doi: 10.1136/gut.2008.152652 2009 58: 679-687 originally published online January 9, 2009Gut

 D F Calvisi, F Pinna, S Ladu, et al. sustains ERK activity in human HCCsusceptibility to hepatocarcinogenesis and Forkhead box M1B is a determinant of rat

http://gut.bmj.com/content/58/5/679.full.htmlUpdated information and services can be found at:

These include:

References http://gut.bmj.com/content/58/5/679.full.html#ref-list-1

This article cites 47 articles, 19 of which can be accessed free at:

serviceEmail alerting

box at the top right corner of the online article.Receive free email alerts when new articles cite this article. Sign up in the

Topic collections

(790 articles)Hepatic cancer   � Articles on similar topics can be found in the following collections

Notes

http://gut.bmj.com/cgi/reprintformTo order reprints of this article go to:

http://gut.bmj.com/subscriptions go to: GutTo subscribe to

group.bmj.com on August 18, 2010 - Published by gut.bmj.comDownloaded from

Forkhead box M1B is a determinant of ratsusceptibility to hepatocarcinogenesis and sustainsERK activity in human HCC

D F Calvisi, F Pinna, S Ladu, R Pellegrino, M M Simile, M Frau, M R De Miglio,M L Tomasi, V Sanna, M R Muroni, F Feo, R M Pascale

c Additional tables and figuresare published online only athttp://gut.bmj.com/content/vol58/issue5

Department of BiomedicalSciences, Division ofExperimental Pathology andOncology, University of Sassari,Sassari, Italy

Correspondence to:Fr F Feo, Dipartimento di ScienzeBiomediche, Sezione di PatologiaSperimentale e Oncologia,Universita di Sassari, Via P.Manzella 4, 07100 Sassari, Italy;feo@uniss.it

Revised 14 November 2008Accepted 9 December 2008Published online first6 January 2009

ABSTRACTBackground and aim: Previous studies indicate unrest-rained cell cycle progression in liver lesions fromhepatocarcinogenesis-susceptible Fisher 344 (F344) ratsand a block of G1–S transition in corresponding lesionsfrom resistant Brown Norway (BN) rats. Here, the role ofthe Forkhead box M1B (FOXM1) gene during hepatocar-cinogenesis in both rat models and human hepatocellularcarcinoma (HCC) was assessed.Methods and results: Levels of FOXM1 and its targetswere determined by immunoprecipitation and real-timePCR analyses in rat and human samples. FOXM1 functionwas investigated by either FOXM1 silencing or over-expression in human HCC cell lines. Activation of FOXM1and its targets (Aurora Kinose A, Cdc2, cyclin B1, Nek2)occurred earlier and was most pronounced in liver lesionsfrom F344 than BN rats, leading to the highest number ofCdc2–cyclin B1 complexes (implying the highest G2–Mtransition) in F344 rats. In human HCC, the level ofFOXM1 progressively increased from surrounding non-tumorous livers to HCC, reaching the highest levels intumours with poorer prognosis (as defined by patients’length of survival). Furthermore, expression levels ofFOXM1 directly correlated with the proliferation index,genomic instability rate and microvessel density, andinversely with apoptosis. FOXM1 upregulation was due toextracellular signal-regulated kinase (ERK) and glioblas-toma-associated oncogene 1 (GLI1) combined activity,and its overexpression resulted in increased proliferationand angiogenesis and reduced apoptosis in human HCCcell lines. Conversely, FOXM1 suppression led todecreased ERK activity, reduced proliferation and angio-genesis, and massive apoptosis of human HCC cell lines.Conclusions: FOXM1 upregulation is associated with theacquisition of a susceptible phenotype in rats andinfluences human HCC development and prognosis.

Human hepatocellular carcinoma (HCC) is one ofthe most common tumours worldwide, accountingfor ,500 000 deaths annually.1–3 Better under-standing of the molecular mechanisms underlyinghepatocarcinogenesis may hasten the identificationof novel molecular markers for HCC progressionand the development of new diagnostic andtherapeutic strategies.

Studies on rodent hepatocarcinogenesis demon-strated a polygenic predisposition to HCC.4

Numerous hepatocarcinogenesis susceptibility andresistance loci control rodent HCC development,suggesting a genetic predisposition with a majorlocus and various low-penetrance genes, at play indifferent subsets of population.4 This genetic

model is in keeping with human HCC epidemiol-ogy.5 Analysis of the effector mechanisms ofsusceptibility genes indicates that early preneo-plastic liver lesions, induced by chemical carcino-gens, grow and progress autonomously to HCConly in susceptible but not in resistant rat strains.4

Thus, the comparative evaluation of molecularalterations during hepatocarcinogenesis in ratssusceptible or resistant to this disease may helpin elucidating the mechanisms responsible forhuman liver malignant transformation. We pre-viously showed overexpression of c-myc, cyclin D1,cyclin E, cyclin A and E2f1 genes associated withpRb (retinoblastoma protein) hyperphosphoryla-tion in neoplastic nodules and HCCs developed inFisher 344 (F344) rats, susceptible to hepatocarci-nogenesis, but not in corresponding lesions fromresistant Brown Norway (BN) rats.4 6 Theseobservations imply a deregulation of G1 and Sphases in liver lesions of susceptible rats, and ablock of G1–S transition in the lesions of theresistant strain, explaining their limited ability toprogress. A similar deregulation in cell cycleproteins occurs in human and murine hepatocarci-nogenesis, implying the existence of similar mole-cular mechanisms of hepatocarcinogenesis acrossspecies.2–4 6 7 Recent data indicate that unrestrainedcell cycle progression may depend on activation ofthe Ras–mitogen-activated protein kinase (MAPK)cascade in rodent8 and human HCC.9 Indeed,higher expression of extracellular signal-regulatedkinase (ERK) was detected in neoplastic nodulesand HCC of susceptible F344 rats, whereas ERKwas slightly induced in corresponding lesions fromthe resistant BN strain.8 Similarly, ERK activitywas elevated in human HCC, and highest intumours with a poor prognosis, implying its crucialrole in tumour progression.9 In particular, wefound that ERK achieves unrestrained activity inhuman HCC by triggering degradation of itsspecific inhibitor, dual-specificity phosphatase 1(DUSP1), via the synergistic activity of S-phasekinase-associated protein 2 (SKP2), CDC28 proteinkinase 1b (CKS1) and ERK.10 However, the ERKdownstream targets underlying unrestrained cellcycle progression in human and rat HCC remainelusive. A major ERK effector is the Forkhead boxM1B (FOXM1) transcription factor, whose over-expression occurs in various experimental andhuman tumours.11–17 FOXM1 promotes prolifera-tion through its ability to influence various cellcycle phases. Indeed, FOXM1 triggers the activa-tion of SKP2/CKS1 ubiquitin ligase, which targets

