Loss of Hepatocyte Nuclear Factor 1� Function inHuman Hepatocellular Adenomas Leads to Aberrant

Activation of Signaling Pathways Involved inTumorigenesis

Laura Pelletier,1,2* Sandra Rebouissou,1,2* Alain Paris,3 Estelle Rathahao-Paris,3 Elisabeth Perdu,3

Paulette Bioulac-Sage,4,5 Sandrine Imbeaud,6 and Jessica Zucman-Rossi1,2

Hepatocellular adenomas (HCAs) are benign liver tumors that usually develop in womenwho are taking oral contraceptives. Among these tumors, biallelic inactivating mutations ofthe hepatocyte nuclear factor 1� (HNF1A) transcription factor have been frequently iden-tified and in rare cases of hepatocellular carcinomas developed in noncirrhotic liver. BecauseHNF1A meets the genetic criteria of a tumor suppressor gene, we aimed to elucidate thetumorigenic mechanisms related to HNF1� inactivation in hepatocytes. We searched forsignaling pathways aberrantly activated in human HNF1A-mutated HCA (H-HCA) using agenome-wide transcriptome analysis comparing five H-HCA with four normal livers. Wevalidated the main pathways by quantitative reverse transcription polymerase chain reaction(RT-PCR) and western blotting in a large series of samples. Then, we assessed the role ofHNF1� in the observed deregulations in hepatocellular cell models (HepG2 and Hep3B) bysilencing its endogenous expression using small interfering RNA. Along with the previouslydescribed induction of glycolysis and lipogenesis, H-HCA also displayed overexpression ofseveral genes encoding growth factor receptors, components of the translation machinery,cell cycle, and angiogenesis regulators, with, in particular, activation of the mammaliantarget of rapamycin (mTOR) pathway. Moreover, estradiol detoxification activities wereshut down, suggesting a hypersensitivity of H-HCA to estrogenic stimulation. In the cellmodel, inhibition of HNF1� recapitulated most of these identified transcriptional deregu-lations, demonstrating that they were related to HNF1� inhibition. Conclusion: H-HCAshowed a combination of alterations related to HNF1� inactivation that may cooperate topromote tumor development. Interestingly, mTOR appears as a potential new attractivetherapeutic target for treatment of this group of HCAs. (HEPATOLOGY 2010;51:557-566.)

Abbreviations: ACL, adenosine triphosphate citrate lyase; FAS, fatty acid synthase; HCA, hepatocellular adenomas; HCC, hepatocellular carcinoma; H-HCA, humanhepatocyte nuclear factor 1�-mutated hepatocellular adenoma; HNF1�, hepatocyte nuclear factor 1�; mTOR, mammalian target of rapamycin; RT-PCR, reversetranscription polymerase chain reaction; siRNA, small interfering RNA.

From 1Institut National de la Sante et de la Recherche Medicale (INSERM), U674, Genomique fonctionnelle des tumeurs solides, Paris, France; 2Universite ParisDescartes, Paris, France; 3INRA UMR 1089, Xenobiotiques, Toulouse, France; 4INSERM, U889, Universite Bordeaux 2, IFR66, Bordeaux, France; 5CHU Bordeaux,Hopital Pellegrin, Bordeaux, France; 6Array s/IMAGE, Genexpress, Functional Genomics and Systems Biology for Health—UMR 7091, CNRS, Université Paris, 6 Pierreet Marie Curie, Villejuif, France.

*These authors contributed equally to this work.Received June 12, 2009; accepted September 8, 2009.Address reprint requests to: Jessica Zucman-Rossi, INSERM, U674, “Functional Genomic of Solid Tumors,” 27 rue Juliette Dodu, 75010 Paris, France. E-mail:

zucman@cephb.fr.Copyright © 2009 by the American Association for the Study of Liver Diseases.Published online in Wiley InterScience (www.interscience.wiley.com).DOI 10.1002/hep.23362Supported by the INSERM (Reseau de Recherche clinique et en sante des populations) and SNFGE. L.P. and S.R. are supported by a doctoral fellowship from the

Ministere de la Recherche and the Ligue Nationale Contre le Cancer, respectively.Potential conflict of interest: Nothing to report.Additional Supporting Information may be found in the online version of this article.

557

Hepatocellular adenomas (HCAs) are rare benignliver tumors usually occurring in young womentaking oral contraceptives and occasionally as-

sociated with the use of anabolic corticosteroid or withtype I glycogen storage disease.1–3 Recently, HCA hasbeen described as a heterogeneous disease including atleast three main subtypes of tumors in which histologicalphenotypes are closely related to specific genetic alter-ations and clinical features.4–6 The most frequent groupof HCAs harbor biallelic inactivating mutations of hepa-tocyte nuclear factor 1� (HNF1A) gene and are pheno-typically characterized by marked steatosis.4,6,7 In 90% ofthe cases, HNF1�-mutated HCA (H-HCA) are sporadiclesions displaying somatic mutations. However, in rarefamilies with an inherited mutation in one allele ofHNF1A, maturity-onset diabetes of the young type 3 pa-tients are predisposed to develop familial liver adenoma-tosis that is defined by the presence of more than 10 HCAnodules in the liver, which is clinically difficult to man-age.7–10 Thus, HNF1A meets the genetic criteria of a tu-mor suppressor gene.7 However, liver adenomatosis is arare complication of maturity-onset diabetes of the youngtype 3, and its development may be promoted by het-erozygous germline mutation inactivating CYP1B1, a keymetabolism enzyme responsible for the formation of hy-droxylated and genotoxic metabolites of estrogen.11

HNF1� is an atypical homeodomain-containing pro-tein that was originally identified as a hepatocyte-specifictranscriptional regulator.12 In vivo and in vitro models ofHNF1� inactivation demonstrated that this transcriptionfactor plays an important role in hepatocyte differentia-tion and is also crucial for metabolic regulation and liverfunction.13–15 Interestingly, and consistent with the iden-tified tumor suppressor function of HNF1� in humans,hnf1�-deficient mice develop a dramatic liver enlarge-ment that has been associated with increased hepatocyteproliferation, sometimes accompanied by dyspla-sia.14,16,17 However, the role of HNF1� in the control ofcell proliferation or survival in hepatocytes is still poorlyunderstood.

