Differential Effect of NAT2 on the Association between Active and Passive Smoke Exposure and Breast Cancer Risk¹

Cigarette smoking has been examined as a possible risk factor for breast cancer for over a decade. Until recently, the evidence from epidemiological studies considering only active smoking and breast cancer risk has been inconsistent. This has been proposed to be because of the opposing protective effect through lowered estrogen levels because of smoking (1, 2) and possible direct carcinogenic effects of tobacco constituents that have been found in breast fluid (3, 4). However, exposure to smoking by spouse was then found to be associated with a 2-fold increase in breast cancer risk, both in a cohort and a case-control study (5, 6), so that passive smoking was also considered as a possible risk factor for breast cancer. Indeed, subsequent studies consistently showed that women exposed to passive smoking are at increased risk of breast cancer compared with women never exposed to either active or passive smoke (7, 8, 9, 10, 11). The association of breast cancer both with active smoking and with passive was even of a similar order of magnitude of ∼2-fold.

The individual variation in metabolism of tobacco smoke constituents has been postulated to play a role because aromatic amines in tobacco smoke when metabolically activated by N-acetylation can cause DNA damage and be involved in breast carcinogenesis (12, 13). The Phase II enzyme NAT2³ is one of the xenobiotic metabolizing enzymes that detoxify and/or bioactivate aromatic and heterocyclic amines. A low activity level associated with slow acetylation of occupational exposure to aromatic amines has been shown to increase bladder cancer risk (14, 15). Fast NAT2 acetylation, on the other hand, is implicated in colorectal cancer, because this enzyme can also activate amine metabolites from heterocyclic amines by O-acetylation (16).

Ambrosone et al. (17) first found evidence that slow acetylation of carcinogenic tobacco constituents increases susceptibility for breast cancer among active smokers. Additional studies (18, 19, 20), including a study that also accounted for passive smoke (20), yielded inconsistent results. In the latter study, the effect modification by the NAT2 genotype of the postmenopausal breast cancer risk associated with active smoking completely changed when passive smokers were included in the reference group. However, this was based on only 165 postmenopausal women. The authors postulated the involvement of heterocyclic amines in cigarette smoke. Therefore, we investigated the possible impact of ETS when considering the association among NAT2 genotype, cigarette smoking, and breast cancer risk, as well as a possibly differential effect of NAT2 genotype by type of smoke exposure in a case-control study, primarily among premenopausal women.

The present study, conducted in 1999–2000, is based on a population-based breast cancer case-control study (1992–1995) of 706 breast cancer patients and 1381 controls in southern Germany (21). In this preceding study, complete ascertainment of German-speaking women residing in the two study areas with incident in situ or invasive breast cancer diagnosed before the age of 51 years was achieved by surveying all of the hospitals serving the population. For every patient, two controls matched by age and study region were selected randomly from lists of 5000 female residents obtained from the population registries of each study region. Of the women contacted, 70.1% of eligible cases and 61.2% of eligible controls participated. After giving written informed consent, all of the study participants completed a self-administered questionnaire assessing demographic, anthropometric, and other known or putative risk factors, and provided a blood sample. The study was reviewed and approved by the ethics committee of the University of Heidelberg, Heidelberg, Germany.

The initial questionnaire contained five questions pertaining to active smoking. To obtain detailed information on lifetime history of active as well as passive smoking, all of the participants were recontacted by letter in August 1999 and invited to participate in a computer-assisted telephone interview. A follow-up through the population registries to obtain current addresses revealed 115 deceased cases and 3 deceased controls. The interviews were conducted from September 1999 to May 2000 by trained interviewers blinded to the case-control status of the participant. Valid telephone numbers were available for 572 cases and 1353 controls. Of these, 81.8% of the cases and 81.2% of the controls gave an interview, 10.3% of the cases and 12.9% of the controls refused participation, and 7.9% of the cases and 6.0% of the controls could not be reached. In summary, 66.3% of the cases and 78.9% of the controls of the original study population participated in this aspect of the investigation.

We attempted to record a comprehensive smoking history. Women were asked when they began smoking, the type of product, amount and frequency of tobacco usage, the intensity of inhalation, and the date of cessation or change in their smoking habit. The same questions were posed for up to eight different phases of active smoking habits.

Passive smoking in childhood, in adult life, and at work was assessed. For childhood, questions pertained to the number of smokers who had lived in the household, onset and age of exposure, as well as number of hours and days each of the persons had smoked in the presence of the participant during weekdays and on the weekend, and the smokiness of the room. Childhood exposure was truncated at 18 years, and exposure data beyond this age were added to the adult household exposure.