Hepatology

Gut 2009;58:679–687. doi:10.1136/gut.2008.152652 679

group.bmj.com on August 18, 2010 - Published by gut.bmj.comDownloaded from

p21Cip1 and p27Kip1 proteins for degradation during the G1–Stransition.11–17 Furthermore, FOXM1 transcriptionally activatesCDC2, CYCLIN B1, AURORA KINASE A (AURKA), AURORAKINASE B, SURVIVIN, NIMA-related kinase 2 (NEK2) andcentromere protein A and B, thus allowing G2–M progression.11–

17 In the mouse liver, FOXM1 depletion results in block ofproliferation and resistance to hepatocarcinogenesis.18–20

However, the role of FOXM1 in rat liver with respect to geneticsusceptibility to hepatocarcinogenesis and its prognostic sig-nificance in human HCC have not been elucidated to date.Here, we evaluated the influence of susceptible and resistantgenotypes to hepatocarcinogenesis on FOXM1 expression inneoplastic rat liver lesions differently prone to progress to moremalignant stages. Next, we assessed the correlation betweenFOXM1 deregulation and patients’ clinicopathological featuresand prognosis in human HCC. Our results indicate thatFOXM1 expression is highest in liver neoplastic lesions fromsusceptible F344 rats and is associated with a poor prognosis inhuman HCC. Furthermore, FOXM1 triggers degradation of theERK inhibitor DUSP1 via transcriptional activation of SKP2 andCKS1, thus determining a positive loop reinforcing ERK activityin human HCC.

MATERIALS AND METHODS

Animals and treatmentsF344 and BN rats (Charles-River-Italia, Calco, Italy) were fed,housed, and treated according to the ‘‘resistant hepatocyte’’protocol,21 consisting of a 150 mg/kg intraperitoneal dose ofdiethylnitrosamine followed by 15 days of feeding a 0.02% 2-acetylaminofluorene-containing hyperprotein diet, with a par-tial hepatectomy at the midpoint of this feeding regime.6

Preneoplastic liver (4–6 weeks after initiation), early nodules(mostly clear/eosinophilic cell nodules; 12 weeks), neoplasticnodules (32 weeks) and HCCs (57–60 weeks) were collected.Animals received human care, and study protocols were incompliance with the National Institutes of Health guidelines foruse of laboratory animals.

Human tissue samplesSix normal livers, 26 HCCs with poor (HCCP) and 32 HCCswith better (HCCB) prognosis, with ,3 and .3 years survivalfollowing partial liver resection, respectively,7 and correspond-ing surrounding non-tumour livers were used. Patients’ clin-icopathological features are shown in Supplementary table 1.Liver tissues were kindly provided by Dr Snorri S Thorgeirsson(NCI, Bethesda, Maryland, USA). Institutional Review Boardapproval was obtained at participating hospitals and from theNational Institutes of Health.

Cell lines and treatmentsHuH6, HLE (exhibiting high FOXM1 levels) and SNU-182(exhibiting low FOXM1 expression) human HCC cell lines weremaintained as monolayer cultures in Dulbecco’s modifiedEagle’s medium supplemented with 10% fetal bovine serum(FBS). A total of 2.06106 HuH6 and HLE cells in 10 cm disheswere incubated for 12 h at 37uC before transfection with smallinterfering RNA (siRNA) duplexes specific to human FOXM1,ERK2 and glioblastoma-associated oncogene 1 (GLI1), asdescribed.17 22 23 siRNAs and scramble oligonucleotides (finalconcentration 100 nmol/l) were transfected using the siPORTNeoFX system (Ambion, Austin, Texas, USA). For transienttransfection experiments, SNU-182 cells were transfected withERK2 (wild-type) in a pUSEamp plasmid (Millipore, Millerica,

Massachusetts, USA), and FOXM1 and GLI1 (wild-type) cDNAin a pCMV6-XL vector (OriGene Technologies, Rockville,Maryland, USA) following the manufacturer’s protocol. Forflow cytometry experiments, 86105 HuH6 cells in 10 cm disheswere cultured for 12 h in Optimem medium (Invitrogen,Carlsbad, California)/10% FBS, and then for 48 h in mediumwith 0.2% FBS. After synchronisation (T0), the cells weretransfected with siRNA anti-FOXM1 or scramble oligonucleo-tide as above, incubated for an additional 48 h and thencollected, washed with phosphate-buffered saline (PBS) andfixed with 70% ethanol at 4uC for 12 h. Fixed cells wereincubated with RNase A and propidium iodide prior to flowcytometric analysis (BD FACSCalibur, Becton Dickinson, Italia,Buccinasco, Milan, Italy).

Proliferation and apoptotic indicesProliferation was determined in human HCC by counting Ki-67-positive cells, and apoptotic figures were stained with theApoTag peroxidase in situ apoptosis kit (Millipore), on at least3000 hepatocytes. Viability and apoptosis of cell lines in vitrowere determined by the WST-1 Cell Proliferation Reagent andthe Cell Death Detection Elisa Plus kit (Roche Diagnostics,Indianapolis, Indiana), respectively.

