Research Article

Serum-based tracking of de novo initiated liver cancerprogression reveals early immunoregulation and response to therapy

Jeff J. Subleski1, Anthony J. Scarzello1, W. Gregory Alvord2, Qun Jiang1, Jimmy K. Stauffer1,Anthony Kronfli1, Bahara Saleh1, Timothy Back1, Jonathan M. Weiss1, Robert H. Wiltrout1,⇑

1Cancer and Inflammation Program, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Frederick, MD 21702,United States; 2Statistical Consulting, Data Management Services, Inc., Frederick National Laboratory for Cancer Research, Frederick, MD,

United States

Abstract

Background & Aims: Liver inflammatory diseases associatedwith cancer promoting somatic oncogene mutations are increas-ing in frequency. Preclinical cancer models that allow for thestudy of early tumor progression are often protracted, whichlimits the experimental study parameters due to time and expense.Here we report a robust inexpensive approach using SleepingBeauty transposition (SBT) delivery of oncogenes along withGaussia Luciferase expression vector GLuc, to assess de novo livertumor progression, as well as the detection of innate immuneresponses or responses induced by therapeutic intervention.Methods: Tracking de novo liver tumor progression with GLucwas demonstrated in models of hepatocellular carcinoma (HCC)or adenoma (HCA) initiated by hydrodynamic delivery of SBToncogenes.Results: Rising serum luciferase levels correlated directly withincreasing liver tumor burden and eventual morbidity. Earlydetection of hepatocyte apoptosis from mice with MET+CATtransfected hepatocytes was associated with a transient delayin HCC growth mediated by a CD8+ T-cell response against trans-formed hepatocytes. Furthermore, mice that lack B cells ormacrophages had an increase in TUNEL+ hepatocytes followingliver MET transfection demonstrating that these cells provideprotection from MET-induced hepatocyte apoptosis. Treatmentwith IL-18+IL-12 of mice displaying established HCC decreasedtumor burden which was associated with decreased levels ofserum luciferase.

Journal of Hepatology 20

Keywords: Sleeping Beauty Transposition; HCC; HCA; Gaussia Luciferase;Real-time tracking; Treatment.Received 2 July 2014; received in revised form 15 June 2015; accepted 20 June 2015;available online 2 July 2015⇑ Corresponding author. Address: Experimental Therapeutics Section, Cancer andInflammation Program, Center for Cancer Research, National Cancer Institute,National Institutes of Health, Frederick, MD 21702, United States. Tel.: +1 (301)846 1584; fax: +1 (301) 846 6016.E-mail address: wiltrour@mail.nih.gov (R.H. Wiltrout).Abbreviations: SBT, Sleeping Beauty Transposition; GLuc, Gaussia Luciferaseexpression SBT vector; HCC, hepatocellular carcinoma; HCA, hepatocellularadenoma; GLuc+ hepatocytes, GLuc vector integrated in hepatocytes; HSB2,Sleeping Beauty transposase; MET, human c-Met; PT3, empty SBT vector; CAT,D90Nb-catenin; AKT, Myr-AKT; ConA, concanavalin A; WT, wild-type.

Conclusions: Hydrodynamic delivery of the SBT vector GLuc tohepatocytes serves as a simple blood-based approach forreal-time tracking of pathologically distinct types of liver cancer.This revealed tumor-induced immunologic responses and wasbeneficial in monitoring the efficacy of therapeutic interventions.Published by Elsevier B.V. on behalf of the European Associationfor the Study of the Liver. Open access under CC BY-NC-ND license.

Introduction

Up to 70–90% of all liver cancers are detected in patients suffer-ing from chronic inflammatory liver diseases due to obesity,chronic hepatitis C and liver fluke infections that increase the fre-quency of deleterious somatic mutations in oncogenes and tumorsuppressor genes [1–9]. Because liver inflammatory diseases dueto obesity [10] and hepatitis C infections [11] have a globallyincreasing frequency, it is important to understand whichsomatic mutations in oncogenes drive liver cancer developmentand/or progression and how inflammation facilitates this process.

The advent of Sleeping Beauty transposition (SBT) technologyhas provided experimental tools to investigate how mutationsof oncogenes lead to the development of liver cancer [12,13]. Thisnon-viral-based gene transfection system can be used to induceliver tumors by stably integrating oncogenes into hepatocytes.The induction of tumors using the SBT system allows the analysisof how inflammation and tumor immunity regulate the entirespectrum of tumor growth including the early stages of neoplasia,which cannot be addressed using transplantable tumors.Although carcinogens such as diethylnitrosamine (DEN) alsoallow us to study liver tumor formation, butl often induces amyriad of undefined mutations and off target events, while theSBT system can be best used to study how specific oncogenescollaborate in the formation of tumors.

Mouse cancer models using SBT, carcinogens and transgenicmice are amenable for the study of tumor formation at the earlystages of neoplasia but are limited by the variable and often pro-tracted time for development/progression of tumors, dependingon the oncogene combinations used for initiation. In vivo imagingcould be used to follow tumor progression but this approach canbe prohibitively expensive, limited by the number of mice that

15 vol. 63 j 1181–1189

Research Article

can be scanned and requires dedicated staff to run complexinstrumentation. Thus, a tumor-tracking method that could beused sequentially in individual mice over a long period of timewould allow an assessment of tumor status in real-time. Suchan approach may reveal therapeutic windows, be used to extrap-olate tumor burden and survival times and potentially elucidateimmunoregulatory events.

In this study, we investigated the use of the SBT GLuc vectorfor tracking nonlinear oncogene-induced liver tumor progression.Using two distinct de novo liver tumor models, we demonstratedthat the level of serum luciferase released by GLuc vectorintegrated in hepatocytes (GLuc+ hepatocytes) significantlycorrelated with the volume of liver tumors, thereby providing aminimally-invasive method to track in living mice the growthof de novo-generated liver tumors. Using this minimally-invasive method, parameters can be measured that would other-wise require euthanasia of mouse cohorts with an associatedhigher experimental cost. The evaluation of immunogenic denovo liver tumors using GLuc+ hepatocytes also allowed us toidentify inflection points that revealed CD8+ T-cell mediatedtransient tumor inhibitory effects during tumor progression.Concordantly, we show for the first time that stable expressionof luciferase by hepatocytes can also be used to efficiently track,in real-time, the outcome of therapies against oncogene-drivende novo liver tumors.

