Analyzing large-scale samples confirms the association between the rs1051730 polymorphism and lung cancer susceptibility

The early genome-wide association studies (GWAS) found a significant association between lung cancer and rs1051730 (15q25) polymorphism. However, the subsequent studies reported consistent and inconsistent results in different populations. Three meta-analysis studies were thus performed to reevaluate the association. But their results remain inconsistent. After that, some new GWAS studies reported conflicting results again. We think that the divergence of these results may be due to small-scale samples or heterogeneity among different populations. Therefore, we reevaluated the association by collecting more samples (N = 33,617 cases and 116,639 controls) from 31 studies, which incorporate 8 new studies and 23 previous studies used by one or more of the three meta-analysis studies. We observed a significant association between lung cancer and rs1051730 in pooled population by using allele (OR = 1.30, 95% CI = 1.27–1.34, P  <  0.0001), dominant (OR = 1.41, 95% CI = 1.29–1.55, P < 0.0001), recessive (OR = 1.53, 95% CI = 1.42–1.65, P < 0.0001) and additive (OR = 1.75, 95% CI = 1.61–1.90, P < 0.0001) models. Through the subgroup analysis, we observed a significant heterogeneity only in East Asian population (P = 0.006, I2 = 66.9%), and the association is significant in all subgroups (OR = 1.2976, 95% CI = 1.2622–1.3339 (European ancestry), OR = 1.5025, 95% CI = 1.2465–1.8110 (African), OR = 1.7818, 95% CI = 1.3915–2.2815 (East Asian), P < 0.0001). We believe that these results will contribute to understanding the genetic mechanism of lung cancer.

. The flow chart of collecting articles for analyzing the association between rs1051730 polymorphism and lung cancer. The criteria of selecting articles are: (1) the study was designed according to the method of case-control. (2) the study evaluated the association of rs1051730 polymorphism and lung cancer. (3) the number of cases and controls was provided in the study. (4) the study provided the population of each individual. (5) the study provided the number of rs1051730 genotypes both in cases and controls or provided enough data to calculate the genotypes. (6) the study provided the OR value with 95% CI and the P value or provided enough data to calculate them. OR = 1.31, 95% CI is from 1.19 to 1.44, and P = 1.5E-08. Shiraishi et al. analyzed 1,250 cases and 936 controls from the Japan, and found the OR = 2.3, 95% CI is from 1.5 to 3.7, and P = 0.0028.
Facing the inconsistent results above, Gu et al. 21 , Hu et al. 22 and Zhan et al. 23 collected different samples to re-assess the association between rs1051730 and lung cancer by meta-analysis, respectively.  Table 1. Main information of these studies used to analyze the association between rs1051730polymorphism and lung cancer. "AA/GA/GG" means the study have offered the data of genotypes AA/GA/GG both in cases and controls. "AA/GA and GG" means only the data of genotypes AA/ GA and GG both in cases and controls have been offered in the study. "GA/GG" means only the data of genotypes GA/GG both in cases and controls have been offered in the study. "A/G" means only the data of genotypes A/G both in cases and controls have been offered in the study. All the alleles have been used A or G to express in our study to replace T or C in original articles. IARC: International Agency for Research on Cancer; CGQS: Copenhagen General Population Study; GELCC: Caucasians of Genetic Epidemiology of Lung Cancer Consortium; MCC: Mayo Clinic Caucasians; NCI: National Cancer Institute.
However, the results of the three meta-analysis studies are also inconsistent. Gu et al. 21 and Hu et al. 22 found that the association between rs1051730 and lung cancer is not significant, while it is significant according to the study of Zhan et al. 23 .
After these studies, five articles [24][25][26][27][28] , which include eight studies, investigated the association between rs1051730 and lung cancer. But the results of eight studies remain inconsistent. For example, He et al. 27 used 301 cases and 318 controls from China, Hansen et al. 28 used 448 cases and 611 controls from America, and found a negligible association between rs1051730 and lung cancer, while other researches [24][25][26] reported opposite consequences.
We considered that the divergence of these results may be due to small-scale samples or heterogeneity among different populations. Here, we collected all the samples of the three meta-analysis studies and the eight studies from the five new articles, and thus obtained a larger sample sizes (33,617 cases and 116,639 controls) from 31 studies to reevaluate the association between rs1051730 polymorphism and lung cancer based on the method that was frequently used by Liu et al. to study Alzheimer's Disease [29][30][31][32] and colorectal cancer 33 .

Results
Literature and Study acquisition as well as Data extraction. By searching PubMed with keywords (details shown in the method section), we obtained 20 articles, which include 16 articles used in the three previous meta-analysis researches and 4 new articles. Moreover, we obtained another article through the reference search in Google Scholar. Finally, we got 31 studies from the 21 articles according to the six inclusion criteria (details shown in the method section). The workflow was showed in Fig. 1   After that, we collected the relative data according to 11 terms for each study, and 7 of the 11 terms were listed in Table 1.
Heterogeneity Test. According to the kind of genotype shown in Table 1 Subgroup Analysis. Because the allele model included the maximum number of studies, we further used it to perform the subgroup analysis. We found that there is no significant heterogeneity in European ancestry (P = 0.2296 and I 2 = 18.4%) and African (P = 0.1586 and I 2 = 42.2%) populations, while in East Asian population the heterogeneity is significant (P = 0.006 and I 2 = 66.9%). So we further split East Asian population into Japanese and Chinese subgroups. The heterogeneity was not found in the Japanese population (P = 0.8387 and I 2 = 0), but it remains significant in the Chinese population (P = 0.0753 and I 2 = 56.5%). Then, we removed each study from Chinese population orderly, and found that there was no significant heterogeneity after the study of Wu et al. had been removed. After meta-analysis and Z test, we found the association between the rs1051730 and lung cancer is relatively strong in all populations. The results were described in Table 2. Forest plots of each subgroup meta-analysis were shown in Figure  S1-S7.  (Fig. 6) reflected the result directly. And then, the result of sensitivity analysis showed that the association between rs1051730 and lung cancer doesn't extremely change when we removed each of the studies in the four models orderly. The detail information was in supplement materials (Table S5-S8).

