratio of the number of two-hit genes identified to the total number of genes involved in the pathway. (c) Disruption of Wnt/b-catenin pathway components in PDAC lines. The occurrence of DNA alterations (hypermethylation and copy- number loss) and underexpression are indicated by red boxes for each of the 20 PDAC lines assessed in this study. Blue boxes represent two-hits and concurrent underexpression within a single cell line.

consistent with activation of the Wnt pathway, which is well been described as a TSG.18,20,38,39 We proceeded to investigate the documented in PDAC.5,6,25–27 role of SOX15 as a novel TSG in PDAC, because (1) we observed Four of our twenty-four two-hit genes were identified as being strong genomic evidence of its inactivation, (2) other SOX involved in Wnt signaling, all with inhibitory roles, and two-hit members function as Wnt antagonists, (3) the three other Wnt inactivation of at least one of the four genes occurred in 12 of the pathway genes (SFRP1, DKK2 and WNT7A) we identified have 20 cell lines (60%) (Figure 1c). SFRP1, DKK2 and WNT7A act already been shown to be TSGs and (4) SOX15 has never been upstream in the pathway at the level of receptor–ligand binding, described as a TSG before. whereas SOX15 is predicted to act downstream at the level of We observed two-hits affecting SOX15 in 9 of 20 (45%) PDAC transcription (Supplementary Figure 1). To our knowledge, this is lines (Table 2). Of the 24 candidate genes identified, SOX15 had the first report to describe two-hit inactivation of the Wnt the second highest two-hit frequency (45%), second only to suppressors SFRP1 and DKK2 in cancer, as silencing of these TSGs SERPINB13 that showed two-hits in 10 of 20 (50%). However, is predominantly described as occurring through aberrant DNA SERPINB13 transcript levels were underexpressed in 70% of the cell methylation.28–37 lines, whereas SOX15 was underexpressed in 100%, providing further rationale for investigation of SOX15 as our top candidate. Genomic DNA sequencing of the two exons comprising SOX15 for Identification of SOX15 as a novel candidate TSG the 20 PDAC lines did not reveal any mutations, suggesting DNA Whereas SOX family members, such as SOX17 and SOX6, are deletions and hypermethylation are the primary mechanisms of established TSGs in other cancer types, SOX15 has not previously SOX15 inactivation in PDAC.

& 2014 Macmillan Publishers Limited Oncogene (2014) 279 – 288 SOX15 inactivation in pancreatic cancer KL Thu et al 282 Table 2. Frequencies of disruption and U test P-values for 24 candidate TSGs in pancreatic cancer

Symbol Hypermethylation Copy- Underexpression Concurrent Concurrent copy- Two-hit GSE11838 GSE15471 number hypermethylation number loss and underexpression P-valuea P-valuea loss and underexpression underexpression

SERPINB13 0.850 0.850 0.700 0.550 0.650 0.500 0.004 0.000 SOX15 0.800 0.550 1.000 0.800 0.550 0.450 0.002 0.000 TMEM40 0.950 0.350 0.950 0.950 0.350 0.350 0.003 0.004 CALML3 1.000 0.500 0.500 0.500 0.300 0.300 0.012 0.000 ALOX12 0.600 0.600 0.700 0.450 0.350 0.250 0.001 0.000 ELAVL2 0.450 0.500 0.800 0.400 0.500 0.250 0.028 0.014 FBXO2 0.700 0.300 0.600 0.450 0.250 0.250 0.041 0.000 HIST1H1A 0.650 0.400 1.000 0.650 0.400 0.250 0.006 0.000 ARMC4 0.600 0.600 0.700 0.450 0.350 0.200 0.006 0.000 DKK2 0.400 0.300 0.950 0.350 0.300 0.200 0.006 0.001 DMRT2 0.500 0.600 0.650 0.350 0.400 0.200 0.004 0.000 GPR74 0.500 0.200 0.850 0.500 0.200 0.200 0.003 0.007 HBQ1 0.700 0.250 0.800 0.600 0.250 0.200 0.019 0.000 LAMA1 0.700 0.350 0.950 0.650 0.350 0.200 0.009 0.000 ALDH1A2 0.450 0.350 0.500 0.300 0.150 0.150 0.007 0.002 CLDN11 0.450 0.250 0.700 0.400 0.200 0.150 0.041 0.000 CNTNAP2 0.650 0.150 0.400 0.300 0.150 0.150 0.016 0.002 FGF5 0.450 0.150 1.000 0.450 0.150 0.150 0.006 0.000 KCNQ1DN 0.700 0.300 0.350 0.300 0.150 0.150 0.002 0.000 PAK6 0.900 0.150 0.800 0.800 0.150 0.150 0.004 0.004 SFRP1 0.450 0.250 1.000 0.450 0.250 0.150 0.046 0.000 SLC6A15 0.650 0.250 1.000 0.650 0.250 0.150 0.025 0.000 UCK1 0.450 0.300 0.950 0.450 0.300 0.150 0.036 0.000 WNT7A 0.350 0.450 0.650 0.350 0.350 0.150 0.022 0.004 Frequencies based on 20 PDAC cell lines. Italics indicate genes involved in Wnt/b-catenin signaling. aP-values for U test comparing expression in pancreatic tumor and non-malignant pancreas tissues.