Materials and methods

Plasmid constructs

Plasmids encoding the Sleeping Beauty transposase (HSB2) and transposons con-taining human c-Met (MET) and empty vector control (PT3), were a generous giftfrom Dr. Xin Chen (UCSF, San Francisco, Ca.). Transposon constructs containingD90Nb-catenin (CAT) and Myr-AKT (AKT) were previously reported [14]. Thegene for Gaussia Luciferase was PCR-amplified and cloned into pENTR(Invitrogen) and subsequently Gateway-cloned into the PT3 destination vectorto produce a Gaussia luciferase transposon (GLuc) vector.

Experimental model

C57/BL6J (wild-type [WT]), B6.129S2-Cd8a (CD8�/�), B6.129S2-Cd4/J (CD4�/�),B6.129S2-lgh-6 (Bcell�/�) and CD1d�/� (NKT�/�) mice were purchased from TheJackson Laboratory and were bred and maintained at the Frederick NationalLaboratory in a specific pathogen free facility in accordance with an approvedAnimal Care and Use Protocol. Oncogene constructs were delivered to hepato-cytes using the hydrodynamic transfection technique. Briefly, constructscontaining either MET+CAT or AKT+CAT along with HSB and GLuc vector weremixed in saline at a volume of 10% vol/mouse weight and injected via tail veininto aged and sex matched 8–12 week old C57/BL6J or B6.129S2-Cd8a mice over5–7 seconds as previously described [12]. MET+CAT induces liver tumors with apathology consistent with hepatocellular carcinoma (HCC) while AKT+CATinduces liver tumors histopathologically characterized as hepatocellular adenoma(HCA) [14].

Immunohistochemistry (IHC) staining and analysis

Liver tissues were fixed with 10% normal buffered formalin overnight and thentransferred to 70% ethanol. Paraffin blocks were made from the fixed tissues.Slides cut from the blocks were analyzed for apoptosis using TUNEL staining, pro-liferation using BrdU at 1:50 (Invitrogen Life Technologies), b-catenin at 1:200(Abcam) and N-ras (F155 Santa Cruz). For IHC slide analysis, fivenon-overlapping pictures were taken from at least two separate lobes and imageswere analyzed using CellProfiler (http://www.cellprofiler.org/) with modifiedKi67 pipeline.

1182 Journal of Hepatology 2015

Immunotherapeutic regimen

Recombinant murine IL-12 and IL-18 was purchased from PeproTech Inc. (RockyHill, NJ). Stock aliquots of cytokines were diluted with HBSS and mice wereinjected intraperitoneally with vehicle control (VC) or IL-12 (0.3 lg) + IL-18(0.3 lg) on days 20–24 and 28–31 post-initiation of liver tumors with MET+CAToncogenes.

qPCR quantitation of oncogenes in the liver

Livers were harvested and snap frozen on a dry ice and 2-Methylbutane mixturefor later manipulations. Frozen livers were mechanically shattered into smallfragments and genomic DNA was isolated using a Gentra Puregene Tissue Kit(Qiagen) according to manufactures instructions. Oncogene copy number per100 ng of liver DNA was determined using qPCR analysis with TaqManprobes MET (Hs 01564484_m1), CTNNB1 (CAT) (Hs 00355049_m1), and AKT(mm 01331626_m1) (ABI) and comparing it to a standard curve made usingthe oncogene plasmids.

Tumor assessment by magnetic resonance imaging (MRI)

Animal imaging was performed by Frederick National Laboratory Small AnimalImaging Program. MRI was performed with a 3.0T clinical scanner (Philips InteraAchieva, Philips Medical Systems, Einthoven, NL) to detect the onset of tumors,monitor their progression, and track the changes in the liver volume biweekly.Data was acquired using a 40-mm diameter solenoid receiver coil (PhilipsResearch, Hamburg, Germany). The mice were anesthetized in an inductionchamber with 3% isoflurane in O2 at 1 L/min flow rate and then placed in a custommade mouse holder within the imaging coil. During an imaging session, theanesthesia level was modified between 1.5–2.0% isoflurane to maintain a40 bpm pulmonary rate, and their physiology monitored (Biopac System Inc.,Goleta, CA). Body temperature was maintained at 37 �C by supplying warm air(Small Animal Instruments, Inc., Stony Brook, NY) around the mouse holder. Multislice T2 weighted turbo spin echo (T2w-TSE) sequence was applied in coronalview with respiratory triggering to minimize the motion artifacts. An imagingvolume of 36 � 27 � 18 mm was chosen to cover the mouse abdominal cavity.The images were acquired with a repetition time (TR) 4500 ms, echo time (TE)65 ms, with an in plane resolution of 0.190 � 0.190 mm, and slice thickness0.5 mm. Serial images obtained in 8 biweekly consecutive imaging sessions wereused to calculate the changes in tumor and liver volumes.

Gaussia Luciferase assay

Mouse blood was collected using retro orbital bleeds in a serum separator tube(Becton Dickinson co.) and then serum was recovered by centrifuging tubes for3 minutes at 20,000 g and then 5 ll of serum was added to 100 ll of PBS and pla-ted on a 96 well flat bottom white polystyrene assay plate (Costar 3362). Serumluciferase levels were determined using the BioLux Gaussia Luciferase Assay kit(New England BioLabs) according to manufacturer’s direction and luminescencemeasurements were acquired using FLUOstar Omega microplate reader (BMGLABTECH) after controlled injection of 50 ll of substrate mixture into platescontaining the serum PBS mixture.

Liver leukocyte isolation and flow cytometric analysis

Liver leukocytes were isolated as previously described [15]. Isolated leukocyteswere counted and 1 � 106 cells were Fc-blocked with monoclonal 2.4G2 priorto staining with appropriately titered monoclonal antibodies (CD8a PerCP andPD-1 PE; eBioscience) used in conjunction with Live/Dead reagent; Invitrogen).Data was collected using a LSRII Special Order System equipped with solid stateblue (488 nm), red (640 nm) and violet (405 nm) lasers and then analyzed usingFloJo Vx0.6.

Transfection of mouse hepatoma Hepa1–6 cells

Hepa1–6 cells were maintained in DMEM containing 1.45 g of glucose/liter(Sigma), supplemented with 2 mM glutamine, 50 lg/ml gentamycin sulfate,and 10% (v/v) heat-inactivated fetal calf serum. Targefect-Hepatocyte reagent(Targeting Systems, CA) was used for PT3-GFP or PT3-MET plasmid transfectioninto Hepa1–6 cells according to the manufacturer’s protocol. Cells were harvested

vol. 63 j 1181–1189

JOURNAL OF HEPATOLOGY

24 h or 72 h post-transfection for apoptosis detection. Cell apoptosis was detectedusing PE Annexin V Apoptosis Detection Kit I according to manufacturer’s proto-col (BD Bioscience).

