A COL11A1-Correlated Pan-Cancer Gene Signature Of

ARTICLE IN PRESS

Cancer Letters ■■ (2016) ■■–■■

Contents lists available at ScienceDirect

Cancer Letters

journal homepage: www.elsevier.com/locate/canlet

1 Q2 Original Articles 2 3 A COL11A1-correlated pan-cancer gene signature of activated 4 ﬁbroblasts for the prioritization of therapeutic targets 5 6 Q1 Dongyu Jia a, Zhenqiu Liu b, Nan Deng b, Tuan Zea Tan c, Ruby Yun-Ju Huang c, 7 Barbie Taylor-Harding a, Dong-Joo Cheon d, Kate Lawrenson a, Wolf R. Wiedemeyer a, 8 Ann E. Walts e, Beth Y. Karlan a,f, Sandra Orsulic a,f,*

9 a Women’s Cancer Program, Samuel Oschin Comprehensive Cancer Institute, Cedars-Sinai Medical Center, Los Angeles, CA, USA 10 b Biostatistics and Bioinformatics Research Center, Cedars-Sinai Medical Center, Los Angeles, CA, USA 11 c Cancer Science Institute of Singapore, Center for Translational Medicine, National University of Singapore, Singapore 12 d Center for Cell Biology and Cancer Research, Albany Medical College, Albany, NY, USA 13 e Department of Pathology and Laboratory Medicine, Cedars-Sinai Medical Center, Los Angeles, CA, USA 14 f Department of Obstetrics and Gynecology, David Geffen School of Medicine, University of California at Los Angeles, Los Angeles, CA, USA 15 16 17 ARTICLE INFO ABSTRACT 18 2019 Article history: Although cancer-associated fibroblasts (CAFs) are viewed as a promising therapeutic target, the design 2221 Received 10 July 2016 of rational therapy has been hampered by two key obstacles. First, attempts to ablate CAFs have re- 2423 Received in revised form 30 August 2016 sulted in significant toxicity because currently used biomarkers cannot effectively distinguish activated 25 Accepted 1 September 2016 26 CAFs from non-cancer associated fibroblasts and mesenchymal progenitor cells. Second, it is unclear whether 2827 CAFs in different organs have different molecular and functional properties that necessitate organ- 3029 Keywords: specific therapeutic designs. Our analyses uncovered COL11A1 as a highly specific biomarker of activated 31 Cancer-associated fibroblasts 32 CAFs. Using COL11A1 as a ‘seed’, we identified co-expressed genes in 13 types of primary carcinoma in 33 Myofibroblasts 3534 Pan-cancer The Cancer Genome Atlas. We demonstrated that a molecular signature of activated CAFs is conserved 3736 Therapeutic targets in epithelial cancers regardless of organ site and transforming events within cancer cells, suggesting that 3938 Tumor microenvironment targeting fibroblast activation should be effective in multiple cancers. We prioritized several potential 40 pan-cancer therapeutic targets that are likely to have high specificity for activated CAFs and minimal tox- 41 icity in normal tissues. 42 © 2016 The Author(s). Published by Elsevier Ireland Ltd. This is an open access article under the CC BY- 43 NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

45 Introduction nisms by which activated CAFs contribute to such diverse aspects 66 46 of cancer progression are unclear, it is thought that fibroblasts 67 47 Under normal physiological conditions, collagen-rich fibro- together with increased collagen deposition and altered extracel- 68 48 blasts maintain tissue architecture and serve as a barrier to epithelial lular matrix (ECM) remodeling serve as a rich depot of cancer- 69 49 cell migration. However, cancer cells have the ability to convert promoting growth factors, cytokines, and chemokines [1,2]. 70 50 the surrounding fibroblasts into activated CAFs, which secrete spe- Additionally, altered levels of enzymes responsible for collagen 71 51 cific collagens, growth factors, and enzymes that promote cancer cross-link formation, such as lysyl oxidase (LOX) [4], increase tissue 72 52 growth, angiogenesis, invasion, and metastasis [1–3]. At the same stiffness and modify mechanotransduction resulting in the reorga- 73 53 time, these CAFs suppress anticancer immunity, confer drug resis- nization of loose connective tissue into tense linear tracks of fibers 74 54 tance and/or limit the access of chemotherapies, anti-angiogenic that serve as highways to promote chemotaxis of cancer cells 75 55 therapies, and immunotherapies [1–3]. Although the exact mecha- [5,6]. 76 56 Recognizing the crucial role of CAFs in most aspects of cancer 77 progression, it has been proposed that rational anticancer therapy 78 57 58 Abbreviations: αSMA, α-smooth muscle actin; CAFs, cancer-associated fibro- design should not only target the cancer cells but also the CAFs [7,8]. 79 59 blasts; CSPG4, chondroitin sulfate proteoglycan 4; ECM, extracellular matrix; LOX, Unlike cancer cells, CAFs are genetically stable [9], which reduce the 80 60 lysyl oxidase; EMT, epithelial–mesenchymal transition; FAP, fibroblast activation risk of therapy-induced clonal selection, resistance, and cancer re- 81 61 protein; PALLD, palladin; PDGFRα, platelet-derived growth factor receptor α; PDPN, currence. Furthermore, targeting CAFs could potentially affect 82 62 podoplanin; TGFβ, transforming growth factor β; TNC, tenascin-C; PRECOG, PREdiction multiple biochemical pathways to prevent cancer progression 83 63 of Clinical Outcomes from Genomic Profiles. 64 * Corresponding author. Fax: +1 310 423 9537. and recurrence. CAF-targeting therapeutic approaches in different 84 65 E-mail address: [email protected] (S. Orsulic). experimental mouse cancer models have been shown to improve 85

http://dx.doi.org/10.1016/j.canlet.2016.09.001 0304-3835/© 2016 The Author(s). Published by Elsevier Ireland Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/ by-nc-nd/4.0/).

2 D. Jia et al. / Cancer Letters ■■ (2016) ■■–■■

86 tumoral immune response, intratumoral drug delivery, and thera- Materials and methods 152 87 peutic efficacy [10–15]. These studies confirm the key role of CAFs Human tissues 153 88 in cancer progression and demonstrate their effectiveness as a ther- 89 apeutic target. However, targeting CAFs in some cancer models Studies involving human tissue samples were approved by the Cedars-Sinai In- 154 90 actually promoted cancer progression. For example, depletion of stitutional Review Board (IRB 15425). The samples included a tissue microarray from 155 91 αSMA+ stroma in a mouse pancreatic cancer model resulted in in- 42 patients with matched primary, metastatic, and recurrent ovarian cancer. 156 92 creased cancer aggressiveness, enhanced hypoxia and epithelial– In situ hybridization 157 93 mesenchymal transition (EMT), suppressed anticancer immunity, 94 and reduced survival [16]. The RNA hybridization kit (RNAscope 2.0 FFPE Assay) and probes for COL11A1, 158 95 The contradictory results in different cancer models could be ex- the bacterial gene dapB (negative control), and the housekeeping gene HPRT (pos- 159 96 plained by different roles of CAFs in different cancer types, i.e. CAFs itive control), were from Advanced Cell Diagnostics, Inc. Formalin-fixed, paraffin- 160 97 could be promoting breast cancer and inhibiting pancreatic cancer. embedded tissue section slides were processed by the Cedars-Sinai Biobank and 161 Translational Research Core following the protocol provided with the RNAscope In 162 98 Alternatively, in all cancer types CAFs prevent cancer progression Situ Hybridization kit from Advanced Cell Diagnostics, Inc. The slides were coun- 163 99 until they receive activating signals from cancer cells and convert terstained with Mayer’s hematoxylin. 164 100 into ‘activated CAFs’, which in turn confer invasive and metastatic 101 abilities upon cancer cells [7]. Therapies that target all CAFs are coun- Immunohistochemistry 165 102 terproductive and likely to result in the death of normal fibroblasts Immunohistochemical detection of αSMA was performed on formalin-fixed, 166 103 and significant toxicity. Preferential targeting of activated CAFs has paraffin-embedded tissue sections using the protocol provided with the pre- 167 104 been challenging because activated CAFs are poorly understood at diluted asm-1 clone antibody from Leica Microsystems. Staining was done by the 168 105 the molecular level. During activation, CAFs exhibit phenotypic Cedars-Sinai Pathology Service on the Ventana Benchmark Ultra automated slide 169 106 changes that partially overlap with myofibroblastic changes during stainer. The staining was visualized using the Ventana OptiView DAB Detection System. 170 The slides were counterstained with Mayer’s hematoxylin. 171 107 wound healing, inflammation, and fibrosis, including secretion of

108 specific ECM components, cytokines and growth factors [1,17]. In vitro co-culture experiments 172 109 Several markers have been used to distinguish activated from non- 110 activated CAFs: α-smooth muscle actin (αSMA, encoded by gene The ovarian cancer cell lines OVCAR3-GFP, KURAMOCHI-GFP and OVSAHO- 173 111 ACTA2) [3], fibroblast activation protein (FAP) [12], podoplanin GFP were maintained in RPMI-1640 (Corning) supplemented with 10% fetal bovine 174 serum (FBS) and 1x penicillin/streptomycin (Corning). Cell line authenticity was con- 175 112 (PDPN) [18], palladin (PALLD) [19,20], tenascin-C (TNC) [21], platelet- firmed by Laragen using the short tandem repeat (STR) method. The immortalized 176 113 derived growth factor receptor α (PDGFRα) [22,23], and chondroitin normal ovarian fibroblasts INOF-tdTomato cell lines [34] were maintained in a 1:1 177 114 sulfate proteoglycan 4 (CSPG4) [24]. However, these markers are fre- ratio of MCDB 105 (Sigma-Aldrich) and Medium 199 (GIBCO) with 10% FBS, 50 U/ 178 115 quently expressed in other cells within the cancer stroma, such as ml penicillin and 50 μg/ml streptomycin. Immortalized normal ovarian fibroblasts 179 116 vascular smooth muscle cells, pericytes, and mesenchymal stem cells. and ovarian cancer cells were co-cultured in 1% FBS supplemented media (1:1:2 ratio 180 of MCDB 105, Medium 199 and RPMI-1640) using 6-well plates, either by directly 181 This lack of specificity could pose problems in therapeutic target- 117 plating ovarian cancer cells (105 cells/well) onto a 70% confluent layer of normal 182 118 ing and underscores the need to better understand the molecular ovarian fibroblasts or onto a 0.4 μm Transwell membrane. Media were replaced every 183 119 characteristics of activated CAFs in order to develop more precise 2 days. After 4 days of co-culture, GFP-labeled ovarian cancer cells and tdTomato- 184 120 and less toxic targeted therapies. labeled fibroblasts were separated by FACS in PBS with 0.5% FBS. RNA extraction from 185 fibroblasts and ovarian cancer cell lines was performed using the RNeasy Mini Kit 186 121 COL11A1 encodes the α1 chain of collagen XI, a minor fibrillar (Qiagen) and reverse transcribed to cDNA using the Quantitect Reverse Transcrip- 187 122 collagen expressed by chondrocytes and osteoblasts but not qui- tion Kit (Qiagen). For qRT-PCR, 50 ng of cDNA was mixed with COL11A1 primers 188 123 escent fibroblasts [25,26]. The absence of functional collagen XI leads (Forward: 5′-GACTATCCCCTCTTCAGAACTGTTAAC-3′; Reverse: 5′- CTTCTATCAAGTGG 189 124 to abnormally thickened cartilage and tendon fibrils, suggesting the TTTCGTGGTTT-3′) and the iQ SYBR-Green Supermix (BioRad) and run on the CFX96 190 -ΔCT 125 role of collagen XI in maintaining proper fibril diameter [27,28]. Real-Time System (BioRad). Data were analyzed using the 2 method and nor- 191 malized to INOF-tdTomato control to present the fold change ratios. All mRNA data 192 126 Studies have demonstrated that COL11A1 mRNA is markedly el- were normalized to RPL32 expression (Forward: 5′-ACAAAGCACATGCTGCCCAGTG- 193 127 evated in cancers of the oral cavity/pharynx, head and neck, breast, 3′; Reverse: 5′-TTCCACGATGGCTTTGCGGTTC-3′). The statistical analyses were 194 128 lung, esophagus, stomach, pancreas, colon, and ovary, but not in performed using GraphPad Prism (version 6.0; GraphPad Software). The unpaired 195 129 matched normal tissues (reviewed in [25,26]). COL11A1 has been t test was used for data analyses. 196 130 identified as part of gene signatures associated with adverse clin- Public database portals and dataset analyses 197 131 ical outcomes including resistance to neoadjuvant therapy in breast 132 cancer [29], time to recurrence in glioblastoma [30], poor survival Data from public portals were used as provided by individual portals without 198 133 in kidney cancer [31], and time to recurrence and overall survival additional processing or normalization, unless otherwise indicated. Box plots of 199 134 in ovarian cancer [32,33]. In situ hybridization in ovarian cancer COL11A1 expression in normal tissues and cancers were generated using the Gene 200 Expression across Normal and Tumor tissue (GENT) portal (medical- 201 135 and immunohistochemistry in pancreatic cancer revealed that genome.kribb.re.kr/GENT) in which data from multiple datasets were processed and 202 136 COL11A1 mRNA and pro-protein are primarily expressed in CAFs normalized as previously described [35]. COL11A1 expression level diagrams for in- 203 137 [25,32]. The restricted expression of COL11A1 in normal tissues and flammatory bowel disease, lung fibrosis, and cancers of the colon and lung were 204 138 its enrichment in CAFs during cancer progression combined with generated using the R2 MegaSampler public portal (hgserver1.amc.nl/cgi-bin/r2/ 205 139 its association with adverse clinical outcomes in multiple types of main.cgi). A description of the methods used for data processing and normalization 206 is available through the portal. Survival z-scores for individual genes and cancer types 207 140 cancer support its candidacy as a specific marker of fibroblast ac- were obtained from the PREdiction of Clinical Outcomes from Genomic Profiles 208 141 tivation in diverse cancers. Here, we explore the suitability of COL11A1 (PRECOG) portal (precog.stanford.edu). Methods for calculating PRECOG z-scores have Q4 209 142 as a pan-cancer marker of activated CAFs and use it as a ‘seed’ to been published [36]. Ranking of the COL11A1-correlated genes in 13 TCGA carcino- 210 143 identify the transcription signature of activated CAFs in 13 epithe- ma types was determined using data from individual cancer datasets that were 211 144 lial cancer types. We show that the COL11A1-coexpressed gene set processed by cBio Portal (cbioportal.org) as previously described [37]. Kaplan– 212 Meier survival plots and plots of COL11A1 expression in individual molecular subtypes 213 145 is highly conserved in these 13 cancer types, indicating that the fi- of ovarian carcinoma were generated using the ovarian cancer microarray gene ex- 214 146 broblast reaction to cancer cells is independent of the organ site- pression database CSIOVDB (csibio.nus.edu.sg/CSIOVDB/CSIOVDB.html), which has 215 147 of-origin and of the transforming events within cancer cells. Finally, been previously described [38]. The dataset for fibroblast and ovarian epithelial cell 216 148 by combining drug target databases with cancer vs. normal tissue co-culture was imported from the Gene Expression Omnibus (ncbi.nlm.nih.gov/ 217 geo). The Euclidean distance clustering analysis heatmap for the e-mtab-991 [39] 218 149 expression databases, we identify several potential therapeutic and GSE40595 [40] datasets was generated using the public R2 GeneSet Clustering 219 150 targets that should have high specificity for activated CAFs and Analysis portal (hgserver1.amc.nl/cgi-bin/r2/main.cgi), which also describes methods 220 151 minimal toxicity in normal tissues. that were used to process and normalize data from datasets included in the portal. 221

