Oxidative DNA damage is unavoidably and continuously generated by oxidant byproducts of normal cellular metabolism. The DNA damage repair genes, mutY and mutM, prevent G to T mutations caused by reactive oxygen species in Escherichia coli, but it has remained debatable whether deficiencies in their mammalian counterparts, Myh and Ogg1, are directly involved in tumorigenesis. Here, we demonstrate that deficiencies in Myh and Ogg1 predispose 65.7% of mice to tumors, predominantly lung and ovarian tumors, and lymphomas. Remarkably, subsequent analyses identified G to T mutations in 75% of the lung tumors at an activating hot spot, codon 12, of the K-ras oncogene, but none in their adjacent normal tissues. Moreover, malignant lung tumors were increased with combined heterozygosity of Msh2, a mismatch repair gene involved in oxidative DNA damage repair as well. Thus, oxidative DNA damage appears to play a causal role in tumorigenesis, and codon 12 of K-ras is likely to be an important downstream target in lung tumorigenesis. The multiple oxidative repair genes are required to prevent mutagenesis and tumor formation. The mice described here provide a valuable model for studying the mechanisms of oxidative DNA damage in tumorigenesis and investigating preventive or therapeutic approaches.
Oxidative DNA damage has been considered a key factor in tumorigenesis for decades (1) , but direct causative evidence is lacking and the molecular mechanisms remain obscure. The most frequent mutagenic base lesion caused by oxygen free radicals is 7,8-dihydro-8-oxoguanine (8-oxoG or GO) that can result in adenine mismatches during DNA replication (2 , 3) . The genes that prevent the mutagenic effect of GO are highly conserved from Escherichia coli to human cells. The E. coli MutM DNA glycosylase removes GO directly from GO:C pairs. The MutY DNA glycosylase excises an adenine mismatched with GO, allowing DNA polymerases to restore a GO:C pair that can be acted on by the MutM enzyme (3 , 4) . Genetic approaches indicated that these two DNA glycosylases synergistically prevent G to T mutations (3) . Mammalian OGG1 and MYH carry out the same in vitro reactions as their E. coli counterparts, MutM and MutY, respectively (5, 6, 7) , but the biological importance of these two genes in vivo has remained unclear. OGG1 knockout mice exhibit elevated levels of GO lesions in liver, with tissue-specific accumulation in older mice and a modest increase in mutation frequency (8, 9, 10) . However, they have no marked tumor predisposition according to two independent studies (9 , 10) and our more recent observations (D. Barnes and T. Lindahl, unpublished data), even after exposure to chronic oxidative stress (11) . A third group reported that OGG1 inactivation might be associated with lung tumorigenesis at ∼18 months, but the effect was dependent on a second uncharacterized gene locus. Moreover, this apparent increase in lung tumorigenesis in a small group of animals puzzlingly disappeared when the cellular load of 8-oxoG in DNA was increased (12) . It is unclear whether Ogg1 is simply not involved in tumor prevention or whether other gene deficiencies are required simultaneously for tumorigenesis. Inherited defects in the human MYH gene have been associated with somatic G to T mutations in the adenomatous polyposis coli (APC) gene and multiple colorectal tumors (13 , 14) . However, it is not yet certain whether the genetic background and environmental factors also contribute to the tumorigenesis and if Myh defects are associated with the formation of other tumors. Appropriate animal models are needed to clarify these questions and investigate the roles and mechanisms of oxidative DNA damage in tumorigenesis.
In addition to Myh and Ogg1, the mismatch repair gene products also prevent GO accumulation and its mutagenic effects in eukaryotes (15, 16, 17) . Recognition and repair of either GO or A or both in GO:A mismatches during DNA replication may be an alternative means of minimizing the mutations caused by GO lesions. Msh2-deficient mice are susceptible to lymphoma at an early age, but Msh2+/− mice show no tumor predisposition (18 , 19) . However, it is unknown whether mismatch repair genes act synergistically with oxidative DNA damage repair genes in tumor prevention.
