Genetic Polymorphism of CYP Genes, Alone or in Combination, as a Risk Modifier of Tobacco-related Cancers¹

Epidemiological and human genetic studies have identified different types of population “at risk,” one consisting of individuals with heavy exposure to carcinogens, such as smokers and exposed workers, and the other consisting of carriers of cancer-predetermining germ-line mutations in genes that because of high penetrance confer a very high risk for cancer per se(1). There is also another group of predisposing polymorphic, low-penetrance genes, i.e., those involved in carcinogen metabolism and DNA repair, which modestly increase the risk for cancer in exposed individuals, perhaps at low doses of carcinogens (2, 3). In the latter case, the proportion of cancers attributable to such genetic traits may be high, because the frequency of “at risk” alleles in the population is high.

Drug-metabolizing enzymes, which often display genetic polymorphism, convert many tobacco carcinogens into DNA-binding metabolites in target cells and can thereby modulate intermediate effect markers such as DNA adducts and ultimately, the risk for cancer. The development of simple assays based on the PCR has allowed identification of individual genotypes for a variety of metabolic polymorphisms and studies on the modulation of cancer risk by environmental exposures, such as tobacco smoke, which are the subject of this review. Given the great number of carcinogen-activating and-detoxifying enzymes, the variation in their expression, and the complexity of exposures to tobacco carcinogens, assessment of a single polymorphic enzyme or genotype may not be sufficient to assess their role in carcinogenesis (reviewed in Ref. 4). Tobacco smoking is the major cause of lung cancer and is associated with risks for cancers of the larynx, mouth, esophagus, urinary bladder, and kidney (5). Breast cancer in women is at best weakly associated with cigarette smoking. We have included this site because recent studies suggested that postmenopausal women who are carriers of the CYP1A1 or NAT2 variant alleles may be at increased risk for breast cancer in a smoking-dose-related manner (6, 7).

Because major classes of tobacco carcinogens are converted to DNA-reactive metabolites by the oxidative, mainly CYP⁴-related enzymes, we have summarized studies of the effect of genetic polymorphism of CYPs in humans, alone or in combination with phase II enzymes, as risk modifiers of some major tobacco-related cancers. Our analysis includes case-control studies published from 1990 to May 1999 on cancers of the upper aerodigestive tract, urinary tract,and breast. To be included, a study had to have been published in a full article in English, cited in MedLine, and have involved a case-control design and adequate methods for analysis of CYPgenotype. When several overlapping reports on one study population were available, we included the most recent publication, which usually covered a larger number of study subjects. To obtain insights into mechanisms, we also briefly reviewed studies of gene-gene interactions and the dependence of the formation of smoking-related PAH-DNA adducts on genotype. Because of space limitations, the references cited are not exhaustive, and the reader is referred to review articles marked in the text.

These subjects have been reviewed by Hoffmann and Hecht (8, 9), McClellan (10), and Bergen and Caporaso (11). Processed tobacco contains over 3000 compounds including 30 carcinogens. The mainstream and sidestream smoke generated when tobacco in cigarettes is burnt contains more than 4000 constituents including about 50 carcinogens. The diversity of carcinogenic and toxic compounds in tobacco smoke leads to ambiguity about which are the most important; however, studies on the mechanisms of tobacco carcinogenesis and dosimetry in smokers and tobacco chewers indicate that three major classes of carcinogens, PAHs, TSNAs, and aromatic amines, play important roles in tobacco-associated cancers.

The mechanism by which PAHs such as B[a]P interact with DNA, activate oncogenes, and initiate the carcinogenic process involves the formation of bay-region diolepoxides as the major ultimate carcinogens. B[a]P is converted into phenolic metabolites and B[a]P-7,8-diol by a CYP-mediated process. Secondary metabolism, mainly involving epoxide hydrolase and other CYP isoforms,leads to the formation of the highly reactive (+)-anti-BPDE. Several carcinogens present in tobacco smoke are inactivated by GSTs. The most frequently studied carcinogenic PAH diolepoxide, BPDE, is a relatively good substrate for GSTM1, M2, and M3 and better still for GSTP1 (12).

Sensitive detection methods have been used to demonstrate the presence of smoking-related bulky (PAH)-DNA adducts in virtually all target organs of tobacco carcinogenesis. The amounts of(+)-anti-BPDE bound to DNA can be quantified by high-performance liquid chromatography with fluorescence detection by measuring the release of B[a]P-tetrols both from lung tissue and lymphocyte DNA (13). Subsequently, the complex interrelationship between PAH-DNA adduct levels, daily or total smoking dose, genotype, and cancer risk was studied (reviewed in Ref. 14).