D. Jia et al. / Cancer Letters ■■ (2016) ■■–■■ 3

222 Results In sections of metastatic ovarian cancer, we observed that 249 223 COL11A1-positive cells are confined to the intratumoral and imme- 250 224 COL11A1 is expressed in a subset of αSMA-positive CAFs and can be diate peritumoral CAFs (Fig. 1B), suggesting that COL11A1 expression 251 225 induced in normal fibroblasts by the presence of cancer cells may be induced by cues received from epithelial cancer cells. To test 252 226 if cancer cells can induce COL11A1 expression in fibroblasts, we 253 227 To determine if COL11A1 expression is associated with fibro- co-cultured immortalized normal ovarian fibroblasts (INOFs) with 254 228 blast activation, we used αSMA as a marker of activated CAFs [3]. three different ovarian cancer cell lines (OVSAHO, OVCAR3, and 255 229 Comparison of αSMA immunohistochemistry and COL11A1 in situ KURAMOCHI). COL11A1 expression in INOFs was most strongly 256 230 hybridization in a tissue microarray consisting of primary, meta- induced by direct co-culture with ovarian cancer cell lines al- 257 231 static and recurrent ovarian cancers from 42 patients showed that though weak induction occurred by indirect co-culture on a 258 232 COL11A1 is expressed in a subset of αSMA+ CAFs (Fig. 1A). Unlike Transwell membrane (Fig. 1C). The induction of COL11A1 in fibro- 259 233 αSMA, COL11A1 was not expressed in blood vessels (red arrows) or blasts in the presence of cancer cells was confirmed by analysis 260 234 in fibroblasts surrounding the cancer (blue arrows) (Fig. 1A). of the public expression dataset GSE52104 in which two types of 261

A α-SMA IHC COL11A1 ISH

100 μm

BV 235 BV NC

1 mm 1 mm 100 μm

B Fat C D 0%

Peritumoral ~1% stroma

Tumor ~50% Intratumoral 100μm stroma Cultured alone Cultured alone Co-culture (Transwell) Transwell co-culture with normal ovarian epithelial cells Co-culture (direct contact) Transwell co-culture with ovarian carcinoma cell line

236 Fig. 1. COL11A1 is expressed in CAFs. (A) Comparison of αSMA immunohistochemistry and COL11A1 in situ hybridization in a metastatic ovarian cancer sample. Red arrows 237 indicate blood vessels. Blue arrows indicate fibroblasts surrounding the tumor nodule. A high magnification of peritumoral and intratumoral regions in the black rectangles 238 is shown in panels on the left. IHC, immunohistochemistry; ISH, in situ hybridization; BV, blood vessel; NC, necrosis. (B) Distribution of in situ hybridization COL11A1- 239 positive CAFs in relation to cancer cells. The estimated percent of COL11A1-positive CAFs is shown on the right. The image is representative of metastatic and recurrent 240 ovarian cancer samples, which typically express higher levels of COL11A1 than primary ovarian cancers. (C) Quantitative RT-PCR levels of COL11A1 in sorted (FACS) immor- 241 talized normal ovarian fibroblasts (INOFs) grown alone or co-cultured with ovarian cancer cell lines (OVSAHO, OVCAR3, KURAMOCHI) that were either separated from INOFs 242 by a Transwell membrane or directly mixed with INOFs. Statistical analyses were performed between INOFs grown alone and INOFs co-cultured with ovarian cancer cells 243 (*p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant). Error bars indicate standard deviation. (D) Levels of COL11A1 in the GSE52104 expression dataset in which mesen- 244 chymal stem cells (MSCs) or immortalized normal ovarian fibroblasts (INOFs) were either cultured alone or co-cultured with IOSE4 normal epithelial cells (IOSE) or HEYA8 245 epithelial ovarian cancer cells (EOC) using a Transwell membrane. Inverse-log2 values of the Robust Multi-array Average (RMA) scores from different COL11A1 probes were 246 averaged, then log2-transformed. The data were extracted for statistical analyses using GraphPad Prism 6. Data are represented as the mean ± SEM. Intergroup differences 247 were assessed by the Student’s t-test. *p < 0.05; ****p < 0.0001. (For interpretation of the references to color in this figure legend, the reader is referred to the web version 248 of this article.)

4 D. Jia et al. / Cancer Letters ■■ (2016) ■■–■■

262 presumptive cancer-associated fibroblast precursor cells, mesen- adverse outcomes is also not restricted to ovarian cancer. We show 295 263 chymal stem cells (MSCs) and immortalized normal ovarian that in 1820 colon cancers [42], increased expression of COL11A1 296 264 fibroblasts (INOFs), were either cultured alone or co-cultured with mRNA is associated with poor disease-specific and disease-free sur- 297 265 normal ovarian surface epithelial cells (IOSE) or epithelial ovarian vival as well as with clinical and molecular parameters, such as 298 266 cancer cells (EOC) using a Transwell membrane [41]. COL11A1 mRNA increased cancer stage and microsatellite instability and CMS4 (mes- 299 267 was statistically significantly upregulated when MSCs and INOFs enchymal) molecular subtype (Fig. S1A, B). 300 268 were co-cultured with EOC but not IOSE (Fig. 1D), indicating that To systematically investigate an association between COL11A1 301 269 cancer cells have a greater capacity than normal cells to induce mRNA expression and survival in various solid and liquid cancers, 302 270 COL11A1 expression in fibroblasts. we plotted COL11A1 z-score values as determined by the pan- 303 271 cancer PREdiction of Clinical Outcomes from Genomic Profiles 304 272 COL11A1 is associated with cancer progression and poor survival (PRECOG) analysis of ~18000 cases in 166 cancer datasets [36].In 305 273 most epithelial cancers, COL11A1 expression was associated with 306 274 COL11A1 mRNA expression has been associated with poor sur- poor survival (Fig. 3A). Associations between expression of 43 col- 307 275 vival in ovarian cancer [32,33] and kidney cancer [31]. To elucidate lagen genes and survival z-scores in 12 common epithelial cancer 308 276 the underlying biology that could result in poor survival, we inves- types revealed that for the majority of collagens, increased expres- 309 277 tigated its expression in ovarian and colon cancers. Using a sions of mRNA were associated with poor survival, with COL11A1 310 278 comprehensively annotated microarray database for 3431 human having the strongest association (Fig. 3B). 311 279 ovarian cancers [38], we show that increased expression of COL11A1 312 280 mRNA is associated with overall survival and disease-free survival COL11A1 is among the most differentially expressed genes between 313 281 (Fig. 2A) as well as with clinical and molecular parameters such as cancers and corresponding benign tissues 314 282 increased cancer stage and grade and mesenchymal molecular 315 283 subtype (Fig. 2B). The association of COL11A1 expression with poor In the Genotype-Tissue Expression (GTEx) project database [43], 316 284 survival is unlikely to be a manifestation of the total amount of COL11A1 mRNA is expressed at appreciable levels in transformed 317 285 stromal fibroblasts because a general marker of fibroblasts, vimentin skin fibroblasts but not in non-transformed skin fibroblasts or other 318 286 (VIM), is not associated with poor survival in the same cohort of normal tissues (Fig. S2). Additional analyses of various expression 319 287 ovarian cancer patients (Table S1). The association of COL11A1 with datasets containing normal adult mouse and human tissues 320

Overall Survival Overall Survival Disease-Free Survival Disease-Free Survival A (n = 1868) Q1 vs Q4 (n = 934) (n = 1516) Q1 vs Q4 (n = 758)

COL11A1 - low COL11A1 - Q1 COL11A1 - low COL11A1 - Q1 COL11A1 - high COL11A1 - Q4 COL11A1 - high COL11A1 - Q4 Percent Survival Percent Survival Percent Survival Percent Survival

Month Month Month Month HR = 0.7613 (0.6720 - 0.8624) HR = 0.6453 (0.5422 - 0.7680) HR = 0.7739 (0.6850 - 0.8743) HR = 0.5909 (0.4966 - 0.7033) Log-rank p = 0.0001 Log-rank p = 0.0001 Log-rank p = 0.0001 Log-rank p = 0.0001

COL11A1 low high COL11A1 Q1 Q4 COL11A1 low high COL11A1 Q1 Q4 Median Survival 52.77 42.17 Median Survival 53.17 36.67 Median Survival 25.00 18.10 Median Survival 27.20 14.67

288 B Tumor Stage Tumor Grade Molecular Subtype 11.2 11.2 11.2

9.8 9.8 9.8

8.4 8.4 8.4

7 7 7

5.6 5.6 5.6 COL11A1 Expression COL11A1 COL11A1 Expression COL11A1 COL11A1 Expression COL11A1

I II III IV G1 G2 G3 Epi-A Epi-B Mes Stem-A Stem-B Mann Mann Mann Whitney I II III IV Rest Whitney Grade 1 Grade 2 Grade 3 Rest Whitney Epi-A Epi-B Mes Stem-A Stem-B Test Test Test I 2.41E-01 6.96E-21 2.03E-13 2.74E-19 Grade 1 4.03E-04 1.31E-06 8.86E-06 Epi-A 1.15E-12 6.06E-120 7.33E-28 2.04E-01 II 4.19E-08 7.99E-07 2.42E-06 Grade 2 3.83E-02 3.97E-01 III 3.00E-01 9.35E-09 Grade 3 2.46E-05 Epi-B 2.47E-135 2.01E-06 1.52E-09 IV 5.32E-03 Mes 1.59E-85 4.31E-102 Stem-A 1.72E-23 Stem-B

289 Fig. 2. COL11A1 expression is associated with adverse clinical parameters. (A) Kaplan–Meier analyses of overall survival (left two panels) and disease-free survival (right 290 two panels) based on COL11A1 expression in ovarian carcinoma. Disease-free survival includes progression- and recurrence-free survival. Patients were stratified to COL11A1- 291 high (red) or COL11A1-low (blue) based on the median expression of COL11A1, and to COL11A1-Q4 (highest 25% expression; red) or COL11A1-Q1 (lowest 25% expression; 292 blue). (B) COL11A1 expression profiles were stratified by FIGO stage (left), FIGO grade (middle), and molecular subtype (right). Data were obtained from the ovarian microarray 293 gene expression database CSIOVDB (csibio.nus.edu.sg/CSIOVDB/CSIOVDB.html). HR, hazard ratio. (For interpretation of the references to color in this figure legend, the reader 294 is referred to the web version of this article.)

etr 21) doi: (2016), Letters 10.1016/j.canlet.2016.09.001

o fteaetctres Cancer targets, therapeutic of ion prioritizat the for ﬁbroblasts activated of signature gene pan-cancer COL11A1-correlated A al., et Jia, Dongyu as: press in article this cite Please