To systematically address the role of the mammalian Myh, Ogg1, and Msh2 genes in oxidative DNA damage repair and tumorigenesis, we generated and studied Myh single, Myh, Ogg1 double, and Myh, Ogg1, Msh2 triple gene knockout mice.
MATERIALS AND METHODS
Generation of Myh Knockout Mice.
A genomic clone containing the entire coding sequence of murine Myh was isolated from a 129/Svj λ Fix II phage library (Stratagene) using a human MYH cDNA fragment as a probe. The exon-intron structure of the Myh gene was determined by sequencing the Myh genomic DNA clone and comparing it to the Myh cDNA sequence (7) . The targeting vector carried a neomycin (neo) cassette from the pMC1NEOpolyA vector (Stratagene), inserted into exon 6 of the Myh gene. The vector also carried a herpes simplex virus thymidine kinase cassette from the pMC1TKpolyA vector (20) , flanking the 3′-segment of homology (Fig. 1) ⇓ . The linearized targeting vector (NotI digestion) was electroporated into RW-4 129/Svj mouse ES cells (Genome Systems). G418/gancyclovir-resistant clones were screened by Southern blot hybridization using a 5′-flanking (5′-probe) and neo probe. ES cells carrying the disrupted allele were microinjected into C57BL/6J blastocysts to obtain chimeric mice. Chimeric males were bred with C57BL/6J mice and then intercrossed to produce mice homozygous for the Myh mutation. Genotyping was done by Southern blot hybridization with a 5′-probe or by PCR with a common forward primer (P1: 5′-CAAGTGCTGGGATCAAAGGTG-3′) and specific reverse primers against exon 6 (P2: 5′-GCTCCTTCTTGTAGCCGACG-3′) and Neo (P3: 5′-TCCTCGTGCTTTACGGTATCG-3′).
Generation of Myh−/−Ogg1−/− and Myh−/−Ogg1−/−Msh2−/− Mice.
Heterozygotes for Myh were crossed with Ogg1 knockouts (9) and siblings were intercrossed to generate Myh−/−Ogg1−/− mice in a mixed C57BL/6J and 129 background. The Myh−/−Ogg1−/− mice were bred with Msh2 heterozygotes (19) in a similar mixed background to produce Myh+/−Ogg1+/−Msh2+/− mice. Myh+/−Ogg1+/−Msh2+/− mice were backcrossed with Myh−/−Ogg1−/− to produce Myh−/−Ogg1−/−Msh2+/− mice. Myh−/−Ogg1−/−Msh2+/− mice were intercrossed to produce littermates: Myh−/−Ogg1−/−Msh2+/+ (referred to as Myh−/−Ogg1−/−); Myh−/−Ogg1−/−Msh2+/−; and Myh−/−Ogg1−/−Msh2−/−, our three study groups. We also generated three control groups: (a) Myh+/−Ogg1+/−Msh2+/− mice from crossing Myh−/−Ogg1−/− and Msh2+/−; (b) Msh2+/−; and (c) Msh2−/− mice from intercrossing the parental Msh2+/− mice. All viable 1-month-old mice were maintained under the same conditions according to Animal Research Committee regulations and observed for 18 months, after which, all surviving mice were sacrificed within a month. It was impossible to cross Myh−/− or Myh−/−Ogg1−/− mice on to the Big Blue strain (9) to assay spontaneous mutation frequencies in vivo due to linkage of the Myh locus and the integrated reporter transgene on murine chromosome 4.
Genotypes were determined at weaning and were also confirmed postmortem by specific PCR-based assays for Myh, Ogg1, and Msh2 using tail-snip DNA. Three-primer assays specific for Myh or Msh2 were carried out as described in the above section or as reported previously (19) . Two-primer assays specific for wild-type Ogg1, using primers Ogg1–2 (5′-GCCTTGTGGGCCTCTTCATA-3′ forward) and Ogg1–12 (5′-CACCTGAGGAAGTTGGGCC-3′ reverse), and for mutant Ogg1, using primers Ogg1–3 (5′-CAAGACCCCACTGAGTGCC-3′ forward) and N10 (5′-GAGAACCTGCGTGCAATCCA-3′ reverse), were carried out to yield 0.14 kb (Ogg1–2, Ogg1–12) and 1.1 kb (Ogg1–3, N10) fragments (data not shown). All PCR reactions were carried out in 25-μl buffers with 0.1–1 μg DNA, 0.2 pmol of each primer, 2.5 mm MgCl2, 0.1 mm each deoxynucleoside triphosphate, and 1 unit of TaqDNA polymerase (Invitrogen) at 35 cycles of 94°C for 30 s, 57°C for 30 s, and 72°C for 1.5 min.
