Genetic susceptibility of lung cancer associated with common variants in the 3′ untranslated regions of the adenosine triphosphate-binding cassette B1 (ABCB1) and ABCC1 candidate transporter genes for carcinogen export

Abstract

BACKGROUND:

Tobacco-specific nitrosamine 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NKK) is a well defined carcinogen that can induce lung cancer. Genetic polymorphisms in its disposition pathways could modify the risk of developing lung cancer. The authors of this report previously catalogued the sequence variations of the adenosine triphosphate-binding cassette B1 (ABCB1) and ABCC1 candidate transporter genes for carcinogen export in the Chinese population and screened out common variants with potential function in their 5′ flanking and 3′ untranslated regions. The objective of the current study was to test the hypothesis that these common variants are associated with lung cancer risk.

METHODS:

The genotyping analyses for 6 common regulatory variants (reference single-nucleotide polymorphism 4728709 [rs4728709] and rs2188524 in the 5′ flanking region of ABCB1 and rs3842 in its 3′ untranslated region; rs3743527, rs212090, and rs212091 in the 3′ untranslated region of ABCC1) was conducted in a case-control study of 500 patients with incident lung cancer and 517 cancer-free controls in a Chinese population.

RESULTS:

Compared with the wild adenosine/adenosine (A/A) genotype, the variant rs3842 genotype (adenosine/guanosine [A/G] + G/G) of ABCB1 was associated with a statistically significant increased risk of developing lung cancer (odds ratio [OR]. 1.36; 95% confidence interval [95% CI], 1.06-1.76). Also evident was the association between cancer susceptibility and the variant rs212090 genotype (adenosine/thymidine [A/T] + T/T) of ABCC1 (OR, 1.37; 95% CI, 1.03-1.83). Haplotype-based association analysis also emphasized that 2 common haplotypes carrying the culprit alleles of the 2 single-nucleotide polymorphisms were associated with an increased risk of cancer. In addition, stratification analysis demonstrated a remarkable association of ABCB1 rs3842 with the risk of cancer manifested in women (OR, 2.57; 95% CI, 1.36-4.85), in the histologic type of adenocarcinoma (OR, 1.42; 95% CI, 1.03-1.99), and in individuals aged <60 years (OR, 1.50; 95% CI, 1.05-2.14).

CONCLUSIONS:

The current study demonstrated that common polymorphisms in the 3′ untranslated region of ABCB1 and ABCC1 may contribute to the etiology of lung cancer, providing further support for the hypothesis that genetic components in the metabolism and the disposition of NNK may modify the risk of lung cancer, especially in lung adenocarcinoma among women. Functional studies are warranted to elucidate whether aberrant expression and dysfunction of ABC transporters for carcinogen export may play a role in the development of lung cancer. Cancer 2009. © 2008 American Cancer Society.

Lung cancer is the single most common cause of cancer-related death worldwide, mostly because of tobacco smoking. Of the hundreds of chemicals and carcinogens contained in cigarette smoke, the nicotine-derived tobacco-specific nitrosamine 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NNK) is the most potent carcinogen that reproducibly can induce a high incidence of lung cancer in rodents.1, 2 Studies of tobacco smoke also have confirmed NNK as the most potent ingredient responsible for lung carcinogenesis.3 Much of the research on the role of NNK in lung cancer etiology has focused on the mechanisms of its enzymatic activation and the ability of its metabolites to induce DNA adducts and gene mutations. The biotransformation of NNK from its procarcinogen form into a final, effective carcinogen is mediated by phase I enzymes, such as cytochrome P450, family 2, subfamily A, polypeptide 13 (CYP2A13), the most active cytochrome P450 for the metabolic activation of NNK and other nitrosamines in tobacco smoke that is expressed predominantly in human respiratory tract, including peripheral lung.4 Moreover, NNK also has nongenotoxic effects on pulmonary cancer cells by functioning as an agonist for β-adrenergic receptors and stimulates arachidonic acid release, leading to the formation of mitogenic metabolites that stimulate DNA synthesis and cell proliferation and, thus, therefore is believed to cause lung cancer.5, 6 Conversely, as protection from the carcinogen attack, the carcinogenic metabolites, such as 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol (NNAL), are glucuronidated to the nontoxic NNAL-O-glucuronide by the phase II enzyme of uridine diphospate-glucuronosyltransferase.7 And the conjugated metabolites, such as NNAL-Gluc, finally are transported and eliminated by the phase III adenosine-5′-tripohsphate-binding cassette transporters, such as adenosine triphosphate-binding cassette B1 (ABCC1).8, 9

It has been well documented that difference in genetic susceptibility to lung cancer among individuals is caused in part by their varied capacity for metabolism and the disposition of xenobiotics, such as the tobacco-specific carcinogen NNK. For example, we and other groups have documented that the interaction of CYP2A13 and NNK is associated with lung cancer.4, 10, 11 In addition, we previously analyzed the possible association between lung cancer and variants of the β-2 adrenergic receptor gene ADRB2, which is the key mediator of the NNK nongenotoxic pathway.12 It also has been reported that polymorphisms in the phase II conjugation enzyme genes, such as uridine diphosphate glucuronosyltransferase 1 family, polypeptide A7 (UGT1A7) and UGT2B17, have implications in lung cancer etiology.13, 14 However, few studies have addressed the relation between genetic predisposition to lung cancer and polymorphisms in phase III transporter genes.

