Characterization of adductomic totality of NNK, (R)-NNAL and (S)-NNAL in A/J mice, and their correlations with distinct lung carcinogenicity

Abstract

Lung cancer is the leading cause of cancer-related deaths. While tobacco use is the main cause, only 10–20% of smokers eventually develop clinical lung cancer. Thus, the ability of lung cancer risk prediction among smokers could transform lung cancer management with early preventive interventions. Given that DNA damage by tobacco carcinogens is the potential root cause of lung carcinogenesis, we characterized the adductomic totality of 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (a potent lung carcinogen in tobacco, commonly known as NNK) in the target lung tissues, the liver tissues and the peripheral serum samples in a single-dose NNK-induced lung carcinogenesis A/J mouse model. We also characterized these adductomic totalities from the two enantiomers of 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol (NNAL, the major in vivo metabolite of NNK) given their distinct carcinogenicity in A/J mice. With these adductomic data, we demonstrated that tissue protein adductomics have the highest abundance. We also identified that the adductomic levels at the 8 h time point after carcinogen exposure were among the highest. More importantly, the relationships among these adductomics were characterized with overall strong positive linear correlations, demonstrating the potential of using peripheral serum protein adductomics to reflect DNA adductomics in the target lung tissues. Lastly, we explored the relationships of these adductomics with lung tumor status in A/J mice, providing preliminary but promising evidence of the feasibility of lung cancer risk prediction using peripheral adductomic profiling.

NNK-induced adductomics in different A/J mouse biospecimens show strong and positive linear correlations with peripheral adductomics revealing the potential to predict lung carcinogenesis risk, which may translate to better and more feasible lung cancer risk prediction among human high-risk populations.

Graphical Abstract

Graphical Abstract.

Introduction

Lung cancer is an aggressive cancer with high incidence and death rates. There are over 200 000 lung cancer cases with >130 000 deaths annually in the USA (1–3). Currently, there are no effective approaches in lung cancer early diagnosis. Even with the advent of targeted therapies and immunotherapies, the 5-year overall survival rate for patients with lung cancer has languished between 10 and 20% (4). Thus, there is an urgent impetus for paradigm-shift strategies to improving lung cancer management. Although multiple factors could increase lung cancer risk, tobacco smoking remains the main cause (5,6). Cessation therefore is the ideal preventive strategy. Quitting, however, is difficult because of the addictive nature of nicotine in tobacco products that only a small portion of smokers succeed (7,8). There are 34 million adult smokers in the USA (9) and this number may not decrease significantly in the near future (10). Since 80–90% of smokers would not develop lung cancers clinically (11), one strategy is to predicting which smoker has a higher risk of lung cancer early on with early intervention.

Tobacco smoke contains various carcinogens and co-carcinogens, which collectively initiate and promote lung carcinogenesis. Mechanistically, these carcinogens covalently modify DNA, resulting in DNA damage. Such DNA damage, if not repaired, would increase the chance of mutation and cancer risks. Carcinogen-induced DNA damage, therefore, has been proposed as the root cause of carcinogenesis. The DNA damage profiles from tobacco carcinogens in the lung tissue among smokers thus have the potential to predict lung cancer risk. Among carcinogens in tobacco smoke, 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (commonly known as NNK) is a tobacco-specific N-nitrosamine with potent lung carcinogenicity, inducing lung adenoma and adenocarcinoma in multiple species (12). Ample evidence suggests that NNK may be key to lung cancer formation among US smokers. First, the lung adenocarcinoma case increase in the USA over the past few decades appeared to correlate with the increased content of NNK in tobacco products (13,14), indicating a potential connection. Second, a number of studies discovered that polymorphism of multiple genes involved in NNK metabolism, including its bioactivation and detoxification, are associated with differential lung cancer risks (15–19), revealing molecular mechanisms of NNK carcinogenesis risk. Lastly, several prospective studies showed that higher levels of urinary NNK metabolites were associated with increased lung cancer risk among smokers, linking its exposure to lung cancer (20–23). NNK therefore is a logical model lung carcinogen to explore its DNA damage profiles in lung cancer risk prediction.

NNK carcinogenesis has been well characterized (Figure 1): α-hydroxylation of NNK generates two reactive species, leading to two types of DNA damages—pyridyloxobutylation (POB) and methylation (12,24). NNK has a very short in vivo half-life and is dominantly converted to 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol (NNAL) via enzyme-mediated carbonyl reduction (25). NNAL can be oxidized back to NNK in rats and mice (26–28). NNK and NNAL, therefore, would interconvert in vivo. NNAL is also a lung carcinogen (26) and can be activated via α-hydroxylation to generate two reactive species, leading to DNA pyridylhydroxybutylation (PHB) and methylation. The POB reactive intermediates from NNK also modify plasma proteins (27,29,30). Protein modifications from the methyl and PHB reactive intermediates are also expected. Another major metabolism of NNAL is glucuronidation, which is proposed as a detoxification mechanism. Although both the hydroxy functional group and the pyridine nitrogen of NNAL can be glucuronidated in human smokers, producing NNAL-O-glucuronide (NNAL-O-Gluc) and NNAL-N-glucuronide (NNAL-N-Gluc), respectively (31–33), our recent study detected only NNAL-O-Gluc in A/J mice (34).

