Introduction

Lung cancer is the leading cause of cancer death worldwide including Korea (Jee et al. 1998; Marugame and Hirabayashi 2009; Parkin et al. 2005). While cigarette smoking is the main cause of lung cancer, 10% of all patients with lung cancer worldwide are considered never smoker which is defined as less than 100 cigarettes in their lifetime (Subramanian and Govindan 2007). Especially, the proportion of never smokers in Asia including Korea is about 30–40% (Koo and Ho 1990), indicating other environmental factors might be involved in carcinogenesis of never smoker lung cancer.

Never smoker non-small cell lung cancer (NSCLC) has distinct clinical features and outcomes, in which the incidence of activating mutations in epidermal growth factor receptor (EGFR) tyrosine kinase is nearly 10 times higher than smokers, and dramatic response rate to EGFR tyrosine kinase inhibitors (Shigematsu et al. 2005). Recently, echinoderm microtubule-associated protein-like 4-anaplastic lymphoma kinase (EML4-ALK) fusion gene was commonly found in never smokers (Soda et al. 2007). So far, these two molecular pathways account for 40–50% of cases of lung cancer in never smokers.

Despite the striking demographics and high prevalence of never smoker NSCLC, the exact causes still remain undetermined. Until now, several factors, such as second-hand tobacco smoke, radon exposure, environmental pollutants, and cooking oil fumes have been considered as the major causes (Brennan et al. 2004; Darby et al. 2005; Lan et al. 2002). Given that, it can be hypothesized that there might be a genetic susceptibility to develop lung cancer even after trivial exposure to environmental carcinogen or toxin.

Recently, genome-wide association (GWA) studies have demonstrated that three human genomic regions at chromosomes 5p15, 15q25, and 6p21 are associated with susceptibility to lung cancer in European and American populations (Amos et al. 2008; Hung et al. 2008; McKay et al. 2008; Rafnar et al. 2009; Spitz et al. 2008). The 15q25 region encoding nicotinic acetylcholine receptor subunit genes (CHRNA5, CHRNA3, and CHRNB4) was associated with the risk of lung cancer and also considered to be related to nicotine dependence in smokers. Another 6p21 chromosome region was associated with risk of lung cancer. The third region at 5p15 contains two genes; the human telomerase reverse transcriptase gene (TERT) and cleft lip and palate transmembrane 1-like gene (CLPTM1L). The replication study with a small series of never smoker lung cancer has shown statistically significant association between lung cancer risk and 5p15.33 genotypes (Wang et al. 2008). Another GWA study of lung adenocarcinoma in never smoking Asian women also showed that common genetic variants in the TERT-CLPTM1L locus on chromosome 5p15.33 are associated with risk for lung adenocarcinoma (Hsiung et al. 2010). However, the other GWA study of never smoker lung cancer in mostly American population demonstrated that novel genetic variants at 13q31.3 are associated with susceptibility to never smoker lung cancer but the genetic variants at 5p15.33 was not replicated in this study (Li et al. 2010).

Although several studies were conducted to find susceptibility loci for lung cancer in never smokers, no regions were replicated except for 5p15.33, suggesting locus heterogeneity and different environmental toxic effects. In this study, to identify genetic loci associated with susceptibility of lung cancer in never smokers in Korea, we conducted a genome-wide association analysis using the Affymetrix 6.0 single nucleotide polymorphism (SNP) array (Affymetrix, Inc., Santa Clara, CA).

Materials and methods

Study population

For discovery set, a total of 446 patients who were never smokers and diagnosed with histologically confirmed NSCLC at Samsung Medical Center were enrolled in this study. Never smokers were defined as individuals who had smoked less than 100 cigarettes during their lifetime. Four hundred and ninety seven control subjects were recruited from the Korea Association REsource (KARE) project, which has prospectively collected more than 10,000 subjects and has been described in a previous study (Cho et al. 2009). The control subjects can be regarded as healthy individuals because patients with an apparent disease status were excluded from the cohort.