ABCB1, also known as multidrug resistance 1 (MDR1) or P-glycoprotein (P-gp), and ABCC1, also known as multidrug resistance protein 1 (MRP1), are expressed at relatively high levels in both normal15 and cancerous lung tissue.16, 17 It has been demonstrated clearly that ABCC1 mediates the transport and elimination of metabolites of the tobacco-specific carcinogen NNK.8 The transport substrates of ABCB1 include environmental carcinogens that are present in cigarette smoke or diet, such as benzo(a)-pyrene,18 7,12-dimethylbenzanthracene,19 or 2-amino-1-methyl-6-phenylimidazo [4,5-b] pyridine.20 The 2 ABC transporters are supposed to be the most important transporters in normal lung physiology protecting the lungs against inhaled toxicants. Their dysfunction, which may result in part from polymorphisms in their coding genes, could have implications for the development of lung cancer.21

The identification of sequence variants of ABCB1 and functional characterization of their biologic significance and clinical relevance with respect to drug disposition has stimulated interest in the pharmacogenetics of xenobiotic phase III transporters.22 However, the relevance of the results from many previous studies that focused primarily on common single-nucleotide polymorphisms (SNPs) in coding exons remains to be corroborated, because there have been some conflicting results from many population-based association studies.23 With regard to the ABCC1 gene, mutations in its coding region largely are infrequent in the general population, and their phenotypic implications in drug response and in common diseases are limited.24 In our previous environmental genomic studies, we systemically cataloged SNPs of the ABCB1 and ABCC1 genes and comprehensively characterized their haplotype structures in the Chinese population.25, 26 The extent and strength of linkage disequilibrium (LD) in the intragenic region of ABCB1 are remarkably strong in the Chinese population, especially throughout its exons and its 3′ flanking region.26 For the ABCC1 gene, however, extensive intragenic recombination and weak allelic association throughout this locus are manifest in the Chinese population.25 In those previous studies, we characterized an array of common SNPs, such as reference SNP 3743527 (rs3743527), rs212090, and rs212091 located at 543 base pairs (bp), 866 bp, and 1513 bp, respectively, downstream from the stop codon of ABCC1, which may serve as tag SNPs that are enabled to capture a large fraction of diversity information for the 5′ or 3′ flanking regions of the 2 ABC transporter genes. We also screened out a few common regulatory variants of potential functional significance. Bioinformatic analysis and computer prediction indicated that the common 5′ flanking SNPs of ABCB1, rs4728709 and rs2188524, located at 3502 bp and 235 bp, respectively, upstream from its major transcription initiation site,27 and the common rs3842, located at its 3′ untranslated region (UTR) 202 bp downstream from the stop codon, are embodied in or close to putative cis-regulatory elements targeted by potential trans-acting factors (transcription factor and microRNA [miRNA]). However, the implications and roles in the etiology of lung cancer for these common regulatory variants of the ABC transporter genes for carcinogen export remain largely unknown.

In the current case-control investigation, we also demonstrate that common variants in the 3′ UTR of ABCB1 and ABCC1 may be correlated with an increased genetic susceptibility to lung cancer.

MATERIALS AND METHODS

Study Patients

This study included 500 patients with lung cancer and 517 cancer-free controls. All participants were genetically unrelated ethnic Han Chinese and were from Nanjing City and surrounding regions in southeast China. Patients with histopathologically confirmed incident lung cancer were recruited between July 2003 and August 2004 at the Cancer Hospital of Jiangsu Province (Nanjing, China) and the First Affiliated Hospital of Nanjing Medical University, (Nanjing, China) and had a response rate of 90.5% (500 of 552 patients). Cancer-free controls were selected randomly from 10,500 individuals who participated in a community-based screening program for noninfectious diseases that was conducted in Jiangsu Province during the same period when the cases were recruited and had a response rate of 83.8%. These control individuals had no history of cancer and were frequency-matched to the cases on age (±5 years), sex, and residential area (urban or rural). Each participant was scheduled for an interview after written informed consent was obtained, and a structured questionnaire was administered by interviewers to collect information on demographic data and environmental exposure history, including tobacco smoking. A family history of cancer was defined as any self-reported common cancers in first-degree relatives (parents, siblings, or children), such as liver cancer, colorectal cancer, esophageal cancer, breast cancer, and so on. After the interview, an approximately 5-mL venous blood sample was collected from each participant. The study was approved by the institutional review boards of Nanjing Medical University.

