integratedgenomiccomparisonofmousemodels reveals their … · sun young yim1,2, jae-jun shim1,3,...

Genomics

IntegratedGenomic Comparison ofMouseModelsReveals Their Clinical Resemblance to HumanLiver CancerSun Young Yim1,2, Jae-Jun Shim1,3, Ji-Hyun Shin1, Yun Seong Jeong1, Sang-Hee Kang1,4,Sang-Bae Kim1, Young Gyu Eun1,5, Dong Jin Lee1,6, Elizabeth A. Conner7, Valentina M.Factor7, David D. Moore8, Randy L. Johnson9, Snorri S. Thorgeirsson7, and Ju-Seog Lee1

Abstract

Hepatocellular carcinoma (HCC) is a heterogeneous dis-ease. Mouse models are commonly used as preclinical mod-els to study hepatocarcinogenesis, but how well these modelsrecapitulate molecular subtypes of human HCC is unclear.Here, integration of genomic signatures from molecularlyand clinically defined human HCC (n ¼ 11) and mousemodels of HCC (n¼ 9) identified the mouse models that bestresembled subtypes of human HCC and determined theclinical relevance of each model.Mst1/2 knockout (KO), Sav1KO, and SV40 T antigen mouse models effectively recapitu-lated subtypes of human HCC with a poor prognosis, where-as the Myc transgenic model best resembled human HCCswith a more favorable prognosis. The Myc model was alsoassociated with activation of b-catenin. E2f1, E2f1/Myc, E2f1/Tgfa, and diethylnitrosamine (DEN)-induced models wereheterogeneous and were unequally split into poor and favor-

able prognoses. Mst1/2 KO and Sav1 KO models best resem-ble human HCC with hepatic stem cell characteristics. Apply-ing a genomic predictor for immunotherapy, the six-geneIFNg score, the Mst1/2 KO, Sav1 KO, SV40, and DEN modelswere predicted to be the least responsive to immunotherapy.Further analysis showed that elevated expression of immune-inhibitory genes (Cd276 and Nectin2/Pvrl2) in Mst1/2 KO,Sav1 KO, and SV40 models and decreased expression ofimmune stimulatory gene (Cd86) in the DEN model mightbe accountable for the lack of predictive response toimmunotherapy.

Implication: The current genomic approach identified themost relevant mouse models to human liver cancer andsuggests immunotherapeutic potential for the treatment ofspecific subtypes. Mol Cancer Res; 16(11); 1713–23. �2018 AACR.

IntroductionHepatocellular carcinoma (HCC) represents 75% of cases of

primary liver cancer (1), which is the seventh most commoncancer globally (2). Despite the implementation of surveillanceprograms for at-risk populations, 30% to 60% of HCCs aredetected at an advanced stage (3), resulting in a dismal prognosis

(5-year survival rates of 0% to 10%; ref. 1). Currently, sorafeniband regorafenib are the only approved molecular-targeted ther-apies for HCC (4, 5). Therefore, improved treatments are stillneeded, and much remains to be discovered in clinical andexperimental studies.

Hepatocarcinogenesis is a multistep process involving theaccumulation of genetic changes that result in altered expressionof cancer-related genes, such as oncogenes and tumor-suppressorgenes, and their related molecular signaling pathways (6). Genet-ically engineered mouse models recapitulate the complex multi-step process of hepatocarcinogenesis so that researchers canunderstand it and design therapeutic experiments. Althoughmouse models are extensively used in cancer research, no onemouse model fits all purposes, and each model can only recapit-ulate part of the process of hepatocarcinogenesis in humans.Furthermore, genomic studies have identified molecularly dis-tinct subtypes of HCC with different clinical outcomes (7–12).Therefore, establishing the molecular and clinical resemblance ofmouse cancer models to these subtypes of human HCC (or viceversa) will enable the appropriate selection of mouse models foruse in investigations of the functional roles of newly discoveredcancer genes and validation of potential therapeutic targets.However, there have been no systematic comparisons of humanHCCs and mouse models at the molecular, genomic, and clinicallevels.

Currently, genomic studies of humanHCCare progressing, andsystematic analyses of gene expression patterns are providinginsight into the biology and pathogenesis of HCC. Many gene

1Department of Systems Biology, The University of Texas MD Anderson CancerCenter, Houston, Texas. 2Department of Internal Medicine, Korea UniversityCollege of Medicine, Seoul, Korea. 3Department of Internal Medicine, School ofMedicine, Kyung Hee University, Seoul, Korea. 4Department of Surgery, KoreaUniversity College of Medicine, Seoul, Korea. 5Department of Otolaryngology-Head andNeck Surgery, School of Medicine, Kyung HeeUniversity, Seoul, Korea.6Department of Otolaryngology-Head and Neck Surgery, Hallym UniversityMedical Center, Seoul, Korea. 7Laboratory of Experimental Carcinogenesis,National Cancer Institute, National Institutes of Health, Bethesda, Maryland.8Department ofMolecular andCell Biology, Baylor CollegeofMedicine, Houston,Texas. 9Department of Cancer Biology, The University of Texas MD AndersonCancer Center, Houston, Texas.

Note: Supplementary data for this article are available at Molecular CancerResearch Online (http://mcr.aacrjournals.org/).

S.Y. Yim, J.-J. Shim, and J.-H. Shin contributed equally to this article.

Corresponding Author: Ju-Seog Lee, The University of Texas MD AndersonCancer Center, 6565 MD Anderson Blvd, Houston, TX 77054. Phone: 713-834-6154; Fax: 713-563-4235; E-mail: [email protected]

doi: 10.1158/1541-7786.MCR-18-0313

�2018 American Association for Cancer Research.

MolecularCancerResearch

www.aacrjournals.org 1713

on September 4, 2020. © 2018 American Association for Cancer Research. mcr.aacrjournals.org Downloaded from

Published OnlineFirst August 6, 2018; DOI: 10.1158/1541-7786.MCR-18-0313

http://crossmark.crossref.org/dialog/?doi=10.1158/1541-7786.MCR-18-0313&domain=pdf&date_stamp=2018-10-10

http://mcr.aacrjournals.org/

signatures associated with HCC progression and recurrence havebeen and are being explored (8, 11, 13). In previous studies, weand others demonstrated that gene expression signatures reflect-ing the clinical and molecular characteristics of tumors are highlyconserved in human andmouse tumors (7, 9). The purpose of thepresent study, therefore, was to determine how well mousemodels recapitulated human HCCs. We determined whether11 human gene signatures associated with prognosis, activationof oncogenes, and activation of the immune system inHCCswerefound in gene expression data from nine mouse HCC models.