For the adult household, information was sought on whether the women had lived with a smoking partner, onset and end or change of smoking exposure, daily amount and type of tobacco product smoked, number of hours and days of passive exposure, as well as the smokiness of the room. These questions were posed for up to eight various phases. Exposures because of other household members as well as up to eight different phases of smoke exposure at the workplace were assessed with the same questions as for childhood exposure. Adult exposure comprised of childhood exposure over the age of 18, exposure through partner or other household members, and at work.

All of the information was truncated at the reference date, which was date of diagnosis for cases and date of completion of the first questionnaire for controls. Menopausal status was assigned according to the reported state a year before the reference date. Women with previous hysterectomy not accompanied by bilateral oophorectomy were classified as unknown.

Blood samples was available from 95% of cases and 82% of controls in the original study population. Restricting to women who have at least one parent of German nationality, comprising 91% of cases and 96% of controls, and for whom genotyping was successful, this analysis was based on 422 cases and 887 controls.

Genomic DNA was extracted from the EDTA blood samples using Blood and Cell Culture DNA kits as described by the manufacturer (Qiagen GmbH, Hilden, Germany) and by a standard salt precipitation-based method. The analysis of NAT2 was based on PCR amplification and the use of specific hybridization probes to detect the point mutations at nucleotide positions 481, 590, 803, and 857 in two different amplification products as described previously in Wikman et al. (22). NAT2 polymorphic sites were detected by capillary-based real-time PCR followed by melting curve analysis. PCR and melting curve analyses were performed in 10 μl volumes in glass capillaries (Roche Diagnostics) using: 1× PCR buffer, 2.5 mm MgCl₂, 200 μm deoxynucleotide triphosphate, 0.1% BSA, 0.5 units Taq-polymerase, 0.15 μm of each probe, and 0.5 μm or 1 μm of each primer (TIB MOLBIOL, Berlin, Germany). Approximately 10 ng of genomic DNA (gDNA) was used as template. The cycling conditions were as follows: initial denaturation at 95°C for 2 min following 45 cycles of denaturation at 95°C for 0 s, annealing at 52°C (NAT2: 803 and 857) or 58°C (NAT2: 481 and 590) for 10 s, and elongation at 72°C for 12 s with a ramping rate of 20°C/s. Melting curve analyses were performed with an initial denaturation at 95°C for 60 s, 60 s at 40°C, followed by slow heating of the samples to 80°C with a ramping rate of 0.3°C/s and stepwise fluorescence detection. The melting curves were converted to melting peaks by plotting the negative derivatives of fluorescence against temperature (−dF/dT).

Alleles with the point mutations at nt 590 were classified as NAT2∗6A, at nt 857 as NAT2∗7B, at nt 481 alone as NAT2∗5A, at nt 803 alone as NAT2∗5C, and alleles with both the nt 481 and nt 803 polymorphisms as NAT∗5B. Genotyping for the above-mentioned alleles results in >95% correlation to described phenotyping methods (23, 24). In a previous study of 1088 unrelated individuals in Germany, only the NAT2∗4, ∗5A, ∗5B, ∗5C, ∗6A, and ∗7B alleles could be detected at a frequency of 1% or more (25).

The association between active/passive smoking and breast cancer risk by acetylator status was assessed by maximum likelihood estimates of ORs and their 95% CIs obtained with multivariate conditional logistic regression analysis using the PHREG procedure of the statistical software package SAS release 6.12 (SAS Institute, Cary, NC). Analyses were stratified for age in 5-year intervals to ensure adequate numbers in subgroups of the study population according to genotype and smoking status.

An ever-active smoker was defined to have smoked >100 cigarettes in her lifetime. Among ever smokers, current smokers were women who had smoked regularly within the 1 year preceding the interview. Ever-passive smokers were women with an average exposure to passive smoke of >1 h a day for at least 1 year either in childhood or in adulthood. This average exposure was obtained by multiplying the average hours per day of each exposure phase with the duration in years of that phase and dividing the sum overall phases separately for childhood and adulthood by the total years of passive smoke exposure. Missing data on hours per day for 36 cases and 62 controls were replaced with the mean hours per day of the exposed controls for the particular source of exposure. All of the analyses redone using a strict yes/no definition of ever-passive exposure did not essentially alter the estimates.