This topic has been reviewed by McClellan (10) and Hecht (9, 15). NNK and NNN are the most important TSNAs. They originate mostly from unburned tobacco and are also pyrosynthesized during smoking. The exposure of smokers to TSNAs is much higher than that to other environmental nitrosamines. The evidence that TSNAs are causative in tobacco-induced cancers of the upper aerodigestive tract in humans is highly suggestive: NNK is a powerful lung carcinogen in all species tested; human exposure is comparable with the dose that causes tumors in laboratory animals; and the metabolic activation pathways of NNK are similar in humans and laboratory animals. NNK and NNN require metabolic activation to bind to DNA and express their carcinogenic effects. The metabolism of NNK includes α-methylhydroxylation, α-methylenehydroxylation,pyridine-N-oxidation by CYP-mediated reactions, and reduction to 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol and its conjugation as glucuronide (reviewed in Refs. 9, 10, 11, 12, 13, 14, 15). The last compound can be detected in human urine and is a good indicator of exposure to NNK. N- and O-methylated DNA bases have been detected in many exposed tissues. In addition,pyridyl-oxobutylation of DNA and globin occurs after exposure to NNK or NNN. The keto alcohol released from globin or DNA after hydrolysis allows sensitive human dosimetry of these TSNAs in tobacco users. NNK-derived DNA adducts in humans have been characterized only partially (15).

4-Aminobiphenyl and other aromatic amines are the components of smoke that appear to be primarily responsible for urinary bladder cancer in smokers. The key reactions by which the compounds are metabolized and produce DNA adducts in the bladder epithelium involve N-hydroxylation (CYP1A2) and N-acetylation (NAT1 and NAT2). The resulting hydroxylamine also reacts with hemoglobin to form 4-aminobiphenyl-hemoglobin adducts in smokers or may be further activated by O-acetylation to N-acetoxyarylamines, reactions that are also catalyzed by NAT1 and NAT2. Molecular dosimetry in smokers of black (air-cured) and blond (flue-cured) tobacco provided further evidence that aromatic amines induce bladder cancer in smokers. Smokers of black tobacco, who have a risk for bladder cancer that is two to three times that of smokers of blond tobacco, have 2-fold higher 4-aminobiphenyl-hemoglobin adduct levels and excrete twice as much mutagens in their urine(reviewed in Ref. 16). The 4-aminobiphenyl-DNA guanine adduct was detected as a major smoking-related adduct in biopsy samples from bladder cancer patients and in exfoliated urothelial cells of volunteers. The levels of adducts of 4-aminobiphenyl with hemoglobin and with DNA in the bladder were correlated, and both were related to recent cigarette smoking. The metabolic phenotypes of rapid or slow N-acetylator and rapid or slow N-oxidizer, CYP1A2, significantly affected the levels of 4-aminobiphenyl-hemoglobin and 4-aminobiphenyl-DNA adducts in the urothelium of smokers. Studies in aromatic amine-exposed workers and in smokers have shown previously that the slow acetylator phenotype (slow NAT2) is at higher risk for developing bladder cancer than those with the fast phenotype (reviewed in Ref. 17).

Like preformed carcinogens, tobacco smoke contains ROS and reactive nitrogen species that impose oxidative stress on smokers’tissue. As a consequence, oxidative DNA-base damage has been detected in respiratory tract tissue of smokers, with lipid peroxidation products such as malondialdehyde, crotonaldehyde, and trans-4-hydroxy-2-nonenal, the last of which can be further epoxidized by CYP-mediated reactions. These reactive aldehydes have been shown to form promutagenic exocyclic DNA adducts in human tissues (18, 19) and thus could contribute to carcinogenesis in the upper aerodigestive tract (20). Chewing of tobacco alone or with betel quid is causally associated with oral cancer;because chewing of betel quid generates large amounts of ROS in the mouth, TSNAs and ROS are the major genotoxic agents implicated in oral cancer related to chewing (reviewed in Ref. 21).

Because tobacco carcinogens, ROS, and lipid peroxidation products are likely to be substrates for GSTT1 or M1, the extent of DNA damage and ultimately the cancer risk may be affected by polymorphic CYPs and GST detoxifying enzymes (reviewed in Refs. 17 and 22).

Epidemiological studies have incriminated environmental tobacco smoke as a risk factor for lung cancer in nonsmokers. The concentrations of carcinogenic agents in such smoke appear to be low overall in comparison with those in undiluted mainstream smoke. Involuntary inhalation of tobacco smoke can occur over several hours/day. When comparing the ratio of concentration in sidestream and in mainstream smoke, relatively large amounts of carcinogenic, volatile nitrosamines and aromatic amines are released into sidestream smoke, as reflected by the concentrations of cotinine, a crude marker for uptake of tobacco carcinogens, in body fluids of passive smokers, which are about two orders of magnitude lower than those in active cigarette smokers (8). However, some data suggest that people with certain “at risk” genotypes are particularly susceptible to low doses of carcinogens (Ref. 2; reviewed in Ref. 3). This important health issue should be resolved by properly designed studies that would show whether passive smokers who have developed lung cancer are particularly susceptible.

Some of the principal ways in which genetic polymorphism can affect the expression of gene products or the catalytic activity of the respective enzyme can be summarized as follows:

(a) Nucleotide variations in the coding region of the gene result in amino acid substitution and alter enzyme activity or substrate binding (e.g., CYP2D6).