To gain insight into the tumorigenic mechanisms re-lated to HNF1� inactivation, we searched for tumori-genic pathways aberrantly activated in human H-HCA.The role of HNF1� in the observed deregulations wassubsequently assessed in vitro by inhibiting its endoge-nous expression in human liver cancer cell lines by usingsmall interfering RNA.

Patients and Methods

Patients and Samples. Liver tissues were collected innine French surgery departments from 1992 to 2004.

They were immediately frozen in liquid nitrogen andstored at �80°C until used for molecular studies. Thewhole series of HCA used for the different molecular anal-yses included 23 H-HCA previously described4,6 and 17normal livers taken from patients resected with primaryliver tumors developed in the absence of cirrhosis (Sup-porting Table 1). All of the patients were recruited inaccordance with French law and institutional ethicalguidelines. The study was approved by the ethical com-mittee of Hopital Saint-Louis, Paris, France.

Microarray Analysis. Transcriptional profiling offive H-HCA and four normal liver tissues were performedusing Affymetrix oligonucleotide GeneChips HG-U133A (GEO accession number GSE7473), as previ-ously described.5,18

Cell Lines and Small Interfering RNA Transfec-tion. HepG2 and Hep3B cells were obtained from theAmerican Type Culture Collection (Manassas, VA) andwere cultured in Dulbecco’s modified eagle medium withhigh glucose (Invitrogen) supplemented with 10% fetalcalf serum, penicillin 100 IU/mL, and streptomycin 100�g/mL. Small interfering RNA (siRNA) transfectionswere performed according to the manufacturer’s proto-col, in six-well plates using the lipofectamine RNAiMaxreagent (Invitrogen) with two different sequences ofsiRNA duplexes targeting HNF1� (NM_000545) (Am-bion) as follows: GGUCUUCACCUCAGACACUtt(exon 8-9 3544) or GGCAGAAGAACCCUAGCAAtt(ex3 3450). Block-iT Alexa Fluor Red Fluorescent OligosiRNA (Invitrogen) was used as a double-stranded RNA-negative control. In most experiments, 10 nM of eachsiRNA was transfected in triplicate, except for dose–effectstudies, in which several siRNA concentrations weretested (0, 0.01, 0.05, 0.1, 0.2, 0.4, 0.6, 0.8, 1, 5, 10, and50 nM) to obtain different levels of HNF1� expression.Cells were prepared for quantitative reverse transcriptionpolymerase chain reaction (RT-PCR) and western blot-ting analyses either 3 or 7 days after transfection. Theabsence of cross-reaction of the HNF1�–siRNA duplexeswith the HNF1� sequence was checked by comparing theexpression level of HNF1� transcript in cells transfectedwith siRNA targeting HNF1� with the control siRNA-transfected cells (data not shown).

Quantitative RT-PCR. Quantitative RT-PCR wasperformed in duplicate as previously described19 by usingpredesigned primers and probe sets from Applied Biosys-tems (Supporting Table 2). Ribosomal 18S (R18S) wasused for the normalization of expression data, and the2���CT method was applied. The final results were ex-pressed as the fold differences in target gene expression intested samples compared with the mean expression value

558 PELLETIER, REBOUISSOU, ET AL. HEPATOLOGY, February 2010

of normal tissues (for tumor analysis) or with the controlsiRNA in cell lines.

Western Blotting. Western blot analyses were per-formed as described,18 using the primary antibodies spe-cific for ErbB2 (V-ERB-B2 avian erythroblastic leukemiaviral oncogene homolog 2), mammalian target of rapamy-cin (mTOR), phospho-mTOR ser2448, 4E-BP1, phos-pho 4E-BP1 (eukaryotic translation initiation factor 4E-binding protein 1) thr37/46 (Cell Signaling Technology,diluted 1:500, except for phospho-mTOR that was usedat 1:200 dilution); eIF4G3 (eukaryotic translation initia-tin factor 4-gamma, 3) (a gift of Dr. N. Sonenberg,1:500), eEF1A2 (eukaryotic translation elongation factor1, alpha-2) (Abnova, 1:500); HNF1� and cyclin D1(Santa Cruz Biotechnology, 1:100) and L-FABP (Liverfatty acid binding protein) (a gift of Dr. J. Gordon,1:2000). Polyclonal rabbit anti-actin (1:3000, Sigma) wasused as loading control.

Lipid Staining. HepG2 cells were grown on slidesfor 7 days and fixed with 4% formaldehyde in phosphate-buffered saline 1X. After washing, cells were stained withOil Red O (Sigma) dissolved in 60% isopropyl alcohol(20’), and counterstaining was performed with 50% he-matoxylin (Harris, Gill II, PAP1, Labonord).

Estrogen Metabolism. Human liver S9 fractionswere prepared from 100 to 135 mg frozen tissue. Thetissues were mechanically disrupted in an Elvehjem potterin 0.1 M phosphate buffer (pH 7.4), and the homoge-nates were subsequently subjected to centrifugation at600g for 10 minutes. The S9 fraction was obtained fromthe supernatant after centrifugation at 9000g for 20 min-utes, and the upper solid fatty material was discarded atthe same time. Then, S9 fractions were stored at �80°Cuntil analyses.

Estradiol-17� (�2�) 2-hydroxylase and 4-hydroxylaseactivities were determined from 1 mg S9 protein by quan-tifying 2- and 4-o-methyl-ether derivatives of E2�, usingradio–high-pressure liquid chromatography as previouslydescribed in Rizzati et al.20

�2� esterification activity was determined from 2 mgS9 protein, by measuring �2�-17-fatty acid esters forma-tion, as previously described in Paris and Rao.21 Detailedprocedures are provided in the Supporting Methods.

