It is estimated that cigarette smoking kills over 1 000 000 people each year by causing lung cancer as well as many other neoplasmas. p53 mutations are frequent in tobacco-related cancers and the mutation load is often higher in cancers from smokers than from nonsmokers. In lung cancers, the p53 mutational patterns are different between smokers and nonsmokers with an excess of G to T transversions in smoking-associated cancers. The prevalence of G to T transversions is 30% in smokers’ lung cancer but only 12% in lung cancers of nonsmokers. A similar trend exists, albeit less marked, in laryngeal cancers and in head and neck cancers. This type of mutation is infrequent in most other tumors aside from hepatocellular carcinoma. At several p53 mutational hotspots common to all cancers, such as codons 248 and 273, a large fraction of the mutations are G to T events in lung cancers but are almost exclusively G to A transitions in non-tobacco-related cancers. Two important classes of tobacco smoke carcinogens are the polycyclic aromatic hydrocarbons (PAH) and the nicotine-derived nitrosamines. Recent studies have indicated that there is a strong coincidence of G to T transversion hotspots in lung cancers and sites of preferential formation of PAH adducts along the p53 gene. Endogenously methylated CpG dinucleotides are the preferred sites for G to T transversions, accounting for more than 50% of such mutations in lung tumors. The same dinucleotide, when present within CpG-methylated mutational reporter genes, is the target of G to T transversion hotspots in cells exposed to the model PAH compound benzo[a]pyrene-7,8-diol-9,10-epoxide. As summarized here, a number of other tobacco smoke carcinogens also can cause G to T transversion mutations. The available data suggest that p53 mutations in lung cancers can be attributed to direct DNA damage from cigarette smoke carcinogens rather than to selection of pre-existing endogenous mutations.
Cigarette smoking causes 30% of all cancer deaths in developed countries (World Health Organization, 1997). In addition to lung cancer, cigarette smoking is an important cause of esophageal, oral, oropharyngeal, hypopharyngeal, and laryngeal cancers as well as pancreatic cancer, bladder cancer, and cancer of the renal pelvis (International Agency for Research on Cancer, 1986b). Cigarette smoking has also been linked to cancers of the stomach, renal body, liver, colon, nose, and myeloid leukemia, although the connection to these cancers is weaker (Chao et al., 2000; Doll, 1996). Carcinogenic compounds in cigarette smoke are thought to be responsible for these cancers.
The mainstream smoke emerging from the mouthpiece of a cigarette is an aerosol containing about 1010 particles/ml and 4800 compounds (Hoffmann and Hecht, 1990). Experimentally, vapor-phase components of the smoke can be separated from the particulate phase by a glass fiber filter. The vapor phase comprises over 90% of the mainstream smoke weight (Hoffmann et al., 2001). The main constituents of the vapor phase are nitrogen, oxygen, and carbon dioxide. Potentially carcinogenic vapor phase compounds include nitrogen oxides, isoprene, butadiene, benzene, styrene, formaldehyde, acetaldehyde, acrolein, and furan (Hoffmann et al., 2001). The particulate phase contains at least 3500 compounds and many carcinogens including polycyclic aromatic hydrocarbons (PAH), N-nitrosamines, aromatic amines, and metals (Hoffmann et al., 2001). Cigarette smoke condensate can be prepared by trapping non-volatile (mainly particulate phase constituents) in cold-traps. Cigarette smoke condensate reproducibly and robustly causes tumors when applied to mouse skin and implanted in rodent lung (International Agency for Research on Cancer, 1986b). Fractions of the condensate which contain PAH also induce tumors in these models, but the concentrations of the PAH are too low to explain the carcinogenicity (reviewed in Rubin, 2001). Other fractions of the condensate have tumor promoting and cocarcinogenic activity which enhance the carcinogenicity of the PAH-containing fractions. Inhalation experiments using Syrian Golden hamsters demonstrate that whole cigarette smoke and its particulate phase consistently induce preneoplastic lesions and benign and malignant tumors of the larynx (International Agency for Research on Cancer, 1986b). This model system has been widely applied and is the most reliable one for induction of tumors by inhalation of cigarette smoke. Tumors are also observed in hamsters exposed only to the particulate phase of smoke (International Agency for Research on Cancer, 1986b). Recently, studies in A/J mice exposed to environmental tobacco smoke (comprised of 89% mainstream smoke and 11% sidestream smoke in this experimental model) by inhalation demonstrate a small but reproducible increase in lung tumor multiplicity (Witschi, 2000). Tumor induction in this model is due to vapor phase constituents of cigarette smoke. Thus, there is reliable evidence that both particulate phase and vapor phase constituents of cigarette smoke cause tumors in laboratory animals, and that tumor promoters and co-carcinogens are also involved in the observed response.
There are over 60 carcinogens in cigarette smoke that have been evaluated by the International Agency for Research on Cancer, and for which there is ‘sufficient evidence for carcinogenicity’ in either laboratory animals or humans (Hoffmann et al., 2001). They belong to various classes of chemicals, as follows: PAH (10 compounds), aza-arenes (3), N-nitrosamines (8), aromatic amines (4), heterocyclic amines (8), aldehydes (2), volatile hydrocarbons (4), nitro compounds (3), miscellaneous organic compounds (12), and metals and other inorganic compounds (9). Other carcinogens not evaluated by IARC are also likely to be present. For example, among the PAH, multiple alkylated and high molecular-weight compounds have been detected, but are incompletely characterized with respect to their carcinogenicity (Snook et al., 1978). Eighteen N-nitrosamines are present in cigarette smoke (Hoffmann et al., 2001). Published lists also include 106 aldehydes and 138 monocyclic aromatics; some may be carcinogenic (Hoffmann et al., 2001).
Tobacco smoke carcinogens and cancer
The potential role of tobacco smoke carcinogens in smoking-associated cancers can be evaluated by various means, but it is important to consider levels of the compounds in cigarette smoke and their ability to induce tumors in laboratory animals. In the following, we discuss these factors with respect to cancers of the lung, oral cavity, esophagus, pancreas, and bladder.
