**Formalin-Fixed Paraffin-Embedded Tissues—An Untapped Biospecimen for Biomonitoring DNA Adducts by Mass Spectrometry**

#### **Byeong Hwa Yun, Jingshu Guo and Robert J. Turesky \***

Masonic Cancer Center and Department of Medicinal Chemistry, University of Minnesota, 2231 6th St. SE, Minneapolis, MN 55455, USA; bhyun@umn.edu (B.H.Y.); guoj@umn.edu (J.G.)

**\*** Correspondence: rturesky@umn.edu; Tel.: +1-612-626-0141; Fax: +1-612-624-3869

Received: 28 April 2018; Accepted: 25 May 2018; Published: 1 June 2018

**Abstract:** The measurement of DNA adducts provides important information about human exposure to genotoxic chemicals and can be employed to elucidate mechanisms of DNA damage and repair. DNA adducts can serve as biomarkers for interspecies comparisons of the biologically effective dose of procarcinogens and permit extrapolation of genotoxicity data from animal studies for human risk assessment. One major challenge in DNA adduct biomarker research is the paucity of fresh frozen biopsy samples available for study. However, archived formalin-fixed paraffin-embedded (FFPE) tissues with clinical diagnosis of disease are often available. We have established robust methods to recover DNA free of crosslinks from FFPE tissues under mild conditions which permit quantitative measurements of DNA adducts by liquid chromatography-mass spectrometry. The technology is versatile and can be employed to screen for DNA adducts formed with a wide range of environmental and dietary carcinogens, some of which were retrieved from section-cuts of FFPE blocks stored at ambient temperature for up to nine years. The ability to retrospectively analyze FFPE tissues for DNA adducts for which there is clinical diagnosis of disease opens a previously untapped source of biospecimens for molecular epidemiology studies that seek to assess the causal role of environmental chemicals in cancer etiology.

**Keywords:** carcinogen; DNA adducts; biomonitoring; formalin-fixed paraffin-embedded tissues; biomarker; mass spectrometry

#### **1. Metabolism, Bioactivation, and DNA Adducts as Biomarkers of Exposure and Health Risk**

#### *1.1. Xenobiotic Metabolism and Bioactivation of Procarcinogens*

Humans are continuously exposed to potentially hazardous chemicals in the environment, diet, medicines, and through occupational exposures. Many of these chemicals undergo biotransformation by phase I and/or phase II enzymes to produce reactive electrophiles that can form adducts with macromolecules [1]. Cytochrome P450s (P450s) are by far the most important Phase I enzymes involved in xenobiotic metabolism [2]. P450s catalyze a variety of reactions, including aliphatic and aromatic hydroxylation, *N*- or *O*-dealkylation, aliphatic desaturation, hetero atom oxidation, and epoxidation reactions [2]. The resulting metabolites can contain functional groups such as –OH, –NH2, and –SH which can undergo conjugation reactions by phase II enzymes, including UDP-glucuronosyltransferases (UGTs), sulfotransferases (SULTs), *N*-acetyltransferases (NATs), glutathione *S*-transferases (GSTs), and methyltransferases [3].

While many Phase I metabolites are detoxification products, some oxidative metabolites are reactive electrophiles, which can induce toxicity or genotoxicity by covalently binding to protein or DNA, or generate free radicals that deplete cellular antioxidants and induce oxidative stress [4,5]. In a similar vein, many phase II enzyme reactions are regarded as detoxification pathways, and the resulting metabolites are efficiently eliminated from the body. However, in some cases, reactive intermediates are generated, and the metabolites can bind to proteins and DNA. The *O*-acetylation or *O*-sulfation of aromatic amines and heterocyclic aromatic hydroxylamines [6], glutathione conjugation of ethylene dibromide [7], *O*-sulfation of hydroxymethyl polycyclic aromatic hydrocarbons [8], and the acyl glucuronidation of carboxylic acid moieties of nonsteroidal anti-inflammatory drugs (NSAIDs) [9] are examples of conjugation reactions leading to reactive intermediates. The metabolic activation of rodent and possible human carcinogens including 2-amino-1-methyl-6-phenylimidazo[4,5-*b*]pyridine (PhIP) [10], aristolochic acid I (AA-I) [11], 5-methylchrysene [12,13] and tamoxifen [14,15], are shown as examples of procarcinogens that require phase I and/or II enzymes to produce penultimate species that bind to DNA (Figure 1).

**Figure 1.** The metabolic activation of aristolochic acid I (AA-I), 2-amino-1-methyl-6-phenylimidazo[4,5 *b*]pyridine (PhIP), 5-methylchrysene, and tamoxifen are shown as prototypes of procarcinogens. Bioactivation is carried out with phase I and/or phase II enzymes, which lead to the formation of DNA adducts. AA-I undergoes nitro-reduction through NAD(P)H:quinone oxidoreductase (NQO1), cytochrome P450s 1A1 and 1A2, NADPH:P450 reductase (POR) or prostaglandin H synthase (COX).

The resulting *N*-hydroxyaristolactam-I is bioactivated by SULTs to form an unstable *N*-sulfoxy ester, which quickly undergoes heterolytic cleavage to produce the reactive nitrenium/carbenium intermediate that forms dA-AL-I and other DNA adducts. PhIP undergoes *N*-hydroxylation by P450s, then it is further bioactivated by NATs or SULTs to form *N*-acetoxy or *N*-sulfoxy esters, which lead to the formation of dG-C8-PhIP through the nitrenium intermediate. 5-Methylchrysene undergoes epoxidation (P450s 1A1 and 1B1) followed by epoxide hydroxylation (epoxide hydrolase) on the bay-region phenyl ring, to form the corresponding *trans*-1,2-dihydrodiol-5-methylchrysene. A subsequent round of monooxygenation leads to the formation of *anti*-1,2-dihydrodiol-3,4-epoxide-5-methylchrysene, which can form a DNA adduct at the *N*2-atom of dG (dG-*N*2-5-methylchrysene-diolepoxide). Two pathways are involved in the DNA adduct formation of the bioactivated tamoxifen. In the first pathway, oxidation of the allylic ethyl side chain results in the formation of *α*-hydroxytamoxifen. The subsequent esterification catalyzed by SULTs leads to the reactive carbenium intermediate and the dG-*N*2-α-hydroxytamoxifen adduct. The second pathway involves aryl-oxidation of one of the phenyl rings to yield 4-hydroxytamoxifen quinone methide, a reactive electrophile that can form the DNA adducts. Both pathways lead to (*Z*)- or (*E*)-dG-*N*2-4-hydroxytamoxifen.

Rodents are often employed as experimental laboratory animals to study metabolism of hazardous chemicals, to screen for DNA adduct formation, and elucidate mechanisms of carcinogenesis [5]. The metabolism of carcinogens and their biological effects in animal models can differ from humans because of species differences in catalytic activities of phase I and II enzymes involved in bioactivation or detoxification [10,16–18]. Thus, animal carcinogen bioassay data may not accurately gauge health risk of some chemicals in humans. However, DNA adducts of carcinogens, which are measures of the biologically effective dose, can serve as biomarkers for the extrapolation of genotoxicity data from animal studies for human risk assessment [19,20].

Epidemiological studies have reported that exposures to different chemicals in the diet and environment, or lifestyle factors, such as tobacco usage and alcohol consumption, are linked to the increased risk of developing certain types of cancers. As examples, polycyclic aromatic hydrocarbons (PAHs) in cigarette smoke are linked to lung cancer [21]; occupational exposures to aromatic amines are linked to bladder cancer [21,22]; usage of traditional Chinese herbal medicines containing AA-I are linked to upper urothelial cancer [23,24]; and consumption of aflatoxin B1 (AFB1) produced by fungi on agricultural crops, is a risk factor for liver cancer [25,26].

The identification and quantitation of DNA adducts is a first step in elucidating the potential role of a genotoxic chemical in the etiology of cancer [19,20]. The identification DNA adducts in human tissues are likely to represent a combination of recent and longer-term exposures to certain hazardous chemicals. The interpretation of negative findings, or the absence of DNA adducts, must be done with caution, since many adducts can undergo repair [27]. Ideally, the biomonitoring of DNA adducts should be conducted when the multistage process of tumorigenesis began, rather than many years later when the cancer is diagnosed. However, life-style factors such as tobacco smoking, diet, and environmental pollution often represent long-term exposures, and current adduct levels of carcinogens from these exposures are likely to correlate with adduct levels that existed during the time of tumor initiation and progression.

#### *1.2. Methods to Measure DNA Adducts*

The measurement of DNA adducts in humans is a challenging analytical task because the levels of DNA adducts generally occur at less than one adduct per 107 nucleotides, and the amount of tissue available for measurement is limited. Even for blood, a readily accessible biofluid, the amount of DNA obtained is usually a few up to several tens of micrograms scale. Thus, highly sensitive and specific methods are required to measure DNA adducts in humans. During the past three decades, the major techniques employed to measure DNA adducts have been 32P-postlabeling [28,29], antibody-based immunoassay/immunohistochemistry (IHC) [30,31], gas chromatography-mass spectrometry (GC-MS) [32], and most recently, liquid chromatography-mass spectrometry (LC-MS) [33–37].

32P-postlabeling is a highly sensitive method to detect DNA adducts. The DNA is enzymatically digested to 3 -phospho-2 -deoxyribonucleotides, and 32P-orthophosphate from [γ-32P] ATP is transferred to the 5 -OH position of the 2 -deoxyribonucleotide adduct, by polynucleotide kinase. The adducted 5 - 32P-labeled nucleotides are resolved by multi-dimensional thin-layer chromatography with polyethylenimine-modified cellulose plate, or by polyacrylamide electrophoresis, using autoradiography for detection, or by HPLC with radiometric detection [28,29,38,39]. The assay only requires 1–10 μg of DNA, and the sensitivity for some adducts can reach a limit of detection as low as one adduct per 1010 nucleotides [29]. Studies in rodents and humans employing 32P-postlabeling methods have shown that many genotoxic chemicals undergo metabolism and covalently adduct to DNA in many organs [29,40,41]. However, there are several limitations of the 32P-postlabeling assay. The technique is labor intensive and its usage requires large amounts of hazardous phosphorous radioactivity. Moreover, the technique is not quantitative [42], and structural information about the identity of the adduct is uncertain, particularly in humans where many overlapping lesions are present [29,40]. Thus, epidemiology studies employing 32P-postlabeling often provide equivocal data about chemical exposures linked to DNA adducts and cancer risk [43–46].

Immunodetection relies on the generation of monoclonal or polyclonal antibodies raised against modified-DNA adducts coupled to carrier proteins, or carcinogen-treated DNA, where usually very high levels of modification, about one modified base to 100 nucleotides, are required for successful generation of a titer [30,47]. The sensitivity of the method depends on the affinity of the antibody, but a detection limit of about one adduct per 108 nucleotides for certain DNA adducts can be reached, when detected by fluorescence or chemiluminescence spectroscopy [48,49]. IHC detection of DNA adducts in tissue section-cuts mounted on slides is generally less sensitive than immunoassays performed on isolated DNA; however, IHC allows the visualization of the DNA adduct within specific cell types of a tissue, and is especially suitable for archived human formalin-fixed paraff in-embedded (FFPE) tissues (Section 3) [50]. Cross-reaction of the antibody with DNA adducts of similar structure or cellular components can occur [30,31], which raises concerns about the specificity of the methodology. Immunodetection methods have made significant contributions to the biomonitoring of DNA adducts; however, similarly to the 32P-postlabeling method, immunodetection does not provide structural information to confirm adduct identity, and the method is semi-quantitative.

GC-MS with electron impact ionization, and more recently, negative ion chemical ionization has been employed to measure DNA adducts (primarily used for oxidized DNA bases) where adduct structures can be corroborated from the MS fragmentation spectra [32]. Often, the DNA is hydrolyzed with formic acid or by elevated temperature under neutral pH conditions. Most DNA adducts require chemical derivatization to increase the volatility required for GC analysis. The derivatization process can complicate the analysis and introduce artifact formation, particularly for oxidized DNA base measurements [51]. In contrast, the online coupling of capillary electrophoresis or LC to electrospray ionization (ESI) MS is a breakthrough technology that can measure many DNA adducts which would otherwise undergo thermal decomposition by GC-MS [52].

Currently, LC-ESI-multistage MS (MSn) is the predominant platform for DNA adduct analyses [33,35,37,53]. The rapidly advancing technologies in LC-MS instrumentation have attained ultra-high sensitivity and selectivity, particularly with ion trap and high resolution accurate mass spectrometry (HRAMS). These platforms include the coupling of nano-flow chromatography and nanoESI source, and versatile and flexible scanning strategies. The detection of DNA adducts at levels as low as one per 1011 nucleotides have been reported using a hybrid Orbitrap MS [54]. Both targeted and non-targeted MS scan approaches have been employed to identify many DNA chemical modifications [35–37,55–58].

