Fibroblast Growth Factor Receptor 2

Human Cancer Biology

Fibroblast Growth Factor Receptor 2–Positive Fibroblasts Provide a Suitable Microenvironment for Tumor Development and Progression in Esophageal Carcinoma Chunyu Zhang,1,3 Li Fu,4 Jianhua Fu,2 Liang Hu,4 Hong Yang,2 Tie-Hua Rong,2 Yan Li,1 Haibo Liu,1 Song-Bin Fu,3 Yi-Xin Zeng,1 and Xin-Yuan Guan1,4

Abstract Purpose: Tumor fibroblasts (TF) have been suggested to play an essential role in the complex process of tumor-stroma interactions and tumorigenesis. The aim of the present study was to investigate the specific role of TF in the esophageal cancer microenvironment. Experimental Design: An Affymetrix expression microarray was used to compare gene expression profiles between six pairs of TFs and normal fibroblasts from esophageal squamous cell carcinoma (ESCC). Differentially expressed genes were identified, and a subset was evaluated by quantitative real-time PCR and immunohistochemistry. Results: About 43% (126 of 292) of known deregulated genes in TFs were associated with cell proliferation, extracellular matrix remodeling, and immune response. Up-regulation of fibroblast growth factor receptor 2 (FGFR2), which showed the most significant change, was detected in all six tested TFs compared with their paired normal fibroblasts. A further study found that FGFR2-positive fibroblasts were only observed inside the tumor tissues and not in tumor-surrounding stromal tissues, suggesting that FGFR2 could be used as a TF-specific marker in ESCC. Moreover, the conditioned medium from TFs was found to be able to promote ESCC tumor cell growth, migration, and invasion in vitro. Conclusions: Ourstudy providesnew candidate genes forthe esophageal cancermi- croenvironment. Based on our results, we hypothesize that FGFR2(+)-TFs might provide cancer cells with a suitable microenvironment via secretion of proteins that could promote cancer development and progression through stimulation of cancer cell proliferation, induction of angiogenesis, inhibition of cell adhesion, enhancement of cell mobility, and promotion of the epithelial-mesenchymal transition.

Cancer has long been considered a cell-autonomous process in Authors' Affiliations: 1State Key Laboratory of Oncology in Southern China which progressive genetic and epigenetic alterations transform 2 and Department of Surgery, Cancer Center, Sun Yat-Sen University, cells independent of the external context. However, a growing Guangzhou, China; 3Medical Genetics Laboratory, Harbin Medical University, Harbin, China; and 4Department of Clinical Oncology, body of evidence suggests that the tumor microenvironment, University of Hong Kong, Hong Kong, China composed of noncancer cells and their stroma, plays an impor- Received 10/29/08; revised 3/24/09; accepted 3/26/09; published OnlineFirst tant role in cancer development and progression (1, 2). The 6/9/09. interaction between tumor and stromal cells has been implicat- Grant support: National Natural Science Foundation of China grants ed in the regulation of cell growth, invasion, metastasis, angio- 30700462 and 30772475, Major State Basic Research Program of China – grant 2006CB910104, China Postdoctoral Science Foundation grant genesis, and the outcome of therapy (3 7). Tumor stroma 20070410861, Sun Yat-Sen University “Hundred Talents Program” grant includes endothelial cells, immune cells, and fibroblasts. The 85000-3171311, and RGC grants HKU 7656/07M and HKUST2/06C. importance of angiogenesis involving endothelial cells has been The costs of publication of this article were defrayed in part by the payment of widely studied, and antiangiogenic strategies targeting endothe- page charges. This article must therefore be hereby marked advertisement in – accordance with 18 U.S.C. Section 1734 solely to indicate this fact. lial cells have been applied to cancer therapy (8 10). Note: Supplementary data for this article are available at Clinical Cancer Recently, tumor fibroblasts (TF) have been suggested to play Research Online (http://clincancerres.aacrjournals.org/). an essential role in the complex process of tumor-stroma coevo- C. Zhang and L. Fu contributed equally to this work. lution and tumorigenesis (6, 11). Studies of different cancer Requests for reprints: Xin-Yuan Guan, Room 605, State Key Laboratory of types have shown that TFs are located in the vicinity of tumor Oncology in Southern China, Cancer Center, Sun Yat-Sen University, 651 cells and are able to enhance tumor growth by secreting growth Dongfeng Road East, Guangzhou 510060, China. Phone: 852-2819-9785; β Fax: 852-2819-9629; E-mail: [email protected]. factors, such as transforming growth factor- , matrix degrading F 2009 American Association for Cancer Research. enzymes, such as matrix metalloproteinases (MMP), and doi:10.1158/1078-0432.CCR-08-2824 angiogenic factors, such as vascular endothelial growth factor

www.aacrjournals.org 4017 Clin Cancer Res 2009;15(12) June 15, 2009 Downloaded from clincancerres.aacrjournals.org on September 28, 2021. © 2009 American Association for Cancer Research. Human Cancer Biology

Interestingly, comparisons of gene lists have indicated great Translational Relevance heterogeneity in the expression patterns of TFs from different cancer types (21). To date, few studies have evaluated the roles Despite compelling evidence showing the crucial of TFs in esophageal squamous cell carcinoma (ESCC). To ob- role of stromal cells in tumorigenesis, our knowledge tain an accurate overview of the broad range of changes occur- of the genes that mediate changes in the tumormicro- ring in TFs during ESCC, we isolated six pairs of TFs and their environment and tumor-stroma interactions in esoph- matched nontumorous fibroblasts (NF) from six ESCCs. mRNA ageal cancer is limited. Here, we describe a broad from TFs and NFs was pooled, and gene expression profiles range of deregulated genes between fibroblasts iso- were compared using the Affymetrix human genome U133 Plus lated from esophageal squamous cell carcinoma and 2.0 GeneChip. their corresponding nontumorous tissues. Our study finds new candidate genes forthe esophageal cancer microenvironment. These data should, therefore, pro- Materials and Methods vide a valuable resource for future basic and clinical ESCC samples. Six pairs of primary ESCC tissues and surrounding studies addressing the role of tumor-stroma interac- nontumorous esophageal tissues were collected at the time of surgical tions in esophageal cancer. resection at the Cancer Center of Sun Yat-sen University. None of the patients received adjuvant therapy before surgery. Specimens used in this study were approved by the Committees for Ethical Review of (12–16). TFs have a distinctive phenotype when compared with Research at Sun Yat-sen University. Isolation of fibroblasts. Tumor tissue and its paired nontumorous quiescent fibroblasts in differentiated adult tissue, but the mor- esophageal tissue were cut into as small pieces as possible in sterile phologic and functional differences are not completely under- PBS solution, followed by collagenase digestion (0.1% collagenase type stood. Gene expression profiling using cDNA microarray IV, Sigma) at 37°C for 30 min. The suspension was filtered through a technology can vastly aid in the characterization of TFs isolated 20-μm stainless steel wire mesh to collect a single-cell suspension. The from a broad range of solid tumors, including pancreas (17), filtrate was centrifuged at 1,500 rpm for 5 min and washed twice with colon (18), breast (19), and basal cell cancer (20). DMEM before being finally plated on 6-cm tissue culture dishes in