Genotype Assay

We genotyped the 6 common regulatory SNPs (rs4728709, rs2188524. and rs3842 of ABCB1 and rs3743527, rs212090. and rs212091 of ABCC1) according to the Illumina GoldenGate Assay protocol at Chinese National Human Genome Center at Shanghai as part of a large-scale project that was studying the candidate gene-based genetic epidemiology of lung cancer. Details regarding the sequences of the allele-specific oligos and the locus-specific oligos for the 6 SNPs were provided upon request. The assay products were hybridized to high-density, bead-based microarrays and imaged on the Sherlock scanner (Illumina, Inc.). The GenCall software (Illumina, Inc.) clustering and calling algorithm was run, and the genotyping results were produced. Quality control was performed as described previously in association with studies in lung cancer.28

Statistical Analysis

Deviations in genotype frequencies in the control and case populations compared with the frequencies expected under Hardy-Weinberg equilibrium were assessed by using the standard chi-square test. The chi-square test also was used to examine differences in demographic variables, smoking, and distribution of genotypes between cases and controls. We estimated haplotypes based on the observed genotypes by using the Bayesian statistical method implemented in PHASE.29 LDA software was used to calculate linkage disequilibrium (LD) index r².30 The risk of lung cancer associated with genotype and haplotype were estimated by calculating the odds ratios (ORs) and accompanying 95% confidence intervals (95% CIs) with unconditional logistic regression models.31, 32 Age, sex, pack-years smoked, and familial history of cancer were included for multivariate adjustment. Information was collected on the number of cigarettes smoked per day, the age at which individuals started smoking, and the age at which the smokers stopped smoking. Individuals were defined as nonsmokers if they never smoked or if they smoked <100 cigarettes before the date of cancer diagnosis (for cases) or the date of interview (for controls) Individuals were defined as smokers if the smoked >100 cigarettes in their lifetime. The light and heavy smokers were categorized by using the 50-percentile pack-year ([cigarettes per day/20] × [years smoked]) values of the control population as the cutoff points (ie, ≤30 pack-years and >30 pack-years). We tested the null hypothesis of additive and multiplicative gene-gene interaction between the polymorphisms in the 3′ UTR of ABCB1 and ABCC1 and evaluated departures from additive and multiplicative interaction models by including main effect variables and their product terms in the logistic regression model. The homogeneity test was used to compare the difference between smoking-related ORs among different genotypes or between the product of related ORs and the joint effect OR. The probability level of <.05 was used as the criterion for statistical significance, and all statistical tests were 2-sided. All statistical analyses were performed by using Stata software (version 8.0).

RESULTS

This study involved 500 patients with lung cancer (cases) and 517 individuals in a cancer-free control population. The distributions according to age, sex, smoking status, and family history of cancer in first-degree relatives among all individuals are summarized in Table 1. There were no significant differences between cases and controls in terms of the distribution according to sex and age, suggesting that the frequency matching was adequate. However, compared with the control population, the cases were more likely to be smokers (ever-smokers: 67% in cases vs 51.8% in controls; P < .001). In addition, 122 cases (24.4%) and 87 controls (16.8%) reported a family history, the difference between the 2 populations was associated with a 1.60-fold (95% CI, 1.17-2.17) increased risk of lung caner. Of the 500 lung cancer cases, 229 adenocarcinomas (45.8%), 141 squamous cell carcinomas, and 130 other subtypes (26%) were classified histopathologically.

Table 1. Distributions of Select Characteristics by Case-Control Status

	No. of Individuals (%)
Variable	Cases, N=500	Controls, N=517	P
SD indicates standard deviation. * The chi-square test was used to determine the difference between cases and controls. † The Student t test was used to determine the difference between cases and controls. ‡ The chi-square test was used to determine the distribution of never smoking and ever smoking between cases and controls. § The chi-square test was used to determine the distribution of never smokers, ≤30 pack-years, and >30 pack-years between cases and controls. ∥ The Student t test was used to determine the difference between cases and controls.
Sex
Men	386 (77.2)	400 (77.4)	.949*
Women	114 (22.8)	117 (22.6)
Mean±SD age, y	59.3±10.4	60±10	.328†
Smoking status
Never smokers	184 (36.8)	267 (51.6)	<.001‡
Ever smokers	316 (63.2)	250 (48.4)
≤30 pack-years	139 (27.8)	153 (29.6)	<.001§
>30 pack-years	177 (35.4)	97 (18.8)
Mean±SD pack-years	38.8±23	30.4±16.3	<.001∥
Family history of cancer, no. of first-degree relatives
≥1	122 (24.4)	87 (16.8)	.003
None	378 (75.6)	430 (83.2)
Histologic type
Adenocarcinoma	229 (45.8)
Squamous cell carcinoma	141 (28.2)
Others	130 (26)

