A Comparison of Urinary Biomarkers of Tobacco and Carcinogen Exposure in Smokers

The use of tobacco products contributes to >30% of human cancers in the United States (1). It is estimated that 90% of lung cancer cases are the result of tobacco use (2). The tobacco-specific carcinogen, 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NNK), and polycyclic aromatic hydrocarbons (PAH) present in tobacco smoke are believed to be the major causative agents for lung cancer in smokers (3, 4). In the past several years, reduction in exposure to tobacco toxins have been under consideration as a means to potentially reduce mortality and morbidity among those who continue to use tobacco products (5-7). This approach has been considered by both pharmaceutical and tobacco industries. To assess the potential to reduce toxin exposure in tobacco users, studies have been carried out to determine the effect of cigarette reduction on carcinogen exposures. In a recently completed study, we have used the NNK metabolites 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol (NNAL) and its glucuronide (NNAL-Gluc) and a PAH metabolite 1-hydroxypyrene (1-HOP) as biomarkers of carcinogen exposure (8, 9). To effectively use these carcinogen biomarkers, it is important to determine the relationship between them and biomarkers of tobacco uptake, such as nicotine, cotinine, and the minor alkaloid, anatabine.

To exert its carcinogenic potential, NNK is metabolically activated by α-hydroxylation to reactive intermediates that are carcinogenic (10). NNK is also metabolized by carbonyl reduction to NNAL, which may also be metabolically activated to carcinogenic intermediates. NNAL is detoxified by glucuronidation (11). NNAL and its glucuronide conjugate, NNAL-Gluc, have been established as excellent biomarkers of NNK exposure (12). In many studies, 1-HOP has been shown to be a good biomarker of exposure to PAH (12, 13). 1-HOP is the major metabolite of the abundant but noncarcinogenic PAH, pyrene, which is always present in mixtures of PAH (13, 14).

Nicotine is the major addictive agent in tobacco, and its metabolite, cotinine, is commonly used as a biomarker of tobacco and/or nicotine exposure (15). Nicotine is metabolized to cotinine in two steps: P450 2A6 catalyzed 5′-oxidation to the corresponding iminium ion and oxidation of the iminium ion to cotinine (16-18). Cotinine is further metabolized to trans-3′-hydroxycotinine, and nicotine, cotinine, and trans-3′-hydroxycotinine are all metabolized to glucuronide conjugates (18). Trans-3′-hydroxycotinine is the major urinary metabolite of nicotine (19). In most smokers, nicotine plus these five metabolites account for >80% of the nicotine dose received (19, 20).

Other potential biomarkers of tobacco exposure that have not received a great deal of attention are tobacco alkaloids other than nicotine. One of the tobacco alkaloids is anatabine. The use of anatabine as a measure of exposure is important because some of the reduced exposure methods involve using medicinal nicotine products. Because these products contain nicotine, use of nicotine or cotinine as a marker for tobacco uptake would not be appropriate. Therefore, Jacob et al. (21, 22) have initiated the use of the minor tobacco alkaloids, anatabine and anabasine, as biomarkers of tobacco use in individuals receiving nicotine replacement therapies. Anatabine and anabasine are structurally similar to nicotine (Fig. 1) and are present in tobacco at levels 40- to 300-fold lower than in nicotine. Nicotine accounts for ∼95% of the alkaloids in tobacco, whereas anatabine and anabasine account for 2% to 3% and 0.2% to 0.5%, respectively. In mainstream tobacco smoke, the levels of anatabine and anabasine are reported to be similar, 3 to 14 μg per cigarette, compared with 800 to 3,000 μg nicotine per cigarette (23). Little is known about the metabolism of either anatabine or anabasine. There is a single publication in the literature on the metabolism of anabasine in which it was reported that hepatic microsomes from rodents metabolize anabasine by N-hydroxylation of the piperidine ring (24).

In our recently completed cigarette reduction study, participants were encouraged to use either nicotine gum and/or nicotine patch as a means by which to reduce their cigarette consumption (8, 9). At baseline smoking levels, prior to entering the cigarette reduction phase of this trial, urine samples were collected from all subjects. All baseline urine samples were analyzed for the concentrations of NNAL, NNAL-Gluc, 1-HOP, anatabine, nicotine, cotinine, total trans-3′-hydroxycotinine (free plus glucuronidated), and the glucuronide conjugates of nicotine and cotinine. The subject of this report is the investigation of intraindividual variation in these biomarkers across a 7-week time period when smoking was reported to be unchanged. In addition, the correlation of these biomarkers with cigarettes per day (CPD) and with each other was examined. The primary aim of these correlations was to determine the validity of using anatabine as a biomarker of tobacco exposure and the relationship of cotinine to other nicotine metabolites in multiple urine samples obtained from a smoker over time.

