Genetic Landscape and Biomarkers of Hepatocellular Carcinoma EEI TESTING GENETIC EEISAND GENETICS

Gastroenterology 2015;149:1226–1239

Jessica Zucman-Rossi1,2,3,4 Augusto Villanueva5,6 Jean-Charles Nault1,2,7 Josep M. Llovet5,8,9

1Inserm, UMR-1162, Génomique Fonctionnelle des Tumeurs Solides, Equipe Labellisée Ligue Contre le Cancer, Institut Universitaire d’Hematologie, Paris, France; 2Université Paris Descartes, Labex Immuno-Oncology, Sorbonne Paris Cité, Faculté de Médecine, Paris, France; 3Université Paris 13, Sorbonne Paris Cité, Unité de Formation et de Recherche Santé, Médecine, Biologie Humaine, Bobigny, France; 4Université Paris Diderot, Paris; 5Liver Cancer Program, Division of Liver Diseases, Department of Medicine, Tisch Cancer Institute, Icahn School of Medicine at Mount Sinai, New York, New York; 6Division of Hematology and Medical Oncology, Department of Medicine, Icahn School of Medicine at Mount Sinai, New York, New York; 7Service d’hépatologie, Hôpital Jean Verdier, Hôpitaux Universitaires Paris-Seine-Saint-Denis, Assistance-Publique Hôpitaux de Paris, Bondy, France; 8Liver Cancer Translational Research Laboratory, Barcelona-Clínic Liver Cancer Group, Institut d’Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), Liver Unit, CIBEREHD, Hospital Clínic, Barcelona, Catalonia, Spain; 9Institució Catalana de Recerca i EstudisAvançats(ICREA),Barcelona, Catalonia, Spain

Hepatocellular carcinoma (HCC) has emerged as a ur understanding of the molecular pathogenesis of major cause of cancer-related death. Its mortality has O hepatocellular carcinoma (HCC) has improved increased in Western populations, with a minority of significantly in the last decade. Certainly, cumulative data patients diagnosed at early stages, when curative from high-throughput analyses of large number of samples treatments are feasible. Only the multikinase inhibitor have provided an accurate landscape of HCC genetic alter- sorafenib is available for the management of advanced ations. This has enabled us to delineate some of the key cases. During the last 10 years, there has been a clear events that might dominate tumor development and pro- delineation of the landscape of genetic alterations in gression. Hopefully, translation of this knowledge into new HCC, including high-level DNA amplifications in chro- VEGFA FGF19/CNND1 targets and biomarkers might impact HCC decision making, mosome 6p21 ( ) and 11q13 ( ), as ’ well as homozygous deletions in chromosome 9 and ultimately improve patient s outcomes. There are, (CDKN2A). The most frequent mutations affect TERT however, some challenges ahead. First, the most prevalent promoter (60%), associated with an increased telo- oncogenic mutations are currently undruggable. Second, in merase expression. TERT promoter can also be affected those cases where a molecular subclass or a dominant by copy number variations and hepatitis B DNA in- pathway can be identified, there have been minimal efforts sertions, and it can be found mutated in preneoplastic to translating this knowledge into clinical trials. Third, the lesions. TP53 and CTNNB1 are the next most prevalent debate on the role of tumor heterogeneity and its assess- mutations, affecting 25%L30% of HCC patients, that, ment in advanced stages has fueled some skepticism on in addition to low-frequency mutated genes (eg, AXIN1, the future of single-biopsy stratified treatments. Finally, our ARID2, ARID1A, TSC1/TSC2, RPS6KA3, KEAP1, MLL2), ability to manipulate the immune system in the therapeutic help define some of the core deregulated pathways in arsenal against HCC is promising but remains to be HCC. Conceptually, some of these changes behave as evaluated. This review will provide an overview of the ge- prototypic oncogenic addiction loops, being ideal bio- netic changes involved in HCC development and progres- markers for specific therapeutic approaches. Data from sion, as well as discuss the role of molecular markers to genomic profiling enabled a proposal of HCC in 2 ma- stratify patients based on their prognosis and response to jor molecular clusters (proliferation and nonprolifera- therapies. tion), with differential enrichment in prognostic signatures, pathway activation and tumor phenotype. fi Translation of these discoveries into speci cthera- Abbreviations used in this paper: AFB1, aflatoxin B1; EGF, epidermal peutic decisions is an unmeet medical need in this growth factor; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; field. HCV, hepatitis C virus; HGDN, high-grade dysplastic nodules; LGDN, low- grade dysplastic nodules; SNP, single nucleotide polymorphism. Most current article Keywords: Liver Cancer; Genomics; Signaling Pathways; © 2015 by the AGA Institute Molecular Therapies. 0016-5085/$36.00 http://dx.doi.org/10.1053/j.gastro.2015.05.061 October 2015 Genetic Landscape in HCC 1227

Genetic Predisposition, Environmental in HCC development in mice and rat models that can be prevented by tyrosine kinase inhibitors targeting EGF re- Factors, and Mechanism of Malignant ceptor, such as gefitinib.20 In addition, EGF is among the top- Transformation in Hepatocellular ranked genes of a signature highly correlated with HCC Carcinoma development in HCV cirrhosis.21 An early clinical chemo- prevention trial is currently evaluating whether erlotinib is Gene Environment Interaction in Hepatocellular able to revert this high-risk HCC signature into a low risk Carcinoma Predisposition (NCT02273362). To date, no drug, including sorafenib, has HCC occurrence results from a complex interplay among been able to decrease HCC incidence in high-risk patients, or genetic and nongenetic host factors, exposure to environ- prevent tumor recurrence after curative treatments.22 mental carcinogens and virus, and development of an un- Most of the polymorphisms associated with HCC devel- derlying chronic liver disease, which, at its ultimate stage opment on chronic liver disease are related to specificrisk (ie, cirrhosis), becomes a certain procarcinogenic field. factors. This observation highlights the close relationships Although cirrhosis is the “soil” where most of HCC grow, its among the nature of the exposure, the genetic background, GENETICS AND development on noncirrhotic liver helps us to reappraise and the mechanism of hepatocyte transformation/prolifera- GENETIC TESTING the risk factors and mechanisms that lead to HCC develop- tion.23 The association between aflatoxin B1 (AFB1), hepatitis ment without background liver damage (Figure 1). B virus (HBV), and SNP of GTSM1 and GSTT1 is the prototype Mendelian genetic predisposition to hepatocellu- of a strong interplay between specific exposure to genotoxic lar carcinoma. The liver is a very rare target of classical contaminant together with viral infection and common poly- cancer predisposition syndromes. The most frequent morphisms to dramatically increase HCC risk.24,25 A PNPLA3 monogenic syndromes that predispose to breast, ovarian, or polymorphism (coding for a lipase that mediates tri- colorectal cancers are usually not associated with develop- acylglycerol hydrolysis) is strongly associated with fatty and ment of liver tumors.1 Exceptionally, HCC can develop in 2 alcoholic chronic liver diseases; it is also associated with HCC patients with APC germ-line mutations. However, HCC occurrence.25–27 In contrast, contribution of PNPLA3 poly- predispositions are observed in several genetic metabolic morphism seems only minor in the risk of HCV-HCC.27–29 diseases, mainly via development of cirrhosis. These rare Recently, genome-wide association studies allowed the monogenic diseases include iron (hemochromatosis, HFE1 analysis of thousands of SNPs in HCC patients compared gene) or copper (Wilson disease, ATP7B gene) overload, with controls. A genome-wide association study performed tyrosinemia type 1 (FAH gene), porphyria acute intermittent in HCV patients in Japan identified a polymorphism in MICA a (HMBS gene) or cutanea tarda (UROD gene), and the 1 (gene involved in immune regulation),30 and another study fi 1 antitrypsin de ciency (SERPINA1 gene). In addition, ge- in the same population identified a different polymorphism netic alterations of glucose metabolism leading to glycogen in DEPDC5 (gene of unknown function),31 associated with storage diseases (particularly type Ia or von Gierke disease, HCC development. In Chinese patients infected by HBV, fi G6PC gene) or to speci c maturity onset diabetes of the SNPs were identified in STAT4 (a key protein of the in- young type 3 (MODY3, HNF1A gene) can promote genetic flammatory pathway),32 TPTE2 (a homolog of PTEN),33 liver adenomatosis occurrence and, in a second step, their 3–5 DCL1 (a tumor suppressor gene implicated in HCC patho- rare malignant transformation to HCC without cirrhosis. genesis),34 and in a region containing the UBEB4, KIF1B, Multifactorial genetic predisposition to hepato- PGD genes35 in 4 different studies as being associated with cellular carcinoma. Several single nucleotide poly- HCC occurrence. All these polymorphisms require validation morphisms (SNPs) were identified to be associated with in patients with various etiologies and at different stages of HCC risk.6 Among them, many polymorphisms alter biologic the liver diseases.6 pathways of carcinogenesis, including inflammation via TNFA, IL1B,orTGFB7–9; oxidative stress through SOD2 or MPO10,11; iron metabolism with a cooperation between alcohol intake and the HFE1 C282Y mutant variant12,13; Early Genetic Alterations in DNA repair (MTHFR or XRCC3)14,15; cell cycle with a major Precancerous Lesions Involved 16 17 18 role of MDM2 and TP53 ; or growth factor with EGF. in Malignant Transformation Polymorphisms modulate HCC risk at different steps of the disease, including predisposition to risk factors (eg, viral Cirrhosis as a Cancer Field hepatitis, alcohol intake, or obesity), to the severity of the There is a key event during malignant transformation in chronic liver disease and its evolution to cirrhosis, or to the cirrhosis that involves damaged cells, possibly hepatocytes, malignant transformation and tumor progression.6 They surrounded by fibrosis and vascularized mainly by the could be used to stratify patients in personalized sur- portal system switching to highly proliferative cells, vascu- veillance policies, and act as candidate targets for chemo- larized by arterial neovessels with an incremental invasive prevention strategies.19 As an example, rs4444903 and metastatic potential.36 This is a multistep process polymorphism in the 50 untranslated region that activates defined by a precise sequence of lesions: cirrhosis / low- EGF was associated with HCC on hepatitis C virus (HCV)- grade dysplastic nodules (LGDN) / high-grade dysplastic related cirrhosis and validated in independent studies.18 nodules (HGDN) / early HCC / progressed HCC and Epidermal growth factor (EGF) signaling is also involved advanced HCC.37 Several lines of evidence underlined the 1228 Zucman-Rossi et al Gastroenterology Vol. 149, No. 5 EEI TESTING GENETIC EEISAND GENETICS