The DNA is typically digested with a cocktail of nucleases prior to adduct measurements by LC-MS. The digestion products contain adducts formed at the DNA bases of the 2 -deoxyribonucleosides, or in rarer cases, adducts are formed at the phosphate backbones [59,60]. A common feature for many DNA adducts is their tendency to lose the deoxyribose moiety (dR, 116 or 116.0473 Da in HRAMS), when subjected to collision-induced dissociation (CID) [61]. The transition between the adduct precursors ([M + H]+) and their aglycones after losing dR ([M + H − 116]+) is commonly targeted to detect and quantify DNA adducts in MSn. The constant neutral loss of molecules, such as dR from the 2 -deoxyribonucleosides, serves as the foundation of the "DNA adductomics" approach [37,55,56,62,63]. Figure 2 shows the fragmentation pathways of modified nucleosides, where the major ions are the chemically modified bases after neutral loss of dR, or in less frequent cases the bases are eliminated as the neutral fragment and the carcinogen moieties retain the charge. These types of MS transitions are usually monitored in the targeted and un-targeted approaches by LC-MS [58,63].

**Figure 2.** The fragmentation pathways of modified nucleosides analyzed by LC-MS. (**A**) The major fragmentation of the modified nucleosides is the neutral loss of deoxyribose. Other common fragmentations include (**B**) the neutral loss of base and (**C**) the neutral loss of the adduct with the formation of base ions [58].

#### **2. Overview and the History of Formalin Fixation Process**

While great strides have been made in the detection of DNA adducts in humans, fresh tissues obtained from biopsies or post-mortem are often not available. The paucity of fresh tissue specimens has hampered the advancement of DNA adduct biomonitoring in human studies. However, archived FFPE tissue specimens with clinical diagnosis of disease are a largely untapped biospecimen and often available for DNA adduct biomarker research.

Formalin, 10% neutral buffered formaldehyde solution, is the most commonly used fixative worldwide [64]. During the process of formalin fixation, formaldehyde undergoes multiple steps of reactions with cellular nucleophilic species to form molecular crosslinks [64,65]. Formaldehyde permeates through the tissue, and the nucleophilic moieties of amino acids and nucleobases attack the formaldehyde yielding unstable intermediates of methylol adducts and Schiff bases [65]. These intermediates are stabilized by forming methylene bridges with a second nucleophilic group, often on another molecule. The methylene bridges formed with DNA and protein are stable crosslinks at room temperature (Figure 3); however, the linkages are reversible by heat treatment and/or under alkaline pH [66,67]. The reversal rate of the crosslink increases exponentially as a function of temperature [66]. The efficacy of reversal of formaldehyde-mediated crosslinks is

the most critical feature that impacts the quantitative analysis of RNA, DNA, and protein biomarkers in FFPE tissues.

**Figure 3.** The reactions of formaldehyde mediated crosslinking of DNA and protein. Formaldehyde diffuses through tissue and reacts with a nucleophilic sites of protein and/or DNA base resulting in unstable intermediates of methylol and Schiff base. Then, a second nucleophile from inter- or intramolecular DNA or protein attacks the Schiff base resulting in a crosslinked complex. A specific example of a protein-DNA crosslink is shown. The atoms are color coded: *cyan*, protein; *red*, formaldehyde; and *black*, DNA. Reproduced with permission from [65]. Copyright ASBMB, 2015.

#### *Technical Challenges and Breakthrough Technology in DNA Recovery from FFPE Tissues*

FFPE tissues are now widely used in high throughput genomic [64,68–71], proteomics [72–74], and to a lesser extent, metabolomics studies [75,76]. The crucial step in these applications is the quantitative extraction of the molecules of interest. In genomic sequencing studies, the conventional method of DNA isolation from FFPE tissues has often employed elevated temperature (up to 100 ◦C) and alkaline pH (>9) to achieve a complete reversal of crosslinks. Many automated methods employed in cancer genomics still use elevated temperature to isolate DNA from FFPE tissues. The recovered DNA can serve as a template for PCR amplification. However, these harsh conditions can cause oxidation of nucleobases or depurination of chemically-modified nucleobases, and thus, are not compatible for quantitative measurements of DNA adducts. Moreover, even though formalin-fixation is the most common method of tissue preservation world-wide, the conditions of fixation can vary in different laboratories. A prolonged time of tissue preservation in formalin results in over-fixation of the tissue and leads to inefficient hydrolysis of crosslinks between DNA and protein. Therefore, the yield and quality of the recovered DNA is decreased [77–79]. Thus, the development of robust analytical methods to quantitatively recover DNA adducts from FFPE tissue has been a challenging endeavor.

#### **3. Measurement of DNA Adducts in FFPE Tissues by IHC, 32P-Postlabeling, and LC-MS**

#### *3.1. IHC Detection of DNA Adducts*

FFPE specimens are often used for immunodetection of DNA adducts, most commonly by IHC methods [30,80]. In contrast to mass spectrometry-based methods, which break down the DNA to the mono 2 -deoxyribonucleoside or DNA base (*vide supra*), IHC methods employ intact DNA. The detection of DNA damage can be carried out on either fixed cells such as lymphocytes, exfoliated oral or bladder cells, or with FFPE tissue section-cuts. The cells or FFPE tissue section-cuts are mounted

on glass slides for IHC analysis. Procedures are often used to increase the accessibility of the antibody to the carcinogen DNA adduct to increase the sensitivity of the assay. These procedures can include treatment with proteases to remove histone and other proteins from the DNA, followed by treatment with RNase to eliminate potential cross-reactivity with RNA adducts. Mild acid or base treatment also may be performed to denature the DNA and further increase the accessibility of the antibody to the adduct. It is imperative that the adduct is stable to the denaturing treatment conditions for validation of the IHC technique. The two most commonly used detection systems for visualization of DNA adduct-antibody complexes are immunofluorescence or chromophores, where the secondary antibody is tagged with a chemically conjugated fluorophore, a peroxidase or alkaline phosphatase enzyme. [30,81].

Table 1 summarizes examples of IHC detection of DNA adducts in FFPE tissues. Santella's group detected and quantified DNA adducts of 4-aminobiphenyl (4-ABP), an aromatic amine and a human bladder carcinogen that is formed in tobacco smoke [21,22], and also occurs as a contaminant in some commercial hair dyes [82]. 4-ABP-adducted DNA was detected in uroepithelium of bladder cancer patients [83]. The level of the 4-ABP adduct was correlated with the smoking status and *p53* overexpression, a response to DNA damage. There was linear relationship between the relative degree of DNA adduct staining and the number of cigarettes smoked. The same group also detected DNA adducts of polycyclic aromatic hydrocarbons (PAHs) in archived breast tissues sections using polyclonal antiserum [84,85]. PAHs are incomplete combustion products of organic matter and found in cereal and grain products, some oils, and also found in charred meat and tobacco smoke [42]. PAHs have been linked to human cancers at multiple sites [21]. The most well studied PAH is benzo[*a*]pyrene (B[*a*]P), a human lung carcinogen found as an environmental pollutant, and it also occurs in tobacco smoke, and charred meat [42,86]. The Poirier laboratory developed an antibody raised against DNA modified with *r*7,*t*8-dihydoxy-*t*-9,10-oxy-7,8,9,10-tetrahydrobenzo[*a*]pyrene (BPDE) [87], which later was shown to cross-react with other structurally similar diol-epoxide-PAH-DNA adducts [88]. This PAH-DNA antiserum has been used to screen for DNA adducts in FFPE tissues from human esophagus [81,89], prostate [90], cervix [91], vulva [47], and placenta [80]. A significantly higher level of staining of presumed B[*a*]P adducts was found in benign breast disease in comparison to the cancerous tissues of patients, possibly due to cellular proliferation and dilution of the adduct in cancerous tissue [84,85]. Rundle et al. employed IHC to measure PAH-DNA adducts and examined the associations with alcohol consumption and the influence of GSTM1 genotype on DNA adduct formation in FFPE breast tissues [92]. Subjects harboring the GSTM1-null genotype, which lacks the expression of GTSM1, an enzyme that detoxicates PAH diol-epoxides [93], had increased levels of DNA adducts among current alcohol consumers, but not among nondrinkers. In contrast, in benign tissues from controls, no association was observed between genotype and adduct levels, regardless of drinking status. Poirier also analyzed tamoxifen-DNA adducts in rat hepatocytes by IHC [94]. A steady increase in adduct levels was observed with chronic exposure.

Shirai et al. developed polyclonal antibodies against DNA adducts of 3,2 - dimethyl-4-aminobiphenyl (DMAB), an aromatic amine that induces tumors at multiple sites in rodent models, and PhIP, a probable human carcinogen formed in cooked meat that induces tumors in colorectum and prostate of rodents [95–97]. Dose-related nuclear staining was observed in various acetone-fixed tissues of rodents 24 h after single exposure of DMAB or PhIP. Using the same polyclonal serum, putative DNA adducts of PhIP were detected, by IHC, at high frequency in mammary tissue of women with breast cancer [98] and in prostate tissue of men with prostate cancer [99]. However, these results are at odds with specific mass spectrometry-based methods, where PhIP DNA adducts were detected at considerably lower frequency and at much lower levels of DNA modification in both tissues [100,101]. The discrepancy between the estimates of the PhIP DNA adduct reported by MS and IHC methods suggest the possible cross-reactivity of the polyclonal antibodies with other DNA adducts of similar structure or endogenous cellular components. There is a need to cross-corroborate the identities and levels of DNA adducts measured by IHC and specific MS-based methods. Aoshiba

and coworkers raised antibodies against 8-hydroxy-2 -deoxyguanosine (8-OHdG), an oxidative DNA adduct, and 4-hydroxy-2-nonenal (4-HNE), a lipid peroxidation adduct, to evaluate the oxidative stress induced by cigarette smoke in paraffin-embedded pulmonary epithelial cells of mice [102]. There was a dramatic increase in the intensity of signals one hour post cigarette smoke exposure, compared to pre-exposure, which confirmed the causal role of cigarette smoking in oxidative damage to respiratory epithelium.



<sup>a</sup> Adduct levels were reported as relative nuclear stain intensity; <sup>b</sup> Adduct levels were reported as a percentage of positive cells; <sup>c</sup> LOD was reported in the citation, which was one per 1010 nucleotides employing 10 μg DNA. NR: Not reported.

#### *3.2. DNA Measurements in FFPE Tissues by 32P-Postlabeling*

There is only one report employing 32P-postlabeling to detect DNA adducts in FFPE tissues [103]. In that study, rat tissues were fixed in formalin and embedded in paraffin after dosing with B[*a*]P or 2-acetylaminofluorene (2-AAF). DNA was extracted from fixed tissues using a modified phenol-chloroform method [106]. The levels of DNA adduct recovered from FFPE tissues were significantly lower than the levels obtained from fresh frozen tissues. The authors concluded that FFPE tissues could be used to screen for DNA adducts but that adduct levels may be underestimated particularly with prolonged time of fixation in formalin.

#### *3.3. Measurement of DNA Adducts in FFPE Tissues of Rodents and Human by LC-MS*

The physio-chemical data provided by MS for proof of DNA adduct structure combined with the robust quantitation and high sensitivity makes MS the technique of choice for DNA adduct biomarker measurements. The DNA adducts must be stable towards both the formalin fixation and DNA retrieval processes. Furthermore, the DNA must be fully digestible by nucleases to monodeoxyribonucleosides. Until recently, the recovery of high quality DNA completely devoid of formalin crosslinks was difficult to achieve under mild hydrolysis conditions. However, commercial kits from several vendors now employ mild retrieval conditions at neutral pH to reverse the crosslinks of FFPE DNA. The DNA recovered was shown to be successfully employed as templates for amplification by PCR. We tested commercial kits from several vendors and found the FFPE miniprep kit from Zymo Research (Irvine, CA, USA), with some modifications in manufacturer's protocol, provided high quality DNA that was fully digestible by nucleases [79,101,105].

Our laboratory established a method to isolate DNA from FFPE liver and kidney tissues of C57BL/6J mice, using aristolochic acid I (AA-I) as the model carcinogen [79,104]. AA-I is an upper urinary tract human carcinogen found in Aristolochia plants, some of which have been used in traditional Chinese herbal medicines [79,104]. DNA was isolated from freshly frozen tissue by the phenol-chloroform method, and DNA from FFPE tissue was isolated with the FFPE miniprep kit (Zymo Research). AA-I DNA adducts were measured by ultra-performance liquid chromatography-electrospray ionization-ion trap-multistage MSn scanning (UPLC-ESI-IT-MS3). Across all dosing levels, the amounts of AA-I DNA adduct in DNA from FFPE tissues were comparable to those of matching freshly frozen tissues (Figure 4) [104].