Materials and MethodsMouse HCC models and tumors

Nine genetically engineered or chemically inducedmouseHCCmodels were used in the current study. A total of 108HCC tumorswere collected for the study. Tumors from transgenic mice expres-sing Myc (n ¼ 15), Myc/Tgfa (n ¼ 15), E2F transcription factor1 (E2f1; n¼ 18), orMyc/E2f1 (n¼ 15) were collected as describedin previous studies (7, 14–16). The diethylnitrosamine (DEN)model (n ¼ 16) was described in an earlier study (17). The SV40T antigen transgenicmodel (n¼ 10)was also described previously(18). Generation of liver-specific Mst1/2 knockout (KO) andSalvador family WW domain containing protein 1 (Sav1) KOmice was described in a previous study (19). HCC tumors werecollected from mice ages 6 to 7 months in theMst1/2 double-KOmodel (n ¼ 4) and ages 13 to 14 months in the Sav1 KO model(n¼11). Themousemodelwith activated catenin beta 1 (Ctnnb1)was generated as described previously (20). Briefly, mice carryingloxP sites flanking Ctnnb1 exon 3 (Ctnnb1loxP(ex3)/loxP(ex3)), whichcontains the negative regulatory phosphorylation sites for degra-dationofb-catenin (21),were treatedwithAd-Cre (Cre-expressingadenovirus) tomake constitutively activatedb-catenin in the liver.Mice were further treated with a single dose of TCPOBOP (3mg/kg bodyweight), which is a constitutive androstane receptor (Car)agonist, and liver tumor promoters and tumorswere collected at 8months of age. All animals received humane care according to thecriteria outlined in the "Guide for the Care and Use of LaboratoryAnimals" prepared by the National Academy of Sciences andpublished by the National Institutes of Health (NIH publication86-23 revised 1985).

Microarrays and preparation of RNA from mouse tissuesMicroarrays for tumors from the DEN, Myc, E2f1, Myc/E2f1,

Myc/Tgfa, and SV40 models were produced in the Laboratory ofMolecular Technology, National Cancer Institute, as described inan earlier study (22). The Mouse OligoLibrary Release 1 plusExtension oligonucleotide set containing 21,997 65-mer oligo-nucleotides representing 19,740 unique genes was purchasedfrom Compugene. Total RNA was extracted from the primarycultures using TRIzol (Invitrogen). Microarray experiments werecarried out as described in a previous study (22). Briefly, labeledRNAs were hybridized against common reference RNAs isolatedfrom B6/129 wild-type normal liver using a reverse-fluor design,and gene expression values were defined as a target-per-referenceratio. Twenty micrograms of total RNA was used to synthesizefluorescently labeled cDNA probes. Preparation of the labeledcDNA samples and hybridization of oligonucleotide microarrayswere performed as previously described, with smallmodifications(22). Microarray experiments with tumors fromMst1/2 KO, Sav1KO, and Ctnnb1 models were carried out with Illumina mouse-6v2 arrays (Illumina). A total of 500 ng of total RNA from tumors

and wild-type normal livers was used for labeling and hybridiza-tion according to the manufacturer's protocol. Collected geneexpression data were transformed and normalized as describedpreviously (7–9). Because gene expression data from tumors fromthe DEN, Myc, E2f1, Myc/E2f1, Myc/Tgfa, and SV40 models weredefined as expression ratio to normal liver, gene expression fromtumors from the Mst1/2 KO, Sav1 KO, and Ctnnb1 models wasrenormalized as expression ratio to normal liver by dividing bythe expression value from normal liver.

Clinically defined liver cancer subtypes and associated geneexpression signatures

The National Cancer Institute proliferation (NCIP; refs. 7, 8,23), hepatic stem cells (HS; ref. 9), Seoul National Universityrecurrence (SNUR; ref. 12),Cholangio-likeHCC(CLHCC; ref. 10),silence ofHippo (SOH; ref. 11), recurrence-risk score (RS; ref. 13),b-catenin (BC; ref. 20), EPCAM (24), and Hoshida (25). HCCsubtypes and associated gene signatures were described in earlierstudies. IDH-like signatures were identified in a recent analysis ofliver cancer genomic data from The Cancer Genome Atlas (26).The six-gene IFNg (IFNG6) composite score was calculated on thebasis of genes reported in a recent study (27). The numbers ofgenes in each signature are listed in Table 1.

Analysis of data and stratification of mouse tumorsThe BRB Array Tools software program (http://linus.nci.nih.

gov/BRB-ArrayTools.html) was used for analysis of the geneexpression data and construction of a prediction model (28). Aheatmap was generated using the Cluster and TreeView softwareprograms (29), and further statistical analysis was performedusing the R language (http://www.r-project.org). Data on geneswithmore than 30%of their expression datamissing across tissuesampleswere excluded and then transformed andnormalized (8).When a gene was represented more than once on the microarrayplatform, the genes with the greatest variance in expression wereselected. To identify mouse genes whose expression changednontrivially, we selected expression ratios with at least a 2-folddifference relative to a reference in at least nine tissues (880genes). The selected geneswere used for unsupervised hierarchicalcluster analysis and principal component analysis for the ninemouse models.

Before collation of human andmouse gene expression data forconstruction of prediction models, expression levels of ortholo-gous genes in themouse and humandata sets were independentlystandardized by transforming each gene's expression level to a

Table 1. Prognostic molecular subtypes of HCC and associated gene expressionsignatures

Signature CharacteristicNumber of genesin signature References

NCIP Prognostic 947 8SNUR Prognostic 628 12CLHCC Prognostic 581 10RS Prognostic 65 13HS Prognostic 907 9IDH-like Prognostic 1,009 21EPCAM Prognostic 793 17BC Prognostic 82 26SOH Prognostic 347 11Hoshida Prognostic 615 31IFNG6 score Benefit from

immunotherapy6 32

Yim et al.

Mol Cancer Res; 16(11) November 2018 Molecular Cancer Research1714



http://linus.nci.nih.gov/BRB-ArrayTools.html

http://linus.nci.nih.gov/BRB-ArrayTools.html

http://www.r-project.org


meanof 0 and an SDof 1, as described in earlier studies (7, 22). Toconstruct the prediction models, we used a Bayesian compoundcovariate prediction (BCCP) model as described previously toestimate the probability that a particular mouse HCC tissuewould have a given gene expression signature (7–9, 30). Geneexpression data in training sets (based on the clinically definedprognostic signatures) were combined to form a classifieraccording to a BCCP. The robustness of the classifier wasassessed using a misclassification rate determined usingleave-one-out cross-validation in the training set. The BCCPclassifier estimated the likelihood that an individual mousetumor would have either subtype of a given gene expressionsignature and dichotomized tumors according to Bayesianprobability (cutoff of 0.5).

The RS for each tumor was generated using the followingprocedure. First, gene expression level was multiplied by a pre-determined coefficient for each gene, as indicated in a previousstudy (13), generating gene values. Second, raw RSs were gener-ated by summation of all gene values. Third, the raw scores wererescaled by multiplying them by 6 and subtracting the minimumscore among the tumors from the rescaled scores. The finalrescaled RSs ranged from 0 to 98.22. Tumors with scores higherthan 40 were considered to have a high risk of recurrence.

Immune activity in mouse HCCs was analyzed by calculatingIFNG6 composite scores, which were previously shown to predictresponse to anti–PD-1 therapy (pembrolizumab) in patients withcancer (27). The IFNG6 composite score for each tumor wascalculated by averaging the expression levels of six genes (CXCL9,CXCL10, IDO1, STAT1, IFNG, and HLA-DR) for each tumor. Themedian value, 1.68, was used as the cutoff demarcating high andlow immune response.

Data accessRaw and processed gene expression data are available under

GSE32510, GSE43628, and GSE110627 in the Gene ExpressionOmnibus database.

ResultsGlobal gene expression patterns reveal differences andsimilarities among mouse liver cancer models

To assess the differences in gene expression patterns among thenine mouse HCC models, we performed hierarchical clusteranalysis of the gene expression data, revealing several clusterswith distinct gene expression patterns (Fig. 1A). As expected,tumors in each model formed tight clusters that were well

DENSV40

Myc/Tgfa

E2f1Myc

Myc/E2f1

Mst1/2 KOCtnnb1

Sav1 KO

0 4-4 2-2

A

B

PC 1

PC 3

PC 2 DENSV40

Myc/Tgfa

E2f1Myc

Myc/E2f1

Mst1/2 KOCtnnb1

Sav1 KO

Figure 1.