Quantitative reverse transcription-PCR (QRT-PCR)Primers for rat FOXM1 and RNR-18 genes were chosen with theassistance of the ‘‘Assay-on-DemandProducts’’ (AppliedBiosystems, Foster City, California, USA). PCRs and quantita-tive evaluation were performed as described.24

Immunoprecipitation analysisTissue samples from human livers, early nodules (12 weeks afterinitiation), nodules and HCCs from F344 and BN rats wereprocessed as reported.24 Membranes were probed with specificprimary antibodies (Supplementary table 2). Cyclin B1–Cdc2complexes were determined through immunoprecipitation withthe anti-cyclin B1 antibody and probing the membranes withthe anti-Cdc2 antibody. Bands were quantified in arbitraryunits by the ImageMaster Total LabV1.11 software, andnormalised to actin levels.

Erythropoietin (EPO) and vascular endothelial growth factor a(VEGFa) assaysHuH6, HLE and SNU-182 cell culture medium was collected,centrifuged to remove cellular debris and stored at 270uC untilassayed for EPO and VEGFa by ELISA following the supplier’sprotocol (R&D Systems, Minneapolis, Minnesota, USA). Datawere expressed in mU/100 mg and in pg/mg proteins per well forEPO and VEGFa, respectively.

Random amplified polymorphic DNA (RAPD) analysisTwenty-two previously designed primers were used to scoregenomic alterations in human HCCs, and the RAPD reactionwas performed as described.25 Differences from correspondingnon-tumorous livers were scored in the case of a change in theintensity, absence of a band or appearance of a new band inHCC. The frequency of altered RAPD profiles was calculated foreach liver lesion as reported.26

Chromatin immunoprecipitation (ChIP) analysisChIP was performed using the ChIP Assay Kit (Millipore)following the manufacturer’s protocol. Rabbit polyclonal anti-FOXM1 (Santa Cruz Biotechnology, Santa Cruz, California,

Hepatology

680 Gut 2009;58:679–687. doi:10.1136/gut.2008.152652

group.bmj.com on August 18, 2010 - Published by gut.bmj.comDownloaded from

USA) antibody was used to immunoprecipitate chromosomalDNA in cross-linked chromatin prepared from exponentiallygrowing HuH6 and HLE cell lines. Immunoprecipitated DNAwas analysed by PCR with primers specific for SKP2, CKS1 andCDC2 promoters.17

Statistical analysisStudent t test and the Tukey–Kramer test were used to evaluatestatistical significance, and Fisher’s exact test was used forcomparative analysis of survival of HCC patient subgroups.Multiple regression analysis was performed to calculate thecorrelation coefficient (R) using GraphPad InStat 3 (www.

graphpad.com). Values of p,0.05 were considered significant.Results are means (SD).

RESULTS

FOXM1 overexpression in rat and human hepatocarcinogenesisFor the experiments on rat liver tissues, frozen archival material,previously collected,6 24 was used. At 4 and 6 weeks afterinitiation, foci of altered hepatocytes occupied 80–97% and40–50% of the liver in F344 and BN rats, respectively. At12 weeks, a pool of nodules was used for analysis. They were,6-fold smaller in size in BN than in F344 rats. Dysplasticnodules and HCCs (70% moderately differentiated and 30%

Figure 1 Expression of FOXM1 (Forkhead box M1B) and related genes in preneoplastic and neoplastic liver of Fisher 344 (F344) and Brown Norway(BN) rats. Preneoplastic liver (4 and 6 weeks after initiation), early nodules (12 weeks), neoplastic nodules and hepatocellular carcinomas (HCCs) wereinduced in rats treated according to the ‘‘resistant hepatocyte’’ model. (A) FOXM1 mRNA levels were determined by quantitative reverse transcription-PCR. N Target = 22DCt; DCt = Ct RNR18–Ct target gene. Data are means (SD) of N target of at least five rats for each time point analysed. C, control(normal) liver; PL; preneoplastic liver, 12, early nodules; N, nodules; H, HCC. (B) Representative immunoprecipitation analysis of FOXM1 and itsupstream activators (phosphorylated extracellular signal-regulated kinase (pERK1/2) and glioblastoma-associated oncogene 1 (GLI1)) and downstreameffectors (CYCLIN B1, CDC2, AURORA KINASE A (AURKA) and NIMA-RELATED KINASE 2 (NEK2)). Protein lysates were immunoprecipitated withspecific antibodies and separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis. The CYCLIN B1–CDC2 complexes were determinedthrough immunoprecipitation (IP) with the anti-CYCLIN B1 antibody and probing the membranes with the anti-CDC2 antibody (immunoblot; IB).(C) Chemiluminescence analysis showing the mean (SD) of at least five rats for each time point investigated. Optical densities of the peaks werenormalised to b-actin values and expressed in arbitrary units. Tuckey–Kramer test: A, (N) carcinogen-treated vs control, p,0.001; (*) BN vs F344,p,0.001. C, (N) carcinogen-treated vs control, at least p,0.01; (*) BN vs F344, p,0.001.

Hepatology

Gut 2009;58:679–687. doi:10.1136/gut.2008.152652 681

group.bmj.com on August 18, 2010 - Published by gut.bmj.comDownloaded from

poorly differentiated) were present at 32 and 57 weeks,respectively, only in F344 rats. Clear/eosinophilic cell nodules,without atypical features, and HCCs (92% well differentiated)developed in BN rats at 32 and 60 weeks, respectively.

FOXM1 mRNA levels were comparatively evaluated by QRT-PCR in preneoplastic and neoplastic liver lesions from F344 andBN rats. No differences were detected when comparing theexpression of FOXM1 in normal livers and in livers at 4 and6 weeks after initiation from F344 and BN rats (fig 1A). FOXM1upregulation occurred as early as 12 weeks after initiation inF344 lesions and progressively increased in dysplastic nodulesand HCCs (fig 1A). In contrast, FOXM1 upregulation occurredonly in nodules and HCCs from resistant BN rats when

compared with control liver values, and always to a lowerextent than corresponding lesions in F344 rats (fig 1A).Equivalent results were observed when assessing the expressionof FOXM1 and its targets at the protein level. Indeed, FOXM1and its targets (Aurka, Cdc2, cyclin B1 and Nek2)11–13 17 27–30 andCdc2–cyclin B1 complexes were significantly higher at all timepoints in F344 than BN rat lesions (fig 1B,C).

