Cytochrome P4501A2 Phenotype and Bladder Cancer Risk: The Shanghai Bladder Cancer Study

Abstract

Cytochrome P450 1A2 (CYP1A2) is hypothesized to catalyze the activation of arylamines, known human bladder carcinogens present in cigarette smoke. The relationship between CYP1A2 phenotype and bladder cancer risk was examined in a case-control study involving 519 patients and 514 controls in Shanghai, China. Both CYP1A2 and N-acetyltransferase 2 (NAT2) phenotypic status were determined by a caffeine-based urinary assay. The present study showed that among smokers at urine collection, bladder cancer patients had statistically significantly higher CYP1A2 phenotype scores compared with control subjects (P = 0.001). The odds ratios (95% confidence intervals) of bladder cancer for the 2^nd, 3^rd, and 4^th quartiles of the CYP1A2 score were 1.31 (0.53–3.28), 2.04 (0.90–4.60) and 2.82 (1.32–6.05), respectively, relative to the lowest quartile (P for trend = 0.003). NAT2 slow acetylation phenotype was associated with a statistically significant 40% increased risk of bladder cancer, and the relationship was independent of subjects’ smoking status. Subjects possessing the NAT2 slow acetylation phenotype and the highest tertile of CYP1A2 scores showed the highest risk for bladder cancer. Their odds ratios (95% confidence intervals) was 2.13 (1.24–3.68) relative to their counterparts possessing the NAT2 rapid acetylation phenotype and the lowest tertile of CYP1A2 scores. The findings of the present study demonstrate that CYP1A2 phenotype may be an important contributing factor in the development of smoking-related bladder cancer in humans.

INTRODUCTION

Tobacco smoking is an established cause of bladder cancer, accounting for approximately 50% of all cases in the US and other Western countries ¹. Smokers experience 2- to-3 fold excess risk of bladder cancer than nonsmokers ². There are over 60 established carcinogens in cigarette smoke ³. Among them arylamines, such as 4-aminobiphenol (4-ABP) and other aromatic amines, have been identified to be carcinogenic to the urinary bladder ⁴.

Arylamines are procarcinogens and require metabolic activation before they can exert their carcinogenic effects ⁵. Hepatic cytochrome P450 1A2 (CYP1A2) isoenzyme is believed to play an important role in the metabolic activation of arylamines ⁶. The activated arylamines can form adducts with large cellular molecules such as DNA, that eventually lead to the development of bladder cancer ⁵. Humans exhibit considerable inter-individual variability in CYP1A2 activity. Liver tissue preparations from different individuals revealed more than 40-fold difference in the CYP1A2 mRNA ⁷. Caffeine-probing studies have shown a more than 44-fold variation in hepatic CYP1A2 activity ⁸. This inter-individual variation is most likely caused by both environmentally and genetically determined factors. For example, cigarette smoking was associated with significantly increased CYP1A2 activity ⁹. Monozygotic twins had a greater concordance of CYP1A2 phenotype index than dizygotic twins, suggesting the existence of an inheritable component of CYP1A2 ¹⁰. However genotype-phenotype association studies have produced mixed results; some found a positive association ^11–13 while others did not ^{14, 15}. The relationship between CYP1A2 genotype and phenotype still is a subject of active research.

Alternatively, N-acetyltransferases (NATs) detoxify various tobacco toxicants including arylamines ⁵. The human NATs are coded by genes NAT1 and NAT2 ¹⁶. Although both NAT1 and NAT2 catalyze N-acetylation of arylamines, human NAT2 has much higher affinity than NAT1 for urinary bladder carcinogens such as 4-ABP ¹⁷. NAT2 phenotypic status has been extensively studied as a risk factor for bladder cancer. The odds ratio of developing bladder cancer among NAT2 slow acetylators was between 1.4 and 1.8 ¹⁸. Moreover, a likely interaction between cigarettes smoking and the NAT2 genetic variants was reported ¹⁹. Some studies reported an elevated risk of bladder cancer for occupational exposure to arylamines in individuals with NAT2 slow acetylation status^20,21 while others reported no or opposite effect of NAT2 acetylation status on the association between occupational exposure to arylamines and bladder cancer risk^{22, 23}.

Given the requirement of metabolic activation of arylamines to their carcinogenic forms, we and others have hypothesized that smokers showing a higher CYP1A2 activity are at an elevated risk of bladder cancer relative to smokers with comparable smoking history but possessing a lower CYP1A2 activity. Two hospital-based case-control studies with a relatively small sample size (100 or fewer bladder cancer cases) tested this hypothesis ^{24, 25}. One study reported a positive association between CYP1A2 phenotype activity index and bladder cancer risk ²⁴ whereas the other found a null association ²⁵. Previously we examined and reported a null association between CYP1A2 phenotype score and bladder cancer risk in a non-Hispanic white population in Los Angeles, California ²⁶. The present study in a Chinese population in Shanghai, China, was a companion study of the Los Angeles Bladder Cancer Study; the common goal of both studies was to clarify the role of CYP1A2 in the development of bladder cancer. Given the important role of N-acetyltransferase 2 (NAT2) isoenzyme in the detoxification pathway of bladder procarcinogens in tobacco smoke and from other environmental sources, we also assessed the joint effect of CYP1A2 and NAT2 phenotypic status on bladder cancer risk in the present study. Here, we report the presence of a positive association between CYP1A2 phenotype and risk of bladder cancer in Shanghai Chinese.