SOX15 is ubiquitously underexpressed in PDAC lines Panc0327-EV) developed tumors; however, Capan2-SOX15 and To validate the ubiquitous loss of SOX15 expression observed in Panc03.27-SOX15 mice showed remarkably suppressed tumor PDAC lines, we performed qRT-PCR to measure SOX15 transcript formation (Figures 3e and f), indicating SOX15 has a significant levels in each line relative to HPDE. Confirming our microarray effect on impeding tumor formation. results, qRT-PCR revealed that SOX15 was underexpressed in all cell lines, with 15 of 20 lines having an expression level less than SOX15 may exert its TSG effects via suppression of Wnt/b-catenin one percent of that observed in HPDE (Figure 2a). Western signaling blotting further corroborated our mRNA expression results, After confirming its tumor suppressor abilities, we investigated the demonstrating a strong correlation between SOX15 mRNA and mechanism through which SOX15 mediates its TSG effects. Our protein expression (Figure 2b). We also observed significant pathway analysis suggested a role for SOX15 in the Wnt pathway underexpression of SOX15 in two distinct clinical PDAC cohorts based on the fact that other SOX proteins function as Wnt (Figures 2c and d). The reduced SOX15 expression levels observed pathway antagonists,14,40 thus, we assessed the effect of SOX15 in both cancer cell lines and clinical tumors suggests that expression on Wnt pathway activity using four different strategies. underexpression is a consequence of the genomic alterations First, we quantified expression levels of six transcriptional targets we observed, characteristic of TSG silencing. of the Wnt pathway, AXIN2, CCND1, CD44, JUN, LEF1 and MYC in Capan2 cells with and without SOX15 expression. qRT-PCR SOX15 exhibits biological characteristics of a TSG in PDAC in vitro confirmed our assumption that expression levels for these genes and in vivo in Capan2-EV exceeded those in Capan2-SOX15 (Figure 4a). The Next, we ectopically expressed SOX15 in PDAC cell lines with average fold change in expression for the Wnt target genes in undetectable SOX15 expression to assess its tumor suppressor Capan2-EV relative to Capan2-SOX15 ranged from 1.323–2.399 capability. A full length SOX15 cDNA expression vector was over three replicates. Second, we performed the TOP-Flash introduced into Capan2 and Panc0327 to generate cells stably reporter assay, considered the gold standard for measuring Wnt expressing SOX15 at levels comparable to those of non-malignant pathway activity. After three independent replicates, the average pancreatic cells (Figure 2b, Figures 3a and b). Capan2- and fold change in Wnt activity in Capan2-EV relative to Capan2- Panc0327-SOX15 cells showed a reproducible and significant SOX15 cells was 2.855 (Figure 4b). Third, we measured active reduction in cell viability relative to Capan2- and Panc0327-EV b-catenin (unphosphorylated at serine 33 or threonine 37) (empty vector control) cells in vitro, as measured over 5 levels.31,41 Corroborating our qRT-PCR and TOP-Flash results, consecutive days using the MTT assay (Student’s t-test, Po0.05, active b-catenin levels were up to 18% lower in Capan2-SOX15 Figures 3c and d). Given that SOX15 negatively affected cell than in Capan2-EV (Figure 4c). Lastly, we performed a gene set viability in vitro, we then asked whether SOX15 could suppress enrichment analysis on Capan2-EV and Capan2-SOX15 ex- growth in vivo. Capan2 and Panc0327 control cells, and cells stably pression profiles (Supplementary Tables 6 and 7) to determine expressing SOX15 were subcutaneously injected into the flank whether genes more highly expressed in Capan2-EV versus of severe combined immunodeficient mice and monitored for SOX15-expressing cells were enriched for TCF and LEF (members 7 weeks. All mice injected with empty vector cells (Capan2-EV and of the b-catenin complex) transcriptional targets. As hypothesized,

Oncogene (2014) 279 – 288 & 2014 Macmillan Publishers Limited SOX15 inactivation in pancreatic cancer KL Thu et al 283

1.4 GSE11838 4 1.2 p = 0.001671 1 3 0.8 0.6 2 0.4 1 0.2 Relative quantification 0 Normalized Expression 0

PL45 HPDE HPAC Panc1 AsPC1BxPC3Capan1 HPAFII Capan2CFPAC1 Hs766T SU8686SW1990 Normals Tumors MIAPaca2Panc02.03Panc02.13Panc03.27Panc04.03Panc05.04Panc08.13Panc10.05

ACTB 45kDa GSE15471 SOX15 28kDa 8.0 p = 0.000278 7.5 7.0 HPDE HPAC PL45 AsPC1BxPC3Capan1Capan2 HPAFIIHs766T + control CFPAC1 6.5 6.0 ACTB 45kDa 5.5

SOX15 28kDa Normalized Expression 5.0

Normals Tumors Panc1 + control SW1990SU8686 Panc02.03Panc04.03Panc10.05Panc08.13Panc05.24Panc03.27 MIAPaca2 Figure 2. SOX15 is underexpressed in PDAC cell lines and tumors. (a) SOX15 is ubiquitously underexpressed in 20 PDAC lines relative to HPDE as determined by qPCR. (b) Low expression levels of SOX15 in PDAC lines was verified at the protein level. Capan2 cells overexpressing SOX15 served as a positive control and ACTB was used as a loading control. (c,d) Publically available gene expression microarray data was used to compare SOX15 expression levels in pancreatic tumors versus non-malignant pancreatic tissues. The mean expression values and s.e.’s of the mean are indicated by the horizontal bars in the dot plots.

Capan2 Capan2 Capan2 ACTB 45kDa 0.40 ) * 3 0.35 EV EV * SOX15 SOX SOX15 28kDa 0.30 * 0.25 * EV SOX15 0.20 * * *

Panc0327 OD Value 0.15 * 0.10 * * ACTB 45kDa * 0.05 0 Average Tumor Volume (cm SOX15 28kDa Day1 Day2 Day3 Day4 Day5 0 10172125283235414752 Days Post Injection EV SOX15

Panc0327 Capan2 Panc0327 Panc0327 700 1.0 )

3 * 0.9 EV * EV 600 0.8 SOX15 SOX * * 500 0.7 * 0.6 400 0.5 * 0.4 *

300 OD Value 0.3 * 200 0.2 * Relative Expression * 100 0.1 * 0 Average Tumor Volume (cm 0 Day1 Day2 Day3 Day4 Day5 EV SOX15 0 1117212528353942 46 50 53 Days Post Injection Figure 3. In vitro and in vivo effects of SOX15 expression. (a) SOX15 protein expression levels in Capan2 and Panc0327 cells modified to express SOX15. ACTB was used as a loading control. (b) Relative SOX15 expression levels in Capan2 and Panc0327 cells as determined by qPCR. (c,d) Cell viability in Capan2 and Panc0327 cells as measured over a 5 day period by the MTT assay. The s.e. for duplicated experiments is shown. (e,f) Effect of SOX15 expression on tumor burden in severe combined immunodeficient mice injected with Capan2-EV or Panc0327-EV and Capan2-SOX15 or Panc0327-SOX15 cells. Tumor measurements were taken by palpating tumors through the skin throughout the experiment, and error bars indicate the s.d. across each group of five mice. Asterisks indicate significant differences. EV ¼ empty vector controls, SOX15 ¼ SOX15 expressing cells.

