Oncogene-specific gene expression signatures at preneoplastic stage in mice define distinct mechanisms of hepatocarcinogenesis

Authors

  • Cédric Coulouarn,

    1. Laboratory of Experimental Carcinogenesis, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD
    Search for more papers by this author
  • Luis E. Gomez-Quiroz,

    1. Laboratory of Experimental Carcinogenesis, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD
    Search for more papers by this author
  • Ju-Seog Lee,

    1. Laboratory of Experimental Carcinogenesis, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD
    Search for more papers by this author
  • Pal Kaposi-Novak,

    1. Laboratory of Experimental Carcinogenesis, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD
    Search for more papers by this author
  • Elizabeth A. Conner,

    1. Laboratory of Experimental Carcinogenesis, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD
    Search for more papers by this author
  • Tatyana A. Goldina,

    1. Department of Cell Biology and Histology, Moscow State University, Moscow, Russia
    Search for more papers by this author
  • Galina E. Onishchenko,

    1. Department of Cell Biology and Histology, Moscow State University, Moscow, Russia
    Search for more papers by this author
  • Valentina M. Factor,

    1. Laboratory of Experimental Carcinogenesis, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD
    Search for more papers by this author
  • Snorri S. Thorgeirsson

    Corresponding author
    1. Laboratory of Experimental Carcinogenesis, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD
    • LEC/CCR/NCI, Building 37, Room 4146A, 37 Convent Drive MSC 4262, Bethesda, MD 20892
    Search for more papers by this author
    • fax: 301-496-0734


  • Potential conflict of interest: Nothing to report.

Abstract

We applied a genome-wide microarray analysis to three transgenic mouse models of liver cancer in which targeted overexpression of c-Myc, E2f1, and a combination of the two was driven by the albumin promoter. Although gene expression profiles in HCC derived in all three transgenic lines were highly similar, oncogene-specific gene expression signatures were identified at an early dysplastic stage of hepatocarcinogenesis. Overexpression of E2f1 was associated with a strong alteration in lipid metabolism, and Srebp1was identified as a candidate transcription factor responsible for lipogenic enzyme induction. The molecular signature of c-Myc overexpression included the induction of more than 60 genes involved in the translational machinery that correlated with an increase in liver mass. In contrast, the combined activity of c-Myc and E2f1 specifically enhanced the expression of genes involved in mitochondrial metabolism—particularly the components of the respiratory chain—and correlated with an increased ATP synthesis. Thus, the results suggest that E2f1, c-Myc, and their combination may promote liver tumor development by distinct mechanisms. In conclusion, determination of tissue-specific oncogene expression signatures might be useful to identify conserved expression modules in human cancers. (HEPATOLOGY 2006.)

Current evidence indicates that human carcinogenesis is a multistep process in which dysregulation of proliferation and apoptosis are principal characteristics.1, 2 Therefore, it is not surprising that altered expression of cell cycle regulators is a recurrent feature in cancer, including hepatocellular carcinoma (HCC).1, 3 The transcription factors E2f1 and c-Myc are prototypes of such regulators that are frequently altered in human cancers.4 E2f1 plays a crucial role in cell cycle control by inducing the expression of genes required for G1/S progression and DNA replication.5–7 Thus, E2f1 activation is tightly regulated during the cell cycle. In the absence of mitotic signals, E2f1 is inactivated by binding to the dephosphorylated form of the retinoblastoma protein. Phosphorylation of retinoblastoma protein by cyclin-dependent kinases (CDKs) at the end of G1 phase releases inhibition of E2f1 to allow S-phase gene transactivation and cell-cycle progression.8, 9 Similarly to E2f1, c-Myc has been shown to drive quiescent cells into S phase in the absence of other mitogenic signals. c-Myc heterodimerizes with its transcriptional partner Max via a helix–loop–helix leucine zipper domain to regulate the expression of target genes involved in cell proliferation, such as Ccnd1, Cdk4, or Cdc25a.10 Furthermore, E2f1 and c-Myc are potent inducers of apoptosis through p53-dependent and independent mechanisms.4, 11, 12

Previously, we generated and extensively characterized several transgenic mouse models of liver cancer that reproduce the broad spectra of pathological and genetic changes in human liver disease.13 In these models, overexpression of either the c-Myc or E2f1 proto-oncogene was sufficient to promote tumor growth. In both cases, the tumorigenesis was characterized by a persistent activation of cell proliferation that surpassed the high rate of c-Myc- or E2f1-mediated apoptosis.14 Conversely, when these two transcription factors were coexpressed in the liver, the ability of c-Myc and E2f1 to sensitize cells to proapoptotic stimuli was greatly reduced, resulting in the acceleration of neoplastic development, compared with either of the parental lines.15

