Integrative Genomic Analysis Identifies theCore Transcriptional Hallmarks of HumanHepatocellular CarcinomaCoralie Allain, Ga€elle Angenard, Bruno Cl�ement, and C�edric Coulouarn
Abstract
Integrative genomics helped characterize molecular heteroge-neity in hepatocellular carcinoma (HCC), leading to targeted drugcandidates for specificHCC subtypes. However, no consensuswasachieved for genes and pathways commonly altered in HCC.Here, we performed a meta-analysis of 15 independent datasets(n¼ 784 human HCC) and identified a comprehensive signatureconsisting of 935 genes commonly deregulated in HCC as com-pared with the surrounding nontumor tissue. In the HCC signa-ture, upregulated geneswere linked to early genomic alterations inhepatocarcinogenesis, particularly gains of 1q and 8q. The HCCsignature covered well-established cancer hallmarks, such as
proliferation, metabolic reprogramming, and microenvironmentremodeling, together with specific hallmarks associated withprotein turnover and epigenetics. Subsequently, the HCC signa-ture enabled us to assess the efficacy of signature-relevant drugcandidates, including histone deacetylase inhibitors that specif-ically reduced the viability of six human HCC cell lines. Overall,this integrative genomics approach identified cancer hallmarksrecurrently altered in humanHCC thatmay be targeted by specificdrugs. Combined therapies targeting common and subtype-spe-cific cancer networksmay represent a relevant therapeutic strategyin liver cancer. Cancer Res; 76(21); 6374–81. �2016 AACR.
IntroductionLiver carcinogenesis is a long process associated with multiple
risk factors that contribute to hepatocellular carcinoma (HCC)heterogeneity, for example, viral hepatitis, alcohol abuse, meta-bolic disorders, and obesity (1).MostHCCs develop in the settingof chronic liver disease associated with cycles of tissue destructionand regeneration that result in the activation of numerous sig-naling pathways and the accumulation of genomic alterations.Lately, next-generation sequencing approaches highlighted thelarge spectrum of mutational processes underlying the develop-ment ofHCC (2, 3).Other factors such as cell plasticity and tumormicroenvironment remodeling contribute to HCC heterogeneity,as well (4).
Over the last 2 decades, gene expression studies in HCC haverevealed the great diversity of transcriptional alterations occurringin liver carcinogenesis (5). By integrating gene expression profilesfrom various sources, a consensus has been achieved successfullyand 3 main HCC subtypes were identified (6). However, trans-lating thesefindings into individualized treatments is still amatterof debates. Surprisingly, the definition of recurrent transcriptionalalterations in HCC has been largely neglected, so far. Failure todefine a robust common transcriptional fingerprint in HCC may
have resulted from technical variabilities and/or from the inherentHCC molecular heterogeneity. The latter even raises the questionabout the existence of a substantial core expression signature inHCC (7). However, genomic characterization of HCC mousemodels demonstrated that various oncogenic pathways couldgenerate similar expression profiles at the tumor stage, suggestingthat a common HCC signature probably exists in humans (8).This observation prompted us to perform a meta-analysis ofpublicly available human HCC datasets that were generated overa period ofmore than 10 years. Starting from rawmicroarray dataandby using the same analysis algorithms to circumvent technicalvariabilities, our aimwas to define a universal and comprehensivetranscriptional signature in HCC and to determine whether thissignature could be useful to identify clinically relevant drugcandidates.
Materials and MethodsAnalysis of microarray datasets
The meta-analysis was performed using gene expression data-sets available from open databases, namely Gene ExpressionOmnibus (GEO; www.ncbi.nlm.nih.gov/geo/) and ArrayExpress(www.ebi.ac.uk/arrayexpress/). Twenty-eight liver-oriented data-sets were retrieved through a systematic review of the literature onPubMed and queries of the microarray databases using PubMedidentifiers or keywords associated with human liver carcinogen-esis and large-scale gene expression profiling, for example, HCC,microarray, and gene profiling (Supplementary Table S1). Beforeperforming the analysis, allmicroarray platforms (n¼19)were re-annotated using the Database for Annotation, Visualization andIntegratedDiscovery (DAVID; https://david.ncifcrf.gov/; ref. 9). Intotal, 24,085 nonredundant annotated genes were present at leastin 1 of 28 retrieved microarray datasets. Statistical analysis ofmicroarray data was performed using R-based BRB-ArrayTools as
INSERM, UMR 991, Liver Metabolisms and Cancer, University ofRennes, Rennes, France.
Note: Supplementary data for this article are available at Cancer ResearchOnline (http://cancerres.aacrjournals.org/).
Corresponding Author: C�edric Coulouarn, INSERM UMR 991, CHU Pontchaillou,2 rue Henri Le Guilloux, F-35033 Rennes, France. Phone: 33-2-23233881; Fax: 33-2-99540137; E-mail: [email protected]
previously described (10). Median-based normalization wasapplied and differentially expressed genes between the tumorand surrounding nontumor tissues were identified by a 2-sampleunivariate t test (P < 0.01) and a random variance model asdescribed (10). Permutation P values for significant genes werecomputed on the basis of 10,000 randompermutations. To definethe core HCC gene signature, only the genes up- and down-regulated in more than 50% datasets were retained. Bias in thedistribution of chromosomal location of genes included in thecore HCC signature was evaluated by c2 testing using the chro-mosomal locationof the genes present on the integrated dataset asa background, that is, 24,085 genes, see above. Clustering analysiswas performed using Cluster and TreeView algorithm as previ-ously described (10).
