DNA adducts as markers of exposure to carcinogens and risk of cancer

Abstract

The usual categorization of biomarkers refers to the 3 categories of exposure (dose), individual susceptibility and early response. However, such categories are blurred and the interplay among them is likely to be crucial for future epidemiological studies. Both protein adducts and DNA adducts are markers of biologically effective dose, but their biological meaning is different. While protein adducts are not repaired, i.e., they reflect external exposure more faithfully, DNA adducts are influenced by the individual repair ability; in fact, if they are not eliminated by the DNA repair machinery, they will induce a mutation. Also, markers of early response are a heterogeneous category, which encompasses DNA mutations and gross chromosomal damage. The main advantage of early response markers is that they are more frequent than cancer itself and can be recognized earlier, thus allowing researchers to identify earlier effects of potentially carcinogenic exposures. Finally, markers of susceptibility include several subcategories, in particular a type of genetic susceptibility that is related to the metabolism of carcinogenic substances and another type that is related to DNA repair. There is increasing evidence that some types of DNA adduct can be considered both as markers of biologically effective dose and as markers of risk (including susceptibility) because (i) they predict the onset of disease independently of exposure levels and (ii) they express an integrated marker of exposure and of the individual ability to metabolize carcinogens and repair DNA damage. Classical markers of susceptibility, in turn, such as genetically based metabolic polymorphisms, can modulate the effect of exposures; but their expression can also be influenced by external determinants, such as dietary habits. A few studies have considered the association between cancer at different sites and the levels of “bulky” DNA adducts, including polycyclic aromatic hydrocarbon (PAH)–DNA adducts (Table I). Most studies (Vulimiri et al., 2000; Peluso et al., 2000; Li et al., 1996; Tang et al., 1995) have shown that cancer cases have higher levels of adducts than non-cancer controls, after adjustment for relevant exposures such as smoking. For example, WBC–DNA adducts were measured by the 32P-post-labeling method in 162 cases with bladder cancer and 104 controls (Peluso et al., 2000). The level of DNA adducts was strongly associated with the case/control status: the age-adjusted odds ratio (OR) for a level of adducts greater than the median was 3.7 [95% confidence interval (CI) 2.2–6.3], and a dose–response relationship with quartiles of adducts was apparent. (OR measures the extent of the risk increase in exposed compared to unexposed subjects. In Table I, the ORs compare different levels of adducts: detectable vs. non-detectable or above vs. below the median level.) In a logistic regression model, the adduct level was the variable most predictive of cancer risk, independently of smoking habits (OR = 5.25, Table I). Other studies (Vulimiri et al., 2000; Li et al., 1996; Tang et al., 1995) have found a similar ability of adduct levels to predict the risk of cancer, after adjustment for risk factors. The study by Li et al. (1996) differs from the others in that adducts were induced in vitro by BPDE treatment of lymphocytes from cases and controls; i.e., it is an experimental study. The other investigations were based on adducts pre-existing in the study subjects. One study did not find any association between lung cancer and adduct levels (Hou et al., 1999). Authors and method Type of cancer (number) Controls (number) OR 95% CI Peluso et al., (2000), 32P-post-labeling1 Bladder (162) Hospital (104) 5.25 (2.2–12.4) Quartiles of DNA adducts2 <0.1 22 45 1.0 — 0.1–0.23 42 29 3.0 (1.45–6.1) 0.23–0.51 44 19 5.4 (2.5–11.7) >0.51 54 11 7.9 (3.4–18.4) Vulimiri et al., (2000), 32P-post-labeling1 Lung (43) Hospital (47) Mean adducts (per 108 nucleotides): cases 6.03 (1.16 SE) controls 2.80 (0.36 SE)3 Hou et al., (1999), 32P-post-labeling1 Lung (179) Population, matched by smoking (161) 4.1 × 10−8 in both cases and controls Li et al., (1996), 32P-post-labeling1 Lung (21) Healthy (41) 6.4 (1.3–29.4)4 Tang et al., (1995), ELISA Lung (119) Hospital (98) 7.7 (1.7–34)5 These results are limited by the fact that in case-control studies the biomarker may reflect the disease rather than the etiology. Nevertheless, on the basis of the preceding examples, we may consider bulky DNA adducts (as measured by the 32P-post-labeling method or ELISA in WBCs) as candidate markers of risk (including susceptibility) in addition to being markers of exposure. Further evidence for this interpretation comes from a case-control study nested in the prospective Physicians Health Study cohort, in which DNA adducts measured by post-labeling in WBCs from current smokers were predictive of cancer outcome (Tang et al., 2000). The importance of the latter study rests on the measurement of adducts in blood samples that were collected years before cancer onset, thus ruling out the possibility that the higher adduct levels were due to the metabolic changes associated with an already existing cancer. Most likely, bulky DNA adducts express cumulative exposure to PAH and other aromatic compounds after the action of metabolizing enzymes and despite the intervention of DNA repair enzymes. They are, therefore, markers of cumulative unrepaired DNA damage. There is a large amount of additional evidence to support this interpretation, based on the observation that the lymphocytes of cancer patients (and of their healthy relatives) have higher levels of DNA adducts when treated with genotoxic chemicals compared to lymphocytes of non-cancerous individuals (Berwick and Vineis, 2000). Other types of adduct are clearly a better expression of external exposure than of cumulative unrepaired DNA damage. This is the case, in particular, with protein markers that are not repaired. Some DNA adducts reflect exposure to specific carcinogens, e.g., aflatoxin–DNA adducts, and express the cumulative, unrepaired and biologically effective dose of those specific carcinogens rather than of a broad category such as bulky aromatic adducts. To make things even more complicated, the level of DNA adducts, which is suggested to predict the risk of cancer, can be modulated by personal habits such as the intake of fruit and vegetables. The relationship of fruit and vegetable consumption to DNA-adduct formation was examined in a case-control study on bladder cancer (Peluso et al., 2000). The level of WBC–DNA adducts decreased with increasing levels of fruit and vegetable consumption; in addition, the association between the case/control status and the level of adducts (below or above the median value) was stronger in subjects who consumed fewer than 2 portions of vegetables per day (OR = 7.80, 95% CI 3.0–20.3) than in heavy consumers (OR = 4.98 for consumers of 2 portions/day; OR = 2.0 for consumers of 3 or more portions/day). Also, in a study of healthy volunteers, conducted in the context of the EPIC Italian cohort, an inverse association between several dietary items and adduct levels was found (Palli et al., 2000). Interestingly, of about 120 food items investigated, only fruit and vegetables were associated in a statistically significant manner with adduct levels. Some items, namely, leafy vegetables and fruit, were associated with an approximately 25% to 30% decrease in adduct levels, while, e.g., meat consumption was associated with a slight and non-significant increase (Palli et al., 2000). When nutrients were assessed by means of food-nutrient conversion tables, associations were found with mono-unsaturated fatty acids and β-carotene. However, 300 samples were too few to completely adjust each association for the role played by the others; the high level of collinearity among nutrients makes the attribution of the effect on DNA-adduct levels to specific nutrients rather arduous. Another series of studies investigated the relationship between plasma levels of micronutrients including anti-oxidants, genetic polymorphisms and carcinogen–DNA adducts in WBCs. In a cross-sectional analysis of 159 heavy smokers from a cohort of subjects enrolled in a smoking cessation program, smokers with the CYP1A1 exon 7 valine polymorphism had significantly higher (2-fold, p = 0.03) levels of PAH-DNA damage than those without the genotype, while PAH-DNA adducts were inversely associated with plasma levels of retinol (β = –0.93, p = 0.01), β-carotene (β = –0.18, p = 0.09) and α-tocopherol (β = –0.28, p = 0.21). The association between smoking-adjusted plasma β-carotene levels and DNA damage was significant only in subjects lacking the GSTM1 detoxification gene (β = –0.30, p = 0.05, n = 75). There was a statistical interaction between β-carotene and α-tocopherol: when β-carotene was low, α-tocopherol had a significant protective effect (β = –0.78, p = 0.04) on adducts. These results suggest that several micronutrients may act in concert to protect against DNA damage and highlight the importance of assessing overall anti-oxidant status (Mooney et al., 1997). In serial samples from a subset of 40 subjects who were able to quit smoking, a strong inverse relationship between adducts and vitamins (retinol, α-tocopherol and zeaxanthin) was observed but only in subjects with the GSTM1 null genotype (Mooney et al., 2000). Since there was no effect of GSTM1 genotype alone, these results suggest that some individuals may be at increased risk of DNA damage due to a combination of low plasma anti-oxidant/micronutrient levels and susceptible genotypes. The use of biological markers to assess the efficacy of interventions and to study mechanisms of micronutrients is timely given the current debate regarding the use of chemopreventive agents in high-risk populations. Also, DNA damage can be modified by fruit or vegetable intake. For example, several functional tests have been developed to explore individual DNA repair capacity (reviewed by Berwick and Vineis, 2000). Classically, lymphocytes from cancer patients and controls are treated with a clastogen and the number of chromosome breaks is counted (mutagen sensitivity test). A greater proportion of breaks is usually found in cases and interpreted as an expression of a more limited repair capacity. However, the repair capacity thus explored appears to be modified by several exposures or personal habits. In cultured lymphocytes, anti-oxidants such as α-tocopherol exhibited a dose-dependent protective effect in preventing bleomycin-induced chromosomal damage (Trizna et al., 1992, 1993). In a study of 25 healthy individuals, Kucuk et al. (1995) found strong inverse correlations between plasma nutrients and the mutagen sensitivity assay. Correlations were –0.76 with β-carotene and –0.72 with total carotenoids (monthly mean levels). In contrast, a positive correlation was found with triglyceride levels (r = 0.60). However, in a randomized double-blind trial of α-tocopherol, Hu et al. (1996) did not find any association between supplementation and DNA repair activity measured by ADPRT or UDS (other tests assessing DNA repair capacity). Similarly, Goodman et al. (1998) found