Figure 1. Interaction between genetic predisposition, environmental factors, and HCC occurrence. Red, Mendelian inheri- tance pattern, blue,polymorphisms involved in HCC occurrence, green, HCC genetic imprint of genotoxic exposures, and purple, the potential therapies used to reduce HCC incidence.

key role of telomere and telomerase in cirrhosis pathogen- Conversely, telomerase-deﬁcient mouse with experimental esis and tumor initiation.38 Telomeres are repeat sequences liver injury develop cirrhosis, showing that telomere of DNA essential to protect chromosome ends where they shortening fosters hepatocytes senescence.42 In contrast, are located.39 They shorten during cell division leading to this model enlightened that telomerase reactivation is cell senescence and apoptosis.36 Telomerase is required for required in a second step to ensure HCC development.43 telomere synthesis and is composed of TERC, the RNA In humans, telomerase is reactivated in >90% of HCC45 template, and of TERT, the enzyme at rate-limiting compo- due to somatic TERT promoter mutations (54%60%),44,45 nent of the complex.40 In human livers, telomerase is not TERT ampliﬁcation (5%6%),45 or HBV insertion in the expressed in mature hepatocytes; cirrhotic tissue exhibited TERT promoter (10%15%)46,47 (Figure 2). Interestingly, telomere shortening with replicative senescence and premalignant lesions exhibit TERT promoter mutations in germline TERT mutations, predicted to decrease telomerase 6% of LGDN and 19% of HGDN, and frequency of TERT activity, are associated with a higher risk of cirrhosis.39,41 promoter mutations dramatically increases in early HCC October 2015 Genetic Landscape in HCC 1229 GENETICS AND GENETIC TESTING

Figure 2. Mechanisms of malignant transformation in HCC.

(61%) and remains stable in progressed and advanced suggests that early HCC, LGDN, and HGDN harboring TERT HCC.48 In addition, whole-exome sequencing of early HCC promoter mutations are at high risk for full malignant and of premalignant lesions revealed no additional recur- transformation/progression to advance HCC, which also rent mutations in classical HCC driver genes.49 These results requires additional hits in other cancer genes. show that TERT promoter mutation is the earliest recurrent somatic genetic alteration, behaving as a “gatekeeper” during the transformation sequence. It also suggests that early Malignant Transformation of Hepatocellular HCC are monoclonal, genetically closer to LGDN and HGDN Adenoma harboring TERT promoter mutations compared with pro- In normal liver, HCC can sometimes arise from the ma- gressed HCC that harbored mutations in many more cancer lignant transformation of hepatocellular adenoma, a rare genes. Concordantly, there is a phenotypic proximity be- monoclonal benign proliferation of hepatocytes usually tween these lesions, what frequently renders pathologic observed in young women taking oral contraception.51 Ade- discrimination between HGDN and early HCC a difﬁcult task, nomas are classiﬁed in 4 molecular subgroups according to even among expert liver pathologists.50 This paradigm mutations in either HNF1A, CTNNB1 coding for b-catenin, or 1230 Zucman-Rossi et al Gastroenterology Vol. 149, No. 5

in the genes activating the inflammatory pathway as IL6ST, maintenance. This could be useful to better understand the FRK, STAT3, JAK1, and GNAS.51–53 Hepatocellular adenoma biologic events that shape HCC development and help with exon 3 b-cateninactivating mutations have a higher identify new risk factors using a molecular-epidemiologic risk of transformation, and TERT promoter mutation are approach.61 involved at the last step of malignant transformation concomitantly with increased global hypomethylation, and Genomic Landscape and Driver Pathways chromosomal aberrations.54,55 However, the sequence of in Liver Carcinogenesis events differs in hepatocellular adenoma compared with Each HCC genome is the result of a unique combination cirrhosis: in normal hepatocytes, CTNNB1 activating of somatic genetic alterations with a mean numbers of fi mutation is rst and associated with monoclonal benign mutations in coding regions varying from 35 to 80 per EEI TESTING GENETIC – EEISAND GENETICS proliferation at risk of transformation, and in cirrhotic tumor.45,49,62 66 If most of the genetic alterations occurred 48,55 hepatocytes, TERT promoter mutation occurs earlier. in passenger genes without any predicted functional carci- Likely, cirrhotic hepatocytes might require early telomerase nogenic consequences, then a small numbers of mutations reactivation to proliferate and escape from senescence. occurred in cancer driver genes that belong to key signaling pathways involved in liver carcinogenesis.67 An integrated Genotoxic Effect as a Cause of Hepatocellular view of recent whole-exome sequencing studies allows identifying the major pathways recurrently mutated in HCC Carcinoma Development (Figure 3, and Table 1). HCC related to HBV can occur as a consequence of viral Telomere maintenance. As telomerase reactivation is proteins, such as HBx with oncogenic properties or due to the key during malignant transformation, 90% of human HCCs viral genome itself with HBV insertional mutagenesis in harbor an increased telomerase expression. The mecha- 23,46,56,57 cancer genes like TERT, CCNE1, and MLL4. Recently, nisms of telomerase reactivation are mutually exclusive and the adeno-associated virus type 2 (AAV2), a DNA defective include TERT promoter mutations (54%60%),44 TERT virus, was associated with HCC development on normal liver amplification (5%6%),45 and HBV insertion in TERT pro- due to insertional mutagenesis in TERT, CCNA2, CCNE1, moter (10%15%).46,47 TERT promoter mutations are 58 TNFSF10 and MLL4. Another layer of complexity emerged frequently associated with CTNNB1 mutations, suggesting from the analysis of the cancer genome that carries the cooperation between telomerase maintenance and b-catenin imprint of exposure to environmental carcinogens during a pathway in liver tumorigenesis.44,45,49 Other mechanisms of 59 patient’s lifetime. Exposure to AFB1, a mycotoxin contam- telomerase re-expression remain to be discovered and the inating food in certain areas of Asia and Africa, is a well- role of alternative pathways leading to telomere synthesis, known HCC carcinogen, a phenomena fostered by concomi- such as alternative lengthening of telomeres, need to be tant chronic HBV infection. AFB1-related HCC have a specific specifically evaluated. mutational signature due to AFB1-DNA adducts at guanine WNT/b-catenin pathway. The WNT/b-catenin path- leading to a high rate of C>A mutations and a specific hot spot way is pivotal in physiologic embryogenesis, zonation, and 23 of mutations R249S in TP53. In addition, null genotypes of metabolic control in the liver. It is the oncogenic pathway GTSM1 and GSTT1, 2 genes important for carcinogen detox- most frequently activated in HCC by activating mutations of ification that belong to glutathione S transferase family, are CTNNB1 (11%37%),68 and inactivating mutations of associated with an increased risk of HCC development in AXIN1 (5%15%),69 or APC (1%2%). CTNNB1 mutations, Asian patients in cooperation with AFB1 exposure and coding for b-catenin, are substitutions or in-frame deletions 24 chronic HBV infection. Analysis of the spectrum of nucleo- in a hotspot located in the domain targeted by the APC/ tide mutations in HCC enabled to identify signatures over the AXIN1/GSK3B inhibitory complex.68 Tumors mutated for 49 tumor genome specific of the AFB1 exposure. CTNNB1 have also a specific transcriptomic profile with Exposure to aristolochic acid, contained in a plant used overexpression of classical target genes like GLUL and in Chinese traditional medicine, is associated with urothelial LGR5,70 and a specific histologic pattern with intratumor cancer development that bears a highly specific high rate of cholestasis.71,72 mutations with T>A transversion.60 Interestingly, the same P53 cell cycle pathway. The P53 cell cycle pathway is mutational signature was identified in a subset of HCC in altered in at least half of HCC patients with frequent TP53 China, suggesting that aristolochic acid could be primarily mutations (12%48%),23,49,62,73,74 the tumor suppressor involved in HCC pathogenesis in a subset of patients.60 In gene more frequently mutated in cancer. Except for the addition, next-generation sequencing identified a group of R249S mutation related to AFB1 exposure, no other recur- HCCs developed on noncirrhotic liver in France with an rent TP53 mutation hotspot has been identified.62,75 The overrepresentation of T>C at ApTpX with transcription retinoblastoma pathway that control progression from G1 to strand bias, a pattern known to be strongly associated with S phase of the cell cycle is inactivated in HCC mainly by genotoxic injury.49 Interestingly, these tumors developed in homozygous deletion of CDKN2A (2%12%)45,62 or RB1 noncirrhotic patients with high alcohol and tobacco mutations (3%8%).65 Interestingly, genetic alterations of consumption.49 CDKN2A and RB1 were enriched in tumors with poor All of these results showed that the analysis of the prognosis suggesting a role of P21 pathway inactivation in cancer genome is an identity card bearing the archeological tumor aggressiveness.49,65 Finally, recurrent HBV insertions features of exposure to carcinogens or perturbance of DNA in CCNE1 (cyclin E1, 5%)46 and amplification of the CCND1/ October 2015 Genetic Landscape in HCC 1231 GENETICS AND GENETIC TESTING

Figure 3. The genetic landscape of HCC. Red, activating mutations of oncogenes, blue, inactivating mutations of tumor suppressors. *Gene was recurrently targeted by HBV integration.