Statistical analysis

Data in this study were analyzed with linear mixed effects hierarchical models,analysis of variance and covariance, Kaplan-Meier and Cox Proportional Hazardsregression methods, correlation and (weighted) regression analysis, post hoc tests(e.g., Tukey’s test for multiple comparisons) and non-parametric methods (e.g.,Wilcoxon’s test). Repeated measures analyses (in which ‘repeated’ measurementsare recorded for the same animal on more than one occasion) and mixed effectsmodels take into consideration the correlation/covariation of responses withinthe same animal over time [16,17]. Homogeneity of variance and covarianceassumptions were routinely tested in model fits. Probability values less than0.05 were considered significant.

Results

Serum GLuc levels correlate with MET+CAT oncogene-induced livertumor burden

Ectopic expression of foreign proteins in mice can lead to animmune response against the protein and the tissues expressingthe protein [18,19]. To determine whether secreted GLuc could bestably expressed in the liver and not immunologically rejected,we delivered GLuc expression vector hydrodynamically to theliver via SBT, with or without transposase (HSB2) [12]. Trans-posase HSB2 causes genomic insertion of the transposon ran-domly at AT-dinucleotide sites. In the absence of HSB2, theGLuc transposon does not integrate into the hepatocyte genome.Serum expression kinetics of luciferase was analyzed in C57/Bl6mice injected with GLuc (non-integrated) or GLuc+HSB2 vectors(integrated) (Supplementary Fig. 1). Stable constitutive expres-sion of luciferase was observed by day 3 in mice givenGLuc+HSB2. The levels of serum Gaussia Luciferase were signifi-cantly higher in mice given GLuc+HSB2 vectors compared to micegiven GLuc vector alone (post hoc Tukey’s test; p <0.0001). Con-versely, there was no significant difference in luciferase levelsafter day 3 between the serum of mice given GLuc alone and con-trol serum from naïve mice. These data show that hepatocyteswith integrated GLuc vector (GLuc+ hepatocytes) stably expressconstitutive levels of luciferase while luciferase levels ofnon-integrated GLuc are rapidly lost, returning to baseline levelby day 3.

Because hepatocytes with integrated GLuc stably expressluciferase under homeostatic conditions, we next tested whetherthis technique could be used to track hepatocyte proliferation.We hypothesized that the constitutively expressed levels ofluciferase should rise when GLuc+ hepatocytes are induced byoncogenes to proliferate. De novo HCC formation was initiatedby hydrodynamic transfection of 4 lg of GLuc vector and escalat-ing concentrations of MET+CAT oncogenes in combination withHSB2 [14]. At days 30 and 56 mice that received MET+CAT dosesof 25, 12.5, and 5 lg showed significantly higher serum luciferaselevels compared to mice injected with PT3 controls or lower con-centration of MET+CAT (Supplementary Fig. 2A). To confirm thatthe significant rise in luciferase levels with the higher MET+CAToncogene groups was due to tumor burden, mice wereeuthanized at day 55 and tumor-bearing livers were weighed(Supplementary Fig. 2B). A proportional relationship wasobserved between the amount of MET+CAT oncogenes used to

Journal of Hepatology 2015

initiate tumors and their respective tumor-bearing liver weights(Supplementary Fig. 2C). Linear regression analysis revealed asignificant correlation between serum luciferase levels and theweights of the tumor-bearing livers. To specifically determine ifserum luciferase levels correlated with liver tumor burden, wequantitated the amount of integrated oncogenes in liver tissueusing qPCR and compared it with the levels of serum luciferase.The copy number of MET and CAT oncogenes in 100 ng of geno-mic DNA was then plotted against serum luciferase levels fromthese same mice at day 55 (Supplementary Fig. 2D, E). Linearregression analysis showed a significant correlation betweenthe liver levels of either oncogene and serum luciferase levels.

Progression of AKT+CAT-initiated liver tumors originated in GLuc+

hepatocytes is accurately tracked by serum luciferase levels

To validate that hepatocytes with integrated GLuc vector couldalso be used to track alternative models of de novo-initiated livertumors, we used AKT+CAT oncogenes that induce liver tumors(HCA) that are histopathologically distinct from those inducedby MET+CAT (HCC) [14]. Furthermore, we wanted to determinewhether GLuc+ hepatocytes could be used to track liver tumorprogression throughout the oncogenic process. This is criticalbecause tumors can grow in complex three-dimensional nonlin-ear tracks [20,21]. Mice were hydrodynamically transfected withGLuc vector along with AKT+CAT oncogenes and HSB2. Tumorprogression was analyzed in real-time by measuring liver sizeusing MRI analysis on days 27, 41, 59, 66, 95, and 103. At day27, normal liver morphology was evident. However, asAKT+CAT-initiated tumors progress to day 103, tumor nodulesdevelop causing the liver to distend (Supplementary Fig. 3A).Serum luciferase levels were quantitated on the same days ofMRI examination and a comparative analysis was done betweenthese two parameters (Supplementary Fig. 3B). Both MRIliver analysis and serum luciferase levels revealed thatAKT+CAT-initiated liver tumors progress in a nonlinear manner.Pearson correlations between serum luciferase levels and tumorsizes for all three animals are extremely high: r4401 = 0.99(p <0.0001), r4403 = 0.98 (p = 0.0005), and r4404 = 0.99(p <0.0001). In addition, analysis of covariance showed that foreach animal, the proportionate relative luminescence units andliver size measures were statistically parallel, demonstratingthe association between induced size changes in liver tumorsand levels of serum luciferase. Taken together, these data demon-strate that a high degree of correlation exists between liver tumorgrowth kinetics and levels of serum luciferase. Moreover, GLuc+

hepatocytes can be used to track the nonlinear progression ofde novo-initiated liver tumors.