**Figure 4.** Mean level of dA-AL-I adducts present in mouse kidney and liver following treatment with AA-I (0.001−1 mg/kg body weight). Adduct levels measured in freshly frozen and FFPE mouse kidney ( and ) and liver ( and ) (mean adduct level, SD, *N* = 4 animals per dose, quadruplicate measurements per animal) were plotted as a function of dose. The overall mean difference in adduct levels between freshly frozen and FFEP kidney and liver tissues across all doses was 21 ± 14% (mean ± SD). dA-AL-I adduct formation was below the limit of detection in liver of mice dosed with AA-I at 0.001 mg/kg body weight. Mean levels of dA-AL-I adducts were significantly statistically different between freshly frozen and FFPE kidney or liver at the following dose treatments of AA-I: kidney, 1 mg/kg, *p* = 0.03; liver, 0.1 mg/kg, *p* = 0.01; unpaired two-tailed *t*-test. Reproduced with permission from [104]. Copyright ACS, 2013.

Then, we examined the effect of duration of formalin fixation on the recovery of DNA and the level of DNA adducts in rodents treated with AA-I [79]. The yield of DNA retrieved from formalin-fixed tissues decreased as a function of fixation time, and only 30% of DNA was recovered from FFPE tissues after one week of fixation in formalin compared to the freshly frozen tissues. However, the DNA retrieved was completely digested by nucleases and the levels of AA-I DNA adduct were relatively constant between the freshly frozen tissues and FFPE tissues. DNA fragments of 184 and 327 bp extracted from FFPE tissues were readily amplified by PCR, and the quality of sequence data was comparable to that obtained from DNA obtained of fresh frozen tissues [79]. Our findings demonstrate that the DNA can be recovered from FFPE tissue to analyze DNA adducts of AA-I in FFPE tissue, and adducts of AA-I or other carcinogens may be correlated with mutational signatures induced in tumor tissue.

Thereafter, we sought to determine if our method of DNA adduct retrieval from FFPE tissues could be employed to measure DNA adducts of other environmental and dietary genotoxicants. We examined DNA adducts of four important classes of environmental and dietary carcinogens: PAHs (B[*a*]P), aromatic amines (4-ABP), HAAs (PhIP), and *N*-nitroso compounds 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NNK), which is found in tobacco and a lung carcinogen [21,107]. The major DNA adducts of these carcinogens studied were: 10-(2 -deoxyguanosin-*N*2-yl)-7,8,9-trihydroxy-7,8,9,10-tetrahydrobenzo[*a*]pyrene (dG-*N*2-B[*a*]PPDE), *N*-(2 -deoxyguanosin-8-yl)-4-ABP (dG-C8-4-ABP), *N*-(2 -deoxyguanosin-8-yl)-PhIP (dG-C8-PhIP), O6-[4-oxo-4-(3-pyridy)-butyl]-2 -deoxyguanosine (*O*6-POB-dG), *O*6-methyl-2 -deoxyguanosine (*O*6-methyl-dG) (Figure 5) [105]. All of these adducts and dA-AL-I were measured by UPLC-ESI-IT-MS<sup>3</sup> with the stable isotope dilution method. The levels of DNA adducts in FFPE tissues of rodents preserved in formalin for 24 h were at levels comparable to those levels measured in freshly frozen tissues [105].

**Figure 5.** Structures, names, and abbreviations of carcinogens and their adducts used for quantitation of multiclass carcinogenic DNA adducts in freshly frozen and FFPE tissues of rodents. Reproduced with permission from [105]. Copyright ACS, 2016.

*Toxics* **2018**, *6*, 30

The recent improvements in sensitivity of mass spectrometry instrumentation has allowed us to use only 10 to 20 mg of tissue to screen for DNA adducts of environmental and dietary carcinogens in human biopsy samples [100,101,104]. We sought to determine if DNA extraction kits devoted to genomics, such as the FFPE miniprep kit from Zymo Research, could be employed to screen for DNA adducts in human FFPE biospecimens. We applied the method of DNA isolation to assay tissue section-cuts of human FFPE kidney specimens (1.5 cm<sup>2</sup> × <sup>10</sup> <sup>μ</sup>m) from the patients with upper urinary tract carcinoma, who were exposed to AA-I [79,104]. The levels of AA-I DNA adduct measured in FFPE tissues were comparable to those of matching frozen tissues (Figure 6). Some of these FFPE blocks had been stored at room temperature for four to nine years. This was the first report of quantitative measurement of a carcinogen DNA adduct in human FFPE tissue by mass spectrometry. We subsequently showed that DNA adducts of PhIP can be recovered in high yield from human FFPE prostate tissue blocks of prostate cancer patients stored at room temperature for at least 6 months (Figure 7) [101,108]. These findings show that FFPE tissues can be used to retrospectively screen for multiple classes of carcinogen DNA adducts.

**Figure 6.** Extracted ion chromatograms of dA-AL-I from human kidney cortex of patients with upper urothelial cancer from Taiwan at levels (**A**) below the LOQ, and positive samples at (**B**) 0.4 adducts, and (**C**) 5.9 adducts per 10<sup>8</sup> bases. The product ion spectra of dA-AL-I obtained from panel C is depicted in (**D**) along with the internal standard [15N5]-dA-AL-I (E, 15N labels of the internal standard of dA-AL-I are depicted with asterisks). Insert (**F**) dA-AL-I adduct levels in matching fresh frozen and FFPE kidney samples, containing both renal cortex and medulla, obtained from 11 individuals residing in endemic regions of Croatia and Serbia who underwent nephroureterectomy for upper urothelial cancer. Reproduced with permission from [104]. Copyright ACS, 2013.

**Figure 7.** Extracted ion chromatograms of dG-C8-PhIP and 13C-labeled dG-C8-PhIP of DNA from fresh frozen and FFPE human prostate tissues at the MS3 scan stage. (**A**) Fresh frozen prostate and (**B**) paired FFPE block of a patient who was negative for dG-C8-PhIP; (**C**) fresh frozen prostate and (**D**) paired FFPE block of a patient who was positive dG-C8-PhIP at MS<sup>3</sup> scan stage. The structure and proposed fragmentation mechanism of aglycone of dG-C8-PhIP are depicted fresh frozen prostate and (**D**) paired FFPE block of a patient who was positive for dG-C8-PhIP. (**E**) The product ion spectra at MS3 of unlabeled and 13C-labeled dG-C8-PhIP are shown. (**F**) Levels of dG-C8-PhIP in paired fresh frozen prostate and FFPE blocks of six patients are shown in (**G**). The levels of adducts are reported as adducts per 10<sup>8</sup> nucleotides. \* *p* < 0.05; n.s., statistically not significant. Reproduced with permission from [108]. Copyright ACS, 2017.

#### *3.4. Rapid Throughput Method of DNA Extraction from FFPE Tissue*

The method of DNA isolation from FFPE tissues employing the FFPE miniprep kit (Zymo Research) is robust; however, it is a manual and labor-intensive technique and cannot facilely process the large number of samples required for epidemiological studies. We developed a rapid throughput method of DNA isolation from FFPE tissue employing a semi-automated commercial DNA isolation system, Promega Maxwell**®** 16 MDx system, which is used for genomic research [108]. The system employs silica-magnetic beads technology for DNA isolation and can process 32 samples per hour compared to 4–6 samples per hour by the manual method. The DNA recovered from FFPE tissues using the Promega Maxwell**®** 16 MDx is fully digestible by nucleases [108]. The levels of dA-AL-I, dG-C8-4-ABP, and dG-C8-PhIP recovered from DNA of FFPE tissues extracted by rapid throughput method are comparable to those levels measured from DNA isolated by phenol-chloroform in matching frozen tissues, and in DNA of FFPE tissues isolated by the commercial manual Zymo kit [108]. With this advancement in DNA isolation technology, we believe that archived FFPE tissues can be used to screen for DNA adducts in large population studies. A scheme and the time of duration of the procedure to isolate DNA from FFPE section cuts or whole tissues, and ensuing chemical analysis by mass spectrometry, are depicted in Figure 8. The recovery of DNA from FFPE tissues and DNA digestion steps require overnight incubation with enzymes to achieve optimal digestion efficiency. The targeted and simultaneous quantification of a selected number of DNA adducts, by UPLC-ESI-IT-MS3, can be achieved in a 10 to 15 min run time.

**Figure 8.** Scheme of FFPE tissue processing for DNA adduct measurements. DNA is extracted from section cuts or excised whole tissues by the FFPE Miniprep kit (Zymo Research) or Promega Maxwell® automated system. After nuclease digestion, the DNA adducts are measured by UPLC-ESI-IT-MS3. The estimated times of the different processes are reported.

#### *3.5. Future Applications of DNA Adduct Measurements in Human Tissues*

Although this review has focused on DNA adducts of environmental and dietary carcinogens, the measurements of DNA adducts of chemotherapeutic agents, such as platinum drugs and nitrogen mustards used to treat cancer, also can be measured in fresh frozen and FFPE tissues by mass spectrometry. Drugs that modify the structure of DNA and target cancer cells by interfering with DNA synthesis and cell replication often remain first line of medications used in cancer treatment. Thus, the efficacy of many anticancer drugs is thought to be linked to the levels of specific DNA adducts formed during drug treatment, and the quantitative measurements of the DNA adducts may be used as predictive markers in precision medicine to identify individuals who are most likely to benefit from treatment from those patients who may be less responsive to the therapy [109]. The assessment of DNA adducts of chemotherapeutic drugs and their cellular biological responses has been mostly performed in surrogate specimens, such as white blood cells or in vitro using cell lines rather than in the target cells or tumors, because of the invasiveness of biopsy sampling [109]. However, the exquisite sensitivity of current mass spectrometry instrumentation can allow for measurements that characterize the relationships between level of anticancer drug DNA adducts and pharmacodynamic response in patients using only 10 mg of tissue. As the sensitivity of MS instrumentation continues to improve, the amount of tissue specimen required for analysis will be further reduced, and the application of DNA adduct monitoring of chemotherapeutic drugs in clinical settings can be achieved.

The screening of DNA from FFPE tissues shows great promise to measure DNA adducts of multiple classes of carcinogens and anticancer drugs [37,55,105,109]. While most analyses have focused on one to several adducts, different types of MS scanning approaches are being developed to simultaneous scan for multiple types of DNA adducts in the field of DNA adductomics [63]. Triple quadrupole, quadrupole time-of-flight, ion trap or Orbitrap mass spectrometry instrumentation are being employed in DNA adductomics [55,58,62,63,110]. Our laboratory is developing unbiased non-targeted ion trap and Orbitrap scanning methods to screen for an array of DNA adducts in the

human genome in a single assay [55,58]. Some of these adducts are expected to contribute to the tumor mutation burden [111].

A critical need is the development of accompanying informatic tools for data analysis and statistical tools to screen for covalent DNA damage. These scanning technologies and accompanying data analysis tools will provide a wealth of information about the exogenous and endogenous chemicals that damage the genome and may contribute to cancer risk. The implementation of FFPE tissues in DNA adduct biomarker discovery can provide the clues about the origin of human cancers for which an environmental exposure is suspected.

**Author Contributions:** All authors critically reviewed the literature and contributed to drafting the manuscript.

**Acknowledgments:** The work cited in this review conducted by the Turesky laboratory has been supported by Grants R01ES019564 and R21ES014438 from the National Institute of Environmental Health Sciences, R01CA122320, R01CA220367, and R33CA186795, and the National Center for Advancing Translational Sciences award number UL1TR000114 from the National Cancer Institute of the National Institutes of Health. Mass spectrometry was carried out in Analytical Biochemistry Shared Resources of the Masonic Cancer Center, University of Minnesota, funded in part by Cancer Center Support Grant CA-077598.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Recent Studies on DNA Adducts Resulting from Human Exposure to Tobacco Smoke**

#### **Bin Ma \*,†, Irina Stepanov and Stephen S. Hecht**

Masonic Cancer Center, University of Minnesota, Minneapolis, MN 55455, USA; stepa011@umn.edu (I.S.); hecht002@umn.edu (S.S.H.)

**\*** Correspondence: bma@umn.edu or bin.ma@gilead.com; Tel.: +1-612-625-4925; Fax: +1-612-624-3869

† Current address: Gilead Sciences, 333 Lakeside Dr, Foster City, CA 94404, USA.

Received: 2 February 2019; Accepted: 13 March 2019; Published: 19 March 2019

**Abstract:** DNA adducts are believed to play a central role in the induction of cancer in cigarette smokers and are proposed as being potential biomarkers of cancer risk. We have summarized research conducted since 2012 on DNA adduct formation in smokers. A variety of DNA adducts derived from various classes of carcinogens, including aromatic amines, polycyclic aromatic hydrocarbons, tobacco-specific nitrosamines, alkylating agents, aldehydes, volatile carcinogens, as well as oxidative damage have been reported. The results are discussed with particular attention to the analytical methods used in those studies. Mass spectrometry-based methods that have higher selectivity and specificity compared to 32P-postlabeling or immunochemical approaches are preferred. Multiple DNA adducts specific to tobacco constituents have also been characterized for the first time in vitro or detected in vivo since 2012, and descriptions of those adducts are included. We also discuss common issues related to measuring DNA adducts in humans, including the development and validation of analytical methods and prevention of artifact formation.