Anatabine and anabasine were purchased from Toronto Research Chemicals (North York, Ontario, Canada). 5-Ethylnornicotine was synthesized from 5-bromonicotinic acid (Sigma-Aldrich, Milwaukee, WI) following published procedures (25, 26). N-propyl derivatives of anatabine, anabasine, and 5-ethylnornicotine were prepared as described previously (27) and purified by high-performance liquid chromatography. The high-performance liquid chromatography system was that used previously for nicotine metabolite analysis (28), except for 5-ethylnornicotine in which the mobile phase contained 10% methanol. Anatabine and its N-propyl derivative eluted at 14.7 and 34.2 minutes, respectively, and anabasine and N-propylanabasine eluted at 21.3 and 43.1 minutes, respectively. 5-Ethylnornicotine eluted at 14.1 minutes and its N-propyl derivative at 29.8 minutes.

The individuals included in this study were subjects assigned to a waitlist control group in a smoking reduction study (8). These subjects continued smoking over the course of 8 weeks during which time four baseline urine samples were collected. The first morning urine void was collected at each baseline visit. Subjects who reported smoking between 20 and 40 CPD were recruited. Throughout the study, subjects kept a daily diary of the number of cigarettes they smoked. Of the 151 subjects in the reduction study, 49 were randomly assigned to the waitlist group. Two subjects in the waitlist group dropped out prior to providing four baseline urine samples. The waitlist subjects were asked not to change their cigarette consumption prior to entering the reduction phase of the study. Waitlist individuals provided urine samples on the same schedule as individuals in the reduction study, at week 1 (visit 1), week 2 (visit 2), week 6 (visit 3), and week 8 (visit 4). After this time, they entered the reduction phase of the study. Subject characteristics and urine collection procedures were as described previously (8).

Urinary nicotine, cotinine, and trans-3′-hydroxycotinine levels were quantified by gas chromatography/mass spectroscopy as described previously (29, 30). Urinary anatabine was analyzed using a modification of a method by Jacobs et al. (8, 27). Briefly, 5-ethylnornicotine, as an internal standard, was added to 2 mL urine, which was partially cleaned up on a C18-PrepSep column (Fisher Scientific, Fair Lawn, NJ). The sample was then acidified and derivatized by reductive alkylation with propionaldehyde and NaBH₄, NaOH was added, and the sample was extracted with toluene. The toluene layer was then extracted with acidic H₂O. The H₂O layer was removed, 50% K₂CO₃ was added, and the propylamines were extracted into methylene chloride containing 200 μL methanol. Methylene chloride was removed under a gentle stream of N₂, and the sample in methanol was analyzed by gas chromatography/mass spectroscopy with select ion monitoring for N-propylanatabine, m/z 202 (M⁺) and m/z 173 (M-C₂H₅)⁺, and the N-propyl derivative of 5-ethylnornicotine, m/z 218 (M⁺) and m/z 189 (M-C₂H₅)⁺. The urinary concentrations of anatabine were calculated from the ratio of the integrated peak areas of the (M-C₂H₅)⁺ ions for the N-propyl derivatives of anatabine (m/z 173) and 5-ethylnornicotine (m/z 189). Anabasine (m/z 175) was analyzed in some samples; however, a coeluting peak made it difficult to monitor this alkaloid in all samples.

Urinary NNAL and 1-HOP were analyzed as described previously (29, 31, 32).

Pearson correlation coefficients were calculated between the biomarker measurements at each two visits, between each biomarker and the mean CPD values (mean of the daily cigarettes number recorded in the week prior to the collection of the urine sample), and between each pair of biomarkers (mean of four visits). Because the biomarker data were not normally distributed, log transformations were applied to all the data prior to the correlation analysis.

Forty-seven subjects completed the 7-week study and supplied a urine sample at each of the four visits. Each subject kept a daily log of cigarette use. The average CPD was calculated for the 7 days prior to each visit and compared with the number of cigarettes smoked 1 day prior to each visit. For 70% of the 188 visits (47 subjects, four visits each), this value agreed within 10%. There was no significant difference between the average numbers of CPD across the four visits, whether this value was calculated using the values reported for the day prior to a visit or the mean of the week prior (Table 1). An average number of CPD during the 7-week period was calculated for each subject using either the single value reported for the day prior to a visit or the average of the 7 days prior to a visit. The mean of each of these measures for the 47 subjects was determined to be 24.5 ± 8.04 for the day prior or 24.4 ± 6.9 for the weekly average. The lowest reported CPD (weekly average) was 13.7 and the highest was 42.1. For 46 of the subjects, urinary cotinine levels at individual visits ranged from 800 to 12,000 ng/mL (4.5-68 nmol/mL) or 900 to 15,000 ng/mg creatinine. One subject, who reported smoking 30 CPD at all four visits, had urinary cotinine levels ranging from 1,176 to 126 ng/mL. This subject was excluded from further analysis because a value of 126 ng/mL is inconsistent with active smoking.

