Applying Tobacco, Environmental, and Dietary-Related Biomarkers to Understand Cancer Etiology and Evaluate Prevention Strategies

Environmental and lifestyle factors contribute to the majority of human cancers (1, 2). It is critical to understand the etiologic factors involved so that effective preventive strategies can be implemented. Molecular biomarkers can help identify these etiologic factors and individuals at increased risk of cancer. Biomarkers also provide critical evidence required for policies to protect human health.

Our team has a long history of mechanistic research on chemical carcinogenesis and chemoprevention. We have developed a variety of biomarkers to assess the carcinogenic potential of environmental exposures in humans (3–10). Our approaches are based on understanding the biochemistry of environmentally induced carcinogenesis. The primary paradigm involves covalent interaction of reactive chemicals or metabolites with DNA to form DNA adducts (Fig. 1). If not repaired, DNA adducts may miscode during DNA replication, leading to somatic mutations. When these mutations occur in genes responsible for cell-cycle control and maintenance, loss of normal function leads to abnormal cell growth and cancer. Environmental and lifestyle exposures can also modulate cancer risk through other mechanisms such as immune system suppression, receptor activity modulation, DNA repair inhibition, epigenome alterations, oxidative stress and/or inflammation induction, alterations in cell turnover rate or cell death, and/or nutrient supply disturbances (11).

Different types of biomarkers are employed to study these processes. Exposure biomarkers can provide a more reliable determination of the biologically relevant dose than external measures such as questionnaires, medical records, or environmental exposures. Many exposure biomarkers are metabolites of harmful chemicals and provide a measure of internal exposure to reactive metabolites initiating the carcinogenic process. Biomarkers of intermediate effects such as DNA adducts are measures of both exposure and risk since they form at the time of exposure and are directly relevant to the carcinogenic process. Similarly, biomarkers of inflammation are measures of exposure to inflammatory agents and disease risk.

The biomarkers developed by our team have provided critical data on internal exposure to toxic and genotoxic chemicals and their linkage to cancer risk. Furthermore, we are combining our exposure biomarkers with DNA sequencing and other biological information to obtain a greater understanding of the biological factors that affect an individual's susceptibility to the harmful effects of environmental exposures. These biomarkers are also employed to determine the effectiveness of exposure intervention and chemoprevention protocols. Our research team measures some of these key intermediate biomarkers of cancer risk in target tissues, taking advantage of technological advances in mass spectrometry (MS) and other techniques. Specific examples of the application of these analytic methods are described in the sections below and represent different stages of implementing these biomarkers to understanding cancer risk and prevention in humans.

The relative risk of lung cancer in smokers is primarily a function of smoking duration (years) and cigarettes per day (CPD). However, disease risk varies among different racial/ethnic groups; the differences are most pronounced for light (<10 CPD) or moderate smokers (11-20 CPD; refs. 12–14). Among light smokers in the Multiethnic Cohort (MEC) Study, African Americans (AA) have a 2-fold higher risk of lung cancer than Whites, and Japanese Americans (JA) have about a 50% lower risk. Nicotine is the primary addictive agent in tobacco products and the desire to consume more nicotine leads to increased smoking. By quantifying biomarkers of tobacco smoke exposure in 2,239 MEC smokers, we investigated the hypothesis that differences in nicotine consumption due to variable smoking intensity contribute to the observed differences in lung cancer risk across racial/ethnic groups (15–17).

Reported CPD is a notoriously poor measure of smoking dose (18–22). A better measure of exposure is total nicotine equivalents (TNE), the sum of the urinary concentration of total nicotine, total cotinine, total 3′-hydroxycotinine, and nicotine N-oxide (23) where “total” refers to the sum of the compound and its glucuronide conjugate. In MEC smokers, we found that TNE levels were significantly greater in AA as compared with Whites and significantly lower in JA (15). In contrast to reported CPD, these data are consistent with the relative lung cancer risk among these groups. The urinary concentration of total 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol (NNAL), a biomarker of the tobacco specific lung carcinogen, 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NNK), paralleled that of TNE (14, 16, 24). In Shanghai and Singapore cohorts, total cotinine and NNAL are significantly associated with lung cancer risk after adjustment for self-reported smoking history (25–27).

One factor driving smoking dose is the relative activity of CYP2A6, the primary catalyst of nicotine metabolism. For MEC smokers with biomarker data (n = 2,115), we determined CYP2A6 copy number variants (CNV) and SNPs that are well characterized for their effect on nicotine metabolism (28). These variants defined 13 haplotypes and 68 diplotypes, the frequency of which varies by ethnic group (Fig. 2A). On the basis of the functional activity of each allele, six diplotype categories were defined as normal to low activity (N/N to L/L). In JA and AA, CYP2A6 diplotypes were associated with both TNE and total NNAL (Fig. 2B). These data clearly establish a relationship between CYP2A6 and smoking dose. However, we did not find a significant association of TNE (or NNAL) with CYP2A6 diplotype in Whites. This is likely due to the lower frequency of the variant alleles in Whites, relative to AA and JA (Fig. 2A). There may also be uncharacterized CYP2A6 variants or other environmental or genetic factors that contribute to nicotine uptake and smoking dose.