(b) Deletions in (of) the coding region lead to an inactive enzyme or lack of protein synthesis (e.g.,CYP2A6, CYP2D6, and GSTM1).

(c) Polymorphisms in the noncoding region affect transcriptional control elements involved in basal enzyme expression and induction (e.g., CYP1A1).

(d) Variations in the polyadenylation signal of a gene affect transcript half-life and thus the quantity of enzyme(e.g., NAT1).

(e) Gene amplification increases the quantity of enzyme(e.g., CYP2D6).

(f) Complex interactions of polymorphic genes and/or their enzyme catalysis products (e.g., GSTM1-deficient subjects or cells have greater induction of CYP1A1 and 1A2, probably because of greater bioavailability of inducer compounds).

The role of particular human CYP450 in the metabolism of carcinogens has been reviewed (23, 24, 25). In the following,we evaluate case-control studies on the effect of polymorphic CYPs on the tobacco smoke-associated risk of cancers of the lung, larynx,mouth, esophagus, kidneys, urinary tract, and female breast. The past and recent systematic nomenclature for CYP1A1 polymorphisms is given in Table 1. For the purpose of clarity, we have referred to CYP1A1mutations by the system of Cascorbi et al.(26).⁵

The human enzyme CYP1A1, which is well conserved, is involved in the activation of major classes of tobacco procarcinogens, like PAHs and aromatic amines, and is present in many epithelial tissues. About 10%of the Caucasian population has a highly inducible form of the CYP1A1 enzyme (termed B[a]P-hydroxylase or previously arylhydrocarbon hydroxylase), which is associated with an increased risk for bronchial,laryngeal, and oral cavity tumors in smokers (reviewed in Ref. 27).

The induction of CYP1A1 is initiated by the specific binding of aromatic inducer compounds to the Ah receptor. An Ah receptor nuclear translocator (Arnt) gene is further involved in the CYP1A1 induction pathway. Thus far, no relationship has been found between Ah receptor polymorphism and lung cancer risk (28, 29).

Beginning in 1973 with the pioneering work by Kellerman et al. (30) on B[a]P hydroxylase inducibility and bronchogenic carcinoma, studies on the association of the genetic polymorphism of CYP1A1 and cancer started after cosegregation of the CYP1A1 high inducibility phenotype and polymorphism of the MspI restriction site (31).

The CYP1A1 Ile-Val (m2) mutation in the heme-binding region results in a 2-fold increase in microsomal enzyme activity and is in complete linkage disequilibrium in Caucasians with the CYP1A1 MspI (m1) mutation, which has also been associated experimentally with increased catalytic activity (32). Although the Ile-Val mutation in the CYP1A1 allele did not increase activity in vitro(33, 34), it might be linked to other functional polymorphisms, for example in the regulatory region important for CYP1A1 inducibility. Smokers with the exon 7 Ile-Val mutation were found to have more PAH-DNA adducts in their WBCs than smokers without the variant (35). The amount of these adducts is also elevated in cord blood and placenta of newborns with the CYP1A1-MspI polymorphism (36). In lung parenchymal tissue of smokers, the concentrations of BPDE and bulky(PAH)-DNA adducts were positively correlated with CYP1A1 enzyme activity (13). Significant ethnic differences in the frequency of homozygous CYP1A1 MspI alleles have been observed, and both the MspI and Val alleles are rarer in Caucasian than in Japanese populations (26).

A proposal has been made for a mechanism whereby the CYP1A1genotype and GSTM1 0/0 gene-gene interactions result in a greater-than-additive risk for DNA damage and cancer; in human cells,deletion of GSTM1 is associated with strong inducibility of CYP1A1 gene transcription by 2,4,7,8-tetrachlorodibenzo-para-dioxin (37). When BPDE-DNA adduct levels were measured in lung tissue of smokers, a significant interaction between deficiency of the GSTM1phenotype and high CYP1A1 inducibility or CYP1A1allelic variants was observed (38, 39), leading to very high adduct levels in Caucasians with CYP1A1/MspI/MspI-GSTM1 0/0 [see“ CYP1A1-GSTM1 Genotype Dependence of Bulky (PAH)-DNA Adduct Levels and of Other Effect Markers in Smokers”]. These data suggest that this “at risk” genotype combination predisposes to an increased risk for tobacco-associated DNA damage and lung cancer.