The genotypic frequencies of the 6 common regulatory polymorphisms of ABCB1 and ABCC1 for cases and controls are shown in Table 2. Relatively high frequencies (range, 9.2%-46.2%) were observed for their minor alleles in our study population. For example, the allelic frequencies of the guanosine (G) allele of rs3842 (ABCB1) and the thymidine (T) allele of rs212090 (ABCC1) were 30.8% and 26.6% in cases, respectively, and 15.4% and 11.9% in controls, respectively. The genotype distributions of these common regulatory variants did not deviate from those expected from the Hardy-Weinberg equilibrium in either controls or cases. Differences in genotype distribution between cases and controls with marginal statistical significance were observed for rs3842 of ABCB1 (chi-square test, 5.19; degrees of freedom [df], 2; P = .075) and rs212090 of ABCC1 (chi-square test, 6.29; df, 2; P = .043). With regard to rs3842 of ABCB1, we observed genotype frequencies of 54.6% (adenosine/adenosine [A/A]), 37.7% (A/G), and 7.7% (G/G) in the control population and 47.4%, 43.6%, and 9%, respectively, in the case population. The frequencies of the 3 genotypes (A/A, A/T, and T/T) of rs212090 (ABCC1) were 77.3%, 21.5% and 1.2% in controls, respectively, and 72.2%, 24.8% and 3% in cases, respectively. Because rs3842 and rs212090 homozygotes (G/G and T/T, respectively) were relatively infrequent in our study population, these genotypes were combined in a dominant genetic model with heterozygous genotype (rs3842, respectively A/G or the rs212090 A/T) for the subsequent estimation of cancer risk.

Table 2. Genotype Distribution of Adenosine Triphosphate-binding Cassette B1 (ABCB1) and ABCC1 Transporter Gene Polymorphisms Between Cases and Controls and Their Association With the Risk of Lung Cancer

	No. of Individuals (%)
Genotype	Cases, n=500	Controls, n=517	Adjusted OR [95% CI]*	P†
OR indicates odds ratio; 95% CI, 95% confidence interval; ABCB1, the adenosine triphosphate-binding cassette B1 transporter gene; ABCC1, the adenosine triphosphate-binding cassette C1 gene; rs, reference single-nucleotide; C, cytidine; T, thymidine; A, adenosine; G, guanosine. * ORs and 95% CIs were calculated by logistic regression with adjustment for age, sex, pack-years, and familial history of cancer within the strata. † The chi-square test (with 2 degrees of freedom) was used to determine the difference in genotype distribution between cases and controls. ‡ Reference group.
ABCB1
rs4728709				.953
C/C	342 (68.4)	354 (68.5)	1.00‡
C/T	145 (29)	148 (28.6)	1.02 [0.77-1.35]
T/T	13 (2.6)	15 (2.9)	0.97 [0.66-1.43]
rs2188524				.562
A/A	40 (80)	427 (82.6)	1.00‡
A/G	95 (19)	85 (16.4)	1.25 [0.90-1.74]
G/G	5 (1)	5 (1)	1.11 [0.59-2.09]
rs3842				.075
A/A	237 (47.4)	282 (54.6)	1.00‡
A/G	218 (43.6)	195 (37.7)	1.35 [1.03-1.76]
G/G	45 (9)	40 (7.7)	1.20 [0.95-1.52]
A/G+G/G	263 (52.6)	235 (45.4)	1.36 [1.06-1.76]
ABCC1
rs3743527				.707
G/G	160 (32)	153 (29.6)	1.00‡
G/A	233 (46.6)	250 (48.4)	0.86 [0.64-1.15]
A/A	107 (21.4)	114 (22)	0.94 [0.79-1.12]
rs212090				.043
A/A	361 (72.2)	400 (77.3)	1.00‡
A/T	124 (24.8)	111 (21.5)	1.28 [0.94-1.72]
T/T	15 (3)	6 (1.2)	1.73 [1.07-2.80]
A/T+T/T	139 (27.8)	117 (22.7)	1.37 [1.03-1.83]
rs212091				.174
A/A	290 (58)	270 (52.2)	1.00‡
A/G	177 (35.4)	206 (39.8)	0.79 [0.61-1.03]
G/G	33 (6.6)	41 (8)	0.86 [0.67-1.10]

We also analyzed LD of the common regulatory SNPs of ABCB1 and ABCC1 and predicted their phased haplotypes. In the control population, relatively weak LD (average pairwise D′ = 0.47) and allelic association (average pairwise r² = 0.11) for the 2 5′-flanking SNPs (rs4728709 and rs2188524) and the 3′-UTR SNP (rs3842) of ABCB1 were observed, and moderate LD (all pairwise D′ = 1.00; average pairwise r² = 0.17) for the 3 3′-UTR SNP (rs3743527, rs212090 and rs212091) of ABCC1 was observed. Similar LD patterns were observed for the 2 ABC transporter genes in the case population. We separately estimated haplotype frequencies in the case and control populations and identified all 8 of the theoretically expected haplotypes (3 common haplotypes with frequency >5%) phasing on the 3 SNPs of ABCB1 in both cases and controls. However, only 4 haplotypes were observed in the study population, all of which were common with a frequency >10% (Table 3). The physical distance spans, approximately 100 kb from rs4728709 (5′ flank) to rs3842 (3′ UTR) of ABCB1 and approximately 1 kb from rs3743527 to rs212091 (3′ UTR) of ABCC1, could account in part for their different recombination mode, LD pattern, and haplotype diversity.