CYP2A6 activity influences smoking dose and consequently, carcinogen uptake. An association of CYP2A6 genotype and lung cancer has been observed in Asian populations (24). However, a direct link between the effect of CYP2A6 on smoking dose and lung cancer has only recently been established (29). In collaboration with Jian-Min Yuan, we found that several CYP2A6 genotypes and a CNV were associated with lung cancer risk after adjustment for CPD. When the data were further adjusted for TNE, this relationship was no longer significant, confirming that the influence of CYP2A6 genotype on lung cancer risk is mediated by smoking dose (or intensity). Further evidence for this mechanism was established in a large collaborative study, which used our TNE and CYP2A6-mediated nicotine metabolism data from the MEC and GWAS data from the Transdisciplinary Research in Cancer of the Lung consortium. In this study of over 50,000 smokers, common SNPs in CYP2A6 were associated with both reduced nicotine metabolism and lung cancer risk in smokers of European ancestry (30).

Taken together, the above-described studies support the hypothesis that individual differences in CYP2A6 activity predict tobacco exposure and, therefore, lung cancer risk (Fig. 2C). In the update to the MEC lung cancer study, Stram and colleagues report that the difference in lung cancer risk among Whites, AA, and JA can be explained by the inability of CPD to accurately capture smoking dose (n = 4,993; ref. 13). When predicted CPD was replaced by TNE as the measure of smoking, there was no difference in lung cancer risk among JA, Whites, and AA (Fig. 2D). Therefore, the variable risk of lung cancer for these groups appears to be due to differences in tobacco smoke exposure. In JA, the lower mean TNE level is due to the prevalence of low activity and null CYP2A6 alleles (28). In AA, low-activity CYP2A6 also results in lower TNE; however, as a group, mean TNE levels are higher due to other unidentified factors (28).

Most lung cancer occurs in ex-smokers for whom exposure biomarker measurements are not possible. Therefore, it is important to establish the relationship between tobacco biomarkers (including DNA adducts) and genetic and environmental factors, which could then be used in ex-smokers to assess risk. We propose, as a first step, to develop a model that predicts lifetime exposure to TNE by CYP2A6 genotype. Other biomarkers and genotypes (e.g., nicotinic receptor variants) associated with increased tobacco exposure and/or lung cancer risk could be combined with CYP2A6 genotype to develop a comprehensive model of tobacco smoke exposure and cancer risk assessment.

Polycyclic aromatic hydrocarbons (PAH) are significant contributors to tobacco smoke carcinogenesis (31, 32). Many investigations have focused on benzo[a]pyrene (BaP) as a representative PAH, but there are more than 500 PAH in cigarette smoke including multiple carcinogenic PAH such as certain benzofluoranthenes, methyl chrysenes, and dibenz[a,h]anthracene (33, 34). The central role of PAH in tobacco carcinogenesis has stimulated our research on biomarkers of PAH exposure and metabolic activation, which can be applied not only in studies of tobacco carcinogenesis but also in environmental settings where PAH exposure may be related to cancer. While quantitation of parent PAH can establish potential exposure patterns, PAH metabolites are more suitable for molecular epidemiology studies aimed at determining individual susceptibility to PAH carcinogenesis.

The diol epoxide pathway of BaP metabolism results in the formation of the highly carcinogenic “bay region diol epoxide” BPDE which reacts with DNA to produce BPDE-N²-dG, the major characterized DNA adduct of BaP (Fig. 3; refs. 35–37). BPDE also reacts with H₂O to produce trans, anti-BaPT, which is excreted in the urine and can serve as a biomarker of BaP exposure plus metabolic activation. We have quantified this metabolite in human urine by gas chromatography–negative chemical ionization-tandem mass spectrometry (GC-NCI-MS/MS), after enrichment and silylation (38). Significantly higher levels were detected in 30 smokers (0.71 ± 0.64 fmol/mg creatinine) than in 30 nonsmokers (0.34 ± 0.2 fmol/mg creatinine). While this study established a direct phenotyping approach for metabolism of BaP by the carcinogenic diol epoxide pathway, the levels of trans, anti-BaPT in urine were quite low because human exposure to BaP is low and its metabolites are excreted mainly in the feces. Thus, for practical determination of PAH metabolic activation and detoxification in large studies, we have focused on phenanthrene (Phe, Fig. 3). The metabolism of Phe closely mimics that of BaP, but human exposure to Phe is far greater than to BaP, and in contrast to BaP, most Phe metabolites are excreted in urine (39). The concentration of the Phe diol epoxide metabolite, trans, anti-PheT, in urine is about 10,000 times greater than that of trans, anti-BaPT, and the two metabolites are significantly correlated (40, 41). Thus, trans, anti-PheT is an excellent and practical surrogate for trans, anti-BaPT. Similarly, 3-OHPhe can serve as a surrogate for 3-OHBaP, representing detoxification pathways of these PAH (41).