FGF19 locus (5%14%),76,77 two key proteins involved in methylation writer family mainly in MLL (3%4%), MLL2 cell cycle progression, have been reported in HCC. (2%3%), MLL3 (3%6%), and MLL4 (2%3%) genes by Epigenetic modiﬁers. Epigenetic modiﬁers are mutation or HBV insertions in MLL4 (10%) are also recurrently altered in HCC,49 with inactivating mutations of frequent in HCC.46,49,64 In physiologic state, these genes ARID1A (4%17%)62 and ARID2 (3%18%)78 underlining code for proteins that modify histone methylation by adding the key role of SWI/SNF chromatin remodeling complexes and removing H3K4 methyl. Altogether, the functional (BAF and PBAF) as tumor suppressors. The physiologic role consequences of ARID1A, ARID2, and MLL genes mutations of these complexes is to modify chromatin structure and in hepatocarcinogenesis remain to be further explored. nucleosome position. Indirectly, they modify transcription Oxidative stress pathway. The oxidative stress fate of the cell. Recurrent somatic alterations in the histone pathway is altered by activating mutations of NRF2 (coded 1232 Zucman-Rossi et al Gastroenterology Vol. 149, No. 5

Table 1.Candidate Drivers as Targets for Therapies in Hepatocellular Carcinoma

Experimental evidence of “driver” properties

Somatic GEM impact In vivo tumor Candidate aberration liver tumor response after oncogene in human development / selective induction/ Drug (clinical trials addiction loop samples, % Altered pathway progression blockade in any tumor type)

DNA amplifications a EEI TESTING GENETIC FGF19 514 AKT/MTOR; RAS/MAPK Yes Yes BLU9931 EEISAND GENETICS CCND1 514a Cell cycle Palbociclibb VEGFA 711a AKT/MTOR; RAS/MAPK; NA Yes Bevacizumab, aflibercept angiogenesis TERT 56a Telomere maintenance Yes Yes GX301 Imtelstat GV1001 MYC 4 Cell cycle Yes Yes DCR-MYC (RNAi) FAK 4 RAS/MAPK Yes Yes GSK2256098 Gene mutations TERT promoter 5460 Telomere maintenance Yes Yes GX301 Imtelstat GV1001 CTNNB1 1137 WNT pathway Yes Yes PRI-724 TP53 1248 Cell cycle Yes Yes NA ARID1Ac 417 Epigenetic modifiers NA No GSK126 ARID2 318 Epigenetic modifiers NA No NA AXIN1 515 WNT pathway NA No NA TSC1/TSC2 38 AKT/MTOR Yes Yes Everolimus,c Temsirolimus,c Sirolimusc NFE2L2 36 Oxidative stress Yes Yes HSP90 inhibitors (17-AAG, 17-DMAG) RPS6KA3 29 RAS/MAPK No No Refametinib ATM 26 Cell cycle Yes No NA KEAP1 28 Oxidative stress Yes No HSP90 inhibitors (17-AAG, 17-DMAG) RB1 38 Cell cycle Yes Yes NA NRAS/HRAS/BRAF 2 RAS/MAPK Yes Yes Refametinib Overexpression MET 3050d AKT/MTOR; RAS/MAPK Yes Yes Cabozantinib, tivantinibe IGF2 10d AKT/MTOR; RAS/MAPK Yes Yes IMC-A12,f BI 836845 PD1/PD-L1 NA Immune checkpoint NA Yes Nivolumab

NA, not available. aHigh-level DNA focal gains. bCDK4/6 inhibitor. cTarget is MTOR (not TSC1/TSC2). dFrequency estimation not based on high-throughput RNA sequencing studies. eData suggest that tivantinib might not be a speciﬁc MET inhibitor.146 fTarget is IGF-IR.

by NFE2L2) or inactivating KEAP1 in 5%15% of the cases, RAS/RAF/mitogen-activated protein kinase pathways are preventing proteasome degradation of NRF2 physiologically activated in around 5%10% of HCC by amplification of the induced by KEAP1/CUL3 complex ubiquitinylation.62,79 FGF19/CCND1 locus.76,81 Also, activating mutations of NRF2 pathway activation was previously shown to protect PIK3CA (0%2%) and inactivating mutations of TSC1 or mice against chronic oxidative stress and tumor initiation.79 TSC2 (3%8%) lead to activation of the AKT/MTOR In contrast, recurrent mutations identified in HCC revealed signaling in a subset of HCC.45,49,65 In addition, homozygous NRF2 activation as a driver event in tumor progression.62,64 deletion of PTEN, an inhibitor of the PI3K kinase, has been In vitro study has demonstrated that NRF2 activation pre- identified in 1%3% of the HCC. However, some HCC with serves tumor cells from toxic exposure to reactive oxygen activation of the PI3K/AKT/MTOR cascade have no genetic species and subsequent death.80 alterations in this pathway.82 Indirect upstream activation PI3K/AKT/MTOR and RAS/RAF/mitogen-activated through insulin growth factor pathway has been proposed as protein kinase pathways. The PI3K/AKT/MTOR and an alternative mechanism of AKT/MTOR activation.83 October 2015 Genetic Landscape in HCC 1233

Activating mutations of genes belonging to the RAS family are class, HCC can be roughly divided into 2 major molecular rarely observed in HCC (<2%), and inactivating mutations of subtypes.94 One is broadly characterized by an enrichment RP6SKA3, coding for the RAS inhibitor RSK2, were identiﬁed of signals related to cell proliferation and progression in the in 2%9% of tumors.62 RSK2, is located downstream cell cycle (proliferation class) and is generally associated mitogen-activated protein kinase kinase and extracellular with a more aggressive phenotype, and the second class signal-regulated kinase, and is a known negative control loop generally retains molecular features resembling normal he- of RAS signaling.84 Inactivation of RSK2 released this negative patic physiology (nonproliferation class) (Figure 4). feedback and induced a constitutive activation of the pathway.49 Experimental data also suggest that persistent RAS activation may be a mechanism of HCC resistance to Proliferation Subclass sorafenib.85 The main traits of this class, which account for around 50% of patients, are related to activation of signaling cascades involved in cell proliferation/survival, enrichment of Molecular Classes signatures of poor prognosis, and association with clinical GENETICS AND

We have described the landscape of mutations and characteristics of aggressive tumors and poor outcomes. GENETIC TESTING critical pathways involved in the development and pro- Activation of signaling pathway is remarkably heteroge- gression of HCC. But 2 important questions emerge: Are we neous at the genomic and phenotypic levels. These include able to classify tumors according to the molecular events AKT/MTOR,82 MET,95 TGFB,96 IGF,83 RAS/mitogen- described here? And can we treat HCC patients based on activated protein kinase,97 among others. Interestingly, those biomarkers? Molecular classifications are aimed at this class is also enriched in genomic signals that identify providing a molecular understanding of the different biomarkers of progenitor cells, such as epithelial cell adhesion logic events that drive tumor subclasses and also at defining molecule92,98 or hepatoblastoma-like.91 The cell of origin in specific biomarkers/targets for therapies. Development of HCC is a highly controversial issue, with recent experi- whole-genome expression profiling in the late 1990s had a mental data supporting oncogenic reprogramming of tremendous impact in biomedical research.86 The premise different cellular lineages into cancer stem cells.99 A signa- was that, despite being at the same clinical stage, tumors are ture derived from hepatoblastomas, the most frequent pri- significantly different at a molecular level. Implications are mary pediatric liver tumor, as well as 2 signatures derived huge, because decision making in oncology relies mainly on from cholangiocarcinoma, are also enriched in this clinical staging, and rarely considers tumor molecular sin- class,100–102 as well as Notch,103 which acts as a master gularities. Array-based technologies enabled the analysis of regulator of biliary differentiation. Altogether, these data large numbers of transcripts in multiple samples, setting the further reinforce the notion of increased cellular plasticity in basis for cancer classification based on expression pat- these tumors. Different factors might explain increased terns.87 These molecular classes reflected different biologic heterogeneity of the proliferation class, including higher backgrounds with potential implications in patient selection rates of chromosomal instability or enrichment in aberrant for therapies and prediction of clinical outcomes. For epigenetic changes. For example, high-level DNA amplifica- example, gene signatures in breast cancer identified those tions of chromosome 11q13 (locus for FGF19, CCND1, women who would likely to benefit from adjuvant chemo- ORAOV1, FGF4) are enriched in this class, in addition to a therapy after resection.88 This could provide clinicians with DNA methylation-based prognostic signature that mainly additional tools to prevent patient’s exposure to unnec- correlates with signatures of progenitor cells.104 MicroRNA essary drug toxicity.89 There are 2 layers of information deregulation is also prominent in this class, with a clustered derived from molecular classification in HCC: first, its role as enrichment of a primate-specific family of miRNA located in a prognostic or predictive biomarker, and second, its ability chromosome 19 and microRNA/small nucleolar RNA on to increase our understanding of the molecular pathogen- chromosome 14.105 esis of the disease. From a clinical standpoint, patients in the proliferation In HCC, many groups have reported molecular classifi- class have aggressive tumors, with higher a-fetoprotein – cation based on genomic profiling (Table 2).62,90 92 Most of levels, moderately/poor cell differentiation on histology, these studies were conducted on resection specimens from and frequent vascular invasion.81,93 HBV-related HCC are patients with different etiologies for their underlying liver predominantly tumors belonging to this class. As predicted, disease. Remarkably, despite that marker genes defining patients with tumors from the proliferation subclass present each subclass were different across studies, integrative higher risk of recurrence after resection and lower survival analysis showed that most of them identified common rates.93,106 genomic signals reflecting background biology.93 In other words, different investigators reported molecular subclasses that identified similar patients, probably by spotting core Nonproliferation Class genomic alterations. Figure 4 summarizes our cumulative This molecular subclass contains at least 2 critical fea- understanding of gene expressionbased HCC subclasses tures: molecularly, it is dominated by activation of Wnt and its integration with additional levels of molecular in- signaling in up to 25% of cases, and others are characterized formation. These studies along with a meta-analysis showed by immune response; and from a functional perspective, the that, regardless of the specific nomenclature used for each transcriptome of tumors resemble normal hepatic 1234 Zucman-Rossi et al Gastroenterology Vol. 149, No. 5

Table 2.Prognostic Signatures in Hepatocellular Carcinoma

Predominant Clinical end point REMARK etiology (time-to-event) recommendations First author, year

Advanced developmental stage mRNA-based 5-gene signature Alcohol, Survival OK Nault, 201351 HCV, HBV EpCAM-signature HBV Survival OK Yamashita, 200892 186-gene signature (adjacent tissue) HCV Survival, HCC OK Hoshida, 2008118