Statistical Analysis. All of the values reported aremeans � standard deviation. Statistical analyses were per-formed using GraphPad Prism version 4 software, andsignificance was determined using either the nonparamet-ric Mann-Whitney test for unpaired data or the two-tailed t test. Difference was considered significant at P �0.05. In all graphs, *, **, and *** indicate differencesbetween groups at P � 0.05, P � 0.01, and P � 0.001,respectively.

Results

Gene Expression Profiles in HNF1�-MutatedHCA. To study the role of HNF1� inactivation in hu-man liver tumorigenesis, we used oligonucleotide mi-croarrays to compare expression profiles between H-HCAand control normal liver tissues. Statistical analysis ofgene expression identified 338 and 212 genes that were,respectively significantly underexpressed and overex-pressed in H-HCA compared with normal liver samples(Supporting Table 3). Several classes of genes related tonormal hepatocyte function were altered in this subtypeof adenoma, including amino acid, glucose, lipid, xeno-biotic, and steroid hormone metabolism (Supporting Ta-ble 4). A large fraction of differentially expressed genesalso were involved in transcriptional regulation, and fewgenes encoded receptors, growth factors, or proteins im-plicated in translation and cell proliferation (SupportingTable 4). On examination of these data, we selected a setof deregulated genes that may be relevant for tumorigen-esis, and we further validated their expression using quan-titative RT-PCR and western blotting in a large series ofH-HCA and in hepatocellular cell lines inactivated forHNF1� by siRNA.

HNF1� Inhibition in Human Liver Cancer CellLines. HepG2 cells were transfected with a controlsiRNA and two siRNA targeting exon 3 or exon 8-9 junc-tion, leading to 87% and 98% HNF1A messenger RNA(mRNA) extinction, respectively (Fig. 1A). To modulatethe HNF1� expression and to study dose-dependent re-sponse of candidate genes putatively regulated byHNF1�, we used a range of different concentrations ofHNF1� siRNA (Fig. 1B). Indeed, there was a close cor-relation between the expression of HNF1A and differentgenes known to be directly targeted by HNF1�, as shownfor FGB in Fig. 1B. Using siRNA ex8-9, we observed anextinction of HNF1� protein as soon as 24 hours, withmaximal inhibition of HNF1� and its transactivated geneFABP1 72 hours after transfection (Fig. 1C) that wasmaintained for at least 7 days (Fig. 2). In all the followingexperiments, we used the most efficient siRNA (ex8-9).However, to verify that the observed deregulations wererelated to HNF1� inactivation, we confirmed the resultsusing the ex3 siRNA in HepG2 cells and ex8-9 siRNA ina second hepatocellular cell line (Hep3B), reaching 91%of HNF1A mRNA inhibition (Supporting Fig. 1 andother Supporting data).

Activation of Lipogenesis. Presence of a severe ste-atosis is the most striking phenotype constantly observedin H-HCA4,6 (Fig. 2D). Accordingly, we previously dem-onstrated that H-HCA displayed a strong alteration inglucose and lipid metabolism gene expression profiles,

HEPATOLOGY, Vol. 51, No. 2, 2010 PELLETIER, REBOUISSOU, ET AL. 559

Fig. 1. Inhibition of HNF1� expressionin HepG2 cells. (A) HepG2 cells were trans-fected independently with two different se-quences of siRNA directed against HNF1A(HNF1� ex8-9 and HNF1� ex3) (siH), orwith a control siRNA (siC). Inhibition effi-ciencies were assessed 3 days after trans-fection by measuring the expression level ofHNF1A and two of its transactivated genes(FABP1 and FGB) by quantitative RT-PCR(two-tailed t test). (B) Correlation betweenHNF1A and FGB mRNA expression 3 daysafter transfection of HepG2 cells with dif-ferent concentrations of HNF1� ex8-9siRNA (Spearman’s rank correlation test).(C) Time course analysis of HNF1� andLFABP protein expression after transfectionwith control (siC) and HNF1� ex8-9 (siH)siRNA using western blotting.

Fig. 2. Activation of lipogenesis. (A, B)Expression of glycolysis and lipogenesisgenes in H-HCA and HepG2 cells trans-fected with HNF1� ex8-9 siRNA after 3 (A)or 7 (B) days. mRNA levels were analyzedby quantitative RT-PCR and are expressedas an n-fold difference in gene expressionrelative to the mean expression value ofnormal liver tissues (NL, n � 10, mean �1 dotted line) for H-HCA (T, n � 16) (two-tailed Mann-Whitney test), or relative tocells transfected with control siRNA (siC)(mean � 1 dotted line) for HepG2 cellstransfected with HNF1� siRNA (siH) (two-tailed t test). NE: Not expressed in the cellline. (C) Overexpression of ACL and FASproteins in HepG2 cells transfected withHNF1� ex8-9 siRNA after 7 days comparedwith cells transfected with control siRNAanalyzed by western blotting. (D) Typicalaspect of a H-HCA with marked lipid over-load. (E) HepG2 cells transfected withHNF1� ex8-9 siRNA showing marked lipidoverload compared with HepG2 cells trans-fected with control siRNA, after Oil Red Ostaining and hematoxylin counterstain.

560 PELLETIER, REBOUISSOU, ET AL. HEPATOLOGY, February 2010

leading to an aberrant induction of glycolytic and lipo-genic genes expression with a repression of gluconeo-genesis that predict elevated rates of lipogenesis18

(Supporting Table 4; Fig. 2A, B). Moreover, we foundthat most of these gene deregulations were related toHNF1� inactivation in our cell model.