The relationship between CYP1A1 variants and lung cancer risk in various ethnic populations has been examined in more than 20 studies. Early Japanese studies pointed to an increased risk for lung cancer in association with both the m1 (Table 2, Table 2A study nos. 1 and 10) and m2 mutations (Table 2, study 4); the CYP1A1 genotype was particularly important at a low level of smoking and in the development of squamous cell carcinoma (Table 2,study 3). These findings were not confirmed in studies conducted in Norway (Table 2, study 2), Finland (Table 2, study 5), the United States (Table 2, study 6), and Sweden (Table 2, study 8), perhaps because of the much lower prevalence of the m1 allele in Caucasians. Larger studies in mixed American populations (Table 2,study nos. 16, 17, 18, and 20) do point to an increased risk for lung cancer among carriers of m1 alleles, whereas in Caucasian smokers in France, no significant association was observed for either m1 or m2 (Table 2, study 19). In two Brazilian populations, the presence of the m2 allele was significantly associated with an increased risk for lung cancer (Table 2, study 12). The African-American-specific m3 (*3) mutation was not associated with an increased risk for lung cancer overall in three studies (Table 2, study nos. 9, 13, and 21); however, in one of them, a significantly increased risk for adenocarcinoma was seen for carriers of the m3 mutation (Table 2, study 21), and the OR for lung cancer was 8.4 for the genotype m1/m1,m2 (*2A/*2B). In Chinese, the m1 and m2 mutations were not correlated with either allele or lung cancer risk (Table 2, study 24),whereas results from the Republic of Korea surprisingly showed a significantly decreased risk for lung cancer among carriers of the m2 allele (Table 2, study 25). A recently described m4 mutation in close proximity to m1 has thus far been investigated in only one study of Caucasians, in which it did not correlate with an increased risk for lung cancer (Table 2, study 14). It is possible that individuals carrying the m4 mutation were misclassified as carriers of the m2 mutation in earlier studies, in which allele-specific primers were used. The combined CYP1A1 variants (either m2 or m1) and GSTM1 0/0 genotype have been associated with a significantly increased risk for lung cancer (Table 2, study nos. 4, 7, and 11),especially squamous cell carcinoma (Table 2, study nos. 4 and 20) in Japanese populations.

The CYP1A1 enzyme is present in oral tissue (40), and CYP1A1 variants and cancer risk at these sites have been investigated in 11 studies. There was overrepresentation of the CYP1A1 Ile-Val variant among Caucasian patients with oral cancer, but the prevalence of the Ile-Val variant was significantly higher in nonsmokers than smokers, although the number of nonsmokers was low (Table 3, Table 3Astudy 2). Similarly, an increased prevalence of CYP1A1Val-Val variant was found among Japanese patients with head and neck cancers and especially those with pharyngeal cancer (Table 3, study 5). Individuals with the homozygous CYP1A1 MspI(m1/m1) variant were at significantly increased risk for oral squamous cell carcinoma, in particular after exposure to low concentrations of PAH. The combination of homozygous CYP1A1 MspI (m1/m1) variant and GSTM1 0/0 further increased the risk. The buccal mucosa and upper gingiva appear to be the most susceptible tissues in carriers of the risk genotype (Table 3,study 6). The highest prevalence of p53 mutation was observed in oral tumors from patients with the CYP1A1(Val)/GSTM1 active genotype (41).

Studies in Caucasians and Japanese showed no association with esophageal cancer. Esophageal cancer patients in China who were heavy smokers had a 3-fold higher frequency of the CYP1A1 Val-Val variant. The risk was further increased in patients with the combination of CYP1A1 Val-Val and GSTM1 0/0 genotypes (Table 4, study 4).

Two studies of Japanese and Caucasians showed no significant associations (Table 5, study nos. 1 and 2). High CYP1A1 expression, as determined by immunohistochemistry, correlated with the grade of urinary bladder tumor (42).

None of four studies of the effect of the m2 allele in Caucasian populations found a significant association overall (Table 6, Table 6Astudy nos. 1–4), but a significant increase in risk was found among postmenopausal women who were light smokers and carried the m2 allele (Table 6, study 2). These results were not confirmed in a smaller study of Caucasians, but an increased risk for breast cancer was found among African-American women with the homozygous m1/m1 genotype (Table 6, study 3). Bailey et al. (Ref. 43; Table 6, study 4),investigating m1, m2, m3, and m4 in Caucasian and African-American women, found no significant association, nor did the largest study, nested within the Nurses’ Health Study (Table 6, study 5), although in the latter study an increased risk was found for smoking at a young age and the presence of m1 and/or m2. A recent study of the m2 allele in postmenopausal women (Table 6, study 2) showed a significant correlation between above-median serum concentrations of polychlorinated biphenyls, the m2 allele, and breast cancer risk (Table 6, study 6).

Taken together, there is increasing evidence that the homozygous CYP1A1 (MspI and Ile-Val) genotypes are at higher risk for contracting smoking-associated lung (squamous cell carcinoma),head and neck, and esophageal cancers, as particularly seen in Asian study populations where these “at risk” allele frequencies are 8–18 times higher than in Caucasians. The cancer risk was further increased in the combined “at risk” CYP1A1-GSTM1 0/0 genotypes. The underlying mechanism of this genetic predisposition in these “at risk” smokers is supported by a higher prevalence of p53 mutations and B[a]P-DNA adducts (see “CYP1A1-GSTM1 Genotype Dependence of Bulky (PAH)-DNA Adduct Levels and of Other Effect Markers in Smokers”). For breast cancer risk, the available studies conducted on CYP1A1 variants mostly in Caucasians did not reveal an overall association.