Lifetime passive exposure was quantified in hours per day-years by multiplying the number of hours of ETS exposure per day with the number of years of exposure. The mean daily average hours of ETS exposure during childhood was divided by the number of smokers in the household. For adulthood exposure, only exposures because of partner and at work (and not all of the ETS phases) were considered to limit overlapping of exposure phases and to facilitate comparisons with recent study reports. For the analyses of passive smoking among nonsmokers the median of 47 h/day-years of nonactively smoking controls was used to dichotomize the participants into having been exposed 1–50 or 51+ h/day-years.

The reference group for the analyses of active and passive smoking were women who fell neither in our definition of active nor of passive smoker. In analyses concerning active smoking a separate category of only passively exposed women was always included in the models. Effects of passive smoking were analyzed in detail only among nonactive smokers. Women were considered to be phenotypically “rapid” acetylators if they were carriers of at least one wild-type NAT2∗4 allele. Noncarriers of wild-type NAT2∗4 alleles were classified as slow acetylators.

Adjustment was made for several relevant variables influencing breast cancer risk, such as history of breast cancer in first degree relatives, number of full-term pregnancies, total number of months of breastfeeding, menopausal status, average daily alcohol intake, and attained educational level. Other factors such as study region, use of oral contraceptives, age at first full-term pregnancy, and age at menarche did not influence the estimates, and were, therefore, not included in the statistical models. Multiplicative interaction between the genetic and tobacco smoking variables was measured by using multiplicative terms and evaluated by the likelihood ratio test.

The distribution of demographic and risk factors, and of smoking status in the study participants of the telephone interview closely resembles that in the original study population. Fifty-five percent of cases and 52.5% of controls were active smokers in this analysis compared with originally 53.8% and 53.3%, respectively. There was no appreciable difference in the daily number of cigarettes as reported in the first questionnaire. A comparison between the smoking behavior of deceased cases and cases of the original study showed only a slightly higher mean daily number of cigarettes smoked by deceased cases than the complete patient group, mean (SD) = 18.4 (11.4) versus 15.6 (9.8). There were some differences between rapid and slow acetylators in the proportions of cases to controls for the following factors: menopausal status, education level, parity, age at first birth, use of oral contraceptive, and total duration of months of breastfeeding. However, no significant interaction was found between these variables and acetylation status or for family history of breast cancer or average alcohol intake.

The overall proportion of slow acetylators did not differ between cases (57.6%) and controls (60.0%), and was almost the same as in the original study population (56.2% in cases and 60.1% in controls; Table 1). Smoke exposure was associated with increased ORs, being 1.5 (95% CI, 1.0–2.2) for passive smoking only and 1.4 (95% CI, 0.9–2.2) for current active smoking.

Heterogeneity by acetylator status was evident for the tobacco smoking variables. When analysis was performed neglecting the information on passive smoking (which was not available in most previous studies), ever-active smoking compared with never-active smoking was associated with a significantly reduced risk of 35% in rapid acetylators, whereas risk was slightly elevated in slow acetylators (Table 2). The interaction between NAT2 status and ever-active smoking was statistically significant (P = 0.03). When passive smokers were excluded from the reference group, risk for breast cancer associated with active smoking was also higher for slow acetylators than rapid acetylators; however, the reduced risk in rapid acetylators disappeared. More specifically, ORs were 1.22 and 1.02 for current and ex-smoking in rapid acetylators, and 1.67 and 1.16 for current and ex-smoking in slow acetylators. On the other hand, breast cancer risk associated with passive only smoking was greater in rapid than slow acetylators (2.02 and 1.19, respectively). A test for interaction between NAT2 status and smoking status (categorized as never-exposed, passive only, and ever-active smokers) was of borderline statistical significance [χ²_LR = 5.423; P (χ², 2 df) = 0.07].

Active smoking variables, such as pack-years, duration of smoking, and time because cessation of smoking, showed dose-response relationships with breast cancer risk in slow acetylators but not in rapid acetylators (Table 2). Tests for trend among slow acetylators excluding passive smokers were statistically significant for pack-years (P = 0.03) and for duration of smoking (P = 0.006) but not for time since cessation of smoking (P = 0.10). Pack-years was not associated with breast cancer risk in rapid acetylators, but the test for interaction between NAT2 status and pack-years did not reach statistical significance [χ²_LR = 5.753; P (χ², 3 df) = 0.12].