**Keywords:** DNA adducts; tobacco smoke; human carcinogen; biomarkers; cancer risk; mass spectrometry

#### **1. Introduction**

Cigarette smoking causes multiple types of cancers. Despite the advent of advanced cancer genomics and impressive targeted therapies, there were still an estimated 9.6 million cancer deaths worldwide in 2018 with 22% of those deaths being caused by cigarette smoking [1]. In the U.S., where lung cancer is the leading cause of cancer death in both men and women, an estimated 154,000 lung cancer deaths occurred in 2018, with 90% being caused by cigarette smoking [2,3]. Not all smokers develop cancer. While 90% of all lung cancer-related deaths in the U.S. are attributable to cigarette smoking, only 24% of male smokers and 11% of female smokers may die from lung cancer over their lifetime, assuming that there is no competing cause of death [4,5]. A significant research challenge is to identify those smokers who have higher cancer risk, so prevention approaches can be initiated at an early stage before too much damage has been done.

One strategy to address this challenge is to identify and validate smoking-related biomarkers associated with tobacco exposure and cancer risk. We have demonstrated the association of certain urinary tobacco smoke biomarkers with lung cancer risk, independent of smoking intensity and duration. These biomarkers include cotinine and its glucuronide—metabolites of nicotine; 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol (NNAL) and its glucuronides—metabolites of a tobacco specific lung carcinogen 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NNK); and *r*-1-,t-2,3,c-4-tetrahydroxy-1,2,3,4-tetrahydrophenanthrene (PheT), a biomarker of polycyclic aromatic hydrocarbon (PAH) exposure and metabolic activation [6,7]. Similarly, levels of a urinary biomarker of exposure to another tobacco-specific carcinogen -*N* -nitrosonornicotine (NNN)- have been independently and prospectively associated with the risk of esophageal cancer in smokers [8]. In addition, low activity forms of cytochrome P450 2A6, the major nicotine metabolizing enzyme, have been related to decreased lung cancer risk in cigarette smokers [9]. All of these parameters are related to carcinogen uptake by smokers, but do not provide information on the next critical step in the carcinogenic process—DNA adduct formation. DNA adducts associated with tobacco smoke exposure have the potential to significantly contribute to our understanding of the cancer process and could possibly be biomarkers of cancer risk.

DNA adducts are compounds formed when chemicals react with DNA. Figure 1 provides an overview of the central role of DNA adduct formation in tobacco-related cancer. Tobacco smoke contains a highly complex mixture of over 7000 characterized chemical compounds; certain chemicals such as formaldehyde and acetaldehyde are reactive and directly bind to DNA, while some others, including NNK, NNN, and PAH require metabolic activation to reactive intermediates capable of reacting with DNA. The resulting covalent DNA addition products are commonly referred to as "DNA adducts" (Figure 1). Because of the DNA repair enzymes present in the body, the DNA adducts can be removed and the DNA returned to normal. However, if the repair process is overwhelmed or not completely efficient, the DNA adducts can persist and potentially cause miscoding during DNA replication. If the miscoding events occur in critical genes such as *p53* and *RAS*, they can result in the loss of normal growth control mechanisms, and eventually, the development of cancer. Therefore, DNA adduct formation plays a critical role in tobacco smoke-related cancer development. The general protocol to measure DNA adducts includes isolating DNA from biological samples, hydrolyzing the DNA using a cocktail of enzymes, purifying the hydrolyzed samples and enriching the DNA adducts, and then analyzing them by various techniques such as mass spectrometry. The ultimate goal is to quantify these tobacco smoke-related DNA adducts in human biological samples and to incorporate the results into epidemiologic studies to evaluate and validate their role in cancer risk prediction (Figure 1). While this challenging goal has yet to be reached, constantly improving methods for DNA adduct quantitation, as discussed here, provide at least a potential path. We also note that reactive tobacco chemicals can modify other macromolecules such as protein, which can also result in significant biological consequences and can be used in bio-monitoring [10,11].

In 2002, Phillips reviewed previous studies on smoking-related DNA adduct formation in human tissues [12]. A large number of studies mentioned in that review used 32P-postlabeling or immunochemical approaches for the analysis of DNA adducts. In many cases, "bulky DNA adducts" were reported but it was not clear which specific DNA adducts were being measured. Nevertheless, the effects of smoking were still evident by the detection of elevated levels of DNA adducts in many human tissues [12]. In 2012, Phillips updated this topic and reviewed studies on tobacco smoke-related DNA adducts published since 2002 [10]. That review covered the chemical nature and origins of smoking-related DNA adducts and the effect of DNA repair and gene polymorphisms on the levels of DNA adducts in humans. Many studies cited in that review demonstrated convincing evidence of higher levels of certain adducts in smokers than nonsmokers. In the same published issue as Phillips' paper, we also reviewed and discussed structurally characterized DNA adducts in the lungs of smokers and their potential role in lung carcinogenesis by tobacco smoke [13].

**Figure 1.** An overview of the central role of DNA adduct formation in tobacco-related cancer. DNA damage also leads to cell apoptosis. A more detailed mechanistic framework was illustrated by Hecht S.S. [13].

This review covers 1) studies on formation of tobacco smoke-related DNA adducts in humans (Figure 2) since Phillips' review in 2012; and 2) newly characterized DNA adducts formed by tobacco-specific chemicals (i.e., NNK and NNN) in in vitro and in vivo models (Figure 3). Future directions of applying DNA adducts as biomarkers of tobacco smoke exposure and associated cancer risk are also discussed.

#### **2. Smoking-Related DNA Adducts in Humans**

#### *2.1. Tobacco-specific Nitrosamines (TSNA)*

The most studied TSNA are NNK and NNN, which are considered "carcinogenic to humans" by the International Agency for Research on Cancer [14]. NNK causes lung tumors in all species tested, independent of the route of administration [15]. NNN induces tumors of the oral cavity, esophagus, and nasal cavities in rats and respiratory tract tumors in mice, mink, and Syrian golden hamsters [15]. NNK and NNN are present in all tobacco products. Both compounds are specific to tobacco and are not found in any other product, except that NNN can also be formed endogenously from nornicotine, a minor tobacco alkaloid and a nicotine metabolite [16–18].

NNK and NNN are metabolically activated to intermediate **14** (Figure 3), which reacts with DNA to form pyridyloxobutyl (POB) adducts at the four nucleobases and the oxygens of the phosphate backbone (Figure 3) [7,19–22]. Under acid or neutral thermal hydrolysis, most of the POB base adducts release 4-hydroxy-1-(3-pyridyl)-1-butanone (HPB, Figure 2). Multiple in vitro and animal studies demonstrated the relationship of HPB-releasing DNA adducts to NNK and NNN dose as well as biomarkers of NNK exposure [15]. HPB-releasing DNA adducts were also detected by our group and others in human tissues including lung, tracheobronchus, esophagus, and cardia [23–25]. We demonstrated for the first time the presence of HPB-releasing DNA adducts in human lung, with adduct levels being higher in smokers than nonsmokers [25]. Similar results were observed by Schlobe, et al, with HPB levels in the lung being significantly higher in 21 self-reported smokers compared to that in 11 self-reported nonsmokers [24]. However, another study by the same group showed no difference of adduct levels between smokers and nonsmokers [23]. All these studies had a relatively small sample size, probably due to the limited availability of human tissues, which had to be obtained

by surgery. Obtaining these human tissues from individuals for smoking exposure evaluation and HPB-releasing DNA adduct analysis is highly impractical in larger studies. Surrogate tissues, which can be less invasively obtained, should be explored in such studies.

Exfoliated oral mucosa cells are relatively simple and noninvasive to be collected, and could be an excellent source of surrogate tissue for evaluating smoking exposure and adduct formation [26,27]. We developed a robust and sensitive liquid chromatography (LC)−electrospray ionization (ESI)−tandem mass spectrometry (MS/MS) method for the analysis of HPB-releasing DNA adducts in human oral cells (Table 1) [28]. Oral cells were collected from 30 smokers and 15 nonsmokers using a mouthwash rinse or buccal brushing with a cytobrush. In the oral cells collected by mouthwash, HPB was detected in 20 out of 28 smoker samples with quantifiable DNA yield, and in 3 out of 15 nonsmoker samples. Higher levels of HPB-releasing DNA adducts were observed in smokers with an average of 12 pmol adducts/mg DNA, compared to nonsmokers with an average of 0.23 pmol adducts/mg DNA. In the buccal brushing samples collected from the same 30 smokers, HPB was detected in 24 out of 27 samples with quantifiable DNA yield, averaging 45 pmol adducts/mg DNA. A correlation of adduct levels (*R* = 0.73, *p* < 0.0001) was also observed between buccal burshings and mouthwash samples from smokers [28]. The LC−ESI−MS/MS method we developed in that study demonstrated the applicability to the analysis of oral cell samples collected by mouthwash or buccal brushing. However, application of this method in studies in which only limited oral cells are available for DNA extraction required further optimization to increase its sensitivity and selectivity.

**Figure 2.** Structures of DNA adducts detected in humans. dR, 2 -deoxyribose; HPB, 4-hydroxy-1-(3-pyridyl)-1-butanone.

**Figure 3.** NNK and NNN-derived DNA base and phosphate adducts. NNK and NNN form various DNA adducts, and only newly characterized adducts or adducts detected for the first time in vivo since 2012 are presented. NNN is also metabolized via 2'-hydroxylation to intermediate **14**, which reacts with DNA to form POB base adducts. B1 and B2 represent the same or different nucleobases. NNK, 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone; NNN, *N*-nitrosonornicotine; POB, pyridyloxobutyl; dR, 2-deoxyribose.

We then modified the method by optimizing the sample preparation procedure, and more notably by switching from LC−ESI−MS/MS to LC-nanoelectrospray ionization (NSI)−high-resolution tandem mass spectrometry (HRMS/MS) [29]. The LC-NSI-HRMS/MS-based methods for DNA adduct analysis have been very successful in our studies of other DNA adducts [20,30–36]. Using the optimized method, the sensitivity was improved seven-fold compared to the previous LC−ESI−MS/MS method, achieving a limit of quantitation (LOQ) of 0.35 fmol on-column. More importantly, the baseline noise commonly observed in mass spectrometry without high resolution capability was completely eliminated, greatly improving the selectivity of the method. We then applied this method to the analysis of HPB-releasing DNA adducts in oral cells collected by buccal brushing from 65 smokers, including 30 head and neck squamous cell carcinoma (HNSCC) patients and 35 cancer-free controls. HPB-releasing DNA adducts were detected in 29 out of 30 samples of HNSCC patients averaging 8.2 pmol adducts/mg DNA, compared to 20 out of 35 samples of cancer-free smokers averaging 4.5 pmol adducts/mg DNA. The median HPB-releasing DNA adduct level was 6.6 times greater for smokers with HNSCC than for those without HNSCC (*p* = 0.002) [29]. These results suggest that HPB-releasing DNA adducts may play a critical role in the development of smoking-induced HNSCC, and can potentially be used to identify susceptible smokers.

#### *2.2. Bulky/Aromatic Adducts*

The terminology "bulky DNA adducts" comes from early studies using a 32P-postlabeling approach to measure DNA adducts formed by high molecular weight chemical carcinogens, including PAH and probably some other aromatic and nonpolar chemicals [37,38]. PAH are formed primarily by incomplete combustion of tobacco and other organic components during smoking. PAH are also released from burning coal, oil, gasoline, wood, and are present in air, soil and water. Studies on bulky DNA adducts, sometimes referred to as PAH-DNA adducts, were comprehensively covered by Phillips et al. in 2002 [12] and 2012 [10]. The overall trend was that higher adduct levels were observed in smokers compared to nonsmokers. Since 2012, continuous efforts have been made to measure bulky DNA adducts in human samples including lung and leukocytes (Table 1) [39–52]. Smoking is a contributor to adduct formation in these studies. However, since the 32P-postlabeling technique was exclusively used as the detection method, structural information for these adducts was not available, which prevents one from working backwards to evaluate the exposure to responsible chemicals.