Hierarchical cluster of gene expressiondata from nine mouse HCC models.A, Unsupervised hierarchical clusteranalysis of 108mouse tumors revealedsimilarities and differences among themouse models. Genes with anexpression level that was at least 2-fold different from the median valueacross HCC tissues in at least ninetissues were selected for hierarchicalclustering analysis (880 genes). Thedata are presented in matrix format,where each row represents anindividual gene, and each columnrepresents a tissue type. Each cell inthe matrix represents the expressionlevel for a gene in an individual tissuetype. The red and green colors in thecells reflect the relatively high and lowgene-expression levels, respectively,as indicated by the scale bar(a log2-transformed scale). B,Principal component (PC) analysisof gene expression data from thenine HCC models.

Genomic and Clinical Resemblance of HCC Model Systems

www.aacrjournals.org Mol Cancer Res; 16(11) November 2018 1715




separated from those of the other mouse models. However, geneexpression levels in a few tumors in the samemodels were spreadout across different clusters, suggesting a certain level of hetero-geneity within each model. The simian vacuolating virus 40(SV40) tumors were grouped in a single cluster, with tumorsseparated by the shortest distance in the resulting dendrogram,suggesting that these tumors were themost homogeneous amongthe examined models. Similarly, tumors in the DEN-inducedmodel formed a tight cluster and had the most distinct geneexpression patterns, suggesting that chemically induced tumorsmay be different from those generated in genetically engineeredmouse models. Not surprisingly, tumors from mammalianSTE20-like protein kinase (Mst)1/2 double-KO and Salvadorfamily WW domain containing protein 1 (Sav1) KO mice wereclustered together because these genes are core regulators oftumor-suppressive Hippo pathways (31). SV40 tumors clusteredtogether withMst1/2 KO and Sav1 KO tumors, suggesting that theoncogenic activity of SV40 T antigenmight bemediated in part bysuppressionof theHippopathwayor activationof its downstreamtarget oncogene, Yes-associated protein 1 (Yap1), in hepatocytes.Tumors from the activated Ctnnb1 mouse model were well sep-arated frommost other tumors, indicating that this mouse modelmay represent a unique subtype of HCC. Tumors in the myelo-cytomatosis (Myc) and E2f1 models were spread across differentclusters, and some of them were highly similar to tumors in theother genetically engineeredmodels, suggesting that these tumorswere the most heterogeneous among the examined mouse mod-els. Tumors from the Ctnnb1 model were clustered together withmost tumors from theMycmodel, indicating that the tumors fromthese two models may share similar molecular characteristics inhuman HCC. Reassessment of the gene expression data usingprincipal component analysis showed a similar interrelationshipof tumor gene expression profiles (Fig. 1B). The SV40,Mst1/2 KO,and Sav1 KO models were well separated from the rest of themouse models and formed tight clusters in three-dimensionalspace, providing further evidence of the similarity of these mod-els. Likewise, the DEN model was also well separated from theother models.

Resemblance ofmouse tumors to prognostic subtype of humanliver cancer

NCIP subtypes (A or B) were first discovered using unsuper-vised analysis of genome-level expression data from humanHCCtissues; subtype A represents tumors with a poor prognosis (8).When we stratified the mouse tumors according to their NCIPsignatures, 43 of the 108 tumors were classified as subtype A(Fig. 2A). All tumors in the SV40,Mst1/2KO, Sav1KO, andCtnnb1models were classified in subtype A, suggesting that these modelsmimic the poor prognostic subtype A of human HCC. Asexpected, the double-transgenic models (Myc/E2f1 andMyc/Tgfa)had substantially higher proportions (27% in each group) ofsubtypeA tumors thandid the single-transgenicmodels (Myc: 7%;E2f1: 17%). The fraction of subtype A DEN tumors (13%) wassimilar to that in the single-transgenicmodels. TheMycmodel hadthe lowest number (n ¼ 1) of subtype A tumors, suggesting thatthis model has the least aggressive phenotype among the testedHCC models. In agreement with this observation, HCC develop-ment in Myc mice was previously shown to be less frequent andslower than in E2f1, Myc/E2f1, and Myc/Tgfa mice (32).

Researchers previously defined SNUR subtypes (recurrence-high or recurrence-low) using supervised approaches to select

genes associated with early disease recurrence after curative-intenttreatment (12). Consistent with theNCIP results, all of the tumorsfrom the SV40,Mst1/2 KO, and Sav1 KOmodels were classified asthe high-recurrence subtype (Fig. 2B). In contrast to tumors fromtheMst1/2 KO and Sav1 KOmodels, only 3 of 4 tumors from theCtnnb1 model were classified as the high-recurrence subtype,indicating that tumors in the Ctnnb1 model may not be asaggressive as those in theMst1/2 KO, Sav1 KO, and SV40 models.TheMyc/E2f1model had a higher proportion of recurrence (27%)than did the other transgenic models, suggesting that interactionbetween E2f1 and Myc in tumorigenesis may facilitate earlyrecurrence after treatment.

CLHCC subtypes (C1 or C2) were defined by gene-expressionpatterns similar to those found in cholangiocarcinoma (10). TheC1 subtype represents tumors with poor prognosis. As for theNCIP and SNUR signatures, all tumors from theMst1/2 KO, Sav1KO, Ctnnb1, and SV40 models were classified as C1. However,unlike NCIP and SNUR, the rest of the models had low propor-tions of tumors in the poor prognostic C1 group (Fig. 2C).

The recurrence-risk score (RS) is determined by integratingmultiple prognostic signatures to identify a small number ofgenes that predict recurrence after treatment (13); scores typicallyrange from 0 to 100. In good agreement with the results for theother prognostic signatures,Mst1/2 KO and Sav1 KOmodels hadthe highest median RS (100 for Sav1 KO and 93.4 forMst1/2 KO).The SV40model had the third highest median RS (63.18) amongthemodels, and theCtnnb1model had amoderately highmedianRS (56.7). Among all models, the Myc model had the lowestmedian RS (11.8). This provides further confirmation that theMst1/2 KO and Sav1 KO models recapitulate the most aggressivehuman HCCs and theMycmodel recapitulates human HCCwiththe best prognosis (Fig. 2D). The concordance of the predictedoutcomes for mouse tumor models with the four prognosticsubtypes was further supported by pairwise scatter plots of Bayes-ian probabilities correlating prediction models and risk scores(Supplementary Fig. S1).

We next applied the glycolytic gene expression signature frommouse liver (33) to examinemetabolic activity in the ninemousemodels (Supplementary Fig. S2). Not surprisingly, tumors fromthe most aggressive models—Mst1/2 KO, Sav1 KO, and SV40—had the highest glycolytic activity, whereas tumors from the lessaggressive Myc, E2f1, and DEN models had the lowest glycolyticactivity (Supplementary Fig. S3). Interestingly, glycolytic activitywas significantly higher (P ¼ 0.001) in tumors from theMyc/Tgfamodel than in tumors from the Myc/E2f1 model, whereas theoverall aggressiveness of the tumors was equally moderate, asindicated in the prognostic signatures (Fig. 2), suggesting thatmetabolic activity is not always correlated with aggressiveness oftumors.