The levels of FOXM1 and AURKA, NEK2, SURVIVIN, CDC2,CYCLIN B1, CDC25B, SKP2 and CKS111–13 17 27–30 downstreamtargets were assessed at the protein level in human normal livers,HCCs and the respective non-neoplastic surrounding livers (fig 2).A progressive upregulation of FOXM1 and its targets occurred innon-tumorous tissues and HCCs when compared with normal

Figure 2 Expression of FOXM1 (Forkhead box M1B) and related genes in human neoplastic liver lesions. Left panels: representativeimmunoprecipitation analysis of FOXM1, its upstream activators (phosphorylated extracellular signal-regulated kinase (pERK1/2) and glioblastoma-associated oncogene 1 (GLI1)) and downstream effectors (CYCLIN B1, CDC2, CDC25B, AURORA KINASE A (AURKA), NIMA-RELATED KINASE (NEK2),S-PHASE KINASE-ASSOCIATED PROTEIN 2 (SKP2), SKP2, SURVIVIN and CDC28 PROTEIN KINASE 1b (CKS1)). Protein lysates wereimmunoprecipitated with specific antibodies and separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis. The CYCLIN B1–CDC2complexes were determined through immunoprecipitation (IP) with the anti-CYCLIN B1 antibody and probing the membranes with the anti-Cdc2antibody (immunoblot; IB). Right panels: chemiluminescence analysis showing the mean (SD) of 6 normal livers, 32 HCCBs (hepatocellular carcinomaswith better prognosis) and 26 HCCPs (hepatocellular carcinomas with poor prognosis), and corresponding surrounding livers with better or poorerprognosis. Tuckey–Kramer test: (N) HCC and surrounding subtypes vs normal liver, at least p,0.001; (*) HCC subtypes vs corresponding surroundingliver, p,0.001. ({) Different from HCCB for p,0.0001. SLB and SLP, surrounding liver with better and poorer prognosis, respectively.

Hepatology

682 Gut 2009;58:679–687. doi:10.1136/gut.2008.152652

group.bmj.com on August 18, 2010 - Published by gut.bmj.comDownloaded from

livers, with the highest levels being detected in HCCP tumours.Accordingly, a gradual rise in CDC2–CYCLIN B1 complexes wasobserved from non-neoplastic liver tissues to tumours, reachingthe highest values in HCCP tumours.

These data indicate that activation of the FOXM1 axis isassociated with susceptibility to hepatocarcinogenesis in ratsand clinical outcome in human HCC.

Correlation of FOXM1 levels with clinicopathological parametersin human HCCThe proliferation index, microvessel density and genomic instabil-ity (RAPD) values were 2.5- to 3-fold higher in HCCP than HCCB(Supplementary table 1) and correlated with FOXM1 levels (fig 3).In contrast, an inverse correlation of FOXM1 expression withapoptosis and patients’ length of survival was found (fig 3). Nosignificant correlation between FOXM1 and other clinicopatholo-gical parameters, including aetiology, sex, age, presence of cirrhosis,a-fetoprotein, tumour size and grading was observed.

FOXM1 activation is mediated by ERK and GLI1 in HCC cell linesPrevious reports showed FOXM1 upregulation following eitherERK or GLI1 induction.31 32 To identify the upstream inducer(s) ofFOXM1 in HCC, we first assessed the levels of activated ERK andGLI1 in both rat and human lesions. Phosphorylated ERK (pERK1/2) and GLI1 protein levels rose at 12 weeks after initiation andincreased in neoplastic nodules and HCCs from F344 rats (fig 1B).In BN rats, pERK1/2 and GLI1 upregulation was limited to nodulesand HCCs, but to a lower extent than in corresponding F344lesions. Furthermore, pERK1/2 and GLI1 expression was higher inhuman HCCs than in normal and non-neoplastic surroundinglivers, and most pronounced in HCCP (fig 2). Next, we assessedthe impact of suppressing ERK proteins and GLI1 on FOXM1expression in human HuH6 and HLE HCC cell lines, displayingelevated FOXM1 mRNA levels (not shown). Silencing of eitherERK2 or GLI1 via siRNA led to moderate decreases in GLI1 andERK2 protein levels, respectively, and to stronger decreases inFOXM1 levels. Noticeably, the combined suppression of ERK2 andGLI1 resulted in almost complete inhibition of FOXM1 (fig 4A,Supplementary fig S1a). Conversely, upregulation by transienttransfection with ERK2 or GLI1 resulted in increases in GLI1 andpERK1/2, respectively, and in an elevated FOXM1 level whichfurther increased in doubly transfected cells (fig 4B,Supplementary fig S1b). These findings imply an additive activityof ERK and GLI1 in promoting FOXM1 upregulation.

FOXM1 sustains ERK activity via degradation of DUSP1Recent results showed that human kidney embryonic cells33 andrat and human HCCs8 10 maintain elevated levels of ERK viadownregulation of its specific inhibitor, DUSP1. In these cells,DUSP1 downregulation was achieved by its proteolysismediated by cooperation bertween ERK, SKP2 and CKS1.10

Previous17 and present data suggest that CKS1 and SKP2 areFOXM1 targets. In accordance with this hypothesis, wedetected the functional interaction of FOXM1 with SKP2 andCSK1 proteins in HuH6 and HLE cells by ChIP analysis (fig 5).Thus, we examined the role of FOXM1 in DUSP1 suppression.Strikingly, inhibition of FOXM1 expression by siRNA in HuH6and HLE cell lines led to marked CKS1 and SKP2 down-regulation, with consequent upregulation of DUSP1, a decreasein levels of ubiquitinated DUSP1 and a strong reduction in ERKactivity (fig 4C, Supplementary fig S1c). Conversely, over-expression of FOXM1 in SNU-182 HCC cells (exhibiting lowFOXM1 levels) triggered DUSP1 downregulation, which wasinhibited by siRNA against SKP2 and CKS1 (fig 4D,Supplementary fig S1d). These observations assign a role toFOXM1 in sustaining ERK activation via downregulation of theERK inhibitor DUSP1.

