Defective DNA strand break repair causes chromosomal instability and accelerates liver carcinogenesis in mice


  • Potential conflict of interest: Nothing to report.


Chromosomal instability is a characteristic feature of hepatocellular carcinoma (HCC) but its origin and role in liver carcinogenesis are undefined. We tested whether a defect in the nonhomologous end-joining (NHEJ) DNA repair gene Ku70 was associated with chromosomal abnormalities and enhanced liver carcinogenesis. Male Ku70 NHEJ-deficient (Ku70−/−), heterozygote (Ku70 +/−), and wild-type (WT) mice were injected with diethylnitrosamine (DEN), a liver carcinogen, at age 15 days. Animals were killed at 3, 6, and 9 months for assessment of tumorigenesis and hepatocellular proliferation. For karyotype analysis, primary liver tumor cell cultures were prepared from HCCs arising in Ku70 mice of all genotypes. Compared to WT littermates, Ku70−/− mice injected with DEN displayed accelerated HCC development. Ku70−/− HCCs harbored clonal increases in numerical and structural aberrations of chromosomes 4, 5, 7, 8, 10, 14, and 19, many of which recapitulated the spectrum of equivalent chromosomal abnormalities observed in human HCC. Ku70−/− HCCs showed high proliferative activity with increased cyclin D1 and proliferating cell nuclear antigen expression, Aurora A kinase activity, enhanced ataxia telangiectasia mutated kinase and ubiquitination, and loss of p53 via proteasomal degradation, features which closely resemble those of human HCC. Conclusion: These findings demonstrate that defects in the NHEJ DNA repair pathway may participate in the disruption of cell cycle checkpoints leading to chromosomal instability and accelerated development of HCC. (HEPATOLOGY 2008;47:2078–2088.)

Hepatocellular carcinoma (HCC) is the fifth most common cancer worldwide, with a rising incidence in most countries, including the United States.1 Because no effective therapies are available, the overall prognosis is poor. The risk factors for HCC are well known, and include chronic hepatitis B or C virus infection, alcoholism, aflatoxin exposure, nonalcoholic steatohepatitis, hepatic iron accumulation and, in general, hereditary liver diseases.1, 2 Despite the vast amount of information available regarding these risk factors, the molecular pathogenesis of HCC remains undefined.

There are two salient features in the development of HCC: (1) the great majority of these tumors, regardless of etiology, develop in cirrhotic livers, and (2) more than 90% of HCCs have chromosomal abnormalities.1–4 Cirrhosis is characterized by destruction of the hepatic lobular architecture and its replacement by nodules containing proliferative and apoptotic hepatocytes, in the presence of chronic inflammation and fibrosis. Cycles of apoptosis and regeneration in cirrhotic liver result in the growth of dysplastic nodules with a high frequency of monoclonality.5 Constitutive hepatocyte proliferation in this abnormal environment is believed to result in DNA damage leading to tumorigenesis.6, 7 Genomic instability caused by microsatellite instability that is associated with defects in mismatch repair mechanisms is present in some types of colorectal tumors. However, there is no firm evidence to suggest that microsatellite instability occurs in human or murine HCC. Instead, chromosomal abnormalities are present almost universally in HCC and are frequently detected in dysplastic nodules of cirrhotic livers. Some chromosomal defects found in the dysplastic nodules in cirrhotic liver are also present in HCC, which suggests that chromosomal defects occur at early stages of tumor development. However, the mechanisms by which chromosomal abnormalities occur during human or murine hepatocarcinogenesis are unknown. Here, we propose that defects in DNA strand break repair by nonhomologous end-joining(NHEJ) may lead to chromosomal instability and contribute to the development of HCC.

DNA double-strand breaks (DSBs) can be repaired by two mechanisms: homologous recombination and NHEJ. Homologous recombination leads to accurate repair whereas NHEJ may not.8, 9 The immediate response to DSB through the NHEJ repair pathway involves the binding of proteins at the site of breaks. Ku is a heterodimer of 70 kDA (Ku70) and 86 kDa (Ku86) subunits that bind to DSBs without DNA sequence specificity.8, 10 The DNA end-bound Ku heterodimer can associate with polymerases, ligases (XRCC4; DNA ligase), and the complex formed between the catalytic subunit of the DNA-dependent protein kinase (DNA-PKcs) and Artemis nuclease. The DNA-PK complex may signal DNA damage to the cell cycle/apoptosis machinery, activating cell cycle checkpoints that prevent replication of cells with unrepaired or misrepaired DNA damage.11 Ataxia telangiectasia mutated kinase (ATM), a serine/threonine kinase that binds to DNA breaks, functions as a sensor for cell cycle checkpoints by activating p53 and other cell cycle suppressors in cells with damaged DNA.12

To investigate whether a defect in NHEJ repair in hepatocytes would lead to chromosomal abnormalities and disruption of cell cycle checkpoints resulting in liver carcinogenesis, we analyzed the development of HCC in Ku70 null mice that had been injected with the alkylating agent diethylnitrosamine (DEN) at day 15 of age. Compared to wild-type (WT) littermates, Ku70−/− mice had a marked acceleration of hepatocarcinogenesis, which was associated with multiple chromosomal aberrations, loss of p53 protein expression and changes in cell cycle checkpoints.