Data mining of the core HCC gene signatureSeveral tools dedicated to the discovery of specific enrichments
for biologic functions or canonical pathwayswere used, includingEnrichr algorithm (http://amp.pharm.mssm.edu/Enrichr) andIngenuity Pathway Analysis (IPA). IPA was also used to examinethe functional association between differentially expressed genesand to generate thehighest significant genenetworks using the IPAscoring system. Gene Set Enrichment Analysis (GSEA) was per-formed by using the Java-tool developed at the Broad Institute(Cambridge, MA; www.broadinstitute.org/gsea) as previouslydescribed (11). Connectivity map (cMap) algorithm was used tolink gene expression signatures with putative therapeutic mole-cules (10, 12). Briefly, from the permuted results cMap table, wefocused on negative enrichments to retain only the perturbagens(i.e., molecules) that potentially reverse the expression of genesincluded in the core HCC signature. Only the perturbagens with apermuted P < 0.003 and a number of replicate more than 5 wereretained.
Cell cultureA panel of 6 liver cancer cell lines was purchased from ATCC
(www.lgcstandards-atcc.org), including SNU-475 (ATCCCRL-2236, grade II–IV/V), SNU-449 (ATCC CRL-2234, gradeII–III/IV), SNU-423 (ATCC CRL-2238, grade III/IV), SNU-387(ATCC CRL-2237, grade IV/V), HepG2/C3A (ATCC CRL-10741), and SK-HEP-1 (ATCC HTB-52). ATCC performed celllines authentication by STRDNA profiling. The impact of selectedmolecules was evaluated within 6months after receipt. Cells weregrown in a RPMI-1640 medium supplemented with 100 U/mLpenicillin, 100 mg/mL streptomycin, and 10% FBS. Cultures wereperformed at 37�C in a 5% CO2 atmosphere. a-Estradiol,LY294002, rapamycin, resveratrol, sorafenib, and suberoylanilidehydroxamic acid (SAHA, also known as vorinostat) werepurchased from Santa Cruz Biotechnology. Trichostatin A waspurchased from Sigma-Aldrich. All molecules were solubilized ina DMSO solution. The concentration of each molecule used totreat the cells was determined on the basis of the detailed resultstable providedby cMap anda reviewof the literature. Cell viabilitywas evaluated using a PrestoBlue cell viability reagent (Invitro-gen) after 48 and72hours of treatmentwithDMSO, trichostatinA(1 mmol/L), a-estradiol (1 mmol/L), vorinostat (50 mmol/L),sorafenib (2 mmol/L), rapamycin (2 mmol/L), LY294002(50 mmol/L), and resveratrol (500 mmol/L). Independent cultureexperiments were performed (n ¼ 4 independent biologic repli-cates; n ¼ 4 technical replicates for each biologic replicate). Thesignificance of differences in cell viability between experimental
conditions was determined by a 2-tailed nonparametric Mann–Whitney test. For the microarray experiments, the concentrationswere optimized to induce 50% cell mortality after a 72-hour drugexposure, to allow the extraction of nucleic acids from the remain-ing viable cells. Accordingly, cells were treated with trichostatin A(0.6 mmol/L), a-estradiol (1 mmol/L), vorinostat (3.7 mmol/L),sorafenib (0.65 mmol/L), rapamycin (2 mmol/L), LY294002(45 mmol/L), and resveratrol (166 mmol/L).
Gene expression profilingTotal RNA was purified from SNU-423 cells at 50% confluence
with an miRNAeasy kit (Qiagen). Genome-wide expression pro-filing was performed using the Low-InputQuickAmp Labeling Kitand human SurePrint G3 8 � 60K pangenomic microarrays(Agilent Technologies) as previously described (10). Starting from150 ng total RNA, amplification yield was 7.0� 1.3 mg cRNA andspecific activity was 18.2 � 2.3 pmol Cy3 per mg of cRNA. Geneexpression data were processed using Feature Extraction andGeneSpring softwares (Agilent Technologies). Microarray datahave been deposited in NCBI's GEO and are accessible throughGEO Series accession numbers GSE79246 and GSE85257.
ResultsGeneration of a compendium of gene expression profiles inhuman HCC
The initial step of the study was to assemble well-annotatedhuman HCC gene expression profiles. From the main publicdatabases (i.e., GEO and ArrayExpress), 28 liver-oriented datasetswere retrieved including 1,657HCCgene expression profiles froma total of 3,047 (Fig. 1A). A detailed characterization of HCCdatasets is provided in the Supplementary Table S1. Of these 28datasets, 15 (MDS1–MDS15) included both HCC and sur-rounding nontumor tissues and thus were relevant to derivea core HCC signature defined as a set of genes differentiallyexpressed between HCC and surrounding nontumor tissues.Importantly, the gene expression profiles were generated fromvarious microarray platforms (i.e., from academic or industrialsources, including several updated contents). Consequently,variations in both the number and the nature of interrogatedgene features between platforms greatly impede the integrativeanalysis of the datasets. To overcome this issue, all the plat-forms (n ¼ 19, Fig. 1A) were reannotated by using the DAVIDdatabase (9). In total, 24,085 annotated genes were present atleast in one microarray dataset (Fig. 1A).