Established pulmonary carcinogens in cigarette smoke include PAH, aza-arenes, tobacco-specific nitrosamines, e.g. 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NNK), 1,3-butadiene, ethyl carbamate, ethylene oxide, nickel, chromium, cadmium, polonium-210, arsenic, and hydrazine. These compounds convincingly induce lung tumors in at least one animal species and have been positively identified in cigarette smoke.
Among the PAH, benzo[a]pyrene (BaP) is the most extensively studied compound (Phillips, 1983; Besarati Nia et al., 2002a) and its ability to induce lung tumors upon local administration or inhalation has been convincingly established (Hecht, 1999; International Agency for Research on Cancer, 1983). Lung tumors were not observed when BaP was administered in the diet to B6C3F1 mice (Culp et al., 1998). In studies of lung tumor induction by implantation in rats, BaP is more carcinogenic than the benzofluoranthenes or indeno[1,2,3-cd]pyrene (Deutsch-Wenzel et al., 1983). Extensive analytical data convincingly demonstrate the presence of BaP in cigarette smoke. Its sales-weighted concentration in current ‘full-flavored’ cigarettes is about 9 ng per cigarette (Chepiga et al., 2000). The abundant literature on BaP tends to diminish attention to other PAH such as dibenz[a,h]anthracene, 5-methylchrysene, and dibenzo[a,i]pyrene which are substantially stronger lung tumorigens than BaP in mice or hamsters, but occur in lower concentrations in cigarette smoke than does BaP (Nesnow et al., 1995; Sellakumar and Shubik, 1974).
Among the N-nitrosamines, N-nitrosodiethylamine is an effective pulmonary carcinogen in the hamster, but not the rat (Reznik-Shuller, 1983; International Agency for Research on Cancer, 1978). Its levels in cigarette smoke (up to 3 ng/cigarette) are low compared to those of other carcinogens. The tobacco-specific N-nitrosamine NNK is a potent lung carcinogen in rodents (Hecht, 1998). Its activity is particularly impressive in rats, where total doses as low as 6 mg/kg, administered by s.c. injection, or 35 mg/kg administered in the drinking water, produced significant incidences of lung tumors. Even lower doses induced lung tumors when considered in dose-response trend analyses (Hecht, 1998). It is the only compound in cigarette smoke known to induce lung tumors systemically in all three commonly used rodent models. NNK has a remarkable affinity for the lung, causing mainly adenoma and adenocarcinoma, independently of the route of administration (Hecht, 1998). NNK is the most abundant systemic lung carcinogen in cigarette smoke. Multiple international studies definitively document the presence of NNK in cigarette smoke; its sales-weighted concentration in current ‘full-flavored cigarettes’ is 131 ng/cigarette (Chepiga et al., 2000; Spiegelhalder and Bartsch, 1996; Hecht and Hoffmann, 1988).
Lung is one of the multiple sites of tumorigenesis by 1,3-butadiene in mice, but is not a target in the rat (International Agency for Research on Cancer, 1992). B6C3F1 mice develop lung tumors at exposure concentrations that are three orders of magnitude lower than those that cause cancer in Sprague-Dawley rats. These interspecies differences are likely due to differences in metabolism of 1,3-butadiene. Mice convert a higher portion of the parent compound to highly carcinogenic 1,2,3,4-diepoxybutane, while the detoxification pathway via conjugation with glutathione is more prominent in rats (Thornton-Manning et al., 1995). Ethyl carbamate is a well-established pulmonary carcinogen in mice but not in other species (International Agency for Research on Cancer, 1974a). Ethylene oxide induces pulmonary tumors in mice, but not in rats (International Agency for Research on Cancer, 1986a). Nickel, chromium, cadmium, and arsenic are all present in tobacco and a percentage of each is transferred to mainstream smoke (Hoffmann et al., 2001). Levels of polonium-210 in tobacco smoke are insufficient to have a significant impact on lung cancer initiation in smokers (Harley et al., 1980). Hydrazine is an effective lung carcinogen in mice and has been detected in cigarette smoke (International Agency for Research on Cancer, 1973). Formaldehyde and acetaldehyde induce nasal tumors in rats when administered by inhalation (International Agency for Research on Cancer, 1982, 1985, 1999; Swenberg et al., 1980). Although they are not lung carcinogens, their concentrations in cigarette smoke are so high that they may nevertheless play a significant role. There is approximately 100 000 times more acetaldehyde in a cigarette than BaP (Chepiga et al., 2000).
Collectively, the available data indicate that PAH and NNK are important lung carcinogens in cigarette smoke most likely to be involved in lung cancer initiation in smokers. Their potent carcinogenic activities compensate for their relatively low concentrations in tobacco smoke. Other carcinogens mentioned here, as well as tumor promoters and co-carcinogens, may also play a role as causes of lung cancer in smokers.
The potent PAH carcinogen 7,12-dimethylbenz[a]anthracene (DMBA) is routinely used for induction of oral tumors in the hamster (Solt et al., 1987). However, DMBA is not present in cigarette smoke. Other PAH have been less frequently tested in this model. A mixture of NNK and NNN induced oral tumors in rats treated repetitively by oral swabbing (Hecht et al., 1986). The rat oral cavity is one target of benzene carcinogenecity (National Toxicology Program 1986). The risk for oral cancer is markedly enhanced by alcohol consumption in smokers, perhaps due in part to enhancement of carcinogen metabolic activation by ethanol (Melikian et al., 1990; Mccoy and Wynder, 1979).
Numerous N-nitrosamines are potent esophageal carcinogens in rats (Preussmann and Stewart, 1984). Among these, N′-nitrosonornicotine (NNN) is by far the most prevalent in cigarette smoke. N-nitrosodiethylamine and N-nitrosopiperidine are two other smoke constituents that could be involved in esophageal tumor induction in smokers. BaP induces some esophageal tumors when administered to mice in the diet (Culp et al., 1998). The risk for esophageal cancer in humans is also enhanced by alcohol consumption (Mccoy and Wynder, 1979).