This isoform activates many dietary and tobacco procarcinogens, notably aromatic and heterocyclic amines, and also metabolizes nicotine. Human CYP1A2 has 72% sequence identity with CYP1A1and, in contrast to extrahepatic CYP1A1, CYP1A2 appears to be expressed mainly in the liver and only weakly in the peripheral lung (44). Like CYP1A1, CYP1A2 is regulated in part by the Ah-receptor system and is induced in humans by a variety of chemicals. The activity of this enzyme can be determined in a noninvasive assay involving measurement of caffeine 3-demethylation.

Recently, a genetic polymorphism in the 5′ flanking region of the human CYP1A2 was identified that affects inducibility (45). Another single nucleotide polymorphism was found in intron 1, which is associated with high catalytic activity when subjects are exposed to tobacco smoke (46, 47). Individuals who are homozygous for the high inducibility genotype were shown to account for ∼45% of healthy Caucasians. A subgroup of smokers had a 1.6-fold increase in the caffeine demethylation ratio(ratio of paraxanthine:caffeine in serum) over that in nonsmokers.

Gene-gene interactions between GSTM1 0/0 and CYP1A2 and CYP1A1 enzyme induction have been observed in smokers; GSTM1 deficiency not only led to increased hepatic CYP1A2 activity in current smokers but also to significantly increased levels of bulky PAH-DNA adducts in lung parenchyma of smokers and ex-smokers, over that in individuals with wild-type GSTM1(4, 38, 39, 48). CYP1A2 activity was higher in GSTM1 0/0 subjects after exposure to cigarette smoke and heterocyclic amines from cooked meat. Exposed individuals with CYP1A1 Ile-Val alleles had greater CYP1A2 activity than those with wild-type CYP1A1 (49). GSTM1 0/0 led to higher levels of 4-aminobiphenyl-hemoglobin adducts in smokers (50). Such gene-gene interactions are probably attributable to a greater bioavailability of aromatic inducer compounds in GSTM1 0/0 subjects, leading to a higher rate of induction of CYP1A1 and CYP1A2 in smokers, which in turn increases macromolecular carcinogen binding.

In one preliminary case-control study (Table 5, study 3), patients with the intron 1 variant had an OR of 1.7 for bladder cancer if they were smokers; if they also had the slow NAT2 phenotype, the OR was 2.2.

CYP1B1, 1A2, and 3A4 all catalyze 2- and 4-hydroxylation of 17β-estradiol, but 4-hydroxylation is selectively catalyzed by CYP1B1. This enzyme activates many PAH-dihydrodiols, aromatic amines,and other groups of procarcinogens. CYP1B1 is also induced by Ah-receptor ligands. This enzyme is expressed in human kidney,prostate, ovary, and breast, and three CYP1B1 polymorphisms have been identified in exon 3, two of which are associated with amino acid substitutions, i.e., Val⁴³²Leu and Asn⁴⁵³Ser in the heme-binding domain of the enzyme (51).

In the only study conducted thus far, no association was found with the CYP1B1 genotype. Caucasian patients with the codon 432 Val-Val genotype were more likely to have breast tumors containing estrogen and progesterone receptors (Table 6, study 7).

In humans, CYP2A6 mediates 7-hydroxylation of coumarin, a component of cigarette smoke, and activates several nitrosamines in tobacco smoke,including NNK (15, 52). The catalytic selectivity of CYP2A6 appears to overlap with that of CYP2E1. The location of CYP2A6 and 2E1 in extrahepatic tissues such as lung, nasal and pharyngeal areas is of interest. Two CYP2A6 variant alleles have been identified (*2 and *3). The prevalence of the Leu¹⁶⁰His variant allele in Caucasians is ∼2%,and it is associated with lower coumarin 7-hydroxylation activity. A new allele has been described in which exons 5–9 are deleted (53). Individuals lacking functional CYP2A6have impaired nicotine metabolism and may thus be protected against tobacco dependence. In one study using a disputed genotyping method,carriers of the variant allele were reported to smoke fewer cigarettes (54). These findings were not confirmed in another study (55). The discrepancy could be resolved by newly available genotyping methods (56, 57).

The only case-control study published thus far (Table 7, Table 7Astudy 1) showed no associations with CYP2A6*2.

Individuals with the PM phenotype (i.e., deficient in debrisoquine 4-hydroxylation) were first reported in 1984 to have a lowered risk for lung cancer (58). The PM phenotype,inherited as an autosomal recessive trait, is attributable to several defective allelic CYP2D6 variants, three of which account for >90% of all PM individuals. The mutations that cause loss of gene function are attributable mainly to two point mutations that result in CYP2D6*3 (formerly A) and CYP2D6*4(B) alleles and deletion of the entire CYP2D6*5 allele (D). A CYP2D6*2xn (L2) allele associated with 2–12-fold amplification of the CYP2D6 gene is found in carriers known as UMs (59). Conflicting data exist on whether CYP2D6 is expressed in human lung (60). Most searches for procarcinogens that are activated by CYP2D6 have been unsuccessful. It can activate the tobacco-specific nitrosamine NNK and also nicotine, but other P-450s are more active in this respect.