Identification of a comprehensive 935-gene HCC signaturelinked to recurrent early genomic alterations in livercarcinogenesis
By using the same normalization, filtration, and statisticalalgorithms, a set of genes significantly deregulated between HCCand surrounding nontumor tissues was identified [P < 0.01; falsediscovery rate (FDR) < 1%; n¼ 15 datasets; Fig. 1A]. The so-calleduniversal core HCC signature consisted of 935 genes significantlyderegulated in more than 50% HCC datasets and included 41%upregulated genes (Fig. 1B; Supplementary Table S2). Clusteringanalysis based on the expression of these genes in 15 datasetsrevealed a clear transcriptional homogeneity in the investigatedtumors (n¼ 784 HCC, Fig. 1B). Validating the gene selection, the935-gene HCC signature included well-known HCC biomarkers(e.g., GPC3, PEG10) or HCC suppressor genes (e.g., DLC1;
Transcriptional Hallmarks of Liver Cancer
www.aacrjournals.org Cancer Res; 76(21) November 1, 2016 6375
Supplementary Table S2). In addition, chromosomal mapping ofgenes included in the 935-gene HCC signature highlighted a biasin the distribution of up- and downregulated genes at specificlocations known to be frequently altered in HCC (Fig. 1C;Supplementary Table S3). Thus, upregulated genes were signifi-cantly (P < 0.001) enriched in 1q and 8q whose amplificationswere previously reported as the earliest events that occur inhumanhepatocarcinogenesis (13, 14). Similarly, downregulated genes
were enriched in loci frequently subjected to early deletions inHCC (e.g., 4q). These specific locations also correlated with thederegulation of HCC-associated genes (e.g., LAMC1 at 1q31,RAD21 at 8q24, or ALB at 4q13). Altogether, these observationsdemonstrated that transcriptional changes identified in the935-gene HCC signature significantly correlated with early andrecurrent structural genomic alterations linked to stepwise HCCcarcinogenesis in human.
Figure 1.
Large-scale meta-analysis of publicgenomic data identifies coretranscriptional hallmarks in humanHCC.A,analytic workflow of the meta-analysisthat led to the identification of a coregene expression signature in human HCC(935-gene HCC signature). B, heatmap ofthe 935-gene HCC signature. Induced andrepressed genes in HCC as compared withthe surrounding nontumor tissue arerepresented in red and green,respectively. Missing values aregray. Numbers on the top represent the15 HCC microarray datasets that wereinvestigated to derive the signature. C,specific enrichment of induced (left) andrepressed (right) genes of the 935-geneHCC signature on specific chromosomalarms. D, Gene Ontology analysis ofinduced (left) and repressed (right) genesof the 935-gene HCC signaturehighlighting functional cancer hallmarks.E, top gene networks identified by IPA.Highlighted networks were associatedwith cell cycle and proliferation forinduced genes (left) and with cellularmetabolism for repressed genes (right).ST, surrounding nontumor tissue.
Allain et al.
Cancer Res; 76(21) November 1, 2016 Cancer Research6376
The 935-gene HCC signature covers well-established cancerhallmarks
Data mining of the 935-gene HCC signature (Fig. 1D and E;Supplementary Tables S4 and S5) identified gene networks thatreflect a definite transcriptional reprogramming associated withpreviously described hallmarks of cancer cells, as exemplifiedthereafter (15).
Sustained proliferation and evasion to growth suppressors. Genenetworks and Gene Ontology analysis clearly demonstrated thatactive cell proliferation is the most prominent feature within the935-gene HCC signature (Fig. 1E). Thus, numerous cell cyclinggenes were upregulated (Supplementary Table S2), includingcyclins (e.g., CCNB1, CCNE2), cyclin-dependent kinases andinhibitors (e.g., CDK1, CDK4, CDKN2A), cell-cycle and celldivision checkpoints (e.g., BUB1, MAD2L1, NDC80, RAN),together with well-established proliferation markers (e.g., PCNA,MKI67). Accordingly, numerous genes related to DNA replicationwere induced (e.g., CDC6, MCM2-4, RCF4). These observationscoincided with the induction of proproliferative pathways andgrowth factors (e.g., ERBB3, MAPK6, YWHAH, FIBP) and therepression of negative feedback regulators (e.g., PINK1, DUSP1).
Genome instability, oxidative stress, and apoptosis resistance. Asustained proliferation phenotype generally associates withDNA damages, accumulation of mutations, and genome insta-bility, which ultimately lead to apoptosis (15). Interestingly,the 935-gene HCC signature highlighted gene deregulationsthat clearly sign mechanisms of DNA damages associatedwith DNA hyperreplication together with resistance to celldeath (Supplementary Table S2). Thus, several genes essentialfor DNA double-strand break repair (e.g., RAD21, RUVBL2) andnucleotide excision repair (e.g., ERCC3, FEN1, OGG1) wereinduced. Notably, DNA glycosylase OGG1 is involved in theexcision of 8-oxoguanine, a mutagenic base byproduct thatoccurs as a result of exposure to reactive oxygen species. Furtherimplying oxidative stress in this process, key regulators of redoxhomeostasis were repressed in the 935-gene HCC signature(e.g., CTH, CBS, NFE2L2, PRDX4). Besides, several inducers ofapoptosis were repressed (e.g., CRADD, DAPK1), whereas genesencoding proteins that prevent apoptotic cell death wereinduced (e.g., BIRC5, DAD1).