NNK and its major metabolite 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol (NNAL) are the only known pancreatic carcinogens in cigarette smoke (Rivenson et al., 1988). Low doses of these nitrosamines induce pancreatic tumors in rats, in addition to lung tumors (Hecht, 1998). Pancreatic tumors are also observed in the offspring of pregnant rats treated with NNK, and this effect is markedly enhanced by ethanol (Schüller et al., 1993).
4-Aminobiphenyl and 2-naphthylamine are known human bladder carcinogens (International Agency for Research on Cancer, 1972, 1974b). Both are present in cigarette smoke. Hemoglobin adducts of 4-aminobiphenyl and other aromatic amines are associated with bladder cancer induction in smokers (Castelao et al., 2001). The evidence is strong that aromatic amines play a significant role as causes of bladder cancer in smokers (Vineis et al., 2001).
Cigarette smoke contains free radicals and induces oxidative damage (Arora et al., 2001; Pryor, 1997). The gas phase of freshly generated cigarette smoke has large amounts of nitric oxide and other unstable oxidants (Hecht, 1999). The particulate phase is postulated to contain long-lived radicals that may undergo quinone-hydroquinone redox cycling (Pryor, 1997). The presence of such free radicals and oxidants can lead to oxidative DNA damage. However, the role of oxidative damage in cancer induced by cigarette smoke is unclear.
Formation and repair of DNA adducts
Carcinogens form the link between nicotine addiction and lung cancer (Figure 1) (Hecht, 1999). Nicotine is the reason people continue to smoke in spite of the well-known adverse health effects. Nicotine is not a carcinogen. However, the cigarette is a disastrous nicotine delivery device because the carcinogens discussed above accompany nicotine in each puff. Although the dose of each carcinogen per cigarette is quite small, the cumulative dose in a lifetime of smoking can be considerable.
The response of the organism to carcinogen exposure is similar to that for any other foreign compound or drug. Cytochrome P450 enzymes catalyze addition of an oxygen atom to the carcinogen, increasing its water solubility and converting it to a form that is more readily excretable (Guengerich, 2001). This ‘metabolic detoxification’ process is further assisted by phase 2 enzymes, which convert the oxygenated carcinogen to a form that is highly soluble in water (Armstrong, 1997; Burchell et al., 1997; Duffel, 1997). To the extent that this process is efficient, the organism will be protected. However, some of the intermediates formed by the interaction of cytochrome P450 enzymes with carcinogens are in fact quite reactive, generally possessing an electrophilic (electron-deficient) center. Such intermediates or metabolites can react with DNA, resulting in the formation of DNA adducts. This process which converts an unreactive carcinogen to a form that binds to DNA is known as metabolic activation (Miller, 1994). The balance between metabolic activation and detoxification varies among individuals and is likely to affect cancer risk because DNA adducts are central to the carcinogenic process (Tang et al., 2001; Hecht, 1999). Most cigarette smoke carcinogens require metabolic activation.
Elaborate DNA repair systems have evolved to eliminate DNA adducts from the genome (Hoeijmakers, 2001). For instance, the nucleotide excision repair pathway eliminates DNA adducts consisting predominantly of base-attached larger chemical groups (so-called bulky DNA adducts), as well as intra- and interstrand DNA crosslinks. Adducts of PAH are repaired by nucleotide excision repair. The base excision repair systems are more geared towards removing defective DNA bases characterized by attachment of small chemical groups or bases fragmented by ionizing radiation or chemical oxidation. A specialized direct repair system acts through the enzyme O6-methylguanine DNA methyltransferase. This pathway is important for repair of the miscoding methylated base O6-methylguanine. Effective repair processes should result in the rapid elimination and consistent reduction of the cellular levels of DNA adducts. If unrepaired damage is still present during DNA replication, it may either cause replicative DNA polymerases to stop at the site of a lesion (resulting in arrest of DNA replication, and cell death, or chromosomal aberrations). Alternatively, the polymerases may bypass the altered base, with the possibility of base misincorporation. In fact, this problem is so serious that all cells have evolved specialized DNA polymerases that are able to bypass various types of DNA damage (Livneh, 2001). In some cases, these lesion bypass polymerases are endowed with the property of being able to correctly bypass specific types of lesions. An example is DNA polymerase eta (the product of the XPV gene), which correctly bypasses thymine-thymine dimers (Johnson et al., 1999; Masutani et al., 1999). The importance of these polymerases in mutation avoidance is exemplified in the human genetic disorder xeroderma pigmentosum variant (XPV) in which the XPV gene is defective, resulting in a greatly increased risk of developing sunlight-associated skin cancers. Human DNA polymerase eta can bypass a template containing a (+)-trans-anti-benzo[a]pyrene-N2-dG adduct (BPDE-N2-dG), derived from BaP, and predominantly incorporates an adenine. This specificity of nucleotide incorporation correlates well with the known mutation spectrum of BPDE-N2-dG lesions in mammalian cells (Zhang et al., 2000).
If DNA adducts are bypassed incorrectly by a DNA polymerase, mutations may arise (Figure 1). The resulting alterations may lead to creation of a new phenotype and, if growth controlling genes are involved, to cellular transformation and the development of tumors. Protooncogenes and tumor suppressor genes could be critical targets for carcinogens (Pfeifer and Denissenko, 1998; Hussain and Harris, 1998). In smokers, there is a chronic barrage of metabolically activated carcinogens which cause these multiple changes (Figure 1). This constant assault on genes is completely consistent with genetic derangements leading to six proposed hallmarks of cancer: self-sufficiency in growth signals, insensitivity to anti-growth signals, evasion of apoptosis, tissue invasion and metastasis, sustained angiogenesis, and limitless replicative potential (Hanahan and Weinberg, 2000).
Identification of a link between DNA damage and specific mutations in tumor cells would strengthen the understanding to what extent elements of the environment are responsible for tumorigenesis in humans. This idea is based on the knowledge that DNA adducts induced by different mutagens may have significantly different mutational properties. Adducts may form in a base and sequence-specific context. For example, a particular adduct may induce predominantly G to C transversions within a particular sequence context, and if such mutations were found in one type of human tumor, such a carcinogen would become a suspect, in particular if there is epidemiological evidence that exposure to this agent may be involved in causing this type of cancer.