Levels of trans, anti-PheT in the urine of cigarette smokers in one large study were 1.43 ± 2.16 pmol/mL urine (n = 2,613; ref. 42), while in another study in which Phe metabolites were monitored over a period of 49 days, mean levels of trans, anti-PheT were 3.74 ± 3.85 pmol/mg creatinine in smokers (n = 12) and 1.44 ± 0.97 pmol/mg creatinine in nonsmokers (n = 10; ref. 43). Trans, anti-PheT/3-OHPhe ratio, as a metabolic activation/detoxification ratio, has been monitored in 346 smokers, with a mean molar ratio of 4.08 (44). In this study, 10% of the smokers had ratios ≥ 9.90, suggesting that they belong to a high risk group. Levels of trans, anti-PheT are also affected by air pollution. Thus, urine samples collected in Shanghai, China from 1986 to 1989 had average levels of 19.3 pmol/mg creatinine trans, anti-PheT, approximately 6 times higher than cigarette smokers in Minneapolis in 2003–2004 (45).

It should be noted that urinary biomarker values are commonly reported either per mg creatinine or per mL urine. Specific gravity has also been used. All of these methods have advantages and limitations. Creatinine can be affected by muscle mass, gender, and race while urinary volume is affected by extent of hydration and other factors. Specific gravity of normal human urine may range from 1.016 to 1.024. None of these denominators is perfect and results must be interpreted in study context (46, 47).

The relationship of these PAH metabolites to lung cancer was investigated in the Shanghai Cohort Study. Among cigarette smokers (476 cases, 476 controls), urinary levels of PheT were significantly related to lung cancer incidence, after adjustment for self-reported smoking intensity and duration as well as urinary total NNAL and cotinine (48). Among lifelong never smokers (82 cases, 83 controls), trans, anti-PheT, 3-OHPhe, and total OHPhe were significantly related to lung cancer risk, indicating the potentially important role of PAH exposure as a cause of lung cancer in both smokers and never smokers (49).

Quantitation of DNA adducts in tissues of people who use tobacco products can increase our understanding of tobacco carcinogenesis and aid in the development of cancer prevention strategies (7, 50–52). We have focused on high-resolution MS quantitation of DNA adducts formed by three major classes of tobacco smoke carcinogens and toxicants: tobacco-specific nitrosamines, PAH, and aldehydes. The quantitation of DNA adducts in human tissues is challenging because the adduct levels are typically very low (from 1 per 10⁶ to 1 per 10¹¹ normal bases) and the amounts of DNA available for analysis are generally in the microgram range. Thus, in practice, one is frequently quantifying fmol – amol (10⁻¹⁸ mol) levels of analytes. This can only be done with confidence by liquid chromatography-nanospray ionization-high resolution tandem mass spectrometry (LC-NSI-HRMS/MS) techniques, the methods that have been employed in all of our current human DNA adduct analyses.

The carcinogenic tobacco-specific nitrosamines NNK and N´-nitrosonornicotine (NNN) are metabolically activated via α-hydroxylation and 2′-hydroxylation, respectively, to form reactive pyridyloxobutylating intermediates (6, 53). These intermediates react with DNA to produce well-characterized pyridyloxobutyl (POB)-DNA adducts, some of which have known miscoding properties and activate K-ras (6, 53). Under strong acid hydrolysis conditions, certain POB−DNA adducts decompose to release 4-hydroxy-1-(3-pyridyl)-1-butanone (HPB, Fig. 4A; refs. 54, 55). The importance of HPB-releasing adducts in NNN and NNK carcinogenesis has been demonstrated (6, 56).

Recently, we developed a highly sensitive high resolution LC/MS-MS assay for the measurement of HPB-releasing DNA adducts in oral cells–a noninvasive source of DNA and a promising surrogate tissue for assessing NNN- and NNK-induced molecular alterations in the aerodigestive tract (57, 58). Analysis of HPB-releasing DNA adducts in oral cell DNA from smokers with and without head and neck squamous cell carcinoma (HNSCC) showed a remarkable difference between the two groups: the median HPB-releasing DNA adduct level was 6.6 times greater for those with HNSCC than for smokers without HNSCC (P = 0.002; 30 cases and 35 controls; ref. 58). Additional multivariate regression adjusting for gender, age, and alcohol drinking frequency showed persistent significant difference in HPB-releasing adduct levels between cases and controls with a ratio of geometric means equal to 20.0 [95% confidence interval (CI) = 2.7–148.6; ref. 59]. Together, these findings suggest that levels of HPB-releasing DNA adducts in oral cells may be an independent predictor of HNSCC in smokers and potentially in users of other tobacco products.

As discussed above, PAH are important carcinogens in cigarette smoke (60). The major characterized and most mutagenic BaP-DNA adduct is BPDE-N²-dG (Fig. 3). Analysis of this and related PAH adducts in human tissues has been challenging. While ³²P-postlabeling and immunoassay methods have been widely reported, they are not specific for PAH–DNA adducts (61). We developed an ultrasensitive and specific assay for BPDE-N²-dG in human lung tissue, with a limit of detection of 1 amol of BPDE-N²-dG, corresponding to 1 adduct per 10¹¹ nucleotides (62). Smoker (n = 11) and nonsmoker (n = 18) lung DNA samples contained 3.1 and 1.3 BPDE-N²-dG adducts per 10¹¹ nucleotides, respectively.