Metabolic reprogramming. Deregulation of metabolism-associ-ated genes (e.g., lipid, carbohydrate and amino acid metabo-lisms) was a prominent feature in the 935-gene HCC signature,in agreement with metabolic changes observed in cancer cellsduring tumor onset and progression (16). This was particularlynoticeable for downregulated genes involved in liver-specificmetabolisms (Fig. 1D; Supplementary Tables S4 and S5),including those encoding acute-phase plasma proteins (e.g.,A2M, ALB, CP), components of complement and coagulationcascades (e.g., C5-9, CFB, F2), or detoxication enzymes (e.g.,ADH1A, CYP2E1). While such enrichment may reflect a loss ofhepatocyte differentiation accompanying tumor development,other deregulations affecting key enzymes of bioenergetics andbiosynthesis may be directly linked to a specific reprogrammingof cellular metabolism (Supplementary Table S2). As example,ACLY and CS were recurrently upregulated in HCC. ACLYencodes ATP citrate lyase, the primary enzyme for the synthesis
of acetyl-CoA, a major intermediate for biosynthetic pathwaysincluding lipogenesis. CS encodes the citrate synthase, a keyenzyme in the tricarboxylic acid cycle that contributes tolipogenesis by enhancing the conversion of glucose to lipids.While lipogenesis was enhanced, genes encoding key compo-nents of lipid catabolism were repressed, including genesinvolved in the mitochondrial fatty acid b-oxidation pathway(e.g., ACADM, ACADL, ECHS1). In addition, we observed ashift toward the downregulation of genes involved in gluco-neogenesis (e.g., PCK1, FBP1) that may contribute to enhancethe rate of glycolysis in HCC cancer cells.
Induction of angiogenesis. Fueling proliferative and metabolicallyactive tumor cells frequently correlates with an enhanced angio-genesis to support nutrient supply (15). Accordingly, the935-gene HCC signature included several angiogenesis-relatedgenes. As example, PLG encoding plasminogen was stronglyrepressed in almost all HCC (Supplementary Table S2). Of note,plasminogen is activated byproteolysis and converted to plasmin,an activator of matrix metalloproteases, and angiostatin, a potentinhibitor of angiogenesis. Similarly, AMOTL2 encoding an angio-motin like 2 protein was repressed. Angiomotin is known tomediate the inhibitory effect of angiostatin on tube formation.While these angiogenesis inhibitors were repressed, endothelialcell markers were induced (e.g., ESM1).
Microenvironment remodeling and invasion. In agreement with aloss of hepatocyte differentiation (see above), we observed adecreased expression of the epithelial marker E-cadherin (CDH1)in more than 70% datasets (Supplementary Table S2). Loss ofE-cadherin is frequently observed in cancer and notably contri-butes to tumor progression by increasing proliferation and inva-sion. The 935-gene HCC signature was also significantly enrichedin genes encoding extracellular matrix proteins (e.g., COL4A1,COL4A2, LAMC1). Induction of these genes frequently correlateswith changes in the cellular microenvironment associated withliver fibrosis and tumor invasion, a process largely controlled byTGFb pathway. Interestingly, SMAD2 that mediates TGFb signalswas either induced (80% datasets) or repressed, highlighting thefunctional duality of the TGFb pathway, acting as a tumor-pro-moting or a tumor-suppressing factor in cancer, including HCC(17). SFRP1, acting as a negative modulator of the Wnt/b-cateninpathway frequently activated in HCC microenvironment, wasrepressed.
Avoiding immune destruction. Several immune-associated geneswere identified andmost of themwere repressed (SupplementaryTable S2), including activation markers of cytotoxic T lympho-cytes and natural killer (NK) cells (e.g., CD69, KLRK1, GZMA). Inaddition, immunoregulatory mediators with chemotactic activityfor competent immune cells were repressed (e.g., CCL19, CCL2/MCP1, CXCL12/SDF1).
Emerging hallmarks enriched in HCC. In addition to the well-characterized hallmarks of cancer cells described above, 2 morediscrete but promising functional categories emerged from theanalysis of the 935-gene HCC signature. These specific hallmarkswere associated with protein turnover and epigenetics (Supple-mentary Table S2). In agreementwith an activemetabolic activity,we observed an increased expression of several translation-
Transcriptional Hallmarks of Liver Cancer
www.aacrjournals.org Cancer Res; 76(21) November 1, 2016 6377
associated factors, including genes encoding ribosomal subunits(e.g., RPS5, RPL38) and translational machinery (e.g., EIF4G2).Unexpectedly, genes involved in protein ubiquitination (e.g.,UBE2A, UBE2C) and degradation through the 26S proteasomepathway (e.g., PSMA1, PSMA6, PSMD2, PSMD4) were similarlyoverexpressed, evocative of a proteotoxic stress associated with aprotein hyperproduction and/or mysfolding (18). The secondprominent promising HCC hallmark is the upregulation ofnumerous genes acting at the epigenetic level (SupplementaryTable S2). Thus, the 935-gene HCC signature included importantregulators of chromatin assembly and remodeling (e.g.,CHAF1A,HDAC1,HDAC5,HMGB2), components of the polycomb-repres-sive complex 2 (e.g., EZH2, SUZ12) that catalyzes the trimethyla-tion of H3K27 (H3K27me3), andmaster regulators of microRNAprocessing (e.g., DROSHA).