322 Q7 321 RAepeso fdfeetclae ee.Tedt eeotie rmtePEito fCiia ucmsfo eoi rﬁe PEO)database (PRECOG) Proﬁles Genomic from Outcomes Clinical of PREdiction the from obtained were data The genes. collagen different of expression mRNA i.3. Fig. COL11A1 A xrsini acri soitdwt orsria nmlil acrtps A uvvlzsoe ndfeetcne ye soitdwt expr with associated types cancer different in z-scores Survival (A) types. cancer multiple in survival poor with associated is cancer in expression B COL4A4 COL19A1 COL28A1 COL4A6 COL27A1 COL24A1 COL9A2 COL8A2 COL17A1 COL9A1 COL23A1 COL6A6 COL25A1 COL20A1 COL16A1 COL2A1 COL6A1 COL13A1 COL8A1 COL15A1 COL9A3 COL11A2 COL22A1 COL12A1 COL6A2 COL18A1 COL7A1 COL10A1 COL5A3 COL5A1 COL6A3 COL4A2 COL3A1 COL1A2 COL5A2 COL1A1 COL4A1 COL11A1 COL4A3 COL21A1 COL6A5 COL4A5 COL14A1 ( precog.stanford.edu sinof ession .5-. 32-. .704 .8-. .929 16-1.9 -1.6 2.94 0.29 0.09 -0.9 -1.1 1.28 2.91 0.44 2.18 -1.6 1.67 1.01 -0.4 -0.3 1.71 -0.5 -0.7 -3.2 - 0.73 -2.6 -0.4 -1.9 0.02 0.25 - 0.27 0.39 1.96 -0.3 1.78 0.74 1.71 -2.2 0.22 -0 1.02 - 3.09 -1.2 0.62 0.32 -2 0.59 -0.3 0.31 -3.5 -0.1 0.88 2.59 - 0.38 0.47 -3.6 -0.8 0.18 1.63 0.25 - 0.07 - 0.89 -1.1 -0.4 1.12 -0.3 4.54 -0.4 -0.9 -0.8 0.27 -1.1 -0.8 2.02 0.53 -0 0.75 1.08 0.14 -0 0.56 0.27 0.62 2.18 1.26 0.89 0.32 2.32 0.09 -0.6 -0.6 -1.1 -2.7 -0.2 -1.8 2.21 0.67 -0.2 -0.2 -0.3 0.87 1.74 1.63 2.16 2.18 0.66 2.16 4 1.07 -0.7 2.23 -0.7 2.23 0.64 -1.4 1.15 2.66 0.47 0.41 -0.7 0.43 - 3.53 -0.8 0.02 0.45 2.35 3.67 - 0.48 0.56 -0.4 1.57 1.58 0.58 - 0.4 0.98 1.75 1.24 2.15 -0.5 -0.2 2.17 -0.9 1.34 -2.5 0.31 1.69 - 1.88 0.31 1.06 1.39 1.21 2.38 3.29 5.33 1.97 1.67 2.39 0.58 0.03 3.29 0.38 1.9 0.09 -1.6 1.66 3.65 -0.8 2.22 1.56 0.18 -0.2 - -1.3 3.68 2.51 0.75 3.07 1.57 2.72 -0.2 2.4 0.31 -0.2 3.18 1.42 2.06 1.07 -0.9 4.49 0.93 3.69 1.64 2.11 -0.6 2.81 3.79 2.33 3.32 1.93 3.63 0.63 2.34 -0.2 0.35 0.3 3.14 2.31 1.96 3.77 2.4 2.87 2.83 0.85 0.4 -1 0.65 -0.4 1.63 1.28 5.26 3.29 4.62 2.3 4.87 0.99 2.13 -0.4 1.9 0.26 2.09 2.41 4.15 3.93 0.57 -0.8 2.63 0.88 0.54 2.42 3.49 0.68 0.93 3.13 1.66 -0.2 3.26 1.53 1.5 2.44 2.75 0.55 5.58 0.71 -0.5 3.35 1.56 3.47 1.54 4.85 1.41 0.66 2.34 3.59 1.71 0.86 3.08 2.38 0.03 2.04 2.02 1106 0407 1524 .7-. 1900 .4-0.2 -0.8 1.24 -0.8 0.01 -1.3 -1.9 -0.2 0.06 -5.8 0.49 -1.2 0.53 0.77 0.52 0.61 2.42 -1.2 -1.5 - - 0.22 0.75 0.05 -0.1 -0.4 -2.7 0.19 0.67 - 1.85 -0.7 -1.1 0.45 -1.2 0.42 0.47 0.25 -0.3 0.5 1.08 -0.8 -0.1 -2.2 -0.9 -0.2 0.71 1.89 0.68 -1.1 2.05 -0.9 -0.6 -0.4 0.59 0.5 - -0.6 0.5 1.5 0.34 1.25 0.9 -0.3 0.61 - -0.9 -1.6 -0.3 0.94 0.7 2.01 1.74 0.53 1.85 0.89 -2.9 1.87 1.35 1.56 -0.4 -0.6 2.24 -0.4 -1.1 0.65 1.61 1.77 0.08 0.61 3.35 1.85 -0.9 3.31 -2.6 -0.3 1.66 -0.3 5.4 -1.3 0.83 3.44 1.39 2.75 2.69 -1.6 1.72 -0.6 2.82 0.68 2.57 2.41 3.05 2.64 0.51 2.94 2.4 5.25 2.48 -0.2 08-. 16-. 1-03 5514 2115 -6.5 1.57 -2.1 1.46 -1.8 -5.5 -1.6 0.36 -0.4 - -1.3 -1.8 -2.5 -0.5 -1 -0.8 0.05 1.68 0.12 -0.5 0.03 -0.4 -2.3 -1.6 0.75 -1.8 1.63 -0.1 -1.8 -0.3 -2 -0.8 -2.1 2.33 0.06 0.63 1.3 -1.3 0.77 0.08 -0.1 -5.9 -0 -0.3 -4.3 -1.1 0.93 -1.2 . 1115 .318 01-. 202 .1-. 2.05 -1.5 0.91 0.21 -2 -0.5 -0.1 1.86 0.53 1.56 -1.1 0.6 -0.4 -0.2 0.3 -0.9 2.51 2.01 -0.4 1.04 1.3 0.23 2.83 0.5 1-. 2700 .9--01 1234 10.87 -1 3.47 -1.2 0.19 - - 1.09 0.03 -2.7 -0.1 -1 1-. 1-. .1---. 2121 0.56 - 2.16 -2.1 -3.3 - - 0.21 -0.7 -1 -0.6 -1 .706 . .210 .113 0121 . -1 0.2 2.11 -0.1 1.32 0.91 1.06 3.12 1.3 0.68 2.37 1 .202 01--07 0803 .6-. 1.49 -0.1 1.96 0.37 -0.8 0.79 - - -0.1 0.27 2.32 - Adrenocortical COL11A1 Bladder Good survival survival GoodPoor survival ). Breast

RA B uvvlzsoe soitdwith associated z-scores Survival (B) mRNA. Colon Head and neck Kidney Liver Lung (adeno) Melanoma Ovarian Pancreatic

Prostate

– (2016) Letters Cancer / al. et Jia D. 5 ■■ ■■ ■■ RIL NPRESS IN ARTICLE ARTICLE IN PRESS

6 D. Jia et al. / Cancer Letters ■■ (2016) ■■–■■

323 revealed that COL11A1 is expressed in cartilage and collagen- cancer COL11A1-correlated genes in epithelial cells. This is relevant 389 324 producing cells in the eye and brain, with negligible levels in most because several of the 195 pan-cancer COL11A1-correlated genes have 390 325 other tissues that have been profiled including mesenchymal stem been shown to play a role in EMT [42] and malignant cells under- 391 326 cells in the bone marrow, muscle, and fat (data not shown). going EMT have been proposed as one possible source of CAFs [44]. 392 327 Comparison of COL11A1 expression in 17931 cancers and 3503 To determine if the pan-cancer COL11A1-correlated gene set is pref- 393 328 normal tissues (U133Plus2 platform) and 9258 cancers and 4087 erentially expressed in cancer cells undergoing EMT or in host- 394 329 normal tissues (U133A platform) using the Gene Expression across derived fibroblasts, we used the e-mtab-991 public transcription 395 330 Normal and Tumor tissue (GENT) portal [35] revealed that COL11A1 profile dataset of primary patient-derived colon cancers and their 396 331 mRNA is elevated in most cancers in comparison to their corre- patient-derived xenografts (PDX) in nude mice [39]. Presumably, in 397 332 sponding normal tissues (Fig. 4A). In some cancers, COL11A1 was PDX samples, fast-proliferating human cancer cells continued to grow 398 333 ranked among the most statistically significant differentially ex- in mice while slow-proliferating human CAFs were lost and even- 399 334 pressed genes when cancer and its corresponding normal tissue were tually replaced by mouse fibroblasts, which can be distinguished 400 335 compared. For example, comparison of cancer and normal tissues from human cells by species-specific gene probes [39]. GeneSet clus- 401 336 in The Cancer Genome Atlas (TCGA) datasets for colon cancer and tering analysis showed that most of the pan-cancer COL11A1- 402 337 invasive breast cancer ranked COL11A1 as the first and third most correlated genes had diminished levels in PDX samples in comparison 403 338 differentially expressed gene, respectively (Fig. S3). to primary cancers (Fig. 5A), suggesting that the genes are en- 404 339 As many collagens and collagen-remodeling genes are fre- riched in the CAFs rather than in the cancer cells. However, it is also 405 340 quently upregulated in fibroblast activation associated with possible that the pan-cancer COL11A1-correlated genes are ex- 406 341 inflammation and fibrosis in the absence of cancer, use of these genes pressed in epithelial cells in primary colon tumors but become 407 342 as therapeutic targets in cancer could be problematic. Analysis of silenced upon adaptation of human cancer cells to the mouse mi- 408 343 expression profile datasets show that levels of COL11A1 mRNA in croenvironment. Thus, we conducted GeneSet clustering analysis 409 344 inflamed colonic tissue from inflammatory bowel disease and fi- of the primary ovarian cancer dataset GSE40595 in which ovarian 410 345 brotic lung tissues are not significantly different from those in CAFs and epithelial cancer cells were isolated by laser-capture mi- 411 346 corresponding unaffected colon and lung and that COL11A1 levels crodissection [40]. The pan-cancer COL11A1-correlated genes were 412 347 associated with colon inflammation and lung fibrosis are minimal preferentially expressed in CAFs in this dataset (Fig. 5B). 413 348 and markedly different from those associated with colon and lung In addition to CAFs, immune cells are a major component of the 414 349 cancers (Fig. 4B). In contrast, levels of ACTA2, the gene encoding the tissue microenvironment. To exclude the possibility that the pan- 415 350 prototypical marker of myofibroblast differentiation, αSMA, is ex- cancer COL11A1-correlated gene set represents immune cells in the 416 351 pressed at similar levels in cancers and inflamed or fibrotic tissues tumor microenvironment, we used the expression profile of 230 417 352 (Fig. 4B). mouse hematopoietic cell types generated by the Immunological 418 353 Genome Project (ImmGen) compendium [45]. In addition to he- 419 354 A consistent set of genes is co-expressed with COL11A1 across matopoietic cell lineages, the dataset contains expression profiles 420 355 different cancers of skin fibroblasts and fibroblasts residing in the thymus, lymph 421 356 nodes, and spleen. The pan-cancer COL11A1-correlated gene set was 422 357 To better understand the biology of cancers with high levels of highly represented in fibroblasts but not in hematopoietic cell lin- 423 358 COL11A1, we identified genes that most closely correlate with eages (Fig. 5C). 424 359 COL11A1 mRNA expression in 13 TCGA datasets representing dif- CAFs have a different expression profile than normal fibro- 425 360 ferent cancer types. Spearman’s rank correlations between COL11A1 blasts. Moffitt and colleagues defined a 23-gene signature of ‘normal 426 361 and its co-expressed genes for each cancer type were calculated. stroma’ and a 25-gene signature of ‘activated stroma’ [46] using non- 427 362 The genes were then ranked based on the average correlation of each negative matrix factorization for virtual microdissection of primary 428 363 gene across the 13 cancer types. The top 195 correlated genes were and metastatic pancreatic ductal cancer samples into cell subsets 429 364 selected based on an average correlation of >0.4 COL11A1-correlated with prognostic and biologic relevance. None of the 23 (0%) ‘normal 430 365 genes were then ranked based on the average of the absolute cor- stroma’ genes in contrast to 18 of 25 (72%) ‘activated stroma’ genes 431 366 relation values (Table 1 and Table S2). The top 10% most highly were present in the COL11A1-correlated gene set, respectively 432 367 correlated genes in each cancer type are highlighted in pink (Table 1). (Fig. 5D), suggesting that the COL11A1-correlated gene set repre- 433 368 Notably, COL11A1-correlated genes with high average correlation sents CAFs. 434 369 scores also tended to be among the top 10% highest scored genes 435 370 in each cancer type (indicated in pink in Table 1). In contrast, the Biological processes associated with fibroblast activation in cancer 436 371 top 10% COL11A1-anticorrelated genes were not conserved across 437 372 these cancer types (Table S3). Some of the top ranked COL11A1- The remarkable uniformity of COL11A1-correlated genes across 438 373 anticorrelated genes in individual cancer types were associated with 13 different cancer types suggests involvement of these genes in 439 374 normal functions of these organs suggesting that they may repre- common biological processes that are independent of the organ site 440 375 sent normal tissue or a noninvasive tumor component. For example, and of the phenotypic and genetic diversity observed in individu- 441 376 the ovarian cancer top 100 COL11A1-anticorrelated genes (Table S3) al cancer types. To gain insight into the biology of this conserved 442 377 present in the GSE12172 ovarian cancer dataset were primarily ex- gene set, we conducted several analyses that identified overlap 443 378 pressed in ovarian tumors of low malignant potential (Fig. S4). between the 195 pan-cancer COL11A1-correlated genes and genes 444 379 in various datasets with characterized biological features. The Gene 445 380 Pan-cancer COL11A1-correlated genes are induced in CAFs Ontology (GO) Biological Process (BP) analysis revealed that the pan- 446 381 cancer COL11A1-correlated genes are primarily involved in 447 382 Consistent with the induced expression of COL11A1 in the in vitro extracellular matrix modification and collagen remodeling (Table S4). 448 383 co-culture model (Fig. 1D), the average expression of the pan- Additionally, we used SABiosciences array gene tables, which consist 449 384 cancer COL11A1-correlated gene set was significantly induced in of literature-based curated molecular pathways where each pathway 450 385 mesenchymal stem cells and normal ovarian fibroblasts co-cultured was represented by 84 genes. Analysis of gene overlap between 451 386 with ovarian cancer cells but not with normal ovarian epithelial cells the pan-cancer COL11A1-correlated gene set and genes represen- 452 387 (Fig. S5). Since epithelial cells were not profiled in this experi- tative of 67 different pathways in SABiosciences arrays showed the 453 388 ment, it is unknown if fibroblasts also induce expression of the pan- largest overlap for pathways associated with extracellular matrix, 454