Acrolein is a highly toxic compound that occurs in tobacco smoke in considerable quantities, typically 30.8–82.6 μg per cigarette (63). Its carcinogenicity is questionable and it has been evaluated as “not classifiable as to its carcinogenicity to humans” by the IARC (64). Our group was the first to determine structures of the acrolein DNA adducts α-acrolein-dG and γ-acrolein-dG (Fig. 4B) (65). Many studies have quantified these DNA adducts in human tissues and cells with varying results based on the methods used and the tissues studied (66–73). The levels of acrolein-DNA adducts detected in human lung tissue from smokers (validated by urinary NNAL measurements, n = 19) and nonsmokers (n = 18) were 28.5 ± 14.9 adducts/10⁹ nucleotides and 25.0 ± 10.7 adducts/10⁹ nucleotides, respectively, suggesting rapid removal of acrolein from lung tissue of smokers by glutathione conjugation and other detoxification mechanisms (74). These results do not support the hypothesis that acrolein is a major etiologic agent for cigarette smoking-related DNA damage in lungs, which has been suggested on the basis of its ability to interact with the TP53 gene at similar mutational hotspots as found in smoking-related lung cancer (67, 75).

The levels of DNA adducts are a combination of adduct formation, stability, and repair. Therefore, information on their persistence in DNA is crucial to understand the relevance of the exposure, effect, and risk that they represent. Some adducts are poorly repaired and persist, while other are short-lived. For example, DNA adducts induced by aristolochic acid are detected in renal tissue of patients 20 years after they stop consuming herbal remedies containing this carcinogen (76) Cisplatin-DNA adducts in postmortem human organ tissue were detected up to 22 months after the last treatment (77). On the other hand, the acetaldehyde-derived DNA adduct, N²-ethyl-2´-deoxyguanosine (N²-ethyldG), a potential marker for susceptibility to alcohol-related cancers of the upper aerodigestive tract (78), is quite short-lived following ethanol exposure in humans. We found that the levels of this adduct in oral cell DNA were reduced to baseline levels 6–24 hours following a low, intermediate, or high dose of ethanol, corresponding to roughly 1, 2, and 3 drinks, respectively (Fig. 4C; ref. 79). This study was the first to demonstrate that oral cavity DNA adducts of acetaldehyde are formed upon alcohol exposure and confirmed the importance of the sampling time selection.

Cigarette smoke is a rich source of important pro-oxidants that can induce oxidative damage and lead to an inflammatory state (7, 80). In addition, toxic and carcinogenic effects caused by a number of tobacco constituents and environmental toxicants can trigger inflammatory processes. Inflammation generates a spectrum of reactive oxygen and nitrogen species capable of causing extensive damage to DNA and proteins, directly or via lipid peroxidation, resulting in toxic and mutagenic events (81–86). These processes contribute to the pathogenesis of diseases including cancer and a range of respiratory and cardiovascular diseases associated with cigarette smoking, environmental pollutants, and other harmful exposures (87–91).

Oxidative damage to DNA and other macromolecules is likely involved in tumor promotion and other deleterious effects (92, 93). Guanine is the major target for the direct oxidation by the inflammation-induced radicals, resulting in several DNA adducts including 8-oxo-7,8-dihydro-2′-deoxyguanosine (8-oxo-dGuo, Fig. 4B), a widely used biomarker of oxidative stress that can lead to chromosomal aberrations and the induction of mutations (94). Some methodologies, such as ³²P-postlabeling, capillary electrophoresis, or PCR-based techniques, lack selectivity and/or sensitivity for the analysis of these adducts in the small amounts of DNA available in epidemiologic studies or isolated from biological matrices such as oral cells (95–98). As with other DNA adducts discussed here, LC-NSI-MS/MS methodologies provide the most sensitive and accurate detection of oxidative DNA damage. We recently developed a highly sensitive and selective LC-NSI-MS/MS method for the analysis of 8-oxo-dGuo in less than 100 ng of DNA, with a limit of detection at 0.04 fmol and limit of quantitation at 0.1 fmol on-column (99). Careful attention to artifactual oxidation of dG to produce 8-oxo-dGuo was mitigated with antioxidants and monitored by the addition of [¹⁵N₅]dG at the start of sample processing to detect artifactual formation of [¹⁵N₅]8-oxo-dGuo (99). Formation of 8-oxo-dGuo in the ion source from dGuo present in the purified sample is another concern; therefore, we use column purification during sample processing to remove dGuo along with other potentially interfering molecules.

Inflammation-associated DNA damage can also occur through lipid peroxidation of polyunsaturated fatty acids that produces malondialdehyde. Reaction of malondialdehyde with DNA leads to the formation of various adducts, with 3-(2-deoxy-β-d-erythro-pentafuranosyl)-pyrimido[1,2-α]purin-10(3H)-one deoxyguanosine (M₁dGuo) being the predominant one (Fig. 4B ref. 100). M₁dGuo is also formed from radical reactions with the sugar backbone of DNA via base-propenal formation (101, 102). M₁dGuo induces G→T and G→A mutations, potentially contributing to the etiology of cancer and other inflammation-associated diseases (81, 100, 103–107).