Resemblance of mouse tumors to stem cell subtypes of humanliver cancer

Having determined that the mouse HCC models effectivelyrecapitulated the various prognostic characteristics of humanHCC, we next examined whether the mouse models also reca-pitulated the lineage of tumor cells. The hepatic stem cell (HS)gene signature [HS or hepatocyte (HC) subtypes] was previouslydefined as gene-expression patterns resembling those found infetal hepatic stem cells; the prognosis for patients with the HSsubtype is extremely poor (9). When we compared the HS gene-expression signature with gene expression data from the mouse

Yim et al.





tumors, only tumors from Mst1/2 KO and Sav1 KO models wereclassified as HS subtype (Fig. 3A), suggesting that inactivation ofthe Hippo pathway may lead to activation of stem cell character-istics in hepatocytes.

To validate the association of tumors from the Mst1/2 KO andSav1 KO models with the HS subtype, we applied the previouslydefined EPCAM signature (24). EPCAM is frequently overex-pressed in cancer-initiating cells in multiple cancer types, includ-ing liver cancer (34–36). In good agreement with the HS signa-ture, the vastmajority of tumors from theMst1/2 KO and Sav1KOmodels were grouped as EPCAM-positive tumors (Fig. 3B). Sevenof 10 tumors from the SV40 model were EPCAM-positive, sug-gesting that SV40 T antigen may dedifferentiate hepatocytes tostem cell–like cells during tumorigenesis.

To further validate the association of tumors from the Mst1/2KO and Sav1 KOmodels with stem cell characteristics, we appliedthe isocitrate dehydrogenase (IDH)-like signature, an indepen-

dent gene expression signature that also reflects the hepatic stemcell characteristics of HCC (37), to the mouse tumors. Tumorswith the IDH signature exhibited activatingmutations in the IDH1and IDH2 genes. Activated mutant IDH blocks hepatic differen-tiation of hepatic stem cells through production of 2-hydroxy-glutarate and suppression of the activity of HNF4, a masterregulator of hepatic differentiation (38). Like patients with theHS subtype, the prognosis for patients with the IDH-like subtypeis extremely poor, and these tumors exhibit very high similarity tothe HS subtype (37). Consistent with the results of the HSsignature analysis, only tumors from the Mst1/2 KO and Sav1KOmodels were predicted to have the IDH-like subtype (Fig. 3C).Outcomes of the three predictors were significantly correlated(Supplementary Fig. S4). Although one tumor from theMyc/Tgfamodel was classified as IDH-like, its prediction was inconsistentbecause it was not classified as HS subtype or EPCAM-positive. Inagreement with predictions, the expression of hepatic stem cell

40−4

NCIP

SNUR

CLHCC

A

B

C

D

DENSV40

Myc/Tgfa

E2f1Myc

Myc/E2f1

Mst1/2 KOCtnnb1

Sav1 KO

020

4060

8010

0

Myc DENMyc

/TgfaE2f1

Myc/E2f

1

SV40

Ctnnb1

Sav1 KO

Mst1/2

KO

Rec

urre

nce-

risk

scor

es

Figure 2.

Prognostic characteristics of tumors innine mouse HCC models. Mouse HCCtumors were stratified according toprognostic gene expressionsignatures for human HCCs: NCIPsignature (A), SNUR signature (B), andCLHCC signature (C). D, Recurrence-risk scores for each mouse model.Black bars indicate subtype A in NCIP,high-recurrence in SNUR, and subtypeC1 in CLHCC.






markers such as Krt7, Sall4, and Sox9 was significantly higher inMst1/2 KO and Sav1 KO models than in other mouse models(Supplementary Fig. S5; ref. 39). Expression of Krt7 protein intumors from the Sav1KOmodel further supported our prediction(Supplementary Fig. S6).

Resemblance ofmouse tumors to biological subtypes of humanliver cancer

The oncogene CTNNB1, which encodes for b-catenin, is fre-quently activated via somatic mutations in HCC (40, 41). Theb-catenin gene expression signature [BC or non-BC (NBC) sub-type] was previously defined as activation of b-catenin by somaticmutations inhumanHCC(20). As expected, all tumors frommicewith activated b-catenin (Ctnnb1 model) were classified as BCsubtype (Fig. 4A). A large fraction (67%) of tumors from theMycmodel were also classified as BC subtype (Fig. 4A), suggesting thatb-catenin is highly activated in the Myc transgenic model. Fewertumors were BC subtype in the double-transgenic models withMyc activation: 47% in Myc/E2f1 and 27% in Myc/Tgfa. None ofthe tumors in the DEN model and only one tumor in the SV40model were BC subtype. To directly measure b-catenin activity inthe mouse models, we analyzed the expression of Glul and Rhbg,liver-specific direct target genes of b-catenin (42), in microarrayexperiments. Consistent with results for the b-catenin signature,expression of these geneswas highest in theCtnnb1model, secondhighest in theMycmodel, and lowest in the SV40 andMst1/2 KO

models (Supplementary Fig. S7A). A small fraction (27%) oftumors in the Sav1 KO model had activated Ctnnb1, suggestinga potential interaction between Ctnnb1 and the Hippo pathway.

The Hippo pathway is a major tumor-suppressor pathway inliver cells; its inactivation triggers activation of the oncogeneYAP1, leading to the development of HCC (31, 43). Inactivationof the Hippo pathway is associated with poor prognosis forhuman HCC (11, 44). The SOH signature [SOH or activatedHippo (AH) subtypes] was previously defined as a gene-expres-sion signature associated with silencing of the Hippo pathway(11, 44). As expected, all tumors from theMst1/2KOand Sav1KOmodel were classified as SOH subtype (Fig. 4B). The vast majority(90%) of tumors from the SV40 model were classified as SOHsubtype, suggesting that Yap1 is highly active in the SV40 trans-genic model. Tumors from the Ctnnb1 model were classified asSOH subtype, clearly indicating an interaction between Ctnnb1and Yap1 in the development of HCC. With the exception of oneE2f1 tumor andoneMyc/Tgfa tumor, no tumors in anyof theothermouse models were SOH tumors. To directly measure Yap1activity, we analyzed expression of Ctgf and Cyr61, direct targetgenes ofYap1 (43, 45), frommicroarray experiments. As expected,expression of these Yap1 target geneswas highest in theMst1/2KOand Sav1 KO models and very high in the Ctnnb1 and SV40models (Supplementary Fig. S7B).

In a previous study, Hoshida and colleagues classified HCCsinto three molecular subtypes, S1, S2, and S3 (25). The S1 and S2

HSA

B

DENSV40

Myc/Tgfa

E2f1Myc

Myc/E2f1

Mst1/2 KOCtnnb1

Sav1 KO

IDH-like

40−4

C

EPCAMFigure 3.

Similarity of mouse tumors to HSsubtypes of human HCC. Mousetumors were stratified according togene expression signatures reflectingHS subtypes of human HCC: HSsubtype signature (A), EPCAMsubtype signature (B), and IDH-likesubtype signature (C). Black barsindicate HS subtype (A), EPCAM-high(B), and IDH-like HCC (C).

Yim et al.





subtypes are characterized by poor prognosis, high cellular pro-liferation, and stem cell–like characteristics, whereas the S3 sub-type is associated with better prognosis, and tumors are typicallywell-differentiated. The S2 subtype was most significantly asso-ciated with SOH tumors (Fig. 4C), whereas the S1 subtype wasassociated with the SV40 model, and most tumors from thetransgenic models were classified as S3. The S2 subtype was leastpresent in the mouse models; the number of S2 tumors was verysmall.