METHODS

Subjects

Bladder cancer patients were identified through the Shanghai Cancer Registry, a population-based cancer registry covering the approximately 8 million residents of urban area in Shanghai, China, in the 1990’s. Of the patients diagnosed with bladder cancer from 1^st July 1995 through 30^th June 1998, 708 were 25 to 74 years old who met our eligibility criteria for the study. Among the 708 patients, 56 died before we could contact them, 29 refused to be interviewed, and 42 were unable to be located. We interviewed the remaining 581 (82%) eligible patients between July 1996 and June 1999. The diagnosis of bladder cancer for 531 (91%) patients was made based on histopathological evidence whereas the remaining 50 (9%) patients’ diagnoses were based on positive computerized axial tomography scan and/or ultrasonograph with consistent clinical history.

Control subjects were randomly selected from urban residents of Shanghai and chosen to match the frequency distribution by sex and 5-year age groups of bladder cancer patients as ascertained by the Shanghai Cancer Registry during 1990–1994. Personal identification cards issued by the Shanghai Municipal Government were used to select potential control subjects. These cards, one per resident, were housed in 4,410 file cabinet drawers (which were numbered from 1 to 4,410 at the Resident Registry, Bureau of Public Security of Shanghai). We first generated 750 random numbers between 1 and 4470. The number 750 was our anticipated number of incident bladder cancer cases during the three years of case ascertainment period (7/1/95~6/30/98) of the study. We were unable to locate 74 subjects due to their change of home addresses. Seventy-two subjects refused to participate in the study. We interviewed the remaining 604 (80%) subjects during July 1996 and June 1999.

Data collection

A trained interviewer conducted an in-person interview with each study subject using a structured questionnaire that requested information up to 2 years prior to the diagnosis of bladder cancer for bladder cancer case patients and 2 years prior to the date of interview for control subjects (reference date). The questionnaire included background information; demographic characteristics; history of tobacco use; history of passive smoking (for nonsmokers only); consumption of beverages including alcohol, coffee, tea, soft drinks and plain water; use of hormones (for women only); medical history; usual adult diet; and occupational history. Cigarette smoking was defined as smoking one cigarette per day for at least six months.

All study subjects were asked to donate blood and urine samples at the end of the in-person interview. For the collection of an overnight urine sample, each consenting subject was given two packets of Nestle instant coffee or two cans of Coca-Cola Classic drink (about 70 mg of caffeine) to be drunk between 3 and 6 pm. The subject collected an overnight urine sample. When picking up the collected urine, the interviewer asked the subject about use of tobacco products and consumption of alcoholic and caffeine-containing beverages during the past 60 days. Five hundred and thirty-five (92%) of the 581 interviewed case patients and 543 (90%) out of the 604 interviewed control subjects donated an overnight urine sample. The urine samples were processed on the day of collection and acidified (400 mg of ascorbic acid per 20 mg of urine) before they were stored at −80°C until analysis.

Laboratory Measurement

Urinary caffeine metabolites, namely 5-acetylamino-6-amino-3-methyluracil (AAMU), 1-methylxanthin (MX), 1-methyluric acid (MU) and 1,7-dimethylxanthin (17X), were measured by the following methods. Levels of AAMU in urine were determined by a modified procedure of Tang et al. ²⁷, using high-performance size exclusion chromatography. Quantification of MX, MU and 17X in urine was performed according to a modified procedure of Grant et al. ²⁸. These assays were performed with appropriate internal standards.

Calibration curves were created during the analysis and used for calculation of concentrations of all analytes. Quality control urine samples spiked with a low, intermediate and high range of the calibration concentrations were analyzed intermittently during the sample runs. The CYP1A2 and NAT2 phenotype scores were determined based on ratios of urinary caffeine metabolites, i.e. (AAMU + MX + MU)/17X for CYP1A2 ²⁸ and AAMU/(AAMU+MX+MU) for NAT2 ²⁹. Higher ratio values of the CYP1A2 phenotype score reflect higher CYP1A2 activities. Subjects were classified as either slow (< 0.34) or rapid (≥0.34) NAT2 acetylators by the ratio of urinary caffeine metabolites. Of all urine samples, we were unable to detect caffeine metabolites in 48 samples (16 from case patients and 29 from control subjects). Urinary cotinine was measured by a gas chromatographic-mass spectrometric method described previously ³⁰. After excluding subjects with missing values, we included 519 case patients and 514 control subjects in the present analysis.

Cotinine is a major proximate metabolic by-product of nicotine but has a longer metabolic half-life than nicotine. Therefore, cotinine is a better biomarker for daily use of cigarettes than nicotine itself and other nicotine metabolites. In the present study, urinary cotinine was measured by the standard gas chromatographic-mass spectrometric method (GC-MS) ³¹.

