genomic progression in mouse models for liver tumors

Neoplastic tumors arise from a stepwise sequence ofevents known as tumor progression (Nowell 1976; Kin-zler and Vogelstein 1996). Each step in the sequence isthought to represent a discrete genetic or epigenetic aber-ration, typically affecting the structure or expression ofeither a proto-oncogene or tumor suppressor gene.

This scheme engenders a number of questions. Whatcauses each of the steps in tumor progression? How doeseach step contribute to tumorigenesis? Are there discreteand specific genotypic pathways to each form of neo-plasm? Must the steps in these pathways occur in a par-ticular chronological sequence? How constrained is thetumorigenic pathway, once an initiating event has oc-curred? Which of the steps in any given pathway might besuitable targets for therapeutic intervention?

We cannot answer these questions solely by the studyof human tumors. Instead, we need to reconstruct theevents of tumorigenesis in a prospective manner. Themost desirable means to this end would be the develop-ment of animal models that replicate tumorigenesis as itoccurs in humans. For a host of reasons, the animal ofchoice is the laboratory mouse.

There is a venerable tradition of denigrating mousemodels for cancer, and with good cause, but now the tidehas turned. Recent years have seen the development ofmouse models that are based on the faithful reconstruc-tion of genetic lesions found in human cancer. The resulthas been models that offer both biological and molecularauthenticity. Here we illustrate this advance by describ-ing the development and use of mouse models for benignand malignant tumors of the liver, the genesis of whichresembles that of hepatic tumors in humans.

Cancer of the liver is among the most common andgrievous of human malignancies (Parkin et al. 2005). Al-though relatively infrequent in developed nations, at least600,000 new cases occur worldwide every year. The av-

erage survival time is 6 months. The only therapeutic re-course of note is surgical resection and liver transplanta-tion.

The principal cause of human liver cancer is infectionwith either hepatitis B virus or hepatitis C virus, althoughdietary toxins and chronic alcoholism have also been im-plicated (Llovet et al. 2003). Downstream of these causes,however, lie perturbations that confer gain of function onproto-oncogenes or loss of function on tumor suppressorgenes (Thorgeirsson and Grisham 2002). Frequent amongthese perturbations is overexpression of the proto-onco-gene MET (Tavian et al. 2000), which encodes a receptorprotein-tyrosine kinase (Met), activated by a ligandknown as either hepatocyte growth factor or scatter factor(Trusolino and Comoglio 2002).

We have modeled the pathogenesis of liver tumors byexpressing conditional transgenes of MET in the livercells of inbred mice. The response to the MET transgenevaried with both the magnitude and timing of its expres-sion, but included arrested development of the liver, aswell as benign and malignant liver tumors that displayboth phenotypic and genotypic resemblances to humancounterparts. Our results provide insight into the geno-typic underpinnings of tumor progression. The modelsshould be useful in studying the mechanisms of tumori-genesis and the preclinical testing of new therapeutics.

PATHOGENIC EFFECTS OF MET VARY WITHTHE MAGNITUDE AND DEVELOPMENTAL

TIMING OF GENE EXPRESSION

To replicate the overexpression of MET found fre-quently in human liver tumors, we targeted the expressionof a transgene representing human MET to liver cells byusing the transcriptional control element for the liver-en-riched activator protein gene (Wang et al. 2001). Four in-

Genomic Progression in Mouse Models for Liver Tumors

A.D. TWARD,* K.D. JONES,† S. YANT,‡ M.A. KAY,‡ R. WANG,¶ AND J.M. BISHOP**G.W. Hooper Foundation, †Department of Pathology, ¶Departments of Surgery and Anatomy and

Pacific Vascular Research Lab, University of California at San Francisco, San Francisco, California 94143;‡Departments of Pediatrics and Genetics, Stanford University, Stanford, California 94305

Cold Spring Harbor Symposia on Quantitative Biology, Volume LXX. © 2005 Cold Spring Harbor Laboratory Press 0-87969-773-3. 217

The principal cause of human liver cancer is infection with hepatitis viruses B and C, but tumor progression is fueled by en-suing perturbations that confer gain of function on proto-oncogenes or loss of function on tumor suppressor genes. Frequentamong these perturbations is overexpression of the proto-oncogene MET. We have modeled the pathogenesis of liver tumorsby expressing conditional transgenes of MET in the hepatocytes of inbred mice. The response to the MET transgene variedwith both the magnitude and timing of its expression but included hyperplasia of hepatic progenitor cells, as well as benignand malignant tumors that display both phenotypic and genotypic resemblances to human counterparts. The results revealMET to be a crucial switch in the development of the liver; dramatize how different cellular compartments within a devel-opmental lineage can give rise to distinctive tumor stem cells; delineate rules of tumor progression; provide evidence that theexperimental tumors in mice are authentic models for human tumors; and support a role for MET in the genesis of humanliver tumors. The models should be useful in elucidating the mechanisms of tumorigenesis and in the preclinical testing ofnew therapeutics.

