Combination of DNA Repair Gene Single Nucleotide Polymorphisms and Increased Levels of DNA Adducts in a Population-based Study¹

Considerable interindividual variation in DNA repair capacity has been observed in the general population (1). Only a fraction of carcinogen-exposed individuals develop cancer, suggesting an important role of individual susceptibility and possible gene–environment interactions (2). Exposures to xenobiotics can cause cell cycle delays that allow cells the repair of DNA damage to maintain normal cellular functions (3, 4). Inherited polymorphisms of DNA repair genes may contribute to variations in DNA repair capacity and susceptibility to cancer (1, 5, 6, 7, 8, 9). In the present study, we have measured leukocyte DNA-adduct formation in a large healthy population (n = 628) belonging to the EPIC³ (3) multicenter investigation (10, 11), and we have analyzed three DNA repair gene polymorphisms, representing different repair pathways: (a) XRCC1-Arg399Gln; (b) XPD-Lys751Gln; and (c) XRCC3-Thr241Met. A rapidly growing literature describing their association with cancer risk or DNA adduct/damage has been published (12, 13, 14, 15). Our group recently conducted two preliminary studies whose consistent results suggested that further investigation was worthwhile. In the first study (16), we found an association of XRCC3–241 variant with bladder cancer and a higher risk of having “bulky” DNA-adducts above the median value. Second, “bulky” DNA adducts have been measured in the context of the EPIC–Italy project, where we investigated the same three polymorphisms as above in ∼300 individuals (10); this investigation further supported the association of polymorphic variants with DNA-adduct levels. Thus, we extended the analyses to the 628 individuals of the present study to confirm the previous results.

The individuals analyzed belong to the Italian section of the large EPIC project (17), based on ∼48,000 volunteers of both sexes (age 35–64 years) enrolled from 1993 to 1998 in five Italian centers: (a) Varese; (b) Turin; (c) Florence; (d) Ragusa; and (e) Naples. A random sample of 628 subjects (313 men), stratified by age, sex, and area of residence, was selected from the different centers.

Adducts analyses were carried out in WBCs using the nuclease P1 ³²P-postlabelling technique, as reported (11). Measurements were expressed as RAL × 10⁹ nucleotides calculated as follows:

RAL = (cpm in adduct(s)/cpm in total nucleotides) × (1/dilution factor).

PCR followed by enzymatic digestion or primer extension/denaturing high-performance liquid chromatography analysis was used for the genotyping of the XRCC1-Arg399Gln, XPD-Lys751Gln, and XRCC3-Thr241Met polymorphisms, as described (10).

The significance of the differences among genotypes for ³²P-postlabeling DNA-adduct levels was estimated by Kruskal-Wallis nonparametric and ANOVA F tests; the combination of more than one variant allele was investigated by using the test for trend. A test for interaction between smoking history and the different genotypes has been performed by logistic regression method (DNA-adduct levels above/below the median value as dependent variable) for each of the polymorphisms. Logistic regression analysis was also carried out to calculate ORs adjusted for different covariates (age, sex, centers of origin, and smoking status) categorizing DNA-adduct levels into tertiles versus undetectable measurements and grouping individuals according to the number of at-risk alleles. All of the analyses were performed by the statistical package SPSS (version 11.0).

Genotype and allele frequencies were calculated by direct counting, and genotype distributions were in Hardy-Weinberg equilibrium. Genotype frequencies were, respectively: (a) XRCC1–399 Arg/Arg (GG) = 44.2%, Arg/Gln (AG) = 44.4%, Gln/Gln (AA) = 11.4%; (b) XPD-751 Lys/Lys (AA) = 34.3%, Lys/Gln (AC) = 50.7%, Gln/Gln (CC) = 15%; and (c) XRCC3–241 Thr/Thr (CC) = 33%, Thr/Met (CT) = 49.6%, Met/Met (TT) = 17.4%, not different from those reported previously (10, 18, 19, 20, 21, 22). Allele frequencies were: (a) XRCC1–399Arg/Gln = 0.66/0.34; (b) XRCC3–241Thr/Met = 0.58/0.42; and (c) XPD-751Lys/Gln = 0.6/0.4.

Concerning DNA-adduct levels, as reported in the literature (23), the interindividual variability of ³²P-postlabelling DNA-adduct measurements is high. In our sample of 628 individuals, the range of RAL is 0.1–95.1, and the mean ± SD is 7.82 ± 10.06, coefficient of variation = 128%. When we consider only detectable measurements (474 individuals), the range of RAL is 0.15–95.1, and the mean ± SD is 10.33 ± 10.42, coefficient of variation = 101%.

No difference exists in our sample between DNA-adduct levels among current smokers (7.82 ± 0.76), ex-smokers (7.96 ± 0.74), and never-smokers (7.71 ± 0.62). In univariate analysis, a significant difference (ANOVA test, P = 0.025) for the crude arithmetic mean (±SE) of adduct levels in the overall sample was found only for the XRCC1–399 polymorphism (GG = 6.98 ± 0.51, AG = 7.99 ± 0.6, and AA = 10.60 ± 1.75). The association with XRCC1 polymorphism was particularly strong (P = 0.009) in nonsmokers, whereas there was no significant difference among ex- and current smokers (P = 0.58 and 0.47, respectively). A borderline significant difference (P = 0.052) was observed by using Kruskal-Wallis nonparametric analysis for XRCC3–241 polymorphism in the overall sample (CC = 6.93 ± 0.68, CT = 8.03 ± 0.59, and TT = 9.02 ± 0.92) and for XPD-751 (P = 0.073) in the nonsmoker group (AA = 7.7 ± 1.31, AC = 7.21 ± 0.73, and CC = 9.52 ± 1.35), whereas the Ps were nonsignificant for the remaining comparisons.