Statistical analysis

Given the skewed distribution of CYP1A2 phenotype scores, formal statistical test was performed on logarithmically transformed values, and geometric (as opposed to arithmetic) means were presented. The analysis of covariance (ANCOVA) method was applied to identify determinants of CYP1A2 phenotype in control subjects only. In addition to smoking status, number of cigarettes per day (0, 1-<20, 20+) and urinary levels of cotinine (0, 1-<100, ≥100 ng/ml), age (<50 versus 50+ years), body mass index (BMI) (<18.5, 18.5–24.9, 25+ kg/m²), and daily drinking of green tea (yes/no) were significantly associated with CYP1A2 phenotype score, thus there factors were adjusted for when we examined the CYP1A2-bladder cancer association. Given that cigarette smoking was a strong inducer of CYP1A2, we further assessed the difference in CYP1A2 phenotype score between cases and controls stratified by smoking status at urine collection as well as urinary levels of cotinine.

Unconditional logistic regression models were used to examine the association between CYP1A2 phenotype score, NAT2 phenotypic status or other exposures (e.g., cigarette smoking) and risk of bladder cancer. Study subjects were classified into quartiles or tertiles based on the distributions of CYP1A2 phenotype scores among all control subjects only. The strength of the association between exposure and bladder cancer risk was measured by odds ratios (ORs) and their corresponding 95% confidence intervals (CIs) and P-values. The linear trend test for the association between CYP1A2 phenotype and bladder cancer risk was performed based on the median value within each quartile or tertile of CYP1A2 phenotype score. We assessed the CYP1A2-bladder cancer risk association in total subjects as well as in those stratified by cigarette smoking status at urine collection and over lifetime, urinary cotinine level, and/or NAT2 acetylation status. In all analyses for the CYP1A2-bladder cancer risk association, age at urine collection, gender, BMI, number of cigarettes smoked per day during the past 60 days, daily green tea consumption at urine collection, and urinary cotinine levels were included in the logistic regression models.

We examined the potential interaction effect between CYP1A2 and NAT2 on bladder cancer risk. When examining whether the combined effect of two factors was greater than the multiplicative product of their individual effects, we used the multivariate logistic regression models with the two main effects and their productive terms as covariates. When assessing if the combined effect of the two factors on bladder cancer risk was greater than the sum of the individual effects, we used the method described by Rothman ³².

Statistical analyses were performed using SAS version 9.1 (SAS Institute, Cary, NC) statistical software package. All P values are two sided. P values less than 0.05 were considered statistically significant.

RESULTS

The mean age (±standard deviation) of case patients at diagnosis of bladder cancer was 62.8 (±10.1) years while the mean age of control subjects at interview was 64.2 (±10.0) years (P=0.03). The mean age at urine collection were comparable between case patients (63.4±10.1) and control subjects (64.2±10.0) (P=0.20). The mean (±standard deviation) BMIs of cases and controls were 22.5 (±3.1) and 22.3 (±3.0), respectively (P = 0.30). The percentages of women in cases patients (21%) and control subjects (22%) were comparable (P = 0.65). Bladder cancer patients attained comparable level of education as control subjects (P = 0.34). Among cases, there were 12.7% with college education, 55.9% with middle school education, and 31.4% with primary school or no formal schooling. The corresponding figures among control subjects were 9.5%, 57.6%, and 32.9%. The distributions by consumption of alcohol, tea, and coffee in case patients were comparable with those in control subjects (all P > 0.05) (data not shown). Smoking prevalence was low in Chinese women. Only 18 (16.5%) of 109 female patients with bladder cancer and 6 (5.3%) out of 114 female controls reported having smoked at least one cigarette per day for 6 months or longer over lifetime. In both sexes, cigarette smoking was associated with statistically significantly increased risk of bladder cancer. Compared with never smokers, OR (95% CI) for bladder cancer was 1.62 (1.20–2.19) for ever smokers and 1.77 (1.29–2.43) for current smokers. ORs for bladder cancer increased with increasing numbers of cigarettes per day, years of smoking or pack-years of smoking (all P for trend < 0.01) (Table 1).

Table 1.

History of cigarettes smoking in patients with bladder cancer (cases) and control subjects The Shanghai Bladder Cancer Case-Control Study, 1995–1999

Cigarette smoking	No. of cases (%)(n=519)	No. of controls (%)(n=514)	OR (95%CI)†
Smoking status*
Never Smokers	180 (34.7)	224 (43.6)	1.00 (reference)
Ever Smokers	339 (65.3)	290 (56.4)	1.62 (1.20–2.19)
Former Smokers	79 (15.2)	90 (17.5)	1.27 (0.85–1.89)
Current Smokers	260 (50.1)	200 (38.9)	1.77 (1.29–2.43)
P‡			0.002
Age Starting to Smoke
Never Smokers	180 (34.7)	224 (43.6)	1.00 (reference)
≥27	71 (13.7)	88 (17.1)	1.15 (0.77–1.70)
22-<27	72 (13.9)	70 (13.6)	1.45 (0.95–2.21)
18-<22	141 (27.2)	103 (20.0)	1.91 (1.33–2.75)
<18	55 (10.6)	29 (5.6)	2.71 (1.61–4.56)
Ptrend			<0.001
No. of Cigarettes/day
0 (Never Smokers)	180 (34.7)	224 (43.6)	1.00 (reference)
<20	159 (30.6)	154 (30.0)	1.43 (1.02–2.00)
≥20	180 (34.7)	136 (26.5)	1.91 (1.36–2.69)
Ptrend			0.004
Years of Smoking
0 (Never Smokers)	180 (34.7)	224 (43.6)	1.00 (reference)
<20	42 (8.1)	50 (9.7)	0.91 (0.55–1.50)
20-<40	142 (27.4)	138 (26.9)	1.36 (0.96–1.94)
≥40	155 (29.9)	102 (19.8)	2.45 (1.70–3.55)
Ptrend			<0.001
No. of Pack-Years of Smoking
0 (Never Smokers)	180 (34.7)	224 (43.6)	1.00 (reference)
<20	134 (25.8)	119 (23.2)	1.45 (1.02–2.07)
20-<40	96 (18.5)	99 (19.3)	1.38 (0.94–2.02)
≥40	109 (21.0)	72 (14.0)	2.29 (1.53–3.41)
Ptrend			0.001