217-224_25_Tward_Symp70.qxd 5/12/06 9:57 AM Page 217

dependent strains of transgenic mice were developed, ineach of which expression of the MET transgene could beinactivated by administration of doxycycline (Wang et al.2001). Two strains expressed the transgene at relativelyhigh levels (previously designated lines 1 and 2), whereastwo expressed it at lower levels (previously designatedlines 3 and 4) (for details, see Wang et al. 2001). The phe-notypic response was dependent on both the magnitudeand timing of expression (Fig. 1).

Physiological expression of MET in the normal mouseliver occurs throughout embryogenesis and into adult life(Ishikawa et al. 2001 and data not shown). However, thekinase activity of the gene product subsides soon afterbirth, apparently in response to controls that act directlyon the protein without reducing its abundance (Ishikawaet al. 2001 and data not shown). In contrast, hepatic ex-pression of a MET transgene at a high level in utero andcontinuing after birth produced sustained activity of thekinase (data not shown), and this in turn led to deathwithin 3–8 postnatal weeks (Fig. 1, first protocol). Priorto death, the animals had greatly enlarged, hyperplasticlivers composed of immature precursor cells (Fig. 2 anddata not shown). Death was apparently caused by the ab-sence of mature liver function.

If instead, the same transgene was inactivated 1 monthafter birth (Fig. 1, second protocol), the hyperplastic liv-ers were remodeled into mature tissue, and survival of theanimals was prolonged (Fig. 2 and data not shown).Within several months, however, the mice developed fa-tal hepatic carcinomas composed of both biliary and hep-atic lineages (Fig. 2). The mixed nature of the tumors sug-gests that tumorigenesis was initiated from a relativelyprimitive, bipotential cell. The tumors displayed a marker

of precursor cells (α-fetoprotein, AFP) and a marker ofmature hepatocytes (phenyalanine hydroxylase, PAH). Asecond marker of mature cells, glutamine synthetase(GS), was absent. The extensive proliferation of progeni-tor cells that occurred during the time that the MET trans-gene was active would have provided an opportunity forthe occurrence and fixation of an initiating mutation.

In a third protocol with the same strain of mice, theMET transgene was held inactive until 3 weeks afterbirth, then activated by the withdrawal of doxycycline(Fig. 1, third protocol). Over the ensuing months, the liv-ers first became morphologically dysplastic (data notshown), then developed neoplastic tumors resembling thehepatocellular carcinomas of humans (HCC) without ev-idence of biliary elements (Fig. 2). The tumor cells dis-played molecular markers of both hepatic progenitor cellsand mature hepatocytes (Fig. 2). Thus, the tumors mayhave originated either by transformation of partially dif-ferentiated progenitor cells or by partial dedifferentiationof mature hepatocytes—we cannot presently distinguishbetween these two possibilities.

Expression of a MET transgene at a lower level in uteroand throughout postnatal life elicited a progression ofevents that evolved from focal hyperplasia of mature hep-atocytes (frequently surrounding central veins) to dys-plastic foci and, eventually, to multiple nodules of HCC(Fig. 1, fourth protocol; Fig. 3) (see Wang et al. 2001).The histopathology of the murine tumors closely resem-bled that of human HCC (Fig. 3) (Wang et al. 2001). Nobiliary elements were apparent. The tumor cells displayedmolecular markers of both immature and mature hepato-cytes—AFP for the former, PAH and GS for the latter(data not shown). Similar tumors arose, albeit moreslowly, even if activation of the MET transgene was de-layed until after weaning of the mice (data not shown).

218 TWARD ET AL.

Figure 1. Protocols for the induction of liver tumors with trans-genic MET. The strains of transgenic mice used here have beendescribed previously (Wang et al. 2001). Expression of the METtransgene could be repressed by administration of doxycycline(OFF) or permitted by withholding doxycycline (ON). Twostrains of mice expressed the transgene at relatively high levels(Lines 1 and 2 in Wang et al. 2001), two others expressed it atrelatively low levels (Lines 3 and 4 in Wang et al. 2001). Thefirst three protocols diagrammed in the figure employed Line 2,the fourth protocol employed Line 3. Doxycycline was admin-istered (circles) or withheld (square) at the indicated timepoints. Solid black indicates activity of Met kinase, solid whiteindicates absence of such activity despite the presence of Metprotein.