All dead mice were subjected to pathological examination. Tissue specimens were fixed in 10% buffered formalin overnight and kept in 70% ethanol before embedding in paraffin. Histological analyses were carried out on 3-μm thick sections stained with H&E and examined by two pathologists independently.
Sequencing of Exon 1 of K-Ras and Exons 4–8 of p53.
Genomic DNA was isolated from tumors and normal tissues using phenol-chloroform extraction. Some small tumors and normal regions of lung identified on stained formalin-fixed thin sections were microdissected, and the DNA was extracted as described previously (21) . Exon 1 of K-ras and exons 4–8 of p53 were amplified with Platinum high-fidelity TaqDNA polymerase (Invitrogen). All primer sequences are available upon request. PCR products were cloned onto the pCR2.1-TOPO vector (Invitrogen) according to the manufacturer’s instructions. Both PCR products and subclones were sequenced at the University of California at Los Angeles sequencing facility. To identify mutations, the sequences obtained for K-ras exon 1 and the p53 exons 4–8 were aligned with the K-ras sequence accession No. S39586 in the National Center for Biotechnology Information GenBank and the p53 sequence from online. 5
Kaplan-Meier and Wilcoxon analyses were used for statistical analysis (22) .
Generation and Characterization of Myh Knockout Mice.
The mouse Myh gene contains 15 exons and has a structure similar to the human MYH, except that the sequence corresponding to the first exon of human MYH is apparently absent (Fig. 1A ⇓ and data not shown). Exon 6 includes a MutY family conserved pseudo-helix-hairpin-helix motif, and deletion of exon 6 leads to loss of solubility and catalytic function of the bacterially expressed recombinant protein (data not shown). A neomycin-resistance gene expression cassette (20) was inserted into exon 6 of the mouse Myh gene between codons 153 and 154 by targeted homologous recombination (Fig. 1, A and B) ⇓ . Correct gene targeting in ES cells and mice was verified by Southern blot hybridization and PCR (Fig. 1C) ⇓ . The disruption of the Myh gene was additionally confirmed by reverse transcription-PCR analysis using RNA isolated from mouse liver (data not shown).
Intercrossing Myh+/− mice resulted in Myh−/− mice following a normal Mendelian distribution (Myh+/+: Myh+/−: Myh−/− = 46:115:61). There was no significant difference in survival between wild-type and Myh knockouts after 14 months (93.5% in Myh+/+ and 91.8% in Myh−/− mice; observation terminated) and in tumor incidence based on pathological examinations of littermates sacrificed between 15 and 17 months (19.2%, 5 of 26 in Myh+/+ and Myh+/−; 18.8%, 3 of 16 in Myh−/− mice). Thus, Myh deficiency alone is not sufficient to predispose mice to tumors within 17 months, similar to Ogg1 deficiency.
Generation of Double and Triple Knockout Mice for Myh, Ogg1, and Msh2.
To investigate whether mammalian Ogg1 and Msh2 have any backup role for Myh in preventing mutations and tumorigenesis caused by GO, we generated Myh−/−Ogg1−/− double and Myh−/−Ogg1−/−Msh2−/− triple knockout mice. The ratio of mice obtained in the littermates Myh−/−Ogg1−/−, Myh−/−Ogg1−/−Msh2+/− and Myh−/−Ogg1−/−Msh2−/−, was 35:78:30, consistent with Mendelian segregation. The double and triple knockouts, as with the Myh single knockouts, were viable and fertile through at least two generations and appeared healthy with no obvious differences from wild-type into adulthood (2 months). This indicates that deficiencies in Myh and Ogg1, with or without Msh2, do not affect embryonic and neonatal mouse development.