Figure 1.

NNK and NNAL bioactivation and detoxification pathways in humans and laboratory animals. Modified from ref. (26).

In comparison to NNK, NNAL has a chiral center and its two enantiomers appear to have differences in carcinogenicity and likely metabolism, particularly in A/J mice (26). A/J mice have a predisposed pulmonary adenoma susceptibility (Pas-1) gene (35), making this species highly prone to lung carcinogenesis; even a single dose of NNK reproducibly induces lung tumor formation in 17 weeks (5). This model therefore has been widely used to study lung carcinogenesis and develop lung cancer chemopreventive agent (5). In this model, a single dose of NNK or NNAL enantiomers via i.p. injection (20 µmol/mouse) showed that (S)-NNAL was of comparable carcinogenicity as NNK but (R)-NNAL was much less carcinogenic (26). The analyses of urinary metabolites suggested (R)-NNAL as a better substrate for glucuronidation while lung microsomal metabolism indicated (S)-NNAL as a better substrate for activation (26), which may account for their distinct carcinogenicity in A/J mice.

Upon confirming the distinct carcinogenicity of NNK, (R)-NNAL and (S)-NNAL (26) in this study, we systematically quantified different types of DNA and protein adducts from NNK, (R)-NNAL and (S)-NNAL in the target lung tissues, non-targeting liver tissues, serum samples, characterized their temporal profiles and correlations, and importantly explored their relationships with the distinct lung carcinogenicity of NNK, (R)-NNAL and (S)-NNAL in A/J mice. One key observation is that the serum protein adductomics correlate well with lung DNA adductomics as Hecht et al. observed before (30) and thus demonstrate the potential to predict lung carcinogenesis risk. The results from this study also provide qualitative evidence about the rate of in vivo conversion of NNAL enantiomers to NNK and glucuronidation in A/J mice.

Materials and methods

Caution: NNK and NNAL enantiomers are classified as ‘Group 1’ carcinogens by IRAC and should be handled carefully in well-ventilated hoods with appropriate clothing.

Chemicals, reagents and animal diets

NNK, [CD₃]-O⁶-mG, [D₄]-HPB, [CD₅]-NNAL and [CD₅]-NNAL-O-Gluc were from Toronto Research Chemicals (Toronto, Ontario, Canada). 7-POB-G, 7-PHB-G and their corresponding [pyridine-D₄] analogues were synthesized (36,37). The AIN-93 G and M powdered diets were from Harlan Teklad (Madison, WI). Liquid chromatography–mass spectrometry (LC–MS) grade of water, formic acid, methanol and acetonitrile were from Fisher Scientific (Fair Lawn, NJ). StrataX reverse phase cartridges (33 µm, 30 mg/ml) were from Phenomenex (Torrance, CA). Halt Protease Inhibitor Cocktail (100×) and Pierce Universal Nuclease (250 U/µl) were from Thermo Fisher Scientific (Waltham, MA). All other chemicals were from Sigma–Aldrich unless otherwise stated and used without further purification.

General animal study protocols

Female A/J mice (5–6 weeks of age) were from the Jackson Laboratory (Bar Harbor, ME) and maintained in the specific pathogen-free facilities, according to animal welfare protocols approved by Institutional Animal Care and Use Committee at the University of Florida.

Short-term study to characterize the metabolism and adductomics of NNK and NNAL enantiomers

A/J mice, after 1 week of acclimatization, were randomized into different groups and maintained on powdered AIN-G diet on Day 1. With the exception of mice in the negative control groups, mice in the other groups (n = 20) were given the specific carcinogen NNK, (R)-NNAL or (S)-NNAL at a dose of 8 µmol/mouse on Day 7 in 200 µl sterilized saline via i.p. injection. Control mice were given 200 µl saline via i.p. injection. At different time points after carcinogen administration (0.5, 8, 24 and 48 h), five mice from each group were euthanized with urine, serum, lung and liver tissues collected (34). These samples were snap frozen in liquid nitrogen and stored at −80°C until analyses.

Long-term study to assess carcinogenicity of NNK and NNAL enantiomers

A/J mice, after 1 week of acclimatization, were randomized to four different groups and maintained on AIN-G powdered diet. At the end of Week 1, mice in the negative control group (n = 5) were given saline (200 µl) while mice in the other groups (n = 15–22) were given NNK, (R)-NNAL or (S)-NNAL in saline (200 µl, 8 µmol/ mouse) via i.p. injection. At the end of Week 2, mice were switched to AIN-M powdered diet. Food was replenished twice a week. Mouse bodyweight was measured once a week. At the end of Week 17, mice were euthanized with the number of tumors on the mouse lung surface counted with a microscope.