Please cite this article in press as: Dongyu Jia, et al., A COL11A1-correlated pan-cancer gene signature of activated ﬁbroblasts for the prioritization of therapeutic targets, Cancer Letters (2016), doi: 10.1016/j.canlet.2016.09.001 laect hsatcei rs s oguJa ta. O1A-orltdpncne eesgaueo ciae bolssfrteprioritizat the for ﬁbroblasts activated of signature gene pan-cancer COL11A1-correlated A al., et Jia, Dongyu as: doi: press (2016), in Letters article this cite Please U133A A 14 B COL11A1 (37892_at) ACTA2 (200974_at) 14 14

12 12 12

10 10 10.1016/j.canlet.2016.09.001 10 8 8 log2 log2 6 6 8

log2 4 4

6 2 2

0 0 4 (232) (133) (133) (155)

2 RIL NPRESS IN ARTICLE GSE10616 (58) GSE10616 GSE8671 (32) GSE8671 GSE38713 (43) GSE38713 (148) GSE21510 GSE14333 (290) GSE14333 GSE17538 GSE16879 (26) GSE9452 (133) GSE41568 GSE38713 (43) GSE38713 GSE2109 (315) GSE2109 (32) GSE8671 GSE16879 (315) GSE2109 (290) GSE14333 GSE13294 (148) GSE21510 GSE13294 (155) GSE10616 (58) GSE10616 (26) GSE9452 GSE17538 (232) (133) GSE41568 N N C C N N -C - -N -C - -C -N -C -C -N -N .Jae l acrLetters Cancer / al. et Jia D. COLON datasets Skin -C Skin Skin -N Skin Liver -N Liver -C Lung Lung -N Lung Brain Brain Brain -C Brain Heart -N Blood -C Blood -N Colon Colon Colon Colon Ovary -N Ovary Testis -C Testis -N Cervix -N Cervix Cervix Cervix Breast - Breast -C

Uterus Uterus Normal Kidney -C Kidney Kidney -N Kidney Muscle -N Muscle Thyroid - Thyroid -N Thyroid Tongue -N Tongue Bladder -N Bladder -C Adipose -N Prostate -N Prostate -C Stomach -C Stomach Stomach Pancreas Pancreas Pancreas - Pancreas Inflammation + normal Esophagus Esophagus -N Esophagus

Myometrium Myometrium Cancer Endometrium -N Endometrium Small intestine -N Head and neck neck -C Head and neck - Head and U133Plus2 COL11A1 (37892_at) ACTA2 (200974_at) 15 14 14 ■■ 12 12 (2016) 10 10

8 8 ■■ 10 – log2 log2 6 6 ■■

4 4 log2 2 2 5 0 0

0 GSE21369 (29) GSE21369 (100) GSE33532 GSE43580 (150) GSE43580 (23) GSE24206 GSE24206 (23) GSE24206 (121) GSE2109 (114) GSE3141 (100) GSE33532 (29) GSE21369 (150) GSE43580 (121) GSE2109 (114) GSE3141 o fteaetctres Cancer targets, therapeutic of ion N C N C N N N N C C N C N N C N C N C C C N C C C N C C N N C N C N C N N ------C -C - - - -C ------LUNG datasets Eye Eye Skin -C Skin -N Liver - Liver Lung Lung Brain -N GIST -C GIST Brain - Vulva - Vulva -N Blood Blood Colon -C Colon -N Ovary Ovary Testis -C Testis Ovary -N -N Ovary Testis Testis Cervix -C Cervix - Breast Breast Breast - Breast Uterus - Uterus - Kidney Kidney Spleen Vagina -C Vagina - Muscle Muscle Thyroid -C -C Thyroid -N Thyroid Tongue -N Bladder Bladder - Adipose Adipose Prostate -C Prostate -N Stomach -N Abdomin - Stomach Sarcoma Sarcoma Fibrosis + normal Pancreas -C Pancreas

Esophagus Esophagus Esophagus Cancer Endometrium - Endometrium Endometrium - Endometrium Adrenal gland - Adrenal gland Small intestine intestine Small Small intestine intestine Small Head and neck Head and neck

Fig. 4. Increased expression of COL11A1 in cancer and low expression in normal tissues, inflammation and fibrosis. (A) Comparison of COL11A1 mRNA expression in normal tissues and corresponding cancers. Box plots for two different platforms (U133Plus2 and U133A) were generated using datasets and software available through the Gene Expression across Normal and Tumor tissue (GENT) portal (medical-genome.kribb.re.kr/GENT). The y axis shows log2 mRNA levels. Average expression levels in normal tissues and cancer tissues are indicated by vertical dotted green and red lines, respectively. (B) COL11A1 and ACTA2 mRNA expression in normal, inflammatory and fibrotic conditions in comparison to cancer. The graphs were generated using the public R2 MegaSampler software (hgserver1.amc.nl/cgi-bin/r2/main.cgi) for the processing and normalization of individual datasets imported from the Gene Expression Omnibus (u133p2, MAS5.0 platform). The number of samples in each GSE dataset is indicated in parentheses. (For interpretation of the references to color in this figure legend, the reader is referred

to the web version of this article.) 7 460 459 458 457 456 455 ARTICLE IN PRESS

8 D. Jia et al. / Cancer Letters ■■ (2016) ■■–■■

461 Table 1 462 COL11A1-correlated genes (Spearman’s rank correlation) across 13 different TCGA carcinoma types, each represented by >100 primary tumors from therapy-naive patients. Q8 464463 Pink denotes the top 10% COL11A1-correlated genes in each individual carcinoma type. Rectangles denote genes frequently used as markers of activated CAFs. 465 Pan-cancer COL11A1-correlated genes Continued