DEN, diethylnitrosamine; DNA-PKcs, DNA-dependent protein kinase catalytic subunit; DSB, double-strand breaks; HCC, hepatocellular carcinoma; H&E, hematoxylin and eosin; Ku70−/−, Ku70 NHEJ deficient; Ku70+/−, Ku70 heterozygote; NHEJ, nonhomologous end joining; WT, wild type.

Materials and Methods


All mice were maintained in a specific pathogen-free facility, fed on a commercial pellet diet, and allowed water ad libitum. Experimental animal protocols and animal procedures complied with the International Criteria of Animal Experimentation and were approved by the University of Washington Animal Care Committee. Mice heterozygous for the Ku70 mutation (Ku +/−) were indistinguishable from wild-type (Ku 70 +/+ or WT) mice in all physical aspects. However, Ku70 null (Ku70−/−) mice were significantly growth retarded by 50% compared with their control littermates; these observations corroborate those made by Gu et al.10 We administered DEN at 5 mg/kg intraperitoneally to male mice of all genotypes at day 15 of age13; Ku70−/− and WT control animals were injected with 0.9% saline intraperitoneally. Five and up to 11 mice were studied for each timepoint, treatment group, and genotype. The genotype of each animal was confirmed by reverse transcription polymerase chain reaction (RT-PCR).10 The strain of mice used in these studies were 129Svj.

Tumor Harvest and Histology.

Animals were killed at 3, 6, and 9 months by CO2 narcosis. Tumors, liver tissue, and lungs were removed and fixed in 10% neutral buffered formalin overnight or snap-frozen in liquid nitrogen for RNA and protein analysis. Liver sections (5 μm) were cut from paraffin-embedded blocks and stained with hematoxylin and eosin (H&E) for histological examination.

Karyotype Analysis of Liver Tumors.

Cell cultures were prepared from HCC arising in Ku70−/− (6 and 9 months), Ku70+/−, and WT (9 months) mice. Tumor tissue was finely minced with a razor blade and digested in 0.1% collagenase (Sigma, St. Louis, MO) and Williams E media supplemented with 10% fetal bovine serum at 37°C for 15 minutes with gentle agitation. The cell suspension was washed several times in Seglen's media followed by ethylene glycol tetraacetic acid, plated, and grown on plastic dishes containing Williams E media and 10% fetal bovine serum for 72 hours. Following passage 1, cells were harvested with trypsin and Colcemid treatment (Invitrogen, Carlsbad, CA), and chromosome spreads were prepared and stained with Giemsa for G-banding. At least 10 metaphase cells were scored for each HCC isolated from an individual animal of each genotype (WT, Ku70+/−, Ku70−/−), experimental timepoint (6 and 9 months), and DEN or saline treatment groups.

RNA Isolation and Semiquantitative RT-PCR.

Frozen liver tissue was disrupted using a rotor-stator homogenizer. RNA was extracted and isolated from liver homogenates using the RNeasy Midi Kit (Qiagen, Valencia, CA) according to the manufacturer's instructions. RT-PCR analyses, which were performed on the Superscript One-Step RT-PCR kit (Invitrogen), used 1 μg of purified total RNA. For amplification of p53, reverse transcription was performed at 42°C for 60 minutes, followed by denaturing at 99°C for 10 minutes and amplification for 27 cycles with the following parameters: denaturing at 94°C for 30 seconds, annealing at 62°C for 30 seconds, and extension at 72°C for 1 minute. Amplification conditions for Aurora A kinase were as above, except denaturing temperature was 94°C and for 25 cycles. Primers were as follows: p53: Forward CTACAAGAAGTCACAGCACAT; Reverse GTCTTCCAGTGTGATGATGGT; Aurora A kinase: Forward AAGTTCATCCTGGCTCTGAAG; Reverse CTAGGAACTCATAGCAGAGAAC.

Real-Time RT-PCR.