Immune activity in mouse liver cancer modelsWe next examined immune activity in mouse tumors by

calculating previously validated IFNG6 composite scores, whichreflect overall immune activity and predict response to anti–PD-1(pembrolizumab) therapy in cancer patients (27). When tumorswere dichotomized by mean IFNG6 score using a cutoff of 1.68,the vast majority of tumors from the E2f1 model (88%) wereclassified into the high immune activity subgroup (IFNG6 �1.68), whereas the vast majority of tumors from the Mst1/2 KO(100%), Sav1 KO (91%), Ctnnb1 (100%), SV40 (80%), and DEN(94%) models were classified into the low immune activitysubgroup (Fig. 5A). Furthermore, the Mst1/2 KO and Sav1 KOmodels had the lowest median IFNG6 scores, and the E2f1modelhad the highest median score among the nine models (P¼ 5.4�10�9 in Mst1/2 KO compared with E2f1, 2.9 � 10�9 in Sav1 KOcompared with E2f1, 1.9 � 10�6 in Ctnnb1 compared with E2f1,7.2� 10�7 in SV40 compared with E2f1, and 8.3� 10�10 in DENcompared with E2f1, by Student t test; Fig. 5B), suggesting thatgenetic alterations in mouse cancer cells contribute to the com-position and activity of immune cells in the tumor microenvi-ronment. Consistent with the results of a previous study showingthat IFNG6 scores can predict the response of tumors to pem-brolizumab (23), we found that IFNG6 scores in mouse tumors

were correlated with expression of Pd-1, the target of pembroli-zumab (r¼ 0.41, P¼ 0.0002), and Pd-l1, ligand of Pd-1 (r¼ 0.26,P ¼ 0.02; Supplementary Fig. S8).

Because immune-checkpoint genes control the balance ofstimulatory and inhibitory signals and play critical roles in reg-ulating T-cell activity, we examined the expression of ligands andreceptors for stimulatory and inhibitory signals in tumors in thenine mouse models (Fig. 5C). Of 10 stimulatory signaling genesstudied, expression of Cd86 was significantly lower in the DENmodel than in any of the other models (Fig. 5D; P < 0.001 in allStudent t tests), suggesting that the low immune activity in theDEN model is attributable to a lack of stimulatory signals fromcancer cells. Of 12 inhibitory signaling genes, expression ofCd276and Pvrl2was high inHCCmodelswith highYap1 activity (Mst1/2KO, Sav1 KO, and SV40 models; Fig. 5D), suggesting thatincreased expression of checkpoint-inhibitory genes in thesemodels might be attributable to low immune system activity andYap1might be accountable for increased expression ofCd276 andPvrl2. To determine if these checkpoint-inhibitory genes areregulated by Yap1, we accessed previously published gene expres-sion data from the Yap1mouse transgenicmodel (46). Expressionof Cd276 was significantly increased in YAP1-induced mouselivers (Supplementary Fig. S9).

DiscussionBy systematically integrating genomic signatures from human

HCCs with genomic data from nine mouse HCC models, weassessed the clinical relevance of these mouse models by deter-mining their similarity to human HCCs in terms of clinical andmolecular characteristics (Fig. 6). Of the nine examined mouseHCCmodels,Mst1/2 KO and Sav1 KOmodels well mimicked theHS subtype, which confers the poorest prognosis among the

BC

SOH

A

B DENSV40

Myc/Tgfa

E2f1Myc

Myc/E2f1

Mst1/2 KOCtnnb1

Sav1 KO

S2S1

S3

CHoshida

Figure 4.

Similarity of mouse tumors to molecularsubtypes of human HCC. Mouse tumorswere stratified according to gene-expression signatures reflectingmolecular subtypes of human HCC: BC(A), SOH pathway (B), and Hoshida'sthree subtypes (C). Black bars indicateBC subtype (A) and SOH subtype (B).






examinedhumanHCC signatures. Somemousemodels appearedto be clinically heterogeneous, given that different tumors in thesame mouse models were categorized into human subtypes withdifferent prognostic characteristics. We also showed that themouse tumor models significantly differed in their ability tostimulate host immune activity. In the current study, we demon-

strated that themousemodels effectively recapitulated importantaspects of humanmalignancies at both themolecular and clinicallevels.

HS HCCs were identified by several studies. Current analysiswith genomic signatures showed that two mouse HCC modelswith a disrupted Hippo pathway (Mst1/2 KO and Sav1 KO) are

A

Stim

ulat

ory

Inhi

bito

ry

B

D

C

01

23

HighLow

DENSV40

Myc/Tgfa

E2f1Myc

Myc/E2f1

Mst1/2 KOCtnnb1

Sav1 KO

1.0

2.0

3.0

IFN

G6

scor

es

MycDEN

Myc/Tgf

aE2f1

Myc/E2f

1

SV40

Ctnnb1

Sav1 KO

Mst1/2

KO

Il2rbCd40Cd86IcosCd28Cd80Tnfrsf9Tnfrsf18Tnfsf9Cd226Pdcd1Lgals9Cd274Havcr2Pdcd1lg2BtlaLag3Ido1Adora2aTnfrsf14Cd276Pvrl2

−3−2

−1 0

1

−2−1

01

2

−10

12

34

Cd86 Cd276 Pvrl2

Exp

ress

ion

ratio

s (lo

g 2)

Exp

ress

ion

ratio

s (lo

g 2)

Exp

ress

ion

ratio

s (lo

g 2)

MycDEN Myc/Tgf

a E2f1

Myc/E2f

1

SV40

Ctnnb1

Sav1 KO

Mst1/2

KO

IFN

G6

scor

es

* * * * *

MycDEN Myc/Tgf

a E2f1

Myc/E2f

1

SV40

Ctnnb1

Sav1 KO

Mst1/2

KO MycDEN Myc/Tgf

a E2f1

Myc/E2f

1

SV40

Ctnnb1

Sav1 KO

Mst1/2

KO

Figure 5.

Relative immune activity of tumors innine mouse hepatocellular carcinomamodels. A, Tumors in the nine modelswere grouped according to IFNG6composite score (mean expressionlevel for Cxcl10, Cxcl9, Ido1, Stat1,H2-Ea, and Ifng). Tumors weredichotomized by mean IFNG6 score ata cutoff of 1.68, indicated by the arrow.B, IFNG6 scores in the nine mousemodels. � , P < 0.05. C, Expressionpatterns for immune-checkpointgenes in mouse tumors. Expressionof Cd86, Cd276, and Pvrl2 wassignificantly different in theMst1/2KO,Sav1 KO, DEN, and SV40 models fromthat in other models. D, Box plots ofCd86, Cd276, and Pvrl2 expressionlevels in mouse models. Geneexpression data were normalizedas ratios to expression in normalmouse liver.

Mst1/2Sav1

SV40DEN

Ctnnb1Myc E2f1

Myc/Tgfa

Myc/E2f1

NCIP-ASNUR-highCLHCC-C1

RS-high

BC-BCSOH-SOH

IFNG-high

IDH-likeHS-HS

highmidlowS1S2S3

Glycolysis

Hoshida

EPCAM-like

Figure 6.

Summary of stratification of mousetumors according to clinically definedhuman HCC genomic signatures.