Table 3. Haplotype Frequencies of the Adenosine Triphosphate-binding Cassette B1 (ABCB1) and ABCC1 Transporter Gene Polymorphisms Between Cases and Controls and Their Association With the Risk of Lung Cancer

Haplotype*	Cases, 2n=1000	Controls, 2n=1034	Adjusted OR (95% CI)†	P‡
OR indicates odds ratio; 95% CI, 95% confidence interval; Ref, reference group. * Segregating sites in 5′ to 3′ order, as listed in Table 2. † ORs and 95% CIs were calculated by logistic regression with adjustment for age, sex, pack-years, and familial history of cancer within the strata. ‡ P values were determined with the chi-square test (with 3 degrees of freedom) to determine the difference in the distribution of haplotypes between cases and controls. § Others consist of a haplotype frequency <5%.
ABCB1
CAA	601	678	1.00 (Ref)
CAG	188	149	1.45 (1.14-1.86)	.028
TAG	79	71	1.19 (0.85-1.70)
Others§	132	136	1.08 (0.82-1.41)
ABCC1
AAA	447	478	1.00 (Ref)
GAG	243	288	0.90 (0.73-1.10)	.043
GAA	156	145	1.16 (0.89-1.50)
GTA	154	123	1.37 (1.05-1.80)

By using logistic regression analysis, first, we assessed the association between predisposition to lung cancer and the genotypes of the 6 regulatory SNPs of the ABC transporter genes (Table 2). We observed a significant increase in the risk of lung cancer among individuals who carried the ABCB1 rs3842 A/G or G/G genotype (OR, 1.36; 95% CI, 1.06-1.76; P = .016) compared with individuals who carried the wild A/A genotype. In similar manner, an increased risk of lung cancer also was be associated with the ABCC1 rs212090 A/T or T/T genotype (OR, 1.37; 95% CI, 1.03-1.83; P = .033) compared with the reference genotype of A/A. We also estimated haplotype-specific ORs by using a haplotype-based logistic regression method. For statistical advantage, we combined all haplotypes of ABCB1 that had an allele frequency >5%. Table 3 also summarizes the association between the haplotypes and the risk of lung cancer. The cytidine-adenosine-guanosine (CAG) haplotype of ABCB1 (OR, 1.45; 95% CI, 1.14-1.86; P = .003) and the GTA haplotype of ABCC1 (OR, 1.37; 95% CI, 1.05-1.80; P = .019) were associated with an increased risk of lung cancer compared with the reference of their most prevailing haplotypes. It is noteworthy that a comparison of the sequence constitution of the 2 culprit haplotypes with their reference sequences obviously revealed that their divergence could result mainly from rs3842 and rs212090, further indicating that the 2 common SNPs in the 3′ UTR of the ABC transporter genes may be associated with a genetic susceptibility to lung cancer.

Therefore, we further assessed the risk of cancer related to the 3′-UTR SNPs of rs3842 and rs212090 by stratifying individuals by sex, age, smoking status, family history of cancer in first-degree relatives, and pathologic characteristics (Table 4). Remarkably, a pronounced association of ABCB1 rs3842 A/G or G/G genotypes the with risk of lung cancer was observed in women (OR, 2.57; 95% CI, 1.36-4.85; P = .003) for the adenocarcinoma histologic type (OR, 1.42; 95% CI, 1.03-1.99; P = .034), in individuals aged <60 years (OR, 1.50; 95% CI, 1.05-2.14; P = .025), and in individuals with a family history of cancer in first-degree relatives (OR, 1.91; 95% CI, 1.07-3.41; P = .028). These observations suggested that the common variants in the 3′ UTR of ABCB1 and ABCC1 may have been associated with a genetic susceptibility to lung cancer in some subgroups of the study population.