Fig. 1. Characterization of TFs. A, representative cell morphology of NFs, TFs, macrophages, and tumor cells. Immunofluorescence was used to distinguish TFs, NFs, macrophages, and tumor cells with antibodies targeting CD68 (B), E-cadherin (C), cytokeratin (D), and fibronectin (E).

Clin Cancer Res 2009;15(12) June 15, 20094018 www.aacrjournals.org Downloaded from clincancerres.aacrjournals.org on September 28, 2021. © 2009 American Association for Cancer Research. Fibroblasts in Esophageal Cancer

Fig. 2. TFs grow faster than NFs. A, MTT assay was used to compare cell growth rates between TFs and NFs. Points, mean of three independent experiments; bars, SD. B, flow cytometry histogram shows that the percentages of cells in S and G2-M phases were significantly higher in TFs compared with NFs. C and D, two examples of DNA content comparison between NFs, TFs, and tumor cells by flow cytometry.

5 mL DMEM supplemented with 20% fetal bovine serum. After culturing Detection of DNA content by flow cytometry. For flow cytometry anal- for 30 min at 37°C, unadherent cells (mainly tumor cells) were removed ysis, 1 to 2 × 106 cells were fixed in 70% ethanol, stained with propi- to obtain pure fibroblasts because the adhesion time needed for fibro- dium iodide, and DNA content was analyzed using a Cytomics FC blasts is much shorter (∼20-30 min) than that for tumor cells (usually device (Beckman Coulter). Finally, the cell cycle profiles were analyzed >1 h). The adherent fibroblasts were subcultured for further study. with Modfit LT2.0. Todetermine the ploidy pattern of tested cells (NF, Immunofluorescence. Cells were grown on gelatin-coated coverslips TF, and tumor cells), DNA index was calculated based as the ratio of the for 24 h and then fixed in 4% paraformaldehyde at room temperature G1 peak channel of the tested cells to the G1 peak channel of normal for 20 min. Cells were incubated with PBS-B solution at 37°C for peripheral blood lymphocytes. Triplicate independent experiments 30 min to block nonspecific interactions and stained with various were done. ready-to-use primary antibodies targeting fibronectin, E-cadherin, Cell growth assay. The cell growth rate of fibroblasts derived from CD68, and cytokeratin (Dako) at 4°C overnight. After several washes ESCC tumor tissue and nontumorous tissue were determined by in PBS, cells were incubated with optimal concentrations of FITC- 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) labeled goat anti-rabbit secondary antibody (Dako) at room tempera- assay. Cells were seeded into 96-well plates at a density of 1 × 103 ture for 1 h. Antifade 4′,6-diamidino-2-phenylindole solution was per well. The MTT kit (Sigma) was used according to the manufacturer's added, and images were obtained. instructions. Triplicate independent experiments were done.

www.aacrjournals.org 4019 Clin Cancer Res 2009;15(12) June 15, 2009 Downloaded from clincancerres.aacrjournals.org on September 28, 2021. © 2009 American Association for Cancer Research. Human Cancer Biology