& 2014 Macmillan Publishers Limited Oncogene (2014) 279 – 288 SOX15 inactivation in pancreatic cancer KL Thu et al 284 3 of genes with two-fold expression changes (over or under- expressed) revealed ERK5 signaling as the top pathway affected, a 2.5 pathway known to promote the hallmarks of cancer, including cell proliferation.42,43 The three genes identified in this pathway were 2 FOS, MEF2C (downregulated when SOX15 is expressed) and MAP2K5 (namely MEK5, upregulated when SOX15 is expressed). 1.5 Further exploration of the role of SOX15 in both the Wnt and ERK5 signaling pathways is warranted. 1

Relative Expression 0.5

(to Capan2-SOX15 line) DISCUSSION 0 Whereas numerous approaches can be used for TSG identification, such as methylation profiling alone or mutation or homozygous JUN LEF1 MYC AXIN2 CD44 deletion (HD) screening, we used a two-hit screening approach. CCND1 We acknowledge that some TSGs may be haploinsufficient whereby mono-allelic disruption (that is inactivation by DNA 1.2 Fold Change alteration at only one allele) is enough to be functionally relevant. 1 (EV/SOX15) = 2.9 Nevertheless, we employed our integrative genomics strategy in 0.8 which we simultaneously analyzed multiple dimensions of the cancer genome, and identified SOX15 as a candidate TSG in PDAC. 0.6 Unlike other TSGs, such as RB1, p53 and p16, mutation is not the 0.4 primary mechanism of SOX15 inactivation. Instead, we observed concurrent hypermethylation and DNA copy loss. We confirmed

Top-Flash Activity Top-Flash 0.2 SOX15 inactivation in PDAC, and demonstrated its tumor 0 Capan2 Capan2 suppressor abilities in cell and animal models. Our work EV SOX15 demonstrates the power of a multi-dimensional genomics approach and its utility for the robust identification of novel 1.00 0.82 cancer genes. A caveat of our study worth mentioning is that a single non-malignant cell line (HPDE) was used as a baseline for Active β-Catenin 92 kDa defining PDAC alterations, as opposed to a panel of non- malignant pancreas lines. Despite this limitation, we were able GAPDH 37 kDa to validate our findings in multiple publically accessible PDAC tumor studies where non-malignant pancreas tissues were used as Capan2 Capan2 controls, giving us confidence in the veracity of our data. EV SOX15 Members of the SOX family of transcription factors contain a Figure 4. Wnt/b-catenin activity is higher when SOX15 is not high-mobility group box domain that facilitates DNA binding. expressed. (a) mRNA expression levels of six known Wnt pathway They may act as transcriptional activators or repressors depending target genes were compared in Capan2-EV versus Capan2-SOX15 on their amino acid sequences and specific binding partners, and cells. (b) TOP-FLASH assay results illustrating Wnt activity levels in have emerged as important genes in a variety of tumor types.14 Capan2-EV compared to Capan2-SOX15 cells. (c) Active b-catenin SOX2 and SOX4 are oncogenes in lung and esophageal squamous (dephosphorylated at Ser37 and Thr41) levels in Capan2-EV and cell carcinomas, and prostate and lung cancers, respec- Capan2-SOX15 cells. Relative active b-catenin protein expression tively,44–46 whereas SOX6, SOX11 and SOX17 function as levels are indicated and GAPDH was used as a loading control. Error bars in (a) and (b) indicate the s.e. of three replicate tumor suppressors in esophageal squamous cell carcinoma, experiments. hematopoietic malignancies, and gastric and liver cancers, respectively.18,20,38,39,47 SOX15 exhibits classic features of a TSG. We observed frequent downregulation of SOX15 in two independent clinical pancreatic LEF1, TCF1 and TCF4 transcriptional target gene sets were tumor cohorts in addition to our panel of cell lines, which enriched in cells lacking SOX15 expression, which is consistent suggests SOX15 inactivation is selected for in PDAC. Hyper- with elevated Wnt activity in Capan2-EV relative to Capan2-SOX15 methylation was observed in 80%, copy-number loss in 55% and (FDR qo0.05). underexpression in 100% of the 20 PDAC lines we profiled. These assays revealed a moderate but significant and repea- Aberrant DNA methylation of SOX15 in pancreatic tumors has table decrease in Wnt activity in cells expressing SOX15 compared been reported,16,17 suggesting SOX15 methylation is not limited to with isogenic cells lacking SOX15 expression, supporting an cultured cells. Furthermore, Vincent et al.16 demonstrated that inhibitory role for SOX15 in Wnt signaling. Considering Capan2 DNA methylation contributes to SOX15 silencing, as treatment of has elevated Wnt activity to begin with and it also exhibits PDAC cell lines with the DNA methyltransferase inhibitor 5-aza-20- downregulation of other Wnt inhibitors (SFRP1 and DKK2, deoxycytidine reversed silencing of SOX15 expression. Figure 1c), our observations of reduced Wnt activity when The Catalog of Somatic Mutations in Cancer (COSMIC) database SOX15 is expressed is substantial. Although the specific mechan- reports only one SOX15 mutation in cancer, which was found in ism of SOX15 involvement with Wnt-mediated transcription the breast cancer cell line, HCC1395.48 Although we and others did remains to be elucidated, our results provide evidence to support not detect sequence mutations in SOX15 in PDAC, deletions that SOX15 behaves as a TSG at least in part by repressing Wnt- including HD of SOX15 are present in public genomic profiling driven transcriptional activity in Capan2 cells. data for pancreatic cancers illustrating another mechanism of Considering the profound effect of SOX15 overexpression on SOX15 inactivation typically observed with other TSGs.6,49 We did tumor growth in vivo and the moderate effect of SOX15 on Wnt not observe any HDs of SOX15 in the PDAC cell lines, suggesting activity, we investigated other avenues through which SOX15 may concurrent hypermethylation, and single copy loss is the primary exert its tumor suppressor abilities using expression profiles for mechanism of SOX15 disruption. Of interest, evaluation of public Capan2 cells with and without SOX15 expression. Pathway analysis databases revealed that putative SOX15 HDs also occur in colon