^*Smoking status at reference date, i.e., two years prior to the date of cancer diagnosis for cases or two years prior to the date of interview for controls.

^†Odds ratios (ORs) were adjusted for gender and age at reference date; CI=confidence interval.

^‡P value was based on chi-square test for the distribution by smoking status.

Determinants of CYP1A2 Phenotype Score

Cigarette smoking was a strong inducer of CYP1A2. The geometric mean of CYP1A2 phenotype scores increased with increasing number of cigarettes smoked per day and increasing levels of urinary cotinine (both Ps for trend <0.001, Supplemental table). CYP1A2 phenotype scores were comparable between men and women after adjustment for number of cigarettes smoked per day and urinary levels of cotinine. Control subjects under age 50 years had a higher CYP1A2 phenotype score than those of 50 years or older (P <0.001). There was no discernible difference in CYP1A2 phenotype score between age groups above 50 years (P for trend = 0.53). Both BMI and daily green tea drinking were associated with reduced CYP1A2 phenotype scores (Supplemental Table). We did not find any significant relation of CYP1A2 phenotype score with consumption of alcoholic and caffeine-containing beverages (coffee and coke) or cruciferous vegetables (bok choi, cabbage, and cauliflowers) (data not shown).

CYP1A2 and NAT2 in Relation to Risk of Bladder Cancer

Bladder cancer patients had statistically significantly elevated levels of CYP1A2 phenotype scores (Table 2). The most striking difference in CYP1A2 phenotype scores between case patients and control subjects was seen among current smokers. Bladder cancer patients had a 20% higher in CYP1A2 phenotype score than control subjects after adjustment for potential confounders (P = 0.001). Similarly among subjects with detectable levels of urinary cotinine, cases had significantly elevated levels of CYP1A2 phenotype over control subjects (Table 2). There was no statistically significant difference in CYP1A2 phenotype scores between case patients and control subjects who were nonsmokers at urine collection or over lifetime, or had undetectable cotinine in urine.

Table 2.

Geometric means of CYP1A2 phenotype score in patients with bladder cancer (cases) and control subjects, The Shanghai Bladder Cancer Case-Control Study, 1995–1999

	Cases		Controls		% difference†	2-sided P
	No. of subjects	Geometric Means* (95% CI)	No. of subjects	Geometric Means* (95% CI)
Total subjects	519	6.30 (5.87–6.76)	514	5.84 (5.48–6.22)	7.9	0.006
By smoking status
Smokers at urine collection‡	135	6.60 (5.40–8.07)	200	5.51 (4.53–6.71)	19.8	0.001
1-<20 cigarettes/day	98	6.81 (5.57–8.32)	117	5.88 (4.81–7.21)	15.8	0.021
≥20 cigarettes/day	37	6.94 (4.56–10.58)	83	5.68 (3.76–8.56)	22.2	0.038
Nonsmokers at urine collection	384	5.68 (5.32–6.07)	314	5.51 (5.12–5.93)	3.1	0.344
Lifetime nonsmokers	176	5.65 (5.08–6.29)	214	5.38 (4.84–5.97)	5.0	0.261
By urinary cotinine level (ng/ml)§
Undetectable (0)	218	6.14 (5.63–6.70)	187	6.19 (5.68–6.74)	−0.8	0.864
Detectable (≥1)	262	6.91 (6.42–7.44)	303	6.07 (5.69–6.47)	13.8	<0.001
1-<100	148	6.22 (5.66–6.85)	145	5.68 (5.20–6.21)	9.5	0.066
≥100	114	7.57 (6.90–8.30)	158	6.37 (5.86–6.93)	18.8	0.001

^*Lease-squared geometric means were derived from analysis of covariance (ANCOVA) regression models that also included gender, age at urine collection (continuous), body mass index at reference date (<18.5, 18.5-<25, and ≥25 kg/m²), smoking status at reference date (never, former, and current smoker), pack-years of cigarettes smoked (continuous) until reference date, and daily consumption of green tea in the past 60 days prior to the collection of urine sample (yes, no). For total subjects and smokers, the regression models included following additional variables: number of cigarettes smoked per day in the past 60 days (0, 1-<20, and ≥20 cigarettes) and urinary cotinine levels (0, 1-<100, and ≥100 ng/ml, or unknown); for non-smokers, the regression models included urinary cotinine levels; or for subgroups stratified by urinary cotinine level, the regression models included number of cigarettes smoked per day in the past 60 days; CI = confidence interval.