One microgram of total RNA from livers was used for the RT reaction performed on Applied Biosystems' TaqMan reverse transcription kit, and one-tenth of the reaction was employed for real-time PCR analysis. Assay kits for mouse Ube1A, ubiquilin, and ubiquitin D from Applied Biosystems (ABI, Foster City, CA) were used for the quantification of the expression of these genes in an ABI 7900HT sequence detection system. The housekeeping gene, mouse β-glucuronidase (ABI) was used to normalize the measurements using the delta-delta Ct method. PCR conditions were the standard when using TaqMan hydrolysis probes.

Analysis of Cell Cycle Proteins and Components of DSB Repair Pathways.

Liver protein and tumor cell lysates were prepared as previously described.13 Expression of p53, and of components of DSB repair pathways (DNA-PKcs, ATM, phospho-ATM, Rad51) were analyzed by immunoblotting. Proteins (20 μg per lane except where indicated) were separated on a 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis gel, transferred to polyvinylidene fluoride membranes (Millipore, Billerica, MA) and blocked with 2% milk containing 0.1% Tween-20 prior to incubation with primary antibody and the appropriate secondary antibody. Chemiluminescence was performed with Super Signal reagent (Pierce, Rockford, IL). Primary antibodies used were as follows: murine double minute 2 (mdm2; Santa Cruz Biotechnology, Santa Cruz, CA; #sc-965), DNA-PKcs (Cell Signaling Technology; #4602), p53 (Calbiochem, San Diego, CA; #OP03), cyclin D1 (Santa Cruz Biotechnology; #sc-450), proliferating cell nuclear antigen (PCNA; Santa Cruz Biotechnology; #sc-56), ATM (Genetex, San Antonio, TX; #MS-ATM11-PX1), phospho-ATM (Upstate, Waltham, MA; #05-740), and Rad51 (Upstate; #05-530). Secondary antibodies (Amersham, Piscataway, NJ) were: anti-mouse immunoglobulin G, horseradish peroxidase–linked whole antibody (from sheep) and anti-rabbit immunoglobulin G, horseradish peroxidase–linked whole antibody (from donkey). β-Actin was used as loading control. Relative density of protein bands were determined using Scan Analysis image quantification software shown in arbitrary units and normalized to β-actin. Results were expressed as mean ± standard deviation (SD), n = 4 per group.

p53, Phospho-Histone H3, and Ubiquitin Immunohistochemistry.

Paraffin-embedded liver sections were quenched to remove endogenous peroxidase activity using 3% hydrogen peroxide, after which antigen retrieval was performed with heated citric acid buffer. Slides were incubated with 1:100 dilution of rabbit p53 (Santa Cruz Biotechnology; #sc-6243), 1:100 rabbit phospho-histone H3 antibody (Ser 10) (Cell Signaling Technology; #9701), and 1:250 rabbit ubiquitin antibody (Neomarkers, Fremont, CA; #RB-9203-P) for 1 hour at room temperature. Detection of the primary antibody was carried out using the appropriate biotinylated antibody (Vector Laboratories, Burlingame, CA) and peroxidase DAB kit (Ventana, Concord, CA).

Reagents Used in mdm2 and Proteasome Inhibitor Studies.

The effects of MG-262 (selective 26S proteasome inhibitor; Biomol, Plymouth Meeting, PA) and Nutlin-3-racemic (mdm2 antagonist; Calbiochem) on the levels of p53 and mdm2 were determined by incubation of each reagent at 2, 5, or 10 μM concentrations with Ku70−/− primary HCC cells for 6 hours. Treated Ku70−/− HCC cells were then prepared into lysates for immunoblotting studies.

Statistical Analyses.

The Mann-Whitney test was used for the comparison of data from different treatment groups. Results (mean ± SD) were considered significant when P was less than 0.05. Statistical analyses were performed using GraphPad PRISM software.


Ku70 Deficiency Accelerates Tumor Development.

WT, Ku70 +/ −, Ku70−/− mice injected with DEN, and noninjected WT mice were examined at 3, 6, and 9 months for the presence of liver tumors. A minimum of 5 (and up to 11) animals from each genotype for each treatment or noninjected group were studied. There was no evidence of liver neoplasia at 3 months in mice from any of the groups. However, by 6 months, liver tumors were either grossly or histologically evident in 80% of DEN-injected Ku70−/− mice (Table 1, Fig. 1A,B). These tumors were well-differentiated HCCs of a solid pattern. Histological examination from liver tumors from all mice at 9 months revealed multifocal nodules of poorly differentiated HCC with areas of hemorrhage, necrosis, and hypervascularity. Because it was often difficult to separate each tumor from surrounding liver tissue, we estimated tumor burden by liver:body weight ratios; 60% of DEN-injected Ku70−/− mice that developed tumors at 6 months had a liver-to-body weight ratio (LBWR) exceeding 7% compared to LBWR of 4%-5% in DEN-untreated controls (Fig. 1C). By 9 months, DEN-injected mice from all genotypes developed HCC with lung metastases in 45% in WT, 64% in heterozygous, and 60% in Ku70−/− mice (Fig. 1D). Ascites and weight loss were not observed in DEN-treated mice from all genotypes. No other metastatic deposits were seen in the peritoneum, lymph nodes, and adrenal glands; long bones, vertebrae, and brain were not routinely sampled in these studies. Female Ku70−/− mice injected with DEN also developed accelerated hepatocarcinogenesis at the same rates as male Ku70−/− mice (data not shown). Because of known gender differences that could confound HCC development, subsequent studies were confined to male mice only.