RNA (2 μg) was done using an Advantage RTfor PCR kit (Clontech). For quantitative real-time PCR (qPCR), pooled RNA from NFs and TFs used for microarray analysis was subjected to reverse transcription followed by qPCR using a SYBR Green PCR kit (Applied Biosystems). Am- plification protocol consisted of incubations at 95°C for 15 s, 60°C for 1 min, and 72°C for 1 min for 40 cycles. Quantification was done using the ABI PRISM 7900HTSequence Detection System (Applied Biosys- tems). Three independent reverse transcriptions were done using pooled RNA from NFs and TFs, and at least three qPCR reactions were done using each RTproduct. Thecomparative CT method (ΔÄCT method) was used to determine the quantity of the target sequences in TFs relative both to NFs (calibrator) and to an endogenous control (GAPDH). Rel- ative expression level was presented as the relative fold change and cal- −ΔΔ − Δ − Δ culated using the following formula: 2 CT =2[ CT(TF) CT(NF)] and target GAPDH each ΔCT = Δ CT − ΔCT . Immunohistochemistry. Immunohistochemistry was done using a standard streptavidin-biotin-peroxidase complex method. In brief, paraffin block sections of esophageal cancers were deparaffinized and re- hydrated. Endogenous peroxidase activity was blocked with 0.3% hydrogen peroxide for 20 min. For antigen retrieval, slides were heated in a microwave oven for 10 min in 10 mmol/L citrate buffer at pH 6.0. Nonspecific binding was blocked with 10% normal rabbit serum for 10 min at room temperature. Sections were incubated with a monoclo- nal antibody against fibroblast growth factor receptor 2 (FGFR2; clone 41D, Upstate Biotechnology) at a dilution of 1:500 at 37°C overnight. One percent bovine serum albumin diluted in PBS was included as a negative control. After a brief wash, slides were incubated with a biotinylated rabbit antimouse immunoglobulin at a concentration of 1:100 for 30 min at room temperature. Subsequently, slides were incubated with streptavidin-peroxidase conjugate for 45 min at room temperature and developed with diaminobenzidine tetrahydrochloride. For immunofluorescent double-labeling, paraffin block sections of esophageal cancers were treated as described above. Sections were incubated with a mouse antibody against FGFR2 and a rabbit antibody against fibronectin simultaneously at 37°C overnight. After brief washing, slides were incubated with FITC goat anti-mouse (ZSGB- BIO) and TRITC goat anti-rabbit (ZSGB-BIO) secondary antibodies at Fig. 3. Validation of differentially expressed genes detected by expression room temperature for 1 h. After brief washing, slides were counter- array. A, relative expression levels of FGFR2, WNT2, WNT5A, WISP1, stained with 4′,6-diamidino-2-phenylindole in antifade solution. MMP1, MMP10, IGFBP2, IL8, FZD9, CD14, IL32, CXCL14, PKP2, ISL1, HIST1H4K, BRCA2, BRCA1, CCNA2, CCNB2, and CDC2 analyzed by qPCR in Preparation of conditioned medium. ESCC cells (KYSE30 and TFs relative to that in NFs. NF was used as a calibrator. The relative fold KYSE510) and fibroblasts (TFs and NFs) were seeded into T75 culture changes of down-regulated genes are represented as negative inverse of flask and grown in normal growth media (10 mL of DMEM containing Δ – the ÄCT. Error bar represents SD. B, reverse transcription PCR was 10% fetal bovine serum) for 48 h to obtain approximate 80% conflu- applied to compare expression levels of WNT2, WISP1, WNT5A, IGFBP2, MMP1, MMP10, and IL8 between five pairs of NFs and TFs. 18S rRNA was ence. Culture medium was then collected from each flask and centri- used as an internal control. fuged at 1,000 × g for 30 min, and the supernatant was collected as conditioned medium (CM) for further study. Microarray analysis. Affymetrix human genome U133 Plus 2.0 Effect of CM from fibroblasts on the growth of ESCC cells. To examine GeneChip (Affymetrix), covering 47,000 transcripts and variants, was the effect of CM from fibroblasts on growth of ESCC cells, KYSE30 or used to identify differentially expressed genes between TFs and NFs. Mi- KYSE510 cells were seeded into 96-well plate at a density of 1.5 × 103 croarray reaction was done according to the manufacturer's instructions. per well and cultured in CM from NFs and TFs. CM from KYSE30 or Briefly, 2 μg total RNA pooled from six TFs (or their paired NFs) were KYSE510 cells was used as a control. The cell growth rate was deter- reversed transcripted, labeled, and hybridized to the chip at 42°C for mined by MTT assay as described previously. Triplicate independent ex- 16 h. After wash, the chip was scanned and visualized using a GeneArray periments were done. scanner (Hewlett-Packard). Probe set intensities were calculated using Effect of CM from fibroblasts on the migration of ESCC cells. Cell mo- the Microarray Analysis Suite (MAS v5.0, Affymetrix) software and bility was assessed by a scratch wound-healing assay. KYSE30 or normalized against 100 housekeeping genes to a mean intensity of KYSE510 cells were grown to confluence in a six-well plate with normal 2,000 in mask files of U133 Plus 2.0 before further statistical analysis. growth media. The cell monolayer was mechanically scratched with a Normalized expression values were then compared between TFs and sterile pipette tip. Subsequently, the cells were incubated with CM from NFs. Fold-change differences were calculated to identify up-regulated TFs or NFs instead of normal growth media. CM from KYSE30 or and down-regulated genes. Transcripts with more than a 1.5-fold differ- KYSE510 cells was used as a control. Cell mobility in terms of wound ence in expression level were defined as differentially expressed. Specif- closure was measured by photographing at three random fields at time ically designed online tools, including FatiGO (22), Gene Ontology points 0 and 24 h. (23) provided by the GO Consortium, and NetAffx Analysis Center Effect of CM from fibroblasts on the invasion of ESCC cells. Invasion database (24), were used to classify the functional roles of the identified assay was done with the BioCoat Matrigel Invasion Chamber (BD Bios- differentially expressed genes. ciences) according to the manufacturer's instructions. Briefly, KYSE30 or Quantitative real-time PCR. Total RNA was extracted from fibro- KYSE510 cells (2.5 × 105) were placed in the upper compartment of blasts using TRIzol reagent (Invitrogen). Reverse transcription of total each chamber. Lower compartment was filled with CM from KYSE30

Clin Cancer Res 2009;15(12) June 15, 20094020 www.aacrjournals.org Downloaded from clincancerres.aacrjournals.org on September 28, 2021. © 2009 American Association for Cancer Research. Fibroblasts in Esophageal Cancer

Table 1. Genes differentially expressed between TFs and NFs

Gene name Log ratio* Genbank no.