In addition to their effect on cell proliferation and apoptosis, both c-Myc and E2f1 can significantly alter cellular metabolism that in turn may affect the transformation process. For example, c-Myc enhances glucose uptake and glycolysis via transcriptional activation of different target genes, including lactate dehydrogenase A.16, 17 It is noteworthy that inhibition of lactate dehydrogenase A has no effect on cell proliferation per se but abrogates c-Myc–induced transformation in fibroblasts, suggesting that lactate dehydrogenase A may be specifically required for the transformation process.16 c-Myc is also a central regulator of cell growth and ribosome biogenesis.18, 19

E2F1 regulates peroxisome proliferator-activated receptor γ (PPAR-γ) transcription during adipocyte differentiation.20 The role for hepatic PPAR-γ in the development and maintenance of steatosis in the liver is well established.21, 22 A liver-specific knockout of PPAR-γ in both A-ZIP/F-1 and ob/ob mice results in decreased lipid stores in the liver and reduced expression of several genes that are important for adipocyte differentiation and lipid metabolism.

The metabolic alterations induced by c-Myc and E2f1 recapitulate several of the abnormal hepatocyte functions associated with type 2 diabetes and steatohepatitis, both of which are risk factors for liver cancer.23 To address the potential impact of c-Myc- and E2f1-induced metabolic alterations during the multistep process of liver carcinogenesis, we characterized the global gene expression changes in early dysplastic livers as well as in HCC and surrounding nontumor livers from E2f1, c-Myc, or c-Myc/E2f1 transgenic mice. We found the gene expression profiles to be very divergent at the early dysplastic stage but highly similar at the tumor stage, suggesting that E2f1 and c-Myc may promote malignant transformation by distinct mechanisms. Furthermore, our data show that E2f1 and c-Myc contribute to hepatocarcinogenesis not only by regulating cell proliferation and death but more profoundly by modulating specific metabolic pathways. Thus, we demonstrate that overexpression of E2f1 is associated with an early induction of genes involved in lipid metabolism, whereas c-Myc is a potent inducer of translational machinery. In addition, the concomitant overexpression of E2f1 and c-Myc specifically targets mitochondrial metabolism and enhances production of adenosine triphosphate.

Abbreviations

HCC, hepatocellular carcinoma; CDK, cyclin-dependent kinase; PPAR-γ, peroxisome proliferator-activated receptor γ; ATP, adenosine triphosphate; CoA, coenzyme A; Srebp1, sterol regulatory element binding factor 1.

Materials and Methods

Transgenic Mice.

Generation of Alb/E2f1 and Alb/c-Myc single transgenic and E2f1/c-Myc double transgenic mice has been previously described.14, 24 The protocols of the present study were in accordance with the National Institutes of Health guidelines for animal care. Mice had free access to food and water. Only male mice were used in this study. To avoid the effects of circadian clocks on gene expression, all mice were sacrificed between 10:00 and 11:00 A.M. The tissue samples were divided into two parts; one was fixed in 10% formalin for routine histology, and the other was used for RNA extraction by a CsCl gradient-density centrifugation method. For the dysplastic stage, RNA was isolated from the livers of 5 mice at 3 months of age for each genotype. At the tumor stage, RNAs were isolated from 15 mice 8 to 10 months of age depending on the genotype and 5 surrounding nontumorous livers. All liver samples were subjected to histopathological evaluation before RNA isolation to confirm the absence of focal lesions (dysplastic stage) or contaminating tumors (surrounding tissue). Pooled RNA isolated from 10 wild-type mice at 3 months of age was used as a common reference for all microarray experiments.

Microarrays and Data Analysis.

A genome-wide set (>38K) of longmer mouse oligonucleotides obtained from Illumina (San Diego, CA) and printed at the Advanced Technology Center (National Cancer Institute) was used in this study. This set is based on well-annotated sequence information derived largely from the Mouse Exonic Evidence-Based Oligonucleotide consortium (http://mmc.ucsf.edu/Meebo.html). Target preparation and hybridization on microarrays were performed as previously described.25 Briefly, 20 μg of each target RNA were labeled with Cy-3 or Cy-5 and then hybridized with the same batch of reference RNA in a dye-swap arrangement. Differentially expressed genes were identified using a univariate 2-sample t test with a random variance model. Permutation P values for significant genes (P < .001) were computed based on 10,000 random permutations that result in an estimated false discovery rate below 1%. Data mining was performed from GeneOntology annotations, and the most relevant gene networks were confirmed using the Ingenuity Pathway Analysis server (Ingenuity Systems, Mountain View, CA).

Western Blot Analysis.