DNA adducts: structure, detection, and mutation induction
Table 1 summarizes information on structures of some DNA adducts that are formed from representative tobacco smoke carcinogens and related compounds. Available data on the detection of these adducts in lung DNA from smokers and some likely mutations that may result from their presence are also summarized.
Bay region diol epoxides are among the principal PAH metabolites involved in DNA adduct formation (Szeliga and Dipple, 1998; Conney, 1982). Each diol epoxide metabolite has four isomeric forms. There are two diastereomers in which the benzylic-OH of the diol is either on the opposite side (anti-) or the same side (syn-) of the molecule as the epoxide ring. Each of these diastereomers exists as a pair of enantiomers. Each of the four diol epoxides reacts with DNA to differing extents, either at the exocyclic amino group of deoxyguanosine (N2-) or deoxyadenosine (N6-) (Szeliga and Dipple, 1998). Each reaction of the exocyclic amino group results in either trans- or cis-ring opening of the epoxide ring. Therefore, there are 16 possible adducts of this type from each PAH diol epoxide metabolite (Szeliga and Dipple, 1998). All adducts have been thoroughly characerized for multiple PAH molecules. Table 1 illustrates only one adduct from each of two PAH in tobacco smoke: BaP and 5-methylchrysene (5-MeC). This is quantitatively the major one in each case, but it should be noted that 15 other adducts are formed from BaP-7,8-diol-9,10-epoxide and from 5-MeC-1,2-diol-3,4-epoxide. In addition, another set of 16 adducts can be formed from 5-MeC-7,8-diol-9,10-epoxide (Melikian et al., 1988). The diol epoxide pathway is not the only mechanism for adduct formation from PAH. Depurinating adducts have been detected as a result of one electron oxidation, and adducts resulting from quinone formation have also been characterized (Casale et al., 2001; Penning et al., 1999). Adducts are also formed via 9-hydroxy-BaP-4,5-oxide (Ross and Nesnow, 1999).
N-nitrosamines are metabolized to intermediates that alkylate various positions of the DNA bases (Preussmann and Stewart, 1984). The most thoroughly investigated are 7-alkylguanines and O6-alkylguanines, shown for N-nitrosodimethylamine and NNK in Table 1. Other products include N-1, N-3, and N2-deoxyguanosines, N-1, N-7, N-3, and N6-deoxyadenosines, O2-, O4, and N-3 thymidines, O2-, N-3, and N-4 deoxycytidines, and phosphotriesters (Singer and Grunberger, 1983). NNK and NNN are metabolized to intermediates that pyridyloxobutylate deoxyguanosine (Hecht, 1998). N-nitrosopyrrolidine, a cyclic N-nitrosamine, displays alkylation chemistry that is somewhat different from that of acyclic nitrosamines because the alkylating intermediate is tethered to an aldehyde. A complex mixture of deoxyguanosine adducts is produced, among which the N2-tetrahydrofuranyl structure predominates (Wang et al., 2001b).
Ethylene oxide behaves like a typical alkylating agent, reacting primarily at N-7 of deoxyguanosine, as shown in Table 1, but also at other positions (Zhao et al., 1999). 1,3-Butadiene is metabolized to 3,4-epoxy-1-butene, 3,4-epoxy-1,2-butanediol, and 1,2,3,4-diepoxybutane (Zhao et al., 2000; Koivisto et al., 1999; Koc et al., 1999; Tretyakova et al., 1997). Most of the DNA adducts arise from the reactions of the diol epoxide at the N-7 position of guanine, N-3 adenine, and N6-adenine (Table 1). Multiple stereoisomers are formed. Acetaldehyde reacts with the exocyclic amino group of deoxyguanosine to give a Schiff base as the major adduct (Wang et al., 2000a). Several other adducts including a G-G crosslink, shown in Table 1, have also been identified. Crotonaldehyde produces cyclic 1,N2-deoxyguanosine adducts by Michael addition and Schiff base adducts by reaction of the aldehyde group with the exocyclic amino group of deoxyguanosine (Wang et al., 2001a; Chung et al., 1999). Other adducts are formed, with multiple stereoisomers, by dimers of 3-hydroxybutanal, produced by hydration of crotonaldehyde (Wang et al., 2000b).
Aromatic amines such as 4-aminobiphenyl and heterocyclic aromatic amines react with DNA mainly at C-8 of deoxyguanosine via their N-hydroxy-metabolites (Delclos and Kadlubar, 1997). Adducts have also been observed at N2- of deoxyguanosine, O6- of deoxyguanosine, and N6-of deoxyadenosine.
Vinyl chloride is metabolized to chloroethylene oxide which reacts with DNA giving 7-oxoethyldeoxyguanosine as a major product along with ‘etheno’ adducts such as 3,N2-ethenodeoxyguanosine and 1,N6-ethenodeoxyadenosine, as shown in Table 1 (Swenberg et al., 1999; Nair et al., 1999). Ethyl carbamate is metabolized to vinyl carbamate which similarly reacts giving 1,N6-ethenodeoxyadenosine (Guengerich and Kim, 1991).
2-Nitropropane is metabolized to intermediates that aminate deoxyguanosine at the C-8 and N2-positions (Sodum and Fiala, 1998). Radical oxidants in cigarette smoke are believed to give rise to 8-oxodeoxyguanosine while nitric oxide yields deoxyoxanosine (shown in Table 1) along with other products (Burney et al., 1999; Asami et al., 1997). Recent studies demonstrate that 8-oxodeoxyguanosine is further oxidized by peroxynitrite to a variety of products including those illustrated in Table 1 (Henderson et al., 2002).