Cell proliferation FGFR2 Fibroblast growth factor receptor 2 6.8 NM_022969.1 RGS1 Regulator of G-protein signaling 1 5.4 S59049.1 TFAP2C Transcription factor AP-2 gamma 5.2 U85658.1 FGF18 Fibroblast growth factor 18 4.0 BC006245.1 RSG4 Regulator of G-protein signaling 4 3.6 AL514445 WISP1 WNT1 inducible signaling pathway protein 1 3.1 NM_003882.1 RASGRP2 RAS guanyl releasing protein 2 3.0 NM_005825.1 TAF4 TATA box binding protein (TBP)-associated factor 4 3.0 NM_003185.1 TFAP2A Transcription factor AP-2α 2.6 BF343007 HBEGF Heparin-binding EGF-like growth factor 2.5 NM_001945.1 WNT2 Wingless-type MMTV integration site family member 2 2.4 NM_003391.1 FRS2 Fibroblast growth factor receptor substrate 2 2.4 NM_006654.1 LEF1 lymphoid enhancer-binding factor 1 2.2 AF288571.1 WNT5A Wingless-type MMTV integration site family, member 5A 2.1 NM_003392.1 IGFBP2 Insulin-like growth factor binding protein 2 2.0 NM_000597.1 CDK5R1 Cyclin-dependent kinase 5, regulatory subunit 1 (p35) 2.0 AL567411 PDGFA Platelet-derived growth factor α polypeptide 1.9NM_002607.1 AREG Schwannoma-derived growth factor 1.5 NM_001657.1 RGS5 Regulator of G-protein signaling 5 1.5 AF493929.1 NF2 Neurofibromin 2 −3.6 NM_000268.1 CENPF Centromere protein F −2.7 NM_016343.1 CENPE Centromere protein E −2.0 NM_001813.1 BRCA2 Breast cancer 2, early onset −2.0 NM_000059.1 BRCA1b Breast cancer 1, transcript variant BRCA1b −2.0 NM_007295.1 BRCA1 Breast cancer 1, early onset −1.7 NM_007295.1 CEP55 centrosomal protein 55kDa −1.7 NM_018131.1 RASSF2 Ras association (RalGDS/AF-6) domain family 2 −1.6 NM_014737.1 CENPF Centromere protein F −1.6 NM_005196.1 MAD2L1 MAD2 (mitotic arrest deficient, yeast, homologue)-like 1 −1.5 NM_002358.2 CDKN3 Cyclin-dependent kinase inhibitor 3 −1.5 AF213033.1 Extracellular matrix PKP2 Plakophilin 2 6.3 NM_004572.1 FBN2 Fibrillin 2 4.4 X62009.1 NRXN3 Neurexin 3 3.9AF123462 ICAM1 Intercellular adhesion molecule 1 (CD54) 3.6 AI608725 CLDN1 Claudin 1 3.5 AF101051.1 COL10A1 Collagen, type X, α1 3.4 AI376003 CDH1 Cadherin 1 3.3 NM_004360.1 CLDN14 Claudin 14 3.3 AF314090.1 ICAM4 Intercellular adhesion molecule 4 2.5 NM_001544.2 ADAMTS9ADAM metallopeptidase with thrombospondin type 1 motif, 9 2.5 AB037733.1 TNC Tenascin C 2.3 BF434846 LAMB3 Laminin, β3 2.2 L25541.1 COL11A1 Collagen, type XI, α1 2.1 NM_001854.1 COL4A6 Collagen, type IV, α6 2.0 AI889941 MXRA5 Matrix-remodeling associated 5 1.9AF245505.1 FZD8 Frizzled homologue 8 1.9AB043703.1 MMP10 Matrix metallopeptidase 10 1.8 NM_002425.1 CDH13 Cadherin 13 1.7 NM_001257.1 FN1 Fibronectin 1 1.7 W73431 MFAP2 Microfibrillar-associated protein 2 1.6 NM_017459.1 MMP1 Matrix metallopeptidase 1 1.6 NM_002421.2 MATN2 Matrilin 2 1.5 NM_002380.2 COL7A1 Collagen, type VII, α1 1.5 NM_000094.1 ADAM19ADAM metallopeptidase domain 19 1.5 AF311317.1 COL5A3 Collagen, type V, α3 1.5 NM_015719.1 Immune response IL32 Interleukin 32 5.6 NM_004221.1 CD14 CD14 molecule 5.3 NM_000591.1 IGKC Immunoglobulin kappa constant 4.9BC005332.1 HLA-DRA Major histocompatibility complex, class II, DRα 4.6 M60334.1 IFI6 Interferon, α-inducible protein 6 4.0 M87789.1 NFATC1 Nuclear factor of activated T-cells, calcineurin-dependent 1 3.8 AW027545 IL8 Interleukin 8 3.8 AF043337.1 CXCL14 Chemokine (C-X-C motif) ligand 14 3.8 AF144103.1

(Continued to the following page)

www.aacrjournals.org 4021 Clin Cancer Res 2009;15(12) June 15, 2009 Downloaded from clincancerres.aacrjournals.org on September 28, 2021. © 2009 American Association for Cancer Research. Human Cancer Biology

Table 1. Genes differentially expressed between TFs and NFs (Cont'd)

Gene name Log ratio* Genbank no.

IL24 Interleukin 24 3.6 NM_006850.1 IL1A Interleukin 1, α 3.6 M15329.1 IL1B Interleukin 1, β 3.3 NM_000576.1 S100A8 S100 calcium binding protein A8 3.0 NM_002964.2 TNFSF4 Tumor necrosis factor superfamily, member 4 2.8 NM_003326.1 IGLJ3 Immunoglobulin lambda joining 3 2.8 X57812.1 CF1 Complement factor I 2.7 BC020718.1 CCL5 Chemokine (C-C motif) ligand 5 2.6 NM_002985.1 TNFRSF13B Tumor necrosis factor receptor superfamily, member 13B 2.5 NM_012452.1 FCGR1A Fc fragment of IgG, high affinity Ia, receptor (CD64) 2.5 X14355.1 IL11 Interleukin 11 2.4 NM_000641.1 TNFSF18 Tumor necrosis factor superfamily, member 18 2.1 NM_005092.1 IFNE1 Interferon epsilon 1 2.1 NM_176891.1 CXCL2 Chemokine (C-X-C motif) ligand 2 1.5 M57731.1

*Ratio of TF against NF expressed in log2-tranformed values.