Figure 2. Hepatic phenotypes elicited by overexpression ofMET. Doxycycline was administered according to the first threeprotocols outlined in Fig. 1, with the indicated outcomes. Liverswere harvested at the following times: (Control) 1 month post-partum; (MET ON/ON) 1 month; (MET ON/OFF, Non-Tumor)2 months; (MET ON/OFF, Tumor) 16 months; and (METOFF/ON) 12 months. Specimens were analyzed by microscopyafter staining with hematoxylin and eosin (H+E), and by west-ern blotting for the presence of α-fetoprotein (AFP), phenylala-nine hydroxylase (PAH), and glutamine synthetase (GS). AFPis a marker for hepatic precursor cells, PAH and GS are mark-ers for mature hepatocytes. (+) Strong expression; (–) little ifany expression. The white arrowhead denotes biliary cells, theblack arrowhead denotes hepatocytic cells.

217-224_25_Tward_Symp70.qxd 5/12/06 9:57 AM 18

MOUSE MODELS FOR HEPATIC TUMORIGENESIS 219

mor progression occurring in response to a distinctivegenotypic change (see below, Conclusions).

GENOTYPIC PROGRESSION IN THE GENESISOF LIVER TUMORS

The morphological progression to HCC elicited by rel-atively low expression of transgenic MET had biochemicaland genomic counterparts. Phosphorylation of the Metprotein was used as a surrogate marker for activation of theprotein-tyrosine kinase. Despite abundant expression ofthe transgenic protein (data not shown), no kinase activitycould be detected in normal liver of the adult mice (Fig. 3),so it appears that the transgenic protein was susceptible toposttranslational control of its enzymatic activity, reminis-cent of that observed for endogenous Met in normal mice(see above). Met kinase activity became apparent in cellsof hyperplastic foci, intensified in dysplastic cells, and wasabundant in end-stage malignancies (Fig. 3).

It seemed likely that additional genetic events con-tributed to tumorigenesis in the model. For hints of whatthese might be, we turned to the example of human HCC.Among the more frequent genetic anomalies in such tu-mors is the mutational activation of β-catenin, a multi-functional protein that mediates signaling to the tran-scriptional apparatus through the Wnt pathway (Giles etal. 2003). In the absence of Wnt signaling, β-catenin is lo-cated in the cytoplasm and at the plasma membrane. Cy-toplasmic β-catenin is situated in a protein complex thatfacilitates phosphorylation and consequent destruction ofthe protein. When activated by signaling from Wnt, how-ever, degradation of β-catenin is inhibited and much ofthe protein moves to the nucleus, where it then serves asa transcription factor. Thus, nuclear localization can beused as a surrogate marker for activation of β-catenin inthe Wnt pathway.

Figure 3. Morphological and molecular progression in the gen-esis of hepatocellular carcinoma elicited by MET. Doxycyclinewas withheld throughout life from a strain of mouse that ex-presses a MET transgene at a relatively low level (fourth proto-col in Fig. 1). Stages in tumor progression were evaluated by mi-croscopy (H+E, as in Fig. 2), and by immunohistochemistry forboth Met kinase activity, using phosphorylation of Met as a sur-rogate for enzymatic activity (Phospho-Met), and activation ofβ-catenin, using nuclear location as an indicator of activity intranscription. Livers were harvested according to the followingschedule: (Control) 2 months of age; (Hyperplastic focus) 1month; (Dysplastic focus) 2 months; and (Hepatocellular carci-noma) 4 months.

Figure 4. Morphological and molecularphenotypes. Hepatic tissue and tumorswere obtained from transgenic mice ei-ther subjected to the fourth protocol inFig. 1 or that had received various genesby hydrodynamic transfection at 8 weeksof age. Transfection was performed as de-scribed by Yant et al. (2000). The genesintroduced by transfection includedcDNA representations of a normal alleleof human MET, a mutant allele of β-catenin that constitutively activates thetranscriptional function of the protein(Barth et al. 1997) (ΔNβ-catenin), and adominant negative allele of HNF-1α(dnHNF-1α)(Vaxillaire et al. 1999).Specimens from liver were analyzed bymicroscopy (H+E); and for the presenceof α-fetoprotein (AFP), phenylalaninehydroxylase (PAH), and glutamine syn-thetase, as in Fig. 2. The white arrow-heads indicate foci of dysplastic cells thatappear following transfection of MET,the black arrowhead denotes cells that aredeficient in PAH following transfectionof dnHNF-1α. Fields of normal tissue aredesignated nt, fields of adenoma, ad.