Table 1 summarizes information on the detection of specific adducts in lung DNA from smokers. The BaP adduct shown in Table 1, BPDE-N2-dG, has been the subject of numerous studies. Convincing evidence, obtained by HPLC with fluorescence detection of BaP tetraols released upon acid hydrolysis, clearly demonstrates the presence of BPDE-N2-dG in some samples of human pulmonary DNA (Rojas et al., 1998; Kriek et al., 1998). Methods such as 32P-postlabeling and immunoassay have reported the presence of ‘PAH-DNA adducts’ or ‘aromatic DNA adducts’ in human lung, but these are mainly uncharacterized (Santella, 1999; Kriek et al., 1998). It is not certain that they are in fact derived from PAH. Gupta et al. (1999) have reported that adducts detected by 32P-postlabeling in lung DNA of cigarette smoke-exposed rats were endogenous adducts enhanced by cigarette smoke. No other specific PAH adducts have been detected with certainty in human lung.
7-Methylguanine and 7-hydroxyethylguanine have been detected in human lung DNA by 32P-postlabelling (Zhao et al., 1999). Levels of 7-methylguanine are higher in smokers than in nonsmokers in some but not all studies (Hecht and Tricker, 1999). Evidence has also been presented for O6-methyl- and O6-ethyldeoxyguanosine in human lung DNA, but confirmation by other methods is lacking (Wilson et al., 1989). GC-MS analysis of released 4-hydroxy-1-(3-pyridyl)-1-butanone (HPB) establishes the presence of NNK or NNN-derived pyridyloxobutyl DNA adducts in human lung (Foiles et al., 1991). These adducts, which are produced mainly by reaction with deoxyguanosine, are higher in lung DNA from smokers than nonsmokers, as expected based on the specificity of NNK and NNN to tobacco products.
GC-MS analysis of 4-aminobiphenyl released from human lung DNA provides evidence in support of the presence of the C-8 adduct, as also indicated by 32P-postlabeling and immunoassay (Culp et al., 1997; Lin et al., 1994). Levels of this adduct were not related to smoking. One study demonstrated the presence of 8-oxodeoxyguanosine in human lung DNA using HPLC with electrochemical detection (Asami et al., 1997). Levels were higher in smokers than in nonsmokers.
Collectively, the available data provide convincing evidence for the presence of certain adducts listed in Table 1 in human lung DNA. It is very likely that many of the other adducts are also present, but the available methodology is not sensitive or specific enough to detect them, or has not been applied yet.
Table 1 summarizes data obtained mainly, although not exclusively, from site-specific mutagenesis studies. The major adduct of BaP illustrated in Table 1 produces GC→TA mutations (Kozack et al., 2000; Seo et al., 2000). In site-specific mutagenesis studies, this event was sequence dependent. This adduct induced >95% G→T mutations in one sequence context (5′-TGC) and approximately 95% G→A mutations in another context (5′-AGA). This may results from conformational complexities (Kozack et al., 2000; Seo et al., 2000). In chromosomal genes, racemic benzo[a]pyrene-7,8-diol-9,10-epoxide (BPDE) produces predominantly G to T mutations (Eisenstadt et al., 1982; Mazur and Glickman, 1988; Chen et al., 1990; Wei et al., 1993; Ruggeri et al., 1993; Yoon et al., 2001). Some PAH diol epoxides such as those derived from benzo[c]phenanthrene react extensively at deoxyadenosine in DNA and consequently produce significant levels of A mutations (Szeliga and Dipple, 1998; Bigger et al., 1992).
7-Alkyldeoxyguanosines such as those derived from N-nitrosodimethylamine, NNK, ethylene oxide, 1,3-butadiene, and vinyl chloride readily depurinate giving rise to abasic sites. Replication past abasic sites results predominantly in GC→TA mutations (Kunkel, 1984). Therefore, these cigarette smoke constituents can be expected to produce GC→TA mutations in pulmonary DNA. However, apurinic sites may be repaired rapidly via the base excision repair pathway. Pyridyloxobutylation of DNA gives both GC→TA and GC→AT mutations, as demonstrated by analysis of ras mutations in lung DNA of mice treated with a model pyridyloxobutylating compound, NNKOAc (Ronai et al., 1993).
Site specific mutagenesis experiments in human embryonic kidney cells have shown that the mispairing characteristics of O6-pyridyloxobutyldeoxyguanosine are comparable to that of O6-methyldeoxyguanosine, with a high number of G→A transitions and smaller amounts of G→T transversions observed (Pauly et al., 2002). GC→TA mutations are also the predominant ones observed in studies of mutagenesis by N2-deoxyguanosine adducts of 1,3-butadiene (Carmical et al., 2000), adducts of 4-aminobiphenyl (Melchior et al., 1994), and 8-oxodeoxyguanosine (Moriya, 1993). Other studies of butadiene mutagenesis demonstrate the occurrence of GC→AT, AT→TA, and GC→TA mutations (Recio et al., 2000). Products of the further oxidation of 8-oxodeoxyguanosine by peroxynitrite are highly efficient in producing GC→TA mutations (Henderson et al., 2002). Collectively, the available data indicate that many DNA adducts associated with cigarette smoke exposure may produce GC→TA mutations.
The spectrum of p53 mutations in smoking associated lung cancers – clues to etiology?
Mutations in the p53 gene are a common occurrence in human tumors and are found in approximately 40% of human lung cancers. p53 mutations are generally more common in smokers than in nonsmokers (Greenblatt et al., 1994; Hernandez-Boussard and Hainaut, 1998). Initial studies demonstrated that p53 mutations in lung cancer are different from those in other cancers and that an excess of G to T transversions is characteristic for these tumors (Hollstein et al., 1991). G to T transversions have been described as a molecular signature of tobacco smoke mutagens in smoking-associated lung cancers (Greenblatt et al., 1994; Hainaut and Hollstein, 2000). It was shown that there is an increased frequency of G to T transversions in lung cancers from smokers compared to lung cancers from nonsmokers and compared to most other cancers (reviewed in Greenblatt et al., 1994; Husgafvel-Pursiainen and Kannio, 1996; Hernandez-Boussard and Hainaut, 1998; Bennett et al., 1999; Hainaut and Pfeifer, 2001; Vähäkangas et al., 2001). However, such conclusions have been questioned by two recent reports (Rodin and Rodin, 2000; Paschke, 2000). These reports have prompted us to re-analyse this important issue using the currently available literature (Hainaut and Pfeifer, 2001; Hainaut et al., 2001). These new analyses confirmed and extended the existence of a specific mutation pattern in lung cancers of smokers, and indicate that the recent reports by Rodin and Rodin (2000) and Paschke (2000) are in error.