Table 1. Incidence of Hepatocellular Carcinoma in Ku70 Mice Treated with Diethylnitrosamine or Saline Only According to Genotype and Age at Analysis
Genotype3 Months6 Months9 Months
+/+0 (0/8)0 (0/10)82% (9/11)
+/−0 (0/9)0 (0/10)91% (10/11)
−/−0 (0/5)80% (4/5)100% (5/5)
+/+0 (0/5)0 (0/5)0 (0/8)
+/−0 (0/5)0 (0/5)0 (0/6)
−/−0 (0/5)0 (0/5)0 (0/5)
Figure 1.

Ku70 deficiency confers susceptibility to DEN-induced hepatocarcinogenesis. (A) Gross morphology of liver tumor from DEN-treated Ku70−/− mouse at 6 months. Arrowheads indicate macroscopic HCC. (B) Representative H&E-stained section of Ku70−/− hepatocellular carcinoma (HCC) at 6 months (magnification ×100). (C) Liver tumor burden in DEN-injected Ku70 mice of different genotypes. Increased liver-to-body weight ratios indicative of liver tumor burden (measured by dividing total weight of the liver by total body weight and expressed as a percentage) were noted earlier in Ku70−/− mice. (D) Representative H&E-stained section from DEN-treated Ku70−/− mouse liver at 3 months showing a dysplastic focus at ×400 magnification. (E) Accelerated development of proliferative foci in Ku70−/− DEN-injected mice. The number of foci per square centimeter of liver tissue was counted and expressed as mean ± SD [*†P < 0.01 compared with control (non–DEN-injected), wild-type (+/+), and heterozygote (+/−) mice at 3 and 6 months (mths), respectively).

Tumor development in DEN-injected Ku70−/− mice was preceded by the formation foci of altered hepatocytes (Fig. 1D,E). We examined liver sections from all mice at 3 and 6 months of age to determine the number of abnormal foci present/cm2. At 3 months, dysplastic foci were present only in livers of Ku70−/− mice. By 6 months, the number of foci of altered hepatocytes was at least three-fold higher in Ku70−/− mice compared to WT mice (12 ± 2.6 versus 4.2 ± 2.6 abnormal foci present/cm2 respectively) (Fig. 1D,E). The dysplastic features of such foci include nuclear anisocytosis, hyperchromasia, pleomorphism, and increased nuclear to cytoplasmic ratio. These results provide clear evidence that hepatocarcinogenesis, which is preceded by early dysplastic changes, is accelerated in animals defective for the Ku70 gene of the NHEJ DNA repair pathway, compared with their WT and heterozygote littermates.

Increased Hepatocyte Proliferation and Mitosis in Ku70−/− Mice.

Both cellular proliferation and mitosis were increased in Ku70−/− tumors. In marked contrast to HCC from WT mice, liver tumors from Ku70−/− mice exhibited a striking number of mitotic figures (Fig. 2A), many of which were multipolar with condensed and asymmetric chromatin aggregation (Fig. 2B). Phosphorylation at serine 10 of histone H3 is associated with chromosome condensation during mitosis, and is a good marker for cell replication.15 Immunohistochemistry for phospho-histone H3 in Ku70−/− HCC sections showed significantly increased positive nuclear staining (Fig. 2C; Supplementary Fig. 1A) compared with WT tumors (Fig. 2C; Supplementary Fig. 1B). These findings demonstrate that HCCs from Ku−/− mice have much higher cellular proliferation and mitosis than do WT tumors.

Figure 2.

Increased mitosis and minimal apoptotic cell death in Ku70−/− HCCs. (A) Multiple mitotic figures (indicated by arrowheads) seen in a representative H&E-stained liver tumor section from a Ku70−/− mouse (×200). (B) Abnormal mitoses (arrowheads) with condensed chromatin at higher magnification (×400) in Ku70−/− HCC. (C) Increased phosphohistone H3 immunostaining in Ku70−/− HCCs. The number of positive-staining nuclei was noted and expressed per 3000 hepatocytes per high-power (×200) field. At least 10 high-power fields were examined for n = 4 mice per experimental group (P < 0.01 compared to 6-month WT (*) and Ku70−/− nontumorous (6-month −/−) (†) livers.