llary o cell amous eck mous clear cell bol d n ion papillarydeno ey an ch id d and neck der st rectal ical a tate rostate lo ney clear ng s ma Rank Gene symbol BladderBreast Colorectal CervicalHea Kidn KidneyLung papi adenLung squOvarianP Stomach ThyroidAverageCorrelation ank erv idney u o R Gene sym Blad Brea Co C Head Kid K L Lung squaOvarianPro St ThyroAverageCorrelat 1 COL11A1 101 PODN 0.57 0.37 0.41 0.43 0.62 0.63 0.52 0.36 0.65 0.65 0.15 0.03 0.65 0.465 2 COL1A1 0.87 0.69 0.76 0.72 0.87 0.72 0.67 0.8 0.86 0.83 0.63 0.74 0.85 0.77 102 CD248 0.66 0.25 0.63 0.46 0.64 0.38 0.5 0.51 0.67 0.68 0.04 0.42 0.2 0.465 3 COL1A2 0.85 0.74 0.86 0.71 0.87 0.69 0.58 0.8 0.88 0.8 0.51 0.73 0.75 0.752 103 PLXDC1 0.52 0.36 0.5 0.36 0.59 0.42 0.44 0.3 0.67 0.73 0.48 0.31 0.36 0.465 4 COL3A1 0.85 0.7 0.88 0.64 0.85 0.68 0.63 0.83 0.86 0.85 0.54 0.66 0.79 0.751 104 MMP13 0.22 0.72 0.56 0.23 0.42 0.49 0.26 0.47 0.45 0.68 0.14 0.62 0.77 0.464 5 FAP 0.77 0.72 0.83 0.55 0.69 0.75 0.63 0.77 0.77 0.85 0.59 0.73 0.84 0.73 105 COL16A1 0.65 0.28 0.56 0.21 0.38 0.58 0.48 0.54 0.63 0.64 0.2 0.3 0.57 0.463 6 COL5A1 0.83 0.75 0.82 0.59 0.74 0.71 0.65 0.75 0.82 0.82 0.45 0.66 0.75 0.718 106 CPZ 0.35 0.3 0.61 0.36 0.63 0.54 0.46 0.58 0.69 0.25 0.37 0.44 0.44 0.463 7 CTHRC1 0.8 0.68 0.87 0.37 0.81 0.55 0.54 0.81 0.74 0.69 0.61 0.77 0.8 0.695 107 VGLL3 0.14 0.54 0.78 0.43 0.58 0.53 0.56 0.43 0.65 0.68 -0.2 0.39 0.5 0.463 8 SULF1 0.85 0.76 0.88 0.63 0.79 0.57 0.54 0.84 0.75 0.62 0.47 0.76 0.55 0.693 108 COL15A1 0.56 0.33 0.71 0.45 0.42 0.43 0.51 0.67 0.7 0.48 0.39 0.22 0.13 0.462 9 VCAN 0.76 0.72 0.87 0.67 0.83 0.42 0.2 0.79 0.81 0.83 0.54 0.62 0.84 0.685 109 SGIP1 0.6 0.73 0.69 0.39 0.7 0.24 0.38 0.47 0.72 0.57 0.35 0.3 -0.2 0.461 10 FN1 0.74 0.84 0.69 0.71 0.78 0.56 0.42 0.66 0.78 0.82 0.43 0.61 0.66 0.669 110 CMTM3 0.37 0.52 0.77 0.42 0.6 0.48 0.14 0.54 0.41 0.48 0.17 0.54 0.54 0.46 11 SFRP2 0.82 0.54 0.73 0.57 0.79 0.68 0.54 0.81 0.79 0.81 0.32 0.46 0.84 0.669 111 LOXL1 0.33 0.46 0.55 0.26 0.61 0.61 0.17 0.54 0.58 0.63 0.32 0.29 0.63 0.46 12 OLFML2B 0.78 0.65 0.7 0.59 0.79 0.56 0.57 0.74 0.71 0.77 0.48 0.56 0.66 0.658 112 NALCN 0.37 0.42 0.49 0.32 0.59 0.54 0.51 0.37 0.53 0.44 0.57 0.34 0.48 0.459 13 COL6A3 0.79 0.66 0.82 0.56 0.78 0.69 0.58 0.66 0.81 0.76 0.23 0.52 0.7 0.658 113 COL24A1 0.35 0.46 0.72 0.32 0.74 0.39 0.36 0.4 0.4 0.44 0.28 0.48 0.62 0.458 14 INHBA 0.63 0.78 0.76 0.53 0.37 0.57 0.53 0.77 0.73 0.87 0.52 0.75 0.73 0.657 114 PLAU 0.46 0.8 0.54 0.19 0.29 0.53 0.09 0.57 0.35 0.75 0.14 0.63 0.62 0.458 15 ASPN 0.82 0.71 0.79 0.4 0.75 0.54 0.49 0.71 0.77 0.8 0.61 0.46 0.68 0.656 115 MFAP2 0.37 0.46 0.59 0.36 0.53 0.65 0.35 0.58 0.4 0.36 0.23 0.62 0.43 0.456 16 ADAMTS12 0.76 0.69 0.8 0.61 0.73 0.52 0.47 0.69 0.85 0.8 0.28 0.73 0.53 0.651 116 LTBP2 0.56 0.31 0.53 0.41 0.45 0.51 0.59 0.24 0.57 0.47 0.37 0.32 0.59 0.455 17 LUM 0.73 0.76 0.81 0.46 0.68 0.7 0.51 0.65 0.65 0.85 0.26 0.52 0.84 0.648 117 MATN3 0.38 0.4 0.49 0.37 0.59 0.46 0.15 0.38 0.51 0.56 0.56 0.46 0.61 0.455 18 NTM 0.78 0.74 0.88 0.68 0.8 0.5 0.51 0.58 0.73 0.84 0.4 0.69 0.27 0.646 118 THBS1 0.63 0.52 0.44 0.45 0.34 0.2 0.32 0.54 0.56 0.71 0.16 0.37 0.67 0.455 19 SPARC 0.77 0.69 0.84 0.66 0.83 0.42 0.25 0.73 0.84 0.81 0.51 0.68 0.34 0.644 119 CTGF 0.63 0.36 0.49 0.4 0.71 0.51 0.5 0.39 0.63 0.51 0.21 0.38 0.13 0.45 20 AEBP1 0.81 0.71 0.76 0.54 0.81 0.46 0.39 0.68 0.85 0.78 0.39 0.48 0.66 0.64 120 DPYSL3 0.59 0.65 0.7 0.38 0.39 0.52 0.61 0.54 0.38 0.47 0 0.19 0.43 0.45 21 CDH11 0.74 0.74 0.76 0.52 0.73 0.53 0.55 0.51 0.75 0.7 0.49 0.55 0.75 0.64 121 ARSI 0.69 0.56 0.65 0.23 0.22 0.51 0.14 0.6 0.51 0.24 0.17 0.59 0.73 0.449 22 GREM1 0.74 0.62 0.74 0.56 0.67 0.58 0.54 0.81 0.79 0.79 0.46 0.27 0.71 0.637 122 ADAMTS16 0.69 0.59 0.69 0.57 0.74 -0.1 -0.2 0.59 0.53 0.45 0.41 0.54 0.31 0.448 23 ITGBL1 0.67 0.64 0.71 0.63 0.65 0.55 0.6 0.51 0.67 0.77 0.69 0.42 0.76 0.636 123 GRP 0.48 0.53 0.61 0.51 0.65 0.29 0.33 0.38 0.52 0.54 -0.1 0.31 0.72 0.447 24 FNDC1 0.87 0.72 0.79 0.55 0.72 0.61 0.49 0.74 0.89 0.61 0.61 0.63 0.03 0.635 124 RAB31 0.61 0.58 0.84 0.19 0.21 0.53 0.02 0.4 0.46 0.62 0.31 0.43 0.61 0.447 25 PRRX1 0.79 0.67 0.78 0.47 0.68 0.63 0.55 0.76 0.69 0.79 0.17 0.58 0.61 0.628 125 HTRA1 0.64 0.55 0.69 0.46 0.46 0.51 0.23 0.59 0.69 0.42 0.16 0.3 0.1 0.446 26 MMP11 0.7 0.71 0.58 0.51 0.78 0.48 0.09 0.71 0.69 0.81 0.51 0.76 0.81 0.626 126 CLMP 0.59 0.45 0.73 0.38 0.35 0.58 0.13 0.61 0.73 0.73 0.06 0.09 0.37 0.446 27 FBN1 0.74 0.75 0.88 0.62 0.72 0.58 0.37 0.67 0.85 0.82 0.35 0.45 0.26 0.62 127 IGFL2 0.54 0.53 0.3 0.35 0.3 0.51 0.31 0.4 0.51 0.71 0.17 0.56 0.59 0.445 28 CTSK 0.78 0.65 0.77 0.46 0.78 0.55 0.11 0.75 0.8 0.81 0.35 0.51 0.69 0.616 128 BICC1 0.46 0.53 0.68 0.4 0.63 -0 0.04 0.5 0.84 0.67 0.15 0.35 0.54 0.443 29 COL8A1 0.65 0.75 0.83 0.65 0.7 0.38 0.43 0.52 0.79 0.77 0.46 0.45 0.48 0.605 129 PDGFRL 0.58 0.49 0.56 0.3 0.76 0.54 0.44 0.36 0.7 0.35 0.32 0.28 0.07 0.442 30 MMP2 0.73 0.59 0.79 0.51 0.68 0.61 0.42 0.65 0.83 0.8 0.22 0.41 0.62 0.605 130 FAM101A 0.72 0.42 -0.1 0.41 0.71 0.69 0.5 0.65 0.68 0.68 -0.2 -0.2 0.7 0.442 31 WISP1 0.76 0.66 0.72 0.45 0.63 0.4 0.48 0.73 0.71 0.71 0.24 0.58 0.77 0.603 131 SGCD 0.74 0.58 0.79 0.49 0.7 0.44 0 0.56 0.83 0.34 0.09 0.16 0.01 0.441 32 COMP 0.66 0.62 0.61 0.52 0.58 0.62 0.62 0.59 0.59 0.57 0.5 0.5 0.73 0.593 132 CSMD2 0.71 0.68 0.54 0.4 0.64 0.07 0.43 0.66 0.73 0.39 0.16 0.48 -0.2 0.441 33 SFRP4 0.8 0.39 0.75 0.23 0.74 0.43 0.37 0.7 0.65 0.64 0.68 0.53 0.77 0.591 133 LAMP5 0.56 0.31 0.51 0.21 0.68 0.6 0.41 0.34 0.61 0.08 0.47 0.27 0.67 0.44 34 HTRA3 0.82 0.6 0.64 0.57 0.7 0.41 0.53 0.8 0.75 0.62 0.24 0.52 0.4 0.585 134 SERPINH1 0.52 0.42 0.44 0.37 0.46 0.52 0.23 0.55 0.39 0.46 0.27 0.65 0.41 0.438 35 EPYC 0.77 0.7 0.73 0.41 0.57 0.23 0.24 0.79 0.65 0.79 0.24 0.74 0.72 0.583 135 PDGFRA 0.48 0.38 0.41 0.33 0.53 0.58 0.51 0.58 0.6 0.48 0.17 0 0.64 0.438 36 PCOLCE 0.75 0.51 0.73 0.48 0.8 0.63 0.3 0.63 0.76 0.7 0.18 0.47 0.61 0.581 136 BNC2 0.66 0.64 0.72 0.44 0.72 0.27 -0.1 0.6 0.8 0.27 0.12 0.16 0.34 0.437 37 PODNL1 0.76 0.52 0.5 0.4 0.53 0.71 0.58 0.56 0.7 0.48 0.34 0.66 0.79 0.579 137 COL4A1 0.42 0.39 0.62 0.55 0.51 0.43 0.46 0.4 0.58 0.64 0.4 0.29 -0 0.437 38 ANTXR1 0.72 0.79 0.88 0.47 0.66 0.51 0.25 0.67 0.79 0.68 0.32 0.59 0.18 0.578 138 IBSP 0.45 0.42 0.5 0.43 0.56 0.42 0.44 0.5 0.28 0.28 0.16 0.6 0.61 0.435 39 ZNF469 0.78 0.48 0.72 0.59 0.76 0.46 0.35 0.53 0.75 0.68 0.12 0.61 0.66 0.576 139 ITGA5 0.59 0.52 0.71 0.42 0.34 0.54 0.41 0.59 0.47 0.69 -0.1 0.3 0.12 0.434 40 CORIN 0.55 0.8 0.75 0.55 0.73 0.47 0.21 0.67 0.82 0.67 0.11 0.51 0.64 0.575 140 MMP16 0.45 0.36 0.55 0.51 0.69 0.45 0.41 0.42 0.62 0.25 0.2 0.3 0.43 0.434 41 THY1 0.72 0.64 0.73 0.64 0.8 0.38 0.39 0.75 0.81 0.44 0.49 0.59 0.1 0.575 141 ST6GALNAC5 0.72 0.13 0.8 0.54 0.49 0.47 -0.1 0.55 0.58 0.19 0.25 0.38 0.59 0.434 42 MFAP5 0.75 0.77 0.72 0.55 0.5 0.54 0.34 0.81 0.67 0.52 0.26 0.22 0.77 0.571 142 ADAM19 0.58 0.48 0.41 0.34 0.38 0.56 0.44 0.4 0.49 0.57 0.14 0.28 0.56 0.433 43 GLT8D2 0.7 0.63 0.76 0.53 0.74 0.55 0.51 0.76 0.82 0.7 0.27 0.31 0.08 0.566 143 MEIS3 0.48 0.17 0.64 0.48 0.63 0.54 0.16 0.75 0.64 0.36 -0.1 0.4 0.52 0.433 44 ISLR 0.8 0.49 0.68 0.45 0.81 0.61 0.54 0.78 0.76 0.48 0.24 0.47 0.23 0.565 144 APCDD1L 0.61 0.31 0.45 0.37 0.28 0.65 0.47 0.74 0.57 0.26 0 0.17 0.75 0.433 45 COL6A2 0.75 0.49 0.59 0.55 0.78 0.57 0.46 0.64 0.77 0.71 0.05 0.27 0.58 0.555 145 XIRP1 0.66 0.28 0.45 0.45 0.43 0.36 0.44 0.62 0.48 0.08 0.3 0.7 0.38 0.433 46 PDGFRB 0.7 0.53 0.71 0.5 0.68 0.39 0.56 0.47 0.77 0.77 0.36 0.48 0.21 0.548 146 ADAMTS4 0.5 0.3 0.52 0.44 0.62 0.33 0.39 0.58 0.58 0.59 0.11 0.39 0.27 0.432 466 47 LOXL2 0.6 0.68 0.62 0.5 0.57 0.56 0.27 0.64 0.73 0.7 0.32 0.59 0.3 0.545 147 PDPN 0.62 0.58 0.75 0.28 0.17 0.39 0.14 0.55 0.51 0.67 -0.1 0.53 0.5 0.432 48 OMD 0.76 0.59 0.58 0.34 0.72 0.4 0.42 0.67 0.79 0.68 0.24 0.17 0.72 0.545 148 COLEC12 0.66 0.35 0.71 0.37 0.73 0.46 0.1 0.16 0.6 0.66 -0.1 0.07 0.74 0.428 49 P4HA3 0.68 0.64 0.73 0.53 0.76 0.46 0.36 0.64 0.57 0.67 0.19 0.62 0.22 0.544 149 MAGEL2 0.54 0.52 0.54 0.4 0.81 0.57 0.4 0.55 0.67 0.32 -0 0.4 -0.1 0.427 50 LOX 0.74 0.71 0.76 0.19 0.47 0.34 0.36 0.68 0.58 0.72 0.14 0.7 0.68 0.544 150 C1S 0.65 0.33 0.75 0.24 0.34 0.44 0.37 0.41 0.51 0.43 0.19 0.21 0.67 0.426 51 DCN 0.71 0.58 0.72 0.41 0.67 0.63 0.51 0.5 0.62 0.71 0.2 0.07 0.73 0.543 151 DDR2 0.67 0.44 0.73 0.4 0.59 0.33 0.41 0.35 0.75 0.43 0.11 0.11 0.22 0.426 52 TIMP2 0.67 0.65 0.79 0.47 0.77 0.46 0.14 0.56 0.81 0.49 0.25 0.34 0.61 0.539 152 BMP1 0.57 0.57 0.51 0.3 0.4 0.47 0.2 0.43 0.54 0.46 0.06 0.62 0.4 0.425 53 WNT2 0.71 0.67 0.63 0.44 0.76 0.46 0.23 0.63 0.71 0.22 0.03 0.74 0.78 0.539 153 TENM3 0.67 0.56 0.71 0.38 0.35 0.29 -0 0.6 0.66 0.56 -0 0.45 0.33 0.425 54 NOX4 0.64 0.78 0.77 0.54 0.76 -0 -0.2 0.79 0.67 0.41 0.57 0.71 0.58 0.538 154 DACT3 0.51 0.44 0.57 0.46 0.69 0.38 0.51 0.53 0.66 0.53 0.1 -0.1 0.21 0.425 55 TNFSF4 0.73 0.66 0.75 0.37 0.58 0.28 0.22 0.68 0.6 0.63 0.14 0.62 0.63 0.53 155 DIO2 0.58 0.49 0.65 0.22 0.46 0.53 0.45 0.71 0.69 0.64 0.33 0.15 -0.4 0.425 56 EMILIN1 0.71 0.39 0.64 0.58 0.77 0.6 0.44 0.57 0.74 0.58 0.03 0.13 0.68 0.528 156 SPOCK1 0.51 0.67 0.83 0.46 0.49 0.37 0.12 0.78 0.31 0.56 -0.2 0.5 0.08 0.425 57 NID2 0.7 0.73 0.57 0.64 0.76 0.39 0.37 0.7 0.81 0.57 0.16 0.49 -0 0.527 157 GGT5 0.6 0.05 0.59 0.45 0.66 0.46 0.42 0.29 0.65 0.48 0.24 0.26 0.34 0.422 58 COL6A1 0.72 0.53 0.62 0.48 0.78 0.56 0.32 0.58 0.72 0.62 0.14 0.32 0.45 0.526 158 PALLD 0.68 0.6 0.66 0.34 0.41 0.48 0.1 0.55 0.34 0.63 0.07 0.13 0.5 0.422 59 COL8A2 0.7 0.72 0.77 0.29 0.56 0.42 0.25 0.66 0.73 0.51 0.31 0.32 0.59 0.525 159 NTNG2 0.72 0.55 0.53 0.29 0.61 0.35 0.27 0.56 0.6 0.39 -0.1 0.4 0.29 0.422 60 TNFAIP6 0.77 0.59 0.69 0.52 0.73 -0.3 0 0.75 0.62 0.72 0.34 0.67 0.68 0.522 160 LGALS1 0.57 0.36 0.54 0.43 0.38 0.41 0.19 0.58 0.54 0.58 0.07 0.31 0.51 0.421 61 RCN3 0.71 0.38 0.54 0.38 0.73 0.44 0.32 0.57 0.58 0.58 0.35 0.55 0.61 0.518 161 CPXM1 0.79 0.27 0.52 0.35 0.77 0.66 0.43 0.61 0.67 -0 -0.2 0.48 0.15 0.42 62 FAM26E 0.66 0.71 0.74 0.55 0.57 0.33 0.5 0.59 0.62 0.57 0.39 0.41 0.09 0.518 162 ADAMTS6 0.57 0.59 0.59 0.29 0.5 0.34 0.41 0.29 0.51 0.11 0.38 0.34 0.54 0.42 63 COL5A3 0.57 0.61 0.59 0.4 0.66 0.37 0.39 0.64 0.77 0.75 0.22 0.59 0.14 0.515 163 HMCN1 0.46 0.58 0.69 0.39 0.63 0.1 0.51 0.1 0.54 0.46 0.29 0.26 0.45 0.42 64 DACT1 0.71 0.74 0.47 0.58 0.8 0.46 0.34 0.67 0.78 0.58 0.22 0.34 -0 0.514 164 NNMT 0.68 0.32 0.67 0.29 0.61 0.19 0.23 0.17 0.5 0.56 0.15 0.34 0.75 0.42 65 GFPT2 0.69 0.5 0.7 0.38 0.48 0.54 0.28 0.63 0.63 0.67 0.13 0.45 0.6 0.514 165 PMP22 0.39 0.39 0.63 0.39 0.69 0.36 0.23 0.41 0.65 0.67 0.21 0.09 0.34 0.419 66 MXRA8 0.74 0.45 0.7 0.53 0.79 0.4 0.13 0.43 0.78 0.5 0.31 0.41 0.49 0.512 166 TGFBI 0.54 0.59 0.05 0.29 0.25 0.59 0.23 0.54 0.48 0.61 0.36 0.42 0.5 0.419 67 CCDC80 0.73 0.39 0.74 0.19 0.56 0.6 0.48 0.67 0.64 0.72 0.23 0.2 0.43 0.506 167 CLEC11A 0.71 0.36 0.48 0.35 0.77 0.49 0.19 0.58 0.54 0.16 0.15 0.47 0.19 0.418 68 ANGPTL2 0.67 0.56 0.75 0.46 0.65 0.15 0.2 0.7 0.76 0.67 0.22 0.34 0.45 0.506 168 CCL11 0.68 0.34 0.18 0.36 0.4 0.51 0.3 0.66 0.62 0.7 0.21 -0.1 0.59 0.418 69 MMP14 0.61 0.64 0.64 0.3 0.57 0.43 -0.1 0.6 0.64 0.7 0.15 0.67 0.64 0.502 169 ZFPM2 0.72 0.61 0.7 0.24 0.51 0.41 0.39 0.48 0.56 0.28 0.26 0.13 0.14 0.418 70 KIF26B 0.64 0.67 0.59 0.44 0.64 0.41 0.13 0.57 0.71 0.65 0.26 0.51 0.31 0.502 170 GUCA1A 0.76 0.37 0.51 0.5 0.64 0.22 0.14 0.71 0.64 0.04 0.06 0.66 0.18 0.418 71 FIBIN 0.85 0.65 0.77 0.54 0.75 -0.1 0.01 0.61 0.77 0.53 0.46 0.31 0.29 0.499 171 SPOCD1 0 0.41 0.51 0.36 0.45 0.58 0.3 0.53 0.49 0.41 0.21 0.54 0.62 0.416 72 MRC2 0.64 0.52 0.71 0.46 0.62 0.63 0.43 0.4 0.65 0.47 0.15 0.42 0.39 0.499 172 TGFB1I1 0.56 0.38 0.58 0.36 0.56 0.44 0.33 0.53 0.7 0.64 -0.1 0.08 0.32 0.416 73 KCND2 0.22 0.54 0.65 0.53 0.75 0.4 0.14 0.7 0.67 0.35 0.28 0.56 0.66 0.496 173 KERA 0.57 0.43 0.47 0.24 0.56 0.13 0.26 0.56 0.51 0.39 0.31 0.33 0.64 0.415 74 ALPK2 0.62 0.68 0.7 0.61 0.62 -0.2 0.06 0.7 0.68 0.78 0.15 0.31 0.73 0.493 174 NUAK1 0.4 0.56 0.61 0.22 0.44 0.38 0.38 0.55 0.46 0.71 0.18 0.46 0.05 0.415 75 ADAMTS14 0.72 0.6 0.41 0.37 0.71 0.52 0.1 0.41 0.62 0.44 0.31 0.42 0.73 0.489 175 ALDH1L2 0.68 0.59 0.4 0.33 0.39 0.52 0.37 0.61 0.37 0.39 0.14 0.27 0.34 0.415 76 TWIST1 0.61 0.24 0.69 0.3 0.63 0.45 0.48 0.71 0.48 0.61 0.12 0.45 0.57 0.488 176 RUNX2 0.24 0.6 0.47 0.15 0.57 0.51 0.26 0.53 0.49 0.64 0.04 0.36 0.52 0.414 77 ACTA2 0.59 0.4 0.62 0.61 0.76 0.42 0.52 0.66 0.73 0.74 -0.1 0.05 0.3 0.488 177 CLEC5A 0.61 0.48 0.68 0.26 0.36 0.07 0.19 0.43 0.51 0.26 0.3 0.61 0.62 0.414 78 LAMA4 0.59 0.58 0.72 0.52 0.7 0.41 0.34 0.53 0.77 0.65 0.27 0.19 0.06 0.487 178 CCIN 0.61 0.54 0.42 0.43 0.6 0.48 0.22 0.54 0.55 0.09 0.12 0.32 0.45 0.413 79 TAGLN 0.65 0.36 0.68 0.55 0.75 0.46 0.53 0.66 0.68 0.62 -0 0.02 0.4 0.486 179 MSC 0.66 0.38 0.56 0.42 0.45 0.1 0.07 0.6 0.31 0.62 0.3 0.31 0.59 0.413 80 CERCAM 0.6 0.43 0.73 0.35 0.61 0.51 0.4 0.6 0.61 0.52 0.14 0.49 0.3 0.484 180 PRR16 0.58 0.36 0.61 0.31 0.39 0.45 0.44 0.57 0.49 0.37 0.15 0.4 0.23 0.412 81 FBLN2 0.57 0.4 0.65 0.27 0.67 0.35 0.4 0.59 0.69 0.44 0.31 0.32 0.62 0.483 181 LZTS1 0.39 0.29 0.55 0.43 0.53 0.3 0.33 0.45 0.52 0.5 0.34 0.34 0.37 0.411 82 CALD1 0.67 0.53 0.81 0.53 0.67 0.32 0.07 0.55 0.79 0.61 0.07 0.15 0.49 0.482 182 FILIP1L 0.54 0.55 0.69 0.44 0.5 0.29 0.22 0.38 0.64 0.62 0.19 0.04 0.23 0.41 83 ECM2 0.68 0.56 0.7 0.38 0.73 0.27 0.15 0.47 0.75 0.6 0.34 0.25 0.34 0.478 183 ZNF521 0.62 0.22 0.73 0.47 0.66 0.24 0.42 0.44 0.75 0.38 0.28 0.13 -0 0.41 84 MEDAG 0.54 0.26 0.67 0.3 0.51 0.43 0.48 0.54 0.49 0.64 0.36 0.35 0.62 0.476 184 PMEPA1 0.51 0.47 0.25 0.35 0.41 0.36 0.43 0.5 0.56 0.55 -0.1 0.4 0.63 0.409 85 TMEM119 0.59 0.22 0.65 0.34 0.67 0.59 0.48 0.36 0.63 0.42 0.4 0.25 0.59 0.476 185 TMEM158 0.59 0.44 0.24 0.41 0.32 0.58 0.44 0.66 0.18 0.73 -0.3 0.35 0.67 0.409 86 TGFB3 0.67 0.26 0.65 0.27 0.68 0.58 0.4 0.64 0.66 0.55 -0.2 0.36 0.64 0.475 186 CHN1 0.46 0.27 0.68 0.38 0.58 0.44 0.16 0.48 0.62 0.49 0.16 0.23 0.36 0.408 87 EDNRA 0.55 0.56 0.57 0.44 0.52 0.34 0.52 0.39 0.71 0.71 0.06 0.26 0.54 0.475 187 TSHZ3 0.48 0.55 0.67 0.24 0.44 0.57 0.47 0.63 0.64 0.48 -0 0.15 0.01 0.407 88 FSTL1 0.56 0.59 0.83 0.39 0.66 0.3 0.06 0.56 0.61 0.58 0.1 0.35 0.57 0.474 188 MICAL2 0.37 0.58 0.18 0.26 0.48 0.53 0.3 0.25 0.63 0.54 0.24 0.26 0.66 0.406 89 C1QTNF6 0.41 0.54 0.44 0.27 0.47 0.44 0.14 0.64 0.6 0.69 0.21 0.7 0.59 0.472 189 LMCD1 0.6 0.4 0.51 0.23 0.7 0.41 0.25 0.41 0.5 0.22 0.09 0.37 0.58 0.405 90 SCARF2 0.73 0.46 0.63 0.53 0.76 0.46 0.28 0.53 0.69 0.27 0.21 0.45 0.14 0.472 190 SNAI2 0.28 0.43 0.71 0.18 0.25 0.51 0.48 0.68 0.28 0.76 -0.2 0.45 0.45 0.405 91 RGS4 0.66 0.36 0.61 0.43 0.59 0.48 0.47 0.56 0.48 0.58 0.17 0.17 0.57 0.472 191 GALNT15 0.55 0.46 0.63 0.55 0.68 -0.1 0.16 0.55 0.65 0.46 0.2 0.21 0.24 0.405 92 FAM180A 0.73 0.46 0.68 0.43 0.4 0.52 0.35 0.38 0.61 0.3 0.39 0.22 0.66 0.472 192 CSGALNACT2 0.47 0.43 0.64 0.23 0.42 0.23 0.33 0.37 0.52 0.39 0.37 0.31 0.55 0.405 93 OLFML1 0.66 0.52 0.7 0.37 0.7 0.36 0.38 0.43 0.75 0.59 0.08 0.2 0.39 0.472 193 BCAT1 0.52 0.64 0.76 0.31 0.42 0.45 0.24 0.12 0.31 0.08 0.19 0.62 0.58 0.403 94 MSRB3 0.64 0.54 0.79 0.53 0.68 0.46 0.4 0.45 0.65 0.68 0.05 0.07 0.18 0.471 194 AXL 0.55 0.57 0.66 0.21 0.38 0.43 0.14 0.37 0.54 0.41 0.17 0.22 0.57 0.402 95 CRISPLD2 0.56 0.49 0.66 0.35 0.7 0.35 0.22 0.58 0.7 0.82 -0.1 0.16 0.6 0.47 195 CILP2 0.28 0.46 0.26 0.43 0.46 0.57 0.3 0.71 0.5 0.39 0.19 0.41 0.25 0.401 96 GPC6 0.73 0.68 0.81 0.56 0.73 0.14 -0.1 0.61 0.71 0.41 0.27 0.29 0.24 0.469 97 SOX11 0.49 0.34 0.66 0.62 0.72 0.4 0.35 0.39 0.4 0.33 0.28 0.58 0.52 0.468 98 GPR1 0.66 0.76 0.52 0.25 0.08 0.33 0.3 0.74 0.45 0.53 0.31 0.53 0.61 0.467 99 GAS1 0.74 0.48 0.71 0.2 0.62 0.5 0.27 0.7 0.72 0.45 -0.1 0.28 0.45 0.466 100 SERPINF1 0.61 0.55 0.73 0.28 0.56 0.55 0.09 0.56 0.43 0.74 -0 0.25 0.71 0.465 467