Figure 3 Relationships between FOXM1 (Forkhead box M1B) levelsand proliferation index, microvessel density (MVD), random amplifiedpolymorphic DNA (RAPD), apoptotic index (percentage of apoptoticbodies) and length of survival (months after partial liver resection) ofhuman hepatocellular carcinomas (HCCs).

Hepatology

Gut 2009;58:679–687. doi:10.1136/gut.2008.152652 683

group.bmj.com on August 18, 2010 - Published by gut.bmj.comDownloaded from

FOXM1 promotes HCC cell proliferation, survival andangiogenesisThe role of FOXM1 in HCC cells was investigated by assessingthe consequence of FOXM1 inactivation in HuH6 and HLE cellsby siRNA and of FOXM1 overexpression on SNU-182 cells,respectively (fig 6). FOXM1 suppression markedly reduced

proliferation, and EPO and VEGFa secretion, and inducedapoptosis. In sharp contrast, FOXM1 overexpression led to anincrease in cell proliferation, and in EPO and VEGFa secretion,(potentially) contributing to neovascularisation,34 35 and adecline in apoptosis. The growth suppression by siRNA wasstudied further in HuH6 cells by flow cytometric analysis,which revealed a cell cycle arrest at G0–G1 and G2–M phases.

DISCUSSIONRecent work indicates that a peculiar feature of hepatocarcino-genesis is that changes of signal transduction in autonomouslygrowing preneoplastic and neoplastic cells are under control ofthe genes responsible for susceptibility to HCC development.4

Deregulation of G1 and S phases of the cell cycle, implying fastG1–S transition and elevated DNA synthesis, in lesions of thesusceptible F344 rat, is considerably less remarkable in lesions ofthe resistant BN rat.4 6 Here, we show an earlier and morepronounced FOXM1 induction associated with a faster growthof preneoplastic and neoplastic liver lesions in susceptible F344than in resistant BN rats. FOXM1 upregulation was followed bya very prominent rise in FOXM1 targets in F344 preneoplasticand neoplastic livers, including some proteins implicated in G2–Mtransition, such as Cdc2, cyclin B1, AurkA and Nek2.11–17 27–30

Figure 4 Representative immunoprecipitation analysis of the effects of the inhibition of FOXM1 (Forkhead box M1B) upstream regulators(extracellular signal-regulated kinase (ERK) and glioblastoma-associated oncogene 1 (GLI1)) and changes in FOXM1 protein level on its downstreameffectors S-phase kinase-associated protein 2 (SKP2) and CDC28 protein kinase 1b (CKS1) in human hepatocellular carcinoma (HCC) cell lines.Ubiquitinated dual-specificity phosphatase 1 (DUSP1) was determined by immunoprecipitation (IP) of ubiquitin (UBQ) followed by immunoblotting (IB)with antibodies against DUSP1. (A) The HuH6 cell line was treated for 48 h with small interfering RNA (siRNA) against ERK2, GLI1, or ERK2+GLI1.Equivalent results were obtained using the HLE cell line (data not shown). (B) SNU-182 cells were transfected with ERK2 (wild-type) in a pUSEampplasmid (Millipore), and FOXM1 and GLI1 (wild-type) cDNA in a pCMV6-XL vector. (C) The HuH6 cell line was treated for 12 and 24 h with siRNAagainst FOXM1. Equivalent results were obtained using the HLE cell line (data not shown). (D) The SNU-182 cell line, transiently transfected withFOXM1, was treated for 36 h with siRNA against CSK1 or SKP2. Controls (C) received solvent alone or scramble oligonucleotides.

Figure 5 Functional interaction of FOXM1 (Forkhead box M1B) withSKP2 (S-phase kinase-associated protein 2), CKS1 (CDC28 protein kinase1b) and CDC2 genes in human HuH6 and HLE hepatocellular carcinoma(HCC) cell lines as detected by chromatin immunoprecipitation analysis.M, DNA marker; I, input (aliquot of chromatin prior to immunoprecipitation);N, negative control (absence of DNA in the PCR); 1, HuH6; 2, HLE. Cross-linked chromatin was immunoprecipitated with the anti-FOXM1 antibody.Chromatic immunoprecipitation of the CDC2 gene by FOXM1 was alsoperformed as a positive control.

Hepatology

684 Gut 2009;58:679–687. doi:10.1136/gut.2008.152652

group.bmj.com on August 18, 2010 - Published by gut.bmj.comDownloaded from

Figure 6 Effect of the modulation of FOXM1 (Forkhead box M1B) expression on proliferation and apoptosis of human hepatocellular carcinoma (HCC)cell lines. (A) and (B) Effect of inhibition of FOXM1 by small interfering RNA (siRNA) on erythropoietin and vascular endothelial growth factor a (VEGFa)production, proliferation, and apoptosis in HuH6 and HLE cells, respectively. (C) Effect of FOXM1 overexpression on erythropoietin and VEGFaproduction, proliferation and apoptosis in SNU-182 cells. The cells were transiently transfected with FOXM1 cDNA in a pCMV6-XL vector. Results in A–C aremeans (SD) of five experiments. siRNA-treated or transfected cells and controls are shown by squares and rhombuses, respectively. Controlsreceived scramble oligonucleotides (SC) or plasmid alone. (D) Representative flow cytometric analysis and average quantitative data (SD) of threeindependent experiments with HuH6 cells, transfected with anti-FOXM1 siRNA. The analyses were performed at the end of culturesynchronisation, in low serum medium (T0), and after 48 h incubation with SC or siRNA. Quantitative evaluations were made by ModFit LT (VeritySoftware House, Topsham, Maine, USA). (E) FOXM1 mRNA levels determined by quantitative reverse transcription-PCR. N Target = 22DCt;DCt = Ct RNR18–Ct target gene. Data are means (SD) of N target of three experiments. Tuckey–Kramer test, treated vs control: (A) and (B),p,0.001 at all time points after 6; (C), p,0.001 at the asterisked time points. ‘‘t’’ test: (D) siRNA vs SC, p = 0.0045, p = 0.0075 and p,0.001for G0–G1, S and G2–M, respectively. (E) siRNA vs SC, p = 0.0045.