In humans, M₁dGuo has been detected in blood (leukocytes), various organ tissues and urine, with many studies employing nonselective ³²P-postlabeling and immunochemical techniques, as well as GC-MS (108–111). The effect of smoking on M₁dGuo levels in humans is not clear because of the lack of consistency across published reports, which is potentially due to different biological matrices that have been examined and different analytic methodologies used. Our group recently developed a highly selective and sensitive LC-NSI-HRMS/MS method for the analysis of M₁dGuo in human leukocyte DNA (112), which can be used in future studies to clarify the role of M₁dGuo in smoking-associated and other diseases.

Multiple studies have shown that the urinary metabolites (Z)-7-[1R,2R,3R,5S)-3,5-dihydroxy-2-[(E,3S)-3-hydroxyoct-1-enyl]-cyclopentyl]hept-5-enoic acid (8-iso-PGF_2α), an F2-isoprostane and biomarker of oxidative damage, and “prostaglandin E₂ metabolite” (PGE-M), a biomarker of inflammation, are elevated in smokers (113–120). Both biomarkers decrease only modestly after cessation of smoking reflecting their relationship to complex mechanisms of oxidative damage and inflammation, as opposed to direct toxicant or carcinogen exposures (121). Urinary 8-iso-PGF_2α levels were significantly higher in lung cancer cases (n = 610) than their smoking-matched controls (n = 610) in the prospective Shanghai Cohort Study (122), supporting the important role of oxidative damage and tumor promotion in lung cancer etiology in smokers.

1,3-Butadiene (BD) is a known animal and human carcinogen (123–129). The genotoxic effects of BD are attributed to its epoxide metabolites (Fig. 5A) (123). Cytochrome P450 monooxygenases catalyze the epoxidation of BD to yield 3,4-epoxy-1-butene (EB; refs. 130, 131). EB can be hydrolyzed to 1-butene-3,4-diol (EB-diol) by microsomal epoxide hydrolases (EH), conjugated with glutathione in the presence of GSTs, or subjected to a second oxidation to yield S,S, R,R- and meso-diepoxybutane (DEB; refs. 130, 132–134). EB-diol can in turn be coupled with glutathione (following conversion to hydroxymethyl-vinylketone, HMVK) or further oxidized to form 3,4-epoxy-1,2- butanediol (EBD; ref. 130). All three BD-derived epoxides are direct-acting mutagens (135–137). EBD is the most abundant in vivo (138), while DEB is the most cytotoxic and mutagenic (139). Glutathione conjugation products of EB, HMVK, EB-diol, and DEB are further metabolized and excreted in urine as the corresponding N-acetylcysteine derivatives: MHBMA, DHBMA, THBMA, and bis-BDMA (Fig. 5A; refs. 140–142).

We have developed isotope dilution HPLC-ESI⁻-MS/MS methodology for MHBMA, DHBMA, THBMA, and bis-BDMA in urine (Fig. 5B; refs. 143). Two studies involving current smokers demonstrated significant ethnic differences in the levels of these metabolites. In the first study (n = 584), we found that urinary MHBMA levels and the (MHBMA)/(MHBMA + DHBMA) metabolic ratio were the highest in White smokers and the lowest in JA smokers, while DHBMA values were not significantly different between groups (144). A second study of smokers from the MEC (n = 1,072) demonstrated that Whites excreted the highest levels of MHBMA, followed by AA and JA (145). Copy number of the GSTT1 gene partially explained ethnic differences in MHBMA excretion (12, 145).

The genotoxic effects of BD are mediated by the formation of BD–DNA adducts. EB reacts with DNA to form 7-(2-hydroxy-3-buten-1-yl)guanine (EB-GI) and 7-(1-hydroxy-3-buten-2-yl)guanine (EB-GII; Fig. 5C; refs. 146–150). EBD forms 7-(2,3,4-trihydroxybut-1-yl)-guanine (THB-Gua) adducts, while DEB sequentially alkylates two guanine bases of DNA to generate bis-7G-butanediol cross-links (bis-7G-BD; Fig. 5D). Although 7-alkylguanines have not been shown to be mutagenic, they are the most abundant adducts formed in cells and humans upon exposure to BD (123, 148, 151–153) and are considered biologically relevant biomarkers of carcinogen exposure and risk (133, 146, 149, 154, 155).

Adducts generated by alkylation of the 7-position of guanine in DNA by EB, EBD, and DEB are hydrolytically labile, leading to their spontaneous depurination and excretion in urine (156). We have developed nanoLC/ESI⁺ HRMS³ methodologies for urinary depurinated EB-Gua II and THB–Gua adducts and have shown that their levels were increased in workers occupationally exposed to BD (157, 158) and in smokers as compared with nonsmokers (159). EB-GII quantitation in smokers' urine revealed that White smokers (n = 75) excreted significantly higher amounts of EB-GII than did AA smokers (n = 75; 0.48 ± 0.09 vs. 0.12 ± 0.02 pg/mg creatinine, P = 3.1 × 10⁻⁷; ref. 160). There was no significant correlation between EB-GII and BD-mercapturic acids (R ∼ 0.03). Because urinary EB-GII levels represent the concentration of BD–DNA adducts released from DNA due to spontaneous hydrolysis and/or DNA adduct repair, their increased adduct amounts in urine of White smokers could represent more efficient removal of EB-GII adducts by Whites as compared with AA smokers.