Isolation of DNA and proteins from mouse lung and liver tissues

DNA was isolated from the lung or liver tissues (25 mg) of each individual mouse following Puregene DNA isolation protocol (Qiagen Corp, Valencia, CA) (38). For protein isolation, mouse tissues (25 mg) from each individual animal were mixed with cold RIPA buffer (0.25 ml) and Halt Protease Inhibitor Cocktail (2.5 µl). After tissues homogenization, samples were sonicated for 30 s on ice to break DNA. Pierce Universal Nuclease (2.5 µl) was added to further digest DNA. Samples were then applied to a Millipore centrifugal filter (3 kDa membrane cutoff) and centrifuged at 13 000g for 20 min. The samples retained by the filter were washed with phosphate-buffered saline (3 × 450 µl). The filter was reversed to recover the proteins (~50 µl). The protein concentrations were measured by the standard BCA assay.

Quantification of three representative DNA adducts induced by NNK and NNAL enantiomers in A/J mouse lung and liver tissues

O ⁶-mG, 7-POB-G and 7-PHB-G, representing DNA methylation, POB and PHB, respectively, were hydrolyzed from mouse DNA with methods modified from previous established protocols (36–38). Briefly, DNA (25 µg) from the lung and liver tissues of each individual animal was subjected to acidic thermal hydrolysis at 95°C for 30 min with corresponding isotope-labeled internal standards. The pH of each sample was adjusted to neutral and subjected to solid-phase extraction (SPE) using StrataX cartridges. Samples were analyzed by targeted UPLC–MS², employing a Dionex Ultimate 3000 RS and a Q Exactive Hybrid Quadrupole Orbitrap Mass Spectrometer.

LC–MS/MS analyses of 4-hydroxy-1-(3-pyridyl)-1-butanone and 1-(3-pyridinyl)-1,4-butanediol released from mouse lung and liver DNA

4-Hydroxy-1-(3-pyridyl)-1-butanone (HPB) and 1-(3-pyridinyl)-1,4-butanediol (Diol) hydrolyzed from mouse DNA were quantified following a reported method (39). To the mouse lung/liver DNA (25 µg) from each individual animal, [D₄]-HPB (20 pg) was added and mixed with HCl (0.8 N, 500 µl). The vials were sealed with Parafilm and incubated at 95°C for 5 days. The pH of the samples was neutralized with NaOH (1 M). A portion of each sample (40 µl) was analyzed via high-performance liquid chromatography (HPLC) for guanine quantification. The rest of samples were subjected to SPE with StrataX cartridges, preconditioned with MeOH (1 ml) and water (1 ml). After sample loading, the cartridges were washed with water (1 ml) and 5% MeOH (1 ml) sequentially. Samples were then eluted in MeOH (1 ml) and vacuumed to dryness. The sample was constituted in ammonium acetate (10 mM, 30 µl) with 25 µl for UHPLC–MS/MS analysis. Due to the instability of Diol under such a thermal hydrolysis condition (Supplementary Figure S1, available at Carcinogenesis Online), [D₄]-HPB was used as the internal standard for both HPB and Diol quantification. Diol analysis thus is relative.

LC–MS/MS analyses of HPB and Diol released from mouse serum proteins

Mouse serum sample of each individual animal (25 µl) was mixed with cold MeOH (1 ml) and kept at −20°C to precipitate proteins. The protein pellet was washed with cold MeOH (2 ml) to remove free HPB and Diol in mouse serum. [D₄]-HPB and [D₄]-Diol were added (final concentration: 2 pg/µl serum). The precipitated protein samples were then mixed with NaOH (0.5 N, 100 µl) and incubated on a Thermo Mixer (700 rpm) at 40°C overnight. The solution was neutralized to pH 7 with 1 N HCl and then mixed with cold MeOH (1 ml) to precipitate the proteins. The supernatant was concentrated to ~50 µl via speed vacuum. Upon dilution with water (1 ml), the sample was then applied to the StrataX cartridge, preconditioned with MeOH (1 ml) and water (1 ml). After sample loading, the cartridges were washed with water (1 ml) and 5% MeOH (1 ml). Samples were eluted in MeOH (1 ml), vacuumed to dryness, dissolved in a solution of ammonium acetate (10 mM, 25 µl) for UHPLC–MS/MS analysis following reported LC–MS/MS method (29).

LC–MS/MS analyses of HPB and Diol released from mouse tissue proteins

The proteins (1 mg) in the lung tissue lysate samples were precipitated with the addition of cold MeOH (1 ml) and incubated at −20°C for 20 min. From here, the same procedures as serum proteins hydrolysis were performed for the mouse lung tissue proteins.