no effect of either α-tocopherol or β-carotene on mutagen sensitivity values in a crossover design. One problem, pointed out by the authors, is that the 3-day culture of cells required for the mutagen sensitivity assay is likely to dilute the circulating anti-oxidants in the plasma. What is the biological interpretation of the modulating effect of fruit and vegetables on the adduct levels or mutagen sensitivity (with the limits of the latter type of marker)? A rather obvious explanation for the first association is the induction of enzymes involved in carcinogen detoxification or the repression of enzymes involved in activation (Vineis et al., 1999). Concerning “mutagen sensitivity”, this end point is usually interpreted as an indirect expression of DNA repair capacity. Although it has been suggested that constituents of fruit and vegetables, or vitamins, can interfere with DNA repair enzymes, there is no evidence in favor of this hypothesis. Wienke et al. (1999), in a study on DNA adducts measured by 32P-post-labeling, found statistically significant interactions between adduct levels and smoking variables, indicating that the impact of smoking on DNA adduct levels may be different in current and former smokers. In current smokers, recent smoking intensity (cigarettes smoked/day) was the most important variable. In former smokers, age at smoking initiation was inversely associated with DNA-adduct levels. According to the authors, the results in former smokers suggest that smoking during adolescence might produce physiological changes that lead to increased DNA-adduct persistence. Alternatively, young smokers may be markedly susceptible to DNA-adduct formation and have higher adduct burdens after they quit smoking than those who started smoking later in life. Similar observations were made by Vulimiri et al. (2000) They found that cases with lung cancer who started smoking at the earliest age had the highest levels of aromatic DNA adducts and of 8-oxo-dG. These observations should be interpreted in light of the epidemiology of smoking and lung cancer and of the reported strong effect of age at start of smoking (see below). From the preceding, one can argue that adducts, particularly DNA adducts, can be used as markers of exposure and susceptibility to cancer, modified by determinants such as diet or genetic traits (metabolic polymorphisms, DNA repair capacity). One concept that can prove useful is that of “cumulative unrepaired DNA damage”. Epidemiological studies have shown that in many instances duration of exposure is more important than daily dose. The paradigm for this general relationship is represented by smoking and lung cancer, but experimental evidence has also been produced. In the case of smoking, the incidence of lung cancer increased with the fourth power of duration in one study (Doll and Peto, 1978). Table II shows estimates based on a study of doctors in the United Kingdom. For a duration of smoking 3 times longer (45 vs. 15 years), the excess lung cancer incidence (%) increases by 100 times (from 0.005 to 0.5 in light smokers, from 0.01 to 1 in heavy smokers), i.e., by the fourth power of time. Other investigations did not find a strong discrepancy between dose and duration, but the latter was nevertheless more relevant. Duration is mainly due to age at start; classical studies have shown a very strong association between earlier age at start and the risk of lung cancer (International Agency for Research on Cancer, 1986). In laboratory animals, fractionated and repeated doses induced tumors more frequently than the same total amount administered as a single dose. The latter observation is at odds with the general mechanisms of toxicity, according to which heavy exposure in a single administration has more devastating effects than repeated small doses. In light of such observations, a possible interpretation of the higher levels of bulky adducts among cancer cases compared to controls is the concept of “cumulative unrepaired DNA damage”. What causes cancer would be the total burden of a genotoxic chemical that binds to DNA, thus overcoming the repair processes. Such a burden may be higher (i) because higher levels of carcinogenic metabolites are present (for genetic or acquired reasons), (ii) because DNA repair is impaired or (iii) because repeated exposures to the same agent occur. We also must consider the limitations of this model. In particular, the level of measurement error for bulky adducts is not well known but appears to be high (coefficient of variation around 20% to 30%). However, the effect of measurement error is to attenuate a relationship, if error is evenly distributed in the comparison groups (Copeland et al., 1977). So, measurement error is expected to blur existing associations rather than to reveal false associations. Although adducts have been interpreted mainly as biomarkers of exposure, bulky DNA adducts, such as those measured by 32P-post-labeling or ELISA in WBCs, are more correctly interpreted as markers of “cumulative unrepaired DNA damage”. This concept can prove useful in cancer epidemiology since it is consistent with long-lasting knowledge on the importance of duration of exposure in the etiology of chemically induced cancers. Increasing evidence suggests that in addition to prolonged exposure to genotoxic chemicals, inter-individual variability in carcinogen metabolism and DNA repair predicts the risk of cancer. Also from this point of view, bulky DNA adducts can be a useful biomarker for population studies since they express the amount of carcinogen linked to DNA after repair and incorporate the individual ability in repair.