First, the four genes (PCK1, PCK2, FBP1, andG6PT1) involved in the gluconeogenesis pathway thatwere underexpressed in H-HCA were also significantlyrepressed after HNF1� extinction in HepG2 cells (Fig.2A). Among the five glycolytic genes altered in H-HCA,three of them (GCKR, PKLR, and PKM2) showed thesame profile of deregulation in HNF1�-inhibited HepG2cells. In contrast, glucose-6-phosphate isomerase (GPI)displayed an opposite regulation in HepG2 cells, and glu-cokinase was not expressed (Fig. 2A). All of the transcrip-tional modifications observed in both gluconeogenesisand glycolysis pathways occurred immediately afterHNF1� extinction and were maintained 7 days later, sug-gesting that their expression is directly controlled byHNF1� (Supporting Fig. 2B and Supporting Table 4).

Second, in the lipogenic pathway, 11 genes were sig-nificantly overexpressed in H-HCA compared with nor-mal livers; six of them were also significantly up-regulatedin HNF1�-inhibited HepG2 cells compared with controlsiRNA-transfected cells (Fig. 2B). This included threecrucial genes involved in the first steps of fatty acid syn-thesis: ACLY, ACACA, and FASN, encoding, respectively,the adenosine triphosphate citrate lyase (ACL), acetyl-CoA carboxylase, and fatty acid synthase (FAS) enzymes.ACLY and FASN overexpression was confirmed at theprotein level (Fig. 2C). Notably, these transcriptionalchanges were late events, because they were observed only7 days after HNF1� inhibition except for the two elon-gases (ELOVL1 and ELOVL5) that were already overex-pressed after 3 days (Supporting Fig. 2C and Supporting

Table 4). Furthermore, the overexpression of lipogenicgenes was correlated with lipid overload in HNF1�-in-hibited cells, which showed more and larger lipid vacuolesthan control siRNA-transfected cells (Fig. 2E). These re-sults suggested that several of these lipogenic genes areindirectly regulated by HNF1�.

Finally, in HNF1�-inactivated HepG2 as in H-HCA,we did not find overexpression of several transcriptionfactors known to regulate lipogenic genes expression; inparticular, SREBP1 was not overexpressed (Fig. 2B andSupporting Fig. 2A). Almost all of the results found after3 days of HNF1� extinction were confirmed in Hep3Bcells and in HepG2 cells using ex3 siRNA (SupportingFig. 2B and 2C and Supporting Table 4). However, lipo-genesis could not be studied at 7 days in Hep3B and inHepG2 cells using ex3 siRNA, because in these condi-tions, HNF1� suppression was not maintained.

Overexpression of Receptors, Growth Factors, andTranslational and Cell Cycle Regulators. In H-HCAsamples, our microarray analysis identified an overexpres-sion of several genes encoding critical proteins that areknown to be involved in a variety of tumors. Amongthem, ERBB2 transcript, which encodes a tyrosine kinasereceptor, was fivefold increased in H-HCA comparedwith normal livers. This overexpression was correlatedwith a strong accumulation of the corresponding proteinin tumors (Fig. 3A, B). This transcriptional overexpres-sion was not related to gene amplification, because noERBB2 gene amplification was found in seven H-HCAcases using comparative genomic hybridization based onsingle nucleotide polymorphism genotyping (data notshown). Both PDGFA and PDGFB mRNA, encodingtwo growth factors implicated in angiogenesis and cellproliferation, were also significantly increased in H-HCAcompared with normal livers (Fig. 3A). Moreover, weidentified up-regulation of several members of the trans-

Fig. 3. Overexpression of receptors,growth factors, and translational and cellcycle regulators in H-HCA. (A) QuantitativeRT-PCR validation of gene array expressiondata comparing H-HCA (T; n � 16) withnormal liver tissues (NL; n � 10) (two-tailed Mann-Whitney test). (B) Westernblotting validation of quantitative RT-PCRresults comparing H-HCA (n � 6) with nor-mal livers (n � 4).

HEPATOLOGY, Vol. 51, No. 2, 2010 PELLETIER, REBOUISSOU, ET AL. 561

lational machinery. In particular, mTOR, which isknown to play a pivotal role in the regulation of transla-tion, was overexpressed in H-HCA at both transcriptional(4.5-fold) and protein level (Fig. 3A, B). Accordingly,compared with normal livers, phospho-mTOR was sig-nificantly elevated in H-HCA, as well as 4E-BP1 phos-phorylation, which is a well-known downstream targetphosphorylated by mTOR (Fig. 3B). These results indi-cated that mTOR pathway was aberrantly activated inH-HCA. In addition, these tumors displayed an elevationof eIF-4G3 and eEF1A2 at mRNA and protein levels, thatare involved in protein translation initiation and elonga-tion, respectively (Fig. 3A, B). Finally, H-HCA showedoverexpression of CCND1 mRNA (3.6-fold), which en-codes cyclin D1, a key cell cycle regulatory protein (Fig.3A). This result was confirmed by western blotting anal-ysis, demonstrating a constant accumulation of cyclin D1in H-HCA (Fig. 3B).

Except for mTOR and ERBB2, significant overexpres-sion of five other aforementioned genes (EIF4G3,EEF1A2, PDGFA, PDGFB, and CCND1) was also found

in HepG2 cells 3 and 7 days after HNF1� inhibitioncompared with control-siRNA-transfected cells (Fig. 4Aand Supporting Fig. 3A). These data display a similarpattern of deregulation using HNF1� siRNA targetingexon 3 or in Hep3B cells (Supporting Fig. 3B-D). Re-markably, using the siRNA concentration range, we iden-tified a strong negative correlation between the expressionof HNF1� mRNA and mRNA expression of these fivegenes (Fig. 4B). Moreover, we confirmed that cyclin D1and eIF-4G3 were closely correlated to HNF1� expres-sion even at the protein level (Fig. 4C).