One of the PAH adducts is BPDE-*N*2-dG (Figure 2), which is formed by BPDE, a metabolite of the most studied carcinogenic PAH—benzo[*a*]pyrene (BaP). In humans, BPDE-*N*2-dG was reported to be present in buccal cells [53], sputum [53], lung [53,54], leukocytes [55], lymphocytes [56], tonsil [57], oropharynx [57], gingiva [57], and prostate [58] using immunochemical methods. Another method to measure this adduct is to hydrolyze the DNA in acid and analyze BaP-7,8,9,10-tetraols instead of the intact adducts by LC-fluorescence or gas chromatography/mass spectrometry methods [59,60]. However, the specificity of these methods, although in some cases well designed and validated, has to be considered when interpreting the data. This concern was further highlighted by studies using mass spectrometry-based methods [61–64]. In one study analyzing human lung tissues, BPDE-*N*2-dG was not detected in any of the 10 samples with a limit of detection (LOD) of one−three adducts per 108 nucleotides [61], and the second study showed detectable adducts in only 1 out 26 human lung DNA samples (LOD, 0.3 adducts per 108 nucleotides) [62]. BPDE-*N*2-dG was not detected in salivary or oral cell DNA in another two studies [63,64]. Finally, a recent study failed to detect this adduct in human prostate tissue [65].

We recently developed an ultrasensitive LC-NSI-HRMS/MS method for the analysis of BPDE-*N*2-dG in human lung DNA, with an LOD of 1 adduct per 1011 nucleotides, which is equivalent to 1 adduct per 10 human lung cells (Table 1) [32]. To our knowledge, this is the most sensitive DNA adduct quantitation method yet reported. We applied this method to the analysis of lung samples obtained during surgery for lung cancer from 29 patients. Another unique feature of this study was confirmation of smoking status at the time of lung cancer surgery by the measurement of urinary cotinine and NNAL. BPDE-*N*2-dG was detected in 20 out of 29 samples, with smoker and nonsmoker

DNA containing 3.1 and 1.3 adducts per 1011 nucleotides, respectively. The adduct levels in our study are in contrast to those obtained by the immunochemical approach, with levels typically ranging from 1–30 adducts per 108 nucleotides in lung tissues [53,54]. Compared to mass spectrometry-based methods including our study, significantly higher levels (~1,000-fold) of adducts were observed by immunochemistry-based methods. Although the lung samples analyzed in these studies were from different subjects, the significantly high level observed by immunochemistry-based methods suggests that those methods should be re-evaluated to avoid false-positive results, and the data obtained from those studies should be interpreted with caution. Studies should be conducted using a mass spectrometry-based method and an immunochemistry-based method to analyze the same human lung samples, to compare the levels of BPDE-*N*2-dG adducts assessed by the two methods.

#### *2.3. Aromatic and Heterocyclic Aromatic Amines*

Aromatic and heterocyclic aromatic amines are structurally related classes of chemicals that are present in tobacco smoke, formed in cooked meats, and occur as contaminants in the atmosphere [38,66]. One such compound, an aromatic amine 4-aminobiphenyl (4-ABP) is a known bladder carcinogen in both animals and humans [66]. The major DNA adduct formed by 4-ABP is 4-ABP-C8-dG, and to a lesser extent, 4-ABP-C8-dA and 4-ABP-*N*2-dG (Figure 2) [67].

In a study investigating the mutagenicity of 4-ABP-derived DNA adducts, the three adducts were measured in both normal human urothelial mucosa and bladder tumor tissues by 32P-postlabeling-based methods (Table 1) [68]. All three adducts were detected in the urothelial mucosa samples (*n* = 19) with levels ranging from 2.3–12, 4.6–16, and 2.3–12 adducts per 10<sup>8</sup> nucleotides for 4-ABP-C8-dG, 4-ABP-C8-dA, and 4-ABP-*N*2-dG, respectively. The three adducts were also detected in bladder tumor samples (*n* = 10) with levels ranging from 2.3–9.2, 4.6–25, and 2.3–28 adducts per 108 nucleotides for 4-ABP-C8-dG, 4-ABP-C8-dA, and 4-ABP-*N*2-dG, respectively. No significant difference was observed in either total adduct levels or individual adduct levels between normal human urothelial mucosa and bladder tumor tissue [68].

In contrast to the results obtained by 32P-postlabeling methods, studies using mass spectrometry-based methods showed much lower levels or no detection of 4-ABP adducts in human samples [51,65,67]. Xiao, et al developed and validated an LC-NSI-HRMS/MS method for the analysis of 4-ABP-C8-dG with an LOQ of 0.22 adducts per 108 nucleotides using 2.5 μg DNA. The method was applied to the analysis of normal tumor-adjacent prostate tissues from 35 prostate cancer patients. 4-ABP-C8-dG was detected at a level of 2.8 adducts per 10<sup>8</sup> nucleotides in only one out of 35 samples and that one patient was identified as a nonsmoker [65]. The same research group subsequently applied this method to the analysis of normal tumor-adjacent bladder mucosa tissues from bladder cancer patients. 4-ABP-C8-dG was detected in 12 out of 41 samples with levels ranging from 0.14 to 3.4 adducts per 10<sup>8</sup> nucleotides. 4-ABP-C8-dA and 4-ABP-*N*2-dG were also detected in the same samples of two subjects who had the highest levels of 4-ABP-C8-dG [67]. In another study, Gu, et al applied an LC-MS/MS<sup>3</sup> method to the analysis of tumor-adjacent normal mammary tissues from 70 breast cancer patients. No 4-ABP-C8-dG was detected in any of the samples with an LOQ of 3 adducts per 108 nucleotides [51]. In comparison, an earlier study reported detecting 4-ABP adducts in all the 55 breast tissues by an immunochemical method [69], even though its sensitivity was ~100-fold lower compared to the LC-MS/MS<sup>3</sup> method.

#### *2.4. Methylating Agents*

There are several methylating agents present in tobacco products, including the tobacco-specific nitrosamine NNK, *N*-nitrosodimethylamine (NDMA) and methyl chloride. All of them can lead to formation of methyl DNA adducts. NNK and its metabolite NNAL are metabolized to methane diazohydroxide **15** and then methyldiazonium ion (Figure 3), which reacts with DNA forming methyl DNA base adducts including 7-methylguanine (7-mG), *O*6-methyldeoxyguanosine (*O*6-mdG) (Figure 2) [15] and methyl DNA phosphate adducts [34]. The major mutagenic and toxic methyl base

adduct is *<sup>O</sup>*6-mdG, which induces primarily GC→AT transition mutations, with some other minor methyl base adducts also demonstrating mutagenicity and cytotoxicity [70]. In addition to methyl base adducts, some methylating agents also react with the DNA phosphate backbone to form methyl DNA phosphate adducts—B1pMeB2 (Figure 2, see details in Section 3.1). B1pMeB2 adducts were demonstrated to induce TT→GT and TT→GC mutations [71], inhibit RNA synthesis [72], and affect the binding affinity of DNA to other macromolecules [73].

Methyl DNA base adducts were detected in human lung tissue samples [53,74]. *O*6-mdG was measured in the tumor-adjacent normal lung tissue samples from lung cancer patients by an immunochemical approach, and no difference in adduct levels was observed between smokers (*n* = 41) and nonsmokers (*n* = 13) (Table 1) [53]. In another study, 7-mG was detected by an immunochemical approach in lung samples from 14 former and 6 current smokers undergoing surgery for lung cancer [74]. The lung tissues were collected from five different positions of the lung including central bronchus, lung periphery, and three equidistant points along its length. The levels of 7-mG in all the samples averaged 0.75 ± 0.57 adducts per 10<sup>6</sup> dG. No significant difference in adduct levels was observed at different lung positions. Compared to former smokers, the levels of 7-mG were higher (*p* = 0.047) in current smokers at two lung positions including the lung periphery [74].

Multiple methyl DNA base adducts were also detected in human urine samples [75–78]. Wang, et al developed a capillary LC-HRMS/MS method for the simultaneous analysis of 7-mG, 3-methyladenine (3-mA), and 1-methyladenine (1-mA) in human urine samples (Figure 2 and Table 1). They applied the method to the analysis of urine samples from 20 smokers and 14 nonsmokers. The levels of the three adducts were all significantly higher in smokers than nonsmokers, with the difference in 3-mA levels being the most significant (11-fold, *p* < 0.0001) [75]. Higher levels of 3-mA in smokers compared to nonsmokers were also observed in another two studies that used LC-MS/MS-based analytical methods [76,77]. In one study, the level of 3-mA was also correlated with the level of urinary NNAL in smokers (*n* = 192, *r* = 0.48, *p* < 0.001) [77]. The correlation was also observed between adduct (3-mA and 7-mG) levels and urinary NNAL in another study [78]. However, the levels of these methyl adducts detected in urine are far greater than the levels of adducts possibly formed by NNK, suggesting a contribution of methylating agents from other sources such as diet.

We have developed an ultrasensitive LC-NSI-HRMS/MS method for the analysis of methyl DNA phosphate adducts (B1pMeB2, Table 1 and Figure 2) in human lung DNA [79]. The adduct levels were measured in both tumor and adjacent normal tissues from 30 lung cancer patients, including 13 current smokers and 17 current nonsmokers, as confirmed by measurements of urinary cotinine and NNAL. Levels of total B1pMeB2 in normal lung tissues were higher (*p* < 0.05) in smokers than nonsmokers, with an average of 13 and 8 adducts per 10<sup>9</sup> nucleotides, respectively. More details on DNA phosphate adducts, including B1pMeB2, are discussed in Section 3.1.

#### *2.5. Ethylating Agents*

Studies have demonstrated the presence of direct-acting ethylating agent(s) in tobacco products and tobacco smoke, though without characterized structure(s). These agents can react with DNA forming ethyl DNA base adducts [80,81]. Compared to methyl DNA base adducts, ethyl DNA base adducts are generally more persistent in vivo with more specificity to smoking [82–84]. Certain ethyl DNA base adducts, for example *O*4-ethylthymidine (*O*4-etT) (Figure 2), are promutagenic lesions and may contribute to the initiation of hepatocellular carcinomas in animal models [85]. Ethyl DNA base adducts were reported to be present in human leukocytes [84,86], saliva [83,87], and urine samples [76,77,88].

Chen et al. developed a series of LC-NSI-MS/MS methods for the analysis of multiple ethyl DNA base adducts in human biological samples (Table 1) [83,84,86–88]. Two of their studies measured these DNA adducts in human leukocytes. In one study, three ethyl DNA base adducts, *O*2-ethylthymidine (*O*2-etT, Figure 2), *N*3-ethylthymidine (*N*3-etT), and *O*4-etT, were detected and quantified in leukocyte DNA samples from 20 smokers and 20 nonsmokers [86]. The levels of *O*2-etT, *N*3-etT and *O*4-etT in smokers were 45 ± 52, 41 ± 44, and 48 ± 54 adducts per 108 nucleotides, while nonsmokers

had significantly lower (*p* < 0.001) adduct levels at 0.19 ± 0.87, 4.1 ± 13, and 1.0 ± 2.9 adducts per 108 nucleotides, respectively. Furthermore, the level of *O*2-etT was correlated with the smoking index (number of cigarettes per day × years smoked) (*r* = 0.48, *p* < 0.05) [86]. In the other study, 3-ethyladenine (3-etA), and 7-ethylguanine (7-etG) were detected and quantified in leukocyte DNA samples from 20 smokers and 20 nonsmokers. The levels of 3-etA and 7-etG in smokers were <sup>16</sup> ± 7.8 and 9.7 ± 8.3 adducts per 108 nucleotides, significantly higher (*<sup>p</sup>* < 0.0001) than the levels in nonsmokers at 5.4 ± 2.6 and 0.3 ± 0.8 adducts per 10<sup>8</sup> nucleotides, respectively. Additionally, the levels of both adducts were correlated with the smoking index [84]. Chen et al. also measured ethyl DNA base adducts in human saliva. In one study, *O*2-etT, *N*3-etT and *O*4-etT were detected in saliva samples from 20 smokers at 5.3 ± 6.2, 4.5 ± 5.7, and 4.2 ± 8.0 adducts per 10<sup>8</sup> nucleotides, respectively, while none of the adducts were detected in saliva samples from 13 nonsmokers [83]. In another study, 3-etA and 7-etG were detected and quantified in saliva samples from 15 smokers and 15 nonsmokers. The levels of 3-etA and 7-etG in smokers were 13 ± 7.0 and 14 ± 8.2 adducts per 108 nucleotides, significantly higher (*<sup>p</sup>* < 0.0001) than in nonsmokers with adduct levels at 9.7 ± 5.3 and 3.8 ± 2.8 adducts per 10<sup>8</sup> nucleotides, respectively. In addition, the levels of 7-etG in smokers were correlated with the number of cigarettes per day (*r* = 0.76, *p* < 0.0001) and the smoking index (*r* = 0.85, *p* < 0.0001) [87].