Hepatology

Gut 2009;58:679–687. doi:10.1136/gut.2008.152652 685

group.bmj.com on August 18, 2010 - Published by gut.bmj.comDownloaded from

Accordingly, restriction of FOXM1 expression by siRNA, in theHuH6 cell line, led to G2–M phase arrest. FOXM1 inhibition wasalso associated with G0–G1 arrest which could depend onmodulation of the activity of FOXM1 targets, including pERK1/2 downregulation,36 and possible activation of G1 inhibitors suchas p21WAF1 and p27KIP1.14–17 Thus, the present investigation showsFOXM1 regulation by genes controlling the susceptibility toHCC, and underlines the role of uncontrolled progression throughthe cell cycle as an effector mechanism of these genes, determiningthe susceptible phenotype.

In human samples, FOXM1 was ubiquitously and progres-sively induced from non-tumorous surrounding liver to HCC,with the highest increase in HCCP, substantiating the role ofFOXM1 in both HCC development and progression, inaccordance with mouse18 and rat (present work) hepatocarci-nogenesis. FOXM1 levels directly correlated with genomicinstability, the proliferation index and tumour microvesseldensity, and inversely with the apoptotic index and survival,indicating that FOXM1 contributes to hepatocarcinogenesis viamultiple mechanisms. Indeed, FOXM1 upregulation inducedoverexpression of genes promoting cell cycle progression(AURKA, CDC2, CYCLIN B1, NEK2 and CDC25B), generationof genomic instability (NEK2 and CDC25B), suppressors of cellcycle inhibitors (SKP2 and CKS1) and apoptosis inhibitors(SURVIVIN)11–17 27–30 (Supplementary fig S2). Importantly, thestrong correlation between FOXM1 levels and both genomicinstability rate and adverse outcome in HCC agrees with theexistence of a molecular signature, including FOXM1 over-expression, which is significantly associated with the degree ofgenomic instability and predicts survival of patients withmultiple tumours.37 Furthermore, 9 of 70 (12.9%) genesrepresenting the signature of genomic instability and shortsurvival are direct FOXM1 targets (CYCLIN B1 and B2, CDC2,NEK2, KIF20A, TOP2A, CDC25B, AURORA KINASE A andAURORA KINASE B).11–17 27–30 These data, together with theresults from the comparative analysis of preneoplastic andneoplastic lesions of rats with different genetic predisposition tohepatocarcinogenesis and human HCC with different prog-nosis, suggest a potential prognostic role for FOXM1 signallingin numerous neoplasms. In addition, induction of EPO andVEGFa expression substantiates the role of FOXM1 in HCCneoangiogenesis. Signalling via EPO and the EPO receptor isrequired for angiogenesis in numerous non-haematopoietictissues34 35 and in cancer, including human pancreatic cancer,16

and mouse and human HCC cells.38 39

A link between fast growth and signalling deregulationcharacterises human HCCP, whereas the behaviour of HCCBsis more similar to that of resistant rat lesions4 6 22 (and presentwork). This does not necessarily imply a genetic regulation ofsignalling pathways in humans like that found in rodents, inwhich polygenic inheritance with several low penetrance genesand a main gene regulates the genetic predisposition to HCC.5

Even if according to epidemiological studies a genetic model,similar to that of rodents, can influence human hepatocarcino-genesis,5 further studies are needed to clarify the influence ofsusceptibility genes on signalling pathways supporting tumourgrowth and progression in humans.

Furthermore, we demonstrated the importance of FOXM1 insustaining ERK activity by inducing CKS1 and SKP2 expression.SKP2/CSK1 ligase, which degrades DUSP1, a major ERKinhibitor, contributes to sustained ERK overactivity in HCC.10

On the other hand, sustained ERK2 activation triggers DUSP1degradation via phosphorylation of its Ser296 residue, followedby ubiquitination and proteasomal degradation (Supplementary

fig S2).10 Therefore, FOXM1 is involved in a positive feedbackloop, reinforcing the ERK cascade by its ability to inhibitDUSP1.

The possible upstream inducers of FOXM1 were alsoinvestigated. FOXM1 is a direct transcriptional target ofGLI1.40 GLI family proteins, including GLI1, 2 and 3, are theterminal effectors of Hedgehog signalling.41 42 The interaction ofSonic hedgehog (SHH) with its plasma membrane receptorPTCH1, releases PTCH-induced inhibition of the membraneprotein Smoothened (SMO). This results in activation andnuclear translocation of GLI proteins, where they activate targetgene transcription.43 GLI2 is overexpressed in some HCC celllines, and its inhibition by antisense oligonucleotides inhibitscell proliferation.44 GLI1 overexpression occurs in a lowernumber of HCC cell lines than GLI2 and its inhibition causesa lower decrease in growth rate.44 We did not evaluate the effectof GLI2 on FOXM1 activity. It must be considered that theeffect of GLI proteins on cell proliferation may reflect changes indifferent genes and signalling pathways. Our data suggest thatFOXM1 upregulation results from combined ERK and GLI1activity in HCC. The elevated levels of both ERK proteins andGLI1 in F344 rat liver lesions and HCCP might therefore explainthe highest levels of FOXM1 in these lesions. Our resultssuggest a reciprocal activation of ERK2 and GLI1. GLI1 has aMEK-1-responsive N-terminal domain, and a recent reportindicates that the activation of the ERK pathway by basicfibroblast growth factor stimulates GLI1 activity through thisdomain.45 The mechanism underlying ERK stimulation by theHedgehog pathway is unclear. Indirect ERK activation throughGLI-mediated induction of platelet-derived growth factor(PDGF) receptor a has been postulated.46 Present knowledgesuggests a complex cross-talk between Hedgehog and MEK/ERKsignalling whose role in hepatocarcinogenesis requires furtherinvestigation. Interestingly, a recent report implies the com-bined overexpression of HSP90 and CDC37 proto-oncogenes insustaining elevated Fused Homolog expression.47 Accordingly,our preliminary data indicate the upregulation of the FusedHomolog gene in human HCC (results not shown), and aprevious report showed a strong induction of HSP90 andCDC37 in F344 rat liver lesions and human HCCP.24 Thus, arole for combined activity of HSP90 and CDC37 in the highestactivation of GLI1 observed in F344 neoplastic lesions andhuman HCCPs might be hypothesised.24 Overall it seems thatFOXM1 acts as a pleiotropic regulator of human hepatocarci-nogenesis, playing multiple roles on preneoplastic and neoplas-tic hepatocytes.