EEI TESTING GENETIC development EEISAND GENETICS miRNA-based Down-regulation miR-26a HBV Survival OK Ji, 2009117 Pending further validation mRNA-based 5-gene risk signature HCV Survival OK Villa, 2015152 65-gene signature HBV Survival OK Kim, 2012153 233-gene signature (adjacent tissue) HBV Late recurrence, OK Kim, 2014120 recurrence-free survival G3 subclass HCV, alcohol Survival OK Boyault, 200670 Cluster A/hepatoblast-like HBV Survival OK Lee, 200691 17-gene metastasis predictor HBV Survival OK Budhu, 2006119 (adjacent tissue) Metastasis-related gene signature HBV Survival OK Roessler, 2010154 Cholangiocarcinoma-like HBV Survival OK Woo, 2010101 617-gene SP-ZEB signature HBV Survival — Marquardt, 2011155 miRNA-based 20-miRNA signature HBV Venous metastasis, OK Budhu, 2008116 survival Down-regulation miR-122 HBV, HCV, Survival OK Coulouarn, 2009156 alcohol 20-miRNA signature HBV Survival OK Wei, 2013157 C19MC HCV Survival — Augello, 2012158 Down-regulation Let-7 members HCV Early recurrence — Viswanathan, 2009159 Up-regulation miR-125a HBV Survival — Li, 2008160 19-miRNA signature HCV, alcohol Survival — Jiang, 2008161 Up-regulation miR-221 HCV Recurrence — Gramantieri, 2009162 DNA methylation 36 CpG DNA methylation signature HCV Survival OK Villanueva, 2015104 Methylation signature HBV, HCV Survival — Hernandez-Vargas, 2010163 Genome wide hypomethylation HBV Survival — Calvisi, 2006164 Hypermethylation of E-cadherin or HBV Survival — Lee, 2003165 GSTP1 Hypermethylation of E-cadherin HBV Recurrence — Kwon, 2005166

EpCAM, epithelial cell adhesion molecule; REMARK, Reporting Recommendations for Tumor Marker Prognostic Studies.

physiology. In terms of aberrant signaling, genomic data subclass of intrahepatic cholangiocarcinoma.102 From the suggest a dual modulation of WNT signaling in HCC,71 with clinical standpoint, tumors in this class show less aggressive differential representation in the proliferation and nonpro- phenotype, including better histologic differentiation, lower liferation subclasses. Classical WNT signaling, as defined by a-fetoprotein, and lack of enrichment in poor prognosis up-regulation of well-known target genes, such as GLUL or signatures. Regarding etiologic factors, HCV and alcohol- LGR5, is significantly enriched in this class.70 These data are related HCC are more prevalent in this class. further confirmed when analyzing CTNNB1 mutations and The backbone of HCC classification relies on gene nuclear translocation of b-catenin. A subset of tumors expression profiling. However, there have been recent at- within this class is characterized by broad gains in chro- tempts to enhance cancer classification with combined ge- mosome 7, associated with overexpression of EGF receptor netic and epigenetic features. Despite not yet being and male sex predominance.107 There are also data sug- comprehensively applied in HCC, a curated selection of gestive of immune signals within this class.81 Interestingly, a approximately 500 features, including copy number gains similar predominance in inflammation-related genomic and losses, mutations, and methylation-enabled accurate traits was also found in the less aggressive molecular cancer classification across 12 tumor types.108 Interestingly, October 2015 Genetic Landscape in HCC 1235 GENETICS AND GENETIC TESTING

Figure 4. Summary of molecular classiﬁcation of HCC. Major classes (proliferation and nonproliferation) are depicted based on messenger RNA expression proﬁling. Additional molecular features affecting DNA structure, pathway deregulation and epige- netics are overlapped.

at the top of this hierarchical classiﬁcation, 2 main tumor been described,111 although none has become a tangible tool subclasses with signiﬁcant differential prevalence of muta- in clinical decision making. Many factors could contribute to tions (ie, M) and DNA copy number alterations (ie, C), which this, including that molecular classes have been generated suggests that either mutations or chromosomal instability mostly from surgical specimens, limiting its theoretical can act as drivers of progression in different tumors. applicability to patients at earlier stages. There are recurrent discussions on the impact of tumor heterogeneity, both intra- and internodular, on molecular class predictions 112 Genetic Alterations as Biomarkers based on single biopsies. Despite this, molecular signatures are slowly permeating clinical practice guidelines.109 A Prognosis set of criteria has been proposed to determine formal in- HCC prognosis prediction is currently assessed using the clusion of molecular-based biomarkers in GPC.109 Also, the Barcelona Clinic Liver Cancer algorithm, as endorsed by the use of archived specimens collected in the context of high- American and European Associations for the Study of Liver end clinical trials has been suggested as a legitimate Diseases.109,110 Barcelona Clinic Liver Cancer relies on a source for biomarker development.113 composite of tumor burden, degree of liver damage, and Most prognostic signatures were generated using cancer-related symptoms. In terms of molecular-guided transcriptome data, such as the G3,70,93 5-gene,114 epithe- prognosis prediction, >40 prognostic gene signatures have lial cell adhesion molecule,92,115 cluster A,106 or 1236 Zucman-Rossi et al Gastroenterology Vol. 149, No. 5

hepatoblast-like.91 Others used epigenetic data to predict Studies in solid tumors have provided a broad picture of outcomes, like the 36-CpG DNA methylation signature,104 the mutational profile and identified a wide range of 5150 the 20-miRNA signature,116 or the identification and vali- mutations per tumor, depending on exposure to genotoxics dation of miR-26 in HBV-HCC (Table 2).105,117 Most of them (nonsmall cell lung cancer, melanoma-high mutation rate), identify patients within the proliferation subclass, however, tumors developed in infants or hematologic tumors (low- the 5-gene score is also able to predict prognosis of tumors mutation rate122). In HCC, the mean number of mutations in classified in the nonproliferative subclass (Figure 4). In coding regions per tumor is 3580, among which it has addition to those derived from the tumor, a number of been estimated that there might be 48 drivers of oncogene studies have reported signatures derived from adjacent addiction.49 The oncogene addiction theory postulates that cirrhotic tissue able to classify HCC patients based on their some tumors harbor certain somatic genetic alterations that

EEI TESTING GENETIC 123 EEISAND GENETICS prognosis. Presumably, these classes capture oncogenic dominate the malignant phenotype. Translation of this signals reflecting the so-called “field effect.” A186-gene principle in cancer management provides the rationale for signatureenrichedingenesetsassociatedwithinflamma- personalized care, commonly referred as precision tion, including interferon signaling, activation of EGF, nu- medicine. There are many examples of survival benefits clear factor kB, and tumor necrosis factora was able to achieved after selective inhibition of aberrant molecular identify HCC patients, mainly HCV-related, with poor sur- somatic events in solid tumors, such as vemurafenib in vival after resection, as well was those cirrhotic patients at melanoma with BRAF V600E mutations,124 or crizotinib in higher risk for HCC development.21,118 Differential inflam- lung cancer patients with anaplastic lymphoma kinase mation patterns identified with genomic profiling were also rearrangements.125 Effectiveness of this approach relies on determinant to predict intrahepatic metastasis in a cohort the following factors: dominance of the candidate oncogenic of HBV-HCC.119 More recently, a 233-gene signature iden- addiction loop in tumor progression, existence of a tified in regenerating livers accurately predicts risk of de biomarker able to accurately identify patients with this novo tumor formation (ie, late recurrence) in resected oncogenic loop, and a therapeutic agent able to selectively HCC.120 Patients with poor prognosis signatures from the and effectively abrogate the loop. Ultimately, a well- tumor (eg, G3, Cluster A) were not significantly enriched in designed, properly powered clinical trial enrolling patients those signatures, conferring poor prognosis from the with the activated oncogenic loop and treated with its adjacent tissue (eg, 186-gene signature), as judged by an antagonist will determine its clinical implementation. integrative analysis in a large cohort.93 In other words, and In HCC, the only effective systemic agent is the BRAF/ similar to prognosis prediction using clinical systems, vascular endothelial growth factor receptor/platelet- genomic data from both components (ie, tumor and adja- derived growth factor receptor inhibitor sorafenib.126 No cent tissue) are complementary to maximize prediction oncogenic addiction loop has been validated yet. In addition, accuracy (Table 2). none of the negative phase 3 trials in advanced HCC re- Recent refinements in these studies have enabled pre- ported so far used biomarker-based patient enrollment.127 dictions based on individual risk assessments. For example, It is feasible that not all tumor types can harbor oncogenic a nomogram that combines data from the 5-gene score in addiction loops. In addition, inhibition of a single loop might conjunction with Barcelona Clinic Liver Cancer stage and not suffice to provide enough survival benefits at advanced vascular invasion provides survival estimation on an indi- stages, and could promote growth of resistant clones or vidual basis.114 Similarly, the use of machine learning metastatic sites. Also, a comprehensive signaling blockade techniques for class identification and prediction of a DNA might be needed to achieve clinical responses, as has been methylation-based signature provides a continuous gradient shown in breast cancer,128 and suggested in resistance of risk, which ultimately might increase precision in models of HCC.85 outcome prediction.104 Signature-based prognosis pre- Among the somatic alterations described previously and dictions require tissue, which might constitute a limitation signaling pathway activation, we can define 2 areas where because HCC diagnosis can be confidently made with im- biomarkers can be used: targeting oncogenic addiction aging techniques. The development of technologies that loops based on the 3 requirements mentioned here, based in allow analysis of tumor byproducts in the blood, including human data and experimental models, and targeting cell-free DNA or circulating tumor cells, might facilitate the signaling cascades or immune checkpoints (Tables 1 and 2). implementation of molecular-based biomarkers.121 Oncogenic loops. DNA copy number changes. VEGFA (gains in chromosome 6p21) is recurrently described, correlated with expression, and with an estimated Therapeutic prevalence based on fluorescence in situ hybridization of In oncology, survival benefits from molecular-targeted 7%11%.81,129 Also, high VEGF plasma levels are correlated therapies might be derived as a result of treating all pa- with poor survival130; there is solid evidence of anti-tumor tients with very effective/broad therapies with manageable effect after VEGF inhibition in experimental models. VEGFA toxicity, targeting a specific, generally small, subgroup of could have a double pro-tumorigenic effect by promoting patients whose tumors are addicted to an oncogenic aber- angiogenesis and inducing noncell autonomous over- ration; and targeting a broad number of patients whose expression of hepatocyte growth factor.129 Early trials with tumors present a frequent mechanism of activation VEGFA inhibitor bevacizumab in unselected populations (signaling pathway) or immune response. show modest efficacy and raised some safety concerns.131 October 2015 Genetic Landscape in HCC 1237