Although these biomarkers have been successfully applied to estimate occupational and environmental exposure to BD, their use as biomarkers of cancer risk is still controversial. Genomic BD–DNA adducts would be a more relevant biomarker, but are difficult to measure due to limited access to human tissue samples, requiring the development of more sensitive MS-based methodologies.

A major impediment in molecular epidemiology studies is the lack of fresh frozen biopsy specimens for cancer biomarker research and our understandings of the role of chemical exposures in DNA damage, adduct formation, and human disease risk, such as cancer, have been slow to advance. In contrast, archived formalin-fixed paraffin-embedded (FFPE) tissues with a clinical diagnosis of disease are often accessible for research. Formalin is the most common method of fixation, and FFPE has been used as the standard storage technique for more than a century in laboratories worldwide. The chemical fixation process preserves cell morphology; cellular components remain intact. Thus, details of the preserved cells can be visualized, by various staining techniques, followed by light and electron microscopy, or IHC techniques (161).

IHC techniques have served for the screening of several putative carcinogen DNA adducts in human FFPE tissues (162–164). The applicability of IHC to screen paraffin-embedded tissue-sections makes IHC an attractive method for the analysis of DNA adducts. However, a critical drawback of IHC is that the specificity of many antibodies, even mAbs, for DNA adducts is uncertain as they may cross-react with other DNA lesions or cellular components, leading to errors in identification and quantification. Moreover, the panel of DNA adducts that can be screened by IHC is limited to those lesions for which antibodies are available. A method has been long sought to measure DNA adducts in FFPE specimens by quantitative MS methods. The major obstacle in the use of FFPE tissues for quantitative measurements of DNA adducts has been the development of a technique to fully unravel DNA crosslinks in FFPE specimens under mild conditions that recovers DNA in high yield while preserving the chemical structures of the DNA adducts. Conditions commonly used in genomics to reverse DNA crosslinks employ elevated temperature and alkaline pH (165, 166). These harsh conditions can destroy many classes of DNA adducts (167, 168).

The quantitative measurement of DNA adducts in FFPE specimens requires the retrieval of DNA devoid of cross-links that is fully digestible by nucleases while preserving the chemical structures of the DNA adducts (169). Several vendors have developed kits advertised to retrieve high-quality DNA under mild retrieval conditions that is free of cross-links, which can serve as high-fidelity templates for amplification with PCR. We tailored the DNA retrieval method of commercial FFPE kits to measure DNA adducts of multiple classes of carcinogens by ion trap and high-resolution Orbitrap MS (170, 171).

In collaboration with the Grollman Laboratory, Stony Brook University (Stony Brook, NY), we showed that 7-(2′-deoxyadenosine-N⁶-yl)aristolactam (dA-AL-I, Fig. 4B), a DNA adduct of aristolochic acid-I, a potent upper urothelial carcinogen (UUC) in Aristolochia herbs still found in traditional Chinese medicines, could be retrieved from human FFPE kidney blocks stored at ambient temperature for up to nine years and measured by ion trap MS at levels closely matching those of fresh-frozen specimens (170). The dA-AL-I adduct is responsible for the uncommon A:T-to-T:A transversion mutational signature attributable to aristolochic acid in UUC (172). Thereafter, in collaboration with Scelo and colleagues at IARC, we showed that dA-AL-I was prevalent in kidney tissue of patients with renal cell cancer (RCC) in Romania, but not detected in cohorts from the United Kingdom, the Czech Republic, and Russia (173). The proportion of A:T-to-T:A mutations increased according to the level of dA-AL-I, implicating AA-I in the development of RCC in Romania (173–175).

Our method also detected DNA adducts of the bladder carcinogen 4-aminobiphenyl in FFPE bladder specimens (176) and a DNA adduct of 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP, Fig. 4B), a heterocyclic aromatic amine formed in cooked meats and a potential prostate carcinogen, in FFPE specimens of patients with prostate cancer (177, 178). Thus, the technology can screen for DNA adducts formed with a wide range of carcinogens, using only 10 μg of DNA, a quantity recovered from several FFPE section cuts (∼1 cm² × 10 μm thickness).

The ability to retrospectively analyze FFPE tissues for DNA adducts paves the way for usage of this largely untapped source of biospecimens for molecular epidemiology studies seeking to understand the causal role of environmental chemicals in cancer etiology. The DNA adduct biomarker data obtained from FFPE tissues must be interpreted with caution. The DNA adducts detected may represent recent exposures, and the absence of DNA adducts does not preclude earlier exposure and DNA damage. In the case of chronic exposures such as tobacco, diet, and pollution, the biomonitoring of DNA adducts in FFPE tissue may represent long-term exposure to the chemicals. FFPE tissues also can serve to screen for DNA adducts that are resistant to DNA repair, such as aristolochic acid. Moreover, FFPE tissues can be screened for a wide range of exogenous DNA adducts through our emerging adductomics approaches (179).