473

472

etr 21) doi: (2016), Letters 10.1016/j.canlet.2016.09.001

471

o fteaetctres Cancer targets, therapeutic of ion prioritizat the for ﬁbroblasts activated of signature gene pan-cancer COL11A1-correlated A al., et Jia, Dongyu as: press in article this cite Please

470

469 468 C sn nardtts.Tepo a eeae sn rpPdPimsfwr eso ..Itrru ifrne eeasse yteSuetst-tes Student’s the by assessed were differences Intergroup 6.0. version software Prism GraphPad using generated was plot The test. t unpaired using N eentpeeti h aaae cos20idvda eltps( xs rue no1 angop.()Oelpo h 9 pan-cancer 195 the of Overlap (D) groups. main 10 into grouped axis) (X types cell individual 230 across database) the in present not were endb ot ta.()Uwihe eazsoe f191 of z-scores meta Unweighted (E) al. et Moﬃtt by deﬁned n As(1smls nhg rd eosoaincne nteGE09 aae sn h pan-cancer pan-cancer the the using of dataset Expression GSE40595 (C) (hgserver1.amc.nl/cgi-bin/r2/main.cgi). the tool in Analysis cancer Clustering ovarian GeneSet serous grade R2 high public pan-cancer in the using samples) dataset (31 e-mtab-991 CAFs the and in samples (PDX) xenograft patient-derived 75 5. Fig. ye nteImeepoet( project ImmGene the in types A

= mRNA (log2) 9) mean 191); COL11A1

5.00E+00 5.00E+01 Pan-cancer -correlated genes 5.00E-02 5.00E-01 h pan-cancer The Colon Cancer e-mtab-991 Cancer Colon D (75) PDX (40) Primary tumor ± E falgns(0.09918 genes all of SEM Stem el Bcl edii el arpae oo NaTCell abT Cell B-cell Dendritic Cell Macrophage Mono GN COL11A1 creae eesti xrse nCF n soitdwt orptetsria nmlil acrtps A eee xrsincutrn anal clustering expression GeneSet (A) types. cancer multiple in survival patient poor with associated and CAFs in expressed is set gene -correlated immgen.com ± .Tepo a eeae sn yeee ol( tool MyGeneset using generated was plot The ). .11 N 0.01416

= 38) ****p 23287); COL11A1 creae ee 4gnswr o vial ntePEO aaae eecmae ihtoeo l ee ttl23 (total genes all of those with compared were database) PRECOG the in available not were genes (4 genes -correlated < 0.0001. B -3 -3 0 3 Score COL11A1 rstats.immgen.org/MyGeneSet creae ee.()GnStepeso lseigaayi flsrmcoisce vra acreihla el 3 samples) (32 cells epithelial cancer ovarian laser-microdissected of analysis clustering expression GeneSet (B) genes. -correlated

COL11A1 Pan-cancer COL11A1-correlated genes OvarianGSE40595 Cancer creae ee.TeEciendsac lseigaayi etasi A n B eegnrtduigthe using generated were (B) and (A) in heatmaps analysis clustering distance Euclidean The genes. -correlated Ovarian cancerfibroblasts (31) Ovarian cancerepithelia (32) COL11A1 creae ee apdo h rncitm fidvda uiehmtpitcadsrmlcell stromal and hematopoietic murine individual of transcriptome the on mapped genes -correlated .Bakdt ersn RAlvl yai)o 8 pan-cancer 186 of axis) (y levels mRNA represent dots Black ). COL11A1 creae ee ih2 nra toa n 5‘ciae toa genes stroma’ ‘activated 25 and stroma’ ‘normal 23 with genes -correlated .Mean t. ± E fpan-cancer of SEM -3 0 3 0 -3 Score d el Cl Lymphocyte Cell Cell gdT sso 0piayclncne ape and samples cancer colon primary 40 of ysis E D COL11A1 correlated genes correlated 195 pan-cancer , stroma genes stroma stroma genes stroma 8 ee)i h RCGdatabase PRECOG the in genes) 287 COL11A1 25

creae ee (0.8916 genes -correlated unweighted poor Moffitt, 2015 Moffitt, 2015 23 survival meta z-score COL11A1- activated 0 1 2 3 4 normal COL11A1- carcinoma correlated creae ee 9genes (9 genes -correlated genes Stromal pan- **** 177 genes 18 23 7 0 all ±

0.1641

– (2016) Letters Cancer / al. et Jia D. 9 ■■ ■■ ■■ RIL NPRESS IN ARTICLE ARTICLE IN PRESS