For independent validation set, a total of 434 patients and 1,000 control subjects were collected. Two hundred cases were obtained from Kyungpook National University Hospital (KNUH), 175 from Korea University Medical Center (KUMC), and 59 from Samsung Medical Center (SMC). Written informed consent was obtained from all patients at each of the participating institutions. Research protocol was approved by the institutional review boards at each institution. Individuals who serve as control were obtained from the KARE project.

Genotyping and quality control

An ethylenediamine tetraacetic acid (EDTA) venous blood was collected from the registered patients. Single nucleotide polymorphism (SNP) genotyping was performed using the Affymetrix Genome-Wide Human SNP Array 6.0 which includes more than 906,600 SNPs (Affymetrix, Inc., Santa Clara, CA). Before genotyping, we determined yields of pure double stranded genomic DNA using the Wizard Genomic DNA Purification Kit (Promega, Corp., Madison, WI). Samples were normalized to 50 ng/μl and the normalized genomic DNA (5 μl) from each sample was used as a template for the Affymetrix 6.0 Assays. Genotyping reactions were performed using the Affymetrix Genome-Wide Human SNP Nsp/Sty 5.0/6.0 kit reagents and protocols. Genotypes were called by the Birdseed algorithm of the Affymetrix Genotyping Console version 3.0.2. All procedures were done according to recommended protocols of the manufacturer. After excluding 174,617 SNPs of call rate <99%, 273,926 SNPs of minor allele frequency (MAF) <5%, and 6,532 SNPs showing deviation from Hardy–Weinberg equilibrium (P < 0.001) in controls, 474,503 autosomal SNPs were analyzed. Two patient samples with call rate <95% were excluded. Candidate validation SNPs selected from the discovery analysis were genotyped in the validation sample using the MassARRAY® system (Sequenom, Inc., San Diego, CA).

Statistical analysis

The population structures of our samples were examined to confirm genetic homogeneity and assess stratification using the multidimensional scaling (MDS) method. Affymetrix 6.0 data for East Asian (JPT + CHB), Caucasian (CEU) and African (YRI) populations from the International HapMap Project were used for MDS analysis. Genomic inflation factor (λ) was calculated based on median Chi-squared statistic. An association between each SNP and lung cancer susceptibility was tested using allelic test with the Cochran-Armitage trend test. Among the top 50 SNPs, SNPs that showed erroneous genotype clustering patterns were excluded by visual inspection. For validation, 39 SNPs that passed visual inspection from the top 50 SNPs were selected and five additional SNPs (P < 0.001) were selected in the DAB1 gene region to increase coverage of this region because the peak of the region seemed promising at discovery stage. A total of 44 SNPs were genotyped. All statistical analyses were conducted using the PLINK 1.06 (Purcell et al. 2007), and R 2.9.1 software. Linkage disequilibrium (LD) structure was assessed and SNP tagging was conducted using HaploView version 4.1 (Barrett et al. 2005). We also performed SNP imputation to increase genome-wide coverage for further analyses. IMPUTE program version 1.0.0 was used to impute additional polymorphic SNPs that were not covered by the Affymetrix 6.0 array (Howie et al. 2009). The reference panel used for imputation was composed of 90 known JPT + CHB haplotypes from the International HapMap Project data (Phase II Public Release #22 NCBI Build 36).

Results

Study population description and clinical characteristics

In order to identify common genetic variants associated with never smoker NSCLC, we finally conducted a GWA analysis using a total of 878 lung cancer patients and 1,497 healthy normal control subjects. Due to the high proportion of never smokers in women than in men, the proportion of male patients was low compared to control subjects (48.9% in discovery controls and 58.5% in validation controls) from the KARE cohort (Table 1). We observed no genetic substructure within study populations while there was a partial distinction between the study populations and the International HapMap East Asian (JPT + CHB) populations (Supplementary Fig. 1). The most frequent histologic type in all the cases was adenocarcinoma.