Total and nuclear proteins were extracted using T-PER and NE-PER buffers, respectively (Pierce, Rockford, IL), and processed as previously described.15 Immunodetection was performed using anti-Akt, anti–phospho-Akt, (Cell Signaling, Beverly, MA), anti-actin, anti–cyclin D1, anti-p21, and anti-Srebp1 (Santa Cruz Biotechnology, Santa Cruz, CA) primary antibodies and an anti-rabbit secondary antibody (Amersham, Piscataway, NJ).

Adenosine Triphosphate Assay.

Adenosine triphosphate (ATP) content was measured from 100 mg of fresh tissue homogenized in ice-cold phosphate-buffered saline. Tissue suspensions and ATP standards were then mixed with an ATP-dependent light-emitting luciferase reagent (Celltiter-Glo, Promega, Madison, WI), and ATP content was measured in a luminescent counter (Perkin Elmer, Wellesley, MA). To standardize the unit of ATP concentration, the protein content was determined using a BCA protein assay kit (Pierce, Rockford, IL) and the unit of ATP concentration was expressed as nmol/mg protein.

Electron Microscopy.

Fragments of liver were fixed in a mixture of 2.5% glutaraldehyde and 2% neutral formaldehyde in phosphate-buffered saline (pH 7.4) for 4 hours and then postfixed in 1% osmium tetroxide in phosphate-buffered saline for an additional 2 hours. After embedding in Epon 812, ultrathin sections were cut with a Diatome diamond knife on LKB ULTRATOM III ultratome (LKB Ultratome, Uppsala, Sweden), then contrasted with uranyl acetate and lead citrate and examined on a JEOL 100CX transmission electron microscope (JEOL, Tokyo, Japan).

Results and Discussion

Stage-Specific Gene Expression Alterations.

We used a mouse genome-wide microarray platform for the identification of genes regulated during the course of malignant transformation induced by a targeted overexpression of E2f1 and c-Myc in the liver, either alone or in combination. The expression level of 5,331 unique gene features was found to be significantly (P < .001) altered in transgenic mice compared with wild-type mice (2,541 genes differentially expressed in early dysplastic livers, 2,520 genes in surrounding nontumor livers, and 2,570 genes in HCC).

An unsupervised clustering of all samples based on the expression levels of the significant genes identified 3 major clusters (C1-3) (Fig. 1A). Livers from wild-type mice clustered in C1 with all early dysplastic livers and most of the surrounding nontumor tissues isolated from transgenic mice. Although tumors also clustered together, two groups formed predominantly by c-Myc and E2f1 HCC were clearly separated. Thus, clusters C2 and C3 encompassed 67% and 80% of c-Myc and E2f1 tumors, respectively, whereas HCC from E2f1/c-Myc transgenic mice displayed larger intertumor variance recapitulating either c-Myc-specific (5 HCC) or E2f1-specific (10 HCC) gene expression pattern and reflecting higher genetic heterogeneity. Two surrounding livers from E2f1 overexpressing mice also clustered in C3, most likely due to the contamination by tumor cells given the high multiplicity of neoplastic growth in this model.14

Figure 1.

Hierarchical clustering of genes and mouse tissue samples reveals similarities and differences between transgenic mice at different stages of hepatocarcinogenesis. (A) Unsupervised clustering of mouse tissue samples. Only genes in which expression was significantly (P < .001) different between transgenic (E2f1, c-Myc, and E2f1/c-Myc) and wild-type liver samples were selected for analysis (5,331 gene features). Hierarchical clustering using Pearson's correlation coefficients identified three major clusters (C1-3). Cluster C1 encompassed all dysplastic livers (Dysp. Liv.) and most surrounding nontumor livers (Sur. Liv.), whereas C2 and C3 included all HCC samples. (B) Supervised clustering of genes. For each stage of the malignant transformation, we applied the same statistical parameters (P < .001) to select genes differentially expressed between transgenic mice and wild-type mice (2,541 genes in early dysplastic livers, 2,520 genes in surrounding nontumor livers and 2,570 genes in HCC). The clustering data are presented in matrix format in which each column represents an individual tissue sample and each row represents a gene. This supervised approach identified clear transgene-specific gene expression signature in dysplastic livers that was not present in HCC, which exhibited similar gene expression profiles regardless of the overexpressed transgene. WT, wild-type; HCC, hepatocellular carcinoma.