FGF19, CCND1 (gains in chromosome 11q13) are frequency has been described up to 9% based on whole- described in 5%14% of HCC,45,76,81 and are also corre- genome sequencing data in HBV-HCC.66 This has not been lated with high expression levels and poor prognosis.49 confirmed in subsequent studies.45,49,64,65 Experimental FGF19 blockade with monoclonal antibodies inhibits clonal evidence suggests that its blockage could have antitumor growth and tumorigenicity in vivo.76,132 Strikingly, trans- activity.66 genic mouse overexpressing FGF19, which binds to FGFR4 133 in skeletal muscle, develop HCC, which points toward a Signaling Pathways or Immune Checkpoints dual role of FGF19 in both tumor progression and initiation. MET. Overexpression at the messenger RNA and pro- In terms of FGF blockade, brivanib, a highly active FGFR1-3 tein levels has been described in at least 40%50% of HCC inhibitor, failed to improve survival in unselected advanced 144 134,135 samples. A subanalysis of the phase 2b trial testing HCC populations. CCND1 shares genomic locus with tivantinib vs placebo in second line suggested that the drug fi FGF19, and it is also frequently ampli ed in cancer. CDK4/6 was particularly effective in those patients with high dual inhibitors, such as palbociclib or abemaciclib, have expression of MET on immunohistochemistry.145 There has shown anti-tumoral activity in experimental models, ’ fi

been some concerns about tivantinib s speci city as an MET GENETICS AND

fi GENETIC TESTING although data from breast does not con rm a particular inhibitor,146 but a phase 3 RCT enrolling patients with high fi fi 136 bene t of CDK4/6 inhibition in CCND1 ampli ed tumors. MET is currently evaluating its role in second line FAK, MYC (gains in chromosome 8q24) have broad gains (NCT02029157). described in 4% of HCC, with a single study reporting IGF2. Enrichment of high transcript levels of IGF2 65,66 26%. FAK inhibition induces tumor responses in (>20-fold) in approximately 10% of HCC samples suggest 137 experimental models of HCC, and selective inhibitors of that it could act as a tumor driver.83,147 DNA methylation is FAK are currently under clinical development. MYC trans- a reported mechanism for IGF2 deregulation in HCC,104 and genic mice develop liver cancer and it has a central role it has been described as an oncogenic partner of Notch- 138 during hepatocarcinogenesis. However, the development induced HCC.103 Functional validation indicate that IGF of effective pharmacologic strategies to inhibit MYC has pathway inhibition could be an effective strategy in HCC, but been challenging. results in early clinical trails using IGF-IR antagonists have Gene mutations. CTNNB1 is highly prevalent across shown limited efficacy and significant toxicity.148 45,49,64 fi studies (approximately 25%). Similar to MYC, dif - PD1/PD-L1. Both involved in the inhibitory signals to culties developing selective and nontoxic WNT inhibitors T cells, leading to suppression of anti-tumoral immune 139 have limited its clinical development, but clearly, response. Data in other solid tumors show remarkable re- CTNNB1 mutation is an appealing candidate for selective sponses after selective inhibition with nivolumab,149 with a targeting. potential role of PD1/PD-L1 as a biomarkers of response.150 TSC1/TSC2 are both negative modulators of MTOR In HCC, intratumoral increased expression of PD1/PD-L1 140 cascade, and their inactivation promotes MTOR signaling. correlates with poor outcomes, and its inhibition in ortho- Mutations, as well as DNA copy number alterations are topic HCC models induces tumor responses.151 This repre- 45,49 described in 2% 14% of HCC. Pathway-based analysis sents a novel immune-base approach. Studies with underscores the potential driving role of this cascade in checkpoint inhibitors are currently ongoing in phase 2 in 82 fi 45 HCC, with a signi cant enrichment in Asian populations. HCC. In terms of MTOR inhibition, a phase 1 trial testing ever- In conclusion, the studies reported during the last 10 olimus in combination with sorafenib show subtherapeutic years have been able to delineate the landscape of muta- concentrations for everolimus and precluded further tions and genomic alterations occurring in HCC develop- 141 development of this combination. When tested in second ment and progression. This has certainly changed our fi line, everolimus as a single agent failed to signi cantly understanding of the disease. Nonetheless, we are at a point improve survival in a phase 3 trial without biomarker- that there is a clear unmet need to translate this knowledge 142 selected patients. Subgroup analyses suggested an in actual advantages for patients, either in the arena of increased survival in HBV patients, maybe in relation to a refining at-risk populations, patient prognosis, and re- higher AKT/MTOR pathway activation in this population. sponses to therapies. A huge translational effort needs < 49,65,66 NRAS/KRAS/HRAS are mutated in 3% of HCC. attention. The research community in HCC has to link dis- Preliminary data from a phase 2 trial (ie, BASIL [(Assess- coveries with actual survival benefits and, therefore, it is ing BAY86-9766 Plus Sorafenib for the Treatment of Liver now about time to demonstrate that biomarkers defining Cancer) NCT01204177]) in patients with advanced HCC subclasses of patients ultimately might lead to better ther- receiving a combination of MEK inhibitor refametinib apies that change the decision-making process in this com- (BAY86-9766) plus sorafenib found mutations in approxi- plex and heterogeneous disease. mately 6% of patients (4 of 69)143; analysis was conducted on circulating DNA. Three of these four patients had confirmed partial responses. Inactivating mutations in RAS Supplementary Material inhibitor RP6SKA3 are more prevalent (2%9%)49,62,65 and Note: The first 50 references associated with this article are could also predict response to RAS inhibition. A phase available below in print. The remaining references accom- 2 study testing the combination of refametinib and sor- panying this article are available online only with the elec- afenib in RAS mutant HCCs is ongoing. JAK1 mutation tronic version of this article. To access the supplementary 1238 Zucman-Rossi et al Gastroenterology Vol. 149, No. 5

material accompanying this article, visit the online version and thymidylate synthase genes and risk of hepatocel- of Gastroenterology at www.gastrojournal.org, and at http:// lular carcinoma. Hepatology 2007;46:749–758. dx.doi.org/10.1053/j.gastro.2015.05.061. 16. Dharel N, Kato N, Muroyama R, et al. MDM2 promoter SNP309 is associated with the risk of hepatocellular carcinoma in patients with chronic hepatitis C. Clin References Cancer Res 2006;12:4867–4871. 1. Tommaso AD. Risk of HCC: genetic heterogeneity and 17. Yoon YJ, Chang HY, Ahn SH, et al. MDM2 and p53 complex genetics. J Hepatol 2010;52:252–257. polymorphisms are associated with the development of 2. Li M, Gerber DA, Koruda M, et al. Hepatocelluar carci- hepatocellular carcinoma in patients with chronic hepa- noma associated with attenuated familial adenomatous titis B virus infection. Carcinogenesis 2008;29:

EEI TESTING GENETIC polyposis: a case report and review of the literature. Clin 1192–1196. EEISAND GENETICS Colorectal Cancer 2012;11:77–81. 18. Dayyeh BKA, Yang M, Fuchs BC, et al. A functional 3. Willson JSB, Godwin TD, Wiggins GAR, et al. Primary polymorphism in the epidermal growth factor gene is hepatocellular neoplasms in a MODY3 family with a novel associated with risk for hepatocellular carcinoma. HNF1A germline mutation. J Hepatol 2013;59:904–907. Gastroenterology 2011;141:141–149. 4. Bacq Y, Jacquemin E, Balabaud C, et al. Familia liver 19. Pinyol R, Nault J, Quetglas I, et al. Molecular profiling of adenomatosis associated with hepatocyte nuclear factor liver tumors: classification and clinical translation for 1a inactivation. Gastroenterology 2003;125:1470–1475. decision making. Semin Liver Dis 2014;34:363–375. 5. Calderaro J, Labrune P, Morcrette G, et al. Molecular 20. Schiffer E, Housset C, Cacheux W, et al. Gefitinib, an characterization of hepatocellular adenomas developed EGFR inhibitor, prevents hepatocellular carcinoma in patients with glycogen storage disease type I. development in the rat liver with cirrhosis. Hepatology J Hepatol 2013;58:350–357. 2005;41:307–314. 6. Nahon P, Zucman-Rossi J. Single nucleotide poly- 21. Hoshida Y, Villanueva A, Sangiovanni A, et al. Prognostic morphisms and risk of hepatocellular carcinoma in gene expression signature for patients with hepatitis C- cirrhosis. J Hepatol 2012;57:663–674. related early-stage cirrhosis. Gastroenterology 2013; 7. Tarhuni A, Rufat P, Sutton A, et al. Impact of cytokine 144:1024–1030. gene variants on the prediction and prognosis of hepa- 22. Bruix J, Takayama T, Mazzaferro V, et al. STORM: a tocellular carcinoma in patients with cirrhosis. J Hepatol phase III randomized, double-blind, placebo-controlled 2014;61:342–350. trial of adjuvant sorafenib after resection or ablation to 8. Wang Y, Kato N, Hoshida Y, et al. Interleukin-1beta gene prevent recurrence of hepatocellular carcinoma. J Clin polymorphisms associated with hepatocellular carci- Oncol 2014;32:5s; 2014 (Suppl; Abstr 4006). noma in hepatitis C virus infection. Hepatology 2003; 23. Bressac B, Kew M, Wands J, et al. Selective G to T 37:65–71. mutations of p53 gene in hepatocellular carcinoma from 9. Migita K, Miyazoe S, Maeda Y, et al. Cytokine gene southern Africa. Nature 1991;350:429–431. polymorphisms in Japanese patients with hepatitis B 24. Wang B, Huang G, Wang D, et al. Null genotypes of virus infection–association between TGF-beta1 poly- GSTM1 and GSTT1 contribute to hepatocellular carci- morphisms and hepatocellular carcinoma. J Hepatol noma risk: evidence from an updated meta-analysis. 2005;42:505–510. J Hepatol 2010;53:508–518. 10. Nahon P, Sutton A, Rufat P, et al. Myeloperoxidase and 25. Kozlitina J, Xing C, Pertsemlidis A, et al. Genetic vari- superoxide dismutase 2 polymorphisms comodulate the ation in PNPLA3 confers susceptibility to nonalcoholic risk of hepatocellular carcinoma and death in alcoholic fatty liver disease. Nat Genet 2008;40:1461–1465. cirrhosis. Hepatology 2009;50:1484–1493. 26. Liu YL, Patman GL, Leathart JBS, et al. Carriage of the 11. Nahon P, Sutton A, Rufat P, et al. A variant in myelo- PNPLA3 rs738409 C >G polymorphism confers an peroxidase promoter hastens the emergence of hepa- increased risk of non-alcoholic fatty liver disease asso- tocellular carcinoma in patients with HCV-related ciated hepatocellular carcinoma. J Hepatol 2014; cirrhosis. J Hepatol 2012;56:426–432. 61:75–81. 12. Nahon P, Sutton A, Rufat P, et al. Liver iron, HFE gene 27. Nahon P, Valenti L, Henrion J, et al. Association between mutations, and hepatocellular carcinoma occurrence in the PNPLA3 (rs738409 C >G) variant and hepatocellular patients with cirrhosis. Gastroenterology 2008; carcinoma: Evidence from a meta-analysis of individual 134:102–110. participant data. Hepatology 2014;59:2170–2177. 13. Lauret E, Rodríguez M, González S, et al. HFE gene 28. Singal AG, Manjunath H, Yopp AC, et al. The effect of mutations in alcoholic and virus-related cirrhotic patients PNPLA3 on fibrosis progression and development of with hepatocellular carcinoma. Am J Gastroenterol 2002; hepatocellular carcinoma: a meta-analysis. Am J Gas- 97:1016–1021. troenterol 2014;109:325–334. 14. Long XD, Ma Y, Qu DY, et al. The polymorphism of 29. Guyot E, Sutton A, Rufat P, et al. PNPLA3 rs738409, XRCC3 codon 241 and AFB1-related hepatocellular hepatocellular carcinoma occurrence and risk model carcinoma in Guangxi population, China. Ann Epidemiol prediction in patients with cirrhosis. J Hepatol 2013; 2008;18:572–578. 58:312–318. 15. Yuan J-M, Lu SC, Van Den Berg D, et al. Genetic poly- 30. Kumar V, Kato N, Urabe Y, et al. Genome-wide asso- morphisms in the methylenetetrahydrofolate reductase ciation study identifies a susceptibility locus for October 2015 Genetic Landscape in HCC 1239