Serum luciferase levels detect growth rate changes in oncogene-induced liver tumors

The changes in kinetics of tumor progression relative to morbid-ity can provide insights into the diversity of basic biologicalmechanisms and/or immune recognition of tumors initiated bydifferent oncogenic events. We tested whether integration ofGLuc into hepatocytes provides a useful tool for detecting subtlevariations in tumor progression of pathologically distinct HCA(AKT+CAT) or HCC (MET+CAT) [14]. By day 340, 90% and 70% ofmice transfected with AKT+CAT and MET+CAT, respectively, hadsuccumbed to the tumors (Fig. 1A), while all the mice given the

vol. 63 j 1181–1189 1183

Day

Per

cent

surv

ival

0 100 200 300 4000%

50%

100%

ControlMET-CATAKT-CAT

50 100 150 2000

2×106

4×106

6×106

8×106

107

ControlMET-CATAKT-CAT

Day

RLU

0

106

2×106

3×106

4×106

ControlMET-CATAKT-CAT

Day

RLU

0 10 20 30 40 500

MET CAT

No

.of T

UN

EL+

hepa

tocy

tes

WT

CD1d(-/

-)

CD8(-/

-)

CD4(-/

-)0

5

10

15

20

25

****

WT

asG

M1lgh

- 6

Clodro

nate

0

20

40

60

*

****

A B

C DN

o.o

f TU

NE

L+he

pato

cyte

s

E F

Fig. 1. Morbidity analysis of mice with AKT+CAT and MET+CAT-initiated livertumors using GLuc integrated in the liver. GLuc (4 lg), AKT (20 lg) + CAT(20 lg) and MET (20 lg) + CAT (20 lg) oncogenes along with HSB2 (4.4 lg) weredelivered to C57/Bl6 mice using hydrodynamic tail vein injection. (A) Kaplan-Meier survival plot. (B) Serum luciferase levels days 0–218. (C) Serum luciferaselevels days 0–42. PT3 n = 5, AKT+CAT n = 10, MET+CAT n = 10, mean ± SEM. (D)Liver sections were ApopTag Peroxidase stained using the TUNEL method on day7 following initiation with MET or CAT oncogenes. (E and F) MET washydrodynamically injected into WT, CD1d�/�, CD8�/�, CD4�/� or lgh-6 mice. ForNK cell and macrophage depletion (F) mice were treated with asGM1 orClodronate respectively on days 3 and 1 prior to MET delivery. On day 7 livertissue sections from (E and F) were stained for TUNEL counted on 5 non-overlapping 100� fields from 3 individual mice per group using CellProfiler. n = 3,***p <0.0001 mean ± SEM. Mann-Whitney U test was used to determinesignificance.

Research Article

PT3 control vector survived. In addition, the mortality rate ofAKT+CAT and MET+CAT transfected mice was similar (Fig. 1A),serum luciferase levels measured at various time points indicatedintriguing differences in the kinetics and therefore in the biologyof AKT+CAT and MET+CAT-initiated tumor growth (Fig. 1B). Dur-ing the early phase (days 10–42) of tumor growth, luciferaselevels in the sera from AKT+CAT-induced tumors increased, whilethe levels in mice that received MET+CAT fell slightly below thePT3 control (Fig. 1C). These data suggest the AKT+CAT oncogenecombination initially causes sustained proliferation of trans-formed hepatocytes more effectively than the MET+CAT onco-genes. However, starting at day 59, luciferase levels from bothgroups rise approximately co-linearly (Fig. 1B). It remained tobe determined whether this difference in kinetics was due tointrinsic characteristics in the induction and maintenance of pro-liferation in the transformed hepatocytes, or rather to some otherbiological factors such as differences in the ability of the immunesystem to recognize and respond to these tumors. We speculated

1184 Journal of Hepatology 2015

that the human MET protein might be immunogenic in mice, andwhen expressed on hepatocytes might elicit an immune responseresulting in detectable changes in tumor progression. To deter-mine whether hepatocytes expressing the MET oncogeneundergo apoptosis we injected MET or CAT oncogenes into miceand collected livers on day 7. Apoptotic hepatocytes weredetected by TUNEL staining of liver tissue sections (Fig. 1D).TUNEL staining was more prevalent in the livers of mice thatreceived the MET oncogene compared to those that received CAT.

To determine if the increased MET-induced hepatocyte apop-tosis was immune mediated, we first examined the level ofnon-specific TUNEL+ hepatocytes by immunohistochemistry innaïve WT, NKT deficient (CD1d�/�), CD8�/�, CD4�/�, and B cell�/�

(lgh-6) mice. We also examined the level of non-specific TUNEL+

hepatocytes in naïve mice depleted of NK cells and macrophagesusing asGM1 and clodronate respectively. No change in the fre-quency of TUNEL+ hepatocytes was observed in CD1d�/�, CD8�/�,CD4�/�, B cell�/� or in mice treated with clodronate or asGM1compared to WT mice (Supplementary Fig. 4A). Next to deter-mine if the increased MET-induced hepatocyte apoptosis wasimmune mediated, we injected MET into CD1d�/�, CD8�/�,CD4�/� or B cell�/� mice. We also injected MET into micedepleted of NK cells and macrophages using asGM1 and clo-dronate respectively. Livers were harvested on day 7 and TUNEL+

hepatocytes were quantitated using immunohistochemistry. Asignificant decrease in TUNEL+ hepatocytes was observed in theCD8�/� mice compared to WT mice (Fig. 1E), suggesting CD8 cellsare at least partially required for MET-induced hepatocyteapoptosis. In contrast, the frequency of TUNEL+ hepatocyteswas increased in mice depleted of macrophages and Igh-6deficient mice (Fig. 1F), suggesting these cells may offer a partialprotection from MET-induced hepatocyte apoptosis.

Recently it was shown that oncogenic N-ras can inducesenescence in pre-malignant cells that are eliminated by CD4+ Tcells [22]. However, mice with MET transduced livers did notexpress hepatocyte N-ras (Fig. 2A) or senescence markerbeta-galactosidase at days 10 and 30 (Fig. 2B), suggesting METdoes not induce the senescence pathway in hepatocytes.

The ability of MET oncogene to induce apoptosis directly wasalso examined. Mouse hepatoma Hepa1–6 cells were transfectedwith MET or GFP SBT vectors with greater than 20% transfectionefficiency determined by the number of GFP+ Hepa1–6 cells(Supplementary Fig. 4B). For a cell apoptosis control, Hepa1–6cells were treated for 24 h with 100 nM sorafenib. At 24 and72 h, apoptotic Hepa1–6 cells were detected by Annexin V and7-AAD staining. No increase in the proportion of apoptoticHepa1–6 cells was observed compared to GFP-transfectedcontrols (Fig. 2C), suggesting at least in this setting MET doesnot cause oncogene-induced apoptosis.