Yim et al.





most similar to the HS subtype. The association of these mousemodels with the HS subtype is also supported by their similarityto previously established EPCAM-positive human HCC (24)and newly discovered IDH-like human HCC tumors (37),whose underlying biology well reflects stem cell characteristicsby inhibiting hepatic differentiation of hepatic stem cells throughproductionof 2-hydroxyglutarate (38). Interestingly, theBayesianprobabilities of three independent stem cell predictors weresignificantly correlated with the probability of SOH (Supplemen-tary Fig. S4), indicating that inactivation of the Hippo pathwaytriggers activation of stem cell characteristics in hepatocytes.Further support for this idea is supplied by the strong associationofmouse SOH tumorswithHoshida's S2 subtype in humanHCC,which is best associated with cancer stem cells (25, 47). Ourobservation is in good agreement with a previous study showingthat Yap1 reprogramed mature hepatocytes in adult mice intoprogenitor-like cells that could transdifferentiate into biliaryepithelial cells (46). Similar promotion of stem/progenitor cellswas also observed by inactivation of the Hippo pathway inmultiple organs (48). Twomouse models were also characterizedby a high probability of recurrence after treatment, as reflected inthe highest RS among these examinedmodels, further supportingthe similarity of these models to the HS subtype; previous studiesdemonstrated that HCC with stem cell characteristics has thepoorest clinical outcomes among all HCCs (9, 24, 26).

The SV40 T antigen induces oncogenic transformation ofnormal cells, including hepatocytes, by inactivating the tumor-suppressor genes p53 and Rb and interacting with a number ofsignaling proteins such as HSC70, CBP/p300, CUL7, IRS1,FBXW7, and BUB1 (49). The vast majority (90% to 100%) ofSV40 model tumors were classified as subtypes indicating poorprognosis for most genomic signatures except for stem cell sig-natures, suggesting that the SV40 model had the most aggressiveHCC tumors among the five examined transgenic models. Thiscould be because transformation by the SV40 T antigen is medi-ated by a large number of oncogenes and tumor suppressors.Interestingly, the SV40 model exhibited a significant associationwith inactivation of the Hippo pathway, suggesting that the SV40T antigen is involved in activation of Yap1. In agreement with thisobservation, a recent study showed that the SV40 T antigen canactivate YAP1 through PAK1-mediated inhibition of NF2, anupstream negative regulator of YAP1 in the Hippo pathway(50). In support of this notion, expression of Nf2 was lowestamong the five transgenic mouse models (Fig. 5D). However,unlike tumors from theMst1/2 KO and Sav1 KO models, tumorsfrom the SV40 model lacked stem cell signatures, suggesting thatYap1 activity in SV40 tumors is not as strong as in tumors from theMst1/2 KO and Sav1 KO models. In fact, expression of Ctgf andCyr61, key downstream targets of Yap1, was substantially lower intumors from the SV40model than in tumors from theMst1/2 KOand Sav1 KO models.

CTNNB1 is one of themost frequentlymutated genes inhumanHCC, and these mutations lead to activation of b-catenin (51).Activation of b-catenin was markedly enhanced in theMyc trans-genic model (67%; Fig. 6), suggesting that b-catenin is a frequentcoactivating partner of Myc in the development of HCC. Inter-estingly, coactivation of b-catenin was substantially less commonin double-transgenic models (47% inMyc/E2f1 and 27% inMyc/Tgfa) than in the single-transgenic Myc model, suggesting thatdifferent oncogenes can replace b-catenin in the development ofHCC. Also, b-catenin activation was substantially less frequent

(33%) in the E2f1 model than in the Myc model, suggesting thatE2f1 has greater potential for oncogenic transformation than doesMyc in hepatocytes. This is in agreement with previous observa-tions showing thatCtnnb1 is frequentlymutated inMycmice (32).Most of mutations (S33Y, S35C, T41A, and S45F) are located inphosphorylation sites for negative regulation and lead to consti-tutive activation of b-catenin. The lack of b-catenin activation inthe DEN model was also in agreement with previous studiesshowing that the most prevalent mutations in tumors in DEN-treatedmice were ofHras and detecting nomutations ofCtnnb1 inthese mice (52). All Ctnnb1 tumors showed high Yap1 activitybecause they were classified as SOH subtype, suggesting interac-tion of Ctnnb1 and Yap1 in the development of HCC. Thisobservation is in good agreement with a previous study demon-strating that Yap1 is necessary for survival and tumorigenesis ofb-catenin–driven cancers (53). Unlike tumors from the Ctnnb1model, only a small fraction (27%) of tumors from the Sav1 KOmodel and none from theMst1/2model had activated b-catenin,suggesting unequal interaction of two oncogenes in HCC devel-opment. Yap1 might be a key interacting partner for b-catenin–driven tumorigenesis, whereas b-cateninmight not be an essentialpartner forYap1-driven tumorigenesis in the liver. In fact, previousstudies demonstrated that activation of b-catenin alone is notsufficient to develop HCC (21), whereas activation of Yap1 aloneis sufficient to develop HCC (54).

Recent advances in our understanding of cancer-relatedimmunobiology have led to the development of various immu-notherapeutic strategies, including vaccination, adoptive celltherapy, and immune-checkpoint inhibition. In particular,blockade of checkpoint molecules such as PD-1 and CTLA-4has emerged as a novel therapeutic approach for cancer (55,56). PD-1 is a negative costimulatory receptor and a stronginhibitor of T-cell response. Pembrolizumab, which targets PD-1, is approved for the treatment of several cancers, includingmetastatic melanoma (55). When we assessed immune activityindicative of a response to pembrolizumab in mouse tumorsusing previously validated IFNG6 scores, which can predictresponse of cancer cells to pembrolizumab, we found thattumors in the Mst1/2 KO and Sav1 KO models had low scoreswith concomitant expression of the immune-checkpoint inhi-bitors Cd276 and Pvrl2. Likewise, the SV40 model, which hashigh Yap1 activity, showed very low IFNG6 scores and highexpression of Cd276 and Pvrl2, supporting the notion that Yap1may suppress host immunity through direct or indirect regu-lation of immune-checkpoint inhibitors. This observation sug-gests that Yap1-mediated suppression of host immune activitymay contribute to the poor prognostic characteristics of thesemouse models. In the DEN mouse model, expression of Cd86,an immune-checkpoint stimulator, was markedly lower than itis in other mouse models. Importantly, expression of Pd-1 inmouse tumors correlated with IFNG6 scores, which predictresponse to anti–PD-1 therapy in human cancers (27).

In the current study, we identified the most appropriate mousemodels recapitulating various subtypes of human HCC. Ourfindings elucidate similarities in the underlying biology ofmousemodels and subtypes of human HCC. The ability of mousemodels to represent different levels of immune activity, reflectedin IFNG6 scores,mayprovide a framework for guiding selection ofthemost appropriate mousemodels for preclinical trials of newlydeveloped immunotherapies. However, our study is limited by alack of functional validation of underlying biology in mouse






models associated with clinical outcomes. Therefore, future studyshould be carried out to establish the clinical and molecularrelationships between subtypes of human HCC and their corre-spondingmousemodels to gain better insights into the treatmentand prevention of HCC.

Disclosure of Potential Conflicts of InterestNo potential conflicts of interest were disclosed.