^†Cases’ mean minus controls’ mean divided by controls’ mean, then multiplied by 100.

^‡Smoked cigarettes in the past 60 days prior to urine sample collection.

^§Thirty-nine case patients and 24 control subjects with unknown value of urinary cotinine were excluded.

The association between quartile levels of CYP1A2 and bladder cancer risk was presented in Table 3. Overall, CYP1A2 phenotype scores was associated with risk of bladder cancer (P for trend = 0.01), especially in current smokers. Compared with smokers in the lowest quartile, smokers in the highest quartile of CYP1A2 phenotype score had an OR of 2.82 (95% CI = 1.32–6.05) for bladder cancer, and the association was dose-dependent (P for trend = 0.003). Similarly, a positive association between CYP1A2 phenotype and bladder cancer risk was seen among individuals with detectable levels of urinary cotinine (P for trend = 0.001) (Table 3).

Table 3.

CYP1A2 phenotype score in relation to risk of bladder cancer, The Shanghai Bladder Cancer Case-Control Study, 1995–1999

	CYP1A2 phenotype score in quartiles				Ptrend
	1st (0-<4.29)	2nd(4.29-<5.53)	3rd(5.53-<7.67)	4th(7.67+)
Total Subjects
No. cases/no. controls	110/128	119/129	133/129	157/128
OR (95% CI) *	1.00	0.96 (0.65–1.42)	1.22 (0.83–1.80)	1.60 (1.08–2.39)	0.007
By smoking status
Smokersat urine collection
No. cases/no. controls	14/39	16/35	31/49	74/77
OR (95% CI) *	1.00	1.31 (0.53–3.28)	2.04 (0.90–4.60)	2.82 (1.32–6.05)	0.003
Non-Smokersat urine collection
No. cases/no. controls	96/89	103/94	102/80	83/51
OR (95% CI) *	1.00	0.92 (0.59–1.43)	1.09 (0.69–1.71)	1.31 (0.79–2.16)	0.206
Lifetime nonsmokers
No. cases/no. controls	45/63	47/69	47/47	37/35
OR (95% CI) *	1.00	0.89 (0.51–1.55)	1.39 (0.78–2.48)	1.39 (0.74–2.61)	0.159
By urinary cotinine level (ng/ml) †
Undetectable (0)
No. cases/no. controls	67/53	55/56	50/47	46/31
OR (95% CI) *	1.00	0.74 (0.42–1.30)	0.81 (0.45–1.47)	1.07 (0.56–2.04)	0.790
Detectable (≥1)
No. cases/no. controls	36/69	49/64	76/77	101/93
OR (95% CI) *	1.00	1.33 (0.73–2.44)	1.92 (1.10–3.37)	2.33 (1.33–4.06)	0.001
1-<100
No. cases/no. controls	24/40	39/38	50/40	35/27
OR (95% CI) *	1.00	1.69 (0.79–3.63)	2.31 (1.12–4.81)	2.43 (1.08–5.50)	0.035
≥100
No. cases/no. controls	12/29	10/26	26/37	66/66
OR (95% CI) *	1.00	0.90 (0.31–2.65)	1.69 (0.66–4.32)	2.36 (1.02–5.50)	0.008

^*All odds ratios (ORs) were adjusted for gender, age at urine collection (continuous), body mass index at reference date (<18.5, 18.5-<25, and ≥25 kg/m²), smoking status at reference date (never, former, and current smoker), pack-years of cigarettes smoked (continuous) until reference date, and daily consumption of green tea in the past 60 days prior to the collection of urine samples (yes, no). For total subjects and smokers, the regression models included following additional variables: number of cigarettes per day in the past 60 days (0, 1-<20, and ≥20 cigarettes) and urinary cotinine levels (0, 1-<100, ≥100 ng/ml, or unknown); for non-smokers, the regression models included urinary cotinine levels; or for subgroups stratified by urinary cotinine levels, the regression models included number of cigarettes smoked per day in the past 60 days; CI = confidence interval.

^†Thirty-nine case patients and 24 control subjects with unknown value of urinary cotinine were excluded.

Among lifelong nonsmokers with detectable level of urinary cotinine, above versus below median CYP1A2 phenotype score (5.44) was associated with a statistically significant elevation of bladder cancer risk (OR = 2.30, 95%CI = 1.14–4.64). The difference in the CYP1A2-bladder cancer risk association between subjects with and those without detectable urinary cotinine was statistically significant (P for interaction = 0.04) (data not shown).

NAT2 slow acetylation status was associated with a statistically significantly, 40% increased risk of bladder cancer (OR=1.40, 95%CI = 1.05–1.88). There was no discernable difference in risk of bladder cancer associated with NAT2 slow acetylation status between ever smokers and never smokers (Table 4), or among smokers with different levels of smoking exposure (data not shown).

Table 4.