Humans are exposed to many environmental chemicals as well as a variety of endogenous chemicals derived from cellular metabolism, or endogenous processes including the microbiome. The totality of this exposure has been defined as the exposome (180). This comprehensive assessment parallels the development of powerful technologies such as transcriptomics, proteomics, and metabolomics, and is used to define an individual's response to complex exposures taking into account interindividual variations in bioavailability, absorption, metabolism, and excretion. Drawing upon technological innovations in analytical chemistry, we have been establishing DNA adductomic approaches to comprehensively characterize DNA damage resulting from complex exposures. DNA adductomics is emerging as a promising tool to screen for patterns of DNA modifications. This contrasts with targeted LC/MS methods that focus on measurement of small numbers of anticipated DNA adducts based on a priori assumptions regarding their formation from specific chemicals. This methodology does not account for the complexity of in vivo DNA adduct formation resulting from exposure to complex mixtures, which cannot be completely anticipated or predicted.

Our DNA adductomic studies are performed using LC high-resolution/accurate-mass (HRAM) MSⁿ data acquisition with Orbitrap mass analyzers, allowing for the selectivity and identification capabilities that support both the screening of multiple DNA modifications and the identification of unknowns. Using this technology, we established novel complementary data-independent acquisition (DIA; ref. 181) and data-dependent acquisition (DDA; ref. 182) DNA adductomic methods (Fig. 6). These two approaches both rely on accurate mass measurements that allow for the determination of the likely elemental compositions of both the parent ions and fragment ions of the observed DNA adducts.

Our DIA-WideSIM/MS² approach identified a DNA adduct of PhIP, a mutagen linked to prostate cancer, in prostate DNA from patients with prostate cancer (181) and the DNA adduct of the renal carcinogenic aristolochic acid (dA-AL-l) in renal cortex DNA (181). This method also detected lipid peroxide-derived DNA adducts in these organs (183). Our DDA-CNL/MS³ method was employed to screen lung DNA from A/J mice exposed to a mixture of tobacco smoke chemicals, the lung carcinogen NNK, and the proinflammatory agent, lipopolysaccharide. Both NNK-specific DNA adducts and DNA adducts derived from inflammation induced lipid peroxidation and oxidative stress were identified (184). Similarly, this methodology was applied to better characterize the range of DNA adducts caused by alkylating drugs, with the goal of identifying markers to be used to stratify patients in personalized medicine approaches (185). Finally, this approach recently identified the previously unknown in vivo DNA adducts induced by the E.coli–derived genotoxin colibactin, which is believed to play a role in colorectal cancer (186).

Altogether, these analyses demonstrate the potential of our DNA adductomic methods to provide a broader picture of the overall DNA damage burden induced by complex exposures from known genotoxicants and by identifying new unknown DNA adducts as well as set the stage to investigate the effects derived from a combination of exposures. Challenges remain as this is a developing field. Limitations include isobaric interferences and the lack of automated tailored data analysis tools to detect DNA adducts. Databases are also needed to support the structural elucidation of DNA adducts. Development of software to improve the efficiency of the data analysis and strategies for credentialing the putative adducts are ongoing (187, 188).

Biomarkers such as urinary NNAL, PheT, and mercapturic acids of volatile toxicants including acrolein and benzene have been particularly useful in clinical studies evaluating the relationship between tobacco products use and tobacco carcinogen exposure. For example, we showed clear reductions in NNAL and metabolites of acrolein, benzene and 1,3-butadiene when people quit smoking (189). Significant decreases in NNAL also occurred when smokeless tobacco use stopped (190). These studies established the principle that biomarkers can verify decreased toxicant and carcinogen exposure upon cessation of tobacco use. Biomarkers, particularly of tobacco-specific carcinogens, have been crucial in confirming the link between secondhand smoke exposure and cancer (191–193). These studies played an important role in the clean air regulations that are now practically universal in the United States. A series of studies evaluating tobacco products have provided key data on how tobacco product formulation affects internal exposure to nicotine and tobacco carcinogens (194–198). These data contributed to the FDA proposal of product standards requiring significant reduction of NNN levels in smokeless tobacco and reduction of nicotine in cigarettes to minimally addictive levels (197, 198) More recently, we determined the effects of immediate versus gradual reduction in nicotine content of cigarettes to very low levels relative to usual levels on toxicant exposure biomarkers (199). Significantly lower levels of exposure to nicotine, NNK, PAH, and volatile toxicants were found in the immediate versus the gradual reduction group.

Investigators at the University of Minnesota (Minneapolis, MN) have continued research of the late Dr. Lee Wattenberg on the role of cruciferous vegetables and their constituents in prevention of tobacco and environmental carcinogen–related cancer (200). A primary challenge has been the need for objective biomarkers of cruciferous vegetable consumption and uptake of the glucosinolates contained therein—those natural phytochemicals that give rise to hydrolysis products, several of which have led to remarkable reductions in carcinogen-induced tumors in animal models. Establishing a beneficial effect of glucosinolates on environmental carcinogenesis in humans requires rigorous study using mechanistically relevant biomarker endpoints. Our prior work has focused on developing a biomarker reflective of uptake of glucobrassican, the precursor to indole-3-carbinol, an important constituent of common cruciferous vegetables (200). We are currently conducting a clinical trial to see whether eating Brussels sprouts daily can favorably alter PAH metabolism. This work is just the beginning of the mechanism-based clinical studies required to reach our overarching goal of establishing a simple, dietary intervention to prevent tobacco and environmental carcinogenesis, with a potential global impact.