Authors' ContributionsConception and design: S.Y. Yim, J.-H. Shin, S.S. Thorgeirsson, J.-S. LeeDevelopment of methodology: S.Y. Yim, J.-S. LeeAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): S.Y. Yim, J.-J. Shim, J.-H. Shin, Y.G. Eun, E.A. Conner,V.M. Factor, D.D. Moore, R.L. JohnsonAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): S.Y. Yim, J.-J. Shim, S.-H. Kang, S.-B. Kim, Y.G. Eun,D.J. Lee, J.-S. LeeWriting, review, and/or revision of the manuscript: S.Y. Yim, J.-J. Shim,J.-H. Shin, Y.S. Jeong, S.S. Thorgeirsson, J.-S. Lee

Administrative, technical, or material support (i.e., reporting or organizingdata, constructing databases): S.Y. Yim, J.-H. Shin, Y.S. Jeong, S.-H. Kang,D.J. Lee, J.-S. LeeStudy supervision: J.-S. Lee

AcknowledgmentsThis study was supported in part by the Duncan Cancer Prevention

Research Seed Funding Program at The University of Texas MD AndersonCancer Center (2016 cycle), the MD Anderson Sister Institute Network Fund(2012 and 2016 cycles), and the National Institutes of Health throughCancer Center Support Grant P30 CA016672. The authors thank AmyNinetto of the Department of Scientific Publications at MD Anderson forediting the article.

The costs of publication of this article were defrayed in part by thepayment of page charges. This article must therefore be hereby markedadvertisement in accordance with 18 U.S.C. Section 1734 solely to indicatethis fact.

Received March 29, 2018; revised July 2, 2018; accepted July 26, 2018;published first August 6, 2018.

References1. Forner A, Llovet JM, Bruix J. Hepatocellular carcinoma. Lancet 2012;

379:1245–55.2. Torre LA, Bray F, Siegel RL, Ferlay J, Lortet-Tieulent J, Jemal A. Global cancer

statistics, 2012. CA Cancer J Clin 2015;65:87–108.3. Bruix J, Sherman M, American Association for the Study of Liver D.

Management of hepatocellular carcinoma: an update. Hepatology2011;53:1020–2.

4. Llovet JM, Ricci S, Mazzaferro V, Hilgard P, Gane E, Blanc JF, et al.Sorafenib in advanced hepatocellular carcinoma. N Engl J Med 2008;359:378–90.

5. Bruix J, Qin S, Merle P, Granito A, Huang YH, Bodoky G, et al. Regorafenibfor patients with hepatocellular carcinoma who progressed on sorafenibtreatment (RESORCE): a randomised, double-blind, placebo-controlled,phase 3 trial. Lancet 2017;389:56–66.

6. Pitot HC. Pathways of progression in hepatocarcinogenesis. Lancet2001;358:859–60.

7. Lee JS, Chu IS,MikaelyanA,CalvisiDF,Heo J, Reddy JK, et al. Applicationofcomparative functional genomics to identify best-fit mouse models tostudy human cancer. Nat Genet 2004;36:1306–11.

8. Lee JS, Chu IS, Heo J, Calvisi DF, Sun Z, Roskams T, et al. Classification andprediction of survival in hepatocellular carcinoma by gene expressionprofiling. Hepatology 2004;40:667–76.

9. 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. Nature Medicine 2006;12:410–6.

10. WooHG, Lee JH, Yoon JH, KimCY, Lee HS, Jang JJ, et al. Identification of acholangiocarcinoma-like gene expression trait in hepatocellular carcino-ma. Cancer Res 2010;70:3034–41.

11. Sohn BH, Shim JJ, Kim SB, Jang KY, Kim SM, Kim JH, et al. Inactivation ofhippo pathway is significantly associated with poor prognosis in hepato-cellular carcinoma. Clin Cancer Res 2016;22:1256–64.

12. WooHG, Park ES, Cheon JH, Kim JH, Lee JS, Park BJ, et al. Gene expression-based recurrence prediction of hepatitis B virus-related human hepatocel-lular carcinoma. Clinical Cancer Research 2008;14:2056–64.

13. Kim SM, Leem SH, Chu IS, Park YY, Kim SC, Kim SB, et al. Sixty-five gene-based risk score classifier predicts overall survival in hepatocellular carci-noma. Hepatology 2012;55:1443–52.

14. Santoni-Rugiu E, Nagy P, Jensen MR, Factor VM, Thorgeirsson SS.Evolution of neoplastic development in the liver of transgenic miceco-expressing c-myc and transforming growth factor-alpha. Am J Pathol1996;149:407–28.

15. Conner EA, Lemmer ER, Omori M, Wirth PJ, Factor VM, Thorgeirsson SS.Dual functions of E2F-1 in a transgenic mouse model of liver carcinogen-esis. Oncogene 2000;19:5054–62.

16. Conner EA, Lemmer ER, Sanchez A, Factor VM, Thorgeirsson SS. E2F1blocks and c-Myc accelerates hepatic ploidy in transgenic mouse models.Biochem Biophys Res Commun 2003;302:114–20.

17. Poirier LA. Hepatocarcinogenesis by diethylnitrosamine in rats fed highdietary levels of lipotropes. J Natl Cancer Inst 1975;54:137–40.

18. Held WA, Mullins JJ, Kuhn NJ, Gallagher JF, Gu GD, Gross KW. T antigenexpression and tumorigenesis in transgenicmice containing amousemajorurinary protein/SV40 T antigen hybrid gene. EMBO J 1989;8:183–91.

19. Lu L, Li Y, Kim SM, Bossuyt W, Liu P, Qiu Q, et al. Hippo signaling is apotent in vivo growth and tumor suppressor pathway in the mammalianliver. Proc Natl Acad Sci U S A 2010;107:1437–42.

20. Dong B, Lee JS, Park YY, Yang F, Xu G, Huang W, et al. Activating CAR andbeta-catenin induces uncontrolled liver growth and tumorigenesis.Nat Commun 2015;6:5944.

21. Harada N, Miyoshi H, Murai N, OshimaH, Tamai Y, OshimaM, et al. Lackof tumorigenesis in the mouse liver after adenovirus-mediated expressionof a dominant stable mutant of beta-catenin. Cancer Res 2002;62:1971–7.

22. Kaposi-Novak P, Lee JS, Gomez-Quiroz L, Coulouarn C, Factor VM,Thorgeirsson SS. Met-regulated expression signature defines a subset ofhuman hepatocellular carcinomas with poor prognosis and aggressivephenotype. J Clin Invest 2006;116:1582–95.

23. Thorgeirsson SS, Lee JS, Grisham JW. Functional genomics of hepatocel-lular carcinoma. Hepatology 2006;43:S145–50.

24. Yamashita T, Ji J, Budhu A, Forgues M, Yang W, Wang HY, et al. EpCAM-positive hepatocellular carcinoma cells are tumor-initiating cells withstem/progenitor cell features. Gastroenterology 2009;136:1012–24.

25. 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.

26. Cancer Genome Atlas Research Network. Comprehensive and integrativegenomic characterization of hepatocellular carcinoma. Cell 2017;169:1327–41.

27. Seiwert TY, Burtness B, Mehra R, Weiss J, Berger R, Eder JP, et al. Safety andclinical activity of pembrolizumab for treatment of recurrent or metastaticsquamous cell carcinoma of the head and neck (KEYNOTE-012): an open-label, multicentre, phase 1b trial. Lancet Oncol 2016;17:956–65.

28. Simon R, Lam A, Li MC, Ngan M, Menenzes S, Zhao Y. Analysis of geneexpression data using BRB-ArrayTools. Cancer Inform 2007;3:11–7.