NAT2 acetylation status in relation to risk of bladder cancer by smoking status The Shanghai Bladder Cancer Case-Control Study, 1995–1999

	NAT2 rapid acetylation (≥ 0.34)	NAT2 slow acetylation (<0.34)	OR (95% CI) *	2-sided P
	No. cases/no. controls	No. cases/no. controls
Total subjects	382/409	137/105	1.40 (1.05–1.88)	0.023
Never smokers	139/183	41/41	1.39 (0.85–2.27)	0.188
Ever smokers	243/226	96/64	1.43 (0.98–2.07)	0.061

^*Odds ratio (OR) was adjusted for age, gender, smoking status (never, former, and current), number of cigarettes per day (continuous), and number of years of smoking (continuous) for total subjects; number of cigarettes per day and number of years of smoking for ever smokers; CI = confidence interval.

We further examined the interaction effect of CYP1A2 and NAT2 phenotypes on bladder cancer risk. The relative risk of bladder cancer increased with increasing CYP1A2 score in both NAT2 rapid and slow acetylation status. This positive CYP1A2-bladder cancer risk association was stronger in subjects with NAT2 slow acetylation than those with NAT2 rapid acetylation (Table 5). On the other hand, NAT2 slow acetylation status was associated with increased risk of bladder cancer in higher tertiles of CYP1A2 score, but not in the lowest tertile of CYP1A2 score. The OR of bladder cancer for the NAT2 slow acetylators with the highest tertile of CYP1A2 was 2.13 (95% CI = 1.24 – 3.68) compared with the NAT2 rapid acetylation and the lowest tertile of CYP1A2 phenotype scores. In total subjects, the interaction between CYP1A2 and NAT2 on bladder cancer was not statistically significant on either the multiplicative (p = 0.43) or the additive scales (Rothman synergy index (S) = 1.97 and 95% CI = 0.11–3.83). We further examined the interaction effect between CYP1A2 and NAT2 on risk of bladder cancer among self-reported smokers during the past 60 days only, or subjects with cotinine levels greater than 75 ng/ml only. All tests for interaction were statistically non-significant, with P values greater than 0.10 (data not shown).

Table 5.

CYP1A2 phenotype score in relation to risk of bladder cancer stratified by NAT2 acetylation status The Shanghai Bladder Cancer Case-Control Study, 1995–1999

CYP1A2 phenotype score in tertiles of control subjects	NAT2 rapid acetylation (≥0.34)		NAT2 slow acetylation (<0.34)
	No. cases/no. controls	OR (95%CI) *	No. cases/no. controls	OR (95%CI) *
1st (0-<4.69)	108/135	1.00 (reference)	24/36	0.75 (0.39–1.42)
2nd (4.69-<6.82)	136/138	1.27 (0.87–1.86)	50/34	1.70 (0.98–2.96)
3rd (6.82+)	138/136	1.51 (1.02–2.25)	63/35	2.13 (1.24–3.68)
Ptrend		0.122		0.005

^*All odds ratios (ORs) were adjusted for gender, age at urine collection (continuous), body mass index at reference date (<18.5, 18.5-<25, and ≥25 kg/m²), smoking status at reference date (never, former, and current smoker), pack-years of cigarettes smoked (continuous) until reference date, daily consumption of green tea in the past 60 days prior to the collection of urine samples (yes, no), number of cigarettes per day in the past 60 days (0, 1-<20, and ≥20 cigarettes), and urinary cotinine levels (0, 1-<100, ≥100 ng/ml, or unknown); CI = confidence interval.

Among 519 bladder cancer cases included in the present study, 475 (91%) cases were diagnosed based on histopathological evidence. The results of the analyses after excluding these 44 patients who had no histopathological information remained similar to those based on the entire data set.

DISCUSSION

The present study demonstrates that the CYP1A2 phenotype is a major determinant of bladder cancer risk in those who were exposed, whether actively or passively, to tobacco smoke. Among individuals with detectable urinary cotinine, the highest quartile of CYP1A2 phenotype scores was associated with more than two-fold increased risk of bladder cancer compared with the lowest quartile. On the other hand, there was no association between CYP1A2 and risk of bladder cancer among subjects with undetectable cotinine level. The findings of the present study implicate that the carcinogenic potential of this metabolic gene may depend upon the presence of its major inducer, cigarette smoking, with upregulation of gene expression being a possible causal mechanism behind the observed gene-cancer association.

Tobacco smoke contains considerable amount of arylamines, which require metabolic activation to be transformed into fully carcinogenic agents ⁴. The first step in this metabolic process is N-oxidation, which is catalyzed by CYP1A2. The N-hydroxylamines can form adducts with hemoglobin adducts and/or circulate freely. In the acidic environment of the bladder lumen, a derivative of the glucuronide conjugates of N-hydroxylamines can covalently bind to urothelial DNA and cause malignant transformation of urothelial cells, and ultimately lead to the formation of bladder cancer ⁵. Individuals with high CYP1A2 activity will generate more activated arylamines than their counterparts with low CYP1A2 activity and, in theory, would have increased risk of bladder cancer. High CYP1A2 activity has been linked to elevated urinary mutagenicity in smokers ^{9, 22} and high 4-aminobiphenol hemoglobin (4-ABP-Hb) adduct levels ³³. The elevated 4-ABP-Hb adduct levels have been found to be associated with significantly increased risk of bladder cancer ³⁴. These findings provided a biological possibility that CYP1A2 could play a direct role in the development of bladder cancer.