HCV-induced hepatocellular carcinoma. Nat Genet 2011; mutations in hepatocellularcarcinoma and preneoplastic 43:455–458. lesions. Nat Commun 2013;4:1–6. 31. Miki D, Ochi H, Hayes CN, et al. Variation in the DEPDC5 45. Totoki Y, Tatsuno K, Covington KR, et al. Trans- locus is associated with progression to hepatocellular ancestry mutational landscape of hepatocellular carci- carcinoma in chronic hepatitis C virus carriers. Nat Genet noma genomes. Nat Genet 2014;46:1267–1273. 2011;43:797–800. 46. Sung W-K, Zheng H, Li S, et al. Genome-wide survey of 32. Jiang D-K, Sun J, Cao G, et al. Genetic variants in recurrent HBV integration in hepatocellular carcinoma. STAT4 and HLA-DQ genes confer risk of hepatitis B Nat Genet 2012;44:765–769. virus-related hepatocellular carcinoma. Nat Genet 2013; 47. Paterlini-Bréchot P, Saigo K, Murakami Y, et al. Hepatitis 45:72–75. B virus-related insertional mutagenesis occurs frequently 33. Clifford RJ, Zhang J, Meerzaman DM, et al. Genetic in human liver cancers and recurrently targets human variations at loci involved in the immune response are telomerase gene. Oncogene 2003;22:3911–3916. risk factors for hepatocellular carcinoma. Hepatology 48. Nault JC, Calderaro J, Di Tommaso L, et al. Telomerase 2010;52:2034–2043. reverse transcriptase promoter mutation is an early so-

Chan KYK, Wong C-M, Kwan JS-H GENETICS AND

34. , et al. Genome- matic genetic alteration in the transformation of prema- GENETIC TESTING wide association study of hepatocellular carcinoma in lignant nodules in hepatocellular carcinoma on cirrhosis. Southern Chinese patients with chronic hepatitis B virus Hepatology 2014;60:1983–1992. infection. PLoS ONE 2011;6:e28798. 49. Schulze K, Imbeaud S, Letouze E, et al. Exome 35. Zhang H, Zhai Y, Hu Z, et al. Genome-wide association sequencing of hepatocellular carcinoma identifies new study identifies 1p36.22 as a new susceptibility locus for mutations signatures and potential therapeutic targets. hepatocellular carcinoma in chronic hepatitis B virus Nat Genet 2015;47:505–511. carriers. Nat Genet 2010;42:767–770. 50. Di Tommaso L, Sangiovanni A, Borzio M, et al. Advanced 36. Farazi PA, Depinho RA. Hepatocellular carcinoma path- precancerous lesions in the liver. Best Pract Res Clin ogenesis: from genes to environment. Nat Rev Cancer Gastroenterol 2013;27:269–284. 2006;6:674–687. 37. International Consensus Group for Hepatocellular Neoplasia. Pathologic diagnosis of early hepatocellular Author names in bold designate shared co-first authorship. carcinoma: a report of the international Consensus Received March 19, 2015. Accepted May 20, 2015. Group for Hepatocellular Neoplasia. Hepatology 2009; 49:658–664. Reprint requests Address requests for reprints to: Josep M. Llovet, MD, PhD, Liver Cancer 38. Satyanarayana A, Manns MP, Rudolph KL. Telomeres Translational Research Laboratory, Barcelona Clinic Liver Cancer Group, and telomerase: a dual role in hepatocarcinogenesis. Liver Unit, CIBERehd, Hospital Clínic, IDIBAPS, University of Barcelona, – Villarroel 170, 08036 Barcelona, Catalonia, Spain. e-mail: [email protected]; Hepatology 2004;40:276 283. fax: þ34 93 3129406; or Jessica Zucman-Rossi, MD, PhD, INSERM U 1162, 39. Calado RT, Young NS. Telomere diseases. N Engl J Med Génomique Fonctionnelle des Tumeurs Solides, Paris, France. e-mail: þ 2009;361:2353–2365. [email protected]; fax: 33 01 53725192. 40. Günes C, Rudolph KL. The role of telomeres in stem cells Acknowledgments and cancer. Cell 2013;152:390–393. Author contributions: JZ-R, AV, J-CN, and JML participated in all stages of manuscript production, design, figures, tables, writing, and review of final 41. Hartmann D, Srivastava U, Thaler M, et al. Telomerase version. gene mutations are associated with cirrhosis formation. – Conflicts of interest Hepatology 2011;53:1608 1617. These authors disclose the following: Josep M. Llovet is consultant for Bayer, 42. Rudolph KL, Chang S, Millard M, et al. Inhibition of BMS, Lilly, Blueprint, Celsion, GSK, and Boehringer-Ingelheim, and has experimental liver cirrhosis in mice by telomerase gene received research grants from Bayer, BMS, Blueprint, and Boehringer- Ingelheim. Jessica Zucman-Rossi is consultant for IntraGen. The remaining delivery. Science 2000;287:1253–1258. authors disclose no conflicts. 43. Farazi PA, Glickman J, Jiang S, et al. Differential impact of Funding telomere dysfunction on initiation and progression of he- Jessica Zucman-Rossi has grant from the Ligue Contre le Cancer and the INCa patocellular carcinoma. Cancer Res 2003;63:5021–5027. with the ICGC project. Josep M. Llovet received grants from the European Commission Framework Programme 7 (Heptromic, proposal number 44. Nault JC, Mallet M, Pilati C, et al. High frequency 259744), Samuel Waxman Cancer Research Foundation, Grant IþD Program of telomerase reverse-transcriptasepromoter somatic (SAF2013-41027), and the Asociación Española Contra el Cáncer (AECC). 1239.e1 Zucman-Rossi et al Gastroenterology Vol. 149, No. 5