Alteration of Estrogen Metabolism. Interestingly,the pattern of deregulation observed in genes encodingxenobiotic and steroid hormone-metabolizing en-zymes, suggested that H-HCA may have impaired abil-ity to inactivate the active form of estradiol(Supporting Table 4). Indeed, in these tumors, we con-firmed the dramatic decrease in mRNA level ofCYP1A1, CYP1A2, and CYP3A4, which encodes thethree principal cytochrome P450 isoforms catalyzingE2� oxidation in human liver (Fig. 5A). In addition,

Fig. 4. Overexpression of receptors, growth factors, and translational and cell cycle regulators in HNF1�-inhibited HepG2 cells. (A) MessengerRNA expression levels were compared between HepG2 cells transfected with HNF1� ex8-9 siRNA (siH) and with control siRNA (siC) (two-tailed t test).(B) Correlations between expression of HNF1A mRNA, and the five genes significantly overexpressed in HNF1� inhibited HEPG2 cells were analyzedusing a range of siRNA concentrations, and significance was assessed by Spearman’s rank correlation test. All graphs plot quantitative RT-PCR resultsrelative to cells transfected with control siRNA. (C) Cyclin D1 and eIF-4G3 expression was analyzed using western blotting after transfection withdifferent concentrations of HNF1� ex8-9 siRNA. All analyses were performed 3 days after transfection.

562 PELLETIER, REBOUISSOU, ET AL. HEPATOLOGY, February 2010

compared with normal livers, H-HCA showed a two-fold decrease in HSD17B2 transcript (Fig. 5A) encod-ing the 17�-hydroxysteroid dehydrogenase, which isresponsible for the interconversion of active E2� to theless potent E1. CYP1A1 and HSD17B2 were also down-regulated by HNF1� inactivation in HepG2 andHep3B cells (Fig. 5B and Supporting Fig. 4), whereasCYP1A2 and CYP3A4 were not expressed in these celllines at baseline. According to the observed strongdown-regulation of cytochrome P450 isoforms, activ-ities of 2-hydroxylation and 4-hydroxylation of E2�were suppressed in H-HCA samples as compared withnormal livers (Fig. 5C). Furthermore, in these tumors,a significant threefold increase in activity of E2�-17-acylation to long chain fatty acid was found comparedwith normal livers, predicting elevated concentrationsof estrogen fatty acid esters which are known to behighly potent and long-lived estrogens22 (Fig. 5C).Collectively, these results support the hypothesis thatstimulation of growth by estrogen would be enhancedin H-HCA tissue.

DiscussionBiallelic inactivating mutations of HNF1A have been

identified in 35% to 50% of HCA and in rare cases ofhepatocellular carcinomas (HCC) developed in the ab-sence of cirrhosis.4,6,7 Hence, this genetic event appears tobe an important driver of HCA development and mayrepresent an early step in malignant transformation ofhepatocytes. Using a genome-wide transcriptome analy-sis, we identified altered expression in several genes thatmay contribute to the pathogenesis of H-HCA (Fig. 6).Moreover, we demonstrated that a number of these tran-scriptional changes were related to the loss of HNF1�activity in hepatocellular cells.

A role for HNF1� in the control of liver metabolismhas been well documented. Mice deficient in hnf1� de-velop severe liver steatosis that is very similar to that ob-served in H-HCA. As previously shown, humanadenomas display coordinated aberrant overexpression ofglycolytic and lipogenic genes and inhibition of glucone-ogenesis, which may be responsible for elevated rates oflipid synthesis and hepatocyte fat overload.18 Interest-ingly, our cellular model of HNF1� inactivation recapit-ulates most of the glycolysis and lipogenesis genederegulations and the fat overload observed in H-HCA.In the gluconeogenic and glycolytic pathways, almost allof the down-regulated genes were previously shown astargets or putative targets positively regulated by HNF1�,and the expression was suppressed immediately afterHNF1� extinction. In contrast, most of the significantoverexpressions observed in the lipogenic pathway aroselater, suggesting that they are an indirect consequence ofHNF1� inhibition. These observations suggest a repres-sor function of HNF1� that remains to be demonstratedon the corresponding promoters. Surprisingly, as shownin H-HCA, none of the well-known transcription factorsinvolved in the control of lipogenesis was activated in ourcell model. Our results showed an obvious role of HNF1�in the negative control of lipogenesis that may involve acomplex regulatory network whose understanding needsfurther investigation.

An emerging view suggests that metabolic reprogram-ming commonly observed in tumor cells is required forsurvival and to satisfy increased energetic and anabolicneeds. Particularly, high glycolysis and lipogenesis rates,as shown in H-HCA, are a hallmark of many humantumors, and recent findings demonstrated that these met-abolic adaptations function downstream from oncogenicsignalings.23 Moreover, several works showed that inhibi-

Fig. 5. Impaired estradiol metabolism. (A, B) Expression of estradiol detoxification-related genes was validated using quantitative RT-PCR. (A)H-HCA (T; n � 16) were compared with normal liver tissues (NL; n � 10) (two-tailed Mann-Whitney test) and (B) HepG2 cells transfected with siRNAHNF1� ex8-9 (siH) were compared with cells transfected with control siRNA (siC) 3 days after transfection (two-tailed t test). (C) Activities of E2�-2hydroxylation, E2�-4-hydroxylation, and E2�-17-acylation were compared between H-HCA (n � 3) and normal liver tissues (n � 3) (one-tailed t test).