LC–MS/MS analyses of NNK and free NNAL in mouse serum

Mouse serum (5 µl) was mixed with cold MeOH (0.995 ml) and cooled to −20°C to precipitate proteins. The supernatant (2 µl) was mixed with [D₅]-NNAL (100 pg) and [D₄]-NNK (100 pg) in 2% formic acid (1 ml). MCX cartridges were preconditioned with MeOH (1 ml) and water (1 ml). After sample loading, the cartridge was washed with 2% formic acid (2 ml) and MeOH (1 ml). Then samples were eluted with 2.5% ammonium hydroxide in MeOH (1 ml), vacuumed to dryness and then resuspended in ammonium acetate (10 mM, 100 µl) with 5 µl subjected to UHPLC–MS/MS analysis, following the method published earlier (34).

LC–MS/MS analysis of NNAL-O-Gluc and free NNAL in mouse urine

The analysis followed the method published earlier (34). Briefly, mouse urine samples were diluted 10⁵ times with LC–MS grade water. The diluted samples (100 µl each) were mixed with [D₅]-NNAL-O-Gluc and [D₅]-NNAL at a final concentration of 1 ng/ml, and subjected to targeted UPLC–MS².

Statistical analysis

Data on lung tumor multiplicity, DNA adducts, protein adducts, serum NNK and NNAL, urinary NNAL-O-Gluc and free NNAL and their ratios are reported as mean ± standard deviation. For three groups or more, one-way analysis of variance was used to compare means among different treatment groups followed by the Dunnett test for comparisons between NNK and other groups. For studies with only two groups, two-tailed t-test was used for analysis. A P value ≤0.05 was considered statistically significant. All analyses were conducted in GraphPad Prism 4 (GraphPad Software).

Results and discussion

Lung tumor formation induced by NNK and NNAL enantiomers

At a single dose of 8 µmol/mouse, NNK and NNAL enantiomers revealed distinct carcinogenicity in A/J mice (Figure 2). NNK-induced lung adenoma in every single mouse with 4.86 ± 2.59 tumors/mouse (mean ± standard deviation). (S)-NNAL also induced tumors in every mouse with 4.40 ± 2.10 tumors/mouse, statistically not different from NNK. (R)-NNAL, on the other hand, failed to induce tumors in 7 of 15 mice with 1.07 ± 1.28 tumors/mouse. The tumor multiplicity induced by (R)-NNAL is only 22.0% of that from NNK treatment. This was consistent with the results from Hecht et al. (26) although the doses used in these two studies were different. The distinct carcinogenicity among NNK, (R)-NNAL and (S)-NNAL suggests that their adductomic profiles should be different.

Figure 2.

Number of lung tumors induced by the single-dose NNK, (R)-NNAL and (S)-NNAL enantiomers (8 µmol/mouse, i.p. injection) in A/J mice (n = 15–22). One-way analysis of variance for comparison among groups and Dunnett test for comparisons between groups. ns: non-significant; ∗∗∗P < 0.001; ∗∗∗∗P < 0.0001.

A short-term A/J mouse study was carried out to characterize adductomics from NNK and NNAL enantiomers because these adducts have limited half-lives in vivo. Since NNK, (R)-NNAL and (S)-NNAL interconvert (Figure 1), they all could form three types of reactive intermediates and result in three types of adducts—POB, PHB and methyl adducts. We therefore quantified these adducts in the lung, liver and serum samples from each individual A/J mouse with samples collected over a short period of time (0.5–48 h) after carcinogen administration.

Lung and liver DNA adduct profiles induced by NNK and NNAL enantiomers

Our previous studies showed high correlations among adducts within each type, such as O⁶-mG and 7-mG, in the same biological samples (40). We therefore quantified O⁶-mG, 7-POB-G and 7-PHB-G, respectively, in the lung (Figure 3A) and liver (Supplementary Figure S2, available at Carcinogenesis Online) tissues. The abundance of these DNA adducts in lung and liver all showed time-dependent profiles with the levels at the 8 h time point among the highest. In the target lung tissues, the NNK groups had the highest level of O⁶-mG at all of the time points, followed by (S)-NNAL with the lowest levels in (R)-NNAL groups. For instance, the amount of O⁶-mG from (R)-NNAL treatment at the 8 h time point was only 33.5% of that with NNK treatment at the same time point while that from (S)-NNAL was 67.5%. As expected, the levels of 7-POB-G were the highest with NNK treatment while there was a significant difference between (S)-NNAL and (R)-NNAL. The amount of 7-POB-G from (R)-NNAL treatment at the 8 h time point was only 10.5% of that with NNK treatment while that from (S)-NNAL was 59.9%, suggesting that the formation of NNK from (S)-NNAL was more favorable than from (R)-NNAL. The level of 7-PHB-G was the highest with (S)-NNAL treatment at all of the time points while its levels were similar between NNK and (R)-NNAL. The DNA adduct profiles in the mouse liver tissues were similar as those in the lung tissues except that the abundance was 5–10 times higher. The higher levels of adducts in the liver in this study are probably driven by the i.p. route of carcinogen administration.