The 935-gene HCC signature highlights drug candidates forsystemic therapies
Next, the cMap algorithmwas used to identify relevant drugs inHCC, as previously described (10). We selected drugs that gener-ated gene expression profiles inversely correlated with the935-gene HCC signature with the highest confidence (P <0.003). Six candidate molecules were identified, including twohistone deacetylase (HDAC) inhibitors (trichostatin A and vorino-stat), PI3K inhibitor LY294002, mTOR inhibitor sirolimus (alsoknown as rapamycin), a-estradiol, and resveratrol (Fig. 2A; Sup-plementary Table S6). The impact of these drugs as comparedwithsorafenib that is currently used for the treatment of advancedHCC(19) was evaluated on the viability of 6 HCC-derived cell lines.Remarkably, except for a-estradiol, all the drugs significantly (P <0.05) reduced cell viability (Fig. 2B). The SNU-423 cell line was
Figure 2.
The 935-gene HCC signature highlights relevant drug candidates. A, cMap algorithm applied to the 935-gene HCC signature identified 5 candidate molecules:trichostatin A, LY294002, vorinostat, rapamycin, resveratrol, and a-estradiol. The barview shows the enrichment of treatment instance ordered by theircorresponding connectivity scores. All selected drug candidates exhibit a significant negative enrichment as regard to the 935-gene HCC signature. B, impactof identified drug candidates on the viability of 6 human HCC cell lines. Cell viability was determined after 72 hours of treatment with drugs at variousconcentrations (see Materials and Methods; n¼ 4 independent experiments). Each bar represents cell viability (mean� SD) in the investigated cell lines (SNU-475,SNU-449, SNU-423, SNU-387, HepG2/C3A, SK-Hep-1, from left to right). A 2-tailed nonparametric Mann–Whitney test was used for the comparison betweenexperimental groups for each cell line (DMSOvs. untreated control anddrug vs. DMSO). � ,P<0.05.C,GSEAof the inducedgenes (top three panels) and the repressedgenes (bottom three panels) from the 935-gene HCC signature in the gene expression profiles of the SNU-423 cell line treated with DMSO or vorinostat (left twopanels), resveratrol (middle two panels), and sorafenib (right two panels). Positive and negative enrichment scores (ES) were determined by GSEA.
Allain et al.
Cancer Res; 76(21) November 1, 2016 Cancer Research6378
used as a paradigm to test the drug-induced transcriptional repro-gramming based on the positive response of this cell line to drugtreatments. Thus, gene expression profiling combined with GSEAdemonstrated that thesedrugs could, at least partly, reverse the coretranscriptional programming of HCC cells, as exemplified byvorinostat, resveratrol, and sorafenib (Fig. 2C). Indeed, upregu-lated genes of the 935-gene HCC signature were enriched in thegene expression profiles of DMSO-treated cells (i.e., negativelyenriched in drug-treated cells), whereas downregulated geneswerepositively enriched indrug-treated cells, evidencingadrug-inducedtranscriptional reprogramming (Fig. 2C). These resultswere furthervalidated in SNU-387 and HepG2/C3A cell lines (SupplementaryFig. S1). Finally, data mining of genes included in the coreenrichment for eachdrug (Supplementary Table S7) demonstratedthat the identified drugs are able to reverse the expression ofnumerous genes involved in the functional networks that operateto establish the hallmarks of HCC cancer cells.
DiscussionHCC is a deadly cancer worldwide,mainly due to late diagnosis
and the absence of effective treatments for advanced stages of thedisease. A strategy of HCC stratification intomolecular subgroupshas been developed worldwide with the objective to translatethese knowledge into individualized treatments (6, 20, 21). Thus,Hoshida and colleagues reported the existence of 3 major HCCsubtypes referred to as S1, S2, and S3 (6). These subtypes wereassociated with specific biologic and clinical features (Fig. 3). S1and S2 subtypes included aggressive HCC and were associatedwith an aberrant activation of the WNT signaling pathway byTGFb (S1 subtype) or a progenitor-like phenotype associatedwithMYC and AKT activation (S2 subtype). S3 subtype included goodprognosis HCC that exhibited a hepatocyte-like phenotype andCTNNB1 mutations (Fig. 3). The classification into molecularsubtypes highlighted specific pathways for drug targeting, includ-ing TGFb, WNT, and AKT inhibitors. Conceptually, our study isslightly different, as it hypotheses that optimized treatmentsshould take into consideration not only themolecular alterationsthat occur in specific HCC subtypes but also those recurrentlyaltered in all tumors (Fig. 3). However, the definition of coretranscriptional hallmarks in HCC has never been really exploredin details, so far. Our study fills this gap and demonstrates theexistence of a substantial transcriptional homogeneity in HCC.Hence, we provide a robust, exhaustive, and comprehensivesignature of 935 genes commonly deregulated in human HCCfrom data generated in various laboratories and covering all theknown etiologies. We hypothesized that analyzingmultiple data-sets with the same algorithms significantly increases the accuracyof the findings as compared with conclusions raised from singlestudies. Consistent with the specific objectives of Hoshida's study(i.e., the definitionof signatures formolecularHCCsubtypes) andour study (i.e., the definition of a common transcriptional fin-gerprint for all HCC), there was almost no overlap between thegenes included in the S1–S3 signatures and our core HCC signa-ture. Indeed, in the 935-gene HCC signature, only 3%, 2%, and13% genes overlapped with the S1, S2, and S3 signatures, respec-tively (Supplementary Fig. S2). Besides, most of genes overlap-ping with the S3 signature weremetabolism-associated genes thatwere shown to be relatively upregulated in S3-HCC, as comparedwith poorly differentiated HCC of S1 and S2 subtypes (5, 6).However, regardless of HCC subtypes, these genes were com-
monly repressedwhen comparedwith the surroundingnontumortissues, as observed in our core HCC signature. Each of thesignatures (i.e., S1–S3 and core) are then specific in term ofconstituting genes but could be applied to large HCC cohorts.Altogether, we believe that our signature constitutes a unique andspecific fingerprint of recurrent and common transcriptionalalterations in HCC.