Table 4. The Risk of Lung Cancer Related to the Adenosine Triphosphate-binding Cassette B1 (ABCB1) and ABCC1 Transporter Gene Polymorphisms by Sex, Age, Smoking Status, Family History of Cancer, and Histologic Type

	ABCB1 (rs3842)				ABCC1 (rs212090)
	No. of Cases/Controls				No. of Cases/Controls
Variable	A/A	A/G+G/G	OR (95% CI)*	P	A/A	A/T+T/T	OR (95% CI)*	P
rs indicates reference single-nucleotide polymorphism; A, adenosine; G, guanosine; T, thymidine; OR, odds ratio; 95% CI, 95% confidence interval. * ORs and 95% CIs were calculated by logistic regression with the A/A genotypes of rs3842 and rs212090 as the reference group and adjusted for age, sex, pack-years, and familial history of cancer within the strata.
Sex
Men	197/221	189/179	1.19 (0.89-1.58)	.247	284/313	102/87	1.36 (0.97-1.91)	.075
Women	40/61	74/56	2.57 (1.36-4.85)	.003	77/87	37/30	1.47 (0.77-2.81)	.245
Age, y
<60	111/162	129/127	1.50 (1.05-2.14)	.025	173/222	67/67	1.23 (0.82-1.84)	.319
≥60	126/120	134/108	1.34 (0.91-1.96)	.134	188/178	72/50	1.60 (1.03-2.50)	.036
Smoking status
Nonsmokers	83/139	101/128	1.30 (0.88-1.94)	.185	124/202	60/65	1.53 (0.99-2.35)	.058
Smokers	154/143	162/107	1.39 (0.98-1.97)	.063	237/198	79/52	1.20 (0.79-1.82)	.388
≤30 pack-y	63/85	76/68	1.50 (0.92-2.44)	.101	105/117	34/36	0.88 (0.50-1.56)	.663
>30 pack-y	91/58	86/39	1.31 (0.78-2.18)	.305	132/81	45/16	1.77 (0.93-3.39)	.084
Family history of cancer, no. of first-degree relatives
≥1	55/52	67/35	1.91 (1.07-3.41)	.028	85/69	37/18	1.76 (0.90-3.43)	.097
None	182/230	196/200	1.28 (0.97-1.70)	.085	276/331	102/99	1.28 (0.92-1.77)	.138
Histologic type
Adenocarcinoma	93/282	112/235	1.42 (1.03-1.99)	.034	155/400	50/117	1.11 (0.75-1.63)	.607
Squamous cell carcinoma	65/282	76/235	1.49 (1.00-2.21)	.051	105/400	36/117	1.28 (0.81-2.02)	.290
Others	79/282	75/235	1.22 (0.84-1.76)	.298	101/400	53/117	1.83 (1.22-2.73)	.003

Because the 2 ABC transporters are coexpressed in lung tissue and, in a synergetic mode, may export pulmonary environmental carcinogens such as NNK in tobacco smoking, next, we examined whether there was a statistical joint gene-gene interaction in the etiology of lung cancer (Table 5). We observed that patients with lung cancer who carried the ABCB1 rs3842 A/G or G/G genotypes also were more likely to carry the ABCC1 rs212090 A/T or T/T genotypes than controls (14.2% vs 10.3%; P = .067; Pearson chi-square test with Yates continuity correction). Compared with the homozygous wild-type reference group, for both ABCB1 rs3842 and ABCC1 rs212090, the presence of 1 risk genotype (ABCB1 rs3842 A/G or G/G; or ABCC1 rs212090 A/T or T/T) was associated with a moderate increase in the risk of developing lung cancer (OR, 1.41; 95% CI, 1.05-1.89 or OR, 1.44; 95% CI, 0.96-2.16, respectively). The OR increased to 1.83 (95% CI, 1.20-2.80) among individuals who carried both the ABCB1 rs3842 G risk allele and the ABCC1 rs212090 T risk allele. However, no discrepancy with statistical significance was observed between the OR for individuals who had both risk alleles and individuals with only a risk genotype (P > .05 for both the homogeneity test and the interaction test). These results did not suggest a statistical interaction between the 6 common variants in the 3′ UTR of ABCB1 and ABCC1 in intensifying the risk of lung cancer in the study population.

Table 5. The Risk of Lung Cancer Related to the Joint Effects of Adenosine Triphosphate-binding Cassette B1 (ABCB1) and ABCC1 Transporter Gene Polymorphisms

	ABCC1 (rs212090)
	A/A			A/T+T/T
ABCB1 (rs3842)	No. of Cases/Controls	OR (95% CI)*	P	No. of Cases/Controls	OR (95% CI)*	P
rs indicates reference single-nucleotide polymorphism; A, adenosine; T, thymidine; Ref, reference; OR, odds ratio; 95% CI, 95% confidence interval; G, guanosine. * ORs and 95% CIs were calculated by logistic regression with adjustment for age, sex, pack-years, and familial history of cancer within the strata. † Test for interaction: P > .05.
A/A	169/218	1.00 (Ref)		68/64	1.44 (0.96-2.16)	.076
A/G+G/G	192/182	1.41 (1.05-1.89)	.022	71/53	1.83 (1.20-2.80)†	.005

DISCUSSION

In the current study, we examined whether genetic polymorphisms in ABCB1 and ABCC1 could have an impact on the risk of developing lung cancer. On the basis of our analysis of 500 patients with lung cancer and a control population of 517 healthy individuals in a Chinese population, we observed that common variants in the 3′ UTRs of major transporter genes for carcinogen export influenced the risk of developing lung cancer. Patients who had the variant allele of either ABCB1 rs3842 or ABCC1 rs212090 were at an increased risk of developing lung cancer. These findings provide further insight into the complex role of the tobacco-specific carcinogen NNK in the etiology of lung cancer and extend the carcinogen disposition pathways from phase I and phase II metabolizing enzymes to include the phase III transporters, in which genetic polymorphisms may modify the predisposition to lung cancer.