Ethyl DNA base adducts have also been reported to be present in human urine samples (Table 1) [76,77,88]. In addition to leukocytes and saliva, Chen, et al also developed an LC-NSI-MS/MS method for the analysis of 3-etA and 7-etG in urine samples from 21 smokers and 20 nonsmokers. 3-etA and 7-etG were detected in all smokers with levels at 69 ± 29 and 19 ± 14 pg/mL urine, respectively. In nonsmokers, the adducts were detected in 16 out of 20 samples, with levels at 3.5 ± 3.8 and 2.4 ± 3.0 pg/mL urine for 3-etA and 7-etG, respectively. Higher adduct levels were observed in smokers compared to nonsmokers [88]. Similar results were also observed in another two studies that measured 3-etA and 7-etG in urine samples [76,77].

#### *2.6. 1,3-Butadiene*

1,3-Butadiene is a colorless gas that is widely used in the polymer industry and is present in cigarette smoke, the urban environment, and automobile exhaust. 1,3-Butadiene is metabolized to epoxides, which react with DNA forming DNA adducts with the most abundant one in animals being 7-(2, 3, 4-trihydroxybut-1-yl) guanine (7-THBG, Figure 2) [89]. Sangaraju et al. developed a capillary LC-HRMS/MS method for the analysis of 7-THBG in human leukocytes (Table 1) [90]. The method was first applied to the analysis of leukocyte samples from 13 smokers and 13 nonsmokers. The levels of 7-THBG were 0.82 ± 0.51 and 0.71 ± 0.53 adducts per 10<sup>8</sup> nucleotides in smokers and nonsmokers, respectively, a non-significant difference (*p* = 0.6). To further evaluate the influence of smoking on adduct formation, the adduct levels were determined in 10 individuals who participated in a smoking cessation study. The leukocyte samples were collected before smoking cessation, and 28 and 84 days afterwards. The adduct level was not significantly changed by cessation. In the same study, the influence of occupational exposure to 1,3-butadiene on adduct levels was also investigated. 7-THBG was measured in leukocyte samples from 10 workers with known exposure levels and 10 matched controls. Compared to the controls (0.31 ± 0.22 adducts per 10<sup>8</sup> nucleotides), the levels of 7-THBG were significantly elevated in the exposed workers (0.97 ± 0.38 adducts per 108 nucleotides, *<sup>p</sup>* < 0.001) [90].

#### *2.7. Acrolein*

Human exposure to the α,β-unsaturated acrolein can be from tobacco smoke, automobile exhaust, plastic waste, heated cooking oil and endogenous lipid peroxidation [91]. Acrolein reacts with deoxyguanosine in DNA to form two pairs of regioisomeric 1,*N*2-propanodeoxyguanosine adducts: (6*R*/*S*)-3-(2'-deoxyribos-1'-yl)-5,6,7,8-tetrahydro-6-hydroxypyrimido[1,2-*a*]purine-10(3*H*)one (α-OH-Acr-dG, Figure 2) and (8*R*/*S*)-3-(2'-deoxyribos-1'-yl)-5,6,7,8-tetrahydro-8-hydroxypyrimido[1,2 *a*]purine-10(3*H*)one (γ-OH-Acr-dG). Compared to γ-OH-Acr-dG, α-OH-Acr-dG is more mutagenic and predominantly induces G→T transversions.

We developed an LC-ESI-MS/MS method to analyze Acr-dG adducts in human leukocyte DNA (Table 1) and applied this method to the analysis of leukocyte DNA samples from 25 smokers and 25 nonsmokers [92]. γ-OH-Acr-dG was the predominant isomer in all samples, while α-OH-Acr-dG was detected in only one nonsmoker and two smokers. The total levels of α-OH-Acr-dG and γ-OH-Acr-dG were not different between smokers and nonsmokers, with levels averaging 0.74 ± 0.34 and 0.98 ± 0.55 adducts per 108 nucleotides, respectively [92]. In an earlier study, we also measured Acr-dG adducts in human lung DNA from smokers and nonsmokers using an LC-ESI-MS/MS method [93]. However, the smoking status of some subjects in that study was unknown, which made it difficult to interpret the role of smoking in the adduct formation. Recently, we have developed a new ultrasensitive LC-NSI-HRMS/MS method for the analysis of Acr-dG adducts in human lung DNA samples [94]. In addition to improved sensitivity, we also managed to lower the levels of artifactual formation of adducts during sample preparation and minimized the interference of artifactually formed DNA adducts in quantitation. We analyzed the levels of Acr-dG adducts in lung DNA of 19 smokers and 18 nonsmokers who underwent surgery for lung cancer. The smoking status of these subjects was confirmed by urinary total cotinine and NNAL. The levels of α-OH-Acr-dG averaged 0.86 ± 0.27 and 1.0 ± 0.47 adducts per 10<sup>8</sup> nucleotides in smokers and nonsmokers, respectively, and the levels of <sup>γ</sup>-OH-Acr-dG averaged 2.0 ± 1.4 and 1.5 ± 0.64 adducts per 10<sup>8</sup> nucleotides in smokers and nonsmokers, respectively. There was no significant difference in the levels of α-OH-Acr-dG or γ-OH-Acr-dG between smokers and nonsmokers [94]. While the sample size was relatively small, our results indicate that acrolein is not a major etiological agent for cigarette smoking related DNA damage.

Li et al. developed an LC-ESI-MS/MS-based method for the analysis of Acr-dG adducts in human saliva samples (Table 1). γ-OH-Acr-dG was detected in all samples, and no difference of adduct levels was observed between smokers (*n* = 16) and nonsmokers (*n* = 16). α-OH-Acr-dG was not detected in any of the samples [95]. γ-OH-Acr-dG as the major acrolein-derived DNA adduct was also confirmed in human lung and liver tissues in another LC-ESI-MS/MS-based study [96]. Using an immunochemical approach, Weng et al. measured γ-OH-Acr-dG in multiple human tissues including buccal cells, sputum, and lung tissue samples [53]. The levels of γ-OH-Acr-dG in buccal cells from smokers (*n* = 33) were significantly higher (*p* < 0.0001) compared to nonsmokers (*n* = 17). Similarly, the adduct levels in sputum were higher levels (*p* < 0.05) in smokers (*n* = 22) compared to nonsmokers (*n* = 8). In the same study, γ-OH-Acr-dG was also detected in noncancerous lung tissues obtained from lung cancer patients with higher levels in smokers (*n* = 41) compared to nonsmokers (*n* = 13) [53]. Compared to our study of γ-OH-Acr-dG in human lung tissues [94], the levels obtained by Weng et al. were 10–20 times higher, suggesting possible artifactual formation of γ-OH-Acr-dG, which was not investigated in their study. Similarly, a drastic difference of BPDE-dG levels was also observed between our study and the study by Weng et al. The average level of BPDE-dG in smokers' lung in our study was ~3 adducts per 10<sup>11</sup> nucleotides [32], while the average level in their study was ~1000 times higher [53]. Due to the potential significant artifactual formation of adducts in that study, its results and conclusions are questionable. Another study employed both 32P-postlabelling and immunochemical methods for the analysis of α-OH-Acr-dG and γ-OH-Acr-dG in normal human urothelial mucosa and bladder tumor tissues. γ-OH-Acr-dG was the predominant isomer in both tissues, with higher levels observed in tumor tissue compared to normal tissue [68].

#### *2.8. Formaldehyde*

Formaldehyde is considered a human carcinogen and is widely present in the environment. It is mainly used in the production of industrial resins. It is also produced during cooking and cigarette smoking as well as by endogenous processes. Formaldehyde reacts with DNA and forms several DNA adducts and cross-links, with the most abundant one being *N*6-hydroxymethyldeoxyadenosine (HOMe-dA, Figure 2). We were the first group that demonstrated the presence of this specific formaldehyde-DNA adduct in humans [97].

Li et al. developed an LC-ESI-MS/MS-based method for the simultaneous analysis of two formaldehyde-derived DNA adducts, HOMe-dA and *N*2-hydroxymethyldeoxyguanosine (HOMe-dG, Figure 2), in human saliva samples (Table 1). HOMe-dA and HOMe-dG were measured after sodium cyanoborohydride (NaBH3CN) reduction as *N*6-methyldeoxyadenosine and *N*2-methyldeoxyguanosine, respectively. The levels of HOMe-dA were 996 ± 757 and 670 ± 524 adducts per 108 nucleotides in smokers (*n* = 16) and nonsmokers (*n* = 16), respectively, while HOMe-dG was detected in a much lower level in the same samples, with 26 ± 21 and 20 ± 11 adducts per 108 nucleotides being present in smokers and nonsmokers, respectively. No statistical difference of adduct levels was observed between smokers and nonsmokers [95].

#### *2.9. Acetaldehyde and Crotonaldehyde*

Acetaldehyde associated with the consumption of alcoholic beverages is carcinogenic to humans [91]. Acetaldehyde is ubiquitous in the human environment, and is one of the most prevalent carcinogens in cigarette smoke. It also occurs widely in fruits, vegetables, and cooked meat. Crotonaldehyde is also ubiquitously present in the environment. It is present in mobile source emissions, tobacco smoke, and other thermal degradation mixtures. Crotonaldehyde is mutagenic and carcinogenic. Both acetaldehyde and crotonaldehyde react with DNA forming a pair of diastereomeric adducts, (6*S*,8*S*)- and (6*R*,8*R*)-3-(2 -deoxyribos-1 -yl)-5,6,7,8-tetrahydro-8-hydroxy-6-methylpyrimido[1,2-*a*]purine-10(3*H*)one (Cro-dG, Figure 2). In addition to Cro-dG, acetaldehyde also reacts with DNA to generate its major DNA adduct, *N*2-ethylidene-deoxyguanosine (*N*2-ethylidene-dG). We have previously developed LC-ESI-MS/MS-based methods and detected both Cro-dG and *N*2-ethylidene-dG in human DNA [98,99].

Using an immunochemical approach, Weng et al. detected Cro-dG in human buccal cells, sputum, and lung tissue samples (Table 1) [53]. The levels of Cro-dG in buccal cells from smokers (*n* = 33) were significantly higher (*p* < 0.0001) compared to nonsmokers (*n* = 17). Similarly, the adduct levels in sputum were also higher (*p* < 0.05) in smokers (*n* = 22) compared to nonsmokers (*n* = 8). Cro-dG was also detected in noncancerous lung tissues obtained from lung cancer patients with higher levels in smokers (*n* = 41) compared to nonsmokers (*n* = 13) [53]. However, the levels of Cro-dG reached ~1,000 adducts per 108 nucleotides in buccal cells and sputum. Such high levels are surprising, and the specificity of the method as well as artifactual formation during sample preparation should be evaluated. In a second study using an LC-ESI-MS/MS method, Cro-dG was detected in saliva samples from 16 smokers at 2.6 ± 2.1 adducts per 108 nucleotides, while no Cro-dG was detected in any of the samples from nonsmokers [94]. In a third study, Cro-dG was analyzed in urine samples using an LC-ESI-MS/MS method to investigate exposure to urban air pollution. Higher adduct levels were observed in subjects (*n* = 47) exposed to air pollution with a median value of 21 fmol of Cro-dG per mg creatinine, compared to controls (*n* = 35) with a median value of 8 fmol of Cro-dG per mg creatinine (*p* < 0.05) [100].

The acetaldehyde-derived *N*2-ethylidene-dG is generally measured after reduction (e.g., by NaBH3CN) as *N*2-ethyl-dG (Table 1) [99]. In the second study mentioned above, *N*2-ethylidene-dG (as *<sup>N</sup>*2-ethyl-dG) was also measured in the same saliva samples, with levels of 6.2 ± 3.5 and 0.59 ± 0.89 adducts per 10<sup>8</sup> nucleotides being present in smokers and nonsmokers, respectively. The difference between smokers and nonsmokers was not statistically significant [94]. The formation of *N*2-ethylidene-dG was also investigated in alcohol consumption-related studies [101–103].

#### *2.10. Acrylamide*

Acrylamide is a probable human carcinogen and is present in tobacco smoke and in carbohydrate-rich foods processed at high temperatures. Acrylamide is metabolized to glycidamide, which reacts with DNA to form a major DNA adduct, 7-(2-carbamoyl-2-hydroxyethyl) guanine (7-GAG, Figure 2). Huang, et al developed an LC-ESI-MS/MS method for the analysis of 7-GAG in human urine (Table 1). The method was applied to the analysis of urine samples from 30 smokers and 33 nonsmokers. The levels of 7-GAG ranged from 0.61–6.22 (mean: 1.4) and 0.36–3.0 (mean: 0.93) μg/g creatinine in smokers and nonsmokers, respectively, a non-significant difference. However, the urinary *N*-acetyl-S-(propionamide)-cysteine, a metabolite of acrylamide, showed significantly higher levels (*p* < 0.001) in smokers compared to nonsmokers [104]. The same research group investigated in another study the effect of occupational exposure to acrylamide on the formation of 7-GAG. The adduct was measured in urine samples from eight workers who were exposed to acrylamide and 36 controls. Higher levels (*p* < 0.001) of 7-GAG were observed in exposed workers, with adduct levels ranging from 1.0–13 (mean: 2.5) μg/g creatinine, compared to controls with adduct levels ranging from 0.20–0.93 (mean: 0.36) μg/g creatinine [105].