Figure 3.

Adductomics induced by the single-dose NNK, (R)-NNAL or (S)-NNAL (8 µmol/mouse, i.p. injection) in A/J mice. (A) O⁶-mG, 7-POB-G and 7-PHB-G DNA adducts in the lung tissues. (B) HPB and Diol from serum proteins. (C) HPB and Diol from lung DNAs. (D) HPB and Diol from lung proteins.

Serum protein adducts induced by NNK and NNAL enantiomers

Given the challenge to obtain the target lung tissues for carcinogen adduct profiling, adductomics on serum proteins could be potential surrogates. POB and PHB modifications in serum proteins, upon hydrolysis, release HPB and Diol (Figure 1), which can be quantified to non-invasively assess the level of NNK bioactivation. This approach, however, would not be practical to indirectly quantify methyl modification due to its non-specific nature. Herein we therefore quantified HPB and Diol from the serum samples (Figure 3B). Method development, characterization and application scopes will be detailed in a separate manuscript. For three carcinogens, HPB and Diol hydrolyzed from mouse serum protein peaked at the 8 h time point except that the level of HPB from NNK at the 0.5 h time point was slightly but statistically non-significantly higher than that at the 8 h time point. Overall, the level of HPB was the highest for NNK treated mice, followed by (S)-NNAL with (R)-NNAL being the lowest. For instance, the levels of HPB from (S)-NNAL treated mouse serum proteins at the 8 h time point was 77.8% of that from NNK at the same time point while that from (R)-NNAL was only 20.2%.

Tissue DNA and protein-adducted HPB and Diol induced by NNK and NNAL enantiomers

Given that POB and PHB modifications in mouse serum samples can be indirectly quantified through the hydrolyzed HPB and Diol, we applied the same method to indirectly quantify the POB and PHB DNA adducts and the POB and PHB protein adducts from the lung tissues. The goal of this exploration was to determine the relative sensitivity of these different analyses and their relationships. A more sensitive method could be more applicable in clinical translation when the quantity of the clinical samples is limited, particularly if these adducts correlate positively with each. The profiles of HPB and Diol from lung DNA and lung protein are shown in Figure 3C and D, respectively, which showed similar patterns as 7-POB-G and 7-PHB-G from lung DNA but higher abundance.

Distinct adduct profiles among NNK, (R)-NNAL and (S)-NNAL at the 0.5 h time point

As discussed, NNK and NNAL can interconvert in vivo that they all can form POB and PHB modifications. Their in vivo conversion rates, however, have not been investigated. We envision that the adduct profiles at the earliest 0.5 h time point in this study may provide qualitative information about their conversion rate and the potential differences of (R)-NNAL and (S)-NNAL in bioactivation. As shown in Figure 4A–C, the methyl DNA adducts from NNK treatment were the highest while that from (S)-NNAL was 30.6% and that from (R)-NNAL was only 8.8% relative to NNK treatment in the lung tissue. As there were higher levels of 7-PHB-G adducts in the NNAL treatment groups in comparison to the NNK treatment group, our data raise the question whether the methyl adducts in vivo were mostly derived from NNK bioactivation instead of NNAL, which should be investigated in the future. Similarly, for 7-POB-G, the level from NNK was the most followed by (S)-NNAL with very little from (R)-NNAL, again suggesting that the in vivo conversion from (S)-NNAL to NNK was substantially higher than that from (R)-NNAL. This observation also holds true for the other time points that (S)-NNAL resulted in significantly higher levels of 7-POB-G than (R)-NNAL. At the 0.5 h time point, 7-PHB-G was not detected in the NNK treatment group while the level from (S)-NNAL appeared to be slightly higher than that from (R)-NNAL but not statistically significant. The higher levels of 7-PHB-G from (S)-NNAL were more obvious at the other time points, such as the 8 h time point (Supplementary Figure S3, available at Carcinogenesis Online). These data overall suggest that (R)-NNAL is less prone to be bioactivated to generate DNA damages, which is consistent with its low carcinogenicity.

Figure 4.

Adductomics and metabolites from NNK, (R)-NNAL or (S)-NNAL at the 0.5 h time point in A/J mice (n = 3–5). (A) Lung methyl adducts. (B) Lung 7-POB-G adducts. (C) Lung 7-PHB-G adducts. (D) Serum NNK. (E) Serum NNAL. (F) Urinary free NNAL. (G) Urinary NNAL-O-Gluc. (H) The ratio of urinary NNAL-O-Gluc: free NNAL. One-way analysis of variance for comparison among groups and Dunnett test for comparisons between groups. ns: non-significant; ∗P < 0.05; ∗∗P < 0.01; ∗∗∗P < 0.001; ∗∗∗∗P < 0.0001.