Oncogene (2014) 279 – 288 & 2014 Macmillan Publishers Limited SOX15 inactivation in pancreatic cancer KL Thu et al 285 and prostate cancers and sarcomas.50 Furthermore, querying the complex transcription) levels of the pathway, elucidation of the public data sets in The Cancer Genome Atlas data portal revealed most efficient therapeutic intervention points remains a four colon adenocarcinomas with concomitant copy-number loss, challenge.59 Our study demonstrates yet another mechanism of DNA methylation and low SOX15 expression, highlighting the TSG Wnt pathway disruption, which occurs downstream in the potential of SOX15 in colon cancer. Two cell lines in our study pathway, akin to CTNNB1 mutations in other cancer types.57–59 (HPAFII and SW1990) exhibited SOX15 underexpression in the Thus, the high frequency of SOX15 disruption and consequential absence of mutation, deletion or methylation, suggesting that expression of Wnt target genes warrants consideration of additional unidentified mechanisms capable of silencing SOX15 therapeutics aimed at targeting the nuclear level (for example, exist. b-catenin/TCF complex) of the Wnt pathway in PDAC. Given the In addition to genomic evidence for SOX15 as a TSG, we complexity of Wnt signaling and the multitude of alterations demonstrated the tumor suppressor potential of SOX15 in vitro contributing to aberrant Wnt stimulation, assessment of the and in vivo. Forced expression of SOX15 led to reduced cell association between SOX15 status and response to Wnt-targeted viability in culture, and drastically reduced tumor formation in therapeutics is a topic for future study. mice. These results strongly exemplify the tumor suppressor capacity of SOX15. Stable knockdown of SOX15 in HPDE did not cause a change in cell viability (data not shown), which is not surprising, given that a single molecular change alone is unlikely MATERIALS AND METHODS to transform non-malignant cells. Similar to our findings, over- Cell lines and culture expression of SOX15 in testicular embryonal carcinoma cells The 20 PDAC cell lines were obtained from the American Type Culture caused a reduction in cell proliferation further demonstrating the Collection (ATCC, Manassas, VA, USA) (Table 1). Cells were cultured growth suppressive effects of SOX15 in cancer.51 according to ATCC instructions. HPDE cells were generated by Dr M-S Tsao 52–56 (Princess Margaret Hospital, Canada) and cultured in keratinocyte serum- Besides its established role in muscle development, little is free media supplemented with epidermal growth factor and bovine known about the function of SOX15, especially in cancer. Yan 51 pituitary extract (Invitrogen, Burlington, ON, USA). DNA and RNA were et al. found that overexpression of SOX15 in embyronal testicular isolated using phenol-chloroform and Trizol extractions. carcinoma cells caused an increase of cells in the G1/G0 phase, a decrease of cells in the G2/M phase, and resulted in reduced cell 54 proliferation. In contrast, Meeson et al. found that SOX15 DNA methylation profiling knockdown in myogenic progenitor cells induced G0/G1 arrest HM27 Infinium arrays were used to measure DNA methylation (Illumina, and a decrease in cell proliferation. These studies suggest SOX15 San Diego, CA, USA).63 Data were processed using Bead Studio and the may have a role in cell cycle control that is likely cell type methylation level for each CpG locus, termed the b-value was calculated, dependent. As we demonstrated that SOX15 suppressed Wnt- reflecting the methylated signal as a proportion of total signal (methylated mediated transcription, it is possible that SOX15 downregulation and unmethylated) for each CpG site. b-value differences between each helps to drive transcription of pro-proliferative genes, resulting in PDAC line and HPDE were calculated for every CpG probe, and the probes increased cell proliferation in PDAC cells. with b-value differences in the 95th percentile of all b-value differences Wnt pathway activation is a prominent feature of were considered hypermethylated. PDAC.5,6,14,25–27,40 Genetic alterations promoting Wnt signaling in cancer include activating CTNNB1 (b-catenin) mutations, and inactivating AXIN and APC mutations.57–59 These mutations are DNA copy number profiling Copy number profiles were generated using tiling path arrays, and not often observed in PDACs and none have been reported in the 64,65 cell line panel we studied, suggesting that other mechanisms exist processed and normalized as previously described. Each profile was 6,48,57 then segmented to define the copy number status for each region as to drive Wnt pathway activation. Epigenetic silencing of DKK gained, lost or neutral using the breakpoint algorithm, FACADE.66 and SFRP genes, which encode inhibitors of receptor–ligand binding at the top of the Wnt signaling cascade, is another 28–34,60 mechanism known to drive Wnt signaling in cancer. Gene expression profiling Additionally, a number of SOX family proteins (for example, SOX6, Expression profiles for PDAC and HPDE cell lines were generated using SOX9, SOX17) are known antagonists of the Wnt pathway.14,40 This Agilent 4 Â 44K expression arrays (Agilent, Mississauga, ON, USA). prompted us to investigate a role for SOX15 in modulating Wnt/b- Expression values were normalized by dividing the background subtracted catenin signaling. We showed SOX15 expression had a modest but median intensity value by the median array intensity for each array. Fold reproducible effect on suppressing Wnt activity in PDAC cells. change in expression relative to HPDE was calculated for each probe in Considering PDACs also exhibit downregulation of Wnt inhibitors each PDAC line. Genes were classified as underexpressed if the fold change including DKK2 and SFRP1, which likely contribute to aberrant Wnt was less than twofold. To reduce the potential overestimation of signaling, we were impressed to observe a decrease in Wnt activity underexpression calls due to poor probe performance, the bottom 2% of with the introduction of SOX15.WeproposeSOX15 silencing expression values for HPDE were removed. contributes to activation of Wnt signaling and may act Profiles for Capan2 cells engineered to express SOX15 and correspond- ing controls (see below) were generated on HT-12 v4 Beadchips in cooperatively with genetic and/or epigenetic disruption of other duplicate (Illumina). Data were robust spline normalized using BRB array Wnt pathway components in PDAC cells. To our knowledge, our tools, replicate arrays were averaged, and fold changes between Capan2- study is not only the first to describe SOX15 as a candidate TSG in EV and Capan2-SOX15 cells were calculated for each probe.67 A pre-ranked cancer, but also the first to establish SOX15 as a regulator of Wnt/b- gene set enrichment analysis was performed on fold change data using catenin signaling. In addition to identifying a role for SOX15 in the C3 transcription factor target v2.5 gene set to determine whether TCF/ modulating the Wnt pathway, we also provide evidence suggesting LEF (Wnt responsive) transcription factor target genes were affected by its involvement in ERK5 signaling, although this role requires SOX15 expression. experimental validation. SOX15 expression was assessed in two publically available clinical The discovery that Wnt signaling is a key pathway in multiple cohorts. GSE11838 expression data was downloaded for 28 pancreatic tumors and four normal pancreas tissues using Agilent 60mer Oligo cancers, has made the development of therapeutic Wnt inhibitors arrays. Data was normalized using median array normalization. GSE15471 a priority for several malignancies including colorectal, liver, lung 58,59,61,62 downloaded data was comprised of Affymetrix U133 Plus expression arrays and PDAC. However, as the Wnt pathway is disrupted for 36 pancreatic tumor and 36 normal pancreas tissues. Array data was through a variety of mechanisms at the top (receptor–ligand RMA normalized. Expression in tumors relative to normals was compared binding), middle (b-catenin stability) and bottom (b-catenin/TCF using a U test, with a P-valueo0.05 considered significant.