Discussion
Nicotine receptor protein abnormal expression is one of the reasons for lung cancer occurrence 3 . CHRNA3, a gene coding a part of nicotinic acetylcholine receptor protein subunits, includes a SNP rs1051730. The elder GWAS studies showed that a significant association between the rs1051730 polymorphism and Consequently, the association between rs1051730 polymorphism and lung cancer is significant in all of these populations. In addition, the result of sensitivity analysis reflects the conclusion is robust, and the publication bias isn't significant. Before submitting this paper, we used keyword "rs1051730", "lung cancer" and "meta" to search in PubMed, and obtained six articles [21][22][23][34][35][36] , which include the researches of Gu et al. 21 , Hu et al. 22 and Zhan et al. 23 . Among these three articles, Gu et al. 21 integrated 16 studies to assess the risk of rs1051730 in East Asian, European, and African populations by using allelic and dominant models. Hu et al. 22 also collected 16 studies in the same populations, but merely the allelic model was used. They obtained a similar result that the risk is high in European and African populations, but it is weak in East Asian. Zhan et al. 23 only assessed the East Asian, they combined 4 studies and the result shows a significant association between the rs1051730 and lung cancer. The other three articles didn't research the association between the rs1051730 polymorphism and lung cancer. They evaluated the association between rs1051730 and habit of smoking [34][35][36] , or the association between rs1051730 and cotinine levels 36 by meta-analysis.
Our work is different from the others. We collected all 31 studies, which include 23 studies used by the three previous meta-analysis researches and 8 new studies. We analyzed the association between rs1051730 and lung cancer in European ancestry, African and East Asian populations. The association is significant in European ancestry and African populations, which is consistent with the result of Gu et al. 21 and Hu et al. 22 . However, we found that the association remains significant in East Asian population, which is not consistent with the result of Gu et al. 21 and Hu et al. 22 .
The rs1051730 polymorphism is in CHRNA3 on 15q25. CHRNA3 is one of members in a multigene family of nicotinic acetylcholine receptor (nAChR gene cluster) which can code various nicotinic acetylcholine receptor protein subunits include: α 3, α 4, α 7, α 9, α 10, β 2 and β 4 nAChR subunits 37 . These subunits are expressed on the bronchial epithelial cells of human being and primates 38 . Through combining with the nicotinic acetylcholine receptor protein, nicotine promotes tumor cell proliferation, invasion, migration and induces blood vessel formation. At the same time, it provides a protection for the  Table 1, 13 studies that provided the data of genotype AA/GA/GG were used in the additive model (AA vs GG).
tumor cell to avoid the programmed cell death 39 . In addition, Arredondo et al. 40 found that α 3, α 4, α 7, α 9, α 10, β 2 and β 4 nAChR subunits can form the high-affinity sites to bind 4-(methylnitro-samino)-1-(3-pyridyl)-1-butanone (NNK), a cancerogen produced through nicotine nitrosylation, thus to increase the risk of lung cancer. Moreover, CHRNA3 also contains other two SNP: rs578776 and rs938682 polymorphism. The rs17486278, rs11637635 as well as rs7178270 polymorphism belong to CHRNA5 and CHRNA4, respectively. They may be also associated with the susceptibility of non-small cell lung cancer was reported in a study 41 . In addition, another gene AGPHD1 also in 15q25 contains the rs8034191 polymorphism, and many studies reported its association with lung cancer 5,8,9,20,24 . We expect that more research on them could be performed in the future.

Methods
Literature and Study acquisition. We collected all the articles which were used to perform meta-analysis by Gu et al. 21 , Hu et al. 22 and Zhan et al. 23 . And then, we searched all the possible articles in PubMed (http://www.ncbi.nlm.nih.gov/pubmed) with keywords: "rs1051730" and "lung cancer", or "rs1051730" and "Carcinoma", or "rs1051730" and "tumor", or "CHRNA3" and "lung cancer", or "CHRNA3" and "Carcinoma", or "CHRNA3" and "tumor". All of these literatures had been collected before the PubMed's last update on April 7 2015. In addition, we selected the related references in these articles both from PubMed and the three meta-analysis researches by using Google Scholar (http:// scholar.google.com/). All the selected articles were written in English.
After that, we selected the studies in all obtained articles according to the following criteria: (1) The study was designed according to the method of case-control. (2) The study evaluated the association of rs1051730 polymorphism and lung cancer. (3) The number of cases and controls was provided in the study. (4) The study provided the population of each individual. (5) The study provided the number of rs1051730 genotypes both in cases and controls or provided enough data to calculate the genotypes. (6) The study provided the OR value with 95% CI and the P value or provided enough data to calculate these. Data extraction. We extracted the following information from each study we have selected: (1) The first author of each article. (2) The publication year of each article. (3) The population and ethnicity of individual in each study. (4) The number of cases and controls in each study. (5) The number of rs1051730 genotypes both in cases and controls. (6) The OR value with 95% CI and the P value in each study. (7) The genotyping platform. If the information didn't be provided directly, we used program R (http://www.r-project.org/) to work them out.