Life Span and Tumor Predisposition in Myh and Ogg1 Double Knockouts.
The 50% survival age in Myh−/−Ogg1−/− mice was significantly reduced to 10.3 months, whereas the survival in Msh2+/− and Myh+/−Ogg1+/−Msh2+/− controls did not fall below 50% within 18 months (Fig. 2 ⇓ ; P < 0.0001). Myh−/−Ogg1−/− mice started to have tumor after 2 months, and the tumor incidence increased up to 65.7% ultimately, significantly higher than in Msh2+/− and Myh+/−Ogg1+/−Msh2+/− control mice, of which, 26.1 and 21.7%, respectively, developed tumors after 15 months (Fig. 3A ⇓ and Table 1 ⇓ ; P < 0.0001).
Myh−/−Ogg1−/− mice suffered from a range of different tumors. Notably, 31.4% of Myh−/−Ogg1−/− mice ultimately developed lung tumors after 12 months, whereas we only found one lung tumor case (4.3%) in Msh2+/− mice at 16.8 months and 4 cases (16.7%) in Myh+/−Ogg1+/−Msh2+/− mice after 18 months (P < 0.05; Fig. 3B ⇓ and Table 1 ⇓ ). More significantly, 78.6% (11 of 14) of older Myh−/−Ogg1−/− mice that lived >12 months developed lung tumors, in contrast to 5.3% in Msh2+/− mice and 18.2% in Myh+/−Ogg1+/−Msh2+/−. The lung tumors were solitary and white nodules with a diameter of 1–10 mm (Fig. 4A) ⇓ . Histologically, the lung tumors were diagnosed as adenoma or adenocarcinoma, and most of them (10 of 11) were adenomas (Fig. 4B) ⇓ .
A total of 21.7% of Myh−/−Ogg1−/− mice developed ovarian tumors after the age of 12 months, and 4 of 5 ovarian tumors were bilateral, but no ovarian tumors were found in control groups (P < 0.0001; Fig. 4A ⇓ ). Most ovarian tumors were hemangiomas, which can result in lethal hemorrhaging (Fig. 4D) ⇓ . The lymphoma incidence was 37.1% in Myh−/−Ogg1−/− mice, significantly higher than in either the Myh+/−Ogg1+/−Msh2+/− or Msh2+/− control groups (P < 0.05; Table 1 ⇓ ). Myh−/−Ogg1−/− mice had extensive lymphoma burdens in most organs such as the lung, heart, liver, spleen, kidneys, and lymph nodes, but few enlarged thymuses were involved (Fig. 4, A and E ⇓ , and data not shown). In contrast, lymphomas in Myh−/−Ogg1−/−Msh2−/− and Msh2−/− mice mostly resulted in enlarged thymuses. This might relate to late lymphoma onset and thymus dystrophy in older mice. There were 8.6% (3 of 35) gastrointestinal tract tumors in Myh−/−Ogg1−/− mice and one was a colorectal tumor, in contrast to none in control Msh2+/− and Myh+/−Ogg1+/−Msh2+/− mice (Fig. 4F ⇓ , Table 1 ⇓ , and data not shown). Thus, the oxidative DNA damage repair genes Myh and Ogg1 prevent a variety of tumors, more significantly lung and ovarian tumors, and lymphomas, especially after 12 months of age.
G to T Mutations in Codon 12 of the K-Ras Oncogene in Lung Tumors.