& 2014 Macmillan Publishers Limited Oncogene (2014) 279 – 288 SOX15 inactivation in pancreatic cancer KL Thu et al 286 Quantitative real-time PCR (qRT-PCR) male severe combined immunodeficient mice bred at the Ontario Cancer SOX15 mRNA expression levels were quantified by quantitative real-time Institute animal facility. All manipulations were conducted in accordance PCR (qRT-PCR) using TaqMan gene expression assays for SOX15 with protocols approved by the OCI Animal Care Committee. Subcuta- (Hs00199511_m1) and 18S (Hs99999901_s1) as an endogenous control neous flank injections of two million cells in 50 ml of growth media were 72 (Life Technologies, Carlsbad, CA, USA). Relative expression levels were performed as previously described. After 10 days, tumor size was calculated using the DDCt method. Six Wnt target genes were also measured by palpation through the skin every 3–4 days and tumor volume 2 assessed using the following TaqMan primers: AXIN2 Hs00610344_m1, calculated (length  width  p/6). Experiments were terminated after 3 c-JUN Hs00277190_s1, c-MYC Hs00905030_m1, CD44 Hs01075861_m1, tumors reached 1.0–1.5 cm . Differences in tumor burden were compared CCND1 Hs00765553_m1, LEF1 Hs01547250_m1. between mice injected with cell lines expressing SOX15 (n ¼ 5) and those expressing an empty vector (EV) (n ¼ 5) using a Student’s t-test. Tumors formed were excised to confirm PDAC histology. SOX15 mutation screening The two SOX15 exons were sequenced from genomic DNA for the 20 PDAC lines by the Sanger’s method using primer pairs: exon 1-1 (50-AACTAGGAC Wnt-activity reporter assays ACAGTCGCATCC-30,50-CAAGATGCACAACTCCGAGA-30)’, exon 1-2 (50-GTAG Capan2-SOX15 and Capan2-EV cells were transfected with pTOPFlash or TCGCGCAGGTGTCG-30,50-GGTTTAAGCTCTGGGCTTGG-30) and exon 2 (50-G pFOPFlash plasmids containing multimerized wild-type and mutated TAGTCGCGCAGGTGTCG-30,50-GTGGCACGCTCTACCCTG-30).6 TCF/LEF binding sites, respectively (Addgene plasmids 12456 and 12457, Cambridge, MA, USA). A Renilla reniformis luciferase reporter was used as an internal control for normalization of transfection efficiencies. Transfec- Generation of Capan2 cells stably expressing SOX15 and HPDE tions were performed with Lipofectamine LTX (Invitrogen). Luciferase cells with SOX15 knockdown activities were measured 24 h post transfection using a luciferase reporter SOX15 cDNA from Open Biosystems (MHS1010-57552, Thermo Fisher kit (E1910, Promega). Transfections were performed in triplicate and the Scientific, Waltham, MA, USA) was PCR amplified (primers: 50-CACCATGGCG fold change in luciferase activity of pTOPFlash to pFOPFlash was CTACCAGG-30,50-TTAGAGGTGGGTTAGGGGCA-30), cloned using the pENTR determined for each cell line. Experiments were repeated three times. Directional TOPO Cloning Kit (Invitrogen, Carlsbad, CA, USA) and clones DNA methylation, copy number, and gene expression profiles for PDAC were verified by Sanger sequencing. The lentiviral SOX15 expression and HPDE lines have been deposited in the Gene Expression Omnibus construct was generated using LR Recombinase Mix (Invitrogen) between (GSE40099). Supplementary Information including detailed methods pENTR and the destination vector, pLD_puro_CcF,68 which contains a accompany the paper on the Oncogene website (http://www.nature. puromycin selective marker. Positive pLD_puro_SOX15 clones were com/onc). sequence verified. 293T cells were co-transfected with pLD_puro_SOX15 or pLD_puro_CcF 69 vectors and three packaging plasmids using Fugene transfection reagent CONFLICT OF INTEREST (Promega, Madison, WI, USA).A total of 3 ml of virus was used to infect The authors declare no conflict of interest. 200 000 Capan2 cells. Puromycin (0.5 mg/ml) was added to the infected plates 48 h post infection. Capan2-SOX15 and Capan2-EV (pLD_puro_CcF empty vector control) cells were selected for two weeks. SOX15 expression was confirmed by qRT-PCR and western blotting. ACKNOWLEDGEMENTS PLKO vectors containing short-hairpin RNA targeting SOX15 were KLT, NR and LAP are supported by Vanier Canada Graduate Scholarships, DDBS by purchased from Open Biosystems (Waltham, MA, USA) (catalog RHS4533- University of British Columbia 4YF scholarship and WWL by a CIHR Jean-Francois NM_006942). Individual lentiviruses were prepared in 293T cells co- Saint Denis Fellowship in Cancer Research. This work was supported by grants from transfected with one of the four PLKO plasmids coding a SOX15 short- the Canadian Cancer Society Research Institute (#700809), the Canadian Institutes for hairpin RNA and the viral packaging plasmids VSVG and d8.91 using Health Research, PMH Foundation and Ontario Ministry of Health and Long Term TransIT-LT1 transfection reagent (Mirus, Madison, WI, USA). Virus contain- Care. MST is the M Qasim Choksi Chair in Lung Cancer Translational Research. ing empty PLKO vector served as a control. HPDE cells were infected with 2 ml of virus and selected with puromycin at 0.8 mg/ml, producing stable HPDE-PLKO and HPDE-shSOX15 cells. RNA was extracted from each line REFERENCES and the short-hairpin RNA producing the greatest SOX15 knockdown was used to assess cell viability. SOX15 expression was confirmed by qRT-PCR 1 Jemal A, Bray F, Center MM, Ferlay J, Ward E, Forman D. Global cancer statistics. and western blotting. CA Cancer J Clin 2011; 61: 69–90. 2 Bardeesy N, DePinho RA. Pancreatic cancer biology and genetics. Nat Rev Cancer 2002; 2: 897–909. Western blotting 3 Hezel AF, Kimmelman AC, Stanger BZ, Bardeesy N, Depinho RA. Genetics and Protein lysates were collected and western blots were performed as biology of pancreatic ductal adenocarcinoma. Genes Dev 2006; 20: 1218–1249. previously described.70 SOX15 was assessed using a polycloncal sheep 4 Singh P, Srinivasan R, Wig JD. Major molecular markers in pancreatic ductal anti-human antibody (AF4070, R&D Systems, Minneapolis, MN, USA). Active adenocarcinoma and their roles in screening, diagnosis, prognosis, and treatment. b-catenin was measured with a mouse monoclonal antibody specific for Pancreas 2011; 40: 644–652. dephosphorylated (at residues Ser37 or Thr41) b-catenin (#05-665, Millipore, 5 Samuel N, Hudson TJ. The molecular and cellular heterogeneity of pancreatic Billerica, MA, USA).31,41 ACTB and GAPDH (#3700 and #2118, Cell Signaling, ductal adenocarcinoma. Nat Rev Gastroenterol Hepatol 2012; 9: 77–87. Danvers, MA, USA) were used as loading controls. Secondary antibodies 6 Jones S, Zhang X, Parsons DW, Lin JC, Leary RJ, Angenendt P et al. Core signaling included: antisheep (HAF016, R&D Systems), antirabbit and antimouse pathways in human pancreatic cancers revealed by global genomic analyses. (#7074 and #7076, Cell Signaling). Band intensities were quantified using Science 2008; 321: 1801–1806. Adobe Photoshop CS4 (Adobe Systems, San Jose, CA, USA). 7 Mann KM, Ward JM, Yew CC, Kovochich A, Dawson DW, Black MA et al. Sleeping Beauty mutagenesis reveals cooperating mutations and pathways in pancreatic adenocarcinoma. Proceedings of the National Academy of Sciences of the United Cell viability assays States of America 2012; 109: 5934–5941. MTT assays were used to assess cell viability in stable Capan2-SOX15 and 8 Jiao Y, Shi C, Edil BH, de Wilde RF, Klimstra DS, Maitra A et al. DAXX/ATRX, MEN1, Capan2-EV cells as previously described.71 Briefly, 3000 cells were seeded and mTOR pathway genes are frequently altered in pancreatic neuroendocrine in triplicate in 96-well plates and viability was assessed for 5 consecutive tumors. Science 2011; 331: 1199–1203. days by the addition of MTT reagent (Invitrogen), followed by lysis with 9 Wu J, Jiao Y, Dal Molin M, Maitra A, de Wilde RF, Wood LD et al. 20% SDS and quantification of absorbance with a spectrophotometer. Whole-exome sequencing of neoplastic cysts of the pancreas reveals recurrent Experiments were repeated twice. mutations in components of ubiquitin-dependent pathways. Proceedings of the National Academy of Sciences of the United States of America 2011; 108: 21188–21193. In vivo mouse models 10 Birnbaum DJ, Adelaide J, Mamessier E, Finetti P, Lagarde A, Monges G et al. To assess the effect of SOX15 expression on tumor formation in vivo, Genome profiling of pancreatic adenocarcinoma. Genes Chromosomes Cancer Capan2-SOX15 and Capan2-EV cells were introduced into 4–6 week old 2011; 50: 456–465.