CD8+ T cells regulate early growth of MET+CAT oncogene-inducedliver cancer

For these studies we used a lower dose of oncogenes (10 lg) toallow more time for nuanced immune reactions to occur and toincrease sensitivity for GLuc tracking. Cytotoxic CD8+ T cells thatrecognize tumor antigen can cause tumor cell apoptosis. How-ever, chronic exposure to excessive antigen can lead to CD8+ Tcell exhaustion [23], characterized in part by the upregulationof inhibitory receptors such as PD-1 [24,25]. We hypothesizedthat early CD8-mediated immune responses against MET occur

vol. 63 j 1181–1189

Annexin V7-

AAD

24 h

72 h

MET GFP SorafenibControl

N-ra

sIs

otyp

e

Day

10

WT

Day

30

CD8-/-A

ControlR

OSA

26

Day

10

WT

Day

30

CD8-/-B

C

Fig. 2. MET does not induce N-ras-mediated hepatocyte senescence or oncogene-induced apoptosis. (A) MET was delivered to WT (n = 5) or CD8�/� (n = 5) mice orN-ras (control) was delivered to WT mice and on days 10 and 30 liver tissue sections were stained for N-ras. (B) MET was delivered to WT (n = 5) and CD8�/� mice (n = 5)and on days 10 and 30 liver tissue sections from these mice and B6.129S4-Gt(ROSA)26Sortm1Sor/J (control) mice were stained for beta-galactosidase. (C) Hepa1–6 cellswere transfected with SBT vectors MET or GFP (control) using Targefect-Hepatocyte. At 24 and 72 h cell apoptosis was determined by staining for Annexin V and 7-AAD andanalyzed by flow cytometry. For cell apoptosis control Hepa1–6 cells were treated for 24 h with 100 nM sorafenib. Two independent experiments were performed withsimilar results.

8.18% ± 0.74% 10.17% ± 1.72% 11.77% ± 1.43%

A B

CD8

PD1

Day

Perc

ents

urvi

val

100 150 200 2500%

50%

100%

WTCD8-/-

p = 0.0099

CMET-CAT AKT-CAT Control

**

Day

RLU

0

2×106

4×106

6×106

8×106 WT (PT3 Control)

WT (MET-CAT)CD8-/- (MET-CAT)

CD8-/- (PT3 Control)

WT CD8(-/-)0

20

40

60N

o.of

tum

orno

dule

s

WT CD8(-/-)0

1×1042×1043×1044×1045×104

1020

Mea

ntum

orn o

dule

size

µm2

****

WT CD8(-/-)0%

20%

40%

60%

80%

100%

Tum

orno

dule

area

perf

ield

****ED

WT

CD

8-/-

0 5 10 15 20 25 0 50

×104×104

Fig. 3. CD8+ T cells regulate early MET+CAT-induced liver oncogenesis. (A) C57/Bl6 mice were hydrodynamically transfected with MET+CAT (10 lg), AKT+CAT (10 lg) orPT3 (control) and on day 14 liver leukocytes were isolated and CD8+ T cells were analyzed for PD-1 expression by flow cytometry. n = 4, mean ± SEM (B) Serum luciferaselevels from WT or CD8�/� mice hydrodynamically transfected with GLuc and MET+CAT (10 lg) or PT3 (control). WT (MET+CAT) n = 5, CD8�/� (MET+CAT) n = 4, WT (PT3)n = 5, CD8�/� n = 5, mean ± SEM. (C) Kaplan-Meier survival plot of WT and CD8�/� mice hydrodynamically transfected with MET+CAT oncogenes. WT n = 25, CD8�/� n = 10,mean ± SEM. (D) MET+CAT (10 lg) was hydrodynamically injected into WT and CD8�/� mice and on day 42 liver tissue sections were stained for b-catenin and (E) 5 non-overlapping 100� fields were quantified for tumor area, mean tumor size and number of tumor nodules using CellProfiler. n = 5, ****p <0.00001 mean ± SEM. Mann-WhitneyU test was used to determine significance.

JOURNAL OF HEPATOLOGY

and result in apoptosis of transformed hepatocytes that wasdetected by the early decrease in luciferase following deliveryof MET+CAT oncogenes to the liver. If this theory is correct, wereasoned that as CD8+ T cells become exhausted, the remainingtransformed GLuc+ hepatocytes would begin to proliferate result-ing in rising serum luciferase levels, in association with acceler-ated tumor growth and eventual mortality. To test this

Journal of Hepatology 2015

hypothesis, tumor infiltrating CD8+ T cells from mice injectedwith MET+CAT or AKT+CAT oncogenes were analyzed by flowcytometry for the expression of PD-1 antigen as a marker ofantigen-experienced exhausted T cells (Fig. 3A). At day 14, PD-1expression was induced on 33% of the liver CD8+ T cells frommice where tumors were initiated with the MET+CAT oncogenes.In contrast, no PD-1 was detected on liver CD8+ T cells from mice

vol. 63 j 1181–1189 1185

Day

RLU

20 40 600

5×106

107 VCIL-18+IL-12

Live

rwei

ght(

g)

VC IL-18+IL-12 0

2

4

6

8

MET

copy

#

VC IL-18+IL-12 0

5.0×105

1.0×106

1.5×106

2.0×106

VC IL-18+IL-12 0

5.0×105

1.0×106

1.5×106

A B

C

**

***

******

* *

CAT

cop

y #

Fig. 4. IL-18+IL-12 therapy against MET+CAT oncogene-induced liver tumors.MET (10 lg) + CAT (10 lg) oncogenes along with GLuc (4 lg) and HSB2 (2.4 lg)were hydrodynamically-transfected into C57/Bl6 mice. On days 20–24 and 28–31mice were treated with vehicle control (VC) or IL-18+IL-12. (A) Serum GaussiaLuciferase levels vs. day. (B) Livers were weighed on day 61. (C) DNA was isolatedfrom livers on day 61 and MET and CAT oncogene copy number was qPCRquantitated from 100 ng of DNA. VC n = 5, IL-18+12 n = 10, mean ± SEM. For B andC Wilcoxon Rank Sum test was used to determine significance.

Research Article

where tumors were initiated by the AKT+CAT oncogene combina-tion, or in non-tumor-bearing mice injected with control PT3vector. These data suggest that CD8+ T cells mediate immuneresponses against hepatocytes expressing MET+CAT, but notAKT+CAT oncogenes.