10 D. Jia et al. / Cancer Letters ■■ (2016) ■■–■■

477 fibrosis, osteogenesis, wound healing, EMT, cardiovascular disease, Discussion 543 478 and transforming growth factor β (TGFβ) signaling (Table S5). 544 479 COL11A1-correlated gene set enrichment analysis of chemical and Whereas in the past most therapeutic approaches have focused 545 480 genetic perturbations (CGP) showed the most significant overlap with on the cancer cell and its genetic alterations, it is becoming appar- 546 481 genes up-regulated in association with cancer invasiveness, ad- ent that the microenvironment plays an equally important role in 547 482 vanced stage, stromal cell stemness, and epithelial-mesenchymal cancer evolution. We now recognize that the cancer stroma not only 548 483 transition (EMT) (Table S6). The uniformity of the COL11A1-correlated serves as a scaffold for tissue organization and integrity but also pro- 549 484 genes across different cancers might also indicate that these genes vides key biomechanical and molecular signals that can affect various 550 485 are regulated by a common mechanism. Ingenuity Pathway Anal- aspects of cancer growth and biology, including proliferation, sur- 551 486 ysis showed that transforming growth factor beta 1 (TGFB1) is the vival, metabolism, stem cell fate, and response to chemotherapy 552 487 most strongly associated upstream regulator of the pan-cancer [51,52]. As the genetically stable subpopulations of the cancer mi- 553 488 COL11A1-correlated genes (Table S7). croenvironment are increasingly recognized as potentially effective 554 489 therapeutic targets, a comprehensive definition of their molecular 555 490 COL11A1-correlated genes are associated with poor patient survival characteristics will be a prerequisite for the development of more 556 491 and represent potential therapeutic targets precise and less toxic therapies. Currently, there are no reliable 557 492 methods to distinguish activated CAFs from non-activated CAFs, 558 493 To determine whether the pan-cancer COL11A1-correlated gene which although frequently abundant within cancers do not neces- 559 494 set is associated with patient survival in the ~18,000 cases of liquid sarily contribute to adverse outcome. We identified COL11A1 among 560 495 and solid malignancies in the PRECOG dataset [36], we compared the top differentially expressed genes in multiple cancer types when 561 496 survival z-scores for the 195 pan-cancer COL11A1-correlated genes cancer tissues and their corresponding normal tissues were com- 562 497 with the survival z-scores for all genes in the dataset. This analy- pared. We showed that an increase in COL11A1 expression is 563 498 sis showed that expression of the pan-cancer COL11A1-correlated associated with progression and poor survival in most cancer types. 564 499 gene set is significantly associated with poor survival (Fig. 5E). COL11A1 is a particularly attractive therapeutic target because of its 565 500 Expression profile analyses have identified a mesenchymal mo- restricted expression in normal tissues and non-cancer condi- 566 501 lecular subtype of cancer associated with poor survival in multiple tions, such as inflammation and fibrosis. 567 502 cancers including ovarian [47], pancreatic [46], gastric [48] and colon The identification of a highly conserved set of genes associated 568 503 [49]. In colon cancer, it has been shown that the mesenchymal mo- with COL11A1 expression in breast, lung, pancreas, stomach, urinary 569 504 lecular subtype, which constitutes approximately 23% of colon bladder, colon, thyroid, cervix, head and neck, thyroid, ovary, and 570 505 cancers, has no significant enrichment for targetable mutations or prostate cancers was somewhat surprising in light of the genetic 571 506 copy number changes in candidate driver genes [49]. Even if future and phenotypic diversity among these cancer types. The con- 572 507 research identifies targetable events in cancer cells of the mesen- served expression signature indicates that the reaction of stromal 573 508 chymal subtype, it is predicted that enrichment in CAFs and excessive tissues to invading epithelial cancer cells may be similar regard- 574 509 ECM deposition will reduce therapeutic efficacy by creating a phys- less of the organ of origin or genetic alterations. This has significant 575 510 ical barrier for drug transport. Thus, simultaneous targeting of CAFs implications for the development of pan-cancer therapeutic strat- 576 511 and cancer cells may be necessary for chemotherapeutic accessibility. egies. Our analysis of potential upstream regulators of the pan- 577 512 To identify therapies that preferentially target activated CAFs cancer COL11A1-correlated genes revealed TGFB1 as the most likely 578 513 and spare normal tissues, we combined drug target searches with candidate. Dysregulation of TGFβ signaling is recognized as the main 579 514 expression profile datasets in cancers and normal tissues. Ingenu- driver of fibroblast activation and represents the most logical ther- 580 515 ity Pathway Analysis and searches of the Clinicaltrials.gov apeutic target [53]. In immortalized normal ovary fibroblast cell 581 516 (clinicaltrials.gov) and ChEMBL [50] (ebi.ac.uk/chembl) databases culture, recombinant TGFB1 has been shown to upregulate expres- 582 517 revealed that, of the 195 pan-cancer COL11A1-correlated genes, 16 sion of COL11A1 and several other COL11A1-correlated genes; this 583 518 are targets of drugs used in clinical trials (Table S8) and 30 are targets effect was abrogated by the TGFβ receptor inhibitor A83-01 [32]. 584 519 of bioactive compounds (Table S9). However, the pleiotropic nature of TGFβ signaling carries the risk 585 520 To test whether any of the drug/bioactive compound target genes of adverse effects in patients [54]. In order to abrogate fibroblast 586 521 in Tables S8 and S9 are exclusively expressed in activated CAFs, we activation without the negative effects of pan-TGFβ therapy, it will 587 522 determined the expression levels of each gene in normal tissues in be necessary to design therapies for more specific targets. 588 523 the GTEx database [43] and in normal vs. cancer tissues in the GENT Sixteen of the 195 pan-cancer COL11A1-correlated genes are 589 524 database [35]. Additionally, to test whether selected genes were ex- targets of drugs in clinical trials. These targets include CTGF, a 590 525 clusively overexpressed in cancer tissues and not in non-cancer matricellular protein involved in myofibroblast formation in cancer 591 526 associated pathologies such as inflammation and fibrosis, we com- as a binding factor of fibronectin and a downstream mediator of 592 527 pared expression of these genes in normal tissues, inflamed/ TGFβ. A clinical trial (clinicaltrials.gov; NCT02210559) is currently 593 528 fibrotic tissues and cancer tissues of the colon and lung. Unlike enrolling patients with unresectable pancreatic cancer to test a com- 594 529 COL11A1, which has restricted expression in normal tissues (Fig. S2) bination of conventional chemotherapy and FG-3019, the human 595 530 and is highly elevated in cancer vs. normal tissues (Fig. 4A) and in monoclonal antibody that interferes with the action of CTGF. Our 596 531 cancer vs. inflamed/fibrotic tissues (Fig. 4B) most of the target genes expression analyses, consistent with the published literature (re- 597 532 were expressed at high levels in at least one normal tissue and/or viewed in Ref. [55]), show that CTGF is expressed at similar levels 598 533 exhibited equivalent expression levels in cancers vs. normal tissues, in normal tissues and cancers and therefore unlikely to be a safe 599 534 and cancers vs. inflamed/fibrotic tissues. One example of this pattern therapeutic target for cancer treatment. In contrast, targets such as 600 535 of expression is CTGF (Figs. S6 and S7). However, FN1, MMP13, MMP14, FN1, FAP, MMP13, LOX and COL2A1 are markedly increased in cancer/ 601 536 FAP, LOX and COL1A2 exhibited restricted expression in normal tissues inflammation/fibrosis compared to normal tissues and are thus 602 537 and elevated expression in cancer vs. normal tissues. One example predicted to have a better safety profile than agents targeting CTGF. 603 538 of this pattern of expression is FAP (Figs. S8 and S9). Among FN1, Our assessment is consistent with the published moderate and re- 604 539 MMP13, MMP14, FAP, LOX and COL1A2, FAP was also differentially versible toxicity of the FN1-targeting monoclonal antibody-cytokine 605 540 expressed between inflamed/fibrotic tissues and cancer tissues al- fusion protein L19-IL2, which is in a phase I/II study for patients 606 541 though this difference in expression was not as prominent as for with solid cancers (clinicaltrials.gov; NCT01058538) [56,57]. Our ex- 607 542 COL11A1 (compare Fig. S9B and Fig. 4B). pression analyses show that one of the targets of bioactive 608

D. Jia et al. / Cancer Letters ■■ (2016) ■■–■■ 11

609 compounds, FAP, has low expression in normal tissues and also lower [5] M.J. Paszek, N. Zahir, K.R. Johnson, J.N. Lakins, G.I. Rozenberg, A. Gefen, et al., 676 610 expression in inflamed/fibrotic tissues than in cancer. Yet, this dif- Tensional homeostasis and the malignant phenotype, Cancer Cell 8 (2005) 677 241–254. 678 611 ference in expression may not be sufficient for specific targeting of [6] P.P. Provenzano, D.R. Inman, K.W. Eliceiri, J.G. Knittel, L. Yan, C.T. Rueden, et al., 679 612 activated CAFs as studies in mouse models have shown that de- Collagen density promotes mammary tumor initiation and progression, BMC 680 613 pletion of the FAP+ stroma can induce toxicity due to expression Med. 6 (2008) 11. 681 [7] D.F. Quail, J.A. Joyce, Microenvironmental regulation of tumor progression and 682 614 of FAP in the mesenchymal cells of bone marrow, muscle and adipose metastasis, Nat. Med. 19 (2013) 1423–1437. 683 615 tissue [58,59]. Future efforts to specifically target activated CAFs can [8] R.K. Jain, Normalizing tumor microenvironment to treat cancer: bench to 684 616 be improved by designing novel therapies to target genes that exhibit bedside to biomarkers, J. Clin. Oncol. 31 (2013) 2205–2218. 685 [9] W. Qiu, M. Hu, A. Sridhar, K. Opeskin, S. Fox, M. Shipitsin, et al., No evidence 686 617 restricted expression in nonmalignant tissues. When considering of clonal somatic genetic alterations in cancer-associated fibroblasts from human 687 618 COL11A1 as a cancer-specific biomarker and therapeutic target, it breast and ovarian carcinomas, Nat. Genet. 40 (2008) 650–655. 688 619 is important to note that several normal tissues express COL11A1. [10] J.N. Arnold, L. Magiera, M. Kraman, D.T. Fearon, Tumoral immune suppression 689 by macrophages expressing fibroblast activation protein-alpha and heme 690 620 The potential side effects of COL11A1 targeting can be predicted oxygenase-1, Cancer Immunol. Res. 2 (2014) 121–126. 691 621 based on the phenotypes of mice and humans expressing mutant [11] C. Feig, J.O. Jones, M. Kraman, R.J. Wells, A. Deonarine, D.S. Chan, et al., Targeting 692 622 nonfunctional forms of COL11A1. A homozygous truncating muta- CXCL12 from FAP-expressing carcinoma-associated fibroblasts synergizes with 693 623 tion of COL11A1 in mice results in poorly formed cartilage [60], anti-PD-L1 immunotherapy in pancreatic cancer, Proc. Natl. Acad. Sci. U.S.A. 694 110 (2013) 20212–20217. 695 624 while human COL11A1 mutations are associated with articular [12] M. Kraman, P.J. Bambrough, J.N. Arnold, E.W. Roberts, L. Magiera, J.O. Jones, et al., 696 625 hypermobility, dermal hyperelasticity and widespread tissue fra- Suppression of antitumor immunity by stromal cells expressing fibroblast 697 626 gility [61]. Of note, these collagenopathies are associated with the activation protein-alpha, Science 330 (2010) 827–830. 698 [13] A. Lo, L.C. Wang, J. Scholler, J. Monslow, D. Avery, K. Newick, et al., Tumor- 699 627 absence of COL11A1 function throughout development and are un- promoting desmoplasia is disrupted by depleting FAP-expressing stromal cells, 700 628 likely to manifest upon transient targeting of COL11A1 in adults. Cancer Res. 75 (2015) 2800–2810. 701 629 Additionally, since COL11A1 and many of the pan-cancer COL11A1- [14] M. Loeffler, J.A. Kruger, A.G. Niethammer, R.A. Reisfeld, Targeting tumor- 702 associated fibroblasts improves cancer chemotherapy by increasing intratumoral 703 630 coexpressed genes have multiple tissue-specific mRNA splicing drug uptake, J. Clin. Invest. 116 (2006) 1955–1962. 704 631 isoforms, it will be valuable for future targeting purposes to deter- [15] P.P. Provenzano, C. Cuevas, A.E. Chang, V.K. Goel, D.D. Von Hoff, S.R. Hingorani, 705 632 mine if any mRNA isoforms are specifically expressed in activated Enzymatic targeting of the stroma ablates physical barriers to treatment of 706 pancreatic ductal adenocarcinoma, Cancer Cell 21 (2012) 418–429. 707 633 CAFs [62]. [16] B.C. Ozdemir, T. Pentcheva-Hoang, J.L. Carstens, X. Zheng, C.C. Wu, T.R. Simpson, 708 634 et al., Depletion of carcinoma-associated fibroblasts and fibrosis induces 709 635 Acknowledgments immunosuppression and accelerates pancreas cancer with reduced survival, 710 Cancer Cell 25 (2014) 719–734. 711 636 [17] M.R. Junttila, F.J. de Sauvage, Influence of tumour micro-environment 712 637 We thank S. Swartwood (Biobank and Translational Research heterogeneity on therapeutic response, Nature 501 (2013) 346–354. 713 638 Core) for in situ hybridization; F. Chung (Pathology Service) for im- [18] A. Kawase, G. Ishii, K. Nagai, T. Ito, T. Nagano, Y. Murata, et al., Podoplanin 714 expression by cancer associated fibroblasts predicts poor prognosis of lung 715 639 munohistochemistry; G. Martins (Flow Cytometry Core) for assistance adenocarcinoma, Int. J. Cancer 123 (2008) 1053–1059. 716 640 with FACS analysis; and K. Daniels for assistance with manuscript [19] M.J. Ronty, S.K. Leivonen, B. Hinz, A. Rachlin, C.A. Otey, V.M. Kahari, et al., 717 641 preparation. SO is supported by the Office of the Assistant Secre- Isoform-specific regulation of the actin-organizing protein palladin during 718 TGF-beta1-induced myofibroblast differentiation, J. Invest. Dermatol. 126 (2006) 719 642 tary of Defense for Health Affairs through the Ovarian Cancer 2387–2396. 720 643 Research Program Award No. W81XWH-14-1-0107 and W81XWH- [20] T.A. Brentnall, L.A. Lai, J. Coleman, M.P. Bronner, S. Pan, R. Chen, Arousal of 721 644 16-1-0190, the National Cancer Institute grants R21CA194626-01 cancer-associated stroma: overexpression of palladin activates fibroblasts to 722 promote tumor invasion, PLoS ONE 7 (2012) e30219. 723 645 and 1R01CA208753-01, and the Cantor-UCLA Women’s Health Center [21] O. De Wever, Q.D. Nguyen, L. Van Hoorde, M. Bracke, E. Bruyneel, C. Gespach, 724 646 Executive Advisory Board & NCATS UCLA Clinical and Transla- et al., Tenascin-C and SF/HGF produced by myofibroblasts in vitro provide 725 647 tional Science Institute grant UL1TR000124. BYK is supported by convergent pro-invasive signals to human colon cancer cells through RhoA and 726 648 the American Cancer Society Early Detection Professorship Grant Rac, FASEB J. 18 (2004) 1016–1018. 727 [22] K. Pietras, J. Pahler, G. Bergers, D. Hanahan, Functions of paracrine PDGF 728 649 (SIOP-06-258-01-COUN). BYK, SO, KL and WRW are supported by signaling in the proangiogenic tumor stroma revealed by pharmacological 729 650 Q5 the Ovarian Cancer Research Fund Program Project Development targeting, PLoS Med. 5 (2008) e19. 730 651 Q6 Award and DJC is supported by the Ovarian Cancer Research Fund [23] N. Erez, M. Truitt, P. Olson, S.T. Arron, D. Hanahan, Cancer-associated fibroblasts 731 are activated in incipient neoplasia to orchestrate tumor-promoting 732 652 Ann Shreiber Mentored Investigator Award. Opinions, interpreta- inflammation in an NF-kappaB-dependent manner, Cancer Cell 17 (2010) 733 653 tions, conclusions and recommendations are those of the authors 135–147. 734 654 and are not necessarily endorsed by the funding agencies. [24] H. Sugimoto, T.M. Mundel, M.W. Kieran, R. Kalluri, Identification of fibroblast 735 heterogeneity in the tumor microenvironment, Cancer Biol. Ther. 5 (2006) 736 655 1640–1646. 737 656 Conflict of interest [25] F. Vazquez-Villa, M. Garcia-Ocana, J.A. Galvan, J. Garcia-Martinez, C. 738 657 Garcia-Pravia, P. Menendez-Rodriguez, et al., COL11A1/(pro)collagen 11A1 739 expression is a remarkable biomarker of human invasive carcinoma-associated 740 658 The authors declare no conflict of interest. stromal cells and carcinoma progression, Tumor Biol. 36 (2015) 2213–2222. 741 659 [26] Z. Raglow, S.M. Thomas, Tumor matrix protein collagen XIalpha1 in cancer, 742 Cancer Lett. 357 (2015) 448–453. 743 660 Appendix:: Supplementary material [27] R.J. Wenstrup, S.M. Smith, J.B. Florer, G. Zhang, D.P. Beason, R.E. Seegmiller, et al., 744 661 Regulation of collagen fibril nucleation and initial fibril assembly involves 745 662 Supplementary data to this article can be found online at coordinate interactions with collagens V and XI in developing tendon, J. Biol. 746 Chem. 286 (2011) 20455–20465. 747 663 doi:10.1016/j.canlet.2016.09.001. [28] S.M. Smith, D.E. Birk, Focus on molecules: collagens V and XI, Exp. Eye Res. 98 748 664 (2012) 105–106. 749 665 References [29] P. Farmer, H. Bonnefoi, P. Anderle, D. Cameron, P. Wirapati, V. Becette, et al., A 750 stroma-related gene signature predicts resistance to neoadjuvant chemotherapy 751 666 in breast cancer, Nat. Med. 15 (2009) 68–74. 752 667 [1] R. Kalluri, M. Zeisberg, Fibroblasts in cancer, Nat. Rev. Cancer 6 (2006) 392–401. [30] W.Y. Cheng, J.J. Kandel, D.J. Yamashiro, P. Canoll, D. Anastassiou, A multi-cancer 753 668 [2] B. Hinz, S.H. Phan, V.J. Thannickal, M. Prunotto, A. Desmouliere, J. Varga, et al., mesenchymal transition gene expression signature is associated with prolonged 754 669 Recent developments in myofibroblast biology: paradigms for connective tissue time to recurrence in glioblastoma, PLoS ONE 7 (2012) e34705. 755 670 remodeling, Am. J. Pathol. 180 (2012) 1340–1355. [31] J. Boguslawska, H. Kedzierska, P. Poplawski, B. Rybicka, Z. Tanski, A. 756 671 [3] A. Desmouliere, C. Guyot, G. Gabbiani, The stroma reaction myofibroblast: Piekielko-Witkowska, Expression of genes involved in cellular adhesion and 757 672 a key player in the control of tumor cell behavior, Int. J. Dev. Biol. 48 (2004) ECM-remodelling correlates with poor survival of renal cancer patients, J. Urol. 758 673 509–517. 195 (2016) 1892–1902. 759 674 [4] H.E. Barker, T.R. Cox, J.T. Erler, The rationale for targeting the LOX family in [32] D.J. Cheon, Y. Tong, M.S. Sim, J. Dering, D. Berel, X. Cui, et al., A collagen- 760 675 cancer, Nat. Rev. Cancer 12 (2012) 540–552. remodeling gene signature regulated by TGF-beta signaling is associated with 761