To investigate the molecular mechanism of tumor formation in Myh−/−Ogg1−/− mice, we examined codons 12 (GGT) and 13 (GGC) of the K-ras oncogene in lung tumors. Mutations in these codons are activating hot spots involved frequently in many human tumors, especially in lung adenocarcinomas (∼25–50%; Refs. 23 , 24 ). These GC rich codons could be possible targets for deleterious 8-oxoG formation in repair deficient backgrounds. We amplified and sequenced the K-ras oncogene exon 1, using DNA extracted from lung tumors and adjacent normal lung tissue under the same conditions. No mutations were found in exon 1 of K-ras in 11 tumor-adjacent normal lung tissue samples or 4 lung tissue samples from littermates without lung tumors (Fig. 5 ⇓ and Table 2 ⇓ ). However, G to T mutations at codon 12 were found in 75% of the DNA from lung tumors, and all were additionally confirmed by complementary strand and/or cloning sequencing. The mutations at either the first or second G in codon 12 (GGT) in multiple tumors from the same lung suggest that each tumor arose from an independent mutant clone. These G to T mutations correlate with the in vitro activities of mammalian OGG1 and MYH, which excise GO from GO:C pairs and A from GO:A mismatches, respectively (5, 6, 7) . The specific mutation spectrum is identical to that found in their E. coli counterpart, mutY- and mutM-deficient strain (3) .
In addition to K-ras, the tumor suppressor gene p53 is frequently mutated in human cancers, and 98% of base substitution mutations in p53 are distributed within codons 110–307, which encode four domains highly conserved from fish to mammals (25) . Six mutational hot spots in lung tumors from smokers also fall within this region (26) . To investigate whether p53 was also involved in spontaneous lung tumorigenesis in Myh−/−Ogg1−/− mice, we sequenced the p53 exons 4–8 that include the equivalent codons 110–307 in human, but no mutations were found in 10 lung tumors from Myh−/−Ogg1−/− mice (data not shown). In contrast to mutations at codon 12 of the K-ras oncogene, mutation in p53 exons 4–8 is not apparently a frequent event in spontaneous lung tumorigenesis in Myh−/−Ogg1−/− mice.
Synergistic Effects of Deficiencies in Msh2, Myh, and Ogg1 in Tumorigenesis.
Myh−/−Ogg1−/−Msh2+/− mice showed no significant difference from Myh−/−Ogg1−/− mice in life span, total tumor incidence and spectrum, or K-ras and p53 mutations in lung tumors (Figs. 2 ⇓ and 3 ⇓ , A and B, Tables 1 ⇓ and 2 ⇓ , and data not shown). However, 19 of 31(61.3%) of Myh−/−Ogg1−/−Msh2+/− mice bearing lung tumors had lung adenocarcinomas after 10.6 months, but only 1 of 11 Myh−/−Ogg1−/− mice with lung tumors had lung adenocarcinoma after 18 months (P < 0.0001; Figs. 3C ⇓ and 4 ⇓ A–C and Table 1 ⇓ ). Myh−/−Ogg1−/−Msh2+/− mice had the same type of ovarian tumors as Myh−/−Ogg1−/− mice, but the incidence was significantly higher (57.6 versus 21.7%; P < 0.05), and the earliest onset time was also 7 months earlier than in Myh−/−Ogg1−/− mice (5 versus 12 months). Thus, Msh2 heterozygosity did not affect total lung tumor incidence significantly but did accelerate malignant lung tumor and ovarian tumor formation in a Myh−/−Ogg1−/− background.
Msh2−/− combined with Myh−/−Ogg1−/− significantly reduced the 50% survival age to 4.3 months and all of the mice died within 10.3 months (Fig. 2) ⇓ . The tumor incidence increased steeply after 2 months, eventually totaling 86.7% with the majority being lymphomas (76.7%, Fig. 3A ⇓ and Table 1 ⇓ ). Myh−/−Ogg1−/−Msh2−/− and Msh2−/− mice showed no significant differences in life span and tumor development (Figs. 2 ⇓ and 3 ⇓ A and Table 1 ⇓ ), which were as in previous studies of Msh2−/− mice (18 , 19) . The short life span and the high mutagenic background contributed by Msh2−/− might mask any additional difference due to Myh and Ogg1 deficiencies.