A supervised analysis was next applied to identify oncogene-specific gene expression signatures for each step of the malignant transformation (Fig. 1B). Although the number of genes differentially expressed between transgenic and wild-type mice was quite similar at all stages (2,500 genes as mentioned above), the gene expression profiles were strikingly different between early dysplastic livers and HCC. Thus, in the HCC, the majority of genes (70%) were commonly upregulated or downregulated regardless of the transgenic line, in contrast to the presence of obvious E2f1, c-Myc, and E2f1/c-Myc specific gene expression signatures at the early dysplastic stage (Fig. 1B). The dysplastic liver profiles were also highly specific, because 68% of the 2,541 genes regulated at this stage were not found to be altered in HCC. Genes in which expression remained altered during the course of HCC development were mainly associated with proliferation, apoptosis, and immune response (Table 1; Supplementary Table 1). Moreover, ontology-based classification of genes differentially expressed in HCC failed to identify functional modules specific for the transgenic mouse lines, possibly because both c-Myc and E2f1 are overexpressed in the HCC from both lines. Therefore, the profiling of the early dysplastic stage appeared to be more informative for identification of oncogene-specific signatures than of fully developed HCC.

Table 1. Functional Classification of Genes Regulated in Early Dysplastic Livers of E2f1, c-Myc, and E2f1/c-Myc Transgenic Mice
Functional CategorySource*P
  • *

    Functional classification was based on GO annotations (GO) or derived from Ingenuity pathway analysis server (Ing).

Genes upregulated in all 3 models (72 genes)  
 Cancer (cell-cycle/cell-death)Ing3.44E-04
Genes downregulated in all 3 models (96 genes)  
 Immune responseGO + Ing2.06E-05
 Cancer (cell cycle/cell death)Ing3.93E-05
 Cellular movementIng6.15E-05
 Complement activationGO + Ing5.00E-04
Genes upregulated in E2f1 transgenic mice (637 genes)
 Amino acid metabolismGO + Ing3.65E-05
 Lipid metabolismGO + Ing2.17E-04
 ApoptosisGO + Ing4.00E-04
 Cancer (cell cycle/cell death)Ing9.93E-04
Genes downregulated in E2f1 transgenic mice (373 genes)
 Immune responseGO + Ing2.99E-10
 Cellular movementIng4.93E-05
 Cancer (cell cycle/cell death)Ing9.66E-05
 Cellular growth and proliferationIng1.59E-04
 Primary metabolismGO7.80E-04
Genes upregulated in c-Myc transgenic mice (381 genes)
 Protein synthesisGO + Ing1.15E-14
 Cancer (cell cycle/cell death)Ing1.93E-04
 Genes downregulated in c-Myc transgenic mice (307 genes)
 Cancer (cell cycle/cell death)Ing1.54E-05
 Lipid metabolismIng3.05E-04
 Cellular carbohydrate metabolismGO+Ing7.40E-04
 CytoskeletonGO8.80E-04
 CatabolismGO1.30E-03
Genes upregulated in E2f1/c-Myc transgenic mice (304 genes)
 Protein synthesisGO + Ing2.17E-09
 Oxidative phosphorylationGO + Ing2.54E-09
 Ubiquinone biosynthesisGO + Ing1.44E-05
Genes downregulated in E2f1/c-Myc transgenic mice (371 genes)
 Immune responseIng9.26E-05
 Cancer (cell cycle/cell death)Ing1.41E-03

Given the lack of a prominent functional oncogene-specific signature in HCC, we focused on the dysplastic stage to determine the early specific effects of E2f1, c-Myc, and their combination on the process of malignant transformation. All three transgenic lines showed the presence of widespread large cell dysplasia (Supplementary Fig. 1). Only livers without focal lesions as evidenced by histopathological evaluation were selected for microarray analysis. Functional classification of significant genes in dysplastic livers identified both common and specific functional modules in all three models (Table 1). Alteration of genes associated with cell cycle and cell death appeared as a recurrent feature in all three transgenic models and correlated with higher mitosis and apoptosis frequencies compared with age-matched wild-type mice (Supplementary Table 2). A common repression of genes involved in immune response was also prominent in all models, possibly due to alterations in hepatic function, because expression of innate immunity-related genes is a characteristic of the normal liver.26, 27 In contrast, several functional categories appeared to be differentially regulated by E2f1 and c-Myc. Thus, in c-Myc overexpressing livers, genes involved in protein synthesis and catabolism were respectively induced and repressed, reflecting a specific role for c-Myc in the control of translational machinery. In addition, genes implicated in lipid metabolism were found to be induced in E2f1 but repressed in c-Myc livers, suggesting that these two transcription factors can have opposite functional effects on specific gene modules. However, a combined overexpression of E2f1 and c-Myc specifically induced genes involved in oxidative phosphorylation (Table 1). Taken together, these observations suggest that E2f1 and c-Myc may promote malignant transformation through distinct mechanisms involving alterations in the metabolic functions of the hepatocyte.

Alteration of Genes Involved in Cell Cycle, Cell Death, and DNA Repair as Common Features of E2f1, c-Myc, and E2f1/c-Myc Dysplastic Livers.