One striking observation is that the 935-gene HCC signaturecovers almost all hallmark capabilities of cancer cells describedpreviously (Fig. 3; ref. 15). It is noteworthy that none of the genesincluded in the 935-gene HCC signature were directly related toreplicative immortality. This cancer hallmark is largely controlledby telomerase activity. Actually, telomerase has been shown to bereactivated in more than 90% HCC, mostly due to TERT ampli-fication and somatic mutations or HBV insertion in the TERTpromoter (3). More than being only recurrent in HCC, genetic
Figure 3.
Therapeutic strategies integrating core and subtype-specific transcriptionalhallmarks in humanHCC. Themeta-analysis presented in this study enhances ourknowledge on the molecular characterization of HCC tumors and echoesprevious studies focused on HCC stratification. Recently, a consensus has beenachieved identifying three main HCC molecular subtypes (S1, S2, and S3; ref. 6).S1 and S2 subtypes are associated with a poor prognosis and included highlyproliferative and poorly differentiated tumors. S1 subtype is particularlyassociated with the activation of a prometastatic TGFb signaling and the S2subtype included tumors with a progenitor-like phenotype. S3 subtype isassociated with a better prognosis and low proliferation and includes well-differentiated tumors that retained a hepatocyte-like phenotype. At themolecular level, the S3 subtype is enriched in tumors that exhibit activatingmutations in CTNNB1 gene encoding b-catenin. HCC stratification intohomogeneous subtypes opened new avenues for personalized targetedtherapies. The highlighting of core transcriptional hallmarks in human HCCsuggests that efficient therapies should consider drugs targeting commonHCC hallmarks together with drugs targeting specific HCC subtypes.CTNNB1mut, mutated b-catenin gene.
Transcriptional Hallmarks of Liver Cancer
www.aacrjournals.org Cancer Res; 76(21) November 1, 2016 6379
alterations in TERT represent early events in human hepatocarci-nogenesis (22). Importantly, we show that the 935-gene HCCsignature was associated with early chromosomal alterationsdescribed in human hepatocarcinogenesis (14). Thus, the signa-ture should include relevant candidate biomarkers for early HCCdiagnosis, including secreted biomarkers (e.g., SPINK1). In addi-tion, identifying cell surface markers (e.g., ligands and/or recep-tors) overexpressed in most HCC may lead to new drug targetswith high specificity. Accordingly, innovative nanoparticles maybe formulated including specific peptides derived from the 935-geneHCC signature, to increase drug delivery while reducing drugside effects. Interestingly, the 935-geneHCCsignature highlightedspecific hallmarks associated with protein turnover (i.e., synthesisand proteasomal degradation) and epigenetics (Fig. 3). Over-activation of the ubiquitin–proteasome system has been reportedin several cancers and proteasome inhibitors, for example, borte-zomib provided promising results in treating hematologic malig-nancies (23). In experimentalmodels ofHCC, the combination ofsorafenib and bortezomib demonstrated synergistic antitumoreffects through AKT inactivation (24).
Through cMap algorithm, drug candidates were identified andvalidated in several HCC cell lines, including resveratrol, HDACinhibitors (HDACi), and PI3K/AKT/mTOR inhibitors, in agree-ment with the landscape of genetic alterations frequentlyobserved in HCC (2). It is noteworthy that cMap results weremainly derived from MCF7 breast cancer cells (SupplementaryTable S6) that may harbor a different spectrum ofmutations thanHCC cell lines. Interestingly, the comparison of genetic profilesthough the Cancer Cell Line Encyclopedia project (www.broad-institute.org/ccle) identified a mutation signature consisting in16 genes mutated both in MCF7 and in at least 3 of 6 HCC celllines investigated. This signature included key cancer genesinvolved in EMT, tumor growth, metastasis, and angiogenesis(e.g.,AAK1, CDK11B, ILK,MAP3K1,MAP3K14, NCAM1, PRKDC,and VEGFC). One can hypothesize that the efficacy of the iden-tified drugs may be related to these signaling pathways. Resver-atrol and HDACi were shown to target multiple cancer hallmarksin preclinical models, including effects on growth inhibition, celldifferentiation, angiogenesis, and immunosurveillance (25, 26).At the molecular level, resveratrol was reported to modulatecancer-related signaling pathways, including FAS/FASL, PI3K/AKT, NF-kB, and WNT (26). HDACi-induced growth arrest hasbeen linked to the induction of the CDK inhibitors p21 and p27.HDACi-induced cell death involves caspase-dependent and -inde-pendent pathways. HDACi also cause hyperacetylation and inac-tivation of HSP90, leading to the degradation of proteins thatrequire the chaperone function of HSP90, including some onco-proteins (25). HDACi can block tumor angiogenesis by inhibitinghypoxia-inducible factors and expression of VEGF (27) andimpair immunosurveillance by reducing viability and effectorfunctions of NK cells (28). In liver cancer, we showed that HDACicould interfere with tumor–stroma crosstalk (10) and loss of
hepatocyte differentiation (29). In HCC, the results of a multi-center phase I/II study in patients with unresectable tumorsdemonstrated that HDAC inhibition with belinostat was well-tolerated and associated with tumor stabilization (30).