Figure 2 shows that the mutational spectra in lung cancers from smokers and nonsmokers are clearly different. Twelve per cent of the p53 mutations in nonsmokers are G to T transversions. This figure includes the mutation events scored as G to T and as C to A, since the latter correspond to G to T changes occurring on the non-coding, transcribed strand of genomic DNA (by convention, the base changes induced by a mutation are read on the coding, non-transcribed strand). The difference between 12% G to T mutations in nonsmokers and 30% G to T in smokers is statistically highly significant (P<0.001; Chi square test). It is also important to note that the frequency of G to T transversions is much higher in lung cancers than it is in any other tumor type except for liver cancers associated with geographic areas where contamination of food with aflatoxins has been demonstrated. In most internal cancers, not strongly linked to tobacco consumption, such as brain, colorectal, and breast cancers, the frequency of G to T mutations is between 8 and 10 per cent (Figure 2). This is quite similar to the percentage of G to T mutations found in nonsmokers. Nonsmokers have an increased level of G to A transitions (47% as opposed to smokers 29%), a difference that is also statistically highly significant. In Figure 2, we include categories of both ‘designated smokers’ (where the smoking status is indicated in the literature) and ‘all lung cancer cases minus nonsmokers.’ This is based on the knowledge that 90% or more of all lung cancers occur in smokers (Proctor, 2001). As expected, the proportion of G to T transversions in all lung cancers (minus nonsmokers) is remarkably similar to that observed in designated smokers (Figure 2).
In recent years there has been a rise in the proportion of lung cancers that are of adenocarcinoma histological type. This trend has been attributed to changes in cigarette design which may result in deeper inhalation as smokers compensate for lower levels of nicotine. In addition, levels of NNK in cigarette smoke have increased while those of BaP have decreased. In laboratory animals, NNK induces adenocarcinoma of the lung whereas BaP generally produces squamous cell tumors (Hoffmann et al., 2001).
To address the issue of whether the different histological types of lung cancer show differences in their p53 mutational spectra, we have analysed the IARC TP53 mutation database separately for these tumors (Figure 3). The frequencies of G to T transversions in the p53 database were 31% in adenocarcinomas, 28% in squamous cell carcinomas, 26% in small cell lung cancers, and 34% in large cell carcinomas. Thus, these different lung cancers exhibit similar p53 mutation spectra.
The type of base changes seen along the entire p53 coding sequence is very different in lung cancer compared to other cancers. p53 mutations do not occur in a random fashion along the coding sequence but are typically clustered at so-called mutation hotspots. All these hotspots are within the DNA binding domain of the p53 protein spanning approximately 180 amino acids from codon 120 to 300. Figure 4 shows the distribution of all mutations (upper panels) and, specifically that of G to T transversions (lower panels), along the p53 gene in lung cancer (Figure 4a) and in brain/breast/colorectal cancers (Figure 4b). Different hotspots of G to T mutations are observed in brain/breast/colon compared to lung. These hotspot codons are of particular interest since they may allow a more specific assignment to a particular carcinogen if that site is preferentially damaged or mutated by the compounds in question. However, hotspot codons may exist solely as a consequence of preferential phenotypic selection in tumors. To address this issue, we have compared the mutational events in different types of cancers at a number of common hotspot codons.
Figure 5 shows that the major lung cancer hotspots 158, 245, 248, and 273 are commonly G to T transversions in lung cancer but are generally other mutation types (almost exclusively G to A) in internal tumors not associated with smoking. This does not agree with the preferential selection theory put forward by Rodin and Rodin (2000). In this theory, cigarette smoke acts as a physiological stress to expand a population of cells harboring endogenous mutations. What also does not agree with the Rodin model is that tumor promoters in cigarette smoke, which should act on pre-existing endogenous mutations, do not by themselves cause cancer but require initiators (Rubin 2001).
In Figure 5, we present data for five lung cancer p53 mutational hotspots, codons 157, 158, 245, 248, and 273. The tumor types analysed are lung, breast, colon, and brain. These tumors all have substantial numbers of total mutations in the p53 database (between 750 and 1600). Codons 157 and 158 are common mutation sites in lung cancers (mostly G to T) but are much less frequently mutated in other tumors (see Figure 4). These two codons may be considered as mutational hotspots specific for lung cancers of smokers. As rightly pointed out by Rodin and Rodin (2000), the mutagenic spectrum of codon 157 is quite limited since G to T might be the most common substitution that would result in a p53 protein with a mutant residue at codon 157. This fact is indeed sufficient to explain why codon 157 is more often mutated in lung than in other cancers, since the chances of formation of a G to T mutation at that codon are much greater in lung cancers than in cancers less related to tobacco exposure (see below).
The fact that codons 248 and 273 are often the targets of G to T transversions in lung cancers is a perfect illustration of the respective roles of mutagenesis and selection in shaping the mutation spectrum of p53. These two codons are the most commonly mutated ones in the entire p53 mutation database, and this is true in almost every type of human tumor. Mutation at these residues probably has a drastic effect on p53 function, since the two residues form contacts between the p53 protein and its DNA target (Walker et al., 1999). Mutating these residues abrogates the capacity of p53 to act as a transcription factor to activate certain downstream genes such as p21 or BAX. Importantly, in lung cancers, the mutations at these codons differ dramatically from those in cancers not linked to tobacco smoking (Figure 5). At codons 248 and 273, 35–45% of the mutations are G to T transversions in lung cancer, but this type of mutation is virtually absent in the other tumors. It should be noted that in 1994, Soussi and colleagues have reported that there were no substantial differences between the in vitro functional properties of G to T and G to A mutants at codons 248 and 273 (Ory et al., 1994).