In response to stimuli that induce DNA damage, normal cells activate checkpoint mechanisms that inhibit cell cycle progression or induce apoptosis; p53 is most frequently involved in such checkpoint responses.16, 17 To study changes in cell cycle progression in liver tissue and tumors, we analyzed the expression of several cell cycle proteins. Cyclin D1 and PCNA (Fig. 3A,B, respectively) were highly induced in liver and tumors from Ku70 null mice compared to WT or Ku70 heterozygote animals at 6 and 9 months. The enhanced expression of proteins involved in cell cycle, together with the increased number of mitotic figures in HCC from Ku70−/− mice (Fig. 2A) are consistent with the loss of cell cycle regulation.

Figure 3.

Increased proliferative hepatocyte activity in Ku70−/− liver tumors. Liver and tumor lysates were analyzed by immunoblotting. T denotes tumor tissue. (A) Marked induction of cyclin D1 protein expression in Ku70−/− HCCs compared with WT. (B) Increased PCNA levels in Ku70−/− HCCs. (C) Aurora kinase A is overexpressed earlier in Ku70−/− HCC than in WT HCC: Aurora kinase A mRNA expression was determined by semiquantitative RT-PCR. The 18S mRNA was used as an internal control Abbreviations: HCC, hepatocellular carcinoma; L, liver surrounding HCC. P < 0.01 compared to 6-month WT (*) and Ku70−/− nontumorous (6-month −/− L) (†) livers.

Aurora A Kinase Is Overexpressed Early in Ku70−/− Liver Tumors.

Aurora A kinase (Aurora A) plays an important role in chromosome segregation during cell division.18 Overexpression of this enzyme in mammalian cells induces centrosome amplification and chromosomal instability.19, 20 Thus, overexpression of Aurora A provides a potential mechanism that links mitotic abnormalities resulting in numerical and structural aberrations, cell cycle defects and carcinogenesis.

Semiquantitative RT-PCR of RNA extracted from tumors derived from Ku70−/− and liver from WT mice at 6 months showed that Ku70−/− HCC expressed a three-fold increase in levels of Aurora A compared with WT liver (Fig. 3C). Moreover, in Ku70−/− mice, expression of Aurora A messenger RNA (mRNA) was markedly increased in HCCs in contrast to macroscopically normal liver tissue surrounding tumor. By 9 months, Aurora A mRNA expression was similar in HCCs from mice of both genotypes. These data suggest that Aurora A expression is an early marker for chromosomal instability and mitotic abnormalities in Ku70−/− mice.

Ku70−/− -Derived Liver Tumors Have a High Rate of Chromosomal Abnormalities.

To determine whether deficiency in Ku70 was associated with chromosomal abnormalities, we isolated HCCs from Ku70−/− (6 months) (Fig. 4A), Ku70 +/− (9 months) and WT (9 months) mice, and prepared cell cultures from these tumors. Primary cultures of HCC cells remained stable for up to 28 weeks, in contrast to cultures of normal hepatocytes, which rapidly degenerated after 14 days. We performed cytogenetic analysis of metaphase chromosomes from passage 1 cells derived from HCCs of Ku70−/− mice at 6 months (Fig. 4B), WT mice (Fig. 4C), and Ku70 +/− (Fig. 4D) mice at 9 months.

Figure 4.

High rate of chromosomal abnormalities in Ku70−/− derived hepatocellular carcinomas (HCCs). Primary liver tumor cell cultures were prepared from HCCs arising in Ku70 mice of all genotypes at 6 and 9 months (n = 5-11 per experimental group). (A) Primary HCC cell culture derived from Ku70−/− mouse at 6 months (passage 1; magnification ×1000). (B, C) Primary HCC cells were harvested at passage 1, and metaphase spreads were prepared for karyotype analysis. At least 10 cells were scored for each tumor from an individual animal, with a minimum of 50 karyotypes examined from HCCs derived from each genotype and at each time point. Representative karyotypes from (B) Ku70−/− and (C) WT liver tumors at 9 months. Numerical chromosomal aberrations indicated by arrows. In (B), marker chromosomes are circled. (D) High rate of numerical and structural chromosomal abnormalities observed in Ku70−/− tumors (*†P < 0.01 compared with DEN-injected wild-type (+/+) mice. (E) Enhanced ATM phosphorylation associated with Ku70 deficiency. Liver and tumor homogenates from Ku70 wild-type (+/+), heterozygote (+/−), and null (−/−) mice were analyzed by immunoblotting. T denotes tumor tissue. Phosphorylated ATM was significantly increased in Ku70−/− normal, dysplastic, and neoplastic liver.