Overall, our results indicate for the first time a genetic controlof FOXM1 signalling deregulation during hepatocarcinogenesisand its role in both HCC development and prognosis.Furthermore, we show the involvement of FOXM1 in a positivefeedback loop in which the activating interaction of FOXM1with the SKP2/CSK1 ligase sustains ERK and FOXM1 over-activity. This mechanism may have a central role in thepathogenesis of fast growing HCCs. The association of theblock of FOXM1 signalling by specific siRNA with a consistentdecrease in HCC cell growth and EPO production, and anincrease in apoptosis in vitro, suggests that FOXM1 couldrepresent a therapeutic target that, in association with othertargets, may contribute to create networked biological treat-ments.48

Acknowledgements: We are grateful to Dr Piero Bonelli, Experimental Zoo-prophylactic Institute of Sardinia, for his assistance during the flow cytometryexperiments. Supported by grants from the Associazione Italiana Ricerche sul Cancroand Ministero dell’Istruzione, Universita e Ricerca.

Hepatology

686 Gut 2009;58:679–687. doi:10.1136/gut.2008.152652

group.bmj.com on August 18, 2010 - Published by gut.bmj.comDownloaded from

Competing interests: None.

Ethics approval: Institutional Review Board approval was obtained at participatinghospitals and from the National Institutes of Health

REFERENCES1. Bruix J, Boix L, Sala M, et al. Focus on HCC. Cancer Cell 2004;5:215–9.2. Thorgeirsson SS, Grisham JW. Molecular pathogenesis of human HCC. Nat Genet

2002;31:339–46.3. Farazi PA, DePinho RA. HCC pathogenesis: from genes to environment. Nat Rev

Cancer 2006;6:674–87.4. Feo F, De Miglio MR, Simile MM, et al. HCC as a complex polygenic disease.

Interpretive analysis of recent developments on genetic predisposition. BiochimBiophys Acta 2006;1765:126–47.

5. Hemminki K, Li X. Familial risks of cancer as a guide to gene identification and modeof inheritance. Int J Cancer 2004;110:291–4.

6. Pascale RM, Simile MM, De Miglio MR, et al. Cell cycle deregulation in liver lesionsof rats with and without genetic predisposition to hepatocarcinogenesis. Hepatology2002;35:1341–50.

7. Lee JS, Chu IS, Mikaelyan A, et al. Application of comparative functional genomics toidentify best-fit mouse models to study human cancer. Nat Genet 2004;36:1306–11.

8. Calvisi DF, Pinna F, Pellegrino R, et al. Ras-driven proliferation and apoptosissignaling during rat liver carcinogenesis is under genetic control. Int J Cancer2008;123:2057–64.

9. Calvisi DF, Ladu S, Gorden A, et al. Mechanistic and prognostic significance ofaberrant methylation in the molecular pathogenesis of human HCC. J Clin Invest2007;117:2713–22.

10. Calvisi DF, Pinna F, Meloni F, et al. Dual-specificity phosphatase 1 ubiquitination inextracellular signal-regulated kinase-mediated control of growth in humanhepatocellular carcinoma. Cancer Res 2008;68:4192–200.

11. Costa RH, Kalinichenko VV, Major ML, et al. New and unexpected: forkhead meetsARF. Curr Opin Genet Dev 2005;15:42–8.

12. Wierstra I, Alves J. FOXM1, a typical proliferation-associated transcription factor.Biol Chem 2007;388:1257–74.

13. Kim IM, Ackerson T, Ramakrishna S, et al. The Forkhead Box m1 transcription factorstimulates the proliferation of tumor cells during development of lung cancer. CancerRes 2006;66:2153–61.

14. Kalin TV, Wang IC, Ackerson TJ, et al. Increased levels of the FOXM1 transcriptionfactor accelerate development and progression of prostate carcinomas in bothTRAMP and LADY transgenic mice. Cancer Res 2006;66:1712–20.

15. Yoshida Y, Wang IC, Yoder HM, et al. The forkhead box M1 transcription factorcontributes to the development and growth of mouse colorectal cancer.Gastroenterology 2007;132:1420–31.

16. Wang Z, Banerjee S, Kong D, et al. Down-regulation of Forkhead Box M1transcription factor leads to the inhibition of invasion and angiogenesis of pancreaticcancer cells. Cancer Res 2007;67:8293–300.

17. Wang IC, Chen YJ, Hughes D, et al. Forkhead box M1 regulates the transcriptionalnetwork of genes essential for mitotic progression and genes encoding the SCF(Skp2–Cks1) ubiquitin ligase. Mol Cell Biol 2005;25:10875–94.

18. Wang X, Kiyokawa H, Dennewitz MB, et al. The Forkhead Box m1b transcriptionfactor is essential for hepatocyte DNA replication and mitosis during mouse liverregeneration. Proc Natl Acad Sci USA 2002;99:16881–6.

19. Krupczak-Hollis K, Wang X, Dennewitz MB, et al. Growth hormone stimulatesproliferation of old-aged regenerating liver through forkhead box m1b. Hepatology2003;38:1552–62.