Oncogene (2014) 279 – 288 & 2014 Macmillan Publishers Limited SOX15 inactivation in pancreatic cancer KL Thu et al 287 11 Joergensen MT, Geisz A, Brusgaard K, Schaffalitzky de Muckadell OB, Hegyi P, 36 Tennis MA, Vanscoyk MM, Wilson LA, Kelley N, Winn RA. Methylation of Wnt7a is Gerdes AM et al. Intragenic duplication: a novel mutational mechanism in her- modulated by dnmt1 and cigarette smoke condensate in non-small cell lung editary pancreatitis. Pancreas 2011; 40: 540–546. cancer. PloS one 2012; 7: e32921. 12 Yachida S, Jones S, Bozic I, Antal T, Leary R, Fu B et al. Distant metastasis 37 Rubin JS, Barshishat-Kupper M, Feroze-Merzoug F, Xi ZF. Secreted WNT occurs late during the genetic evolution of pancreatic cancer. Nature 2010; 467: antagonists as tumor suppressors: pro and con. Front Biosci 2006; 11: 1114–1117. 2093–2105. 13 Berger AH, Knudson AG, Pandolfi PP. A continuum model for tumour suppression. 38 Qin YR, Tang H, Xie F, Liu H, Zhu Y, Ai J et al. Characterization of tumor-sup- Nature 2011; 476: 163–169. pressive function of SOX6 in human esophageal squamous cell carcinoma. Clinic 14 Bernard P, Harley VR. Acquisition of SOX transcription factor specificity through Cancer Res 2011; 17: 46–55. protein-protein interaction, modulation of Wnt signalling and post-translational 39 Ye YW, Wu JH, Wang CM, Zhou Y, Du CY, Zheng BQ et al. Sox17 regulates modification. Int J Biochem Cell Biol 2010; 42: 400–410. proliferation and cell cycle during gastric cancer progression. Cancer Lett 2011; 15 Tan AC, Jimeno A, Lin SH, Wheelhouse J, Chan F, Solomon A et al. Characterizing 307: 124–131. DNA methylation patterns in pancreatic cancer genome. Mol Oncol 2009; 3: 40 Kormish JD, Sinner D, Zorn AM. Interactions between SOX factors and Wnt/beta- 425–438. catenin signaling in development and disease. Dev Dyn 2010; 239: 56–68. 16 Vincent A, Omura N, Hong SM, Jaffe A, Eshleman J, Goggins M. 41 van Noort M, Meeldijk J, van der Zee R, Destree O, Clevers H. Wnt signaling Genome-wide analysis of promoter methylation associated with gene controls the phosphorylation status of beta-catenin. J Biol Chem 2002; 277: expression profile in pancreatic adenocarcinoma. Clinic Cancer Res 2011; 17: 17901–17905. 4341–4354. 42 Lochhead PA, Gilley R, Cook SJ. ERK5 and its role in tumour development. 17 Hong SM, Omura N, Vincent A, Li A, Knight S, Yu J et al. Genome-wide CpG island Biochem Soc Trans 2012; 40: 251–256. profiling of intraductal papillary mucinous neoplasms of the pancreas. 43 Drew BA, Burow ME, Beckman BS. MEK5/ERK5 pathway: the first fifteen years. Clinic Cancer Res 2012; 18: 700–712. Biochim Biophys Acta 2012; 1825: 37–48. 18 Du YC, Oshima H, Oguma K, Kitamura T, Itadani H, Fujimura T et al. Induction and 44 Medina PP, Castillo SD, Blanco S, Sanz-Garcia M, Largo C, Alvarez S et al. The SRY- down-regulation of Sox17 and its possible roles during the course of gastro- HMG box gene, SOX4, is a target of gene amplification at chromosome 6p in lung intestinal tumorigenesis. Gastroenterology 2009; 137: 1346–1357. cancer. Hum Mol Genet 2009; 18: 1343–1352. 19 Zhang W, Glockner SC, Guo M, Machida EO, Wang DH, Easwaran H et al. Epige- 45 Liu P, Ramachandran S, Ali Seyed M, Scharer CD, Laycock N, Dalton WB et al. Sex- netic inactivation of the canonical Wnt antagonist SRY-box containing gene 17 in determining region Y box 4 is a transforming oncogene in human prostate cancer colorectal cancer. Cancer Res 2008; 68: 2764–2772. cells. Cancer Res 2006; 66: 4011–4019. 20 Jia Y, Yang Y, Liu S, Herman JG, Lu F, Guo M. SOX17 antagonizes WNT/beta- 46 Bass AJ, Watanabe H, Mermel CH, Yu S, Perner S, Verhaak RG et al. SOX2 is an catenin signaling pathway in hepatocellular carcinoma. Epigenetics 2010; 5: amplified lineage-survival oncogene in lung and esophageal squamous cell car- 743–749. cinomas. Nat Genet 2009; 41: 1238–1242. 21 Sakai M, Hibi K, Koshikawa K, Inoue S, Takeda S, Kaneko T et al. Frequent promoter 47 Gustavsson E, Sernbo S, Andersson E, Brennan DJ, Dictor M, Jerkeman M et al. methylation and gene silencing of CDH13 in pancreatic cancer. Cancer Sci 2004; SOX11 expression correlates to promoter methylation and regulates tumor 95: 588–591. growth in hematopoietic malignancies. Mol Cancer 2010; 9:187. 22 Andreeva AV, Kutuzov MA. Cadherin 13 in cancer. Genes Chromosomes Cancer 48 Forbes SA, Bindal N, Bamford S, Cole C, Kok CY, Beare D et al. COSMIC: mining 2010; 49: 775–790. complete cancer genomes in the catalogue of somatic mutations in cancer. 23 Bachman KE, Herman JG, Corn PG, Merlo A, Costello JF, Cavenee WK Nucleic Acids Res 2011; 39: Database issue D945–D950. et al. Methylation-associated silencing of the tissue inhibitor of metalloprotei- 49 Harada T, Chelala C, Bhakta V, Chaplin T, Caulee K, Baril P et al. Genome-wide DNA nase-3 gene suggest a suppressor role in kidney, brain, and other human cancers. copy number analysis in pancreatic cancer using high-density single nucleotide Cancer Res 1999; 59: 798–802. polymorphism arrays. Oncogene 2008; 27: 1951–1960. 