Supplementary References 67. Stratton MR, Campbell PJ, Futreal PA. The cancer genome. Nature 2009;458:719–724. 51. Nault JC, Bioulac-Sage P, Zucman-Rossi J. Hepatocel- 68. La Coste de A, Romagnolo B, Billuart P, et al. Somatic lular benign tumors-from molecular classification to mutations of the beta-catenin gene are frequent in personalized clinical care. Gastroenterology 2013; mouse and human hepatocellular carcinomas. Proc Natl 144:888–902. Acad Sci U S A 1998;95:8847–8851. 52. Bluteau O, Jeannot E, Bioulac-Sage P, et al. Bi-allelic 69. Satoh S, Daigo Y, Furukawa Y, et al. AXIN1 mutations in inactivation of TCF1 in hepatic adenomas. Nat Genet hepatocellular carcinomas, and growth suppression in 2002;32:312–315. cancer cells by virus-mediated transfer of AXIN1. Nat 53. Rebouissou S, Amessou M, Couchy G, et al. Frequent Genet 2000;24:245–250. in-frame somatic deletions activate gp130 in inflam- 70. Boyault S, Rickman DS, De Reyniès A, et al. Tran- matory hepatocellular tumours. Nature 2009;457: scriptome classification of HCC is related to gene alter- 200–204. ations and to new therapeutic targets. Hepatology 2006; 54. Zucman-Rossi J, Jeannot E, Nhieu JTV, et al. Genotype- 45:42–52. phenotype correlation in hepatocellular adenoma: new 71. Lachenmayer A, Alsinet C, Savic R, et al. Wnt-pathway classification and relationship with HCC. Hepatology activation in two molecular classes of hepatocellular 2006;43:515–524. carcinoma and experimental modulation by sorafenib. fi 55. Pilati C, Nault JC, Imbeaud S, et al. Genomic pro ling of Clin Cancer Res 2012;18:4997–5007. hepatocellular adenomas reveals recurrent FRK- 72. Audard V, Grimber G, Elie C, et al. Cholestasis is a activating mutations and the mechanisms of malignant marker for hepatocellular carcinomas displaying beta- transformation. Cancer Cell 2014;25:428–441. catenin mutations. J Pathol 2007;212:345–352. Avail- 56. Neuveut C, Wei Y, Buendia MA. Mechanisms of HBV- able at: http://doi.wiley.com/10.1002/path.2169. related hepatocarcinogenesis. J Hepatol 2010; 73. Fujimoto A, Furuta M, Shiraishi Y, et al. Whole-genome 52:594–604. mutational landscape of liver cancers displaying biliary 57. Brechot C, Pourcel C, Louise A, et al. Presence of inte- phenotype reveals hepatitis impact and molecular di- grated hepatitis B virus DNA sequences in cellular DNA versity. Nat Commun 2015;6:1–8. of human hepatocellular carcinoma. Nature 1980; 74. Hsu IC, Metcalf RA, Sun T, et al. Mutational hotspot in 286:533–535. the p53 gene in human hepatocellular carcinomas. Na- 58. Nault JC, Datta S, Imbeaud S, et al. Recurrent AAV2- ture 1991;350:427–428. related insertional mutagenesis in human hepatocellular 75. Fujimoto A, Totoki Y, Abe T, et al. Whole-genome carcinomas. Nat Genet 2015 Aug 24. http://dx.doi.org/ sequencing of liver cancers identi fies etiological in- 10.1038/ng.3389. [Epub ahead of print]. fluences on mutation patterns and recurrent mutations in 59. Helleday T, Eshtad S, Nik-Zainal S. Mechanisms un- chromatin regulators. Nat Genet 2012;44:760–764. derlying mutational signatures in human cancers. Nat 76. Sawey ET, Chanrion M, Cai C, et al. Identification of a Rev Genet 2014;15:585–598. therapeutic strategy targeting amplified FGF19 in liver 60. Poon SL, Pang S-T, McPherson JR, et al. Genome- cancer by oncogenomic screening. Cancer Cell 2011; wide mutational signatures of aristolochic acid and its 19:347–358. application as a screening tool. Sci Transl Med 2013; 77. Wang K, Lim HY, Shi S, et al. Genomic landscape of 5:197–201. copy number aberrations enables the identification of 61. Alexandrov LB, Nik-Zainal S, Wedge DC, et al. Signa- oncogenic drivers in hepatocellular carcinoma. Hep- tures of mutational processes in human cancer. Nature atology 2013;58:706–717. 2013;500:415–421. 78. Li M, Zhao H, Zhang X, et al. Inactivating mutations of 62. Guichard C, Amaddeo G, Imbeaud S,etal.Inte- the chromatin remodeling gene ARID2 in hepatocellular grated analysis of somatic mutations and focal copy- carcinoma. Nat Genet 2011;43:828–829. number changes identifies key genes and pathways 79. Sporn MB, Liby KT. NRF2 and cancer: the good, the in hepatocellular carcinoma. Nat Genet 2012; bad and the importance of context. Nature 2012; 44:694–698. 12:564–571. 63. Huang J, Deng Q, Wang Q, et al. Exome sequencing of 80. DeNicola GM, Karreth FA, Humpton TJ, et al. Oncogene- hepatitis B virus-associated hepatocellular carcinoma. induced Nrf2 transcription promotes ROS detoxification Nat Genet 2012;44:1117–1121. and tumorigenesis. Nature 2011;475:106–109. 64. Cleary SP, Jeck WR, Zhao X, et al. Identification of driver 81. Chiang DY, Villanueva A, Hoshida Y, et al. Focal gains of genes in hepatocellular carcinoma by exome VEGFA and molecular classification of hepatocellular sequencing. Hepatology 2013;58:1693–1702. carcinoma. Cancer Res 2008;68:6779–6788. 65. Ahn S-M, Jang SJ, Shim JH, et al. Genomic portrait of 82. Villanueva A, Chiang DY, Newell P, et al. Pivotal role of resectable hepatocellular carcinomas: implications of mTOR signaling in hepatocellular carcinoma. Gastroen- RB1 and FGF19 aberrations for patient stratification. terology 2008;135:1972–1983. Hepatology 2014;60:1972–1982. 83. Tovar V, Alsinet C, Villanueva A, et al. IGF activation in a 66. Kan Z, Zheng H, Liu X, et al. Whole-genome molecular subclass of hepatocellular carcinoma and pre- sequencing identifies recurrent mutations in hepatocel- clinical efficacy of IGF-1R blockage. J Hepatol 2010; lular carcinoma. Genome Res 2013;23:1422–1433. 52:550–559. October 2015 Genetic Landscape in HCC 1239.e2

84. Douville E, Downward J. EGF induced SOS phosphory- 101.Woo HG, Lee JH, Yoon JH, et al. Identification of a lation in PC12 cells involves P90 RSK-2. Oncogene cholangiocarcinoma-like gene expression trait in hepa- 1997;15:373–383. tocellular carcinoma. Cancer Res 2010;70:3034–3041. 85. Rudalska R, Dauch D, Longerich T, et al. RNAi 102.Sia D, Hoshida Y, Villanueva A, et al. Integrative molec- screening identifies a mechanism of sorafenib resistance ular analysis of intrahepatic cholangiocarcinoma reveals in liver cancer. Nat Med 2014:1–11. 2 classes that have different outcomes. Gastroenter- 86. Golub TR. Molecular classification of cancer: class dis- ology 2013;144:829–840. covery and class prediction by gene expression moni- 103.Villanueva A, Alsinet C, Yanger K, et al. Notch signaling toring. Science 1999;286:531–537. is activated in human hepatocellular carcinoma and in- 87. DeRisi J, Penland L, Brown PO, et al. Use of a cDNA duces tumor formation in mice. Gastroenterology 2012; microarray to analyse gene expression patterns in hu- 143:1660–1669. man cancer. Nat Genet 1996;14:457–460. 104.Villanueva A, Portela A, Sayols S, et al. DNA 88. van de Vijver MJ, He YD, Van’t Veer LJ, et al. A gene- Methylation-based prognosis and epidrivers in hepato- expression signature as a predictor of survival in breast cellular carcinoma. Hepatology 2015;61:1945–1956. cancer. N Engl J Med 2002;347:1999–2009. 105.Toffanin S, Hoshida Y, Lachenmayer A, et al. MicroRNA- 89. Solin LJ, Gray R, Baehner FL, et al. A multigene based classification of hepatocellular carcinoma and expression assay to predict local recurrence risk for oncogenic role of miR-517a. Gastroenterology 2011; ductal carcinoma in situ of the breast. J Natl Cancer Inst 140:1618–1628. 2013;105:701–710. 106.Lee J-S, Chu I-S, Heo J, et al. Classification and pre- 90. Hoshida Y, Toffanin S, Lachenmayer A, et al. Molecular diction of survival in hepatocellular carcinoma by gene classification and novel targets in hepatocellular carci- expression profiling. Hepatology 2004;40:667–676. noma: recent advancements. Semin Liver Dis 2010; 107.Keng VW, Sia D, Sarver AL, et al. Sex bias occurrence of 30:35–51. hepatocellular carcinoma in Poly7 molecular subclass is 91. Lee J-S, Heo J, Libbrecht L, et al. A novel prognostic associated with EGFR. Hepatology 2012;57:120–130. subtype of human hepatocellular carcinoma derived from 108.Ciriello G, Miller ML, Aksoy BA, et al. Emerging land- hepatic progenitor cells. Nat Med 2006;12:410–416. scape of oncogenic signatures across human cancers. 92. Yamashita T, Forgues M, Wang W, et al. EpCAM and Nat Genet 2013;45:1127–1133. alpha-fetoprotein expression defines novel prognostic 109.European Association for the Study of the Liver, European subtypes of hepatocellular carcinoma. Cancer Res 2008; Organisation for Research and Treatment of Cancer. 68:1451–1461. EASL-EORTC Clinical Practice Guidelines: management of 93. Villanueva A, Hoshida Y, Battiston C, et al. Combining hepatocellular carcinoma. J Hepatol 2012;56:908–943. clinical, pathology, and gene expression data to predict 110.Bruix J, Sherman M; American Association for the Study recurrence of hepatocellular carcinoma. Gastroenter- of Liver Diseases. Management of hepatocellular carci- ology 2011;140:1501–1512. noma: an update. Hepatology 2011;53:1020–1022. 94. Hoshida Y, Nijman SMB, Kobayashi M, et al. Integrative 111.Hoshida Y, Moeini A, Alsinet C, et al. Gene signatures in transcriptome analysis reveals common molecular sub- the management of hepatocellular carcinoma. Semin classes of human hepatocellular carcinoma. Cancer Res Oncol 2012;39:473–485. 2009;69:7385–7392. 112.Friemel J, Rechsteiner MP, Frick L, et al. Intratumor 95. Kaposi-Novak P. Met-regulated expression signature heterogeneity in hepatocellular carcinoma. Clin Cancer defines a subset of human hepatocellular carcinomas Res 2014;21:1951–1961. with poor prognosis and aggressive phenotype. J Clin 113.Simon RM, Paik S, Hayes DF. Use of archived speci- Invest 2006;116:1582–1595. mens in evaluation of prognostic and predictive bio- 96. Coulouarn C, Factor VM, Thorgeirsson SS. Transforming markers. J Natl Cancer Inst 2009;101:1446–1452. growth factor-b gene expression signature in mouse 114.Nault JC, De Reyniès A, Villanueva A, et al. hepatocytes predicts clinical outcome in human cancer. A hepatocellular carcinoma 5-gene score associated Hepatology 2008;47:2059–2067. with survival of patients after liver resection. Gastroen- 97. Newell P, Toffanin S, Villanueva A, et al. Ras pathway terology 2013;145:176–187. activation in hepatocellular carcinoma and anti-tumoral 115.Yamashita T, Ji J, Budhu A, et al. EpCAM-positive he- effect of combined sorafenib and rapamycin in vivo. patocellular carcinoma cells are tumor-initiating cells J Hepatol 2009;51:725–733. with stem/progenitor cell features. Gastroenterology 98. Yamashita T, Wang XW. Cancer stem cells in the 2009;136:1012–1124. development of liver cancer. J Clin Invest 2013; 116.Budhu A, Jia HL, Forgues M, et al. Identification of 123:1911–1918. metastasis-related microRNAs in hepatocellular carci- 99. Holczbauer Á, Factor VM, Andersen JB, et al. Modeling noma. Hepatology 2008;47:897–907. pathogenesis of primary liver cancer in lineage-specific 117.Ji J, Shi J, Budhu A, et al. MicroRNA expression, sur- mouse cell types. Gastroenterology 2013;145:221–231. vival, and response to interferon in liver cancer. N Engl J 100.Cairo S, Armengol C, De Reyniès A, et al. Hepatic stem- Med 2009;361:1437–1447. like phenotype and interplay of Wnt/beta-catenin and 118.Hoshida Y, Villanueva A, Kobayashi M, et al. Gene Myc signaling in aggressive childhood liver cancer. expression in fixed tissues and outcome in hepatocel- Cancer Cell 2008;14:471–484. lular carcinoma. N Engl J Med 2008;359:1995–2004. 1239.e3 Zucman-Rossi et al Gastroenterology Vol. 149, No. 5

