Every year, approximately 1 million cases of CRC are diagnosed worldwide. Those with a familial component could constitute 10% to 50% of all CRC cases (4
), but the fraction of cases caused by highly penetrant familial cancer syndromes such as Lynch syndrome or familial adenomatous polyposis is relatively small. Indeed, the genetic basis for most cases of familial susceptibility is unknown (32
). It is likely that a combination of low and moderately penetrant genes interact with environmental factors to modify risk for CRC (33
). Approximately 5% of patients with UC develop colon cancer, and no susceptibility genes for inflammation associated CRC have been identified (2
). Being able to identify those in the population who are at increased risk of CRC could reduce the number of CRC-related deaths.
We report here that Aag deficiency in a mouse model grossly increases the extent of pathology in the intestinal mucosa in response to chronic inflammation. This deficiency enhances inflammation-associated colon tumorigenesis in 2 carcinogenesis protocols, treatment with AOM+DSS and treatment with DSS alone. As a measure of the generality of this effect, following infection with H. pylori
, Aag deficiency predisposes to the development of gastric cancer precursor lesions. H. pylori
chronic gastritis is associated with increased production of RONS in the host (reviewed in ref. 34
). This pathogen is an enormous public health concern, with over 50% of the world population being infected and 5% of all human cancers attributed to H. pylori
infection and its associated inflammation (35
). Finally, Aag had no effect on the development of AOM-induced ACF or tumor development and had no effect on spontaneous tumorigenesis in mice on the ApcMin
background, indicating that Aag suppresses chronic inflammation-related cancer development.
The number of ε-base lesions and 8oxoG increased in the colons of mice after a single DSS cycle, consistent with previous reports that inflammation is associated with increased ε-base lesions (10
), that ε-base lesions increase more than oxidized or deaminated lesions in response to inflammation (31
), and that 8oxoG is induced by DSS treatment in rats (12
). Moreover, inflammation-induced εA, εC, and 8oxoG increased more dramatically in the Aag-deficient animals, consistent with the ability of Aag to excise both εA and 8oxoG. Interestingly, although Aag binds tightly to εC, it does so without catalyzing its excision (38
); it seems likely that the binding of Aag to εC in vivo facilitates its repair via an alternative repair pathway. In summary, the accumulation of DNA base damage in Aag–/–
mice in response to DSS provides a causal link between repair deficiency and carcinogenesis. Our data demonstrate that DNA repair under chronic inflammatory conditions is an important suppressor of cancer development.
Chronic inflammation produces a complex milieu of growth factors and cytokines that cause general hyperplasia, and we propose that the proliferative and antiapoptotic signals generated during chronic inflammation act synergistically with DNA damage to enhance cancer development (6
). Our results are consistent with 2 different but not mutually exclusive effects of RONS-dependent DNA damage that could lead to increased tumor development in a repair-deficient situation (Figure ). One effect involves a snowballing mechanism of increased levels of cellular damage and death leading to more inflammation, which in turn produces more tumor-promoting cytokines and more RONS (Figure ). Consistent with this model is our observation that Aag–/–
mice had much more severe gross pathological and general histopathological changes than Aag+/+
with DSS treatment. The differences between Aag–/–
mice were evident at 5 cycles but were much more dramatic at 7 cycles. Importantly, after 7 cycles of DSS Aag–/–
mice had significantly more general inflammation than Aag+/+
Model for inflammation-induced tumorigenesis.
Another effect that is likely acting in concert with that described above is that base damage in the presence of proliferative and antiapoptotic signals could serve to fix and expand cancer-causing mutations (Figure ). Of the base lesions found at higher levels with DSS treatment and Aag deficiency, εA induces A:T to G:C and A:T to T:A mutations, εC induces C:G to A:T and C:G to T:A mutations, and 8oxoG induces G:C to T:A transversions (39
). In addition, both εC and εA are replication-blocking lesions (38
). It has been reported that DSS promotion of AOM- or dimethylhydrazine-initiated tumors (AOM is a proximal carcinogen of dimethylhydrazine) results in tumors with predominantly G:C to A:T transitions in Ctnnb1
, consistent with the mutational specificity of these carcinogens (43
). However, Greten et al. (7
) showed that only 3 of 7 tumors from AOM+DSS-treated mice had G:C to A:T transitions, while the rest were mutations at A:T base pairs. In the present study, tumors from AOM+DSS-treated mice harbored predominantly G:C to A:T transitions, and the spectrum was not affected by Aag deficiency. In contrast, the pattern was significantly different in Aag–/–
mice treated only with DSS, in which nearly half the mutations in Ctnnb1
were deletions. Such alterations could arise from replication-blocking ε-base lesions.
Unlike DSS-induced colon tumors in mice, human UC-associated cancers commonly harbor TP53
). Circumstantial evidence that these mutations are caused by RONS lies in the fact that RONS enhance deamination at 5MeC, and most of the TP53
mutations in UC-associated cancers were G:C to A:T transitions at 2 hot-spot CpG dinucleotide sites. Interestingly, these mutations are found in both neoplastic and non-neoplastic tissue in the UC colon (45
A panoply of DNA glycosylases have evolved to repair oxidized bases in DNA. Among them, OGG1, which removes 8oxoG, and MYH, which removes adenine paired with 8oxoG, are important in suppressing G:C to T:A transversions and in preventing CRC in people and mice (reviewed in ref. 41
). These glycosylases suppress CRC in the absence of chronic inflammation, and their association with inflammation-associated cancer is not known. One might expect Ogg1- or Myh-deficient mice to display marked increases in susceptibility to cancer development that is stimulated by inflammation.
In light of our results, AAG
and other BER genes are strong candidates for genetic association studies of human CRC or gastric cancer risk. Although there are no known polymorphisms that alter AAG function, a 20-fold range in activity was observed in peripheral mononuclear cells from over 50 healthy adults (E. Moy, D. Christiani, and L.D. Samson, unpublished observations). Indeed, decreased 8oxoG DNA glycosylase activity in human blood samples has been associated with increased lung cancer risk, which may also have an inflammatory component (47
). Measurements of interindividual differences in the activity of AAG, OGG, and other DNA glycosylases may provide an informative parameter in the multifactorial dissection of cancer risk. This may be particularly important for gene-environment interactions with states of chronic inflammation or with other conditions known to increase oxidative stress such as metal storage diseases, heavy metal exposure, smoking, and chronic infection.