HEPATOLOGY, Vol. 51, No. 2, 2010 PELLETIER, REBOUISSOU, ET AL. 563

tion of lipogenesis is able to reduce cell proliferation andtumor growth in vitro and in human tumor xenograft innude mice.24–26

Importantly, we also identified an aberrant activationof the mTOR signaling in H-HCA. mTOR has beenrecognized as a key regulator of cell proliferation andgrowth, in particular through its ability to stimulate pro-tein translation by phosphorylating S6K and 4E-BP1 pro-teins, leading to selective enhancement in synthesis of asubset of key growth-promoting proteins.27 The abnor-mal activation of the mTOR pathway has been implicatedin a wide variety of tumors.28 Interestingly, in the well-differentiated hepatocellular cell line, HepaRG, increasedmTOR activity was demonstrated to confer a neoplasticphenotype to the hepatocytes by altering the translationof genes necessary for moderating hepatocellular growthand establishing normal hepatic energy homeostasis.29

Moreover, several recent studies identified activation ofthe mTOR pathway in 15% to 50% of HCC, and showedantineoplastic activity of mTOR inhibitors in experimen-tal models of HCC.30–33 Collectively, these findings pin-point a major role for mTOR signaling in the control ofhepatocyte growth and survival. We showed herein thatmTOR pathway activation in H-HCA may occur

through elevated mRNA and protein levels of mTOR aswell as the strong accumulation of the ErbB2 receptor, aspreviously demonstrated in breast cancer overexpressingErbB2.34 In addition, the identified up-regulation ofcomponents of the translational machinery workingdownstream of mTOR, such as the initiation factor eIF-4G3 and the elongation factor eEF1A2, also may contrib-ute to maintaining high levels of protein synthesis inH-HCA, thereby promoting the accumulation of criticalfactors necessary for cell growth and survival. Remark-ably, eEF1A2 has been shown to have transforming activ-ity.35 Recently, frequent EEF1A2 overexpression has beenidentified in HCC, and gene silencing was shown to re-duce cell proliferation and increase apoptosis rates inHepG2 and Hep3B cell lines.36,37

In addition, a loss of critical cell cycle checkpointthrough the abnormal elevated levels of cyclin D1 wasobserved in H-HCA and the increased levels of PDGFAand particularly PDGFB may promote tumor neovascu-larization.38 Although mTOR and ERBB2 overexpres-sion were recurrently observed in H-HCA, thesederegulations were not found after in vitro silencing ofHNF1�. This result may reflect a more complex regula-tion of these two genes involving cooperation of HNF1�

mTOR

ErbB

2E

rbB2

TSC2

TSC1

AKT

PI3K

Rheb

4E-BP1 eIF4G3

eIF4E

S6K

eIF4B

eEF1A2Translation initiation Translation elongation

AMPK

LKB1

Lipogenesis

PDK1

Metformin

RapamycinCCI779RAD001AP23573

PTEN

AminoAcids

G1S

G2 M

Cyclin D1

Cellcycle

PDGFA PDGFB

Overexpression

Underexpression/Inactivation

17 -estradiol (E2)

estrone (E1)

HSD17B2

CYP1A1CYP1A2

CYP3A4

2/4-OH-estradiol

Estrogen metabolism

Acylation

E2-fatty acid

Activates

Inhibits

Glycolysis

Synthesis of proteins involved in cell growth and survival

ATP Nucleotids

Amino acids

Angiogenesis

Cell division

Proliferation

++ E2

Membrane synthesis Lipid modification of

proteins

Fig. 6. Scheme of the different pathways altered in human H-HCA and hypothetical consequences. In adenoma, loss of HNF1� activity leads tomany alterations that may shift cellular activities toward a proliferative state. Indeed, in cooperation with loss of cell cycle control, metabolicreprogramming resulting from elevated levels of glycolysis, and lipogenesis may fuel cell proliferation by providing energy and precursors required forthe synthesis of macromolecules. Furthermore, activation of the mTOR pathway and the translational machinery may promote the accumulation ofcritical proteins necessary for cell growth and survival. Overexpression of proangiogenic factors also may contribute to tumor development bystimulating neovascularization. Finally, suppression of estradiol detoxification activities may lead to estradiol accumulation and growth advantage.Some of these pathways can be targeted by specific pharmacological inhibitors (indicated in italic bold font), such as metformin, which inhibits bothlipogenesis and mTOR pathways, or rapamycin and its derivatives that target mTOR.

564 PELLETIER, REBOUISSOU, ET AL. HEPATOLOGY, February 2010

with other factors that may be altered in our cellular mod-els. In contrast, the strong negative correlation betweenthe expression level of HNF1� and the other aforemen-tioned genes suggests that HNF1� plays an importantrole in directing their basal promoter activity and maydirectly repress their expression in physiological condi-tions.

As for other tumor suppressor genes involved in hu-man benign tumorigenesis, understanding why HNF1�inactivation promotes formation of benign tumors ratherthan malignant tumors remains a difficult challenge. In-deed, it is surprising that many molecular deregulationsidentified in the benign H-HCA are also commonlyfound in malignancies. The effect of any mutation and itsimportance in tumorigenesis are likely to vary, dependingon when and where it occurs and has its effects. One canspeculate that in HNF1�-inactivated hepatocytes, feed-back controls of the cell population are still functional andmay add stability to the system, rendering exponentialexpansion and therefore malignant transformation im-possible unless these regulatory mechanisms are alteredduring progression.

The existence of specific pathways involved in the de-velopment of benign tumors is not clear. However, it isinteresting to note that, as identified in H-HCA, aberrantactivation of the mTOR pathway resulting from the in-activation of many tumor-suppressor genes has been com-monly found in several other benign lesions characterizedby cell growth abnormalities such as hamartomas syn-dromes, as well as cardiac hypertrophy.28

The last point of the current study was to investigatethe link between estrogen and H-HCA development.Several lines of evidence coming from both animal mod-els and human epidemiological studies indicated that es-trogens may act as liver tumor promoters.1,39,40 Inhumans, the relationship between HCA and oral contra-ceptives has been well established, and several reports havedescribed tumor regression after the withdrawal of hor-monal agents.41,42 In addition, some cases of HCA en-largement have been reported during pregnancy,presumably as a result of estrogen stimulation.43 Despitesome discrepancy between the studies, overall, the weightof evidence suggested that estrogen receptors are presentin HCA.44 In our series of H-HCA, we also found detect-able levels of ER� receptors using western blotting, sug-gesting that these tumors may respond to estrogenstimulation (data not shown). Our findings suggestedthat H-HCA have a severely impaired ability to eliminatethe active form of estradiol and may present an increasedcapacity to generate E2�-17-fatty acid esters that are po-tent long-acting estrogens. The ability to increase E2�-17-acylation may be a direct consequence of fatty acids

accumulation in these adenomas. Hence, estrogens mayconfer growth advantage to the HNF1�-inactivatedhepatocytes, as a result of an abnormal prolonged time ofhormonal exposure.