The levels of nicotine and cotinine were measured in all samples before and after treatment with base. Base treatment cleaves the N-glucuronide conjugates of nicotine and cotinine. Therefore, urinary levels of nicotine and cotinine are reported as free and total, where total is a measure of the free plus glucuronide conjugates. In addition, total trans-3′-hydroxycotinine levels were determined after treatment of urine with β-glucuronidase. β-glucuronidase cleaves both N-linked and O-linked glucuronides of trans-3′-hydroxycotinine. The O-linked glucuronide is believed to be the major glucuronide of trans-3′-hydroxycotinine in smokers' urine (18). However, the formation of a N-glucuronide by human liver microsomes has been reported (33). The mean values for nicotine and its metabolites for each of the four visits are reported in Table 1. An average concentration of each biomarker for the four visits was calculated for each subject and the mean of those values, called the mean of means, was determined (Table 1). None of the five biomarkers of nicotine exposure varied significantly across the four visits. However, there was significant variation among subjects. As noted above, average reported CPD varied 3-fold; however, the urinary concentration of cotinine varied 7-fold and trans-3′-hydroxycotinine >60-fold among subjects.

The variation in the urinary concentration of nicotine metabolites is in part due to individual differences in nicotine metabolism. These differences are reflected in the data presented in Table 2. Data in Table 2 are the mean urinary concentrations of nicotine and each of the major nicotine metabolites expressed as a percentage of the total metabolites, which were quantified plus nicotine. The major urinary metabolite was total trans-3′-hydroxycotinine, which on average accounted for 51.6% of the total excreted; however, the range was from 3.7% to 84.4%. Cotinine and its glucuronide accounted for 11.4% and 21.6% of the total, respectively. In contrast, free nicotine was excreted at a higher concentration than was nicotine glucuronide (9.33% versus 6.19%).

Data in Table 2 were calculated from the average percentages determined for each subject from the four visits. Similar values were obtained when analyzing each of the four visits independently (data not shown). To further analyze the consistency of nicotine and nicotine metabolite urinary concentrations across visits, the correlation of each biomarker concentration between each pair of four visits was determined (Table 3). The concentrations of free cotinine and total cotinine, as well as the percent cotinine glucuronidation, were significantly correlated (P < 0.0001) between all four visits (Table 3). Trans-3′-hydroxycotinine was also significantly correlated between all visits (P < 0.0001 or P < 0.001, visits 2 and 4). Free nicotine and total nicotine were less well correlated (P > 0.5 for some pairs of visits). These data are consistent with the relatively short half-life of nicotine and the greater dependence of nicotine concentration on the time at which the last cigarette was smoked prior to urine collection.

In our smoking reduction study, nicotine replacement therapies were used and anatabine was employed as a measure of reduced smoking within a subject. Therefore, it was important to determine the stability of urinary anatabine concentrations across the same time frame when smoking levels were reported to be unchanged. The mean urinary concentration of anatabine was not significantly different across the four visits (Table 1). The mean of the average concentration for the four visits was 44.1 ± 28.1 pmol/mL or 7.06 ± 4.48 ng/mL (n = 46 subjects). The distribution of the average anatabine concentrations is illustrated in Fig. 2. The mean concentrations ranged from <1.0 to 20.1 ng/mL. Anatabine was not detected, <0.8 ng/mL (4.9 pmol/mL) in urine samples from eight visits of nine subjects. The highest urinary concentration of anatabine at a single visit was 44 ng/mL. There was a significant correlation for anatabine between visits (P < 0.001 to P < 0.05; Table 3).

The urinary concentrations of the carcinogen biomarker NNAL and the PAH 1-HOP were also determined at each of the four visits and the mean levels were compared. The levels of total NNAL (free plus glucuronidated) were not significantly different at visit 1, 3, or 4. However, at visit 2, the mean concentration was ∼11% lower (Table 1). Similar results were obtained when the values were expressed per milligram creatinine (data not shown). The correlations between total NNAL concentrations for visits 1, 3, and 4 were statistically significant (P < 0.0001) and very similar and NNAL concentrations were less well correlated between visit 2 and other visits (Table 3). The mean concentration of 1-HOP varied from 1.4 to 2.42 pmol/mL across the four visits (Table 1). However, 1-HOP concentrations were significantly correlated between visits (P < 0.0001 to P < 0.005).