Serum levels of NNK and NNAL and the ratio of urinary NNAL-Gluc and free NNAL at the 0.5 h time point

Given the distinct adduct profiles at the 0.5 h time point for NNK and NNAL enantiomers, we quantified NNK and NNAL in the serum samples at this time point without being able to distinguish NNAL enantiomers (we have tried several reported methods to separate and quantify these enantiomers but not successful). As hypothesized, there were substantially higher levels of serum NNK upon (S)-NNAL treatment in comparison to (R)-NNAL treatment (Figure 4D), confirming that both (S)-NNAL and (R)-NNAL can be converted to NNK with the conversion more favorable from (S)-NNAL. At this time point, NNAL was detectable in the serum samples with NNK treatment (Figure 4E). Although there was no detectable PHB adduct in the lung tissues, we reason that the PHB adducts in the lung tissue has not been accumulated to the detectable level yet. Since the levels of the three types of adducts were lower in (R)-NNAL in comparison to (S)-NNAL, we quantified urinary free NNAL, NNAL-O-Gluc and calculated their ratio to explore if there is a difference in their clearance. The urinary levels of free NNAL in (R)-NNAL treated mice were the highest but not statistically significant from (S)-NNAL treated mice (Figure 4F). The urinary levels of NNAL-O-Gluc were also the highest in (R)-NNAL treated mice while the levels in (S)-NNAL treated mice were surprisingly the lowest (Figure 4G). The ratio of urinary NNAL-O-Gluc and free NNAL was also calculated as an indirect measure to evaluate glucuronidation-mediated NNAL detoxification, which was 0.70, 1.8 and 0.1, respectively, for NNK, (R)-NNAL and (S)-NNAL treated mice (Figure 4H). These distinct differences suggest that (R)-NNAL is more prone to glucuronidation than (S)-NNAL, supporting that (R)-NNAL is more efficiently detoxified in vivo, consistent with its lowest levels of adducts and carcinogenicity in A/J mice. (S)-NNAL, on the other hand, appeared to have the lowest clearance, particularly based on the amount of NNAL-O-Gluc and the ratio between NNAL-O-Gluc and free NNAL. In combination with its higher rate to be converted back to NNK, the adduct profiles and carcinogenicity of (S)-NNAL therefore were more comparable to NNK in A/J mice.

Global analyses of different adducts and their relationships in A/J mice

The abundance of POB and PHB adducts in the same lung tissues through three different analyses was compared—lung DNA 7-POB-G and 7-PHB-G, lung DNA HPB and Diol and lung protein HPB and Diol. The limit of detection for these analytes was similar. Figure 5A summarizes their abundance in 25 mg lung tissues from A/J mice with NNK treatment and lung tissues collected at the 8 h time point. The abundance of lung DNA HPB and Diol is about 4 times of 7-POB-G and 7-PHB-G, respectively, probably because there are other POB and PHB DNA adducts as we have observed before (40) that can contribute to the released HPB and Diol. Interestingly, the abundance of lung protein HPB and Diol is about 10–20 times of lung DNA HPB and Diol, probably because of the high abundance of proteins relative to DNA and the closer proximity of proteins to the reactive intermediates in comparison to DNA. As demonstrated later, these adducts have high linear positive correlations. These data suggest that HPB and Diol from lung proteins could be used as surrogates of lung DNA adducts, which may be more feasible for future clinical translation when the amount of clinical sample is limited.

Figure 5.

Global analyses of different adducts and their relationships in A/J mice treated with NNK, its metabolites (R)-NNAL or (S)-NNAL. (A) Comparison of the abundance of 7-POB-G, 7-PHB-G, HPB and Diol from lung DNAs and HPB and Diol from lung proteins in 25 mg lung tissues. (B) Correlations of O⁶-mG, 7-POB-G and 7-PHB-G adducts between paired lung and liver tissues in A/J mice (n = 37–44). (C) Correlations of POB adducts and PHB adducts in the paired lung DNA and lung proteins in A/J mice (n = 51). (D) Correlations of POB and PHB adducts in paired lung DNA and serum protein samples in A/J mice (n = 59). (E) Correlations between O⁶-mG, 7-POB-G and 7-PHB-G adducts in lung tissues in A/J mice (n = 56). For (B)–(E), the round dots represent the data from NNK group, the triangle-shape dots represent the data from (S)-NNAL group, the square-shape dots represent the data from (R)-NNAL group and the hexagon-shape dots represent the data from NC group. One-way analysis of variance for comparison among groups and Dunnett test for comparisons between groups. ns: non-significant; ^∗∗∗∗P < 0.0001.