24 MacGrogan D, Levy A, Bova GS, Isaacs WB, Bookstein R. Structure and methyla- 50 Cerami E, Gao J, Dogrusoz U, Gross BE, Sumer SO, Aksoy BA et al. The cBio cancer tion-associated silencing of a gene within a homozygously deleted region of genomics portal: an open platform for exploring multidimensional cancer human chromosome band 8p22. Genomics 1996; 35: 55–65. genomics data. Cancer discov 2012; 2: 401–404. 25 McCleary-Wheeler AL, McWilliams R, Fernandez-Zapico ME. Aberrant signaling 51 Yan HT, Shinka T, Sato Y, Yang XJ, Chen G, Sakamoto K et al. Overexpression of pathways in pancreatic cancer: a two compartment view. Mol Carcinog 2012; 51: SOX15 inhibits proliferation of NT2/D1 cells derived from a testicular embryonal 25–39. cell carcinoma. Mol Cells 2007; 24: 323–328. 26 Morris JPt, Wang SC, Hebrok M. KRAS, Hedgehog, Wnt and the twisted devel- 52 Beranger F, Mejean C, Moniot B, Berta P, Vandromme M. Muscle differentiation is opmental biology of pancreatic ductal adenocarcinoma. Nat Rev Cancer 2010; 10: antagonized by SOX15, a new member of the SOX protein family. J Biol Chem 683–695. 2000; 275: 16103–16109. 27 Pasca di Magliano M, Biankin AV, Heiser PW, Cano DA, Gutierrez PJ, Deramaudt T 53 Lee HJ, Goring W, Ochs M, Muhlfeld C, Steding G, Paprotta I et al. Sox15 is et al. Common activation of canonical Wnt signaling in pancreatic adenocarci- required for skeletal muscle regeneration. Mol Cell Biol 2004; 24: 8428–8436. noma. PloS one 2007; 2: e1155. 54 Meeson AP, Shi X, Alexander MS, Williams RS, Allen RE, Jiang N et al. Sox15 and 28 Suzuki H, Watkins DN, Jair KW, Schuebel KE, Markowitz SD, Chen WD et al. Epi- Fhl3 transcriptionally coactivate Foxk1 and regulate myogenic progenitor cells. genetic inactivation of SFRP genes allows constitutive WNT signaling in colorectal EMBO J 2007; 26: 1902–1912. cancer. Nat Genet 2004; 36: 417–422. 55 Savage J, Conley AJ, Blais A, Skerjanc IS. SOX15 and SOX7 differentially regulate 29 Suzuki H, Toyota M, Carraway H, Gabrielson E, Ohmura T, Fujikane T et al. the myogenic program in P19 cells. Stem Cells 2009; 27: 1231–1243. Frequent epigenetic inactivation of Wnt antagonist genes in breast cancer. 56 Ito M. Function and molecular evolution of mammalian Sox15, a singleton Br J Cancer 2008; 98: 1147–1156. in the SoxG group of transcription factors. Int J Biochem Cell Biol 2010; 42: 30 Kongkham PN, Northcott PA, Croul SE, Smith CA, Taylor MD, Rutka JT. The SFRP 449–452. family of WNT inhibitors function as novel tumor suppressor genes epigenetically 57 White BD, Chien AJ, Dawson DW. Dysregulation of Wnt/beta-catenin signaling in silenced in medulloblastoma. Oncogene 2010; 29: 3017–3024. gastrointestinal cancers. Gastroenterology 2012; 142: 219–232. 31 Licchesi JD, Westra WH, Hooker CM, Machida EO, Baylin SB, Herman JG. Epige- 58 Yao H, Ashihara E, Maekawa T. Targeting the Wnt/beta-catenin signaling pathway netic alteration of Wnt pathway antagonists in progressive glandular neoplasia of in human cancers. Expert Opin Ther Targets 2011; 15: 873–887. the lung. Carcinogenesis 2008; 29: 895–904. 59 Polakis P. Drugging Wnt signalling in cancer. EMBO J 2012; 31: 2737–2746. 32 Gumz ML, Zou H, Kreinest PA, Childs AC, Belmonte LS, LeGrand SN et al. Secreted 60 Klarmann GJ, Decker A, Farrar WL. Epigenetic gene silencing in the Wnt pathway frizzled-related protein 1 loss contributes to tumor phenotype of clear cell renal in breast cancer. Epigenetics 2008; 3: 59–63. cell carcinoma. Clin Cancer Res 2007; 13: 4740–4749. 61 Dihlmann S, von Knebel Doeberitz M. Wnt/beta-catenin-pathway as a molecular 33 Rajan N, Burn J, Langtry J, Sieber-Blum M, Lord CJ, Ashworth A. Transition from target for future anti-cancer therapeutics. Int J Cancer 2005; 113: 515–524. cylindroma to spiradenoma in CYLD-defective tumours is associated with reduced 62 Moon RT, Kohn AD, De Ferrari GV, Kaykas A. WNT and beta-catenin signalling: DKK2 expression. J Pathol 2011; 224: 309–321. diseases and therapies. Nat Rev Genet 2004; 5: 691–701. 34 Sato H, Suzuki H, Toyota M, Nojima M, Maruyama R, Sasaki S et al. Frequent 63 Bibikova M, Lin Z, Zhou L, Chudin E, Garcia EW, Wu B et al. High-throughput epigenetic inactivation of DICKKOPF family genes in human gastrointestinal DNA methylation profiling using universal bead arrays. Genome Res 2006; 16: tumors. Carcinogenesis 2007; 28: 2459–2466. 383–393. 35 Tennis MA, Vanscoyk M, Freeman S, Winn RA. Promoter hypermethylation leads 64 Ishkanian AS, Malloff CA, Watson SK, DeLeeuw RJ, Chi B, Coe BP et al. A tiling to loss of wnt7a in non-small cell lung cancer. Proceedings of the American resolution DNA microarray with complete coverage of the human genome. Nat Thoracic Society 2012; 9: 83–84. Genet 2004; 36: 299–303.