Some of these animals also developed benign adeno-mas interspersed with nodules of HCC (Figs. 1 and 4 anddata not shown), but in fewer numbers and over a longercourse of time. In contrast to the HCC described above,the adenomas contained no detectable AFP, PAH, or GS.Appearance of the adenomas required relatively ex-tended survival of the animals and may have been limitedby death from concurrent HCC. The nature of the stemcell from which the adenomas arose is not clear. Tumori-genesis could conceivably have originated with the sametype of cell that engendered HCC, with a diversion of tu-


Immunostaining for β-catenin revealed intense con-centration of the protein at the periphery of normal livercells, and the cells of both hyperplastic and dysplasticfoci (Fig. 3). In contrast, the protein was abundant in thenuclei of most cells in the full-blown HCC (Fig. 3). Theactivation of β-catenin in malignant cells was attributableto mutations found in the β-catenin gene (Fig, 5). Themutations were analogous to those found in human HCCand included both point mutations and small deletions.These changes preclude the phosphorylations that elicitthe degradation of β-catenin (Giles et al. 2003) and thusserve to constitutively activate the protein.

Multiple nodules of HCC developed in most affectedlivers. Individual nodules always contained a single mu-tant allele of the β-catenin gene (CTNNB1), indicatingthat the nodule was probably clonal in origin (data notshown). However, the mutations varied from one noduleto another, signaling independent origins of each nodule.The consistency with which the mutations were found intumors suggests that they confer some selective advan-tage during clonal evolution of the emerging tumor cellsand provides strong circumstantial evidence for a role intumorigenesis.

The adenomas in transgenic animals displayed activa-tion of the Met kinase, but not of β-catenin (data notshown). Sequencing of the β-catenin gene in adenomasalso failed to uncover any mutant alleles (data notshown). In search for other possible genetic lesions, weturned again to the example of human tumors. Hepaticadenomas of humans occur in both congenital and spo-radic forms (Zucman-Rossi 2004). In the congenitalform, tumor susceptibility is apparently transmitted by aheterozygous deficiency in the HNF-1α gene (TCF1)(Bacq et al. 2003), and homozygous deficiencies of HNF-1α are frequent in sporadic adenomas (Bluteau etal. 2002). We have failed to find any genomic damage toHNF-1α in the adenomas of the mouse model (data not

shown), but the signaling pathway to which HNF-1α con-tributes appears to be deficient, as manifested by an ab-sence of PAH (Fig. 4), production of which is dependenton signaling from HNF-1α (Pontoglio et al. 1997).

RECONSTRUCTION OF TUMORIGENESIS BYHYDRODYNAMIC TRANSFECTION

Knowing that abnormalities of both Met and β-cateninapparently contributed to the genesis of HCC in the trans-genic mice, we sought to explore their individual roles intumor progression. To this end, we used hydrodynamictransfection to introduce and express ectopic genes in thelivers of adult mice (Liu et al. 1999; Zhang et al. 1999;Yant et al. 2000). We used a wild-type allele of MET, asin the transgenic model, and a mutant allele of the β-catenin gene that encodes a constitutively active protein(Barth et al. 1997). The results are summarized in Figure6. Introduction of MET alone gave rise to microscopic fociof slightly dysplastic cells (Fig. 4), whereas transfection ofthe β-catenin gene had no apparent effect (data notshown). When the two genes were introduced and ex-pressed together, the recipient animals developed HCCwithin a month or so (Figs. 4 and 6). Moreover, sequentialintroduction of the two genes at an interval of one monthproduced tumors, and the sequence of introduction had noeffect on the outcome (data not shown). Thus, cooperationbetween MET and the β-catenin gene in the genesis ofHCC appears not to be strictly dependent on the chrono-logical sequence with which the two genes go awry.

We were also able to reconstruct the genesis of hepaticadenomas by hydrodynamic transfection. To do this, wecreated a dominant negative allele of HNF-1α found inhuman hepatic adenomas (Vaxillaire et al. 1999), whichcan mimic the effect of a genetic deficiency in the en-dogenous gene. Transfection of the inhibitory allele ofHNF-1α produced numerous individual cells that were

220 TWARD ET AL.

Figure 5. Mutations in the β-catenin gene of human and mouse hepatocellular carcinomas. HCC was elicited by the protocol de-scribed for Fig. 3. The β-catenin gene was sequenced in 21 HCCs obtained from 15 mice. An activating mutation was found in all butone HCC. Representative results are shown in yellow. Representative data reported by others for human tumors are given in red. Ser-ine and threonine residues whose phosphorylations elicit degradation of β-catenin are shown in green. The data for human tumorswere derived from Buendia (2000).


deficient in PAH (Fig. 4), but no tumors (Fig. 6). WhenMET and the dominant negative HNF-1α were introducedand expressed together, however, they rapidly gave rise toadenomas (Figs. 4 and 6). The tumors displayed histolog-ical morphology and differentiation markers identical tothose of the adenomas found in transgenic mice (Fig. 4).