The expression of more than 130 genes linked to the regulation of proliferation, apoptosis, and DNA repair was altered in dysplastic livers of all transgenic mice consistent with increased rate of proliferation and apoptosis as compared with age-matched wild-type controls (Supplementary Tables 2 and 3). Overrepresentation of these functional modules is consistent with the current view that E2f1 and c-Myc possess dual roles in regulating both proliferation and apoptosis.4, 14

Thus, two members of the CDK inhibitor family (p18 and p21) were induced in all models. These genes encode proteins that prevent the activation of CDKs, and therefore function as negative cell growth regulators by blocking G1-S progression.28 Expression of several CDK genes—namely Cdk2, Cdkl2, Cdk3, and Cdk7—as well as several cyclin genes—such as Ccnd1, Ccne1, Ccng2, Ccnm1, and Ccnm3—essential for cell cycle progression was also found to be altered in all 3 transgenic models.

Similarly, upregulation of genes involved in DNA surveillance and repair was not restricted to a particular transgenic model, and was found in dysplastic livers of E2f1 (Mlh1, Msh5, Msh6, Rad51ap1), c-Myc (Prp19, Rad21, Rad51l3), and E2f1/c-Myc (Rad51l3, Xrcc1) transgenic mice. However, expression of DNA replication regulators, such as replication factors Rfc2, Rfc4, and Rpa3, was increased predominantly in c-Myc-overexpressing mice.

Also, inducers and repressors of apoptosis were strongly affected by overexpression of E2f1 and c-Myc. Interestingly, at the early stage of the c-Myc-induced malignant transformation, positive (Trail or Trib3) and negative (Abl1, Bcl2, or Bcl-XL) regulators of apoptosis were respectively induced and repressed, a result consistent with a high frequency of apoptosis in this model. Binding of Trail to its receptor triggers apoptosis via the caspase pathway,29 whereas activation of the protein kinase Trib3 has been shown to negatively regulate Akt cell survival kinase, to inhibit nuclear factor κB, and to sensitize cells to tumor necrosis factor-induced and Trail-induced apoptosis.30, 31 In addition, the expression of 2 master inhibitors of apoptosis, Bcl2 and Bcl-XL, as well as the protein kinase Abl1, a product of the c-Abl proto-oncogene involved in the inhibition of mitochondrial cytochrome c release and caspase activation,32 was significantly reduced. In contrast, dysplastic livers from E2f1 mice showed activation of survival pathways, as illustrated by the induction of Akt, Fgf4, Hyou1, Map3k7, Peli2, or Smad4 and the repression of Atm, Casp1, Btg2, Nfkbiz, Tnfaip2, or Tnfrsf19, along with downregulation of apoptosis and faster tumor development compared with c-Myc transgenic mice.

Of significance, E2f1/c-Myc double transgenic mice not only shared common gene expression signatures with either c-Myc or E2f1 counterparts, but also displayed a unique dysregulation of genes involved in the proliferation and apoptosis consistent with the most rapid hepatocarcinogenesis among the three transgenic lines. For example, the expression levels of Mnat1, an assembly factor of the CDK-activating kinase complex composed of Cdk7 and cyclin H, was significantly induced only in E2f1/c-Myc dysplastic livers. Because deregulation of CDK-activating kinase by abrogation of Mnat1 has been shown to inhibit retinoblastoma protein phosphorylation and cyclin E expression,33 induction of Mnat1 in E2f1/c-Myc transgenic mice may result in the loss of CDK-activating kinase control, thus promoting uncontrolled proliferation. Similarly, downregulation of Dapk1, a tumor suppressor candidate gene, may further contribute to downregulation of apoptosis in this model.34

Because changes in messenger RNA and protein levels do not necessarily correlate, Western blot analysis was also performed for some selected genes to support the microarray data (Fig. 2A).

Figure 2.

Correlation of microarray data with in vivo phenotype. (A) Confirmation of microarray data via Western blot analysis. Whole cell lysates or nuclear proteins (in case of Srebp1) were separated via gel electrophoresis and probed with the indicated antibodies. Western blotting with β-actin antibody demonstrates an equal loading. (B) Increased liver mass in mice overexpressing c-Myc. Mice were sacrificed at 3 months of age (n ≥ 5 per group). Liver weight was expressed as a percentage of body weight (mean ± SE). (C) Increased ATP content in E2f1/c-Myc dysplastic livers. ATP content was determined as described in Materials and Methods at 3 months of age. In B and C, statistical differences were determined via Student t test. (D) Electron micrographs of E2f1 and c-Myc hepatocytes at the dysplastic stage. c-Myc hepatocyte exhibits well developed rough endoplasmic reticulum, frequent mitochondria, and functional nucleoli with or without central vacuole between the loops of nucleolonema (asterisk). E2f1 hepatocyte displays less endoplasmic reticulum, accumulation of free ribosomes, and a few lipid droplets. Scale bars, 2 μm. (E) Electron micrographs of nucleoli in c-Myc and E2f1 dyplastic hepatocytes. c-Myc nucleolus shows well-developed nucleolonema with numerous fibrillar centers and extensive granular component. E2f1 nucleolus is greatly reduced in size and shows features of segregation as well as accumulation of perinucleolar heterochromatin (white arrow) reflecting an impairment of nucleolar function. Scale bars, 1 μm. (F) Type of nucleoli found in c-Myc and E2f1 dysplastic hepatocytes. Nucleoli were divided into nucleolonemic, nucleolonemic-vaculated with central vacuole, and nucleolonemic with features of segregation based on morphometry. Ten to twenty serial sections of at least 45 nucleoli were analyzed for each genotype. WT, wild-type; ATP, adenosine triphosphate.