Associations have been found between nicotine depen-dence and PM (61) and between UM and smoking addiction (62). These findings may make establishing causality between this polymorphism and cancer risk more complex.

Nine genotyping studies have been conducted, including several with large samples and with control for confounding factors. In three studies (Table 7, study nos. 3, 4, and 6), a significant association was found between lung cancer and EM genotype, and in one study an association with UM genotype and lung cancer risk was found for African-Americans (Table 7, study 8). Legrand Andreoletti et al. (Ref. 63; Table 7, study 9) screened over 40 alleles in cases and controls. They found no significant association for lung cancer overall, but the *1A/*2-EM genotype combination was significantly associated with lung cancer. Two meta-analysis showed no association or one of borderline significance between the EM genotype and increased risk (64, 65).

One of three studies of Caucasians showed a significantly higher frequency of the PM genotype among cases (Table 3, study 9).

One of four studies showed an association only with the HEM (Table 7,study 4). The HEM genotype in smokers was associated with more aggressive bladder tumors (66).

One American study (Table 6, study 8) of alleles *3, *4 and *5 showed no association with the CYP2D6 genotype, whereas a slightly larger study in Spain (Table 6, study 9) of alleles *3, *4 and *9 showed a significant association with HEM.

Analyses of all results on CYP2D6 genotype and lung cancer risk found no or a borderline protective effect of the PM genotype. For other cancer sites, the association with disease susceptibility was inconclusive.

The ethanol-inducible CYP2E1 metabolizes many known procarcinogens, including NNN, NNK, and other volatile nitrosamines found in tobacco smoke. Chlorzoxazone 6-hydroxylation is catalyzed by CYP2E1. Wide interindividual variation in the expression of the CYP2E1 gene has been reported in humans, which is possibly attributable to gene-environment interaction. CYP2E1 is induced in mice exposed to cigarette smoke by inhalation (67). Its regulation involves complex transcriptional and posttranscriptional mechanisms. Although in Caucasians no relationship was found between in vivo activity of this enzyme and genotype, in Japanese the presence of the variant c2 alleles resulted in a significant reduction in the oral clearance of chlorzoxazone, after adjustment for age and sex. The mean activity in individuals with the c2/c2 genotype was significantly lower than that in individuals with either the homozygous wild-type or the heterozygote genotype. Body weight and dietary factors were the major modulators of interindividual variation (68).

The human CYP2E1 gene is functionally well conserved, but several polymorphic alleles occurring at low frequency have now been identified. The RFLPs, revealed by either RsaI G_–1259C or PstI C_–1091T, are located in the 5′ flanking transcription region of this gene and appear to be in complete linkage disequilibrium with each other (c1, common allele; c2, rare allele). Although the primary sequence of the enzyme is not altered, increased gene transcription has been suggested (69). A T_-7668A substitution in intron 6 of the CYP2E1 gene is revealed by a DraI RFLP (C, minor allele; D, common allele). The RsaI and the DraI polymorphisms appear to be linked, i.e., individuals with the RsaI polymorphism also had a mutant DraI allele, although the reverse is not true. A TaqI RFLP at position 9930 (intron 7)of the CYP2E1 gene has been reported, but no phenotype has been associated with this mutation.

The wild-type DraI genotype was associated with an increased risk for lung cancer in 3–5 of 16 studies in Japanese,Mexican-Americans, and mixed populations (Table 7, study nos. 11, 13,16, 24, and 25). More conflicting results have been published concerning the RsaI/PstI mutation. The rare PstI/RsaI c2 allele has been associated with decreased risk for cancer in two studies of 11 (Table 7, study nos. 23 and 25), and in one study the c2 allele frequency was significantly lower among cases than controls (Table 7,study 15); however, in an additional study, the c2/c2genotype correlated positively with p53 mutations and with squamous cell carcinoma (Table 7, study 21). Additionally, in one small study the c2/c2 genotype was associated with adenocarcinoma(Table 7, study 22).

Five studies showed no association between head and neck cancer and CYP2E1 variants; however, Chinese patients who were not betel quid chewers had a higher prevalence of the c2 allele(PstI/RsaI; Table 3, study 11).

The c2 allele was overrepresented among Chinese esophageal cancer patients (Table 4, study 9).

Three studies found no association with bladder cancer. One study in Caucasian women revealed a higher risk for renal carcinomas among those with DraI variants (OR, 8.0; Table 5, study 10).

The only study of the DraI polymorphism in Caucasians (Table 6, study 10) found a significant association among premenopausal smokers only (OR, 11.1).

Because the frequencies of variant alleles are very low in Caucasians and African-Americans, the statistical power of the studies is low. Altogether, conflicting results have been reported on the importance of CYP2E1 genotypes in well-documented tobacco-related cancers.

The levels of all smoking-related DNA adducts in the larynx were correlated with the presence of P4502C protein, suggesting a role of CYP2C9 in DNA adduction of PAH-type tobacco carcinogens (70). Two mutant alleles, CYP2C9*2(Arg¹⁴⁴Cys) and *3(Ile³⁵⁹Leu), have been described, and CYP2C9 has a specific substrate, tolbutamide. The level of bulky DNA adducts in normal bronchial tissue of smokers was found to be higher in individuals with the homozygous CYP2C9*3/*3 genotype (71).