20. Kalinichenko VV, Major ML, Wang X, et al. FOXM1b transcription factor is essentialfor development of HCCs and is negatively regulated by the p19ARF tumorsuppressor. Genes Dev 2004;18:830–50.

21. Solt DB, Medline A, Farber E. Rapid emergence of carcinogen-induced hyperplasticlesions in a new model for the sequential analysis of liver carcinogenesis. Am J Pathol1977;88:595–618.

22. Feng YZ, Shiozawa T, Miyamoto T, et al. Overexpression of hedgehog signalingmolecules and its involvement in the proliferation of endometrial carcinoma cells. ClinCancer Res 2007;13:1389–98.

23. Dimitri CA, Dowdle W, MacKeigan JP, et al. Spatially separate docking sites onERK2 regulate distinct signaling events in vivo. Curr Biol 2005;15:1319–24.

24. Pascale RM, Simile MM, Calvisi DF, et al. Role of HSP90, CDC37, and CRM1 asmodulators of P16(INK4A) activity in rat liver carcinogenesis and human liver cancer.Hepatology 2005;42:1310–9.

25. Calvisi DF, Factor VM, Ladu S, et al. Disruption of beta-catenin pathway or genomicinstability define two distinct categories of liver cancer in transgenic mice.Gastroenterology 2004;126:1374–86.

26. Luceri C, De Filippo C, Caderni G, et al. Detection of somatic DNA alterations inazoxymethane-induced F344 rat colon tumors by random amplified polymorphic DNAanalysis. Carcinogenesis 2000;21:1753–6.

27. Laoukili J, Stahl M, Medema RH. FOXM1: at the crossroads of ageing and cancer.Biochim Biophys Acta 2007;1775:92–102.

28. Wonsey DR, Follettie MT. Loss of the forkhead transcription factor FOXM1 causescentrosome amplification and mitotic catastrophe. Cancer Res 2005;65:5181–9.

29. Zhao YY, Gao XP, Zhao YD, et al. Endothelial cell-restricted disruption of FOXM1impairs endothelial repair following LPS-induced vascular injury. J Clin Invest2006;116:2333–43.

30. Laoukili J, Kooistra MR, Bras A, et al. FOXM1 is required for execution of the mitoticprogramme and chromosome stability Nat Cell Biol 2005;7:126–36.

31. Major ML, Lepe R, Costa RH. Forkhead box M1B transcriptional activity requiresbinding of Cdk–cyclin complexes for phosphorylation-dependent recruitment of p300/CBP coactivators. Mol Cell Biol 2004;24:2649–61.

32. Teh MT, Wong ST, Neill GW, et al. FOXM1 is a downstream target of GLI1 in basalcell carcinomas. Cancer Res 2002;62:4773–80.

33. Lin YW, Yang JL. Cooperation of ERK and SCFSkp2 for MKP-1 destructionprovides a positive feedback regulation of proliferating signaling. J Biol Chem2006;281:915–26.

34. Arcasoy MO. The non-haematopoietic biological effects of erythropoietin.Br J Haematol 2008;141:14–31.

35. Watanabe D. Erythropoietin as a retinal angiogenic factor in proliferative diabeticretinopathy. Nippon Ganka Gakkai Zasshi 2007;111:892–8.

36. Klein EA, Campbell LE, Kothapalli D, et al. Joint requirement for Rac and ERKactivities underlies the mid-G1 phase induction of cyclin D1 and S phase entry in bothepithelial and mesenchymal cells. J Biol Chem 2008;283:30911–8.

37. Carter SL, Eklund AC, Kohane IS, et al. A signature of chromosomal instabilityinferred from gene expression profiles predicts clinical outcome in multiple humancancers. Nat Genet 2006;38:1043–8.

38. Gusarova GA, Wang IC, Major ML, et al. A cell-penetrating ARF peptide inhibitor ofFOXM1 in mouse HCC treatment. J Clin Invest 2007;117:99–111.

39. Ribatti D, Marzullo A, Gentile A, et al. Erythropoietin/erythropoietin-receptor systemis involved in angiogenesis in human hepatocellular carcinoma. Histopathology2007;50:591–6.

40. Teh MT, Wong ST, Neill GW, et al. FOXM1 is a downstream target of Gli1 in basalcell carcinomas. Cancer Res 2002;62:4773–80.

41. Pasca di Magliano M, Hebrok M. Hedgehog signalling in cancer formation andmaintenance. Nat Rev Cancer 2003;3:903–11.

42. Kasper M, Regl G, Frischauf AM, et al. GLI transcription factors: mediators ofoncogenic Hedgehog signalling. Eur J Cancer 2006;42:437–45.

43. Crompton T, Outram SV, Hager-Theodorides AL. Sonic hedgehog signalling in T-celldevelopment and activation. Nat Rev Immunol 2007;7:726–35.

44. Kim Y, Yoon JW, Xiao X, et al. Selective down-regulation of glioma-associatedoncogene 2 inhibits the proliferation of hepatocellular carcinoma cells. Cancer Res2007;67:3583–93.

45. Riobo NA, Haines GM, Emerson CP Jr. Protein kinase C-delta and mitogen-activatedprotein/extracellular signal-regulated kinase-1 control GLI activation in hedgehogsignaling. Cancer Res 2006;66:839–45.

46. Xie J, Aszterbaum M, Zhang X, et al. A role of PDGFRalpha in basal cell carcinomaproliferation. Proc Natl Acad Sci USA 2001;98:9255–9.

47. Kise Y, Takenaka K, Tezuka T, et al. Fused kinase is stabilized by Cdc37/Hsp90 andenhances Gli protein levels. Biochem Biophys Res Commun 2006;351:78–84.

48. Epstein RJ, Leung TW. Reversing hepatocellular carcinoma progression by usingnetworked biological therapies. Clin Cancer Res 2007;13:11–7.

Hepatology

Gut 2009;58:679–687. doi:10.1136/gut.2008.152652 687

group.bmj.com on August 18, 2010 - Published by gut.bmj.comDownloaded from