& 2014 Macmillan Publishers Limited Oncogene (2014) 279 – 288 SOX15 inactivation in pancreatic cancer KL Thu et al 288 65 Khojasteh M, Lam WL, Ward RK, MacAulay C. A stepwise framework for the 69 Dull T, Zufferey R, Kelly M, Mandel RJ, Nguyen M, Trono D et al. A third-generation normalization of array CGH data. BMC bioinformatics 2005; 6: 274. lentivirus vector with a conditional packaging system. J Virol 1998; 72: 8463–8471. 66 Coe BP, Chari R, MacAulay C, Lam WL. FACADE: a fast and sensitive algorithm for 70 Thu KL, Pikor LA, Chari R, Wilson IM, Macaulay CE, English JC et al. Genetic the segmentation and calling of high resolution array CGH data. Nucleic Acids Res disruption of KEAP1/CUL3 E3 ubiquitin ligase complex components is a key 2010; 38: e157. mechanism of NF-kappaB pathway activation in lung cancer. J Thorac Oncol 2011; 67 Simon R, Lam A, Li MC, Ngan M, Menenzes S, Zhao Y. Analysis of gene expression 6: 1521–1529. data using BRB-ArrayTools. Cancer Inform 2007; 3: 11–17. 71 Lockwood WW, Thu KL, Lin L, Pikor LA, Chari R, Lam WL et al. Integrative genomics 68 Mak AB, Ni Z, Hewel JA, Chen GI, Zhong G, Karamboulas K et al. A lentiviral identified RFC3 as an amplified candidate oncogene in esophageal adenocarci- functional proteomics approach identifies chromatin remodeling complexes noma. Clin Cancer Res 2012; 18: 1936–1946. important for the induction of pluripotency. Mol Cel Proteomics 2010; 9: 72 Radulovich N, Qian JY, Tsao MS. Human pancreatic duct epithelial cell model for 811–823. KRAS transformation. Methods Enzymol 2008; 439: 1–13.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

Oncogene (2014) 279 – 288 & 2014 Macmillan Publishers Limited