The temporal profiles of these adducts in the lung and liver tissues and the serum samples showed that they do not synchronize perfectly within tissues or across tissues, which is not a surprise given the potential differences of NNK, (R)-NNAL and (S)-NNAL in absorption, distribution, metabolism and excretion in vivo and other variables. Another potential explanation for these variations is that there might be in vivo reservoirs, which could be specific for particular carcinogens as Hecht et al. has demonstrated for (S)-NNAL in F344 rats (41) and in smokeless tobacco users (42). Such carcinogen-specific reservoirs could be tissue dependent so that their metabolism does not follow the simple one-compartment pharmacokinetic model in some tissues. At the same time, the levels of these adducts typically were the highest at the 8 h time point with only a few exceptions that the levels at the 0.5 or 24 h time points were slightly higher but not statistically significant. This was consistent with our previous temporal data of specific DNA adducts (43,44). The temporal profiles between 0.5 and 8 h time points, however, are needed to comprehensively understand the pharmacodynamics of these different adducts. At the same time, the urine samples, if collected comprehensively, will provide a cumulative view of carcinogen metabolism that will be complementary to the adductomics in the tissues and serum samples, which should be explored in the future.

The correlation of the same type of adducts between liver and lung tissues was first explored, because peripheral adducts may be more feasible for future clinical translation than the tissue adducts while the peripheral adducts likely reflect the collections of adductomics from different tissues. The relationships of the adductomics between lung and liver tissues would provide preliminary evidence whether the peripheral adductomics likely resembles the adductomics in the target lung tissues. We therefore explored the correlations of three types of DNA adducts in the liver tissues with the corresponding adducts in the lung tissues. As shown in Figure 5B, all of the three adducts in the paired lung and liver tissues have moderate to high positive linear correlations (45,46). The correlation of 7-POB-G between lung and liver tissues was stronger than those of O⁶-mG and 7-PHB-G, which remain to be confirmed in future studies. Hecht et al. also observed similar DNA adduct profiles in different tissues in F344 rats upon NNK exposure, including lung, liver, oral cavity and pancreas, although paired analyses were not performed (36,47–50).

The correlations between DNA adductomics and protein adductomics in the lung tissues were also analyzed. Given the higher abundance of lung protein adducts in comparison to lung DNA adducts (Figure 5A), this correlation will provide justification whether lung protein adductomics can be employed as a biomarker for lung DNA adductomics particularly when the tissue samples are limited. Strong and positive correlations were observed for both POB and PHB adducts between the lung DNA samples and the paired lung protein samples (Figure 5C), supporting the use of lung tissue protein adductomics as surrogates for its DNA adductomics.

The correlation of the same type of adducts between plasma protein adductomics and lung DNA adductomics was analyzed as well. Such an analysis will provide information about how predictive plasma samples are for lung tissue adductomics. As shown in Figure 5D, again positive correlations were observed between paired lung DNA adducts and serum protein adducts in this mouse cohort, supporting the potential of serum adductomics in resembling the adductomics in the target lung tissues, which is of much greater translational potential. Better correlation was once again observed for HPB than Diol.

The relationships among different adducts within the same biological samples were then explored. We reason that such analysis will provide information whether different adductomics in the same sample may be complementary (of low correlation) or redundant (of high correlation) in cancer risk stratification. The correlations among methyl, POB and PHB DNA adducts in the target lung tissues are shown in Figure 5E. The POB adducts and the methyl adducts had a much better correlation than the PHB adducts and the methyl adducts while the POB and PHB adducts had the worst correlation. Does it suggest that the methyl adducts in vivo may be dominantly derived from NNK while the contribution from NNAL is minor? The adduct profiles at the 0.5 h time point also support this hypothesis. The rather poor correlations between POB and PHB adducts are expected since their formation are completely independent and may have some levels of competition given the interconversion between NNK and NNAL. Interestingly, the sum of POB and PHB had the best correlation with the methyl adducts, potentially because the methyl adducts could be formed from both NNK and NNAL.

Lastly, the correlations between the levels of different lung DNA adducts and serum protein adducts at the 8 h time point from NNK, (R)-NNAL and (S)-NNAL and their respective long-term tumor multiplicities were preliminarily explored with Pearson r and P values summarized in Table 1. It should be noted that because the short-term adduct profiles and the long-term tumor multiplicity data were obtained from two different mouse cohort, only the mean values for adducts and tumor multiplicity among NNK, (R)-NNAL and (S)-NNAL could be used for this correlation exploration, which resulted in limited sample size that these observations are preliminary and validation is absolutely needed. Among the lung DNA adducts, although none of these correlations were statistically significant, lung DNA methyl and POB adducts appeared to have decent correlations with tumor multiplicity while PHB adducts have the weakest correlation. Interestingly, total adducts appeared to have the best correlation. It should be noted that our results are consistent with the reports by Peterson et al. (51), showing a strong positive correlation between the levels of lung O⁶-mG and tumor multiplicity in A/J mice because the abundance of O⁶-mG is much higher than those of POB and PHB adducts. Similar correlations were observed between serum protein adducts and tumor multiplicity with better correlations for HPB than Diol. Again, the total adducts have the best correlation with the carcinogenicity with a P value of 0.019. Rigorous validation is ongoing to determine the correlations between the short-term adductomics and the long-term tumor status.