One of two large (Table 7, study 27), well-designed studies in African-Americans and Caucasians (Table 7, study nos. 27 and 28)revealed an association of the *2 allele with borderline-increased risk(OR, 1.6).

Members of the human CYP2C gene subfamily are constitutively expressed, and at least seven human CYP2C genes may exist. Several defective CYP2C19 alleles are the basis for the(S)-mephenytoin 4′-hydroxylase polymorphism. In addition to(S)-mephenytoin, CYP2C19 also metabolizes a variety of clinically used drugs. The most common variant allele, *2, has an aberrant splice site in exon 5 (72). The premature stop codon mutant *3 allele has thus far been found only in Asians (73).

A very small study on Japanese patients (Table 7, study 29) revealed a significant association of the PM genotype with squamous cell carcinoma of the lung, but the association with bladder cancer seen in Caucasians was not found (Table 5, study 11).

This isoform is the major P450 expressed in human liver and small intestine. It can activate numerous procarcinogenic PAH dihydrodiols,such as BPDE, and also metabolizes NNN (74). Whether genetic or solely environmental factors are responsible for the wide variation in human 3A4 activity is unknown. Although the three CYP3A genes, 3A4, 3A5, and 3A7, are expressed at widely different levels, polymorphism has been found only for CYP3A4 and CYP3A5 to date. Several allelic variants of the CYP3A4 gene were reported (75), but none was apparently related to catalytic activity in the liver samples from which the DNA was derived. No extensive studies on CYP3A4 polymorphism have been reported.

This gene codes for the cytochrome P450C17α-enzyme, which mediates both steroid 17α-hydroxylase and 17,20-lyase activities and functions at key branch points in human steroidogenesis. The 5′-U terminal repeat of CYP17 contains a 1-bp polymorphism that creates a recognition site for the MspAI restriction enzyme and has been used to designate two alleles, A1 and A2. Premenopausal women with CYP17A2 variants have higher serum concentrations of estradiol and progesterone (76).

Four studies in mainly Caucasian populations (Table 6, study nos. 12–15) and one in a mixed Hawaiian population (Table 6, study 11)showed no association with the CYP17 genotype.

This gene encodes aromatase, which is responsible for the rate-limiting step in the metabolism of C19 steroids to estrogens. Aromatase activity has been found in a number of tissues, including normal and transformed breast tissue. A common, high-heterozygosity tetranucleotide simple tandem repeat polymorphism in intron 4 has been described (77), but it is not known whether this polymorphism is associated with a specific phenotype.

Two studies of the tandem repeat polymorphism in Caucasian women showed that an increased proportion of breast cancer patients had the short alleles; however, the allele frequencies varied greatly between the two study populations (Table 6, study nos. 16 and 17).

The genotype dependence of various effect markers, such as DNA adducts, cytogenetic damage and p53 mutations, have been studied, most effort having been focused on PAH-(BPDE)-DNA adducts related to tobacco smoking. Overall, the data indicate that smokers have higher PAH-DNA adduct levels in target tissues and leukocytes than ex- and nonsmokers (Table 8). The increased formation and the wide variation in levels of PAH-DNA adducts in some smokers occupationally exposed to PAH suggest that polymorphisms in genes related to PAH metabolism lead to increased DNA binding and cancer risk (reviewed in Ref. 14).

The relationship between CYP1A1 variants, alone or in combination with GSTM1, and the formation of bulky (PAH)-DNA adducts in human tissues (autopsy tissues excluded) and leukocytes remained controversial for some time (Table 8). Several studies showed a weak or no effect of m1 and m2 on adduct levels(Table 8, study nos. 3, 4, 6, 8 and 9). More recent studies, in which specific, sensitive detection methods were used (such as for(+)-anti-BPDE-DNA adducts), have shown clearly the dependence of adduct levels on CYP1A1 genotype, which is most pronounced in GSTM1-deficient smokers. Lung and leukocytes of Caucasian smokers with the CYP1A1m1/m1-GSTM1 0/0 combination clearly contained more BPDE-DNA adducts (Table 8, study nos. 1, 2, and 5). One study in which an ELISA was used (Table 8, study 7) showed a similar effect of m2-GSTM1 combinations.

In bronchial tissues of smokers with highly induced CYP1A1 enzyme and GSTM1 0/0, the BPDE-DNA levels were 100-fold higher than in subjects with active GSTM1 (Table 8, study 2). Another study found no effect of the CYP1A1 [MspI(m1) or Ile-Val (m2) mutations] or GSTM1 0/0 genotypes on bulky DNA adduct levels, when the ³²P-postlabeling method was used (Table 8, study 3); however, carriers of homozygous CYP1A1 m1 had higher BPDE-DNA adduct levels than individuals with the wild type(Table 8, study 1).