The phenotype of CYP1A2, as assessed by the caffeine metabolite ratio, is capable of capturing all aspects of genetic variance across individuals, including recognized inter-individual variability in gene expression. The results of a twin study in a Caucasian population suggested that heritability could account for 73% of the phenotypic variation in CYP1A2 while unique environmental exposures may explain the remaining 27% of the inter-individual variation ¹⁰. However, all single nucleotide polymorphisms (SNPs) identified so far in the CYP1A2 gene are located in either the 5′-flanking regulatory region or the introns of this gene ^{9, 11–15} and could not consistently predict the CYP1A2 phenotypic status. A few studies found decreased CYP1A2 activity in individuals possessing certain genetic polymorphisms of the CYP1A2 gene. For example, a point mutation at −2964 base in the 5′-flanking region of the human CYP1A2 gene was reported to be associated with a statistically significantly reduced CYP1A2 phenotype score in a Japanese population ¹¹. Another study found significantly lower CYP1A2 activity in Chinese with homozygous mutant allele of a point mutation at −3113 base of 5′-flanking region (G>A) of the CYP1A2 gene ¹³. Other studies did not detect such associations ^{15, 35, 36}. These data suggested the complexity of the genetic effect on CYP1A2 activity in humans.

Given the unsuccessful search for DNA sequence variations associated with CYP1A2 phenotype, numerous studies have been carried out to search for environmental factors that could influence the hepatic CYP1A2 expression and activity. Consistent with previous studies ^{10, 37}, the present study demonstrated a statistically significant increase in CYP1A2 phenotypic measures in smokers than in nonsmokers. In addition, our study demonstrated that the CYP1A2 phenotype scores increased with increasing level of cotinine in urine, an objective biomarker of nicotine uptake from tobacco smoke. Daily green tea drinkers showed a statistically significantly lower level of CYP1A2 phenotype scores. Several epidemiological studies have reported an inverse association between tea drinking and bladder cancer risk ^{38, 39}. The findings of the inverse tea-CYP1A2 association provide a possible mechanism for the potential protective effect of green tea consumption on bladder cancer risk. In the present study population we also noted that CYP1A2 phenotype score was associated with age and BMI but not with consumption of cruciferous vegetables as reported by a previous study ⁴⁰. The diverse environmental factors that could influence an individual’s CYP1A2 phenotype score render more challenges to study the role of CYP1A2 in bladder cancer development and greater difficulties in comparing findings across different populations possessing varying profiles of lifestyle/environmental exposures.

A few epidemiologic studies examined, but failed to identify a statistically significant association between the DNA sequence variations associated with the CYP1A2 gene and bladder cancer risk ^{41, 42}. However, a recent, hospital-based case-control study with a relatively small sample size reported that patients with bladder cancer who were heavy smokers were more likely to carry the deletion polymorphism at −2467 locus of the CYP1A2 gene than control subjects who were heavy smokers ⁹. It is interesting to note that this SNP was associated with an elevated CYP1A2 phenotype score in a previous study by the same authors ⁹.

Epidemiological data on the relationship between CYP1A2 phenotype and risk of bladder cancer are scarce. A hospital-based study involving 100 bladder cancer patients and 84 control subjects from a Korean population reported a statistically significantly higher capacity for 3-demethylation of theophylline, a marker of CYP1A2 phenotype, in cases than control subjects ²⁴. Two case-control studies, both conducted in Caucasians in North America, did not find any difference in plasma or urinary indices of CYP1A2 activity between bladder cancer patients and control subjects ^{25, 26}. The one by us showed that the geometric means (95% confidence intervals) of CYP1A2 phenotype scores were 7.60 (6.31–9.16), 7.57 (6.09–9.41), and 8.78 (7.78–9.90), respectively, for current smokers of <10, 10-<20, and 20+ cigarettes per day among non-Hispanic whites (whites) in Los Angeles ²⁶. One should note that CYP1A2 phenotype scores among the Los Angeles subjects were significantly higher than those shown among their Chinese counterparts in the present study (see Supplemental Table). The considerably higher CYP1A2 activity among Los Angeles whites could be the underlying reason, at least partially, for the lack of an observed association between CYP1A2 phenotype score and bladder cancer risk in non-Hispanic whites of Los Angeles. If the relationship between CYP1A2 and bladder cancer risk is curvilinear across the spectrum of CYP1A2 phenotype score, with relative linearity in the low range of CYP1A2 scores but exhibiting increasingly small incremental changes in bladder cancer risk with increasingly higher values of CYP1A2 scores, then a study conducted in a low-risk population such as the Chinese (with relatively low CYP1A2 levels) would possess a much higher statistical power to detect an association between CYP1A2 phenotype score and bladder cancer risk than a similar study in a high-risk population such as the Caucasians (with higher CYP1A2 activity levels). Earlier ⁴³, we conjectured that variability in genetic factors governing the activation/detoxification of tobacco carcinogens might explain, at least partly, the paradoxical observation that Chinese smokers showed only one-third the risk for bladder cancer relative to their Caucasian counterparts with similar lifetime history of tobacco smoking. In fact, a major goal of the present study in Shanghai, China, serving as a comparative study of our Los Angeles Bladder Cancer Study ^{26, 44}, was to put this hypothesis of ours (as articulated above) to the test.