Previous work demonstrated that the targeted expres-sion of transgenic normal MYC in liver cells gave rise toa primitive tumor known as hepatoblastoma (Shachaf etal. 2004). The same tumor occurred when hydrodynamictransfection was used to express a normal allele of humanMYC in the liver of adult mice (Fig. 6). The tumors aroserapidly, without any readily apparent intermediate stage,and contained neither detectable Met kinase activity normutations in the β-catenin gene (data not shown).

REGRESSION OF HCC AFTER INACTIVATIONOF THE INITIATING TRANSGENE

The prevailing scheme of tumor progression raises thequestion of which steps in that progression remain essen-tial for maintenance of the eventual malignancy and, thus,might be suitable targets for therapeutic intervention. Re-cent efforts to address this question with mouse modelshave shown that many forms of malignancy initiated bytransgenes will regress when the responsible transgene isinactivated, even if the tumor has reached an advancedstage (Felsher 2004). We previously reported that this istrue for HCC elicited by MET (Wang et al. 2001). The re-gression was mediated by both a cessation of tumor cellproliferation and widespread apoptosis, accompanied byregeneration of normal liver tissue.

The regenerated liver contained microscopic foci ofresidual tumor surrounded by normal liver tissue (Fig.7A). As anticipated, there was no detectable Met proteinor kinase activity in either the residual tumor cells or thenormal tissue (Fig. 7C and data not shown). Nuclear β-catenin was apparent in the residual tumor cells, but notin the normal tissue (Fig. 7B). Thus, the liver had appar-ently been reconstituted from normal stem cells or hepa-tocytes, rather than by differentiation of malignant cellsin the manner reported for the hepatoblastomas elicitedby a MYC transgene (Shachaf et al. 2004).

When held for 6 months or more with the MET trans-gene inactivated, the animals often suffered a clinical re-lapse, again developing fatal HCC (data not shown). Thetumor cells still contained activated β-catenin, but therewas no indication that the MET transgene had been reacti-vated (data not shown). We conclude that the relapseswere driven by a new and different genetic event that by-passed the requirement for Met kinase and presumably co-operated with the activity of β-catenin in tumorigenesis.

CONCLUSIONS

We have developed mouse models for both benign andmalignant liver tumors. The results support a role forMET in the genesis of liver tumors by complementingprevious circumstantial evidence from studies of humantumors; provide insight into the rules of tumor progres-sion; dramatize how different cellular compartmentswithin a developmental lineage can give rise to distinc-tive tumors in response to the same tumorigenic influ-ence, presumably by generating distinctive tumor stemcells; and provide evidence that the experimental tumorsin mice are authentic models for human tumors.

We initiated tumorigenesis by overexpression of a nor-mal allele of the MET proto-oncogene. Although thisanomaly is common in human HCC, it apparently liesdownstream of an external cause, such as chronic infec-tion with either hepatitis B or C virus. It is not presentlyclear how viral infection might lead to the genetic abnor-malities required for tumorigenesis. One hypothesisholds that ongoing death and regeneration of liver tissuein response to chronic infection sets the stage for the oc-currence and fixation of genetic damage (Thorgeirssonand Grisham 2002). This view is consistent with thelengthy time that separates the onset of infection from theappearance of malignant tumors. In contrast, tumorigen-esis in the mouse models is relatively quick, with tumorsarising after a few months at the most. At least three fac-tors may contribute to this difference.

First, by introducing the overexpression of MET intonumerous, if not all, liver cells, we have both bypassedearlier events that would otherwise be necessary for tu-morigenesis and increased the likelihood of subsequent


Figure 6. Tumorigenesis following hydrodynamic transfection.Hydrodynamic transfection was used to achieve expression ofectopic genes in the livers of 8-week-old mice, as described forFig. 4. Transfection was also performed with a normal allele ofMYC. The resulting tumors included HCC, benign adenomas,and a primitive tumor known as hepatoblastoma (HBL), previ-ously elicited with a transgene of MYC (Shachaf et al. 2004).