Multiple chromosomal aberrations were evident in HCC cells cultured from Ku70−/− mice, with 95% of metaphases examined displaying aneuploidy and marker chromosomes (Fig. 4B,D) which were indicative of profound chromosomal structural changes. Clonal aberrations were detected in chromosomes 4, 5, 7, 8, 10, 14, and 19 (Supplementary Table 1). By contrast, only 2.5% and 3% of metaphases from HCCs of WT and heterozygous mice, respectively, had abnormal karyotypes (Fig. 4C,D). Hence, the absence of the Ku70 NHEJ pathway greatly increased the frequency of chromosomal changes in liver tumors. Notably, the chromosomal abnormalities detected in HCC from Ku70−/− mice recapitulate many of the changes described in human HCC, as shown in Supplementary Table 1.3, 21, 22

Increased ATM Phosphorylation in Ku70-Deficient Mice.

The cellular responses to DSBs are mediated in part by ATM, a protein kinase of the phosphoinositide 3-kinase family. In response to DSB, ATM undergoes autophosphorylation and initiates a complex response, which is in part mediated by p53.9, 12 We found enhanced expression of phosphorylated ATM (Ser 1981) (phospho-ATM) in DEN-untreated Ku70−/− liver, suggesting that these animals may have an increased but low, basal level of DNA damage (Fig. 4E). More strikingly, in DEN-injected Ku70−/− mice, there was an at least two-fold increase in phospho-ATM (over basal levels) in liver tissue at 3 months, and in HCCs at 6 and 9 months (Fig. 4E). These findings suggest that the alkylating effect of DEN enhances and perpetuates DNA damage in Ku70−/− mice.

The NHEJ pathway for DSB repair requires multiple gene products that include Ku70, Ku86, XRCC4, DNA ligase IV, DNA-PKcs, and Artemis. The Ku70/Ku86 heterodimer binds to DNA ends and associates with other repair complexes. Among these is the complex formed between DNA-PKcs, the catalytic subunit of DNA-dependent protein kinase, and the Artemis nuclease, which trims the ends of DSBs.23 We investigated whether DNA-PKcs would be altered in Ku70−/− mice. DNA-PKcs protein expression was unmodified in all liver samples irrespective of genotype, age, treatment with DEN, or tumor phenotype (Supplementary Fig. 2A), suggesting that this arm of the NHEJ pathway is intact in both normal liver tissues and HCCs from Ku70−/− mice. We also examined whether expression of Rad51, an important component of the homologous recombination DNA repair system would be affected in Ku70−/− mice. Rad51 was uniformly detectable by immunoblotting in whole liver homogenates from all mice aged 3, 6, and 9 months, and its expression was increased in HCCs of mice of all genotypes compared to nontumorous liver (Supplementary Fig. 2B). These observations imply that the overexpression of Rad51 in Ku70-deficient mice did not fully compensate for the NHEJ repair defect.

p53, mdm2 Expression and Ubiquitination in Ku70−/− Liver Tumors.

Deficiency of p53 is commonly found in human HCCs, thereby facilitating the expansion of preneoplastic lesions.24, 25 On the basis of these observations, we examined p53 expression in normal liver and tumor tissue from WT, Ku70 +/−, and Ku70−/− mice. The p53 protein expression was intact in DEN-treated WT and Ku70 +/− mice up to 9 months (Fig. 5A). However, expression of p53 protein was lost by 3 months in Ku70−/− mice and was undetectable in tumors obtained at 6 and 9 months (Fig. 5A). In some animals, the paucity of p53-positive cells in Ku70−/− HCCs was confirmed by immunohistochemical staining. (Supplementary Fig. 3A,B). These findings suggest that the loss of p53 may precede the clonal expansion of preneoplastic hepatocytes and could be permissive for the perpetuation of cells with DNA damage likely to form HCC in Ku70−/− mice.

Figure 5.

Changes in p53 expression in Ku70−/− liver tumors. (A) Loss of p53 protein in Ku70−/− preneoplastic liver and HCCs. Expression of p53 protein determined by immunoblotting. PC and T indicate positive control and tumor tissue, respectively. (B) Preserved p53 mRNA expression in Ku70−/− liver and tumor tissue determined by semiquantitative RT-PCR. β-Actin was used as an internal control Abbreviations: HCC, hepatocellular carcinoma; L, liver surrounding HCC.

The loss of p53 protein detected in Ku70−/− mice could result from inhibition of transcription or defects in posttranscriptional mechanisms. To answer this question, we performed semiquantitative RT-PCR to determine p53 mRNA expression in Ku70−/− liver and tumor tissue. As shown in Fig. 5B, p53 mRNA was present in HCCs from all mice, indicating that changes in p53 protein expression in Ku70−/− HCCs were most likely a consequence of altered posttranscriptional regulation of p53. However, we note that the intact expression of p53 mRNA does not exclude the possibility of point mutations in the coding region of the p53 gene.