To determine whether the early differences in serum levelsof luciferase were reflecting tumor control by a CD8-mediatedresponse, luciferase levels after injection of MET+CAT oncoge-nes were compared in CD8-deficient and WT mice. CD8 defi-ciency did not affect the ability of hepatocytes to take upand express SBT vectors. When CD8�/� and WT mice werehydrodynamically transfected with GLuc+HSB2 and MET+CAT,serum luciferase levels at day 10 were indistinguishable butby days 15 and 22 they increased in CD8�/� mice but not inWT mice (Fig. 3B), confirming that serum luciferase levelsaccurately captured the early CD8-mediated immune responseagainst MET+CAT transformed hepatocytes. Since the data pre-sented above confirms that MET+CAT transduced hepatocytesare immunogenic, we next wanted to determine whether thetransient CD8-mediated response confers a survival advantage.When CD8�/� and WT mice were hydrodynamically transfectedwith MET+CAT oncogenes CD8�/� mice had a significantlyshorter survival time than WT mice (Fig. 3C). The shorter sur-vival from MET+CAT-induced liver tumors in CD8�/� mice sig-nified that tumor load developed more quickly in these mice.However, gross morphology at time of morbidity did not revealif the shortened survival time was due to quicker tumor devel-opment and growth or if more tumor nodules were formed. Todetermine if CD8+ T cells regulated tumor size (growth) and/orthe number of tumor nodules that develop, we analyzed micewith MET+CAT-initiated livers at day 42, before frank tumoris observed, by staining for the surrogate oncogene b-catenin(CAT) (Fig. 3D). Tumor nodules were quantitated for percentfield (100� magnification), mean tumor nodule size in micronsand the number of tumor nodules. CD8�/� mice exhibitedsignificantly more tumor area and mean tumor nodule sizecompared to WT mice (Fig. 3E). However, no difference wasobserved in the number of tumor nodules between the twostrains of mice (Fig. 3E), suggesting CD8+ T cells regulate tumorgrowth rather than the number of malignant foci.

MET+CAT oncogene-induced liver tumor response to IL-18+IL-12therapies can be accurately tracked by luciferase expression

MET is a broadly expressed oncogene in patients with numer-ous malignancies [26] and is recognized by cytotoxic T cells[27]. We reported previously that IL-18+IL-12 therapy waseffective against transplantable tumors seeded in the liver[28]. Since MET+CAT induced a CD8+ T cell immune responsewe hypothesized that IL-18+IL-12 therapy would augment nat-urally occurring responses against MET+CAT-induced tumors.Furthermore, we postulate that serum luciferase could be usedto track in real-time the effectiveness of IL-18+IL-12 therapyagainst MET+CAT-induced liver tumors originated from GLuc+

hepatocytes. Mice were hydrodynamically transfected withMET+CAT, HSB2 and GLuc initiating de novo liver tumor forma-tion in GLuc+ hepatocytes. Liver tumors were allowed to formand on days 20–24 and 28–31 mice were treated with VC orwith IL-18+IL-12. Serum luciferase levels were not significantlydifferent between the two groups of mice for the first 19 days(Fig. 4A). However, during days 31–61 post-therapy, serum

1186 Journal of Hepatology 2015

luciferase levels for the VC-treated group increased while theIL-18+IL-12 treatment group significantly decreased to nearundetectable levels (p <0.0001). This result suggests that ther-apy induced a loss of MET+CAT transformed GLuc+ hepatocytes.To confirm whether the decreased serum luciferase levels fol-lowing IL-18+IL-12 therapy were due to reduced MET+CATtumor burden, mice were euthanized at day 61. Liver weights,as well as genomic copy number of MET and CAT oncogenesmeasured by qPCR, were quantitated as parameters of tumorsize (Fig. 4B, C). VC-treated mice showed a significantly higherliver weight and MET and CAT copy number thanIL-18+IL-12-treated mice. (Fig. 4B, C). Gross morphologicalexamination of livers reveals a broad spectrum of tumor bur-den from nothing observable to substantial burden in the VCgroup. On the other hand, no visible tumor nodules wereobserved on the livers from mice treated with IL-18+IL-12(Fig. 5A). Histological analysis reveals livers with frank tumorin the VC group, which had extensive regions that stained pos-itively for CAT, while none of the livers in IL-18+IL-12 groupstained positive for CAT (Fig. 5B). H&E analysis of theIL-18+IL-12 group shows a normal liver structure (Fig. 5C).H&E analysis of the VC group shows disrupted liver structurewith a thick trabecular pattern containing large eosin+ tumorcells displaying prominent nucleoli. Serum luciferase levelswere significantly higher in mice with morphologically-distinguishable gross tumors (5 out 8 mice in VC group), whichreflected liver weight (Fig. 5D). Serum luciferase levels inIL-18+IL-12-treated mice were at background levels and liverweights were normal (1 g or less). Linear regression analysiscomparing serum luciferase levels and liver weights for allmice were highly significant (Fig. 5E). Collectively we demon-strate for the first time that SBT-delivered oncogene-inducedliver tumors are amenable for studying the efficacy ofimmunotherapeutic approaches and that GLuc vector can beused to assess the effectiveness of these therapies.

vol. 63 j 1181–1189

6696

6790

6777

6778

6779

6787

6789

6796

6763

6764

6765

6757

6758

6759

6770

0

5

10

15

0

2

4

6

Animal #

RLU

(106

)Li verw

eight(g)

RLULiver weight (g)

6696 6690

6777 6778

6779 6787

6789 6796

6763 6764

6765 6757

6758 6759

6770

B D

E

6789

6796

6763

6764

6696 6690 6777 6778

6779 6787 6789 6796

6763 6764 6765 6757

6758 6759 6770

VC IL18 + IL-12A C

VC IL18 + IL-12

VC IL18 + IL-12

0 2 4 60

5

10

15

Liver weight(g)

RLU

(106 )

N = 15R2 = 0.9908p <0.0001

Fig. 5. Liver morphology analysis from mice with MET+CAT initiated liver tumors treated with IL-18+IL-12. MET+CAT was hydrodynamically transfected and on days22–24, 28–31 mice were treated with VC or IL-18+IL-12. (A) Livers were harvested on day 55 (B) and tissue sections were stained for b-catenin surrogate marker or (C) H&E.(D) Serum luciferase levels (left y-axis) and liver weight (right y-axis) was compared per mouse. (E) Linear regression analysis was performed on serum luciferase levels vs.tumor weight dotted line represents 95% confidence band.