These observations provide direct evidence that p53 mutations in lung cancer occur by a distinct mechanism, and cannot be explained simply by selection.
The distribution of G to T transversions in lung cancer is consistent with the adduct spectrum and precise mutational specificity of PAH compounds
As indicated in the previous section, a characteristic mutational fingerprint in smoking-associated lung cancers is the high frequency of G to T transversions. PAHs are one class of carcinogens in tobacco smoke that produce predominantly this type of mutation in various experimental systems (Eisenstadt et al., 1982; Mazur and Glickman, 1988; Chen et al., 1990; Wei et al., 1993; Ruggeri et al., 1993; Yoon et al., 2001). As shown in Table 1, there are a number of other types of DNA adducts derived from agents in tobacco smoke that also can give rise to G to T transversions. These include apurinic sites resulting from depurination of 7-alkylguanines, pyridyloxobutylated DNA, aromatic amine-DNA adducts, and 8-oxodeoxyguanosine and related products of oxidative damage.
The lung cancer p53 spectrum is consistent with the mutational patterns induced by certain PAHs. The distribution of benzo[a]pyrene-7,8-diol-9,10-epoxide (BPDE) and other PAH diol epoxide adducts was mapped at nucleotide resolution along exons of the p53 gene in PAH-treated normal human bronchial epithelial cells (Denissenko et al., 1996; Smith et al., 2000). Frequent adduct formation occurred at guanine positions in codons 157, 158, 245, 248, and 273. These data are schematically summarized in Figure 6. These same positions of preferential PAH adduct formation are major mutational hotspots in human lung cancers from smokers (see Figure 4a).
The distribution of BPDE-N2-dG within p53 exon five was analysed using stable isotope labeling liquid chromatography- electrospray ionization tandem mass spectrometry (HPLC-ESI-MS/MS) (Tretyakova et al., unpublished results). In this approach, specific guanine nucleobases within p53 gene sequences are labeled with 15N, so that the BPDE adducts originating from these positions can be distinguished from the lesions formed at other sites. An excellent agreement with the data obtained by the UvrABC incision method was observed. Although the number of adducts did not directly correlate with mutations (for example, the highest number of adducts was produced at codon 156, which is not a hotspot for mutations in lung cancer), these data did demonstrate an increased formation of BPDE adducts at codons 156, 157, and 158. Other factors, e.g. DNA sequence effects on repair or mispairing, and lack of selection at codon 156 (there is only a single G to T transversion reported at codon 156 in the entire p53 database of more than 16 000 mutations), may explain the predominance of mutations at codons 157 and 158. Alternatively, these mutations may originate from DNA modification by other tobacco carcinogens, in particular other PAH.
The mechanistic basis for the selective occurrence of such adduct hotspots in p53 gene is the enhancement of adduct formation by 5-methylcytosine bases present at CpG dinucleotide sequences (Denissenko et al., 1997; Chen et al., 1998; Weisenberger and Romano, 1999; Das et al., 1999; Pfeifer, 2000). All CpG sequences in the p53 coding exons five through nine are completely methylated in all tissues examined including the lung (Tornaletti and Pfeifer, 1995). Methylation at CpG sites may increase the binding of planar carcinogen compounds at the intercalation step (Geacintov, 1986), although the precise mechanism by which cytosine methylation at CpG sites enhances carcinogen binding still needs to be determined. Our preliminary data indicate that the increased reactivity of guanines in methylated CpG runs is primarily due to the 5-methyl group at the base-paired cytosine (Matter et al., unpublished results). It has been shown that the preferential formation of BPDE adducts at methylated CpG sites is reflected in strongly enhanced mutagenesis at CpG sequences after treatment of cells with BPDE. This was demonstrated with three different CpG-methylated mutational reporter genes including two chromosomal genes that contain methylated CpG sequences (Yoon et al., 2001). The striking sequence specificity of BPDE for producing G to T transversion hotspots at methylated CpG sequences is similar to the distribution of G to T transversion hotspots in smoking-associated lung tumors (see Figure 4). In the p53 gene of lung cancer, five major G→T mutational sites (in codons 157, 158, 245, 248, and 273) consist of methylated CpGs (Yoon et al., 2001).
It has been argued that methylated CpG sites are preferentially modified and mutated by a range of different carcinogens including aromatic amines and aflatoxins (Chen et al., 1998; Hecht, 1999), however, the exact range of such compounds targeting methylated CpGs is not known. We have not seen preferential mutagenesis at methylated CpGs by the aromatic amine 4-aminobiphenyl (Besarati Nia et al., 2002b).
Two recent studies lend support to the PAH – p53 - lung cancer connection. First, using a very sensitive assay to detect mutations in the absence of selection, Hussain et al. (2001) have shown that exposure of bronchial epithelial cells to BPDE produces G to T transversions in the p53 gene at lung cancer hotspot codons 157, 248 and 249. Moreover, non tumorous lung tissues from smokers with lung cancer carry a high p53 mutational load at these codons, even when another mutation is present in the tumor itself. This observation provides strong support to the idea that tobacco carcinogens can directly induce G to T transversions in exposed bronchial cells.
Second, Demarini et al. (2001) determined p53 and KRAS mutations in lung tumors from 24 nonsmoking Chinese women whose tumors were associated with exposure to smoky coal in unventilated homes. These women have high levels of exposure to various PAHs. The tumors showed a high percentage of mutations that were G to T transversions at either KRAS (86%) or p53 (76%). The mutations clustered at the CpG-rich codons 153 through 158 of the p53 gene, and at codons 249 and 273 and had 100% of the guanines of the G to T transversions on the nontranscribed strand. This mutation spectrum is consistent with an exposure to PAHs (Smith et al., 2000; see also Figure 6), which are present in substantial concentrations in smoky coal emissions.