Alteration of Lipid Metabolism in E2f1 Transgenic Mice.

In addition to the genes involved in proliferation and apoptosis, gene expression profiling identified lipid metabolism as one of the most prominent functional modules that was altered in dysplastic livers of E2f1 single transgenic mice (Table 1). Specifically, most of the key enzymes of lipogenesis and cholesterogenesis were induced. These included: (1) ATP citrate lyase, which catalyzes the formation of acetyl-coenzyme A (CoA), the precursor of both fatty acids and cholesterol; (2) acetyl-CoA carboxylase, which is involved in the carboxylation of acetyl-CoA to malonyl-CoA, the primary rate-limiting step in fatty acid synthesis; (3) fatty acid synthase, which catalyzes the NADPH-dependent condensation of acetyl-CoA and malonyl-CoA into the long-chain saturated fatty acid palmitate; (4) isoforms 1 and 2 of fatty acid desaturase; (5) mevalonate pyrophosphate decarboxylase, which converts mevalonate pyrophosphate into isopentenyl pyrophosphate, one of the early steps in cholesterol biosynthesis; (6) lanosterol synthase; (7) sterol-C5-desaturase, which is involved in the conversion of lanosterol into cholesterol; and (8) glycerol-3-phosphate acyltransferase, which catalyzes the initial and committing step in glycerolipid biosynthesis and is predicted to play a pivotal role in the regulation of cellular triacylglycerol and phospholipid levels.

Among other induced genes were Insig1 and Insig2, both of which play a critical role in regulation of cellular cholesterol levels, and genes encoding enzymes of the fatty acid beta-oxidation pathway (enoyl-CoA hydratase 1, CoA dehydrogenase, carnitine acetyltransferase); members of ATP-binding cassette transporters involved in the peroxisomal import of fatty acids (Abcd3, Abcg8); apolipoprotein A-IV; acetyl-CoA acetyltransferase; carboxylesterase 3; hepatic lipase; phospholipid transfer protein, which transfers phospholipids from triglyceride-rich lipoproteins to high-density lipoprotein; and cytochrome P450 proteins (Cyp2a5, Cyp27a1, Cyp2b9, Cyp2c39).

Although previous large-scale studies have not revealed the genes involved in metabolism as primary targets of E2f1,5–7, 35 we identified sterol regulatory element binding factor 1 (Srebp1) as a candidate transcription factor that may relay the effects of E2f1 on lipid metabolism. Srebp1 is a member of the basic helix-loop-helix leucine zipper transcription factor family known as one of the master regulators of lipid metabolism.36, 37 It is synthesized as a precursor that upon cleavage translocates to the nucleus and activates transcription by binding to the sterol regulatory element 1 (SRE1). Several observations support the role for Srebp1 as a transcriptional effector of E2f1 on lipid metabolism. First, in E2f1 dysplastic livers messenger RNA levels of both Srebp1 and its numerous target genes involved in synthesis of cholesterol, fatty acid, triacylglycerol, and phospholipids were significantly increased (Table 2). Second, upregulation of the mature Srebp1 nuclear form was confirmed by Western blotting (Fig. 2A). Third, it has been recently shown that the kinase Akt induces transcription of genes involved in lipogenesis via activation of Srebp1.38 The active phosphorylated form of Akt, which is known to be transcriptionally regulated by E2f1,39 was increased in E2f1-overexpressing dysplastic livers (Fig. 2A). Fourth, several regulators of the Srebp pathway were also induced in E2f1 mice, including Mbtps1, a membrane-bound transcription factor protease site 1 gene, as well as Insig1 and Insig2, acting as an activator and sensor of Srebp, respectively. Mbtps1 cleaves endoplasmic reticulum membrane-bound Srebp and initiates the two-step proteolytic process required for Srebp nuclear translocation.40 In contrast, Insig proteins block the processing of Srebp by binding to the sterol-sensing domains of Srebp cleavage-activating protein, preventing the latter from escorting Srebp to the Golgi apparatus. Thus, upregulation of Insig genes may provide a feedback mechanism to reduce activation of Srebp.41 Upregulation of Srebp at the preneoplastic stage may also support the late induction of PPAR-γ in HCC and surrounding nontumorous liver (data not shown), because PPAR-γ is known to be transcriptionally regulated by both E2f1 and Srebp.20, 42

Table 2. Known Srebp1 Target Genes Upregulated in Dysplastic Livers of E2f1 Overexpressing Mice
Gene DescriptionSymbol
  • *

    Genes also upregulated in E2f1/c-Myc transgenic mice.