Cancer has been proposed to be a consequence of multistep mutagenesis and defects in certain DNA repair genes may initiate or accelerate the process (18 , 27 , 28) . Therefore, it is important to identify genes involved in the process and define the pathways of tumorigenesis in detail. Here, we show that oxidative DNA damage repair genes Myh and Ogg1 are involved in tumor prevention, but the effects of a deficiency in one of them could apparently be compensated by the other in mice, suggesting a synergistic effect of Myh and Ogg1 in tumor prevention, especially in older mice. Our studies establish a causal link between deficiencies in base excision repair, DNA glycosylases correcting oxidative DNA damage, and tumorigenesis in mice. Furthermore, they may reveal a pathway of lung tumorigenesis through K-ras activation resulting from oxidative DNA damage in a Myh- and Ogg1-deficient background.
Oxidative DNA damage has been proposed as one mechanism that contributes to human lung tumor formation, especially in smokers (29, 30, 31, 32) . In support of this model, elevated GO content in peripheral leukocyte and lung tissue DNA was observed in lung cancer patients and smokers, suggesting a general increase of oxidative DNA damage (30, 31, 32) . Increased G to T mutations (∼30%) in p53 in lung tumors from smokers are also observed (33) . Here, we demonstrate that deficiencies in the oxidative DNA damage repair genes, Myh and Ogg1, result in lung tumor formation and K-ras activation through G to T mutations at codon12. Thus, oxidative DNA damage, arising from endogenous and environmental mutational load and/or GO repair gene deficiencies, can be a causative factor in lung tumorigenesis.
The frequent and unique G to T mutations at codon 12 of K-ras from lung tumors are striking in contrast to previous studies of K-ras mutations in spontaneous lung tumors in humans and mice that showed G to T mutations in only one-third of cases (23 , 34 , 35) . This specificity is identical to that found in E. coli deficient in mutM and mutY and correlates with the biochemical activities of mammalian OGG1 and MYH enzymes (3 , 5, 6, 7) . Additionally, expression of the human MYH cDNA in an E. coli mutY- mutant can suppress its mutator phenotype (6) . Thus, mammalian MYH and OGG1 may have similar in vivo functions in mutation avoidance as their bacteria counterparts, accounting for their involvement in tumor prevention. A previous study indicates that a mutation in codon 12 of K-ras is sufficient to cause lung tumorigenesis in mice (24) . So, codon 12 of K-ras appears to be an important downstream target of oxidative DNA damage resulting from Myh and Ogg1 deficiencies, transforming a normal lung cell to a tumor cell.
Msh2 heterozygosity increased malignant lung tumor incidence in a Myh−/−Ogg1−/− background. This may be because of the lowered Msh2 gene dosage and the increased mutation load caused by higher GO accumulation. In support of this possibility, Msh2 gene dosage affects GO accumulation levels in both mouse embryonic fibroblasts and embryonic stem cells (16 , 17) . Haploinsufficiency at tumor suppressor loci may result in a growth advantage and allow for the manifestation of neoplastic phenotypes after mutation of only a single allele (36) . The effects also possibly operate through secondary changes such as loss of heterozygosity or other gene mutations. The apparent lack of lung tumors in Myh−/−Ogg1−/−Msh2−/− mice, which occurred in Myh−/−Ogg1−/− mice after 12 months, suggests the importance of GO accumulation and age in lung tumorigenesis.
There have been very few spontaneous lung tumor mouse models available until now (37) and, in particular, no appropriate animal model for the study of oxidative DNA damage and lung tumorigenesis. These mouse models will facilitate tumorigenesis studies with regard to evaluating the contribution of oxidative stress and carcinogens and the efficacy of prevention and treatment strategies.
We thank Dr. Gregory W. Lawson for pathology diagnoses; Dr. Diana Shi for assistance in constructing the Myh knockout mice; Dr. Jiuyong Xie for helpful discussion and critical reading of the manuscript; and Jennifer H. Tai and Isabella T. Phan for technical assistance.
Grant support: National Cancer Institute Grant CA 85952 (J. Miller).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Requests for reprints: Jeffrey H. Miller. E-mail:
↵5 Internet address: http://www.ensembl.org/Mus_musculus/exonview?transcript=ENSMUSG00000005371&db=core.
- Received December 8, 2003.
- Revision received February 13, 2004.
- Accepted February 17, 2004.
- ©2004 American Association for Cancer Research.