One of the important mechanisms of posttranslational control of p53 is its degradation through ubiquitination.26, 27 We examined several components of ubiquitin pathways and detected differences between livers of WT and Ku 70−/− mice. We also evaluated the extent of ubiquitination in tumors of WT and Ku70−/− mice by immunohistochemistry (Supplementary Fig. 4) and real-time PCR for Ube1A and ubiquilin (Fig. 6A,B). As shown in Fig. 6, the degree of ubiquitination was significantly higher in Ku70−/− HCCs; these tumors showed increased Ube1A ubiquitin-like activating enzyme E1A (also known as SUMO-1 activating enzyme, a sentrin activating enzyme involved in sumoylation)28 and ubiquilin, a mammalian target of rapamycin (mTor)-interacting protein.29 In addition, ubiquitin D expression was significantly increased in livers and tumors of Ku 70−/− mice compared to the same tissues in WT mice (Fig. 6C). Ubiquitin D, also known as FAT10, participates in the regulation of chromosomal stability, is cell cycle regulated and up-regulated in human HCC and other cancers.30–32 Ku70−/− HCCs also demonstrate nuclear and cytoplasmic staining for ubiquitin compared to WT HCC and normal liver (Supplementary Fig. 4).

Figure 6.

Ku70-defective HCCs display increased ubiquitination transcript profiles compared to wild-type liver tumors. Real-time PCR for (A) Ube1A, (B) ubiquilin, and (C) ubiquitin D was performed on HCCs derived from WT and Ku70−/− mice. P < 0.01 for 6-month Ku70−/− HCCs compared with 9-month WT HCCs.

If posttranslational modifications such as ubiquitination could account for diminished levels of p53 protein, we reasoned that chemical inhibition of the 26S proteasome or mdm2 should counteract p53 protein loss. We therefore examined the effect of MG-262, a selective 26S proteasome inhibitor,33 and Nutlin-3, an mdm2 antagonist34 on the levels of p53 and mdm2 in Ku70−/− HCC cells in primary culture. Incubation of these cells with MG-262 led to the restoration of p53, without affecting mdm2 levels (Fig. 7A,B). However, exposure of Ku70−/− HCC cells to Nutlin-3 did not modify p53 protein expression, despite adequate inhibition of mdm2 expression with higher concentrations of Nutlin-3 (Fig. 7B). Taken together, these results suggest that the regulation of p53 in Ku70−/− HCCs occurs posttranslationally, whereby increased proteasomal degradation of p53 is critical.35–38 Although mdm2, (an E3 ubiquitin-protein ligase) is a candidate pathway for p53 proteasomal destruction, our results suggest that it may not be the main mediator in Ku 70−/− mice.39, 40

Figure 7.

Changes in p53 and mdm2 expression in Ku70−/− liver tumors. (A) Expression of mdm2 in liver and tumor homogenates from Ku70 wild-type (+/+), heterozygote (+/−), and null (−/−) mice by immunoblotting. T denotes tumor tissue. (B) Increased p53 stability with proteasome inhibition in Ku70−/− primary HCC cells. The effects of MG-262 (selective 26S proteasome inhibitor) and Nutlin-3-racemic (mdm2 antagonist) on the levels of p53, mdm2 in Ku70−/− primary HCC cells were determined by immunoblotting. Each lane contained 30 μg of cell lysate. Incubation of Ku70−/− HCC cells for 6 hours with MG-262 led to an accumulation of p53 with stable mdm2 levels. Treatment with Nutlin-3 did not restore p53 protein expression.


The pathogenesis of HCC remains undefined, despite detailed knowledge about its etiology. The two most conspicuous features of HCC development include its association with cirrhosis and the development of chromosomal abnormalities. A plausible view is that cells in dysplastic foci in cirrhotic livers acquire DNA damage as a consequence of their proliferation in an oxidative environment caused by chronic inflammation. Damaged cells, at least some of which have chromosomal defects, proliferate to form dysplastic nodules that are precursors of HCC. In these studies, we describe a novel somatic mouse model for HCC whereby a deficiency in Ku70, a gene that participates in DSB repair by the NHEJ pathway,10 causes widespread chromosomal aberrations, disrupts cell cycle checkpoints, and accelerates liver carcinogenesis.