Table 1 Patients characteristics
Full size table

Association in the discovery set

In a discovery stage, we evaluated 474,503 common SNPs (MAF > 5%). The MDS analysis demonstrated that the genetic variations exhibited by our Korean subjects overlap with those from JPT + CHB and are clearly distinct from CEU and YRI according to the International HapMap Project data (Supplementary Fig. 2). The distributions of observed P values for association tests across all SNPs tested showed no evidence of overall systematic bias (λ = 1.042) from the expected P values, and the excess of low P values was consistent with the presence of true associations (quantile–quantile plot; Supplementary Fig. 3). These observations indicate that our samples are genetically homogeneous and the associations will be attributable to genetic difference of lung cancer susceptibility.

In the GWA analysis for 444 cases and 497 controls, we selected top 50 SNPs (Cochran-Armitage trend test; Supplementary Table 1). Manhattan plot of genome-wide association results illustrated clusters of significant association peaks in several regions (Fig. 1). Approximately 20 annotated genes, including possible cancer-related candidate genes, CCND2, DAB1, FRK, HDAC9, and IKBKAP, were located on or nearby these loci. A remarkable association peak was observed nearby the DAB1 gene on 1p32.2, where several consecutive SNPs show low P values.

Fig. 1
figure 1

Manhattan plot of genome-wide association results. P values were calculated using the Cochran-Armitage trend test. The validated region, 18p11.22, in the replication set was presented (see Fig. 2 for details)

Full size image

The association of 18p11.22 was replicated in a different validation set

Among the top 50 SNPs, we excluded 11 SNPs with bad clustering through visual inspection and performed a validation study using the remaining 39 SNPs in another set consisting of Korean lung cancer patients (434 cases and 1,000 controls) recruited from KNUH, KUMC and SMC. While the association of 1p32.2 and most other loci in the discovery set was not replicated in the validation set, two SNPs, rs11080466 and rs11663246, in 18p11.22 showed a significant association in the validation set (P = 2.60 × 10−3 and 6.34 × 10−3, respectively) (Table 2). They were even more significant when analyzed in combination with the discovery set (P = 1.08 × 10−6 and 2.32 × 10−6, respectively). When patients with lung adenocarcinoma were analyzed, the association was still significant (Table 2). The two replicated SNPs were in the same LD block (r 2 = 0.987 and D′ = 1). The regional association plot of the 18p11.22 region in the original discovery set illustrated that the significant peak was located in intron of the FAM38B gene (Fig. 2). The 27 SNPs with P values < 0.05 and their tagging SNPs in the three HapMap populations are listed in Supplementary Table 2. For rs11080466, the allele C was underrepresented in cases when compared to controls with an odds ratio (OR) of 0.68 (95% confidence interval [CI], 0.58–0.79; Table 2) in the combined set. For rs11663246, the minor T allele was underrepresented in cases with an OR of 0.69 (95% CI, 0.59–0.80). The MAF of rs11080466 was consistently low in cases regardless of their origin (Fig. 3).

Table 2 Two SNPs in the 18p11.22 region associated with lung cancer susceptibility
Full size table
Fig. 2
figure 2

Regional association plot of the 18p11.22 region. Manhattan plot of genome-wide association results. P values were calculated using the Cochran-Armitage trend test. Filled circles and gray shaded circles indicate genotyped SNPs and imputed SNPs in the discovery set, respectively. Recombination rates are shown as a gray line and were estimated using the HapMap combined data. Arrows indicate the locations of genes. The lower panel presents an LD map based on r 2 values computed using the HapMap JPT + CHB data

Full size image
Fig. 3
figure 3

Minor allele frequency of rs11080466 for each study data. SMC Samsung Medical Center, KNUH Kyungpook National University Hospital, KUMC Korea University Medical Center