(or KYSE510), NFs, and TFs. After 22 h, cancer cells that had penetrated Genes differentially expressed between TFs and NFs. The spe- the Matrigel-coated membrane and passed into the lower compartment cific roles of TFs in ESCC development and progression were were stained and counted. Triplicate independent experiments were explored by comparing genes differentially expressed between done. TFs and NFs. Gene expression profiles of pooled TFs (from Statistical analysis. Statistical analysis was done using SPSS standard six ESCCs) and their pooled normal counterparts were obtained version 13.0 software. Data are expressed as means ± SD from at least by microarray analysis using Affymetrix human genome U133 three independent determinations. Significance of difference was analyzed using Student's t tests. Differences were considered significant for plus 2.0 arrays. A total of 25,206 probe sets (transcripts) were P < 0.05. present in ESCC-derived fibroblasts relative to their normal controls. Compared with the expression profile of NFs, 423 up-regulated genes (0.9%) and 214 down-regulated genes Results (0.45%) were detected in TFs using an arbitrary cutoff line of signal log ratio of ≥1.5 or ≤−1.5. The functional roles of 292 Isolation of TFs and NFs. TFs and their paired NFs were suc- of these genes (181 up-regulated and 111 down-regulated) cessfully isolated from six primary ESCCs by short-term pri- are known (see Supplementary Table S1). mary culture in DMEM supplemented with 10% fetal bovine The function of 126 differentially expressed genes (of 292, serum. Both TFs and NFs could be grown for at least 10 pas- 43.2%) in TFs were associated with cell proliferation (61 genes; sages. TFs and NFs both showed spindle-like morphology Supplementary Table S2), extracellular matrix remodeling (40 that was different from that of macrophages and tumor cells genes; Supplementary Table S3), and the immune response (25 genes; Supplementary Table S4). To validate the microarray (Fig. 1A). data, the expression level of 20 randomly selected genes, in- Characterization of TFs. To confirm that TFs derived from tu- cluding 13 up-regulated genes and 7 down-regulated genes, mor tissue were pure fibroblasts without macrophage or tumor were compared by qPCR between TFs and NFs. Compared with cell contamination, several cell markers, including the macro- NFs, the expression pattern of all differentially expressed genes phage-specific membrane marker CD68, epithelial cell markers in TFs detected by qPCR, except CCNB2, was consistent with E-cadherin and cytokeratin, and the fibroblast marker fibronec- the microarray results (Fig. 3A). tin, were used to distinguish fibroblasts from macrophages and TF is able to stimulate tumor cell growth. Most genes up-reg- tumor cells. The results showed that the TFs were only stained ulated in TFs have oncogenic function, including growth factors by the fibroblast-specific marker fibronectin (Fig. 1B-E). Flow (FGF18 and HBEGF), transcription regulators (TFAP2A, cytometry analysis also indicated that the DNA ploidy pattern TFAP2C, and TAF4), and members of the Wnt pathway (WISP1, of both NF and TF are diploid based on DNA index calculation. WNT2, WNT5A,andLEF1). Additionally, several tumor sup- The average DNA indices for NF (1.024) and TF (1.036) were pressor genes were down-regulated in TFs, including BRCA1, obviously lower than that from their paired tumor cells BRCA2, NF2,andMAD2L1 (Table 1). Several genes encoding (1.306). All these data showed that fibroblasts isolated from tu- secreted proteins, such as WNT2, WNT5A, WISP1, and IGFBP2, mor tissue were pure TFs. were up-regulated in TFs (Fig. 3B). All of these secreted proteins TFs grew faster than NFs. The cell growth assay revealed that are able to induce cell proliferation, implying that TFs may pro- the cell growth rate of TFs was significantly higher than that of mote tumor cell growth. NFs in the same culture conditions (P < 0.05; Fig. 2A). Flow TF is able to facilitate tumor cell invasion and metastasis. Our cytometry study showed that the percentage of cells in the S microarray results suggest that TFs may facilitate invasive and G2-M phases was significantly higher in TFs (S, 23.8 ± growth and metastasis of ESCC tumor cells mainly via the fol- 6.1; G2-M,20.6±7.8)comparedwithNFs(S,7.2±2.3; lowing two mechanisms: (a) altering tumor microenvironment G2-M, 9.8 ± 2.2, P < 0.05), suggesting that TFs had a stronger interactions (tumor-stroma and tumor-tumor) by decreasing capacity for proliferation than their paired NFs (Fig. 2B-D). cell adhesion (MMP1, MMP10, and MXRA5) and increasing cell

Clin Cancer Res 2009;15(12) June 15, 20094022 www.aacrjournals.org Downloaded from clincancerres.aacrjournals.org on September 28, 2021. © 2009 American Association for Cancer Research. Fibroblasts in Esophageal Cancer mobility (LAMB3, CLDN1,andCLDN14)and(b)promoting FGFR2 is a TF-specific marker. FGFR2 showed the greatest the epithelial-mesenchymal transition (EMT) in tumor cells expression difference between TFs and NFs, which was con- (LEF1, FN1, TNC, and FZD8; Table 1). firmed by reverse transcription–PCR (Fig. 4A), Western blot TF provides a suitable microenvironment for tumors. Immune- analysis (Fig. 4B), and immunostaining with an anti-FGFR2 related genes, including various chemokines, are also differen- antibody (Fig. 4C). We further investigated the distribution tially expressed in TFs. Besides participating in immune of FGFR2-positive cells in ESCCs and normal esophageal tis- responses and inflammatory processes, chemokines secreted sues. FGFR2(+) cells were not observed in any of 12 normal by stromal cells surrounding tumor tissue have been associated esophageal tissues (Fig. 4D). In ESCCs, FGFR2(+) cells were with angiogenesis, tumor cell growth, invasion, and migration detected in 18 of 20 tumor tissues, but not in surrounding via cross-talk between tumor and stromal cells. In this study, stromal tissues including fibroblasts (Figs. 4E and 5A). The several up-regulated genes in TFs, including interleukin 8 (IL8), distribution of FGFR2-positive cells in tumor tissue was S100A8, CXCL14, and CCL5, have been shown to contribute to scattered, and only 0.2% to 2% cells in tumor tissues were suitable microenvironments for tumor cells (Table 1). FGFR2-positive (Fig. 5A). Fibroblasts, positive with fibronectin

Fig. 4. FGFR2 is a TF-specific marker. A, reverse transcription–PCR was used to compare FGFR2 expression between six pairs of NFs and TFs. 18S rRNA was used as an internal control. B, Western blot analysis was used to compare FGFR2 expression between six pairs of NFs and TFs. β-Actin was used as an internal control. C, FGFR2 was only detected in TFs and not in NFs by immunostaining with a FGFR2 antibody (left). The fibroblast-specific marker fibronectin was detected in both NFs and TFs (right). D, fibroblasts in normal esophageal tissue expressed fibronectin (right) but not FGFR2 (left). E, representative distribution of FGFR2-positive cells (left, arrows) and fibronectin-positive cells (right, indicated by arrows) in ESCC tumor tissues.

www.aacrjournals.org 4023 Clin Cancer Res 2009;15(12) June 15, 2009 Downloaded from clincancerres.aacrjournals.org on September 28, 2021. © 2009 American Association for Cancer Research. Human Cancer Biology