119.Budhu A, Forgues M, Ye Q, et al. Prediction of 136.Finn RS, Crown JP, Lang I, et al. The cyclin-dependent venous metastases, recurrence, and prognosis in hepa- kinase 4/6 inhibitor palbociclib in combination with tocellular carcinoma based on a unique immune response letrozole versus letrozole alone as first-line treatment of signature of the liver microenvironment. Cancer Cell 2006; oestrogen receptor-positive, HER2-negative, advanced 10:99–111. breast cancer (PALOMA-1/TRIO-18): a randomised 120.Kim JH, Sohn BH, Lee H-S, et al. Genomic predictors for phase 2 study. Lancet Oncol 2015;16:25–35. recurrence patterns of hepatocellular carcinoma: model 137.Shang N, Arteaga M, Zaidi A, et al. FAK is required for c- derivation and validation. PLoS Med 2014;11:e1001770. Met/b-catenin-driven hepatocarcinogenesis. Hepatology 121.Jiang P, Chan CWM, Chan KCA, et al. Lengthening and 2015;61:214–226. shortening of plasma DNA in hepatocellular carcinoma 138.Kaposi-Novak P, Libbrecht L, Woo HG, et al. Central role patients. Proc Natl Acad Sci U S A 2015:201500076. of c-Myc during malignant conversion in human hep- 122.Vogelstein B, Zhou S. Cancer genome landscapes. atocarcinogenesis. Cancer Res 2009;69:2775–2782. Science 2013;339:1546–1558. 139.Anastas JN, Moon RT. WNT signalling pathways as 123.Weinstein IB. Cancer. Addiction to oncogenes—the therapeutic targets in cancer. Nat Rev Cancer 2013; Achilles heal of cancer. Science 2002;297:63–64. 13:1–16. 124.Chapman PB, Hauschild A, Robert C, et al. Improved 140.Inoki K, Corradetti MN, Guan K-L. Dysregulation of the survival with vemurafenib in melanoma with BRAF V600E TSC-mTOR pathway in human disease. Nat Genet 2005; mutation. N Engl J Med 2011;364:2507–2516. 37:19–24. 125.Shaw AT, Kim D-W, Nakagawa K, et al. Crizotinib 141.Finn RS, Poon RTP, Yau T, et al. Phase I study investi- versus chemotherapy in advanced ALK-positive lung gating everolimus combined with sorafenib in patients cancer. N Engl J Med 2013;368:2385–2394. with advanced hepatocellular carcinoma. J Hepatol 126.Llovet JM, Ricci S, Mazzaferro V, et al. Sorafenib in 2013;59:1271–1277. advanced hepatocellular carcinoma. N Engl J Med 2008; 142.Zhu AX, Kudo M, Assenat E, et al. Effect of everolimus on 359:378–390. survival in advanced hepatocellular carcinoma after fail- 127.Llovet JM, Villanueva A, Lachenmayer A, et al. Molecular ure of sorafenib. JAMA 2014;312:57. targeted therapies in hepatocellular carcinoma. Nat Rev 143.Lim HY, Heo J, Choi HJ, et al. A phase II study of the Clin Oncol 2015;12:408–424. efficacy and safety of the combination therapy of the 128.Swain SM, Baselga J, Kim S-B, et al. Pertuzumab, MEK inhibitor refametinib (BAY 86-9766) plus sorafenib trastuzumab, and docetaxel in HER2-Positive metastatic for Asian patients with unresectable hepatocellular car- breast cancer. N Engl J Med 2015;372:724–734. cinoma. Clin Cancer Res 2014;20:5976–5985. 129.Horwitz E, Stein I, Andreozzi M, et al. Human and mouse 144.Goyal L, Muzumdar MD, Zhu AX. Targeting the HGF/c- VEGFA-amplified hepatocellular carcinomas are highly MET pathway in hepatocellular carcinoma. Clin Cancer sensitive to sorafenib treatment. Cancer Discov 2014; Res 2013;19:2310–2318. 4:730–743. 145.Santoro A, Rimassa L, Borbath I, et al. Tivantinib for 130.Llovet JM, Pena CEA, Lathia CD, et al. Plasma biomarkers second-line treatment of advanced hepatocellular car- as predictors of outcome in patients with advanced he- cinoma: a randomised, placebo-controlled phase 2 patocellular carcinoma. Clin Cancer Res 2012; study. Lancet Oncol 2013;14:55–63. 18:2290–2300. 146.Michieli P, Di Nicolantonio F. Targeted therapies: 131.Siegel AB, Cohen EI, Ocean A, et al. Phase II trial eval- Tivantinib—a cytotoxic drug in MET inhibitor’s clothes? uating the clinical and biologic effects of bevacizumab in Nat Rev Clin Oncol 2013;10:372–374. unresectable hepatocellular carcinoma. J Clin Oncol 147.Breuhahn K, Vreden S, Haddad R, et al. Molecular 2008;26:2992–2998. profiling of human hepatocellular carcinoma defines 132.Desnoyers LR, Pai R, Ferrando RE, et al. Targeting mutually exclusive interferon regulation and insulin-like FGF19 inhibits tumor growth in colon cancer xenograft growth factor II overexpression. Cancer Res 2004; and FGF19 transgenic hepatocellular carcinoma models. 64:6058–6064. Oncogene 2008;27:85–97. 148.Abou-Alfa GK, Capanu M, O’Reilly EM, et al. A phase II 133.Nicholes K, Guillet S, Tomlinson E, et al. A mouse model study of cixutumumab (IMC-A12, NSC742460) in of hepatocellular carcinoma: ectopic expression of advanced hepatocellular carcinoma. J Hepatol 2014; fibroblast growth factor 19 in skeletal muscle of trans- 60:319–324. genic mice. Am J Pathol 2002;160:2295–2307. 149.Robert C, Long GV, Brady B, et al. Nivolumab in previ- 134.Llovet JM, Decaens T, Raoul JL, et al. Brivanib in pa- ously untreated melanoma without BRAFMutation. tients with advanced hepatocellular carcinoma who were N Engl J Med 2015;372:320–330. intolerant to sorafenib or for whom sorafenib failed: re- 150.Shi F, Shi M, Zeng Z, et al. PD-1 and PD-L1 upregulation sults from the randomized phase III BRISK-PS study. promotes CD8þ T-cell apoptosis and postoperative J Clin Oncol 2013;31:3509–3516. recurrence in hepatocellular carcinoma patients. Int J 135.Johnson PJ, Qin S, Park JW, et al. Brivanib versus sor- Cancer 2010;128:887–896. afenib as first-line therapy in patients with unresectable, 151.Chen Y, Ramjiawan RR, Reiberger T, et al. CXCR4 advanced hepatocellular carcinoma: results from the inhibition in tumor microenvironment facilitates anti-PD- randomized phase III BRISK-FL study. J Clin Oncol 1 immunotherapy in sorafenib-treated HCC in mice. 2013;31:3517–3524. Hepatology 2015;61:1591–1602. October 2015 Genetic Landscape in HCC 1239.e4

152.Villa E, Critelli R, Lei B, et al. Neoangiogenesis-related 159.Viswanathan SR, Powers JT, Einhorn W, et al. Lin28 pro- genes are hallmarks of fast-growing hepatocellular car- motes transformation and is associated with advanced cinomas and worst survival. Results from a prospective human malignancies. Nat Genet 2009;41:843–847. study. Gut 2015 (in press, doi: 10.1136/gutjnl-2014- 160.Li W, Xie L, He X, et al. Diagnostic and prognostic im- 308483). plications of microRNAs in human hepatocellular carci- 153.Kim JH, Sohn BH, Lee HS, et al. Genomic predictors for noma. Int J Cancer 2008;123:1616–1622. recurrence patterns of hepatocellular carcinoma: model 161.Jiang J, Gusev Y, Aderca I, et al. Association of Micro- derivation and validation. PLoS Med 2015 (in press, doi: RNA expression in hepatocellular carcinomas with hep- 10.1371/journal.pmed.1001770). atitis infection, cirrhosis, and patient survival. Clin 154.Roessler S, Budhu A, Forgues M, et al. A unique Cancer Res 2008;14:419–427. metastasis gene signature enables prediction of tumor 162.Gramantieri L, Fornari F, Ferracin M, et al. MicroRNA-221 relapse in early-stage hepatocellular carcinoma patients. Targets Bmf in Hepatocellular Carcinoma and Correlates Cancer Res 2010;70:10202–10212. with Tumor Multifocality. Clin Cancer Res 2009; 155.Marquardt JU, Raggi C, Andersen JB. Human hepatic 15:5073–5081. cancer stem cells are characterized by common stem- 163.Hernandez-Vargas H, Lambert MP, Le Calvez-Kelm F, ness traits and diverse oncogenic pathways. Hepatology et al. Hepatocellular carcinoma displays distinct DNA 2011;54:1031–1042. methylation signatures with potential as clinical pre- 156.Coulouarn C, Factor VM, Andersen JB, et al. Loss of dictors. PLoS ONE 2010;5:e9749. miR-122 expression in liver cancer correlates with sup- 164.Calvisi D, Ladu S, Gorden A, et al. Ubiquitous activation pression of the hepatic phenotype and gain of metastatic of Ras and Jak/Stat pathways in human HCC. Gastro- properties. Oncogene 2009;28:3526–3536. enterology 2006;130:1117–1128. 157.Wei R, Huang GL, Zhang MY, et al. Clinical Significance 165.Lee S, Lee HJ, Kim J-H, et al. Aberrant CpG island and prognostic value of microRNA expression signatures hypermethylation along multistep hepatocarcinogenesis. in hepatocellular carcinoma. Clin Cancer Res 2013; Am J Pathol 2003;163:1371–1378. 19:4780–4791. 166.Kwon GY, Yoo BC, Koh, et al. Promoter methylation of 158.Augello C, Vaira V, Caruso L, et al. MicroRNA profiling of E-cadherin in hepatocellular carcinomas and dysplastic hepatocarcinogenesis identifies C19MC cluster as a nodules. J Korean Med Sci 2005;20:242–247. novel prognostic biomarker in hepatocellular carcinoma. – Liver Int 2012;32:772 782. Author names in bold designate shared co-first authorship.