However, the picture is obviously not so simple given thatbesides sorafenib, most monotherapies evaluated in phase IIIclinical trials failed to improve the survival of patients withadvanced HCC (31). In addition, recent results of combinedtherapies, for example, sorafenib associated with erlotinib ordoxorubicin failed to demonstrate meaningful clinical benefits(32, 33). This raises questions about the optimal backbone notonly for drug combinations (31, 32) but also for combinedtreatment modalities, including surgery with adjuvant multidrugchemotherapies and biotherapies, personalized radioemboliza-tion, and immune-based therapies. Our comprehensive 935-geneHCC signature may help resolve this issue by identifying combi-nations of treatments to target various signaling pathways alteredin HCC. We believe that evaluating these strategies in combina-tion with molecules targeting pathways deregulated in specificHCC subtypes may represent promising approaches, particularlyin clinical trials where patients for each subtypes are selected onthe basis of the expression-specific biomarkers.
Disclosure of Potential Conflicts of InterestNo potential conflicts of interest were disclosed.
Authors' ContributionsConception and design: C. CoulouarnDevelopment of methodology: C. Allain, C. CoulouarnAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): C. Allain, C. CoulouarnAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): C. CoulouarnWriting, review, and/or revision of themanuscript: B. Cl�ement, C. CoulouarnAdministrative, technical, or material support (i.e., reporting or organizingdata, constructing databases): C. Allain, G. Angenard, B. Cl�ementStudy supervision: C. Coulouarn
AcknowledgmentsThe authors thank all the researchers who contributed to the genomic
characterization of HCC over the last two decades allowing this meta-analysisto be performed. We thank Dr. Snorri S. Thorgeirsson (NCI, Bethesda, MD) forcritical reading of this article.
Grant SupportThis research was supported by INSERM and University of Rennes 1. C.
Coulouarn is supported by grants from the National Cancer Institute(Canc�eropoles Ile-de-France & Grand-Ouest), Ligue contre le cancer (CD35,44, 49), and Association Francaise pour l'Etude du Foie, France.
The costs of publication of this articlewere defrayed inpart by the payment ofpage charges. This article must therefore be hereby marked advertisement inaccordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Received June 7, 2016; revised August 8, 2016; accepted September 1, 2016;published OnlineFirst September 12, 2016.
References1. Thorgeirsson SS, Grisham JW. Molecular pathogenesis of human hepato-
et al. Exome sequencing of hepatocellular carcinomas identifies newmutational signatures and potential therapeutic targets. Nat Genet 2015;47:505–11.
3. Totoki Y, Tatsuno K, Covington KR, Ueda H, Creighton CJ, Kato M, et al.Trans-ancestry mutational landscape of hepatocellular carcinoma gen-omes. Nat Genet 2014;46:1267–73.
4. Marquardt JU, Andersen JB, Thorgeirsson SS. Functional and geneticdeconstruction of the cellular origin in liver cancer. Nat Rev Cancer 2015;15:653–67.
Allain et al.
Cancer Res; 76(21) November 1, 2016 Cancer Research6380
5. Hoshida Y, Toffanin S, Lachenmayer A, Villanueva A,Minguez B, Llovet JM.Molecular classification and novel targets in hepatocellular carcinoma:recent advancements. Semin Liver Dis 2010;30:35–51.
6. Hoshida Y, Nijman SM, KobayashiM, Chan JA, Brunet JP, ChiangDY, et al.Integrative transcriptome analysis reveals commonmolecular subclasses ofhuman hepatocellular carcinoma. Cancer Res 2009;69:7385–92.
7. van Malenstein H, van Pelt J, Verslype C. Molecular classification ofhepatocellular carcinoma anno 2011. Eur J Cancer 2011;47:1789–97.
8. Coulouarn C, Gomez-Quiroz LE, Lee JS, Kaposi-Novak P, Conner EA,Goldina TA, et al. Oncogene-specific gene expression signatures at pre-neoplastic stage in mice define distinct mechanisms of hepatocarcinogen-esis. Hepatology 2006;44:1003–11.
9. Huang daW, ShermanBT, Lempicki RA. Systematic and integrative analysisof large gene lists using DAVID bioinformatics resources. Nat Protoc2009;4:44–57.
10. Coulouarn C, Corlu A, Glaise D, Guenon I, Thorgeirsson SS, Clement B.Hepatocyte-stellate cell cross-talk in the liver engenders a permissiveinflammatorymicroenvironment that drives progression in hepatocellularcarcinoma. Cancer Res 2012;72:2533–42.
11. Coulouarn C, Factor VM, Andersen JB, Durkin ME, Thorgeirsson SS. Loss ofmiR-122 expression in liver cancer correlates with suppression of the hepaticphenotype and gain of metastatic properties. Oncogene 2009;28:3526–36.