In conclusion, this study provides new insights into thetumorigenic mechanisms related to HNF1� inactivationin the liver. In H-HCA, we identified alterations that maycooperate to reorganize cellular activities supportingbioenergetics, macromolecular synthesis, and ultimatelycell division (Fig. 6). Although the exact role of HNF1�in the control of these processes remains to be more accu-rately studied, our results suggest that this transcriptionfactor may act as a transcriptional repressor. Importantly,our findings give a rationale for targeting the mTORpathway in H-HCA that may be useful for patients withliver adenomatosis or multiple HCA, when surgical resec-tion or liver transplantation is precluded.

Acknowledgment. The authors thank Dr. B. Brad-ford for critical reading of this manuscript. We thank allthe participants to the GENTHEP (Groupe d’etude Ge-netique des Tumeurs Hepatiques) network. We are grate-ful to Emmanuelle Jeannot and Lucille Mellottee for theirhelp in mutation screening. We thank Dr J. Gordon andDr. N. Sonenberg for providing us, respectively, the L-FABP and eIF-4G3 antibodies.

References1. Edmondson HA, Henderson B, Benton B. Liver-cell adenomas associated

with use of oral contraceptives. N Engl J Med 1976;294:470-472.2. Howell RR, Stevenson RE, Ben-Menachem Y, Phyliky RL, Berry DH.

Hepatic adenomata with type 1 glycogen storage disease. JAMA 1976;236:1481-1484.

3. Sale GE, Lerner KG. Multiple tumors after androgen therapy. Arch PatholLab Med 1977;101:600-603.

4. Bioulac-Sage P, Rebouissou S, Thomas C, Blanc JF, Saric J, Sa Cunha A,et al. Hepatocellular adenoma subtype classification using molecular mark-ers and immunohistochemistry. HEPATOLOGY 2007;46:740-748.

5. Rebouissou S, Amessou M, Couchy G, Poussin K, Imbeaud S, Pilati C, etal. Frequent in-frame somatic deletions activate gp130 in inflammatoryhepatocellular tumours. Nature 2009;457:200-204.

6. Zucman-Rossi J, Jeannot E, Nhieu JT, Scoazec JY, Guettier C, Rebouissou S,et al. Genotype-phenotype correlation in hepatocellular adenoma: new classi-fication and relationship with HCC. HEPATOLOGY 2006;43:515-524.

7. Bluteau O, Jeannot E, Bioulac-Sage P, Marques JM, Blanc JF, Bui H, et al.Bi-allelic inactivation of TCF1 in hepatic adenomas. Nat Genet 2002;32:312-315.

8. Bacq Y, Jacquemin E, Balabaud C, Jeannot E, Scotto B, Branchereau S, etal. Familial liver adenomatosis associated with hepatocyte nuclear factor1alpha inactivation. Gastroenterology 2003;125:1470-1475.

9. Flejou JF, Barge J, Menu Y, Degott C, Bismuth H, Potet F, et al. Liveradenomatosis: an entity distinct from liver adenoma? Gastroenterology1985;89:1132-1138.

10. Reznik Y, Dao T, Coutant R, Chiche L, Jeannot E, Clauin S, et al. Hepa-tocyte nuclear factor-1 alpha gene inactivation: cosegregation between liveradenomatosis and diabetes phenotypes in two maturity-onset diabetes ofthe young (MODY)3 families. J Clin Endocrinol Metab 2004;89:1476-1480.

11. Jeannot E, Poussin K, Chiche L, Bacq Y, Sturm N, Scoazec JY, et al.Association of CYP1B1 germ line mutations with hepatocyte nuclear fac-

HEPATOLOGY, Vol. 51, No. 2, 2010 PELLETIER, REBOUISSOU, ET AL. 565

tor 1alpha-mutated hepatocellular adenoma. Cancer Res 2007;67:2611-2616.

12. Courtois G, Morgan JG, Campbell LA, Fourel G, Crabtree GR. Interac-tion of a liver-specific nuclear factor with the fibrinogen and alpha 1-anti-trypsin promoters. Science 1987;238:688-692.

13. Odom DT, Zizlsperger N, Gordon DB, Bell GW, Rinaldi NJ, MurrayHL, et al. Control of pancreas and liver gene expression by HNF transcrip-tion factors. Science 2004;303:1378-1381.

14. Pontoglio M, Barra J, Hadchouel M, Doyen A, Kress C, Bach JP, et al.Hepatocyte nuclear factor 1 inactivation results in hepatic dysfunction,phenylketonuria, and renal Fanconi syndrome. Cell 1996;84:575-585.

15. Shih DQ, Bussen M, Sehayek E, Ananthanarayanan M, Shneider BL,Suchy FJ, et al. Hepatocyte nuclear factor-1alpha is an essential regulator ofbile acid and plasma cholesterol metabolism. Nat Genet 2001;27:375-382.

16. Lee YH, Sauer B, Gonzalez FJ. Laron dwarfism and non-insulin-depen-dent diabetes mellitus in the Hnf-1alpha knockout mouse. Mol Cell Biol1998;18:3059-3068.

17. Servitja JM, Pignatelli M, Maestro MA, Cardalda C, Boj SF, Lozano J, etal. Hnf1{alpha} (MODY3) controls tissue-specific transcriptional pro-grams and exerts opposed effects on cell growth in pancreatic islets andliver. Mol Cell Biol 2009;29:2945-2959.

18. Rebouissou S, Imbeaud S, Balabaud C, Boulanger V, Bertrand-Michel J,Terce F, et al. HNF1alpha inactivation promotes lipogenesis in humanhepatocellular adenoma independently of SREBP-1 and carbohydrate-re-sponse element-binding protein (ChREBP) activation. J Biol Chem 2007;282:14437–14446.