The remarkable site specificity of mutagenesis by PAH compounds strongly suggests that targeted adduct formation in addition to phenotypic selection is responsible for shaping the p53 mutational spectrum in lung tumors. Furthermore, it is important to note that the vast majority (90%) of G to T transversions in lung cancers are targeted to guanines on the non-transcribed DNA strand (IARC p53 mutation database). This is not the case for other mutation types, as for example G to A transitions at CpG sites occur at the same frequency on both strands. This observation implies that, in the case of G to T transversions, a strand-specific DNA repair process plays a role in the preferential repair of DNA lesions occurring on the transcribed strand. DNA repair experiments analysing BPDE adducts in the p53 gene have shown that the nontranscribed strand is indeed repaired more slowly than the transcribed strand (Denissenko et al., 1998). These findings support the proposal that both the initial DNA adduct levels and a strand bias in repair contribute to the mutational spectrum of the human p53 gene in lung cancer.
The spectrum of p53 mutations in other smoking associated cancers
In addition to lung cancers, several common neoplasms are strongly associated with tobacco use. This is the case for squamous cell carcinomas of the oral cavity, larynx and esophagus, and for cancers of the bladder (both squamous cell carcinomas and transitional cell carcinomas). It is important to note that these cancers occur at variable incidences in different regions of the world, and that not all of these cancers are a direct consequence of tobacco use. We have carried out an analysis of the p53 mutation spectra in tumors, where there is strong evidence that tobacco smoking is an important factor of risk (Figure 7).
Of the four cancer types, cancer of the larynx shows the strongest similarities with lung cancers, with a high prevalence of G to T transversions (27%), many of them occurring at PAH-target codons (157,245). Strikingly, this is not the case for cancers at other organ sites. In oral and esophageal cancers, the p53 mutation load has been shown to be proportional to the extent of tobacco consumption, with an almost fourfold increase in mutation prevalence in heavy smokers as compared to nonsmokers (Brennan et al., 1995). In both cancers, however, the prevalence of G to T transversions is only slightly higher than in cancers not strongly related to tobacco smoke (breast, colorectal and brain cancers) and this difference is of borderline statistical significance. A recent study notes the similarity in mutational spectrum induced by acetaldehyde in the HPRT gene of human T lymphocytes to that of the p53 gene in esophageal cancers (Noori and Hou, 2001). Moreover, these transversions do not preferentially occur at PAH-target codons. The patterns of mutations in both oral and esophageal cancers are extremely heterogeneous, in agreement with epidemiological data showing that multiple factors may act in conjunction in the pathogenesis of these cancers (including in particular tobacco and alcohol). Given the heterogeneity of the mutation patterns, the data available at present do not allow the unambiguous identification of molecular signatures of tobacco carcinogens in the p53 mutational spectrum of oral and esophageal cancers.
In the case of bladder cancers, the mutation pattern shows an unusually high prevalence of G to A transitions at non CpG sites. These mutations are not distributed at random, and bladder-specific mutation hotspots have been reported at codons 280 and 285. Both of these codons occur within the same primary sequence context (5′AGAG), raising the possibility that this sequence is a preferential binding site for an exogenous agent involved in bladder carcinogenesis. It has been postulated that these mutations may represent a fingerprint of aromatic amine which are the most potent class of bladder carcinogens in tobacco smoke. Indeed, codo 285 is a preferential binding site for N-hydroxy-4-amino=biphenyl (Feng et al., 2002). However, aromatic amines produce mainly G to T mutations Besarati Nia et al., 2002).
The effect of individual exogenous agents in tobacco carcinogenesis is difficult to assess at the molecular level because there is chronic exposure to a complex mixture of carcinogens, tumor promoters, and co-carcinogens. In the case of lung cancers, however, there is strong evidence for the involvement of tobacco smoke compounds at all steps of this chain of events. First, tobacco smoke compounds are absorbed and metabolized in smokers. Second, many of these compounds produce DNA adducts in smokers’ lungs. Third, there is evidence that experimental exposure to metabolites of PAHs, one group of tobacco smoke carcinogens, can induce the same type of adducts in cultured, normal bronchial cells. Fourth, G to T transversion mutations, consistent with the observed type of adduct damage are detectable in both lung cancers and in adjacent, non-involved lung tissues in smokers. The position of these mutations often coincides with that of adducts detected in vitro. Fifth, the overall prevalence of p53 mutations is lower in lung cancers of nonsmokers than in smokers, and this is particularly true for G to T mutations. Sixth, even at codons that are common hotspots in all types of cancers, there is an excess of G to T transversions in lung cancers of smokers as compared to nonsmokers or to cancers not directly related to tobacco use. Seventh, methylated CpG dinucleotides are the preferred sites for G to T transversions, accounting for more than 50% of such mutations in lung tumors. The same dinucleotide, when present within mutational reporter genes, is the target of G to T transversion hotspots in cells exposed to benzo[a]pyrene-7,8-diol-9,10-epoxide.
The fact that an apparent molecular signature of PAHs is found in lung cancers does not rule out an important role for other tobacco components in carcinogenesis. First, in lung cancers, other tobacco carcinogens may produce a similar molecular signature, or may be responsible for non-G→T mutations. Second, different tobacco smoke compounds may exert carcinogenic, co-carcinogenic, or tumor promoting effects in an organ- and tissue specific manner, depending on the rate of accumulation and metabolism at various sites in the body. It is, however, interesting to note that, in the upper respiratory tract, there is a gradient in the prevalence of p53 G to T transversions in cancers of smokers, from low in the oral cavity, to intermediate in the larynx and high in various histological types of lung cancers. This situation may reflect the existence of an underlying, parallel gradient in the extent of exposure of cells of the respiratory tract to PAHs and other tobacco smoke carcinogens. The studies on PAH-induced DNA damage and p53 mutations provide a compelling link between a group of exogenous carcinogens and human cancer. In the future, similar approaches on other tumors and other mutagens are expected to reveal further clues for the role of environmental mutagens in human carcinogenesis.
The work of the authors was supported by grants from the National Institutes of Health (CA84469 to GP Pfeifer; CA-81301 and DA 13333 to SS Hecht). The IARC p53 database is supported by a grant from the European Community (QLG-1999-00273).