JOURNAL OF HEPATOLOGY

Discussion

Hydrodynamic transfection of SBT vectors provides an efficientand rapid system to functionally evaluate oncogenic collabora-tion in the formation of de novo liver tumors. Furthermore, thistechnique permits the assessment of inflammation-influencedsomatic mutations, delineating cancer ‘‘drivers’’ from ‘‘passengers’’[7,29]. SBT delivery of oncogenes initiates tumor developmentand subsequent progression, which permits the study of inflam-matory events driving liver tumor formation, progression andimmune escape. However, the time to morbidity for tumors initi-ated by the SBT system can vary significantly. Thus, a trackingapproach that quantifies tumor growth patterns and therapeuticinterventions in living mice, without the need for costly andtime-consuming imaging, or euthanasia and analysis of cohorts,would allow for efficient delineation of molecular and immuneregulatory events that regulate tumor initiation and growth.We found co-delivery of a GLuc expression vector in combinationwith oncogenes accurately tracked liver tumor growth initiatedusing the SBT system. Decreasing the amount of GLuc decreasesthe background serum luciferase levels. Reducing the ratio ofoncogene to GLuc allows a higher frequency of hepatocytes totake up both oncogenes and GLuc allowing for a higher frequencyof luciferase expressing transformed hepatocytes and increasedsensitivity for tracking tumor growth with GLuc.

Solid tumors often grow in a complex nonlinear fashion[20,21]. Tumor proliferation and apoptosis can be influenced invarious ways by different factors, such as immune recognition[30,31] and the presence of angiogenic factors needed to formtumor vasculature [32]. Determining these tumor growth

Journal of Hepatology 2015

inflection points could offer valuable biological insights intotumor growth as well as informing the timing and outcome ofpossible therapeutic interventions. We determined that serumluciferase levels closely tracked tumor-induced changes in liversize as detected and quantitated by MRI analysis. Thus, we showhere for the first time that the nonlinear growth patterns ofoncogene-induced liver tumors can be tracked in real-time byserum luciferase levels from hepatocytes using an integratedGLuc expression vector. In addition to tracking SBT-inducedtumor growth, the stable expression of integrated GLuc shouldalso be useful in tracking tumors initiated by carcinogens/mutagen such as DEN. Furthermore, the changes in stableluciferase expression from GLuc could be useful in monitoringliver toxicity or studies employing partial hepatectomy.

MET, which is widely expressed in numerous malignancies[26], was also demonstrated to be a tumor antigen recognizedby cytotoxic T cells (CTLs) [27]. In our model we found humanMET, but not human CAT, induced hepatocyte apoptosis thatwas at least partially mediated by CD8+ T cells. This result sug-gested human MET in a mouse model may also induce cytotoxicT cells. Some MET-induced TUNEL staining was still present inCD8�/� mice suggesting that MET may also induce someoncogene-driven apoptosis. Interestingly delivering MET to Bcell�/� mice or depleting macrophages with clodronate increasedthe number of TUNEL+ hepatocytes suggesting these cells offerpartial protection from MET-induced apoptosis. Examination ofdifferences in early growth patterns revealed that CD8+ T cellsregulated the persistence of MET+CAT transduced hepatocytesduring the time span shortly after transformation. Furthermore,CD8-deficient mice succumbed to MET+CAT-induced liver tumors

vol. 63 j 1181–1189 1187

Thvie

Research Article

more rapidly than did their WT counterpart. Histological exami-nation of MET+CAT-initiated tumors shows CD8�/� mice had lar-ger tumor nodules, as well as a similar number of tumor nodules.Taken together, this suggests CD8 T cells regulate the growth ofMET+CAT-induced liver tumors. However, the larger tumornodules could also be masking the total number of nodules inCD8 deficient mice as they may be coalescing in some cases.

To our knowledge this is the first report to use immunother-apy to treat SBT-induced liver tumors. We previously reportedthat IL-18+IL-12 was a potent immunotherapeutic regimenagainst implanted tumors [28]. Because human CTLs have beenshown to respond to MET [27] and we found mouse CD8+ T cellsresponded to human MET, we examined the immunotherapeuticpotential of treating mice bearing MET+CAT tumors withIL-18+IL-12. We found treating mice with IL-18+IL-12 ablatedMET+CAT-initiated liver tumors and restored liver morphologythat was accurately tracked by serum luciferase quantitation,suggesting that this technology can be used to monitor preclini-cal immunotherapeutic approaches against liver cancer.Immunohistochemical analysis of CAT staining demonstratedthat IL-18+IL-12 also cleared all detectable microscopic tumors,in addition to the absence of grossly detectable visible tumorsas described above. This data suggests IL-18+IL-12 therapy mayelicit an anti-tumor response in humans mediated by METresponding CTLs.

In summary, we have shown that hydrodynamic delivery ofGLuc vector can be used as a simple cost-effective, blood-basedapproach for tracking the progression of pathologically distinctliver tumors, evaluation of immunologic responses and efficacyof therapeutic interventions in individual mice. This approachcan be used to track nonlinear tumor progression in real-timeand identify potentially important biological regulatory pointsat which therapeutic intervention may hold greater potentialfor controlling liver tumor burden. Furthermore, IL-18+IL-12 isan effective therapy against antigenic MET-induced liver tumors.

Financial support

This work was supported by the Intramural Research Program ofthe National Cancer Institute at the National Institutes of Health(NCI/NIH), United States.

Conflict of interest

The authors who have taken part in this study declared that theydo not have anything to disclose regarding funding or conflict ofinterest with respect to this manuscript.

Author’s contributions

Study conception: Subleski and Stauffer.Acquisition of data: Subleski, Scarzello, Kronfili, Saleh, Back,Jiang.Analysis and interpretation of data: Subleski, Scarzello, Alvord,Kronfili, Saleh Weiss, and Wiltrout.Drafting of manuscript: Subleski, Weiss, and Wiltrout.Critical revision: Subleski, Scarzello, Weiss, and Wiltrout.

1188 Journal of Hepatology 2015 vo

e content of this publication does not necessarily reflect thews or policies of the Department of Health and Humanvices, nor does mention of trade names, commercial products,organization imply endorsement by the U.S. government.

Seror

Acknowledgements

We thank Dr. Giorgio Trinchieri for his critical editing of themanuscript and Megan Karwan for her assistance withanimal experiments. This work was supported by the IntramuralResearch Program of the National Cancer Institute, NationalInstitutes of Health (NCI/NIH).

Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.jhep.2015.06.021.

References

[1] Totoki Y, Tatsuno K, Yamamoto S, Arai Y, Hosoda F, Ishikawa S, et al. High-resolution characterization of a hepatocellular carcinoma genome. Nat Genet2011;43:464–469.

[2] Thorgeirsson SS, Grisham JW. Molecular pathogenesis of human hepatocel-lular carcinoma. Nat Genet 2002;31:339–346.