Fig. 5. Detection of FGFR2-positive cells in tumortissue. A, FGFR2(+) cells were only detected inside tumor tissue (red arrows) but not in tumor-surrounding stromal tissue (black arrows). B, expression of fibronectin could be detected in both TFs inside tumor tissue and NFs in tumor-surrounding stromal tissue. C, representative images of two TFs and one NF labeled by antibodies against fibronectin (red) and FGFR2 (green). Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (blue). D, representative immunofluorescent image of an ESCC stained with antifibronectin (green) and anti-FGFR2 (red) antibodies. FGFR2(+) TFs could be only observed in TFs inside tumor tissue (yellow color signal indicated by an arrow). Fibroblasts in tumor-surrounding stromal tissue all expressed fibronectin. E, representative immunofluorescent image of an ESCC stained with anti-CD68 (green) and anti-FGFR2 (red) antibodies. FGFR2(+) TFs could not be stained by macrophage marker CD-68. staining, were detected in both tumor tissues and their sur- The effect of CM from fibroblasts on the cell mobility of rounding stromal tissues (Fig. 5B). Interestingly, the distribu- ESCC cells was tested by wound-healing scratch assay. As tion of fibronectin-positive cells in tumor tissue was also shown in Fig. 6B, KYSE30 or KYSE510 cells cultured with TF scattered with the frequency of 1% to 5%, similar to that of medium covered the scratched area with migrating cells within FGFR2-positive cells. To determine whether FGFR2-positive 24 hours, but NF medium did not. Subsequently, we measured cells in tumor tissue were fibroblasts, double immunofluores- the capacity of ESCC cells to invade through Matrigel, an artifi- cence staining with antibodies against FGFR2 (green) and cial extracellular matrix, after incubation with TF medium or NF fibronectin (red) was done. Most FGFR2-positive cells were medium. KYSE30 and KYSE510 cells cultured with TF medium also stained by the antifibronectin antibody (Fig. 5C and showed an increased invasion by 79% and 78%, respectively D). To rule out that FGF2-positive cells within tumors could (Fig. 6C). be macrophages, double immunofluorescence staining with antibodies against FGFR2 (red) and CD68 (green) was done. Discussion The result showed that FGFR2-positive cells could not be stained by the anti-CD68 antibody (Fig. 5E), suggesting that Increasing evidence suggests that the cancer microenviron- FGFR2-positive cells were TFs. ment plays important roles in tumor development and progres- CM from TFs promotes cell growth, migration, and invasion in sion via cross-talk between neoplastic and stromal cells (3–7, ESCC cells. To study if the secreting proteins produced by TFs 11–16). In the present study, we isolated and characterized could promote tumorigenic behaviors of ESCC cells, we cul- TFs from esophageal cancer tissue. The cell growth rate of TFs tured the ESCC cells (KYSE30 or KYSE510) with CM from TFs was significantly higher (P < 0.05) than that of paired fibro- (TF medium) or NFs (NF medium). Cell growth assay revealed blasts from nontumorous tissue. Flow cytometry revealed that that the TF medium could significantly increase cell growth the percentage of proliferating cells (S and G2-M phases) was rates of KYSE30 or KYSE510 cells compared with the NF medi- also significantly higher in TFs than in NFs (P < 0.05). um. In contrast, control CM from KYSE30 or KYSE510 did not To explore the role of TFs in tumor development and progres- show significant difference in cell growth rate compared with sion, universal gene expression profiles were compared between NF medium (Fig. 6A). TFs and NFs by expression array, and 637 genes were found to

Clin Cancer Res 2009;15(12) June 15, 20094024 www.aacrjournals.org Downloaded from clincancerres.aacrjournals.org on September 28, 2021. © 2009 American Association for Cancer Research. Fibroblasts in Esophageal Cancer be differentially expressed in TFs. Of 292 differentially ex- Because WNT2, WNT5A, and WISP1 are genes encoding secret- pressed genes with known function, 43.2% were associated ed proteins, it is likely that up-regulation of these genes in TFs with cell proliferation, extracellular matrix remodeling, and promotes tumor cell proliferation and metastasis. the immune response. Several members of Wnt pathway, in- The second group of up-regulated genes in TFs was associated cluding WISP1, WNT2, WNT5A, and LEF1, were up-regulated with extracellular matrix remodeling, including genes whose in TFs, suggesting that Wnt signaling might be involved in TF products can decrease cell adhesion (e.g., MMP1, MMP10, proliferation. Constitutive activation of the Wnt/β-catenin sig- and MXRA5; refs. 29, 30), increase cell mobility (e.g., LAMB3, naling pathway has been found in most colorectal tumors due CLDN1, and CLDN14; refs. 31, 32), and induce the EMTin to the mutation of the APC gene (25). WISP-1, a CCN family tumor cells (LEF1, FN1, TNC,andFZD8;refs.33–35). The member regulated by the Wnt/β-catenin signaling pathway, disruption of tissue architecture is a key step in metastasis, can activate the antiapoptotic Akt/PKB signaling pathway and which involves the modulation of cell-matrix and cell-cell up-regulate antiapoptotic Bcl-X (26). Other members of contacts by dysregulation of extracellular proteinases, includ- Wnt pathway, such as WNT5A, WNT2, and WNT2B, have also ing MMPs (36). Another key issue of metastasis is EMT, been associated with cancer invasion and metastasis (27, 28). which causes tumor cell becoming migratory and invasive

Fig. 6. Effects of CM from fibroblasts on ESCC tumor cell growth, migration, and invasion. A, MTT assay was used to compare effect of TF medium, NF medium, and control medium on the growth of ESCC tumor cells. Points, mean of three independent experiments; bars, SD. B, wound healing assay showed that closing of scratch wound in TF medium incubated KYSE30 cells (left) or KYSE510 cells (right) was nearly completed within 24 h but not in NF medium cultured tumor cells. C, representative images showed the KYSE30 or KYSE510 cells that invaded through the Matrigel when incubated with TF medium or NF medium. Representative histogram of invaded tumor cells was displayed and number of invaded tumor cells quantified. *, P < 0.05, NF medium compared with TF medium.

www.aacrjournals.org 4025 Clin Cancer Res 2009;15(12) June 15, 2009 Downloaded from clincancerres.aacrjournals.org on September 28, 2021. © 2009 American Association for Cancer Research. Human Cancer Biology