Restricting the analysis to the 542 nonsmoking women (174 cases and 368 controls), ever-passive smoking was associated with a higher risk in rapid than in slow acetylators, with 1.98 (95% CI, 0.96–4.09) versus 1.16 (0.66–2.04; Table 3). The test for interaction between genotype and passive exposure was not statistically significant (P = 0.36). The increased risk because of smoke exposure in rapid acetylators was stronger during adulthood than in childhood; however, CIs are wide. Duration of passive exposure in years did not seem associated with breast cancer risk. It likewise made no difference whether the passive smoke exposure was current or had taken place in the recent past (data not shown). However, lifetime passive smoke exposure in hours per day-years, showed a positive dose-response association with breast cancer risk in rapid acetylators but also a weak association for slow acetylators (test for trend in three categories, never exposed, 1–50, and 51+ h/day years in rapid acetylators: P = 0.05; in slow acetylators: P = 0.21).

Overall NAT2 genotype was not associated with breast cancer risk, nor did any interaction appear to be present between acetylation status and other relevant study population characteristics. However, our results suggest that the NAT2 acetylator status has a differential effect on the association of active and passive smoking with breast cancer.

The finding of an increased risk of breast cancer associated with active smoking in slow acetylators but not in fast acetylators is consistent with the hypothesis of a higher exposure to carcinogenic tobacco constituents such as aromatic amines as a result of slow acetylation of these tobacco carcinogens. The modification of the effect of active smoking on breast cancer risk by NAT2 status was evident regardless of whether or not passive smokers were included in the reference group. After taking passive smoking into account, there was a statistically significant trend of increase in risk with increasing pack-years and duration of cigarette smoking, as well as a reduction in risk with time since cessation among women with NAT2 slow acetylation status. These observations support a causal relationship between active cigarette smoking and risk for breast cancer because of retarded detoxification of tobacco carcinogens such as aromatic amines.

This is the first study to report a modifying effect of NAT2 status on the association of smoking and breast cancer risk among premenopausal women. Other studies have included predominantly older women. Significantly elevated breast cancer risks for active compared with never-active smoking in postmenopausal slow acetylators have been reported in two case-control studies previously (17, 20). A nested case-control study in the prospective Nurses’ Health Study also found higher risks associated with active smoking in postmenopausal slow acetylators which were, however, not statistically significant. In an additional case-control study, which included both Caucasian and African-American women, the modifying effect of NAT2 was also only observed among postmenopausal women; however, the risk associated with smoking was greater for rapid rather than slow acetylators (19).

In contrast to active smoking in our study, passive smoking was associated with a 2-fold increased risk for breast cancer with borderline significance among nonactive smoking women who were rapid acetylators but did not affect risk in slow acetylators. Intensity of passive smoke exposure, rather than duration of exposure, affected more strongly the risk for breast cancer, the highest OR of 2.24 being associated with 51+ h/day-years of lifetime exposure among rapid acetylators.

The only other study, which collected detailed information on passive smoking found that the risk of breast cancer was higher in rapid than in slow acetylators both for active and for passive smoking compared with nonactive/passive smokers (20). Thus, the effect of NAT2 status for active smoking was reversed after excluding passive smokers from the reference group. This finding was unexpected and was based on only 165 postmenopausal women (81 cases and 84 controls). Two other studies either did not have sufficient information on passive smoke exposure (19) or did not use an appropriate control group (26) to be considered in this regard.

The stronger effect of passive smoke exposure on breast cancer risk in fast acetylators suggests that the carcinogenic substrate could be heterocyclic amines, which are at least 10 times more present in sidestream than mainstream smoke (27). Analyses of urinary mutagens excreted by active tobacco smokers, implicated a heterocyclic amine as one major DNA-reactive aromatic amino compound (28, 29), and mutagens have also been found in the urine of passive smokers (30). Heterocyclic amines are thought to be N-hydroxylated by CYP1A1 and CYP1B1 in mammary epithelial cells (31), and subsequently O-acetylated by N-acetyltransferases (32) and/or conjugated by sulfotransferases (33), forming electrophilic intermediates and DNA adducts, which initiate cancer. The quantitatively most important heterocyclic amine present in tobacco smoke, 2-amino-α-carboline (34; reviewed in Ref. 35), is activated by both NAT2 and NAT1 (36). Both mRNAs and proteins of NAT1 and NAT2 have been shown to be present in human mammary epithelial cells; however, only NAT1 but not NAT2 N-acetylation activity could be detected in cytosols from whole human mammary tissue (33, 37). Stone et al. (38), on the other hand, report higher levels of heterocyclic amine-DNA adduct formation in primary cultures of human mammary epithelial cells from NAT2 fast acetylators compared with slow acetylators. In the light of the ability of 2-amino-α-carboline, present in tobacco smoke, to be bioactivated by NAT2 and NAT1, this heterocyclic amine may be a significant source of DNA damage in human tissues; more studies in breast tissue of passive smoke-exposed women are warranted (36). Endogenous NAT substrates such as serotonin may also affect carcinogenesis (39).