To generate HCCs in Ku70−/− mice, we injected newborn animals with the alkylating agent DEN and studied tumor development for a period of 9 months. DEN-induced hepatocarcinogenesis was chosen because of its relevance to human HCC. Lee et al.21 studied global gene expression patterns of HCCs from seven different mouse models and compared them with human HCCs from predefined subclasses in order to examine how well these models resemble human HCC phenotypes. They found gene expression patterns from DEN-induced mouse HCCs were most similar to those of the poorer survival group of human HCCs.

Typical HCCs appeared at least 3 months earlier in Ku70−/− mice compared to WT or Ku70 heterozygous animals. In addition to numerical and structural chromosomal aberrations, HCCs from Ku70−/− animals had a higher rate of cell proliferation, abundant mitotic figures, and increased expression of Aurora A kinase, a protein that links defects in chromosome segregation during mitosis with cell cycle abnormalities and carcinogenesis.

The chromosomal aberrations detected in this mouse model of hepatocarcinogenesis recapitulate a significant number of changes described in human HCC. Deletions of chromosome arms 1p, 4q, 6q, 9p, 13q, 16q, and 17p are the most commonly described abnormalities in analyses of human HCC, suggesting the presence of important tumor suppressor genes in some these loci.1, 3 The chromosomal changes in humans are orthologous to mouse chromosomes 4, 5, 8, 10, 11, 14, and 17. We found clonal loss of chromosomes 4, 5, 7, 8, 10, 14, and 19 in liver tumors from Ku70-defective mice, demonstrating a potential similarity with human HCC. Further studies are underway to identify specifically which deleted genes are relevant to the initiation and progression of HCC in the Ku 70−/− mouse model.

Loss of heterozygosity at chromosome 17p13, the p53 locus, occurs in 25%-60% of human HCCs.41 Analysis of “nodule-in-nodule” development of human HCC showed that p53 mutation was associated with the progression of HCC from dysplasia to carcinoma. To date, p53 mutations have not been described in carcinogen-induced mouse liver tumors.42–44 We suggest that defective DNA repair in Ku70−/− causes chromosomal instability, and is coupled with a failure in cell cycle checkpoint control resulting from loss of p53. These changes allow cells with DNA damage that have a proliferative advantage to replicate and eventually produce tumors. In preliminary studies to elucidate whether Ku70 deficiency alone in the absence of DNA damage can accelerate tumorigenesis, we crossbred Ku70−/− mice with single transforming growth factor-α transgenic animals; Ku70-deficient/transforming growth factor-α +/− naïve mice (non-DEN injected) did not display accelerated HCC development, which suggests that DNA damage may be required in addition to NHEJ deficiency to enhance hepatocarcinogenesis (Supplementary Table 2). Notably, there is mounting evidence that abnormalities in telomere length and telomerase activity in response to DNA damage can trigger genomic instability in cell cycle checkpoint defective cells.45 The role of telomere dysfunction mediating chromosomal instability and hepatocarcinogenesis may well be pertinent to HCC development in Ku70−/− liver in light of aberrant p53 regulation and deficient NHEJ repair pathways.

Posttranslational modifications of the p53 product can regulate the levels of the protein.26, 35–37 An important posttranslational regulatory mechanism is ubiquitination. Using real-time RT-PCR, we found that ubiquitination transcripts, Ube1A, and ubiquilin were significantly increased in HCC of Ku70−/− compared to tumors from WT mice. Increased expression of ubiquitin genes or related family members have been found in various gene-expression studies in human HCC, and in some cases, expression of these genes correlated with the disease prognosis.21, 30, 46, 47 We also detected an increase in the expression of ubiquitin D (FAT10) transcripts in livers and tumors of Ku70−/− mice, compared to the same tissues in WT mice. The high expression of ubiquitin D in Ku70−/− mice is consistent with the conclusion that chromosomal instability is a prominent factor in the development of HCC in these animals.30–32

The mdm2-negative regulatory feedback loop with p53 is an important regulatory mechanism of p53 levels.38, 39 It is possible that increased expression of mdm2 could be in part responsible for the decrease of p53 expression in Ku70−/− mice. However, studies with cells from Ku70−/− HCCs in culture showed that blockage of mdm2 activity did not change the levels of p53. However, p53 expression was restored to normal levels by MG-262, a proteasome inhibitor, suggesting that other mechanisms of ubiquitination may be involved.

In summary, we show that disruption of NHEJ DNA strand break repair by Ku70 deficiency predisposes hepatocytes to chromosomal instability, disrupts cell cycle checkpoints, and accelerates liver carcinogenesis. In this mouse model, the initial DNA injury was produced by injection of an alkylating agent. It remains to be determined whether similar DNA damage and repair deficiency occurs in the oxidative environment of human cirrhotic livers.


We thank G. C. Farrell for critical review of the manuscript and R. Bauer, J. Brooling, J. Campbell, and B. Molloy for technical support, helpful comments, and discussion.