Studies of leukocytes from mostly Caucasians exposed to PAH, including smokers and nonsmokers, gave contradictory results (Table 8). No effect of m1 and m2 was observed in leukocyte DNA from lung cancer patients (Table 8, study 11), but a significant, 2-fold increase in PAH-DNA adduct level was found when m2 variants were combined with GSTM1 0/0 or wild type (Table 8, study 7). Heterozygous m1 or m2 variants were associated with an increase in the median BPDE adduct level when compared with the wild type (180). One Caucasian subject with the very rare m1/m1 genotype in combination with GSTM1 0/0 had an extremely high level of BPDE adducts:44/10⁸ nucleotides (Table 8, study 5).

Overall, the data are compatible with the assumption that GSTM1 0/0 is a moderately strong susceptibility factor but may become a dominant risk factor in the presence of certain gene-gene combinations. This results in increased DNA damage and mutational events in target and surrogate tissues (leukocytes). These findings provide a mechanistic background why such “at risk” genotypes correlate with increased risks for tobacco-related lung cancer, even at a low level of cigarette smoking (Ref. 78, Table 2). This was seen more clearly in Japanese populations where the “at risk” alleles occur much more frequently than in Caucasians.

A suggested gender difference in lung cancer susceptibility was supported by a study showing that female smokers had a significantly higher level of aromatic DNA adducts in lung tissue than males at an equal cigarette smoking dose (pack-year; Ref. 79).

Lung tumors in Japanese smokers were found to harbor significantly more p53 mutations in people who had the susceptible CYP1A1 genotype. Individuals with the combination of CYP1A1 m2/m2 and GSTM1 0/0 genotypes had an 8-fold greater frequency of p53 mutations than persons with neither genotype (80). Also, lung cancer patients with this “at risk” genotype combination who had undergone an operation had a remarkably shortened survival (81). Shorter survival of operated lung cancer patients with high pulmonary CYP1A1 enzyme inducibility was reported previously (82).

Taken together, because an elevated DNA and mutational damage associated with the “at risk” alleles has been found in both Asian and Caucasian smokers, large-scale studies should prove the prediction that carriers of (homozygous) CYP1A1variants/GSTM1 0/0 combinations of any ethnicity could be at an increased risk for tobacco-related (lung, head, and neck) cancers.

Molecular epidemiology has contributed to a growing awareness of the importance of relatively common genetic and acquired susceptibility factors in modulating risks associated with exposure to environmental carcinogens. Because cancer is largely a preventable disease, the future challenge of molecular epidemiology is to analyze individuals who are exposed to carcinogens for a combination of genotypes associated with susceptibility to cancer. It is evident that use of more precisely measurable intermediary risk markers, like DNA adducts,cytogenetic damage, and mutations, rather than cancer as an end point,will allow the identification of combinations of cancer-relevant genes that positively or negatively affect cancer outcome in humans. Such associations could then be verified in epidemiological studies designed to address the association or hypothesis to be confirmed. Thus, further progress is expected from studies in which biomarkers for carcinogen exposure, early biological effects, and susceptibility are integrated,which should allow establishment of the risk profiles of individuals and subgroups in given exposure situations.

Many of the published studies listed in Tables 2,3,4,5,6,7,8 have shortcomings that should be avoided in future, if possible. Furthermore, there is a bias against publishing (and citing) the absence of correlations. IARC (83, 84) provided state-of-the-art reviews of the application of biomarkers and the design and analysis of molecular epidemiological studies. The prerequisites for proper study design and conduct include: (a) clear definition of representative study populations and controls; (b) a sample size adequate to provide enough statistical power; (c) proper documentation (or measurement) of exposure; (d) avoidance of confounding because of use of study subjects of mixed ethnic background; and (e) study only of gene polymorphisms that have been shown to lead to altered phenotypic expression.

Rapid advances in high-throughput gene analysis by DNA chip technology will speed up the identification of new mutations in predisposing cancer genes. The main task, however, will be to characterize the functional significance of these gene variants in humans. Such efforts are under way, e.g., the Environmental Genome Project pursued by the National Institute for Environmental Health Sciences in the United States (85). The aims are to define genetic variation in a selected number of (∼200) genes in the American population and to relate them to disease risk and individual susceptibility, particularly in combination with specific chemical and physical exposures.

Knowledge of the prevalence and distribution of common genetic susceptibility factors and the ability to identify susceptible individuals or subgroups will have substantial preventive implications,in particular if more data are collected to show that people with certain “at risk” genotypes are more susceptible to low levels of exposure (see “Environmental Tobacco Smoke and Exposure to Low Doses of Carcinogens”). It is conceivable that such subjects could be:(a) more easily persuaded to avoid hazardous exposure like tobacco use; (b) targeted for intensive smoking cessation programs; (c) be enrolled in chemoprevention trials; and(c) be involved in cancer screening programs that are not appropriate for the general population. However, before results of individual screening for genetic traits can be used efficiently to implement preventive measures, more cancer-predisposing genes need to be studied and gene-environment and gene-gene interactions elucidated. To this purpose, the need of well-designed, large-scale studies is emphasized.