#### *2.11. Oxidative Damage*

Reactive oxygen species (ROS) constitute an important class of DNA damaging agents, and their generation can be elevated due to exposure to tobacco smoke. In addition to damaging DNA directly, ROS also cause DNA damage indirectly by reacting with other macromolecules [106]. For example, ROS can lead to lipid peroxidation of polyunsaturated fatty acids and produce malondialdehyde, which reacts with DNA to form DNA adducts, with the predominant one being 3-(2-deoxy-β-d-erythro-pentafuranosyl)pyrimido[1,2-α]purin-10(3*H*)-one deoxyguanosine (M1dG, Figure 2) [107]. In addition to M1dG, some representative oxidative damage-associated DNA adducts include 8-oxo-dG, 1,*N*6-etheno-2-deoxyadenosine (εdA), and 3,*N*4-etheno-2 -deoxycytidine (εdC). A more comprehensive summary of the formation and biological consequences of oxidative stress-induced DNA damage was covered by a recent review [106].

M1dG is a premutagenic lesion and induces G→T and G→A mutations, which could be an important step in the etiology of oxidative associated-diseases. M1dG was detected in human leukocytes [33,48,108–111] and nasal epithelium (Table 1) [112,113]. Studies suggest that smoking contributes to the levels of M1dG in humans. In a study of workers occupationally exposed to silica dust, the levels of M1dG in nasal epithelium of smokers (*n* = 58) and former smokers (*n* = 63) were <sup>78</sup> ± 9.8 and 81 ± 9.7 adducts per 10<sup>8</sup> nucleotides, respectively, significantly higher (*<sup>p</sup>* < 0.05) compared to nonsmokers (*<sup>n</sup>* = 132) with M1dG levels of 57 ± 6.2 adducts per 10<sup>8</sup> nucleotides [112]. In a study of occupational exposure to asbestos, higher M1dG levels (*p* = 0.005) were observed in leukocytes of heavy smokers (>40 packs/year, *n* = 12) compared to nonsmokers (*n* = 136), but the difference was not observed among moderate smokers (20.1–40 packs/year, *n* = 29), light smokers (0.1–20 packs/year, *n* = 20), and nonsmokers [108]. In another study investigating the effects of diet on M1dG levels among industrial estate workers, no difference was observed between smokers (*n* = 46) and nonsmokers (*n* = 17) among those workers. However, in the control group, the adduct levels were significantly higher (*p* < 0.05) in smokers (*n* = 64) compared to nonsmokers (n = 53) [110]. In addition to the findings that the levels of M1dG were elevated due to smoking, some studies also demonstrated the correlation of M1dG and DNA methylation in smokers [48,111].

We developed an LC-NSI-HRMS/MS method for the analysis of M1dG in human leukocyte DNA, and measured M1dG in buffy coat samples (leukocyte-containing fraction) from 25 smokers and 25 nonsmokers (Table 1). The adduct levels in smokers and nonsmokers averaged 2.2 ± 2.4 and 1.9 ± 2.0 adducts per 10<sup>8</sup> nucleotides, respectively, a non-significant difference [33]. Similarly, the effect of smoking on M1dG formation was not significant in another two studies [109,113].

8-Oxo-dG is the most prominent and widely studied oxidative damage-related DNA adduct. It is a highly mutagenic lesion and mispairs with A during DNA replication leading to a GC→AT conversion. 8-Oxo-dG was detected in various human sample types, including semen [114,115], retina [116], leukocytes [117–120], and urine samples (Table 1) [121]. In a study investigating the effect of tobacco use on the possible etiology of childhood cancer, 8-oxo-dG was measured by an immunochemical method in semen samples from the children's fathers. The levels of 8-oxo-dG were 66 ± 2.9, 54 ± 4.4, and 179 ± 20 ng/mL in smokers (*n* = 33), tobacco chewers (*n* = 31), and subjects who both smoked and chewed tobacco (*n* = 41), respectively, all significantly higher than nonsmokers (*n* = 33) with adduct levels of 34 ± 1.1 ng/mL [115]. The effect of smoking on 8-oxo-dG formation was either not significant or not investigated in other studies [114,116–121].


*Toxics* **2019**

**Table 1.**

Detection of tobacco

smoke-related

 DNA adducts in human samples. , *7*, 16

εdA and εdC are two representative DNA adducts formed by 4-hydroxy-2-nonenal, a product of lipid peroxidation. Both adducts are promutagenic and could be useful markers to assess oxidative stress-derived DNA damage. Using an LC-ESI-MS/MS method, Bin, et al measured εdA and εdC in urine to investigate the effect of occupational exposure to diesel engine exhaust on adduct formation (Table 1) [122]. The levels of εdA averaged 0.19 nmol/g creatinine in exposed workers (*n* = 108), significantly higher (*p* < 0.001) than the control group (*n* = 109) with εdA levels averaging 0.09 nmol/g creatinine. The levels of εdC showed no difference between the two groups. The contribution of smoking to the formation of εdA or εdC was not significant [122].

#### **3. Newly Characterized Tobacco-Specific DNA Adducts**

Since 2012, multiple DNA adducts formed by tobacco-specific carcinogens (i.e., NNK and NNN) have been characterized and quantified in in vitro models and in NNK- or NNN-treated animals [19–21,31,34–36,123,124]. In addition to DNA base adducts, which have been extensively studied, we have recently characterized and measured a panel of DNA phosphate adducts formed by NNK and NNN [20,34–36]. The biological significance of these newly identified DNA adducts is still largely unknown and further studies need to be conducted to better understand the formation, removal, and biological implications of these DNA adducts to facilitate their use as biomarkers in exposure evaluation and cancer risk assessment.

#### *3.1. DNA Phosphate Adducts*

In addition to DNA base moieties, some alkylating agents also react with the oxygen of the phosphate backbone to form DNA phosphate adducts [125]. Earlier studies have demonstrated that phosphate adducts formed by certain carcinogens persisted in vivo and had longer half-lives than their corresponding base adducts [126,127], suggesting DNA phosphate adducts may serve as better biomarkers of chronic exposure to those carcinogens. For the formation of DNA phosphate adducts by the tobacco-specific carcinogens NNK and NNN, there was only one study using a transalkylation approach to indirectly measure the adduct levels in [3H]NNK-treated mice [128]. However, that study only provided indirect proof of the presence of NNK-derived DNA phosphate adducts, and the chemical structures were not characterized for individual adducts.

One major challenge of direct measurement of DNA phosphate adducts is their structural complexity. After enzyme hydrolysis of DNA samples, DNA phosphate adducts are measured as phosphotriesters (PTE)—B1p(alkyl)B2. B1 and B2 are the same or different nucleosides, which can be 10 different combinations of the four nucleosides (Table 2) [125]. Due to the tetrahedral phosphate group in B1p(alkyl)B2, there can be *R*p or *S*p diastereomers present. With the same nucleosides, there can be two isomers depending on which oxygen is alkylated. With different nucleosides, because of the connection of sugar moieties, it can be either B1-5 -alkyl-3 -B2 or B1-3 -alkyl-5 -B2. Therefore, there can be 32 different isomers of B1p(alkyl)B2 (Table 2). In order to characterize and measure all of the 32 possible combinations, a specific and powerful analytical method is required. We have developed a series of LC-NSI-HRMS/MS-based methods and characterized four different types of NNK- and NNN-derived DNA phosphate adducts, including pyridyloxobutyl DNA phosphate adducts [B1p(POB)B2], pyridylhydroxybutyl DNA phosphate adducts [B1p(PHB)B2: B1p(PHB)sB2 + B1p(PHB)bB2], and methyl DNA phosphate adducts [B1pMeB2] [20,34–36,124].

\* Chiral center.

#### 3.1.1. B1p(POB)B2

One metabolic pathway of NNK proceeds via α-hydroxylation of its methyl group to produce the unstable intermediate **6** (Figure 3). This intermediate spontaneously yields **11** and then the alkylating agent **14**, which reacts with DNA to form POB DNA base adducts and POB DNA phosphate adducts—B1p(POB)B2 (Figure 3) [20]. We first characterized and detected B1p(POB)B2 in calf thymus DNA (CT-DNA) treated with 4-(acetoxymethylnitrosamino)-1-(3-pyridyl)-1-butanone (NNKOAc, **2**), a regiochemically activated form of NNK. The structures of two combinations of B1p(POB)B2, Cp(POB)C and Tp(POB)T, were confirmed by comparing with the synthetic standards [20,36], and the identities of the other B1p(POB)B2 adducts were confirmed on the basis of the accurate masses of the precursor ions and the corresponding fragment ions. A total of 30 out of 32 possible B1p(POB)B2 adducts were detected in NNKOAc-treated CT-DNA. The B1p(POB)B2 adducts were also detected and quantified in rats treated with NNK acutely (0.1 mmol/kg once daily for 4 days by subcutaneous injection) and chronically (5 ppm in drinking water for 10, 30, 50, and 70 weeks). In the chronically treated rats, some B1p(POB)B2 adducts are persistent and abundant over 70 weeks, which suggested that those adducts could potentially be detected in smokers and be used as biomarkers to investigate chronic exposure to NNK [20]. The B1p(POB)B2 adducts were also detected in rats treated with NNAL [35] or NNN [124].

#### 3.1.2. B1p(PHB)B2

NNK is extensively metabolized to NNAL, which undergoes α-hydroxylation to produce diazonium ion **16**. Diazonium ion **16** reacts with DNA to form PHB base adducts and PHB DNA phosphate adducts, B1p(PHB)sB2, where the subscript s represents 'straight chain' (Figure 3). Diazonium ion **16** further rearranges to carbocation **17**, which also reacts with DNA to form another type of PHB DNA phosphate adducts—B1p(PHB)bB2, where the subscript b represents 'branched chain' (Figure 3). Since both B1p(PHB)sB2 and B1p(PHB)bB2 are derived from NNAL, they are collectively called B1p(PHB)B2 phosphate adducts [36]. For B1p(PHB)sB2, because of the two chiral centers, the phosphorus and carbinol carbon, there can be four and eight isomers with the same and different nucleoside combinations, respectively. Therefore, a total of 64 different possible B1p(PHB)sB2 adducts can be formed (Table 2). Compared to B1p(PHB)sB2, an additional chiral center at the methyl-bearing carbon is present in the structure of B1p(PHB)bB2 adducts, so there can be eight and 16 isomers with the same and different nucleoside combinations, respectively. Consequently, a total of 128 different possible B1p(PHB)bB2 adducts can be formed (Table 2). Therefore, a total of 192 possible B1p(PHB)B2 adducts can be formed from the α-hydroxylation of NNAL's methyl group. Using a similar strategy as in B1p(POB)B2 characterization, a total of 107 out of 192 possible B1p(PHB)B2 adducts were detected in the NNK-treated rat lung DNA samples. Similar to B1p(POB)B2 adducts, certain B1p(PHB)B2 adducts demonstrated persistence for over 70 weeks, suggesting that they could be potential biomarkers of chronic exposure to NNK and NNAL [36]. Both B1p(POB)B2 and B1p(PHB)B2 adducts were also detected in NNAL-treated rats [35].

#### 3.1.3. B1pMeB2

In addition to the methyl group, α-hydroxylation also occurs on the α-methylene carbon of NNK to produce methane diazohydroxide **15**, which reacts with DNA to form methyl DNA base and phosphate adducts (B1pMeB2) [34]. Similar to B1p(POB)B2 adducts, there are 32 possible isomers of B1pMeB2 phosphate adducts (Table 2). We characterized and detected B1pMeB2 adducts in lung DNA of rats treated with NNK in their drinking water (5 ppm) for 10, 30, 50, and 70 weeks, and a total of 23 out of 32 possible isomers were detected. Thus far, we have identified 30 B1p(POB)B2 adducts and 107 B1p(PHB)B2 adducts, resulting in a total of 160 different structurally unique DNA phosphate adducts in NNK-treated rats. This is by far the most structurally diverse panel of DNA adducts identified from any carcinogen. Consistent with the observations of B1p(POB)B2 and B1p(PHB)B2

adducts, certain B1pMeB2 phosphate adducts showed persistence in NNK-treated rats. B1pMeB2 adducts were also detected in NNAL-treated rats in our study [34]. The relative levels of B1p(POB)B2, B1p(PHB)B2 and B1pMeB2 phosphate adducts, and their corresponding base adducts in lung DNA of rats treated with NNK or NNAL were summarized in that study.