NAT enzymes probably play a greater role in activation rather than detoxification of heterocyclic amines because these are poorer substrates for N-acetylation than aromatic amines (32). Thus, the association between red meat consumption with increased colon cancer risk in rapid acetylators has been attributed to heterocyclic amines consumption (16, 40). Recently, an interaction between NAT2 status and consumption of well-done meat for risk of breast cancer was reported (41), which was not observed previously (42). Dietary exposure to heterocyclic amines was not considered in Morabia et al. (20) and our analysis. However, we observed a positive dose relationship with lifetime passive smoke exposure among rapid acetylators. Because it is unlikely that the women with greater exposure to passive smoking are also those with higher dietary heterocyclic amines exposure, these results suggest that there is a real effect of passive smoking that may be because of heterocyclic amines such as 2-amino-α-carboline in tobacco smoke.

Nevertheless, we cannot easily explain the equally strong risks associated with active smoking among slow acetylators and with passive smoking among rapid acetylators. More than 90% of the women in our study who actively smoked were also passively exposed through other sources. Clearly the model for lung cancer cannot be relevant for the mammary gland, which is exposed not to direct but to circulated tobacco carcinogens that are then retained in the mammary epithelial cells. It has been hypothesized that the modifying effect of genotype/phenotype can be more evident at low dose, possibly because of saturation of enzyme activity at high dose. The physicochemical compositions of mainstream and sidestream smoke are complex and differ in concentration for many components (27). Therefore, it is reasonable to conceive that at high levels of exposure to mainstream smoke constituents, the relative exposure to different carcinogens weighs toward a greater importance of N-acetylation for the detoxification of metabolites. For exposure solely to sidestream tobacco smoke, DNA adducts may only accumulate among fast acetylators because of the relative higher concentration of substrates, which can be activated by O-acetylation. The understanding of the complete underlying biological mechanism will require that the individual variations in activity level of other enzymes involved in the metabolism of carcinogenic tobacco smoke constituents are also taken into account.

We cannot exclude that selection bias may have contributed to the study results because women were recontacted for a telephone interview and not all of the women had provided blood samples. However, the distribution of all of the relevant variables were neither different between the new and the original study population nor between those with and without DNA samples. Smoking has been associated with worse survival (43). In this case, the attrition because of mortality would tend to make our estimates more conservative. Recall bias of data would have to be differential for active and passive smoking to result in the observed differential effect of NAT2, and this seems rather unlikely.

Misclassification of NAT∗12A rapid alleles as the NAT∗5C slow allele may have occurred because of our genotyping method. Because both alleles are rare in Caucasians with NAT∗12A being especially rare (0.1%; Ref. 25) any effect is likely to be negligible. NAT1 and NAT2 have distinct but overlapping substrate specificity; however, the phenotype-genotype correlation for NAT1 is less clear than for NAT2. Nevertheless, an analysis including both NAT1 and NAT2 genotypes would be important to better understand the modifying effect of NAT gene polymorphisms on breast cancer risk associated with tobacco smoke exposure.

In conclusion, NAT2 genotype alone was not associated with breast cancer risk, whereas risk was increased for passive and current smoking independent of NAT2 genotype. However, NAT2 acetylator status can influence susceptibility to breast cancer after exposure to tobacco smoke carcinogens, and the effect appears to be differential for active and passive smoke exposure. The study demonstrates the crucial importance of separating passive smokers from the nonexposed group in molecular genetic analyses investigating gene-environment interactions. The differential effect of acetylation status on active and passive smoking with respect to breast cancer risk emphasizes the need to consider possibly different mechanisms involved in malignancies associated with exposure to main and sidestream tobacco smoke. Therefore, studies of gene-environment interaction can confirm but also improve the knowledge of biochemical or physiological pathways and stimulate additional mechanistic studies.

We thank the many gynecologists and oncologists of the study regions “Rhein-Neckar-Odenwald” and “Freiburg” who endorsed the original study; the many women who participated in this research; Dr. B. Spiegelhalder for administrative and IT-related support of the laboratory analysis; Ursula Eilber, German Cancer Research Center, Heidelberg, for competent data coordination and management; P. Galmbacher, German Cancer Research Center, Heidelberg, for technical assistance; and Zentrum für Umfragen, Methoden und Analysen (ZUMA) in Mannheim for efficient conduct of the telephone interviews.