12. Lamb J, Crawford ED, Peck D, Modell JW, Blat IC, Wrobel MJ, et al. Theconnectivity map: using gene-expression signatures to connect smallmolecules, genes, and disease. Science 2006;313:1929–35.
13. Longerich T, Mueller MM, Breuhahn K, Schirmacher P, Benner A, Heiss C.Oncogenetic tree modeling of human hepatocarcinogenesis. Int J Cancer2012;130:575–83.
14. Midorikawa Y, Yamamoto S, Tsuji S, Kamimura N, Ishikawa S, Igarashi H,et al. Allelic imbalances and homozygous deletion on 8p23.2 for stepwiseprogression of hepatocarcinogenesis. Hepatology 2009;49:513–22.
15. Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. Cell2011;144:646–74.
16. Cairns RA, Harris IS, Mak TW. Regulation of cancer cell metabolism.Nat Rev Cancer 2011;11:85–95.
17. Coulouarn C, Factor VM, Thorgeirsson SS. Transforming growth factor-beta gene expression signature in mouse hepatocytes predicts clinicaloutcome in human cancer. Hepatology 2008;47:2059–67.
18. Luo J, Solimini NL, Elledge SJ. Principles of cancer therapy: oncogene andnon-oncogene addiction. Cell 2009;136:823–37.
19. European Association For The Study Of The L, European Organisation ForR, Treatment Of C. EASL-EORTC clinical practice guidelines: managementof hepatocellular carcinoma. J Hepatol 2012;56:908–43.
20. Lee JS, Heo J, Libbrecht L, Chu IS, Kaposi-Novak P, Calvisi DF, et al. A novelprognostic subtype of human hepatocellular carcinoma derived fromhepatic progenitor cells. Nat Med 2006;12:410–6.
21. Yamashita T, Ji J, Budhu A, Forgues M, Yang W, Wang HY, et al.EpCAM-positive hepatocellular carcinoma cells are tumor-initiatingcells with stem/progenitor cell features. Gastroenterology 2009;136:1012–24.
22. Zucman-Rossi J, Villanueva A, Nault JC, Llovet JM. Genetic landscape andbiomarkers of hepatocellular carcinoma. Gastroenterology 2015;149:1226–39e4.
23. Richardson PG, Barlogie B, Berenson J, Singhal S, Jagannath S, Irwin D,et al. A phase 2 study of bortezomib in relapsed, refractory myeloma.N Engl J Med 2003;348:2609–17.
24. Chen KF, Yu HC, Liu TH, Lee SS, Chen PJ, Cheng AL. Synergisticinteractions between sorafenib and bortezomib in hepatocellular car-cinoma involve PP2A-dependent Akt inactivation. J Hepatol 2010;52:88–95.
25. Minucci S, Pelicci PG. Histone deacetylase inhibitors and the promise ofepigenetic (and more) treatments for cancer. Nat Rev Cancer 2006;6:38–51.
26. Shukla Y, Singh R. Resveratrol and cellular mechanisms of cancer preven-tion. Ann N Y Acad Sci 2011;1215:1–8.
27. Xu WS, Parmigiani RB, Marks PA. Histone deacetylase inhibitors: molec-ular mechanisms of action. Oncogene 2007;26:5541–52.
28. Rossi LE, Avila DE, Spallanzani RG, Ziblat A, Fuertes MB, Lapyckyj L, et al.Histone deacetylase inhibitors impair NK cell viability and effector func-tions through inhibition of activation and receptor expression. J LeukocBiol 2012;91:321–31.
29. Dubois-Pot-Schneider H, Fekir K, Coulouarn C, Glaise D, Aninat C,Jarnouen K, et al. Inflammatory cytokines promote the retrodifferentiationof tumor-derived hepatocyte-like cells to progenitor cells. Hepatology2014;60:2077–90.
30. Yeo W, Chung HC, Chan SL, Wang LZ, Lim R, Picus J, et al. Epigenetictherapy using belinostat for patients with unresectable hepatocellularcarcinoma: a multicenter phase I/II study with biomarker and pharma-cokinetic analysis of tumors from patients in the mayo phase IIconsortium and the cancer therapeutics research group. J Clin Oncol2012;30:3361–7.
31. Llovet JM, Hernandez-Gea V. Hepatocellular carcinoma: reasons for phaseIII failure and novel perspectives on trial design. Clin Cancer Res2014;20:2072–9.
32. Lencioni R, Llovet JM, Han G, TakWY, Yang J, Guglielmi A, et al. Sorafenibor placebo plus TACE with doxorubicin-eluting beads for intermediate-stage HCC: phase II, randomized, double-blind SPACE trial. J Hepatol2016;64:1090–8.
33. Zhu AX, Rosmorduc O, Evans TR, Ross PJ, Santoro A, Carrilho FJ, et al.SEARCH: a phase III, randomized, double-blind, placebo-controlled trialof sorafenib plus erlotinib in patients with advanced hepatocellular car-cinoma. J Clin Oncol 2015;33:559–66.
www.aacrjournals.org Cancer Res; 76(21) November 1, 2016 6381
2016;76:6374-6381. Published OnlineFirst September 12, 2016.Cancer Res Coralie Allain, Gaëlle Angenard, Bruno Clément, et al. Hallmarks of Human Hepatocellular CarcinomaIntegrative Genomic Analysis Identifies the Core Transcriptional
Updated version
10.1158/0008-5472.CAN-16-1559doi:
Access the most recent version of this article at:
