Hepatocyte–Stellate Cell Cross-Talk in the Liver Engenders a Permissive Inﬂammatory Microenvironment That Drives Progression in Hepatocellular Carcinoma

Published OnlineFirst March 14, 2012; DOI: 10.1158/0008-5472.CAN-11-3317

Cancer Microenvironment and Immunology Research

Cedric Coulouarn1,2, Anne Corlu1,2, Denise Glaise1,2, Isabelle Guénon1,2, Snorri S. Thorgeirsson3, and Bruno Clément1,2

Abstract Many solid malignant tumors arise on a background of inflamed and/or fibrotic tissues, features that are found in more than 80% hepatocellular carcinomas (HCC). Activated hepatic stellate cells (HSC) play a critical role in fibrogenesis associated with HCC onset and progression, yet their functional impact on hepatocyte fate remains largely unexplored. Here, we used a coculture model to investigate the cross-talk between hepatocytes (human hepatoma cells) and activated human HSCs. Unsupervised genome-wide expression profiling showed that hepatocyte–HSC cross-talk is bidirectional and results in the deregulation of functionally relevant gene networks. Notably, coculturing increased the expression of proinflammatory cytokines and modified the phenotype of hepatocytes toward motile cells. Hepatocyte–HSC cross-talk also generated a permissive proangiogenic microenvironment, particularly by inducing VEGFA and matrix metalloproteinase (MMP)9 expression in HSCs. An integrative genomic analysis revealed that the expression of genes associated with hepatocyte–HSC cross-talk correlated with HCC progression in mice and was predictive of a poor prognosis and metastasis propensity in human HCCs. Interestingly, the effects of cross-talk on migration and angiogenesis were reversed by the histone deacetylase inhibitor trichostatin A. Our findings, therefore, indicate that the cross-talk between hepatoma cells and activated HSCs is an important feature of HCC progression, which may be targeted by epigenetic modulation. Cancer Res; 72(10); 2533–42. 2012 AACR.

Introduction tosis, and maintain epithelial cell polarity and differentiation Extensive evidence from genetics, genomics, and cell (4). In cancer, the microenvironment, which is also referred biology showed that cancer onset and progression is not to as stroma, experiences drastic changes including the only determined by tumor cells but also influenced by the recruitment and the activation of stromal cells and the microenvironment (1–3). Microenvironment is a complex remodeling of ECM. Importantly, coevolution of tumor cells system, which largely consists of extracellular matrix (ECM) with their microenvironment during tumorigenesis suggests – fl proteins and proteoglycans, soluble factors, and small sig- that tumor stroma cross-talk may likely in uence the phe- naling molecules such as cytokines and chemokines, along notype of tumor cells and may provide a selective pressure with a variety of cell types such as fibroblasts, immune for tumor initiation, progression, and metastasis (1, 5, 6). cells, and endothelial cells. Under normal conditions, the Hepatocellular carcinoma (HCC) is the most common pri- microenvironment constitutes an important modulator of mary tumor of the liver. The incidence of HCC is increasing in epithelium cell fate and a barrier to cell transformation (4). many countries and its prognosis is typically poor. Thus, with Dynamic communications between the epithelium and 550,000 cases newly diagnosed and 600,000 deaths annually, the microenvironment notably modulate cell growth, apop- HCC ranks among the deadliest forms of human malignancies worldwide (7). The remodeling of liver microenvironment is a hallmark of HCC pathogenesis (8). Indeed, more than 80% Authors' Affiliations: 1INSERM, UMR991, Liver Metabolisms and Cancer; HCCs develop in the setting of chronic hepatitis, fibrosis, and 2University of Rennes 1, Rennes, France; and 3Laboratory of Experimental cirrhosis, conditions in which inflammation and ECM depo- Carcinogenesis, Center for Cancer Research, National Cancer Institute, NIH, Bethesda, Maryland sition profoundly alter the hepatic microenvironment (9). However, the functional impact of the disrupted microenvi- Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/). ronment on hepatocyte biology remains poorly understood (10). Over the last decade, high-throughput genomic studies Corresponding Author: Cedric Coulouarn, INSERM, UMR991, Pontch- aillou University Hospital, 2 rue Henri Le Guilloux, Rennes F-35033, France. provided important insights to classify HCCs at a molecular Phone: 33-223233-881; Fax: 33-299540-137; E-mail: level and to identify deregulated gene networks and signaling [email protected] pathways (11). However, most of gene expression signatures in doi: 10.1158/0008-5472.CAN-11-3317 HCCs have been derived from whole-tumor tissues consisting 2012 American Association for Cancer Research. of both the cancer epithelial cells and their surrounding

www.aacrjournals.org 2533

Coulouarn et al.

microenvironment, a strategy which rendered elusive to eval- human SurePrint G3 8 60K pangenomic microarrays (Agilent uate the specific contribution of each compartment. Technologies) as previously described (18). Starting from 150 Activation of hepatic stellate cells (HSC) is a key feature of ng total RNA, amplification yield was 9.7 0.6 mg cRNA and liver fibrosis and cirrhosis (12). Following liver injury, qui- specific activity was 20.3 1.3 pmol Cy3 per mg cRNA. Gene escent HSCs become activated and convert into highly expression data were processed using Feature Extraction and proliferative myofibroblast-like cells, which express inflam- GeneSpring softwares (Agilent Technologies) and further ana- matory and fibrogenic mediators responsible for ECM accu- lyzed using R-based BRB-ArrayTools. Briefly, differentially mulation within the microenvironment (12, 13). Therefore, expressed genes were identified by a 2-sample univariate t the present study was specifically designed to address the test and a random variance model (P < 0.01; false discovery rate functional impact and the clinical relevance of the cross-talk < 1%) as described (19). Permutation P values for significant between tumor hepatocytes and activated HSCs. By using a genes were computed on the basis of 10,000 random permuta- coculture model system, genome-wide expression profiling, tions. Class prediction was conducted using 7 algorithms and and functional assays, we showed that the cross-talk misclassification rate was computed using a leave-one-out between hepatocytes and activated HSCs (i) is bidirectional, cross-validation method (20). Clustering analysis was done (ii) induces an alteration of hepatocyte phenotype toward using Cluster 3.0 and TreeView 1.6 using uncentered correla- migration, (iii) generates a permissive proinflammatory and tion and average linkage options. MIAME compliant micro- proangiogenic microenvironment, (iv) is predictive of a poor array data have been deposited into Gene Expression Omnibus prognosis in human HCC, and (v) could be targeted by (GEO) database (GSE32565). epigenetic modulation. Data mining and integrative genomics Materials and Methods Gene annotation was based on Gene Ontology and enrichment for specific biologic functions or canonical pathways was Cell lines and coculture experiments evaluated using FuncAssociate 2.0 program (21). Ingenuity HepaRG and HuGB cell lines were established in our labo- Pathway Analysis (IPA) was used to examine the functional ratory and maintained as previously described (14, 15). association between differentially expressed genes and to HepaRG cells were grown in William's E medium supplemen- generate the highest significant gene networks (Ingenuity). m ted with 10% FBS, 100 U/mL penicillin, 100 g/mL strepto- Relevant networks were identified using the scoring system m m mycin, 5 g/mL insulin, and 50 mol/L hydrocortisone hemi- provided by IPA. Gene set enrichment analysis (GSEA) was succinate. Differentiation of HepaRG from progenitors to conducted by using the Java tool developed at the Broad mature well-differentiated hepatocytes was achieved in 4 Institute (Cambridge, MA) as previously described (22). Unsu- weeks by culturing the cells in the supplemented medium in pervised GSEA was done with the whole C2 collection of presence of 2% dimethyl sulfoxide (DMSO) for the last 2 weeks curated gene sets from the molecular signatures database as previously described (ref. 16; Supplementary Fig. S1A). All (MSigDB). Enrichment score was determined after 1,000 per- experiments hereinafter referred to as HepaRG were carried mutations. Connectivity map algorithm was used to link gene out using mature hepatocytes selectively isolated by mild expression signatures with putative therapeutic molecules trypsinization from DMSO-treated cultures. LX2 cells (Supple- (23). Integration of genomic data was conducted as previously mentary Fig. S1B) were provided by S.L. Friedman (Mount Sinai described (20) using publicly available gene expression data School of Medicine, New York, NY) and were maintained in sets downloaded from GEO. supplemented Dulbecco's Modified Eagles' Media (DMEM) as described (17). HepaRG–LX2 cocultures were conducted in Real-time reverse transcriptase PCR serum- and DMSO-free William's E medium using 6-well plates Expression of relevant genes was measured by quantita- and 1-mm pore size Transwell inserts, which allow diffusion tive real-time PCR as previously described (22). Quantitative of media components but prevent cell migration (BD Bio- 2DDCt sciences; Supplementary Fig. S1C). Huh7 and HepG2 cell analysis of PCR data was conducted with the method lines were obtained from the European Collection of Cell using glyceraldehyde-3-phosphate dehydrogenase (GAPDH) Cultures, which conducted cell lines authentication by DNA Ct values for normalization. Melting analysis was conducted fi barcoding. Primary human umbilical vein endothelial cells to validate the speci city of PCR products. PCR and micro- (HUVEC) were purchased from Invitrogen and were main- array analysis were conducted using RNA extracted from ¼ tained in 200PRF medium supplemented with a low-serum independent culture experiments (n 3). growth supplement (LSGS). All cell cultures were conducted at 37 C in a 5% CO2 atmosphere. Trichostatin A (TSA) was Cell proliferation purchased from Sigma-Aldrich. Independent culture experi- HepaRG (10,000 cells per well) were seeded onto 96-well ments were carried out at least in triplicate. plates. Following 4-hour incubation at 37 C, the medium was replaced by serum-free medium supplemented with 50% (v/v) Microarray analysis conditioned medium derived from culture and coculture of Total RNA was purified from cells at 80% confluence with an HepaRG and LX2. Proliferation was evaluated after 24, 48, and RNeasy kit (Qiagen). Genome-wide expression profiling was 72 hours using a CyQuant Cell Proliferation Assay Kit (Invitro- conducted using the low-input Quick Amp Labeling Kit and gen). Experiments were carried out in triplicate.

2534 Cancer Res; 72(10) May 15, 2012 Cancer Research

Hepatocyte–Stellate Cell Cross-Talk in Liver Cancer

Cell migration Inﬂuence of conditioned medium on HepaRG migration was determined using a 2-dimensional gap closure radius 96-well migration assay, according to manufacturer's instructions (Cell Biolabs). Cell migration was independently evaluated from scratch-wounded conﬂuent monolayers of HepaRG incu- bated in presence of serum-free medium supplemented with 50% conditioned medium as above. Migration was evaluated up to 72 hours in triplicate.

In vitro angiogenesis HUVECs (30,000 cells per well) were seeded onto 48-well plates previously coated with Geltrex-reduced growth factor basement membrane matrix (100 mL/cm2) using nonsupplemented 200PRF medium (Invitrogen). Endothelial tube formation was monitored after 6 hours in the presence of 50% (v/v) serum-free conditioned medium from culture/coculture of LX2 and HepaRG. LSGS-supplemented HepaRG medium (2% FBS; 3 ng/mL basic fibroblast growth factor) was used as a positive inducer control and nonsupplemented HepaRG medium was used as a negative control. Triplicate experiments were carried out. Figure 1. Genome-wide expression profiles changes in cocultures of HepaRG and LX2. HepaRG and LX2 cell lines were cultured alone or side Gel zymography by side using Transwell inserts (n ¼ 3 independent culture experiments). Matrix metalloproteinase (MMP) activity in conditioned After 48 hours, total RNA was extracted from culture and coculture medium was evaluated in triplicate by gelatine zymography experiments and subjected to a microarray analysis. A, volcano plot (left) as described (8). Recombinant human MMP2 and MMP9 were and clustering analysis (right) of 212 genes differentially expressed in 3 used as positive controls. After scanning, images were analyzed independent experiments using HepaRG cultured alone (HepaRG) or in presence of LX2 (HepaRG–LX2). B, volcano plot (left) and clustering by densitometry using ImageJ (NIH, Bethesda, MD). analysis (right) of 123 genes differentially expressed in 3 independent experiments using LX2 cultured alone (LX2) or in presence of HepaRG Statistical analysis (LX2-HepaRG). In A and B, RNAs were selected on the basis of the fi Quantitative results were expressed as mean and SD and the signi cance of the differential gene expression in coculture versus culture fi conditions (horizontal red line; P < 0.01) and the level of induction or signi cance was evaluated by Student t test. repression (vertical red lines; fold-change > 1.5); examples of main changes in steady-state levels of mRNAs are indicated on the right. Results Coculturing hepatocytes with activated HSCs results in a highly expressed genes in LX2 were significantly enriched in a bidirectional cross-talk rat model of hepatic fibrosis and were characteristic of the HepaRG and LX2 cells were used as a paradigm to model the transformation of quiescent HSCs into myofibroblasts (Sup- cross-talk between transformed but differentiated human plementary Fig. S2B). hepatocytes and activated HSCs in the context of liver cancer Hepatocyte–HSC cross-talk was next addressed by analyz- (Supplementary Fig. S1). HepaRG cell line was established in ing cocultures of HepaRG and LX2 separated by a Transwell our laboratory from a well-differentiated Edmondson grade I insert (Supplementary Fig. S1C). Microarray experiments were HCC and was shown to possess the unique property to carried out after 48 hours, when cells reached 80% confluency. spontaneously differentiate into functional mature hepato- In HepaRG, the analysis of gene expression profiles by means of cytes and biliary cells (14, 16). LX2 cells were reported to class comparison and class prediction algorithms identified greatly recapitulate the in vivo phenotype of primary human 212 genes whose expression was significantly modulated (P < activated HSC (17). However, as gene expression profiles 0.01; 1.5 fold change) by the presence of LX2 (Fig. 1A and during HSC activation may differ in culture and in vivo (24), Supplementary Table S1). More than 83% genes were upregu- we first conducted a gene expression profiling to validate the lated suggesting that coculture with LX2 induced a global shift molecular phenotype of LX2 cells in our culture conditions. IPA toward transcriptional activation in HepaRG (Fig. 1A). In LX2, showed that highly expressed genes in LX2 (top 1%, 163 genes) the expression of 123 genes was significantly altered by the were significantly linked to hepatic fibrosis and HSC activation coculture condition (Fig. 1B and Supplementary Table S2), (Supplementary Fig. S2A). As expected, these genes included including the upregulation of master genes involved in ECM major regulators of ECM synthesis and degradation (e.g., remodeling and angiogenesis (e.g., MMP9 and VEGFA). Impor- COL1A1 and MMP2), markers of HSC activation (e.g., ACTG2 tantly, the unsupervised analysis of genes differentially and VIM), as well as proinflammatory cytokines (e.g., IL1B). expressed in coculture versus culture condition revealed a Further validating the choice of this cell line, GSEA showed that bidirectional cross-talk between HepaRG and LX2.

www.aacrjournals.org Cancer Res; 72(10) May 15, 2012 2535

Coulouarn et al.

Coculture with LX2 induces an inflammatory response Inflammation is thought to play a key role in cancer initiation and a motile phenotype in HepaRG and progression by fostering multiple hallmarks of cancer Gene Ontology and ingenuity analysis showed that the genes including tumor cell proliferation and motility (1). To evaluate related to cell chemotaxis, motility, and inflammation were whether the coculture condition had any impact on the phe- significantly enriched in the HepaRG–LX2 signature (Supple- notype of HepaRG, mature hepatocytes were isolated from new mentary Table S3). Notably, several important proinflammatory HepaRG cultures and were exposed to conditioned media and profibrogenic cytokines [e.g., interleukin (IL)-1B, IL-6, and derived from the initial cultures and cocultures of HepaRG and IL-8], acute-phase proteins (e.g., CP and SAA1), and growth LX2. Gap closure assay showed that exposing fresh HepaRG factors (e.g., AREG and EREG) were upregulated in HepaRG hepatocytes to conditioned medium derived from HepaRG– when the 2 cell types were cultured together (Fig. 1A and LX2 coculture significantly induced cell migration (Fig. 3). Supplementary Table S1). Accordingly, a well-organized gene Of note, cell proliferation remained unaffected by condition- network linked to IL-1B, IL-6, IL-8, and CCL2, which is also ed medium treatment, suggesting that gap closure was not known as the monocyte chemoattractant protein 1 (MCP1)was secondary to enhanced cell proliferation (data not shown). identified by IPA (Fig. 2A and Supplementary Fig. S3). The Collectively, these data indicated that coculturing hepatocytes expression of IL-6, IL-8, and CCL2 was further evaluated by with activated HSCs resulted in the production of soluble quantitative reverse transcriptase PCR (Q-RT-PCR) using RNA factors, including proinflammatory signals, which were able extracted from independent cell culture experiments. As to modify the phenotype of hepatocytes toward migration. shown in Fig. 2B, all genes were significantly upregulated in HepaRG after 48-hour coculture with LX2. Together, these data HepaRG–LX2 cross-talk generates a permissive suggested that the cross-talk between HepaRG and LX2 proangiogenic microenvironment by modulating MMP9 resulted in the establishment of a proinflammatory microen- and VEGFA in LX2 cells vironment. To validate this observation, we conducted a GSEA Data mining of 112 genes differentially expressed when LX2 using an independent gene set that covered the whole response were cocultured with HepaRG (LX2–HepaRG signature; Fig. of Hep3B hepatocytes to proinflammatory cytokines (25). This 1B) identified a gene network linked to VEGFA and MMP9 (Fig. approach unambiguously showed that coculture with LX2 in- 4A). Upregulation of VEGFA and MMP9 genes was validated by duced a prominent inflammatory response in HepaRG (Fig. 2C). Q-RT-PCR using RNA derived from independent experiments

Figure 2. The HepaRG–LX2 gene signature is related to inflammation. A, ingenuity analysis of upregulated genes identified a gene network centered on IL-1B, IL-6, IL-8,andCCL2.B, comparison of IL-6, IL-8, and CCL2/MCP1 mRNA levels detected by microarray and Q-RT- PCR in HepaRG using independent 48-hour cultures (white bar, HepaRG cultured alone; black bar, HepaRG cocultured with LX2; n ¼ 3). Both microarray and Q-RT-PCR showed an upregulation of these cytokines in cocultured HepaRG (, P < 0.01). C, GSEA analysis using the gene expression profiles of HepaRG cultured alone (HepaRG; right) or in presence of LX2 (HepaRG–LX2; left) and a proinflammatory gene signature established in Hep3B cell line (25). GSEA showed a significant enrichment of the proinflammatory gene signature in HepaRG–LX2 gene expression profiles (P < 0.01).

2536 Cancer Res; 72(10) May 15, 2012 Cancer Research

Hepatocyte–Stellate Cell Cross-Talk in Liver Cancer

activated HSCs. The HepaRG–LX2 signature (i.e., 212 genes differentially expressed in HepaRG by the coculture condition; Fig. 1A) was first integrated with the gene expression profiles of 139 cases of human HCCs, which were extensively characterized (20, 26–28). Hierarchical clustering analysis of the integrated data set identified 2 robust clusters whose organization was driven by the culture condition of HepaRG: cluster 1, HepaRG–LX2 coculture; cluster 2, HepaRG culture (Fig. 5A). Interestingly, we observed that clinical and biologic parameters of HCCs were not randomly distributed between the clusters 1 and 2. Strikingly, cluster 1 included significantly more tumors, which were previously assigned to a poor prognosis group, than cluster 2 (Fig. 5A). As shown in Fig. 5B, cluster 1 HCCs were previously defined by a bad survival (27), hepatoblast traits (28), along with the activation of oncogenic MET/ hepatocyte growth factor (HGF; ref. 26) and TGFb pathways (20). In addition, they exhibited a poorly differentiated phenotype (Edmondson grade >III), signs of vascular invasion, and were significantly bigger in size (Fig. 5B). More importantly, the survival of patients included in cluster 1 was significantly reduced (P < 0.01). Interestingly, if these observations showed that the HepaRG–LX2 cross-talk signature is predictive of a poor prognosis when evaluated as a unique gene set, we also Figure 3. Conditioned medium from HepaRG–LX2 cocultures induces reported that 85% of genes included in the signature and HepaRG migration. Cell migration was analyzed by a gap closure assay. expressed in primary HCCs were clinically relevant when After seeding (T0), HepaRG were cultured in presence of conditioned medium derived from HepaRG cultures (HepaRG-CM) or HepaRG–LX2 analyzed individually (Supplementary Table S4). Unsupervised cocultures (HepaRG–LX2-CM). After 48 hours, cells were fixed and nuclei GSEA further supported the relevance of HepaRG–LX2 signa- were stained with a DAPI fluorescent dye (48h-DAPI). Image analysis ture in predicting a poor prognosis phenotype not only in HCCs showed that HepaRG–LX2-CM induces a migration of HepaRG (white but also in other cancers. In the context of HCCs, we found that bar, HepaRG-CM; black bar, HepaRG–LX2-CM; n ¼ 3; , P < 0.01). DAPI, fi a 40,6—diamidino-2-phenylindole. speci c signatures for recurrence (29), c-Myc/TGF aggressive mouse model of HCC (30) or TGFb (20) were significantly enriched in the HepaRG–LX2 gene profiles (Supplementary (Fig. 4B). Together with Gene Ontology and ingenuity analysis Table S5). This approach also showed that gene signatures (Supplementary Table S3), these results suggested that associated with a bad prognosis in cancers other than HCCs HepaRG–LX2 cross-talk may have profound impact on angio- (e.g., invasive breast cancer, advanced gastric cancer, meta- genesis and ECM remodeling. This hypothesis was first tested static stromal cells, and highly metastatic pancreatic cancer by exposing HUVECs to conditioned medium derived from the cells) were significantly (P < 0.01) enriched both in HepaRG and culture or the coculture of LX2 and HepaRG. As shown in Fig. LX2 under coculture conditions (Supplementary Fig. S4). To 4C, conditioned medium (CM) issued from coculture experi- further investigate the relevance of the HepaRG–LX2 cross-talk ments (LX2–HepaRG-CM) induced the formation of tubule signature in HCC progression, we conducted a cross-species complexes by HUVECs. In contrast, conditioned medium integrative genomic approach as described previously (20). issued from LX2 cultured alone (LX2-CM) failed to induce in First, we integrated the HepaRG–LX2 signature with the gene vitro angiogenesis (Fig. 4C). Absence of tube formation was also expression profiles of 80 cases of HCCs derived from 4 trans- noticed when HUVECs were exposed to conditioned medium genic mouse models of HCC (c-Myc, E2f1, c-Myc/E2f1, and derived from HepaRG cultured alone (data not shown). Next, c-Myc/TGFa; refs. 19, 30). Consistent with the unsupervised MMP activity in conditioned medium was evaluated by gelatin GSEA, the analysis of the integrated data sets showed that the zymography. In agreement with the increased expression of HepaRG–LX2 signature clustered specifically with HCCs MMP9 (Fig. 4B), we showed that MMP9 activity was signifi- derived from c-Myc/TGFa (Supplementary Fig. S5A). This cantly higher (P < 0.01) in LX2–HepaRG-CM than in LX2-CM observation pointed out c-Myc/TGFa transgenic mice as the (Fig. 4D). Similar observation was made for MMP2. These data best in vivo model for evaluating the HepaRG–LX2 signature in suggested that hepatocyte–HSC cross-talk resulted in the HCC onset and progression. This was investigated by using the establishment of a permissive proangiogenic microenviron- gene expression profiles characteristic of c-Myc/TGFa– ment that may facilitate the migration of tumor cells. induced hepatocarcinogenesis. These profiles were established previously using liver samples collected at various time points HepaRG–LX2 cross-talk signals a poor prognosis in of tumor onset and progression in transgenic mice, ranging human HCCs and correlates with tumor progression from 3 weeks (moderate dysplasia), 3 months (severe dyspla- Integrative genomics was next used to evaluate the clinical sia), and 9 months (HCCs; ref. 30). By using a multidimensional relevance of the cross-talk between cultured hepatocytes and scaling approach, we showed that the HepaRG–LX2 signature

www.aacrjournals.org Cancer Res; 72(10) May 15, 2012 2537

Coulouarn et al.

Figure 4. HepaRG–LX2 cross-talk induces in vitro angiogenesis and MMP expression. A, IPA of genes upregulated in LX2 cocultured with HepaRG identiﬁed a gene network connected to VEGFA and MMP9. B, both VEGFA and MMP9 mRNA levels were upregulated in LX2 after 48-hour coculture with HepaRG (black bar) as compared with culture alone (white bar); gene expression analysis was conducted in genome-wide array and Q-RT-PCR using independent culture experiments (n ¼ 3). C, in vitro angiogenesis assay using HUVECs grown on a Geltrex matrix in presence of conditioned medium derived from the culture of LX2 (LX2-CM) or the coculture of LX2 with HepaRG (LX2–HepaRG-CM). After 6 hours, tube formation by HUVECs was observed in wells corresponding to positive control and treatment with LX2–HepaRG-CM (n ¼ 3; representative images are shown). D, detection of MMPs by gelatin zymography in culture-CM and coculture-CM. Signiﬁcant increase in MMP2 and MMP9 expression was measured in LX2–HepaRG-CM. B and D, n ¼ 3; , P < 0.01.

effectively discriminated the mouse samples on the basis of the genes that signed the cross-talk between HepaRG and LX2 HCC progression (Supplementary Fig. S5B). were significantly enriched in the expression profiles of cir- Because HSC activation is an early event in the pathogenesis rhosis tissue from patients with metastasis (Fig. 5C). Altogeth- of HCCs, we asked whether the HepaRG–LX2 signature within er, these results showed that the expression of genes embedded cirrhosis tissues from patients with HCCs could predict any in the HepaRG–LX2 signature, either in cirrhosis or in HCCs, specific clinical outcome. To test this hypothesis, we used a was predictive of a poor prognosis and was associated with publicly available gene expression data set established from the metastasis propensity. noncancerous hepatic tissue of patients with primary HCCs (GSE5093 in GEO database). In this cohort, patients were Epigenetic modulation of HepaRG–LX2 cross-talk by TSA divided in 2 groups on the basis of the presence or absence The clinical relevance of HepaRG–LX2 signature in predict- of venous metastasis (31). Accordingly, cirrhosis tissues iso- ing a poor prognosis in human HCCs suggested that molecules lated from patients with or without metastasis were, respec- targeting hepatocyte–HSC cross-talk may represent a prom- tively, termed metastasis-inclined microenvironment (MIM) ising therapeutic strategy. In this context, connectivity map and metastasis-averse microenvironment (MAM; ref. 31). On was used to identify molecules that could reverse the global the basis of the HepaRG–LX2 signature, we showed that MIM gene expression profile induced by LX2 on HepaRG. Interest- and MAM samples coclustered with HepaRG–LX2 and ingly, the top 10 ranked molecules identified by this approach HepaRG samples, respectively (Fig. 5C). GSEA confirmed that included 3 inhibitors of histone deacetylases, vorinostat, TSA,

2538 Cancer Res; 72(10) May 15, 2012 Cancer Research

Hepatocyte–Stellate Cell Cross-Talk in Liver Cancer

Figure 5. Clinical relevance of HepaRG–LX2 signature in human HCCs. A, dendrogram overview of HepaRG and HepaRG–LX2 experiments integrated with 139 cases of human HCCs. Clustering analysis was based on the expression of 212 genes differentially expressed in HepaRG cocultured with LX2. Two major clusters (1 and 2) were identified. Distribution of human HCC samples between previously described subgroups with respect to survival (good vs. bad prognosis; ref. 27), cell origin (hepatoblast vs. hepatocytes; ref. 28), activation of MET/HGF ( vs. þ; ref. 26), and TGFb signaling pathway (early vs. late; ref. 20) is indicated on the left. B, statistical analysis of HCC distribution between clusters 1 and 2 based on previous gene signatures and clinical parameters. Cluster 1, which is defined by the HepaRG–LX2 coculture signature, shows a significant enrichment in HCCs with the following features: bad survival, hepatoblast traits, activation of MET/HGF, and late TGFb pathways, higher differentiation grade, and serum a–fetoprotein (AFP) level. Tumors size was significantly higher for HCCs included in cluster 1. Kaplan–Meier plots and log-rank statistics analysis revealed a significant decrease in overall survival for patients included in cluster 1. C, integrative genomics using HepaRG–LX2 signature and gene expression profiles of peritumoral cirrhotic tissues from patient with (MIM) or without (MAM) metastasis (31). Clustering analysis (top) and GSEA (bottom) show that the HepaRG–LX2 signature was significantly enriched in the gene expression profiles of cirrhotic tissues from patients with metastasis (MIM). and valproic acid (Supplementary Table S6). TSA was chosen produced conditioned medium from culture and coculture of for further experiments on the basis of the higher number of HepaRG and LX2 cells exposed to 500 nmol/L TSA or DMSO hits obtained with a connectivity map (n ¼ 182 hits, Fig. 6A) control. As shown in Fig. 6B, the migration of freshly isolated and the results of the unsupervised GSEA, which showed that HepaRG hepatocytes in response to HepaRG–LX2-CM was genes silenced by TSA in pancreatic cancer were significantly completely abrogated in presence of TSA. We further showed (P < 0.01) enriched in the gene profiles of both HepaRG and LX2 in HepaRG that LX2 induced upregulation of amphiregulin under coculture conditions (Supplementary Fig. S4). To test and epiregulin, 2 genes that have been linked to invasion and whether TSA could modulate the HepaRG–LX2 cross-talk, we metastasis and that were abolished by TSA (Supplementary

www.aacrjournals.org Cancer Res; 72(10) May 15, 2012 2539

Coulouarn et al.

Figure 7. Proposed model for molecular cross-talk between hepatocytes and activated HSCs in HCCs.

cross-talk between LX2 and other hepatoma cells, namely, Huh7 and HepG2, showed that the regulation of some important factors, such as IL-8, was conserved (Supplementary Fig. S7). More broadly, the results also suggested that the cross-talk Figure 6. Inhibition of LX2-induced migration of HepaRG by TSA. A, with LX2 may depend on the differentiation status of the cells. connectivity map identification of TSA and vorinostat as candidate Indeed, HepaRG cell line was established from an Edmonson molecules to target HepaRG–LX2 cross-talk. B, cell migration was grade I, well-differentiated HCCs, and showed a unique pro- analyzed by a gap closure assay. After seeding, HepaRG were cultured in perty to differentiate into mature hepatocytes. Taking this presence of conditioned medium derived from the culture (HepaRG-CM) or the coculture (HepaRG–LX2-CM) of HepaRG and LX2 exposed to feature into account was particularly relevant given that either 500 nmol/L TSA or DMSO control. After 72 hours, cells were fixed several studies reported an accumulation of activated HSCs and nuclei were stained with a DAPI fluorescent dye. Image analysis in dysplastic nodules at early stages of HCC development. fi – con rmed that HepaRG LX2-CM induced migration of HepaRG (top 2 Unsupervised analysis of genes deregulated in HepaRG in micrographs) and showed that migration was abolished in presence of fl TSA (bottom 2 micrographs). Histogram, quantification of HepaRG response to the coculture with LX2 highlighted proin amma- migration (white bar, HepaRG-CM; black bar, HepaRG–LX2-CM; n ¼ 3; tory cytokines (e.g., IL-1B and IL-6) and chemokines (e.g., IL-8, , P < 0.01). DAPI, 40,6-diamidino-2-phenylindole; ns, not significant. and CCL2) as key orchestrators of the cross-talk between hepatocytes and activated HSCs, consistent with previous observations (32, 33). Inflammatory cells and mediators are Fig. S6A). Our data also suggested that TSA was able to inhibit frequent in the local environment of tumors and several lines of coculture-induced angiogenesis as evidenced by the absence of evidences suggested that inflammation contributes to the tube formation by endothelial cells and the reduced VEGFA acquisition of core hallmark capabilities in cancer (1). Notably, expression by LX2 cells under coculture condition (Supple- epidemiologic studies suggested that inflammatory diseases mentary Fig. S6B). predispose individuals to cancer (34). Thus, an increase expression of IL-6 was reported in the serum of most patients with Discussion cancer. In HCCs, IL-6 expression was found to correlate with a The tumor microenvironment contributes in the acquisition rapid progression from hepatitis to HCCs (35), and an activa- of multiple hallmarks of cancer (1). However, the underlying tion of IL-6 pathway was reported in HCCs with a poor molecular mechanisms involved in the interactions between prognosis (36). The invasive capacity of malignant cells has the tumor cells and the microenvironment remain poorly been shown to be increased in presence of IL-1B and IL-6 (34). understood. We investigated the molecular mechanisms Recently, somatic alterations of gp130 and STAT3, which are involved in the cross-talk between tumor cells and their required for IL-6 signaling, have been reported in inflammatory microenvironment in liver cancer, by analyzing cocultures of hepatocellular adenomas (37, 38). hepatoma cells and activated HSCs. We propose a model in Previous studies by Omenetti and colleagues described a which this cross-talk is bidirectional and leads to a permissive similar bidirectional cross-talk between HSCs and cholangio- microenvironment through ECM remodeling and angiogene- cytes and provided evidence supporting the importance of the sis, along with the alteration of hepatocyte phenotype toward Hedgehog signaling (39). This pathway has been shown also to motile cells (Fig. 7). Thus, the data suggest that the dynamic promote the viability of HSCs (40). In the HepaRG–LX2 cocul- interactions between hepatocytes and activated HSCs through ture model, no significant difference was observed in the soluble mediators play an important role in the progression of expression of Hedgehog ligands (e.g., SHH), receptors (e.g., HCCs. Supporting this hypothesis, we further established by PTCH1), inducible transcription factors (e.g., GLI2) or inhibi- integrative genomics that hepatocyte–HSC cross-talk in vitro is tors (e.g., HHIP; Supplementary Fig. S8), suggesting that the clinically relevant and is associated with a poor prognosis in Hedgehog is not the prominent signaling pathway involved in human HCCs. More investigations will be needed to fully the cross-talk between mature hepatocytes and activated establish the contribution of each specific genes and/or path- HSCs. Given the fundamental role of this pathway in the ways in HepaRG–LX2 cross-talk. Interestingly, analyzing the cross-talk between immature liver epithelial cells and HSCs

2540 Cancer Res; 72(10) May 15, 2012 Cancer Research

Hepatocyte–Stellate Cell Cross-Talk in Liver Cancer

(39), it is plausible that Hedgehog signaling may play a role in group in HCCs, which was characterized by the presence of the fate of progenitor HepaRG cells. venous metastasis and reported to be significantly associated Consistent with previous observations using human or rat with an increase in TH2 cytokines and a decrease in TH1 HSCs (32, 33), our results also show that hepatocyte–HSC cytokines (31). cross-talk may greatly impact the local tumor microenviron- Treatment of HCCs represents an important clinical chal- ment, particularly ECM turnover by MMPs and angiogenesis, 2 lenge. Here, we provide evidences that targeting tumor–stroma main mechanisms promoting tumor growth and metastasis. cross-talk by epigenetic modulation may represent a promising Previously, we have shown that hepatocyte–HSC interplays therapeutic strategy (Fig. 7). Notably, TSA was able to inhibit induce enhanced ECM remodeling though MMP2 activation the coculture-induced migration of HepaRG hepatocytes as (41). In the present study, in addition to the increased expres- well as angiogenesis and VEGFA expression in LX2. Interest- sion of VEGF and MMPs, we report that the enhanced expres- ingly, TSA has been also reported to abrogate TGFb1-induced sion of chemoattractant chemokines may also contribute to epithelial–mesenchymal transition (EMT) in hepatocytes and the establishment of a permissive microenvironment by to reverse EMT-induced fibrosis by epigenetic modulation of recruiting other cell types than endothelial cells, notably type I collagen (44). Besides acting on hepatocytes, TSA could immune cells. Particularly, we show that mRNA levels for also mediate beneficial effects through HSCs. Indeed, TSA was IL-8, CCL2, CCL20, and CXCL2 genes were significantly induced shown to strongly suppress the proliferation of rat HSCs and to in the coculture condition. These soluble mediators are potent inhibit their conversion into myofibroblasts (45). chemoattractants for monocytes and lymphocytes. IL-8 is also In conclusion, the study highlights the central role of the a potent inducer of angiogenesis and metastasis. Interestingly, cross-talk between hepatocytes and their microenvironment emerging evidences indicate that CCL2 may also modulate the in HCCs and suggests that targeting tumor-stroma cross-talk T-helper (TH)1/TH2 immune responses. Indeed, CCL2 has been by epigenetic modulation may represent a promising thera- shown to induce the expression of IL-4, a potent TH2 cytokine, peutic strategy. whereas decreasing the expression of IL-12, a major TH1 cytokine (42). Thus, the induction of CCL2, IL-6, and IL-8 Disclosure of Potential Conflicts of Interest fl suggests that the cross-talk between hepatocytes and HSCs No potential con icts of interests were disclosed. may compromise a TH1 polarization and switch the expression fi Acknowledgments of cytokines toward a TH2 pro le. Differential expression of The authors thank Dr. S.L. Friedman, Mount Sinai School of Medicine, New TH1 and TH2 cytokines has been reported in the microenvi- York, NY, for his generous gift of LX2 cells, Dr. M. Ravache and K. Jarnouen, ronment of various types of cancer in vivo. In breast cancer, INSERM UMR991, for their help in cell migration assays and Q-RT-PCR, and the microarray core facility team from plateforme genomique sante, Biosit, Rennes. analysis of gene expression profiles of tumor stroma identified a gene signature that predicted clinical outcome indepen- Grant Support dently of existing standard clinical prognosis factors (43). This research was supported by INSERM, CNRS, University of Rennes 1, Importantly, tumor stroma derived from patients with a poor Institut National du Cancer Agence Nationale pour la Recherche and Association pour la Recherche sur le Cancer, France. outcome was characterized by an increased expression of The costs of publication of this article were defrayed in part by the genes linked to hypoxia and angiogenesis along with a payment of page charges. This article must therefore be hereby marked decreased expression of genes characteristic of a T 1 response advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this H fact. (43). Further supporting the hypothesis that a TH2 signature within tumor stroma was predictive of a poor prognosis, we Received October 7, 2011; revised February 8, 2012; accepted February 27, 2012; show that the HepaRG–LX2 signature recapitulates the MIM published OnlineFirst March 14, 2012.

References 1. Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. tumor progression in human hepatocellular carcinomas. Hepatology Cell 2011;144:646–74. 2001;34:82–8. 2. Bhowmick NA, Chytil A, Plieth D, Gorska AE, Dumont N, Shappell S, 9. Thorgeirsson SS, Grisham JW. Molecular pathogenesis of human et al. TGF-beta signaling in fibroblasts modulates the oncogenic hepatocellular carcinoma. Nat Genet 2002;31:339–46. potential of adjacent epithelia. Science 2004;303:848–51. 10. Yang JD, Nakamura I, Roberts LR. The tumor microenvironment in 3. Trimboli AJ, Cantemir-Stone CZ, Li F, Wallace JA, Merchant A, Crea- hepatocellular carcinoma: current status and therapeutic targets. sap N, et al. Pten in stromal fibroblasts suppresses mammary epithelial Semin Cancer Biol 2011;21:35–43. tumours. Nature 2009;461:1084–91. 11. Thorgeirsson SS, Lee JS, Grisham JW. Functional genomics of hepa- 4. Nelson CM, Bissell MJ. Of extracellular matrix, scaffolds, and signal- tocellular carcinoma. Hepatology 2006;43:S145–50. ing: tissue architecture regulates development, homeostasis, and 12. Hernandez-Gea V, Friedman SL. Pathogenesis of liver fibrosis. Annu cancer. Annu Rev Cell Dev Biol 2006;22:287–309. Rev Pathol 2011;6:425–56. 5. Polyak K, Haviv I, Campbell IG. Co-evolution of tumor cells and their 13. Loreal O, Levavasseur F, Fromaget C, Gros D, Guillouzo A, Clem- microenvironment. Trends Genet 2009;25:30–8. ent B. Cooperation of Ito cells and hepatocytes in the deposition of 6. Joyce JA, Pollard JW. Microenvironmental regulation of metastasis. an extracellular matrix in vitro. Am J Pathol 1993;143:538–44. Nat Rev 2009;9:239–52. 14. Gripon P, Rumin S, Urban S, Le Seyec J, Glaise D, Cannie I, et al. 7. El-Serag HB, Rudolph KL. Hepatocellular carcinoma: epidemiology Infection of a human hepatoma cell line by hepatitis B virus. Proc Natl and molecular carcinogenesis. Gastroenterology 2007;132:2557–76. Acad Sci U S A 2002;99:15655–60. 8. Theret N, Musso O, Turlin B, Lotrian D, Bioulac-Sage P, Campion JP, 15. Rumin S, Loreal O, Drenou B, Turlin B, Rissel M, Campion JP, et al. et al. Increased extracellular matrix remodeling is associated with Patterns of intermediate filaments, VLA integrins and HLA antigens in a

www.aacrjournals.org Cancer Res; 72(10) May 15, 2012 2541

Coulouarn et al.

new human biliary epithelial cell line sensitive to interferon-gamma. of aggressive human liver cancer. Carcinogenesis 2011;32: J Hepatol 1997;26:1287–99. 1434–40. 16. Cerec V, Glaise D, Garnier D, Morosan S, Turlin B, Drenou B, et al. 31. Budhu A, Forgues M, Ye QH, Jia HL, He P, Zanetti KA, et al. Prediction Transdifferentiation of hepatocyte-like cells from the human hepatoma of venous metastases, recurrence, and prognosis in hepatocellular HepaRG cell line through bipotent progenitor. Hepatology 2007;45: carcinoma based on a unique immune response signature of the liver 957–67. microenvironment. Cancer Cell 2006;10:99–111. 17. Xu L, Hui AY, Albanis E, Arthur MJ, O'Byrne SM, Blaner WS, et al. 32. Sancho-Bru P, Juez E, Moreno M, Khurdayan V, Morales-Ruiz M, Human hepatic stellate cell lines, LX-1 and LX-2: new tools for analysis Colmenero J, et al. Hepatocarcinoma cells stimulate the growth, of hepatic fibrosis. Gut 2005;54:142–51. migration and expression of pro-angiogenic genes in human hepatic 18. Lavergne E, Hendaoui I, Coulouarn C, Ribault C, Leseur J, Eliat PA, stellate cells. Liver Int 2010;30:31–41. et al. Blocking Wnt signaling by SFRP-like molecules inhibits in vivo cell 33. Xia Y, Chen R, Song Z, Ye S, Sun R, Xue Q, et al. Gene expression proliferation and tumor growth in cells carrying active beta-catenin. profiles during activation of cultured rat hepatic stellate cells by tumoral Oncogene 2011;30:423–33. hepatocytes and fetal bovine serum. J Cancer Res Clin Oncol 19. Coulouarn C, Gomez-Quiroz LE, Lee JS, Kaposi-Novak P, Conner EA, 2010;136:309–21. Goldina TA, et al. Oncogene-specific gene expression signatures at 34. Mantovani A, Allavena P, Sica A, Balkwill F. Cancer-related inflamma- preneoplastic stage in mice define distinct mechanisms of hepato- tion. Nature 2008;454:436–44. carcinogenesis. Hepatology 2006;44:1003–11. 35. Wong VW, Yu J, Cheng AS, Wong GL, Chan HY, Chu ES, et al. High 20. Coulouarn C, Factor VM, Thorgeirsson SS. Transforming growth serum interleukin-6 level predicts future hepatocellular carcinoma factor-beta gene expression signature in mouse hepatocytes predicts development in patients with chronic hepatitis B. Int J Cancer 2009; clinical outcome in human cancer. Hepatology 2008;47:2059–67. 124:2766–70. 21. Berriz GF, Beaver JE, Cenik C, Tasan M, Roth FP. Next generation 36. Calvisi DF, Ladu S, Gorden A, Farina M, Conner EA, Lee JS, et al. software for functional trend analysis. Bioinformatics 2009;25:3043–4. Ubiquitous activation of Ras and Jak/Stat pathways in human HCC. 22. Coulouarn C, Factor VM, Andersen JB, Durkin ME, Thorgeirsson SS. Gastroenterology 2006;130:1117–28. Loss of miR-122 expression in liver cancer correlates with suppression 37. Pilati C, Amessou M, Bihl MP, Balabaud C, Nhieu JT, Paradis V, et al. of the hepatic phenotype and gain of metastatic properties. Oncogene Somatic mutations activating STAT3 in human inflammatory hepato- 2009;28:3526–36. cellular adenomas. J Exp Med 2011;208:1359–66. 23. Lamb J, Crawford ED, Peck D, Modell JW, Blat IC, Wrobel MJ, et al. The 38. Rebouissou S, Amessou M, Couchy G, Poussin K, Imbeaud S, Pilati C, Connectivity Map: using gene-expression signatures to connect small et al. Frequent in-frame somatic deletions activate gp130 in inflam- molecules, genes, and disease. Science 2006;313:1929–35. matory hepatocellular tumours. Nature 2009;457:200–4. 24. De Minicis S, Seki E, Uchinami H, Kluwe J, Zhang Y, Brenner DA, et al. 39. Omenetti A, Porrello A, Jung Y, Yang L, Popov Y, Choi SS, et al. Gene expression profiles during hepatic stellate cell activation in Hedgehog signaling regulates epithelial-mesenchymal transition dur- culture and in vivo. Gastroenterology 2007;132:1937–46. ing biliary fibrosis in rodents and humans. J Clin Invest 2008;118: 25. Coulouarn C, Lefebvre G, Daveau R, Letellier F, Hiron M, Drouot L, et al. 3331–42. Genome-wide response of the human Hep3B hepatoma cell to proin- 40. Yang L, Wang Y, Mao H, Fleig S, Omenetti A, Brown KD, et al. Sonic flammatory cytokines, from transcription to translation. Hepatology hedgehog is an autocrine viability factor for myofibroblastic hepatic 2005;42:946–55. stellate cells. J Hepatol 2008;48:98–106. 26. Kaposi-Novak P, Lee JS, Gomez-Quiroz L, Coulouarn C, Factor VM, 41. Theret N, Musso O, L'Helgoualc'h A, Clement B. Activation of matrix Thorgeirsson SS. Met-regulated expression signature defines a subset metalloproteinase-2 from hepatic stellate cells requires interactions of human hepatocellular carcinomas with poor prognosis and aggres- with hepatocytes. Am J Pathol 1997;150:51–8. sive phenotype. J Clin Invest 2006;116:1582–95. 42. Coussens LM, Werb Z. Inflammation and cancer. Nature 2002;420: 27. Lee JS, Chu IS, Heo J, Calvisi DF, Sun Z, Roskams T, et al. Classi- 860–7. fication and prediction of survival in hepatocellular carcinoma by gene 43. Finak G, Bertos N, Pepin F, Sadekova S, Souleimanova M, Zhao H, expression profiling. Hepatology 2004;40:667–76. et al. Stromal gene expression predicts clinical outcome in breast 28. Lee JS, Heo J, Libbrecht L, Chu IS, Kaposi-Novak P, Calvisi DF, et al. A cancer. Nat Med 2008;14:518–27. novel prognostic subtype of human hepatocellular carcinoma derived 44. Kaimori A, Potter JJ, Choti M, Ding Z, Mezey E, Koteish AA. Histone from hepatic progenitor cells. Nat Med 2006;12:410–6. deacetylase inhibition suppresses the transforming growth factor 29. WooHG,ParkES,CheonJH,KimJH,LeeJS,ParkBJ,etal.Gene beta1-induced epithelial-to-mesenchymal transition in hepatocytes. expression-based recurrence prediction of hepatitis B virus-related Hepatology 2010;52:1033–45. human hepatocellular carcinoma. Clin Cancer Res 2008;14: 45. Niki T, Rombouts K, De Bleser P, De Smet K, Rogiers V, Schuppan D, 2056–64. et al. A histone deacetylase inhibitor, trichostatin A, suppresses 30. Coulouarn C, Factor VM, Conner EA, Thorgeirsson SS. Geno- myofibroblastic differentiation of rat hepatic stellate cells in primary mic modeling of tumor onset and progression in a mouse model culture. Hepatology 1999;29:858–67.

2542 Cancer Res; 72(10) May 15, 2012 Cancer Research

Hepatocyte−Stellate Cell Cross-Talk in the Liver Engenders a Permissive Inflammatory Microenvironment That Drives Progression in Hepatocellular Carcinoma

Cédric Coulouarn, Anne Corlu, Denise Glaise, et al.

Cancer Res 2012;72:2533-2542. Published OnlineFirst March 14, 2012.

Updated version Access the most recent version of this article at: doi:10.1158/0008-5472.CAN-11-3317

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2012/03/14/0008-5472.CAN-11-3317.DC1

Cited articles This article cites 45 articles, 6 of which you can access for free at: http://cancerres.aacrjournals.org/content/72/10/2533.full#ref-list-1

Citing articles This article has been cited by 10 HighWire-hosted articles. Access the articles at: http://cancerres.aacrjournals.org/content/72/10/2533.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 Department at Subscriptions [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/72/10/2533. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.