ABC transporters are recognized increasingly for their ability to modulate the absorption, distribution, metabolism, excretion, and toxicity of xenobiotics. In addition to their originally revealed role in multidrug resistance in chemotherapeutics, there is cumulating evidence that these efflux pumps are capable of transporting a vast and chemically diverse array of toxicants, including dietary and environmental carcinogens, and, thus, are supposed to play an important role in preventing the accumulation of potentially harmful xenobiotics in sensitive tissues such as the lung.9 ABCB1 lies on the apical surface of bronchial and bronchiolar epithelium and functions to remove toxicants to the lung lumen.15 ABCC1 is located with relatively high expression on the basolateral surface of the epithelia layer, and it can transport toxins into the lung interstitial space.15 This case-control association study, together with the biochemical observation that the metabolites of tobacco-specific NNK are transported and eliminated by ABC transporters such as ABCC1, highlights the possibility that they also may influence individual susceptibility to lung cancer.

The common variants of ABCB1 rs3842 and ABCC1 rs212090 that are associated with the risk of lung cancer are located at the 3′ UTR, and they are not in linkage disequilibrium with other potentially functional SNPs in the Chinese population. as we reported previously.25, 26 By using computational alignment and miRNA target prediction (available at: http://www.targetscan.org accessed March 1, 2008), we demonstrated that rs3842 lies at a conserved region of the 3′ UTR of ABCB1, very close to the potential targeting stem for has-miR-374, which also is expressed in lung tissue (according to the GeneNote microarray data; available at: http://www.genecards.org accessed March 1, 2008). These observations indicate that the noncoding polymorphism may have phenotypic implications for the regulation of gene expression, which may be mediated by miRNA. There is cumulating evidence that sequence variations in the 3′ UTR of human genes may be able to alter the content of putative miRNA target sites, and miRNA-target interactions with allelic differences may provide a novel source of phenotypic variation.33 For example, the differential interaction between has-miR-155 and its polymorphic targets in the 3′ UTRs of the angiotensin II type 1 receptor gene (AGTR1) may play a role in the development of hypertension.34, 35 An SNP near the has-miR-24 binding site in the 3′ UTR of the dihydrofolate reductase gene (DHFR) interferes with has-miR-24 function and is associated with drug resistance. It also has been also reported that SNPs in miRNA-binding sites affect miRNA target expression and function and, thus, potentially are associated with cancer.36 Although it remains to be determined whether the common variant of rs3842 in the 3′ UTR of ABCB1 could affect gene regulation, the current case-control study indicates that this SNP may be associated with lung cancer through a possible miRNA-mediated mechanism.

It is noteworthy that, in our stratification analysis, a substantial increase in the risk of lung cancer among women (OR, 2.57; 95% CI, 1.36-4.85) had a remarkable association with the variant genotypes of ABCB1 rs3842. This observation may have biological plausibility and epidemiological implications. In 3 large case-control studies, it was argued that women are more susceptible to the carcinogens in cigarette smoke than men.37-39 Although this originally observed potential difference between the sexes in the risk of lung cancer because of smoking was not confirmed in a follow-up cohort study,40 the epidemiological observation that women who smoke and use estrogen-replacement therapy are at significantly increased risk of developing lung cancer suggested that hormone factors such as estrogen status may play a role in the development of lung cancer.41 Some reproductive factors have been associated with the risk of lung cancer in Chinese women.42 Estrogens may act directly as carcinogens through metabolism to catechol estrogens and the formation of DNA adducts; alternatively, they may act as tumor promoters through a receptor-mediated mechanism. The α and β estrogen receptors (ERs) are expressed abundantly in human nontumor and tumor lung tissue and produce biologic responses to estrogen; and ER-α expression occurs more often in the lungs of women compared with men.43, 44 One of the possible mechanisms for ER-mediated carcinogenesis is its modulation of the expression and activity of carcinogen metabolism and its mechanism of disposition, including P-450 enzymes such as CYP1A1 and CYP1B1.45, 46 It also was reported that the interaction of estrogens and ER-α can down-regulate the expression and activity of ABC transporters such as ABCG2 and ABCB1 in breast cancer cells, which may induce a reduction in the cellular resistance to anticancer agents and xenobiotics.47, 48 Partly consistent with these observations, women have lower hepatic expression and activity of ABCG2 compared with men.49 In the current study, the association between rs3842 at the 3′ UTR of ABCB1 with the risk of lung cancer in women suggests that estrogen and ER may affect the expression of ABCB1 in lung tissue by interacting with rs3842 or other linked, functional SNPs, indicating that their interplay may play a role in the etiology of lung cancer in women.