The average urinary concentration of each biomarker for the four visits was used to determine the correlation between pairs of biomarkers for the 46 subjects. None of the biomarkers correlated with CPD (Table 4). With regard to this lack of correlation, it is important to note that, whereas the range of CPD was from 14 to 44, for 75% of the subjects, the range was much smaller (20-28 CPD). The best correlation for biomarkers was between free and total nicotine (r = 0.875; P < 0.0001). The correlation of anatabine with nicotine, either free or total, was also very good (r = 0.753 and 0.773, respectively; P < 0.001). Anatabine was well correlated with NNAL (r = 0.633; P < 0.0001) and somewhat less well correlated with cotinine, either free (r = 0.465; P < 0.001) or total (r = 0.514; P < 0.001). The correlation between total NNAL and nicotine and all its measured metabolites was statistically significant (P < 0.0001 to P < 0.05). There was no statistically significant correlation between 1-HOP and any of the other biomarkers.

The correlation between the extent of glucuronidation of nicotine, cotinine, and NNAL across the four visits was also determined (Table 5). The average percentage of nicotine glucuronidation and cotinine glucuronidation was well correlated (r = 0.633; P < 0.0001), whereas the percentage of NNAL glucuronidation did not correlate with either nicotine or cotinine glucuronidation.

Recently, several potential harm reduction strategies have been discussed as possible means by which to reduce tobacco-related disease (5). These strategies include reduction in the amount of smoking and the use of smokeless tobacco products or modified cigarette products (5-7, 34, 35). A first step in assessing a potential reduction in harm by any of these approaches is an accurate assessment of tobacco toxin exposure. Cotinine is an accepted biomarker of nicotine exposure, and NNAL and NNAL-Gluc are good biomarkers of NNK exposure (12, 15, 19). We report here a comparison of the urinary levels of these biomarkers and two additional biomarkers, the minor tobacco alkaloid anatabine and a PAH, 1-HOP. To our knowledge, this is the first report to determine the relationships between multiple biomarkers of carcinogen and tobacco alkaloid exposure in the same population of smokers. The relationships were determined at multiple times over a 7-week period. Understanding the individual variability of these biomarkers over time is critical to assessing their usefulness as biomarkers of toxin exposure in individuals using tobacco harm reduction strategies, either limited cigarette consumption or use of modified tobacco products.

There are several key findings from this study. First, most of the measures showed significant intraindividual correlations. The level of cotinine (free or total) was well correlated across visits. On the other hand, nicotine levels were not significantly correlated between all visits. However, this is consistent with the shorter half-life of nicotine compared with cotinine, which would result in the urinary concentration of nicotine being more dependent than cotinine on the time of urine collection relative to when the last cigarette was smoked. Urinary trans-3′-hydroxycotinine levels were significantly correlated between visits; however, the magnitude of that correlation was not as consistent as that for total cotinine. Therefore, total cotinine levels seem to be the most robust urinary measure of nicotine exposure in smokers. Anatabine, 1-HOP, and total NNAL also were significantly correlated between visits.

Second, these biomarkers were not correlated with CPD (Table 4). The lack of correlation of these measures with CPD is not surprising and may even have been predictable given the relatively small number of subjects and the limited range in the number of cigarettes smoked per day. Among the 46 smokers in this study, the reported CPD ranged from 13 to 46; however, 75% of the subjects reported smoking from 20 to 28 CPD. Several factors, which will vary across smokers, including smoking inhalation, biomarker metabolism, and time of urine collection relative to smoking, could have a significant influence on the urinary concentrations of nicotine and its metabolites, NNAL, NNAL-Gluc, and anatabine. In addition, inaccurate reporting of CPD would contribute to the lack of correlation between CPD and nicotine and its metabolites. Inaccurate reporting may have been minimized somewhat in the current study by the use of a smoking diary.