(37). One study found that LEF-1 can induce EMTdirectly an intracellular tyrosine kinase domain (44). Recently, activated by β-catenin and increase tumor metastasis (33). FGFR2 was found to play an important role in mesenchymal- In another study, FN1 was shown to be associated with en- to-epithelialtransition, the critical process of secondary tumor hanced cell invasion during EMT(34). Thesestudies strongly growth in metastatic organs (45). In the present study, suggest that TF may play important roles in helping tumor FGFR2-positive cells were only present in tumor tissues but cell invasion and metastasis. not in surrounding nontumor tissues. This finding suggests that The third group of up-regulated genes in TFs was associated FGFR2-positive cells are TFs, and FGFR2 can be used as TF- with the immune response. Previously, chemokines were be- specific marker. We, therefore, hypothesize that FGFR2-positive lieved to primarily participate in immune and inflammatory fibroblasts are recruited by tumor cells during the early stage of processes and to regulate leukocyte trafficking by stimulating tumor development. Once FGFR2-positive fibroblasts settled in directional chemotaxis. Recently, increasing evidence has sug- the tumor tissue, they interact with their surrounding cancer gested that chemokines, such as IL8, S100A8, CXCL14, and cells and provide these cells with a suitable environment by se- CCL5, produced by stromal cells surrounding tumors are asso- creting various proteins that could promote cancer develop- ciated with tumor development and progression via cross-talk ment and progression. Such activity may occur (a)to between tumor and stromal cells. IL8 has been shown to play promote cell proliferation by secreting growth factors (FGF18 important roles in angiogenesis and metastasis (38), and neu- and HBEGF), transcription regulators (TFAP2A, TFAP2C,and tralization of IL8 by an anti-IL8 antibody has been shown to TAF4), and members of the Wnt pathway (WISP1, WNT2, inhibit angiogenesis, tumor growth, and metastasis in malig- and WNT5A); (b) to induce angiogenesis by secretion of nant melanoma (39). One recent report found that S100A8 IL8; (c) to inhibit cell adhesion by secreting MMP1, MMP10, secreted by a distant primary tumor could attract Mac 1(+)- and MXRA5;(d) to enhance cell mobility by secreting LAMB3, myeloid cells to the premetastatic lung, suggesting that CLDN1, and CLDN14; and (e) to promote the EMTby secretion S100A8 plays a role in long distance metastasis (40). Another of LEF1, FN1, TNC, and FZD8. These findings provide us new study found that up-regulation of S100A8 could be detected lines of evidence that many factors secreted by TFs could lead to only in tumor-associated stroma and not in malignant epithelia a rapid development of esophageal cancer. Further characteriza- (41). These findings suggest that up-regulation of immune re- tion of the roles these proteins play in cancer development and sponse genes in TFs could provide a suitable microenvironment progression will reveal the molecular mechanisms of how TFs for tumor cells. provide cancer cells with a suitable microenvironment and may In the present study, FGFR2 showed the greatest expression result in the development of new therapeutic targets for cancer difference between TFs and NFs. Fibroblast growth factors and treatment. their receptors are believed to play important roles in multiple biological activities, including cell proliferation, differentiation, Disclosure of Potential Conflicts of Interest and migration (42, 43). FGFR is composed of an extracellular immunoglobulin-like domain, a transmembrane segment, and No potential conflicts of interest were disclosed.

References 1. Bissell MJ, Radisky D. Putting tumours in con- proaches for drug delivery. Nat Clin Pract derived stromal fibroblasts. Breast Cancer Res text. Nat Rev Cancer2001;1:46 –54. Oncol 2006;3:682–92. Treat 2008;110:273–81. 2. Kalluri R, Zeisberg M. Fibroblasts in cancer. Nat 11. De Wever O, Mareel M. Role of tissue stroma in 20. Micke P, Kappert K, Ohshima M, et al. In situ Rev Cancer2006;6:392 –401. cancercell invasion. J Pathol 2003;200:429 –47. identification of genes regulated specifically in 3. Gleave M, Hsieh JT, Gao CA, von Eschenbach 12. Tlsty TD. Stromal cells can contribute oncogen- fibroblasts of human basal cell carcinoma. J In- AC, Chung LW. Acceleration of human prostate ic signals. Semin CancerBiol 2001;11:97 –104. vest Dermatol 2007;127:1516–23. cancergrowth in vivo by factors produced by 13. Elenbaas B, Weinberg RA. Heterotypic signal- 21. Micke P, Ostman A. Exploring the tumour envi- prostate and bone fibroblasts. Cancer Res ing between epithelial tumorcells and fibro- ronment: cancer-associated fibroblasts as targets 1991;51:3753–61. blasts in carcinoma formation. Exp Cell Res in cancertherapy.ExpertOpin TherTargets 2005; 4. Bhowmick NA, Chytil A, Plieth D, et al. TGF-β 2001;264:169–84. 9:1217–33. signaling in fibroblasts modulates the oncogenic 14. Nakayama H, Enzan H, Miyazaki E, Naruse K, 22. Al-Shahrour F, Diaz-Uriarte R, Dopazo J. Fati- potential of adjacent epithelia. Science 2004;303: Kiyoku H, Hiroi M. The role of myofibroblasts at GO: a web tool forfinding significant associa- 848–51. the tumorborderof invasive colorectal adeno- tions of Gene Ontology terms with groups of 5. Muerkoster S, Wegehenkel K, Arlt A, et al. Tu- carcinomas. Jpn J Clin Oncol 1998;28:615–20. genes. Bioinformatics 2004;20:578–80. mor stroma interactions induce chemoresis- 15. Tuxhorn JA, Ayala GE, Smith MJ, Smith VC, 23. Gene Ontology Consortium. Creating the gene tance in pancreatic ductal carcinoma cells Dang TD, Rowley DR. Reactive stroma in human ontology resource: design and implementation. involving increased secretion and paracrine ef- prostate cancer: induction of myofibroblast phe- Genome Res 2001;11:1425–33. fects of nitric oxide and interleukin-1β.Cancer notype and extracellular matrix remodeling. Clin 24. Liu G, Loraine AE, Shigeta R, et al. NetAffx: Af- Res 2004;64:1331–7. CancerRes 2002;8:2912 –23. fymetrix probesets and annotations. Nucleic 6. Micke P, Ostman A. Tumour-stroma interac- 16. Orimo A, Tomioka Y, Shimizu Y, et al. Cancer Acids Res 2003;31:82–6. tion: cancer-associated fibroblasts as novel tar- associated myofibroblasts possess various fac- 25. Segditsas S, Tomlinson I. Colorectal cancer gets in anti-cancertherapy?Lung Cancer2004; tors to promote endometrial tumor progression. and genetic alterations in the Wnt pathway. 45(Suppl 2):163–75. Clin CancerRes 2001;7:3097 –105. Oncogene 2006;25:7531–7. 7. MuellerMM, Fusenig NE. Friendsorfoes — 17. Friess H, Ding J, Kleeff J, et al. Microarray- 26. Su F, OverholtzerM, BesserD, Levine AJ. bipolar effects of the tumour stroma in cancer. based identification of differentially expressed WISP-1 attenuates p53-mediated apoptosis in re- Nat Rev Cancer2004;4:839 –49. growth-and metastasis-associated genes in pan- sponse to DNA damage through activation of the 8. Nikitenko L, Boshoff C. Endothelial cells and creatic cancer. Cell Mol Life Sci 2003;60:1180–99. Akt kinase. Genes Dev 2002;16:46–57. cancer. Handb Exp Pharmacol 2006;176:307–34. 18. Nakagawa H, Liyanarachchi S, Davuluri RV, 27. Pukrop T, Klemm F, Hagemann T, et al. Wnt 5a 9. Lamalice LF, Huot J. Endothelial cell migration et al. Role of tumor stromal fibroblasts in me- signaling is critical for macrophage-induced in- during angiogenesis. Circ Res 2007;100:782–94. tastatic colon cancerto the liverand theirex- vasion of breast cancer cell lines. Proc Natl Acad 10. Cooney MM, van Heeckeren W, Bhakta S, pression profiles. Oncogene 2004;23:7366–77. Sci U S A 2006;103:5454–9. Ortiz J, Remick SC. Drug insight: vascular dis- 19. Singer CF, Gschwantler-Kaulich D, et al. Differ- 28. Katoh M. Epithelial-mesenchymal transition in rupting agents and angiogenesis-novel ap- ential gene expression profile in breast cancer- gastric cancer. Int J Oncol 2005;27:1677–83.