[3] Thamavit W, Bhamarapravati N, Sahaphong S, Vajrasthira S, Angsubhakorn S.Effects of dimethylnitrosamine on induction of cholangiocarcinoma inOpisthorchis viverrini-infected Syrian golden hamsters. Cancer Res1978;38:4634–4639.

[4] Levrero M. Viral hepatitis and liver cancer: the case of hepatitis C. Oncogene2006;25:3834–3847.

[5] Khandekar MJ, Cohen P, Spiegelman BM. Molecular mechanisms of cancerdevelopment in obesity. Nat Rev Cancer 2011;11:886–895.

[6] Jemal A, Siegel R, Xu J, Ward E. Cancer statistics, 2010. CA Cancer J Clin2010;60:277–300.

[7] Guichard C, Amaddeo G, Imbeaud S, Ladeiro Y, Pelletier L, Maad IB, et al.Integrated analysis of somatic mutations and focal copy-number changesidentifies key genes and pathways in hepatocellular carcinoma. Nat Genet2012;44:694–698.

[8] El-Serag HB, Rudolph KL. Hepatocellular carcinoma: epidemiology andmolecular carcinogenesis. Gastroenterology 2007;132:2557–2576.

[9] Bhamarapravati N, Virranuvatti V. Liver diseases in Thailand. An analysis ofliver biopsies. Am J Gastroenterol 1966;45:267–275.

[10] Ng M, Fleming T, Robinson M, Thomson B, Graetz N, Margono C, et al. Global,regional, and national prevalence of overweight and obesity in children andadults during 1980–2013: a systematic analysis for the Global Burden ofDisease Study 2013. Lancet 2014;384:766–781.

[11] Mohd HK, Groeger J, Flaxman AD, Wiersma ST. Global epidemiology ofhepatitis C virus infection: new estimates of age-specific antibody to HCVseroprevalence. Hepatology 2013;57:1333–1342.

[12] Bell JB, Podetz-Pedersen KM, Aronovich EL, Belur LR, McIvor RS, Hackett PB.Preferential delivery of the Sleeping Beauty transposon system to livers ofmice by hydrodynamic injection. Nat Protoc 2007;2:3153–3165.

[13] Chen X, Calvisi DF. Hydrodynamic transfection for generation of novelmouse models for liver cancer research. Am J Pathol 2014;184:912–923.

[14] Stauffer JK, Scarzello AJ, Andersen JB, De Kluyver RL, Back TC, Weiss JM, et al.Coactivation of AKT and beta-catenin in mice rapidly induces formation oflipogenic liver tumors. Cancer Res 2011;71:2718–2727.

[15] Subleski JJ, Hall VL, Wolfe TB, Scarzello AJ, Weiss JM, Chan T, et al. TCR-dependent and -independent activation underlie liver-specific regulation ofNKT cells. J Immunol 2011;186:838–847.

[16] Crowder MJ, Hand DJ. Analysis of repeated measures. 1st ed. Boca Raton,Florida: CRC Press; 1990.

[17] Pinheiro JC, Bates DM. Mixed-effects models in S and S-PLUS. NewYork: Springer-Verlag New York, Inc.; 2000.

l. 63 j 1181–1189

JOURNAL OF HEPATOLOGY

[18] Bonini C, Ferrari G, Verzeletti S, Servida P, Zappone E, Ruggieri L, et al.

HSV-TK gene transfer into donor lymphocytes for control of allogeneic graft-versus-leukemia. Science 1997;276:1719–1724.

[19] Riddell SR, Elliott M, Lewinsohn DA, Gilbert MJ, Wilson L, Manley SA, et al. T-cell mediated rejection of gene-modified HIV-specific cytotoxic T lympho-cytes in HIV-infected patients. Nat Med 1996;2:216–223.

[20] Shimizu S, Shirato H, Xo B, Kagei K, Nishioka T, Hashimoto S, et al. Three-dimensional movement of a liver tumor detected by high-speed magneticresonance imaging. Radiother Oncol 1999;50:367–370.

[21] Wise SM, Lowengrub JS, Frieboes HB, Cristini V. Three-dimensional multi-species nonlinear tumor growth–I Model and numerical method. J Theor Biol2008;253:524–543.

[22] Kang TW, Yevsa T, Woller N, Hoenicke L, Wuestefeld T, Dauch D, et al.Senescence surveillance of pre-malignant hepatocytes limits liver cancerdevelopment. Nature 2011;479:547–551.

[23] Zajac AJ, Blattman JN, Murali-Krishna K, Sourdive DJ, Suresh M, Altman JD,et al. Viral immune evasion due to persistence of activated T cells withouteffector function. J Exp Med 1998;188:2205–2213.

[24] Akbay EA, Koyama S, Carretero J, Altabef A, Tchaicha JH, Christensen CL, et al.Activation of the PD-1 pathway contributes to immune escape in EGFR-driven lung tumors. Cancer Discov 2013;3:1355–1363.

[25] Fourcade J, Sun Z, Pagliano O, Guillaume P, Luescher IF, Sander C, et al.CD8(+) T cells specific for tumor antigens can be rendered dysfunctional by

Journal of Hepatology 2015

the tumor microenvironment through upregulation of the inhibitoryreceptors BTLA and PD-1. Cancer Res 2012;72:887–896.

[26] Sierra JR, Tsao MS. C-MET as a potential therapeutic target and biomarker incancer. Ther Adv Med Oncol 2011;3:S21–S35.

[27] Schag K, Schmidt SM, Muller MR, Weinschenk T, Appel S, Weck MM, et al.Identification of C-met oncogene as a broadly expressed tumor-associatedantigen recognized by cytotoxic T-lymphocytes. Clin Cancer Res 2004;10:3658–3666.

[28] Subleski JJ, Hall VL, Back TC, Ortaldo JR, Wiltrout RH. Enhanced antitumorresponse by divergent modulation of natural killer and natural killer T cellsin the liver. Cancer Res 2006;66:11005–11012.

[29] Reimand J, Bader GD. Systematic analysis of somatic mutations in phos-phorylation signaling predicts novel cancer drivers. Mol Syst Biol2013;9:637.

[30] Dunn GP, Old LJ, Schreiber RD. The immunobiology of cancer immuno-surveillance and immunoediting. Immunity 2004;21:137–148.

[31] Pardoll D. Does the immune system see tumors as foreign or self? Annu RevImmunol 2003;21:807–839.

[32] Ferrara N, Kerbel RS. Angiogenesis as a therapeutic target. Nature2005;438:967–974.

vol. 63 j 1181–1189 1189