NAT2 slow acetylation was associated with statistically significant 40% increase in risk of bladder cancer in this Chinese population, consistent with previous epidemiology studies ²⁶. Individuals possessing a NAT2 slow acetylation phenotype are less efficient in detoxifying the arylamines and other environmental carcinogens. Given a similar NAT2-bladder cancer risk association in smokers versus nonsmokers, our data suggest that NAT2 phenotype is less likely to be influenced by smoking. Indeed, several SNPs of NAT2 have been identified in humans ¹⁷, and these SNPs have been found to relate to NAT2 acetylation status ¹⁷ as well as risk of bladder cancer ^{21, 45}.

The present study also demonstrates an independent effect of CYP1A2 and NAT2 on bladder cancer risk. High CYP1A2 phenotype score was associated with an increased risk of bladder cancer regardless of NAT2 phenotype status. The seemingly additive effects of CYP1A2 and NAT2 on bladder cancer risk are consistent with the notion that CYP1A2 and NAT2 may exert their influence via a common pathophysiological pathway that results in the development of bladder cancer.

The present study had several strengths including the population-based study design, a relatively large sample size, and the homogeneous nature of the study population (Han Chinese only). In addition to the collection of smoking history at two different time points (reference date and time of urine collection), urinary cotinine was quantified to correctly determine the smoking status (active, passive, or non-smoking) of subjects at the time of urine collection. Another strength of the present study was the determination of NAT2 and CYP1A2 status based on phenotype-based assays, which capture the total impact of both endogenous (i.e., genetic) and exogenous (i.e., environmental) factors on the levels of enzymatic activities of these two metabolic genes.

The present study has several potential limitations. One of the limitations was the retrospective study design, i.e., urine collection among patients with bladder cancer occurred after their cancer diagnosis. Given that a proportion of bladder cancer patients quit smoking following their cancer diagnosis, the present study might underestimate the overall effect of CYP1A2 phenotype on bladder cancer risk. The stratified analysis by smoking status largely mitigated this potential underestimation of risk. Another concern about the retrospective study design was the potential impact of disease status on the CYP1A2 measured. However, our data did not show much influence of pathological stage of bladder cancer on CYP1A2 phenotype scores. Among smokers, the geometric means of CYP1A2 for patients with stage I, II and III bladder cancer were 6.09, 6.86, and 4.74, respectively, after adjusting for potential confounders (P for trend = 0.29). A final concern of the present study is the use of caffeine metabolite ratios to assess CYP1A2 phenotype, given that NAT2, NAT1, and CYP2A6 also are involved in the metabolism of caffeine. In other words, these extraneous genes had the potential to confound the association between CYP1A2 and bladder cancer risk. However, as detailed below, there is no evidence that these extraneous genes have contributed to a potentially spurious association between CYP1A2 and bladder cancer risk noted in the present study. We assessed the NAT2 phenotype via a different set of caffeine metabolites from those used in assessing CYP1A2 phenotype (see Methods). The NAT2 score was found to be independent of smoking status whereas the CYP1A2 score was highly related to smoking status. The NAT2 score shows no association with the CYP1A2 phenotype score (Pearson’s product-moment correlation coefficient= − 0.09). Finally, adjustment for NAT2 acetylator phenotype did not materially change the CYP1A2-bladder cancer risk association. Although a previous report suggested that NAT1 may contribute to the metabolism of classical NAT2 substrates including caffeine in the slow acetylator phenotype ⁴⁶, epidemiological studies so far failed to establish a relationship between NAT1 and bladder cancer risk ⁴⁷. CYP2A6 catalyzes the metabolism of coumarin, a major caffeine metabolite 1,7-dimethylxanthin ⁴⁸. CYP2A6 also is the major catalyst for the metabolism of nicotine, accounting for approximately 90% of the conversion from nicotine to cotinine ⁴⁹. However, the urinary cotinine levels were measured and adjusted for in the statistical analysis for the association between CYP1A2 phenotype score and bladder cancer risk in the present study.

The findings of the present study may shed light on the understanding of the large difference in bladder cancer incidence across different racial/ethnic populations in spite of their comparable levels of tobacco smoking, a recognized cause for bladder cancer. Given the lack of a consistent association between DNA sequence variation in the CYP1A2 gene and phenotype across different studies, future genotype-phenotype studies should focus on tobacco smokers or individuals exposed to other substantial inducers of CYP1A2. Such studies should examine epigenetic markers that reflect an individual’s intrinsic inducibility of CYP1A2. Separately, identification of modifiable factors that lead to reduced CYP1A2 activity may be a productive area of research in developing strategies for bladder cancer prevention.

In summary, the present study demonstrates that CYP1A2 phenotype score was associated with statistically significantly increased risk of bladder cancer in subjects who were exposed to tobacco smoke. Both CYP1A2 and NAT2 phenotypes had independent effects on risk of bladder cancer.