Full size image

Discussion

We found a novel genetic variant locus at 18p11.22 which has not been identified in previous GWA studies of never smoker NSCLC. Recent GWA study of lung adenocarcinoma in never smoking Asian women including 594 cases demonstrated that common genetic variants in the TERT-CLPTM1L locus on chromosome 5p15.33 are associated with risk for lung adenocarcinoma (Hsiung et al. 2010). This finding was independently replicated from East Asia totaling 1,164 lung adenocarcinoma and 1,736 controls, which contains substantially large sample size. The combined replication study confirmed that rs2736100 was associated with risk for lung cancer with P = 5.38 × 10−11 and allelic OR = 1.44. Given that, we tried to confirm the association with our cohort. However, in this study, we only observed P value of 0.008 (top 1% rank, 4,842nd among 474,503 SNPs) for rs2736100 at 5p15.33. It is of note that the results of Korean population did not show significant association for the rs2736100 SNP in the previous study (Hsiung et al. 2010), which is consistent with our result. Although rs2736100 did not reach a genome-wide significance level in our data, allelic frequency pattern of the Korean data (G allele frequency of 0.44 in cases and 0.38 in controls; allelic OR of 1.29 [95% CI, 1.07–1.55]) showed a similar trend with that of Hsiung et al.’s (2010) data (G allele frequency of 0.48 in cases and 0.39 in controls). Therefore, we could not exclude 5p15.33 as a potential risk locus because inconsistent results between the Korean and Chinese populations might be attributed to genetic heterogeneity between the populations, small number of samples in the Korean cohorts and/or different environmental predisposing factors. Moreover, another GWA study of never smoker lung cancer in mostly American population also demonstrated novel genetic variants at 13q31.3 as susceptibility to lung cancer in never smokers which has not been reported yet. However, the genetic variants at 5p15.33 were not replicated in this study (Li et al. 2010). All together, these inconsistent observations might be attributed to the ethnic difference even in East Asian population. It could be suggested that there is a large heterogeneity in genetic background as well as in environmental predisposing factors among different ethnic groups with regard to the susceptibility to NSCLC in never smokers.

Another possible explanation can be the different environmental toxic agents and exposure to second hand smoke. It has long been considered that cooking oil fumes are associated with increased lung cancer risk in Asian never smokers, especially China and Taiwan (Hosgood et al. 2007). However, the relevance of the difference of indoor air pollution via cooking methods amongst different Asian countries with 18p11.22 is not clear, and thus the exact functional relevance of 18p11.22 with tumorigenesis and environmental toxic agents still needs to be further examined.

We identified 50 candidate SNPs. Although this study has limitation due to the lack of information about potential confounding variables such as age, sex, exposure to second-hand smoke, or family history of cancer, two of these candidate SNPs, rs11080466 and rs11663246 were replicated in validation set. These two SNPs are in linkage disequilibrium and located in intron of the FAM38B gene at 18p11.22. Our data showed a relatively weak association of the locus, and this might be attributed to the small number of subjects and modest impact of the responsible gene. Recently, it has been reported that FAM38B (Piezo2) along with FAM38A (Piezo1) are vertebrate multipass transmembrane proteins with homologs in invertebrates, plants, and protozoa and considered as essential components of distinct mechanically activated cation channels (Coste et al. 2010). FAM38B is expressed in various tissues and has potential role in touch and pain sensation, leading to a broad role in mechanotransduction (Coste et al. 2010). The association of FAM38B and cancer has not been explored, but unpublished data suggest that this gene might play a role in the tumorigenesis of lung cancer (Sethi 2008). Nonetheless, the exact role of FAM38B in tumorigenesis still remains largely unknown, and further biological studies would be needed.

In summary, we have identified a novel genetic locus at 18p11.22 region which is associated with susceptibility of never smoker NSCLC in Korean populations. We failed to replicate findings of previous GWA studies which may indicate a large genetic heterogeneity in the genetics of never smoker NSCLC. Further replication studies in larger populations are necessary to clarify our hypothesis.