#### *3.2. DNA Base Adducts*

In addition to DNA phosphate adducts, multiple DNA base adducts that are associated with NNK and NNN were either newly characterized in vitro or detected for the first time in vivo. These DNA adducts include 2-(2-(3-pyridyl)-*N*-pyrrolidinyl)-2 -deoxyinosine (py-py-dI) [31], 3-POB-dC, *N*4-POB-dC [21], *O*4-POB-dT [123], *N*6-POB-dA, and *N*1-POB-dI [19].

#### 3.2.1. Py-py-dI

Metabolic activation of NNN occurs via one of two pathways: 2 -hydroxylation or 5 -hydroxylation (Figure 3), and 5 -hydroxylation is likely to be the major metabolic pathway in humans. The major adduct formed in vitro via 5 -hydroxylation was py-py-dI, which forms upon the reaction of the *N*<sup>2</sup> position of dG with diazonium ion **13** followed by NaBH3CN reduction [129]. Using an LC-ESI-MS/MS method, we have recently detected py-py-dI in rats treated with NNN in the drinking water (7–500 ppm). Py-py-dI was the major DNA adduct resulting from 5 -hydroxylation, with the highest levels being detected in the lung and nasal cavity of NNN-treated rats. Our study also showed that py-py-dI was more abundant than the 2 -hydroxylation of NNN-derived POB DNA base adducts in the in vitro studies in which human liver S9 fraction or human hepatocytes were incubated with NNN (2–500 μM) [31]. The results of this study identified py-py-dI as the major NNN-derived DNA adduct in the human enzyme systems examined. Thus, py-py-dI could be a useful biomarker to investigate the metabolic activation of NNN in humans.

#### 3.2.2. 3-POB-dC and N4-POB-dC

The intermediate **14** from NNK also reacts with dC in DNA and forms *O*2-POB-C, which has been detected in rats treated with NNK or NNN. In addition to the *O*2-position, the 3-position and exocyclic *N*4-amino group of dC are also potentially reactive sites for alkylation, possibly forming 3-POB-dC and *N*4-POB-dC, respectively. We have recently synthesized the chemical standards of 3-POB-dC and *N*4-POB-dC, and confirmed the presence of both adducts in NNKOAc-treated CT-DNA. In agreement with our previous studies, *O*2-POB-C was the most abundant dC adduct, while considerably lower amounts of 3-POB-dC and *N*4-POB-dC were present in NNKOAc-treated CT-DNA [21]. The results of this study provide a more complete picture of dC adduct formation in the reaction of NNKOAc with DNA.

#### 3.2.3. O4-POB-dT

The major dT adduct formed by NNK and NNN is *O*2-POB-dT. Using an LC-NSI-MS/MS method, Leng, et al detected and characterized for the first time a new dT adduct—*O*4-POB-dT, in NNKOAc-treated CT-DNA [123]. The identity of *O*4-POB-dT was confirmed by the chemical standard synthesized in the study. *O*4-POB-dT was also detected in human skin fibroblasts and Chinese hamster ovary cells treated with NNKOAc. *O*2-POB-dT was detected in all the systems with higher levels than *O*4-POB-dT. Results from this study indicated that both *O*4-POB-dT and *O*2-POB-dT were subjected to repair by the nucleotide excision repair pathway [123].

#### 3.2.4. dA Adducts

All the previous studies have identified and measured a panel of POB DNA base adducts of dG, dC, and dT, but not dA. To complete the panel of POB DNA base adducts, we determined the possible formation of POB-dA adducts in the in vitro NNKOAc-treated CT-DNA and in vivo NNK-treated rats [19]. We hypothesized that the initial alkylation of dA would occur at the *N*<sup>1</sup> position (Figure 3), which results in an unstable cationic intermediate adduct. This unstable adduct can either deprotonate and undergo spontaneous Dimroth rearrangement to give *N*6-POB-dA, or deaminate via an addition−elimination mechanism to give *<sup>N</sup>*1-POB-dI. In support of our hypothesis, both *N*6-POB-dA and *N*1-POB-dI were detected in NNKOAc-treated CT-DNA using our newly developed LC-ESI-MS/MS method. In contrast, only *N*6-POB-dA but not *N*1-POB-dI was detected in lung and liver DNA from rats treated with NNK for 50 weeks (5 ppm in the drinking water), possibly due to the rapid repair of *N*1-POB-dI. Similarly, we also detected N6-PHB-dA in the NNAL-treated rats [19].

#### **4. Perspectives**

#### *4.1. Current Limitations*

Among all the adducts described in this review, some have been found to be at the same or overlapping levels in smokers and nonsmokers, while others had significantly higher levels in smokers compared to nonsmokers. One should keep in mind when investigating the effects of smoking on DNA adduct formation, the association may be compromised by a number of confounding variables, such as current smoking status at the time of sample collection, exposures to secondhand smoke, air pollution, occupational exposure, dietary habits, and disease status. For example, self-reported smoking status collected through questionnaires may not be accurate. In addition, in studies which used surgically obtained tissues from cancer patients, smoking status may not be up-to-date if subjects stopped smoking for a few weeks or more before surgery. This would have an unknown effect on DNA adduct levels since their persistence in those human tissues is generally unknown. To overcome this limitation, urine samples can be collected from the same subjects for the measurement of tobacco biomarkers (e.g., NNAL and cotinine) to confirm their current smoking status.

Another potential confounding variable is the presence in the environment of the same chemicals that form DNA adducts. For example, PAH are tobacco smoke carcinogens and DNA adduct formation by PAH is often used to investigate smoking-related cancers. However, PAH are also present in the environment where combustion occurs and in the diet due to high temperature cooking. Therefore, factors such as air pollution and diet need to be taken into account when interpreting the association between PAH-derived DNA adduct formation and tobacco smoke exposure. The only carcinogens that are specific to tobacco smoke are the tobacco-specific nitrosamines such as NNK and NNN [7,15]. DNA adducts formed by NNK and NNN therefore have the potential to be ideal biomarkers to investigate tobacco-related cancer risk.

#### *4.2. Analytical Methodologies*

The chemical analysis of DNA adducts has been extensively reviewed [107,130,131]. To utilize DNA adducts as biomarkers of tobacco smoke exposure and cancer risk assessment, reliable analytical methods have to be developed and fully validated for linearity, accuracy, precision, specificity, and ruggedness. High-throughput methods are preferred and will allow the analysis of DNA adducts to be performed in studies with large sample sizes, which are typical in epidemiology. 32P-postlabeling and immunochemical approaches were extensively used for DNA adduct analysis in the past and are still quite prevalent nowadays. However, the uncertainty in labeling efficiency and the lack of physiochemical structural confirmation of the 32P-postlabeling technique are major disadvantages potentially undermining the adequacy of data produced by this method. A critical drawback of the immunochemical approach is that the specificity of the antibodies for some DNA adducts is uncertain, as they may cross-react with other DNA lesions or endogenous components, resulting in errors in characterization and quantitation. The disadvantages of 32P-postlabeling and immunochemical approaches are evident when the results are compared to those obtained by mass spectrometry-based methods. For example, the levels of BPDE-*N*2-dG and 4-ABP-C8-dG, as discussed in this review, and the levels of DNA adducts formed by 2-amino-1-methyl-6-phenylimidazo[4,5-*b*]pyridine [65], were all generally 10 to 100-fold higher in the same type of tissues when results of 32P-postlabeling or immunochemical methods were compared to those obtained by using mass spectrometry-based methods. This difference suggests that one should be cautious when interpreting data obtained from 32P-postlabeling or immunochemical methods. We recommend that DNA adduct identity and quantity be confirmed by well-validated mass spectrometry-based methods.

During the past few decades, mass spectrometry coupled with the stable isotope dilution methods has evolved to become the gold standard for identification, characterization, and quantitation of DNA adducts. Further development and validation of mass spectrometry-based methods for high-throughput, sensitive and accurate quantitation of DNA adducts in humans are required in future studies. The LC-NSI-HRMS/MS-based technique can achieve detection limits in the low amol (10−<sup>18</sup> mol), or even high zmol (10−<sup>21</sup> mol) range, allowing the highest levels of specificity and sensitivity. To achieve such detection limits in human biological samples, the samples should be thoroughly purified to avoid co-eluting interferences and ion suppression during the mass spectrometry analysis. In addition to the analysis of targeted DNA adducts, mass spectrometry-based DNA adductomics can be used to comprehensively screen for DNA modifications, including both known and unknown/unidentified DNA adducts, and expedite the discovery of novel tobacco smoke-related DNA adducts [132].

#### *4.3. Artifactual Formation*

Particularly for DNA adducts formed endogenously, in addition to the efforts on improving the detection sensitivity and sample purification process, precautions should be taken to eliminate artifactual formation during all the steps of sample analysis, including sample collection and storage, DNA isolation and enzymatic hydrolysis, sample purification and enrichment, and mass spectrometry analysis. For instance, artifactual formation is one likely contributor to the relatively high levels of γ-OH-Acr-dG and Cro-dG measured in one study [53]. Another example is 8-oxo-dG. When the DNA is extracted from the cell, the in vivo protecting enzymes and antioxidants are no longer present to neutralize oxygen radicals and repair DNA damage, and 8-oxo-dG can thus be formed by the oxidation of free dG as well as dG in DNA [133].

We suggest the following strategy to systematically evaluate and reduce the possible artifactual formation of endogenous DNA adducts: 1) Sample collection. The samples should be handled by well-trained personnel to reduce the period between collection and storage. The samples should be placed on ice if they cannot be processed immediately; 2) Sample storage. The storage conditions should be evaluated to determine that no artifactual formation or degradation of DNA adducts occurs. A comparison of DNA adduct levels can be performed between the DNA isolated immediately after sample collection and the DNA isolated from the same sample after a certain period of storage; 3) DNA isolation. Commercially available calf thymus DNA (CT-DNA) contains most of the endogenous DNA adducts at certain levels, and can be used to investigate possible artifactual formation. To perform such an experiment, CT-DNA is dissolved in solution and isolated using the proposed protocol [33,116]. The adduct levels in the isolated CT-DNA are then compared to the levels in the same CT-DNA without isolation to determine whether the adduct levels are impacted by the DNA isolation process. This approach can also be used to investigate whether certain chemicals (e.g., antioxidants) can be used to lower or prevent artifactual formation; 4) DNA enzymatic hydrolysis and sample purification. Stable isotope-labeled nucleotides can be added to samples prior to starting DNA hydrolysis, and the possible artifactual formation can be monitored by checking whether correspondingly labeled DNA adducts are present [116]; 5) Mass spectrometry analysis. Depending on the sample purification approach, the free nucleotides may still be in the final samples and form corresponding adducts in the ion source of the mass spectrometer (e.g., dG forms 8-oxo-dG in the ESI ion source) [116]. If the free nucleotide happens to co-elute with the adduct, additional adducts may be formed from the free nucleotide, resulting in the overestimation of the adduct levels. To avoid this, the free nucleotide should be separated from the adduct, either at the sample purification step or on the LC column during the mass spectrometry analysis.

#### *4.4. Other Considerations*

Stability of DNA adducts over time. The adduct levels are supposed to reflect the balance between DNA adduct formation and repair. If this is the case, it would be expected that the DNA adduct level would remain relatively constant over time in a similar exposure environment. In the case of tobacco smoke-related DNA adducts, the level of an ideal DNA adduct as a biomarker should stay constant if the smoker maintains his/her habit. A longitudinal study can be performed where the same biological samples are collected at different time intervals (days, weeks, or months apart) and the level of this DNA adduct is measured to determine its stability over time.

Surrogate tissues. To conduct a longitudinal study, surgically obtaining human tissues such as lung becomes highly impractical. Instead, surrogate tissues such as bronchoalveolar lavage (BAL), peripheral white blood cells or oral cells can be used. For example, studies have demonstrated a correlation between changes in the oral cavity and lung in smokers [26,134], so oral cells would potentially be a source of readily obtained biological samples.

Computational approach. In addition to experimental techniques, computational approaches have been used to study conformational preference of the adducts in DNA, their potential to cause mismatch and mutations, and their interactions with DNA polymerases and repair enzymes [135–137]. For example, a multiscale computational approach demonstrated that both *O*6-POB-G and *O*6-PHB-G pair with C and T, with the latter being more mutagenic because of the difference in the bulky moiety hydrogen-bonding pattern [135]. Such studies provide important insights into the biological significance of smoking-related DNA adducts, which is critical for understanding their role in tobacco-induced cancers.

**Author Contributions:** All authors critically reviewed the literature and contributed to drafting the manuscript.

**Funding:** The work cited in this review conducted by the Hecht laboratory has been supported by Grants CA-81301 and CA-138338 from the National Cancer Institute; The work cited in this review conducted by the Stepanov laboratory has been supported by Grants CA-179246 and CA-180880 from the National Cancer Institute. Mass spectrometry was carried out in Analytical Biochemistry Shared Resources of the Masonic Cancer Center, University of Minnesota, funded in part by Cancer Center Support Grant CA-077598.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Perspective*