Acetyl-CoA carboxylase betaAcacb
Acyl-CoA synthetase long-chain family member 6Acsl6
ATP citrate lyaseAcly
ATP-binding cassette, subfamily G, member 8Abcg8*
Elongation of very long chain fatty acid-like 2Elovl2
Elongation of very long chain family member 5Elovl5
Enoyl-CoA, hydrataseEhhadh
Fatty acid desaturase 1Fads1
Fatty acid desaturase 2Fads2
Fatty acid synthaseFasn
GlucokinaseGck
Glutathione S-transferase, theta 3Gstt3*
Glycerol-3-phosphate acyltransferaseGpam
Insulin-induced gene 1Insig1*
Insulin-induced gene 2Insig2*
Lanosterol synthaseLss
Malic enzyme, supernatantMod1*
Mevalonate (diphospho) decarboxylaseMvd
Phospholipid transfer proteinPltp
Pyruvate dehydrogenase kinase, isoenzyme 1Pdk1*
Sterol regulatory element binding factor 1Srebf1
Sterol-C5-desaturaseSc5d

A misregulation of key enzymes involved in lipid metabolism is in agreement with fatty changes found in E2f1-transgenic livers.14 In humans, accumulation of fatty acids and triglycerides is a characteristic feature of nonalcoholic fatty liver disease, which may progress to cirrhosis and HCC. Obesity and fatty liver have also been shown to be risk factors for HCC. Thus, clinical and epidemiological data suggest that perturbation of lipid metabolism is strongly associated with hepatocellular damage and development of HCC.43

Upregulation of Genes Involved in Translation in c-Myc Transgenic Mice.

The vast majority of genes induced in dysplastic livers of c-Myc overexpressing mice with a known or inferred function were involved in ribosome biogenesis and control of translation. Accordingly, more than 50 genes encoding distinct components of 40S and 60S ribosomal protein subunits (Rps and Rpl genes, respectively) as well as numerous translation initiation/elongation factors (Eif3s2, Eif4a1, Eif4b, Eif4g3/Eef1b2, Eef2) were induced (Supplementary Table 4). Promoters of the genes encoding ribosomal proteins (e.g., Rpl10, Rpl13, Rpl13a, Rpl26, Rpl37, Rps13, Rps20) and translation initiation factors (Eif3s2, Eif4a1, Eif4b) harbor high-affinity binding sites for c-Myc, as revealed by chromatin immunoprecipiation experiments,44 supporting a direct involvement of c-Myc in the transactivation of these genes. Upregulation of translational machinery was found to correlate with liver mass increase in c-Myc and E2f1/c-Myc but not in E2f1 mice (Fig. 2B). A specific effect of c-Myc on hepatic translation and liver mass was consistent with other in vivo observations. Thus, adenoviral delivery of c-Myc in the liver has been previously shown to result in a specific induction of ribosomal genes along with a hepatocyte hypertrophy and nuclei enlargement.45 The effect of c-Myc on protein synthesis was also evident in electron microscopy studies. The ultrastructure of c-Myc overexpressing hepatocytes, nucleoli in particularly, was consistent with enhanced synthesis and processing of rRNAs (Fig. 2D–F). In contrast, E2f1 dysplastic hepatocytes frequently displayed signs of nucleoli atypia reflecting reduced ribosome biogenesis, again manifesting the profound differences between E2f1- and c-Myc–overexpressing mice.

Many reports have correlated alteration in the expression of ribosomal proteins and initiation translation factors with cancer.46 The upregulation of genes involved in protein synthesis may reflect an increased demand to support a high rate of cell proliferation, because a defective ribosome biogenesis induced by the conditional deletion of a single ribosomal protein-encoding gene has been shown to result in reduced cell proliferation.47 In addition to the principal role in translation, several of these genes have been shown to support additional cellular functions that may contribute to malignant transformation.46, 48

Concomitant Overexpression of E2f1 and c-Myc Enhances Mitochondrial Metabolism.

As expected, E2f1/c-Myc double transgenic mice also displayed misregulation of genes involved in lipid metabolism and translation, albeit to a lesser extent than either E2f1 or c-Myc mice, respectively.