Multiple preclinical studies have shown that 2-phenethyl isothiocyanate (PEITC) is an effective chemopreventive agent against the tobacco-specific lung carcinogen NNK (201–203). PEITC inhibits NNK carcinogenesis by inhibiting the metabolic activation of NNK, as clearly demonstrated by studies in laboratory animals (203). We carried out a clinical trial of PEITC in cigarette smokers (n = 82) to determine whether it could inhibit NNK metabolism as found in laboratory animals (204). This trial required the use of cigarettes spiked with [pyridine-D₄]NNK to distinguish its metabolism from that of a minor nicotine metabolic pathway. The biomarker used in this study was the ratio of two [pyridine-D₄]NNK metabolites - [pyridine-D₄]hydroxy acid:[pyridine-D₄]NNAL. The results of this placebo controlled clinical study demonstrated that PEITC did significantly inhibit the metabolic activation of NNK in cigarette smokers, although the extent of inhibition (7.7%) was modest, and further studies are required. However, this study also showed, in an evaluation of additional biomarkers, that PEITC had a significant and far stronger effect on the detoxification of benzene, acrolein, and crotonaldehyde based on analysis of their mercapturic acid metabolites in urine (205). The effect was particularly notable in individuals who were null for the GST enzymes GSTM1 and GSTT1. This observation is currently being pursued in a clinical trial of watercress, an abundant source of PEITC, in both smokers and nonsmokers.

As demonstrated above, these biomarkers of exposure and intermediate effect are valuable in defining cancer risks associated with environmental and behavioral exposures, particularly for tobacco. Furthermore, these biomarkers confirmed exposure to tobacco carcinogens as well as demonstrated exposure reduction in interventions, providing key data to support public policies that reduced tobacco exposure and, consequently, cancer burden. The reduction in tobacco-related cancers in the United States has been driven by primary prevention methods. Similar tactics can be applied to reduce cancers associated with other environmental exposures.

Our ability to measure DNA damage from both exogenous and endogenous sources in human tissues continues to improve and provides researchers with strong mechanistic links between exposure and cancer risk, as demonstrated above for aristolochic acid. The capability to measure DNA adducts in FFPE tissues expands the ability to explore the causal role of environmental chemicals in cancer etiology for molecular epidemiology studies. DNA adductomics will assist in identifying the etiologic factors responsible for common cancers whose cause is still unknown.

Many of the analytes described in this review will be available to the NIH-funded community through the Human Health Exposure Assessment Resource. This resource will provide opportunities for epidemiologists to explore the connection of environmental exposures to human disease outcomes like cancer.

N. Fujioka reports receiving other commercial research support from MedImmune/AstraZeneca. No potential conflicts of interests were disclosed by the other authors.

This content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

This work was supported by the NCI grants P01 CA138338 (to S.S. Hecht, S.E. Murphy, N.Y. Tretyakova, I. Stepanov, L.A. Peterson, D.K. Hatsukami), R01 CA81301 (to S.S. Hecht), R01 CA222005 (to S.S. Hecht), U19 CA157345 (to D.K. Hatsukami, S.S. Hecht, S.E. Murphy, I. Stepanov), U54 DA031659 (to D.K. Hatsukami, S.S. Hecht, S.E. Murphy, I. Stepanov), R01 CA122320 (to R.J. Turesky), R01 CA220367 (to R.J. Turesky), R01 ES019564 (to R.J. Turesky), R33 CA186795 (to R.J. Turesky), R50 CA211256 (to P.W. Villalta), R01 CA179246 (to I. Stepanov), U01 DA045523 (to I. Stepanov), and R01 CA220376 (to S. Balbo). The authors are also part of the Human Health Exposure Assessment Resource (HHEAR) supported by the National Institute of Environmental Health Sciences grant U2CES026533. The Masonic Cancer Center Analytical Biochemistry Shared Resource is funded in part by NCI grant P30 CA77598. I. Stepanov is supported in part by startup funding from Minnesota Masonic Charities. The authors thank Mr. Bob Carlson for his assistance with the figures and other editorial matters. The authors also thank their many collaborators, including Dr. Loic Le Marchand at the University of Hawaii (Honolulu, HI); Ms. Yesha Patel, Dr. Christopher Haiman, Dr. Daniel Stram, and Dr. Sungshim Park at the University of Southern California (Los Angeles, CA); Dr. Jian-Min Juan at the University of Pittsburgh (Pittsburgh, PA); Dr. James Swenberg at the University of North Carolina (Chapel Hill, NC); Dr. Vernon Walker at the University of Vermont; Dr. Ivan Rusyn at Texas A&M University (College Station, TX); Dr. Thomas A. Rosenquist, Dr. Kathleen G. Dickman, and Dr. Arthur P. Grollman at Stony Brook University (Stony Brook, NY); and Dr. Byeong Hwa Yun, Dr. Medjda Bellamri, Dr. Jingshu Guo, Dr. Paari Murugan, and Dr. Christopher J. Weight at the University of Minnesota (Minneapolis, MN).