12 D. Jia et al. / Cancer Letters ■■ (2016) ■■–■■

762 metastasis and poor survival in serous ovarian cancer, Clin. Cancer Res. 20 (2014) [48] R. Cristescu, J. Lee, M. Nebozhyn, K.M. Kim, J.C. Ting, S.S. Wong, et al., Molecular 808 763 711–723. analysis of gastric cancer identifies subtypes associated with distinct clinical 809 764 [33] Y.H. Wu, T.H. Chang, Y.F. Huang, H.D. Huang, C.Y. Chou, COL11A1 promotes tumor outcomes, Nat. Med. 21 (2015) 449–456. 810 765 progression and predicts poor clinical outcome in ovarian cancer, Oncogene [49] J. Guinney, R. Dienstmann, X. Wang, A. de Reynies, A. Schlicker, C. Soneson, et al., 811 766 33 (2014) 3432–3440. The consensus molecular subtypes of colorectal cancer, Nat. Med. 21 (2015) 812 767 [34] J.A. Beach, P.J. Aspuria, D.J. Cheon, K. Lawrenson, H. Agadjanian, C.S. Walsh, et al., 1350–1356. 813 768 Sphingosine kinase 1 is required for TGF-beta mediated fibroblast-to- [50] A.P. Bento, A. Gaulton, A. Hersey, L.J. Bellis, J. Chambers, M. Davies, et al., The 814 769 myofibroblast differentiation in ovarian cancer, Oncotarget 7 (2016) 4167–4182. ChEMBL bioactivity database: an update, Nucleic Acids Res. 42 (2014) D1083– 815 770 [35] G. Shin, T.W. Kang, S. Yang, S.J. Baek, Y.S. Jeong, S.Y. Kim, GENT: gene expression D1090. 816 771 database of normal and tumor tissues, Cancer Inform. 10 (2011) 149–157. [51] M. Egeblad, M.G. Rasch, V.M. Weaver, Dynamic interplay between the collagen 817 772 [36] A.J. Gentles, S.V. Bratman, L.J. Lee, J.P. Harris, W. Feng, R.V. Nair, et al., Integrating scaffold and tumor evolution, Curr. Opin. Cell Biol. 22 (2010) 697–706. 818 773 tumor and stromal gene expression signatures with clinical indices for survival [52] I. Malanchi, A. Santamaria-Martinez, E. Susanto, H. Peng, H.A. Lehr, J.F. Delaloye, 819 774 stratification of early-stage non-small cell lung cancer, J. Natl. Cancer Inst. 107 et al., Interactions between cancer stem cells and their niche govern metastatic 820 775 (2015). colonization, Nature 481 (2012) 85–89. 821 776 [37] E. Cerami, J. Gao, U. Dogrusoz, B.E. Gross, S.O. Sumer, B.A. Aksoy, et al., The cBio [53] J. Massague, TGFbeta signalling in context, Nat. Rev. Mol. Cell Biol. 13 (2012) 822 777 cancer genomics portal: an open platform for exploring multidimensional cancer 616–630. 823 778 genomics data, Cancer Discov. 2 (2012) 401–404. [54] E.C. Connolly, J. Freimuth, R.J. Akhurst, Complexities of TGF-beta targeted cancer 824 779 [38] T.Z. Tan, H. Yang, J.R. Ye, J. Low, M. Choolani, D.S.P. Tan, et al., CSIOVDB: a therapy, Int. J. Biol. Sci. 8 (2012) 964–978. 825 780 microarray gene expression database of epithelial ovarian cancer subtype, [55] S. Kubota, M. Takigawa, Cellular and molecular actions of CCN2/CTGF and its 826 781 Oncotarget 6 (2015) 43843–43852. role under physiological and pathological conditions, Clin. Sci. 128 (2015) 827 782 [39] S. Julien, A. Merino-Trigo, L. Lacroix, M. Pocard, D. Goere, P. Mariani, et al., 181–196. 828 783 Characterization of a large panel of patient-derived tumor xenografts [56] M. Johannsen, G. Spitaleri, G. Curigliano, J. Roigas, S. Weikert, C. Kempkensteffen, 829 784 representing the clinical heterogeneity of human colorectal cancer, Clin. Cancer et al., The tumour-targeting human L19-IL2 immunocytokine: preclinical safety 830 785 Res. 18 (2012) 5314–5328. studies, phase I clinical trial in patients with solid tumours and expansion into 831 786 [40] T.L. Yeung, C.S. Leung, K.K. Wong, G. Samimi, M.S. Thompson, J. Liu, et al., patients with advanced renal cell carcinoma, Eur. J. Cancer 46 (2010) 2926–2935. 832 787 TGF-beta modulates ovarian cancer invasion by upregulating CAF-derived [57] T.K. Eigentler, B. Weide, F. de Braud, G. Spitaleri, A. Romanini, A. Pflugfelder, 833 788 versican in the tumor microenvironment, Cancer Res. 73 (2013) 5016–5028. et al., A dose-escalation and signal-generating study of the immunocytokine 834 789 [41] K. Lawrenson, B. Grun, N. Lee, P. Mhawech-Fauceglia, J. Kan, S. Swenson, et al., L19-IL2 in combination with dacarbazine for the therapy of patients with 835 790 NPPB is a novel candidate biomarker expressed by cancer-associated fibroblasts metastatic melanoma, Clin. Cancer Res. 17 (2011) 7732–7742. 836 791 in epithelial ovarian cancer, Int. J. Cancer 136 (2015) 1390–1401. [58] E. Tran, D. Chinnasamy, Z. Yu, R.A. Morgan, C.C. Lee, N.P. Restifo, et al., Immune 837 792 [42] T.Z. Tan, Q.H. Miow, Y. Miki, T. Noda, S. Mori, R.Y. Huang, et al., Epithelial- targeting of fibroblast activation protein triggers recognition of multipotent bone 838 793 mesenchymal transition spectrum quantification and its efficacy in deciphering marrow stromal cells and cachexia, J. Exp. Med. 210 (2013) 1125–1135. 839 794 survival and drug responses of cancer patients, EMBO Mol. Med. 6 (2014) [59] E.W. Roberts, A. Deonarine, J.O. Jones, A.E. Denton, C. Feig, S.K. Lyons, et al., 840 795 1279–1293. Depletion of stromal cells expressing fibroblast activation protein-alpha from 841 796 [43] G. Consortium, The Genotype-Tissue Expression (GTEx) project, Nat. Genet. 45 skeletal muscle and bone marrow results in cachexia and anemia, J. Exp. Med. 842 797 (2013) 580–585. 210 (2013) 1137–1151. 843 798 [44] R. Kalluri, R.A. Weinberg, The basics of epithelial-mesenchymal transition, J. [60] Y. Li, D.A. Lacerda, M.L. Warman, D.R. Beier, H. Yoshioka, Y. Ninomiya, et al., A 844 799 Clin. Invest. 119 (2009) 1420–1428. fibrillar collagen gene, Col11a1, is essential for skeletal morphogenesis, Cell 80 845 800 [45] T. Shay, J. Kang, Immunological Genome Project and systems immunology, (1995) 423–430. 846 801 Trends Immunol. 34 (2013) 602–609. [61] J. Spranger, The type XI collagenopathies, Pediatr. Radiol. 28 (1998) 745–750. 847 802 [46] R.A. Moffitt, R. Marayati, E.L. Flate, K.E. Volmar, S.G. Loeza, K.A. Hoadley, et al., [62] U.H. Weidle, D. Maisel, S. Klostermann, E.H. Weiss, M. Schmitt, Differential 848 803 Virtual microdissection identifies distinct tumor- and stroma-specific subtypes splicing generates new transmembrane receptor and extracellular matrix-related 849 804 of pancreatic ductal adenocarcinoma, Nat. Genet. 47 (2015) 1168–1178. targets for antibody-based therapy of cancer, Cancer Genomics Proteomics 8 850 805 [47] R.G. Verhaak, P. Tamayo, J.Y. Yang, D. Hubbard, H. Zhang, C.J. Creighton, et al., (2011) 211–226. 851 806 Prognostically relevant gene signatures of high-grade serous ovarian carcinoma, 807 J. Clin. Invest. 123 (2013) 517–525.