Figure 7. Reconstitution of the normal liver after regression ofHCC in mice. HCC was elicited with the protocol described forFig. 3. When the tumors reached an advanced state, doxycyclinewas administered to elicit regression, as described previously(Wang et al. 2001). Liver was harvested for examination 6months following the initiation of doxycycline administration,then examined by microscopy (A), and by immunohistochem-istry for β-catenin (B) and Met kinase (C). Fields of normal cellsare designated by nl, fields of residual tumor cells by rt.

Gene(s) transfected Tumor

Met –

ΔNβ-Catenin –

dnHNF1α –

Met + ΔNβ-Catenin HCC

Met + dnHNF1α Adenoma

Myc HBL


events by creating a vast number of potential tumor stemcells. It is also possible that sustained activity of METmight destabilize the genome, much as described forMYC (Mai et al. 1996; Felsher and Bishop 1999). Second,the rapidity of tumorigenesis in the models is in accordwith the natural history of cancer in mice, which developspontaneous malignancies much more quickly than dohumans (Campisi 2003). This difference may reflect in-ferior defenses against permanent genetic damage in ro-dents (Ames et al. 1993). Third, cells in the dysplasticfoci induced by MET were proliferating (data not shown),providing a permissive environment for the occurrenceand fixation of mutations.

Although overexpression of MET and mutational acti-vation of β-catenin are common in human HCC, neitheris universal. In contrast, the mouse HCC described herecontained both abundant Met kinase and activating muta-tions of β-catenin, almost without exception. The formerwas by experimental design; the latter must have arisenfrom spontaneous mutagenesis followed by selection dur-ing the course of tumor progression. We suggest that inconstructing the mouse model, we have imposed one ofseveral possible genetic pathways to HCC.

In pursuit of this idea, we have surveyed more than 50human HCCs for the presence of enzymatically activeMet and activating mutations of β-catenin. The results in-dicate that the two were frequently congruent, and in par-ticular, robust Met kinase activity was virtually neverpresent in the absence of a β-catenin mutation (A.D.Tward et al., unpubl.). These correlations have not previ-ously been appreciated. Thus, the findings in the mousemodel have led us to a subset of human tumors in whichthe course of tumorigenesis was apparently akin to that inthe model. The remainder of human HCC would arisefrom at least one additional pathway that has not beenreplicated by our mouse model. A further inferencewould be that activation of Met kinase constrains the sub-sequent pathway of tumorigenesis.

We have not detected Met kinase activity in the normalcells of adult transgenic livers (Fig. 3 and data notshown). Activity became apparent only with the appear-ance of hyperplastic foci and then persisted throughoutthe course of tumorigenesis. We presume that only occa-sional liver cells express the MET transgene above thethreshold required to initiate tumorigenesis, and that suchcells give rise to the hyperplastic foci.

The tumors in our mouse models all arose from overex-pression of the MET proto-oncogene, but the specific out-come of tumorigenesis was determined by which geneticanomalies occur subsequently. We dramatized this con-clusion by using hydrodynamic transfection to introducevarious genetic anomalies into the mouse liver in vivo.The results demonstrated that MET and β-catenin cooper-ate to produce HCC, whereas MET and a deficiency ofHNF-1α cooperate to produce benign adenomas. We havefound no evidence that the abnormalities of β-catenin andHNF-1α ever coincide in tumors, and we cannot presentlysay whether either of these preempts the other, should theyoccur coincidentally in the same cell, but the issue isamenable to study by hydrodynamic transfection.

The determining nature of genetic lesions in tumorige-nesis is further illustrated by the fact that targeted expres-sion of MYC in liver cells, using the same control elementas in the present work, gave rise to hepatoblastomas,rather than either HCC or adenomas (Shachaf et al. 2004and data not shown). We were able to recapitulate thatfinding by using hydrodynamic transfection to express anormal allele of human MYC in the liver of adult mice.The transfected mice developed hepatoblastoma.

The reconstruction of tumorigenesis in the transgenicmouse models places overexpression of Met upstream ofthe other genetic events required for the genesis of eitherHCC or adenomas. We have not conducted an exhaustivesearch for such cooperating events, focusing instead onknown candidates deduced from data obtained with hu-man tumors. However, the rapidity with which mouse tu-mors arise after simultaneous transfection of MET andmutants of either β-catenin or HNF-1α suggests that, inthe mouse at least, relatively few additional spontaneousevents may be required. Our results with hydrodynamictransfection also raise the possibility that HCC will ariseirrespective of the order in which overexpression of Metand mutation of β-catenin occur, but more extensive ex-perimentation will be required to address this issue in adefinitive manner.