The observations that ABCB1 rs3842 also modified the risk of lung cancer for individuals with the adenocarcinoma histologic type (OR, 1.42; 95% CI, 1.03-1.99) and for individuals aged <60 years (OR, 1.50; 95% CI, 1.05-2.14) also were interesting. NNK and polycyclic aromatic hydrocarbons (PAHs), the 2 most important carcinogens among the hundreds of chemicals contained in cigarette smoke, may be etiologic elements correlated with histopathologically different lung cancer. Lung adenocarcinoma most likely is linked to NNK, whereas lung squamous cell carcinoma may be induced by PAHs such as benzo(a)pyrine.50-53 It has been established clearly that PAHs are activated primarily by the polymorphic carcinogen-metabolizing enzyme of CYP1A1, which appears to influence most the development of lung squamous cell carcinoma.54-56 Conversely, the epidemiological data on smoking and lung cancer continue to indicate increasing rates of lung adenocarcinoma, especially in women, over the past decades, compared with squamous cell carcinoma.57, 58 The carcinogen NNK, which increased in cigarette smoke between 1978 and 1992 by approximately 45%, is supposed to be 1 factor that is responsible in part for the increase in incidence of lung adenocarcinoma.51 It also was reported that lung adenocarcinoma is over-represented in middle-aged patients compared with other types of lung cancers, which typically are diagnosed in patients aged ≥60 years.59 In our previous molecular epidemiological investigations on lung cancer, we suggested that genetic polymorphisms in CYP2A13, the most important phase I enzyme for the metabolic activation of NNK, and in ADRB2, the key receptor and mediator in the nongenotoxic effect of NNK, are associated with the risk of developing lung adenocarcinoma.11, 12 In the current study, we also demonstrated that a common variant of ABCB1, 1 of the phase III transporter genes for excretion of the NNK metabolite, may modify the risk of lung adenocarcinoma. Collectively, these observations consistently support the hypothesis that genetic components in the metabolism of NKK and the mechanism of disposition of NNK (phase I, activation; phase II, conjugation; and phase III, excretion), as well as in its nongenotoxic pathway, are modifiers of the risk of lung cancer, and especially lung adenocarcinoma. Furthermore, because tobacco smoking is the leading preventable cause of cancer and the cancer-prone genotypes of these genetic components are relatively prevalent in human population, our findings also have important implications for the prevention tobacco smoking-related cancers.

The current study may have some limitations, because it was a hospital-based, case-control study with a relatively small sample size. However, we used incident cases, and the procedures for the ascertainment of cases and controls were well defined. The finding that genotype frequencies among controls and cases fit the Hardy-Weinberg law further supports the randomness of our patient selection process. Nevertheless, it is at least possible that ascertainment bias and confounding factors may account for the association of the polymorphisms of ABC transporter genes with the risk of lung cancer. We acknowledge that the relatively small sample size did not enable us to address possible gene-gene and gene-environment interactions in the etiology of lung cancer. Furthermore, with a primary focus on common variants in the 5′-flanking region and 3′ UTR of the ABC transporter gene, we did not cover the extensively studied common SNPs in the coding exons of ABCB1, such as C3435T (rs1045642 in exon 26) and G2677T/A (rs2032582 in exon 21). In a case-control study that examined the possible effects of ABCB1 polymorphisms on tobacco-related lung cancer, no clear association was observed between the C3435T genotype and cancer risk in a group of 268 Caucasian men who were current smokers.60 It recently was reported that G2677T/A of ABCB1 is associated with a high risk of lung cancer; however, observations from the original study should be interpreted cautiously, because the study was statistically underpowered (96 cases vs 86 controls), and it included only men (182 Spanish white men).61 Although their phenotypic and clinical relevance in lung cancer remains speculative, these common coding SNPs of ABCB1 should be examined, and more comprehensive haplotype-based association studies in different populations with larger sample size are warranted to validate the current preliminary report.

In conclusion, the current study demonstrated that rs3842 and rs212090, the common polymorphisms in the 3′ UTR of ABCB1 and ABCC1, are associated with an increased risk of developing lung cancer. These results suggest that further functional studies are warranted to elucidate whether aberrant expression and dysfunction of ABC transporters for carcinogen export may play a role in the development of lung cancer.

Conflict of Interest Disclosures

Supported in part by grants from the National Basic Research Program of China (2002CB512902, 2002BA711A10, and 2004CB518605), the Shanghai Science and Technology Developing Program (06DZ19502), the Program for New Century Excellent Talents in University (NCET-07-0204), and the Shanghai Rising-Star Program (07QA14006).