19. Rebouissou S, Vasiliu V, Thomas C, Bellanne-Chantelot C, Bui H, Chre-tien Y, et al. Germline hepatocyte nuclear factor 1alpha and 1beta muta-tions in renal cell carcinomas. Hum Mol Genet 2005;14:603-614.

20. Rizzati V, Rathahao E, Gamet-Payrastre L, Delous G, Jouanin I, GueraudF, et al. In vitro aromatic bioactivation of the weak estrogen E(2)alpha andgenesis of DNA adducts. Steroids 2005;70:161-172.

21. Paris A, Rao D. Biosynthesis of estradiol-17 beta fatty acyl esters by mi-crosomes derived from bovine liver and adrenals. J Steroid Biochem 1989;33:465-472.

22. Hochberg RB. Biological esterification of steroids. Endocr Rev 1998;19:331-348.

23. Swinnen JV, Brusselmans K, Verhoeven G. Increased lipogenesis in cancercells: new players, novel targets. Curr Opin Clin Nutr Metab Care 2006;9:358-365.

24. Gabrielson EW, Pinn ML, Testa JR, Kuhajda FP. Increased fatty acidsynthase is a therapeutic target in mesothelioma. Clin Cancer Res 2001;7:153-157.

25. Knowles LM, Axelrod F, Browne CD, Smith JW. A fatty acid synthaseblockade induces tumor cell-cycle arrest by down-regulating Skp2. J BiolChem 2004;279:30540–30545.

26. Kridel SJ, Axelrod F, Rozenkrantz N, Smith JW. Orlistat is a novel inhib-itor of fatty acid synthase with antitumor activity. Cancer Res 2004;64:2070-2075.

27. Hay N, Sonenberg N. Upstream and downstream of mTOR. Genes Dev2004;18:1926-1945.

28. Inoki K, Corradetti MN, Guan KL. Dysregulation of the TSC-mTORpathway in human disease. Nat Genet 2005;37:19-24.

29. Parent R, Kolippakkam D, Booth G, Beretta L. Mammalian target ofrapamycin activation impairs hepatocytic differentiation and targets genesmoderating lipid homeostasis and hepatocellular growth. Cancer Res2007;67:4337-4345.

30. Huynh H, Chow KP, Soo KC, Toh HC, Choo SP, Foo KF, et al. RAD001(everolimus) inhibits tumor growth in xenograft models of human hepa-tocellular carcinoma. J Cell Mol Med 2008;13:1371-1380.

31. Sahin F, Kannangai R, Adegbola O, Wang J, Su G, Torbenson M. mTORand P70 S6 kinase expression in primary liver neoplasms. Clin Cancer Res2004;10:8421-8425.

32. Semela D, Piguet AC, Kolev M, Schmitter K, Hlushchuk R, Djonov V, etal. Vascular remodeling and antitumoral effects of mTOR inhibition in arat model of hepatocellular carcinoma. J Hepatol 2007;46:840-848.

33. Villanueva A, Chiang DY, Newell P, Peix J, Thung S, Alsinet C, et al.Pivotal role of mTOR signaling in hepatocellular carcinoma. Gastroenter-ology 2008;135:1972-1983, 1983 e1971–1911.

34. Zhou X, Tan M, Stone Hawthorne V, Klos KS, Lan KH, Yang Y, et al.Activation of the Akt/mammalian target of rapamycin/4E-BP1 pathway byErbB2 overexpression predicts tumor progression in breast cancers. ClinCancer Res 2004;10:6779-6788.

35. Anand N, Murthy S, Amann G, Wernick M, Porter LA, Cukier IH, et al.Protein elongation factor EEF1A2 is a putative oncogene in ovarian cancer.Nat Genet 2002;31:301-305.

36. Boyault S, Rickman DS, de Reynies A, Balabaud C, Rebouissou S, JeannotE, et al. Transcriptome classification of HCC is related to gene alterationsand to new therapeutic targets. HEPATOLOGY 2007;45:42-52.

37. Schlaeger C, Longerich T, Schiller C, Bewerunge P, Mehrabi A, Toedt G,et al. Etiology-dependent molecular mechanisms in human hepatocarci-nogenesis. HEPATOLOGY 2008;47:511-520.

38. Battegay EJ, Rupp J, Iruela-Arispe L, Sage EH, Pech M. PDGF-BB mod-ulates endothelial proliferation and angiogenesis in vitro via PDGF beta-receptors. J Cell Biol 1994;125:917-928.

39. Coe JE, Ishak KG, Ross MJ. Estrogen induction of hepatocellular carcino-mas in Armenian hamsters. HEPATOLOGY 1990;11:570-577.

40. Yager JD, Jr., Yager R. Oral contraceptive steroids as promoters of hepa-tocarcinogenesis in female Sprague-Dawley rats. Cancer Res 1980;40:3680-3685.

41. Buhler H, Pirovino M, Akobiantz A, Altorfer J, Weitzel M, Maranta E, etal. Regression of liver cell adenoma: a follow-up study of three consecutivepatients after discontinuation of oral contraceptive use. Gastroenterology1982;82:775-782.

42. Steinbrecher UP, Lisbona R, Huang SN, Mishkin S. Complete regressionof hepatocellular adenoma after withdrawal of oral contraceptives. Dig DisSci 1981;26:1045-1050.

43. Kent DR, Nissen ED, Nissen SE, Ziehm DJ. Effect of pregnancy on livertumor associated with oral contraceptives. Obstet Gynecol 1978;51:148-151.

44. Torbenson M, Lee JH, Choti M, Gage W, Abraham SC, Montgomery E,et al. Hepatic adenomas: analysis of sex steroid receptor status and the Wntsignaling pathway. Mod Pathol 2002;15:189-196.

566 PELLETIER, REBOUISSOU, ET AL. HEPATOLOGY, February 2010