A test for interaction between smoking history and the different genotypes has been performed, by logistic regression, for each of the polymorphisms, using DNA-adduct levels above/below the median value as dependent variable. No evidence of interaction was found (P < 0.05 for all of the comparisons).

The combination of more than one variant allele was investigated, under the assumption that the combination of polymorphisms can have additive or more than additive effects. A significant difference in mean RAL (test for trend, P = 0.0046 for geometric means) was observed considering the combination of variant alleles for XPD-751, XRCC1–399, and XRCC3–241 polymorphisms (Table 1 and Fig. 1), with a dose-response relationship. The trend was maintained after stratification by smoking habits (data not shown). No individual bearing six variant alleles was found.

We performed multiple regression analysis by dividing adduct levels into tertiles versus undetectable measurements and grouping individuals according to the number of at-risk variant alleles (Table 2). Significant associations were observed among individuals who had at least three variant alleles, when comparing the highest tertile with undetectable adduct levels (three alleles, adjusted OR = 5.07, 95% CI = 1.29–19.9; 4 alleles, adjusted OR = 5.03, 95% CI = 1.18–21.45; 5 alleles, adjusted OR = 7.65, 95% CI = 0.94–62.2). However, the risk gradient for the intermediate tertiles is not straightforward.

In the present study, DNA-adduct formation was analyzed using the nuclease P1 assay, a technique effective at detecting bulky adducts, such as those formed by PAHs and some aromatic amines (23, 24). Our findings confirm the possible involvement of XRCC1-Arg399Gln, XPD-Lys751Gln, and XRCC3-Thr241Met polymorphisms in the repair of bulky DNA adducts (10, 16, 25, 26). Smoking did not seem to interact significantly with the different genotypes, although adducts were more strongly associated with the XRCC1 polymorphism in nonsmokers. No overall effect of smoking over the level of DNA adducts was apparent.

The XRCC1 protein participates in the BER pathway (27), acting as a scaffold protein by binding DNA ligase III at its COOH and DNA polymerase β at its NH₂ terminus. XRCC3 participates in DNA double-strand break/recombinational repair and is a member of a family of Rad-51-related proteins that likely participate in homologous recombination to maintain chromosome stability and repair DNA damage (28). XPD is involved in the NER pathway, which recognizes and repairs many structurally unrelated lesions, such as bulky adducts and thymidine dimers (29). XPD functions as an ATP-dependent 5′-3′ helicase joint to the basal transcription factor IIH complex.

It is known that the main pathway by which mammalian cells remove bulky DNA adducts is the NER (3, 4). However, PAH-DNA adducts have been also shown to be repaired by BER mechanisms (30, 31), supporting the possible involvement of the XRCC1 gene. The involvement of BER in repair of PAH-DNA adducts can be explained with the fact that PAHs can be also metabolized via radical cation intermediates to electrophiles that bind to DNA and destabilize the N-glycosyl bond, inducing rapid depurination or depyrimidation of adducted bases (32). Using a human cell-free system, Braithwaite et al. (30) have examined repair of DNA lesions induced by several PAH dihydrodiol epoxides, and two distinct mechanisms of excision repair were observed. The major repair mechanism, as expected, was NER. The other mechanism is independent of NER and correlated with the presence of apurinic/apyrimidinic sites in the damaged DNA, thus presumably reflecting BER. In experimental animal systems, these unstable PAH-DNA adducts have been shown to induce mutations in the H-ras oncogene (33), supporting the importance of this kind of DNA adducts and of their related mechanisms of DNA repair in carcinogenesis processes. In addition, Hang et al. (31) reported that BER may play an important role in repair of certain bulky-induced DNA lesions, observing that a newly synthesized DNA adduct (1,N6-benzetheno-dA) in defined oligonucleotides was a substrate for the major human apurinic/apyrimidinic endonuclease, HAP1, and the Escherichia coli apurinic/apyrimidinic endonucleases, exonuclease III and endonuclease IV.

Cross-links between bases, induced from some aromatic amines and oxidative agents and detectable by ³²P-postlabeling techniques (34), could be partially responsible for the association between DNA adducts and the XRCC3 gene (35). The in vitro experiments by Araujo et al. (36) have recently suggested that the increased cancer risk associated with the XRCC3–241 variant (16, 18, 37, 38) may not be attributable to an intrinsic HDR. However, such experiments cannot definitely rule out the involvement of other XRCC3 variants in linkage disequilibrium, possible genetic interactions between the XRCC3–241 variant and polymorphic alleles of other DNA repair genes that may lead to an HDR defect, nor even an extremely mild HDR defect that would not be detectable in their assay. It is still possible that XRCC3 participates in other cellular pathways not assayed within their model.

This study suggests that, at the individual level, the combined effect of multiple gene variants may be more important than the investigation of single nucleotide polymorphism to define DNA repair capacity. Our work is in progress to analyze additional polymorphisms in these three genes to investigate possible additive or synergistic effects. Because DNA damage phenotypes seem to vary considerably in the general population, our findings may be relevant for risk assessment.

We thank E. Nieddu and M. Manzari for technical assistance, the cooperation of all study participants, and collaborators of EPIC–Italy Study Group.