More significantly, we identified mitochondrial metabolism as one of the most striking functional modules that was specifically increased in E2f1/c-Myc double transgenic livers. Thus, genes encoding various enzymes of mitochondrial metabolism (including Aldh5a1, Aldh6a1, Dhodh, Dmgdh, Etfa, Etfdh, Lias, Maoa, Opa1, Otc, Surf1, and Tomm70a) and mitochondrial ribosomal proteins (in particular Mrpl23, Mrpl34, Mrpl37, Mrps2, and Mrps15) were upregulated. Moreover, a prominent subset of genes associated with oxidative phosphorylation was strongly induced. These included genes encoding proteins in complex I (Ndufa10, Ndufb3, Ndufb5, Ndufc2, Ndufs2, and Ndufs4), which transfers electrons from NADH to the respiratory chain; complex II (Sdha), which oxidates succinate; complex III (Cytc1), which reduces cytochrome c; complex IV (Cox6a1, Cox7b), which catalyzes the electron transfer from reduced cytochrome c to oxygen; and complex V (Atp5d, Atp5g2, Atp5j, Atp5o), which catalyzes ATP synthesis. Moreover, induction of these genes correlated with an increase in ATP content in this model (Fig. 2C). Because ATP deprivation has been shown to result in cell cycle arrest,49 the enhanced mitochondrial activity and ATP production may sustain the increased rate of proliferation in E2f1/c-Myc double transgenic mice. Such a dramatic increase in mitochondrial metabolism might also reflect modifications in other functions of mitochondria in the cell—namely free radical generation and apoptosis50—that might also contribute to the acceleration of neoplastic development in E2f1/c-Myc transgenic mice.

This study provides a comprehensive overview of oncogene-specific gene expression signatures characteristic for various stages of HCC development. Despite the differences in kinetics of liver cancer development, the gene expression profiles of HCCs arising in E2f1, c-Myc, and E2f1/c-Myc mice significantly overlapped. In striking contrast, dysplastic livers were remarkably different at the gene expression level. Consistent with the postulated role for c-Myc and E2f1, both transcription factors regulated expression of genes involved in proliferation, apoptosis, and DNA repair. More significantly, this study revealed associations between overexpression of E2f1 and c-Myc and several unanticipated regulatory modules independent of proliferation and apoptosis. We demonstrated the capacity of E2f1 and c-Myc to induce or repress sets of genes that perturb unique metabolic pathways known to affect oncogenic development in the liver. Thus, at the early dysplastic stage, the E2f1 expression signature included a large number of genes associated with an induction of lipogenic enzymes that were not previously described as E2f1 targets. Srebp1 was identified as the transcription factor that may likely mediate the effect of E2f1 on lipid metabolism. Therefore, the early alterations in lipid metabolism characteristic for E2f1-driven oncogenesis may be responsible for the rapid initiation of liver cancer in this model. Independently, c-Myc activity at the dysplastic stage selectively triggered a global induction of translational machinery, whereas a combined overexpression of E2f1 and c-Myc increased expression of genes associated with mitochondrial respiration, underscoring the cooperative effects of c-Myc and E2f1 on energy production.

In conclusion, we hypothesize that E2f1 and c-Myc could contribute to hepatocarcinogenesis not only by regulating cell proliferation and cell death but by modulating specific metabolic functions (Fig. 3). However, additional experiments targeting key regulators of these pathways are required to formally test whether the metabolic alterations directly contribute to malignant transformation. Nevertheless, the observation that fully developed HCCs in transgenic mouse models no longer retain early oncogene expression signatures supports the critical role of these metabolic alterations in driving the carcinogenic process. Ultimately, identification of oncogene- and stage-specific signatures may be used to define new subsets of human HCC characterized by activation of c-Myc and E2f1, as well as for diagnostic and preventive purposes.

Figure 3.

Common and transgene-specific functional modules activated at the early stage of E2f1, c-Myc, and E2f1/c-Myc mediated hepatocarcinogenesis. Genes involved in the regulation of proliferation, apoptosis, and DNA repair were misregulated in all three transgenic models. In addition, specific metabolic pathways were independently activated in response to E2f1, c-Myc, or their combination. Overexpression of c-Myc affected protein metabolism, whereas E2f1 transgene caused dysregulation of genes involved in lipid metabolism, likely via activation of Srebp1 by Akt. Coexpression of both E2f1 and c-Myc affected energy metabolism in addition to activation of functional modules specific for single transgene. Enhanced ATP production in E2f1/c-Myc transgenic mice may sustain the increased proliferation rate and protein synthesis, and thus contribute to acceleration of neoplastic development in this model. Srebp1, sterol regulatory element binding factor 1; HCC, hepatocellular carcinoma.

Ancillary