A major objective of modeling tumorigenesis in themouse is to identify individual steps in tumor progressionthat are suitable targets for therapeutic intervention. Ourwork with the mouse model for HCC illustrates thisprospect. Inactivation of MET, even in the most advancedtumors, led to prompt and universal tumor regression.Thus, MET not only initiated tumorigenesis, but alsoserved as a tumor maintenance function. The efficiencyand relative durability of regression were presumably dueto the permanent removal of Met kinase from the tumorcells. The mouse tumors do recur, but only after extendedperiods of time, and only after the advent of some eventthat can bypass the need for MET and, perhaps, exploitthe presence of residual cells with mutant β-catenin.These results highlight Met as a possible target in thetreatment of HCC. They also dramatize how occurrenceof mutations that bypass any requirement for the targetcould give rise to drug resistance during the course oftherapy.

In aggregate, this and many other reports of similar re-sults with mouse models further justify the pursuit oftherapeutics that are targeted at single genetic lesions inhuman tumors—therapies of the sort represented by thelandmark pharmaceutical, Gleevec (Sawyers 2004).Mouse models with the authenticity exemplified by thepresent work should be useful in evaluating both the suit-ability of individual targets and the potential efficacy ofadditional targeted therapies, once they are in hand.

It is now apparent that MET is a crucial switch in the de-velopment of the liver (Fig. 8). The function of the gene isrequired to sustain proliferation of hepatic progenitor cells,but it must be inactivated in order to permit differentiationof those cells (A.D. Tward et al., in prep.). Unphysiologi-cal activity of the Met kinase can disturb the differentiationof hepatocytes and engender tumorigenesis from both

222 TWARD ET AL.


bipotential and unipotential cells within the hepatocyte lin-eage, as dramatized by the work presented here.

We cannot discern whether the effects of Met on pro-genitor proliferation and differentiation are independentof one another. Alternatively, one of the effects might fol-low the other. For example, the mere withdrawal of pro-genitor cells from the cell cycle after depletion of Met ki-nase activity might itself trigger differentiation.Whatever its genesis, the massive hyperplasia of progen-itor cells in response to the in utero and postnatal overex-pression of MET could provide an abundant source ofsuch cells for the experimental study of hepatic differen-tiation, tumorigenesis, and tissue engineering.

ACKNOWLEDGMENTS

We thank Linda Prentice for assistance with histology,and Luda Urisman for assistance with animal husbandry.The work was supported by funds from the National Can-cer Institute (CA009043), the George W. Hooper Re-search Foundation, and the National Institutes of Health(DK49022 to M.A.K.). ADT was supported by the UCSFMedical Scientist Training Program NIH NIGMS(5T32GMO7618).

REFERENCES

Ames B.N., Shigenaga M.K., and Hagen T.M. 1993. Oxidants,antioxidants, and the degenerative disease of aging. Proc.Natl. Acad. Sci. 90: 7915.

Bacq Y., Jacquemin E., Balabaud C., Jeannot E., Scotto B.,Branchereau S., Laurent C., Bourlier P., Pariente D., de MuretA., Fabre M., Bioulac-Sage P., and Zucman-Rossi J. 2003. Fa-milial liver adenomatosis associated with hepatocyte nuclearfactor 1 alpha incactivation. Gastroenterology 125: 1470.

Barth A.I., Pollack A.L., Altschuler Y., Mostov K.E., and Nel-

son W.J. 1997. NH2-terminal deletion of beta-catenin resultsin stable colocalization of mutant beta-catenin with adenoma-tous polyposis coli protein and altered MDCK cell adhesion.J. Cell Biol. 136: 693.

Bluteau O., Jeannot E., Bioulac-Sage P., Marques J.M., BlancJ.F., Bui H., Beaudoin J.C., Franco D., Balabaud C., Laurent-Puig P., and Zucman-Rossi J. 2002. Bi-allelic inactivation ofTCF1 in hepatic adenomas. Nat. Genet. 32: 312.

Buendia M.A. 2000. Genetics of hepatocellular carcinoma.Semin. Cancer Biol. 10: 185.

Campisi J. 2003. Cancer and ageing: Rival demons? Nat. Rev.Cancer 3: 339.

Felsher D.W. 2004. Reversibility of oncogene-induced cancer.Curr. Opin. Genet. Dev. 14: 37.

Felsher D.W. and Bishop J.M. 1999. Transient excess of MYCactivity can elicit genomic instability and tumorigenesis.Proc. Natl. Acad. Sci. 96: 3940.