Clin Cancer Res 2009;15(12) June 15, 20094026 www.aacrjournals.org Downloaded from clincancerres.aacrjournals.org on September 28, 2021. © 2009 American Association for Cancer Research. Fibroblasts in Esophageal Cancer

29. Goerge T, Barg A, Schnaeker EM, et al. Tumor- panied by N-cadherin and fibronectin expres- tants and recruitment of myeloid cells predeter- derived matrix metalloproteinase-1 targets endo- sion. Biochem Biophys Res Commun 2007;358: mines lung metastasis. Nat Cell Biol 2006;8: thelial proteinase-activated receptor 1 promoting 925–30. 1369–75. endothelial cell activation. CancerRes 2006;66: 35. Katoh M. Networking of WNT, FGF, Notch, 41. Sheikh AA, Vimalachandran D, Thompson CC, 7766–74. BMP, and Hedgehog signaling pathways during et al. The expression of S100A8 in pancreatic tu- 30. Van Themsche C, Alain T, Kossakowska AE, carcinogenesis. Stem Cell Rev 2007;3:30–8. mormonocytes is associated with the Smad4 et al. Stromelysin-2 (matrix metalloproteinase 36. Munshi HG, Stack MS. Reciprocal interactions status of pancreatic cancer cells. Proteomics 10) is inducible in lymphoma cells and accel- between adhesion receptor signaling and MMP 2007;7:1929–40. erates the growth of lymphoid tumors in vivo. regulation. Cancer Metastasis Rev 2006;25:45–56. 42. Dickson C, Spencer-Dene B, Dillon C, Fantl V. J Immunol 2004;173:3605–11. 37. Tse JC, Kalluri R. Mechanisms of metastasis: Tyrosine kinase signalling in breast cancer: 31. Calaluce R, Bearss DJ, Barrera J, et al. Lami- epithelial to mesenchymal transition and contri- fibroblast growth factors and their receptors. nin-5 β3A expression in LNCaP human prostate bution of tumor microenvironment. J Cell Bio- Breast Cancer Res 2000;2:191–6. carcinoma cells increases cell migration and tu- chem 2007;101:816–29. 43. Klint P, Claesson-Welsh L. Signal transduction morigenicity. Neoplasia 2004;6:468–79. 38. Kim SW, Hayashi M, Lo JF, et al. Tid1 nega- by fibroblast growth factor receptors. Front Bios- 32. Dos Reis PP, Bharadwaj RR, Machado J, et al. tively regulates the migratory potential of cancer ci 1999;4:165–77. Claudin 1 overexpression increases invasion and cells by inhibiting the production of interleukin- 44. McKeehan WL, Wang F, Kan M. The heparan is associated with aggressive histological fea- 8. CancerRes 2005;65:8784 –91. sulfate-fibroblast growth factor family: diversity tures in oral squamous cell carcinoma. Cancer 39. Huang S, Mills L, Mian B, et al. Fully human- of structure and function. Prog Nucleic Acid 2008;113:3169–80. ized neutralizing antibodies to interleukin- Res Mol Biol 1998;59:135–76. 33. Kim K, Lu Z, Hay ED. Direct evidence for a role 8 (ABX-IL8) inhibit angiogenesis, tumorgrowth, 45. ChafferCL, Brennan JP, Slavin JL, Blick T, of β-catenin/LEF-1 signaling pathway in induc- and metastasis of human melanoma. Am J Thompson EW, Williams ED. Mesenchymal-to- tion of EMT. Cell Biol Int 2002;26:463–76. Pathol 2002;161:125–34. epithelial transition facilitates bladder cancer 34. Yang Z, Zhang X, Gang H, et al. Up-regulation 40. Hiratsuka S, Watanabe A, Aburatani H, Maru Y. metastasis: role of fibroblast growth factor re- of gastric cancer cell invasion by Twist is accom- Tumour-mediated upregulation of chemoattrac- ceptor-2. Cancer Res 2006;66:11271–8.

www.aacrjournals.org 4027 Clin Cancer Res 2009;15(12) June 15, 2009 Downloaded from clincancerres.aacrjournals.org on September 28, 2021. © 2009 American Association for Cancer Research. Fibroblast Growth Factor Receptor 2−Positive Fibroblasts Provide a Suitable Microenvironment for Tumor Development and Progression in Esophageal Carcinoma

Chunyu Zhang, Li Fu, Jianhua Fu, et al.

Clin Cancer Res 2009;15:4017-4027.

Updated version Access the most recent version of this article at: http://clincancerres.aacrjournals.org/content/15/12/4017

Supplementary Access the most recent supplemental material at: Material http://clincancerres.aacrjournals.org/content/suppl/2009/06/15/1078-0432.CCR-08-2824.DC1

Cited articles This article cites 45 articles, 13 of which you can access for free at: http://clincancerres.aacrjournals.org/content/15/12/4017.full#ref-list-1

Citing articles This article has been cited by 10 HighWire-hosted articles. Access the articles at: http://clincancerres.aacrjournals.org/content/15/12/4017.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/15/12/4017. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.