A relatively weak correlation between reported CPD and cotinine levels in urine, blood or saliva has been reported in many studies (15, 36). A recent Australian study reported an ∼20-fold variation in urinary cotinine levels in smokers (n = 40) who reported smoking 20 CPD (36). A similar magnitude of variation has been reported for saliva cotinine concentrations in smokers of 20 CPD (n = 55; ref. 37). Clearly, 20 CPD is a convenient value for subject to report, so inaccuracy in this number may influence the variability in cotinine concentration that was observed. However, it is unlikely that this accounts for the majority of the large variation observed. The metabolism of nicotine to cotinine is reported to vary >2-fold among smokers (38), and in a controlled smoking study, the daily nicotine intake for the same number of CPD varied up to 4-fold (39). In a study of 116 smokers, in which cigarette consumption was recorded by the smoker on a handheld computer comparing, a nonlinear relationship between saliva cotinine levels and daily smoking levels was observed (40). In that study, cotinine levels rose rapidly up to 10 CPD, then leveled off, and rose again above 20 CPD. These data reemphasize the importance of not relying on CPD as an exposure measure (41).

On the other hand, most biomarkers correlated well with each other. Of note, urinary anatabine concentrations were significantly correlated with both free and total cotinine (Tables 4 and 5). Furthermore, among the biomarkers we measured, anatabine was most closely correlated with nicotine, suggesting its urinary concentration, like that of nicotine may be more sensitive to when the last cigarette was smoked. Little is known about the metabolism and clearance of anatabine in smokers. However, the relative consistency in anatabine concentrations over time and its correlation with urinary nicotine and cotinine concentrations support the use of this minor alkaloid as a biomarker for tobacco exposure. Additionally, the strong correlation of anatabine with the tobacco-specific nitrosamine, NNAL, strengthens the argument that anatabine is a good biomarker for tobacco exposure in individuals using nicotine replacement therapies. As shown in previous studies, total NNAL was significantly correlated with nicotine and cotinine (10, 12, 29, 31). 1-HOP did not correlate with any of the other biomarkers. This is not unexpected because from 30% to 50% of the dose of pyrene does not come from tobacco exposure but from other environmental sources including diet (12).

Finally, the results provide additional evidence that the percent glucuronidation of cotinine measured in urine is reproducible and potentially useful as a reflection of UDP-glucuronosyltransferase activity. The relative urinary concentrations of nicotine and its metabolites reported here for an average of four spot urine samples are similar to those reported previously for 24-hour urine samples (Table 2; refs. 19, 20). The relative percentage of cotinine glucuronide was greater in the present study compared with the other two. However, the percentage of cotinine excreted as its glucuronide conjugate was quite similar, 63% in our study versus 54% and 60% in the other two studies. In our study, this percent cotinine glucuronidation was strongly correlated between visits (Table 3). In addition, as reported by Benowitz et al. (20), the percentage of cotinine excreted as a glucuronide was well correlated with the percentage of nicotine excreted as a glucuronide conjugate (Table 5). These data are consistent with our in vivo studies that support UGT1A4 and UGT1A9 as common catalysts of nicotine and cotinine glucuronidation in human liver microsomes (42). The extent of NNAL glucuronidation was not correlated with either nicotine or cotinine glucuronidation. This may suggest a different enzyme catalyst or may simply be due to a greater intraindividual variation in the percentage of NNAL glucuronidation. Evidence of a significant amount of intraindividual variation in the percentage of NNAL excreted as its glucuronide is reflected in the lack of correlation for this variable across visits (Table 3). This increased intraindividual variation may be due to the much lower dose of NNAL compared with nicotine and cotinine.

It should be noted that the mean urinary levels of anatabine we report here, 7.1 ng/mL (44.1 pmol/mL; Table 1), were somewhat lower than the 11.5 or 12 ng/mL reported previously in two small studies by Benowitz et al. (22, 27). However, it was 3-fold lower than a more recent study in which the mean urinary level of anatabine was reported to be 22 ± 23 ng/mL (19 ± 14 ng/mg creatinine; ref. 21). It is unclear why this difference is so large. A higher level of smoking may have contributed because the majority of the subjects in that study smoked >20 CPD. In addition, it is possible that the time of urine collection may have contributed.

We report here the first study in which the relationships between several carcinogen and tobacco alkaloids biomarkers are investigated over time. In summary, the important findings are the following: (a) Demonstration of a strong correlation between anatabine and urinary NNAL, a metabolite of the tobacco-specific carcinogen NNK, confirming the potential usefulness of anatabine as a biomarker of tobacco exposure for smokers using nicotine replacement therapy; (b) Support for the use of total cotinine as the most consistent and reliable urinary marker of nicotine exposure; and (c) Reiteration of the findings that CPD is not an adequate measure of tobacco toxin exposure. In addition, the data provide additional evidence that the percent glucuronidation of cotinine measured in urine is reproducible and potentially useful as a reflection of UDP-glucuronosyltransferase activity.