Giles R.H., vas Es J.H., and Clevers H. 2003. Caught up in a Wntstorm: Wnt signalling in cancer. Biochim Biophys Acta. 1653:1.

Ishikawa K.S., Masui T., Ishikawa K., and Shiojiri N. 2001. Im-munolocalization of hepatocyte growth factor and its receptor(c-Met) during mouse liver development. Histochem. CellBiol. 116: 453.

Kinzler K.W. and Vogelstein B. 1996. Lessons from hereditarycolon cancer. Cell 87: 159.

Liu F., Song Y., and Liu D. 1999. Hydrodynamics-based trans-fection in animals by systemic administration of plasmidDNA. Gene Ther. 6: 1258.

Llovet J.M., Burroughs A., and Bruix J. 2003. Hepatocellularcarcinoma. Lancet 362: 1907.

Mai S., Fluri M., Siwarski D., and Huppi K. 1996. Genomic in-stability in MycER-activated Rat1A-MycER cells. Chromo-some Res. 4: 365.

Nowell P.C. 1976. The clonal evolution of tumor cell popula-tions. Science 194: 23.

Parkin D.M., Bray F., Ferlay J., and Pisani P. 2005. Global can-cer statistics, 2002. CA Cancer J. Clin. 55: 74.

Pontoglio M., Faust D.M., Doyen A., Yaniv M., and Weiss M.C.1997. Hepatocyte nuclear factor 1alpha gene inactivation im-pairs chromatic remodeling and demethylation of the pheny-lalanine hydroxylase gene. Mol. Cell. Biol. 17: 4948.


Figure 8. Pathways of tumor progression. The diagram is based on results for tumor progression reported in the present paper andelsewhere (Wang et al. 2001; Shachaf et al. 2004) and on results describing the role of Met in hepatic development, which demon-strate that the physiological activity of Met both drives proliferation and blocks differentiation of hepatic precursor cells (Suzuki etal. 2003, and in prep.).


Sawyers C. 2004. Targeted cancer therapy. Nature 432: 294.Shachaf C.M., Kopelman A.M., Arvanitis C., Karlsson A., Beer

S., Mandl S., Bachmann M.H., Borowsky A.D., Ruebner B.,Cardiff R.D., Yang Q., Bishop J.M., Contag C.H., and FelsherD.W. 2004. MYC inactivation uncovers pluripotent differen-tiation and tumour dormancy in hepatocellular cancer. Nature431: 1112.

Suzuki A., Iwama A., Miyashita H., Nakauchi H., and TaniguchiH. 2003. Role of growth factors and extracellular matrix incontrolling differentiation of prospectively isolated hepaticstem cells. Development. 130: 2513.

Tavian D., De Petro G., Benetti A., Portolani N., Giulini S.M.,and Barlati S. 2000. u-PA and c-MET mRNA expression isco-ordinately enhanced while hepatocyte growth factormRNA is down-regulated in human hepatocellular carci-noma. Int. J. Cancer 87: 644.

Thorgeirsson S.S. and Grisham J.W. 2002. Molecular pathogen-esis of human hepatocellular carcinoma. Nature Genetics 31:339.

Trusolino L. and Comoglio P.M. 2002. Scatter-factor and

semaphorin receptors: Cell signalling for invasive growth.Nat. Rev. Cancer 2: 289.

Vaxillaire M., Abderrahmani A., Boutin P., Bailleul B., FroguelP., Yaniv M., and Pontoglio M. 1999. Anatomy of a homeo-protein revealed by the analysis of human MODY3 muta-tions. J. Biol. Chem. 274: 35646.

Wang R., Ferrell L.D., Faouzi S., Maher J.J., and Bishop J.M.2001. Activation of the Met receptor by cell attachment in-duces and sustains hepatocellular carcinomas in transgenicmice. J. Cell Biol. 153: 1023.

Yant S.R., Meuse L., Chiu W., Ivics Z., Izsvak Z., and Kay M.A.2000. Somatic integration and long-term transgene expres-sion in normal and haemophilic mice using a DNA transposonsystem. Nat. Genet. 25: 35.

Zucman-Rossi J. 2004. Genetic alterations in hepatocellular ade-nomas: Recent findings and new challenges. J. Hepatol. 40:1036.

Zhang G., Budker V., and Wolff J.A. 1999. High levels of for-eign gene expression in hepatocytes after tail vein injectionsof naked plasmid DNA. Hum. Gene Ther. 10: 1735.

224 TWARD ET AL.


genomic progression in mouse models for liver tumors

Documents