29. EisenMB, SpellmanPT, BrownPO, BotsteinD.Cluster analysis anddisplayof genome-wide expression patterns. Proc Natl Acad Sci U S A 1998;95:14863–8.

30. Kim JH, Sohn BH, Lee HS, Kim SB, Yoo JE, Park YY, et al. Genomicpredictors for recurrence patterns of hepatocellular carcinoma: modelderivation and validation. PLoS Med 2014;11:e1001770.

Yim et al.





31. Yu FX, Zhao B, Guan KL. Hippo pathway in organ size control, tissuehomeostasis, and cancer. Cell 2015;163:811–28.

32. Calvisi DF, Factor VM, Ladu S, Conner EA, Thorgeirsson SS. Disruption ofbeta-catenin pathway or genomic instability define two distinct categoriesof liver cancer in transgenic mice. Gastroenterology 2004;126:1374–86.

33. Iroz A, Montagner A, Benhamed F, Levavasseur F, Polizzi A, Anthony E,et al. A specific ChREBP and PPARalpha cross-talk is required for theglucose-mediated. Cell Rep 2017;21:403–16.

34. Dittmer J, Rody A. Cancer stem cells in breast cancer. Histol Histopathol2013;28:827–38.

35. Imrich S, Hachmeister M, Gires O, et al. EpCAM and its potential role intumor-initiating cells. Cell Adh Migr 2012;6:30–8.

36. Pang RW, Poon RT. Cancer stem cell as a potential therapeutic target inhepatocellular carcinoma. Curr Cancer Drug Targets 2012;12:1081–94.

37. Cancer Genome Atlas Research Network. Comprehensive and integrativegenomic characterization of hepatocellular carcinoma. Cell 2017;169:1327–41.

38. Saha SK, Parachoniak CA,Ghanta KS, Fitamant J, Ross KN,NajemMS, et al.Mutant IDH inhibits HNF-4 alpha to block hepatocyte differentiation andpromote biliary cancer. Nature 2014;513:110–4.

39. Oikawa T, Kamiya A, Kakinuma S, Zeniya M, Nishinakamura R, Tajiri H,et al. Sall4 regulates cell fate decision in fetal hepatic stem/progenitor cells.Gastroenterology 2009;136:1000–11.

40. CleversH,Nusse R.Wnt/beta-catenin signaling anddisease. Cell 2012;149:1192–205.

41. HsuHC, Jeng YM,Mao TL, Chu JS, Lai PL, Peng SY. Beta-cateninmutationsare associatedwith a subset of low-stage hepatocellular carcinomanegativefor hepatitis B virus and with favorable prognosis. Am J Pathol 2000;157:763–70.

42. Benhamouche S, Decaens T, Godard C, Chambrey R, Rickman DS,Moinard C, et al. Apc tumor suppressor gene is the "zonation-keeper"of mouse liver. Dev Cell 2006;10:759–70.

43. Lu L, Li Y, Kim SM, Bossuyt W, Liu P, Qiu Q, et al. Hippo signaling is apotent in vivo growth and tumor suppressor pathway in the mammalianliver. Proc Nat Acad Sci USA 2010;107:1437–42.

44. Park YY, Sohn BH, Johnson RL, Kang MH, Kim SB, Shim JJ, et al. Yes-associated protein 1 and transcriptional coactivator with PDZ-binding

motif activate the mammalian target of rapamycin complex 1 pathwayby regulating amino acid transporters in hepatocellular carcinoma.Hepatology 2016;63:159–72.

45. Chan SW, Lim CJ, Chong YF, Pobbati AV, Huang C, Hong W. Hippopathway-independent restriction of TAZ and YAP by angiomotin.J Biol Chem 2011;286:7018–26.

46. Yimlamai D, Christodoulou C, Galli GG, Yanger K, Pepe-Mooney B,Gurung B, et al. Hippo pathway activity influences liver cell fate. Cell 2014;157:1324–38.

47. Goossens N, Sun X, Hoshida Y. Molecular classification of hepatocellularcarcinoma: potential therapeutic. Hepat Oncol 2015;2:371–9.

48. Mo JS, Park HW, Guan KL. The Hippo signaling pathway in stem cellbiology and cancer. EMBO Rep 2014;15:642–56.

49. Ali SH, DeCaprio JA. Cellular transformation by SV40 large T antigen:interaction with host proteins. Semin Cancer Biol 2001;11:15–23.

50. Nguyen HT, Hong X, Tan S, Chen Q, Chan L, Fivaz M, et al. Viralsmall T oncoproteins transform cells by alleviating hippo-pathway-mediated inhibition of the YAP proto-oncogene. Cell Rep 2014;8:707–13.

51. 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.

52. Jaworski M, Buchmann A, Bauer P, Riess O, Schwarz M. B-raf and Ha-rasmutations in chemically induced mouse liver tumors. Oncogene 2005;24:1290–5.

53. Rosenbluh J, Nijhawan D, Cox AG, Li X, Neal JT, Schafer EJ, et al. beta-Catenin-driven cancers require a YAP1 transcriptional complex for survivaland tumorigenesis. Cell 2012;151:1457–73.

54. Camargo FD, Gokhale S, Johnnidis JB, Fu D, Bell GW, Jaenisch R, et al.YAP1 increases organ size and expands undifferentiated progenitor cells.Curr Biol 2007;17:2054–60.

55. Garon EB, Rizvi NA, Hui R, Leighl N, Balmanoukian AS, Eder JP, et al.Pembrolizumab for the treatment of non-small-cell lung cancer. N Engl JMed 2015;372:2018–28.

56. Hodi FS,O'Day SJ,McDermott DF,Weber RW, Sosman JA,Haanen JB, et al.Improved survival with ipilimumab in patientswithmetastaticmelanoma.N Engl J Med 2010;363:711–23.






2018;16:1713-1723. Published OnlineFirst August 6, 2018.Mol Cancer Res Sun Young Yim, Jae-Jun Shim, Ji-Hyun Shin, et al. Clinical Resemblance to Human Liver CancerIntegrated Genomic Comparison of Mouse Models Reveals Their

Updated version

10.1158/1541-7786.MCR-18-0313doi:

Access the most recent version of this article at:

Material

Supplementary

http://mcr.aacrjournals.org/content/suppl/2018/08/04/1541-7786.MCR-18-0313.DC1

Access the most recent supplemental material at:

Cited articles

http://mcr.aacrjournals.org/content/16/11/1713.full#ref-list-1

This article cites 56 articles, 10 of which you can access for free at:

Citing articles

http://mcr.aacrjournals.org/content/16/11/1713.full#related-urls

This article has been cited by 1 HighWire-hosted articles. Access the articles at:

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

Subscriptions

Reprints and

[email protected]

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://mcr.aacrjournals.org/content/16/11/1713To request permission to re-use all or part of this article, use this link



http://mcr.aacrjournals.org/lookup/doi/10.1158/1541-7786.MCR-18-0313

http://mcr.aacrjournals.org/content/suppl/2018/08/04/1541-7786.MCR-18-0313.DC1

http://mcr.aacrjournals.org/content/16/11/1713.full#ref-list-1

http://mcr.aacrjournals.org/content/16/11/1713.full#related-urls

http://mcr.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://mcr.aacrjournals.org/content/16/11/1713


integratedgenomiccomparisonofmousemodels reveals their … · sun young yim1,2, jae-jun shim1,3,...

Documents