Table 1.

The correlations between the levels of different lung and serum adducts at the 8 h time point and the long-term tumor multiplicity

Tumor multiplicity versus	Lung DNA adducts				Serum protein adducts
	Methyl	POB	PHB	Total adducts	HPB	Diol	Total adducts
Pearson r	0.926	0.938	0.347	0.964	0.987	0.475	0.9991
P	0.247	0.224	0.775	0.172	0.103	0.516	0.019

Conclusion

Our results confirmed that NNK is the most carcinogenic and (R)-NNAL is the least carcinogenic while (S)-NNAL has similar carcinogenicity as NNK in A/J mice. Our study also systematically quantified three types of adducts from these carcinogens in the target lung DNA and proteins, the non-target liver DNA and serum proteins in a separate short-term study. Overall, NNK exposure results in the highest levels of adducts followed by (S)-NNAL with (R)-NNAL generating the least amount of adducts, which appeared to correlate well with their carcinogenicity although the sample size is small (Table 1). There are several additional explorations worth reiterating. First, we demonstrate that hydrolyzed HPB and Diol from lung tissue proteins are alternative to lung DNA POB and PHB adducts with increased abundance and sensitivity. Such analyses may be more practical for clinical translation when the clinical samples are limited in quantity. Second, this is the first study to characterize these adducts among liver, lung and serum samples (adductomic totality) and importantly demonstrate their overall positive correlations, suggesting that the peripheral serum adductomic profiling may be a practical alternate to estimate the adductomic profiles in the target lung tissues, which again may be translatable into cancer risk prediction. This study also provides preliminary qualitative data, demonstrating that (S)-NNAL is more prone to be converted to NNK than (R)-NNAL while (R)-NNAL is more prone to glucuronidation than (S)-NNAL in A/J mice, consistent with the results from Hecht et al. (26).

This study has several limitations as well. First, the A/J mice from the inbred mouse strain do not recapitulate the genetic and environmental heterogeneity among human smokers. Secondly, the single-dose i.p. carcinogen exposure in A/J mice has minimal physiological relevance to lung carcinogen exposure among smokers. The level of NNK used herein is higher than the exposure level of NNK among smokers (52) and the i.p. route of administration is different from tobacco inhalation. Thus, the adductomic profiles characterized herein could be substantially different from the adductomics in human smokers, particularly in the lung tissues. In addition, the dose of NNK used herein resulted in the formation of multiple lung tumors in every single mouse while cancer risk prediction in humans should be ‘Yes’ and ‘No’ status. The dose of NNK should be optimized in future studies so that some mice would develop lung tumor while some mice would not. Lastly, tumor multiplicity was used as the only criteria in this study while their pathological status has not been evaluated, which may also be related to the adductomic profile. Despite these limitations, we also rationalize that the single-dose NNK-induced lung carcinogenesis A/J mouse model is probably the ideal starting point to characterize NNK-derived adductomic totality and explore its potential in lung cancer risk prediction, because (i) the use of inbred mouse strain reduces genetic and environmental confounding variables; (ii) the level of NNK exposure among individual mice is more uniform by the single-dose i.p. NNK regimen than any other routes of administration, reducing exposure variations and (iii) the single-dose NNK regimen likely initiates carcinogenesis during a short period of time. These features likely amplify the carcinogenic risk signals, reduce the noises and therefore increase the chance of success to establish the relationship between the short-term adductomics and the long-term lung tumor status. Future studies are needed to rigorously explore whether the short-term non-invasive adductomic profiles, particularly the peripheral serum and potentially urinary adductomics, have the potential to predict the long-term tumor status, the outcome of which may transform and significantly improve lung cancer management.

Glossary

Abbreviations

Diol

1-(3-pyridinyl)-1,4-butanediol

HPB

4-hydroxy-1-(3-pyridyl)-1-butanone

NNAL

4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol

NNAL-N-Gluc

NNAL-N-glucuronide

NNAL-O-Gluc

NNAL-O-glucuronide

NNK

4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone

O ⁶-mG

6-O-methylguanine

Pas-1

pulmonary adenoma susceptibility

PHB

pyridylhydroxybutylation

POB

pyridyloxobutylation

Funding

The research reported in this publication was financially supported in part by the grants R01 CA193278 (C.X.) from National Institutes of Health, Lung Cancer Research Foundation Research Grant on Disparities in Lung Cancer (C.X. and Z.H.), Frank Duckworth Endowment College of Pharmacy University of Florida (C.X.) and Startup Fund University of Florida Health Cancer Center (C.X.). The content is solely the responsibility of the authors and does not necessarily represent the official views of any funding agencies.

Conflict of Interest Statement

None declared.