Pleiotropic Analysis of Lung Cancer and Blood Triglycerides

Abstract

Epidemiologically related traits may share genetic risk factors, and pleiotropic analysis could identify individual loci associated with these traits. Because of their shared epidemiological associations, we conducted pleiotropic analysis of genome-wide association studies of lung cancer (12 160 lung cancer case patients and 16 838 control subjects) and cardiovascular disease risk factors (blood lipids from 188 577 subjects, type 2 diabetes from 148 821 subjects, body mass index from 123 865 subjects, and smoking phenotypes from 74 053 subjects). We found that 6p22.1 (rs6904596, ZNF184) was associated with both lung cancer (P = 5.50x10^-6) and blood triglycerides (P = 1.39x10^-5). We replicated the association in 6097 lung cancer case patients and 204 657 control subjects (P = 2.40 × 10^-4) and in 71 113 subjects with triglycerides data (P = .01). rs6904596 reached genome-wide significance in lung cancer meta-analysis (odds ratio = 1.15, 95% confidence interval = 1.10 to 1.21, P_combined = 5.20x10^-9). The large sample size provided by the lipid GWAS data and the shared genetic risk factors between the two traits contributed to the uncovering of a hitherto unidentified genetic locus for lung cancer.

Genetic heritability of lung cancer is estimated to be 14% (1), but only a few genetic risk loci have been identified to date in genome-wide association studies (GWAS) of lung cancer in Europeans (2). Epidemiological studies have shown associations between lung cancer and cardiovascular disease (CVD) risk factors related to the metabolic syndrome (3,4). There is also substantial evidence that lipid metabolism and innate immunity evolved from common pathways, and consequently genes that influence lipid traits may also influence inflammation and subsequent cancer development (5–7). Lung cancer is also well known to be strongly associated with tobacco smoking. Predicated on the hypothesis that investigating shared genetic risk factors across these traits could enhance the possibility of identifying new genetic loci for lung cancer, we used quantile-quantile (Q-Q) plots (8) (Supplementary Methods, available online) to assess potential polygenic enrichment of single-nucleotide polymorphisms (SNPs) associated with lung cancer given association with each CVD risk factor or smoking phenotypes (Figure 1; Supplementary Figure 1, available online).

Figure 1.

Conditional Q-Q plots: lung cancer (LuCa) | cardiovascular disease (CVD) factors (triglycerides [TG] and low-density lipoproteins-cholesterol [LDL-C]). ‘Conditional Q-Q plot’ of theoretical vs empirical -log₁₀ P values (corrected for genomic control $\lambda$) in LuCa below the standard GWAS threshold of -log10 P values equal to 7.3 (equals P values above 5 x 10^-8) as a function of statistical significance of association with (A) TG and (B) LDL-C at the levels of P < 1.00, P < .10, P < .01, P < .001, P < 1.00x10^-4, P < 1.00x10^-5, respectively. Dotted lines indicate the theoretical line in case of no association. All statistical tests were two-sided.

The analysis was based on the TRICL consortium meta-analysis of lung cancer GWAS, including 12 160 lung cancer case patients and 16 838 control subjects (2) (Supplementary Table 1, available online); the meta-analysis data of blood lipids from the Global Lipids Genetics Consortium (GLGC; including genetic association with triglycerides [TG], and high and low density lipoproteins-cholesterol [HDL-C and LDL-C]) from 188 577 subjects (9), of type 2 diabetes (T2D) from 148 821 subjects (10), and of body mass index (BMI) from 123 865 subjects (11); and the meta-analysis of cigarettes per day (CPD) and never vs ever smoking (SMOKER) data from the Tobacco, Alcohol and Genetics (TAG) consortium, including 74 053 subjects (Supplementary Table 2, available online). The Supplementary Materials (available online) contain additional details on the contributing studies, statistical analyses, and functional tests. All statistical tests were two-sided unless otherwise specified. P values of less than .05 were considered statistically significant.

The Q-Q plots show enrichment between lung cancer and LDL-C and between lung cancer and TG blood lipid traits across multiple P value thresholds, up to 10^-5 (Figure 1, A and B), verified by an adaptive permutation procedure (Supplementary Table 3, available online). In contrast, we observed no statistically significant enrichment (Bonferroni-corrected significance threshold P < .0036) between lung cancer and HDL-C, BMI, T2D, or smoking phenotypes (the analysis of smoking excluded the SNP markers mapping to chr15:78,686,690-79,231,478, which are known to be associated with lung cancer and smoking (Supplementary Table 3 and Supplementary Figure 1, available online) (12–13). Thus, we excluded these traits from further analysis.

Cross-phenotype-associated loci between lung cancer and TG and between lung cancer and LDL-C were assessed by conjunction false discovery rate (FDR) (Supplementary Materials, available online) (8). Because controlling FDR is heavily affected by the number of identified SNPs, we pruned SNPs in linkage disequilibrium (LD; r² > 0.8) and excluded the major histocompatibility complex (MHC; genomic position (hg 19): chr6:29,528,318-33,373,649 [14]), which harbors established lung cancer susceptibility SNPs and is known for long-range LD. By controlling conjunction FDR, we identified one genetic locus at 6p22.1, rs6904596 (A>G, minor allele frequency in caucasians = 0.094), associated with both lung cancer and blood triglycerides (conjunction FDR = 1.24E-02; P = 5.50x10^-6 for lung cancer; P = 1.39x10^-5 for TG) (this locus and additional genetic loci shared between lung cancer and lipid traits are shown in Supplementary Table 4 and Supplementary Figures 2-6, available online). This locus remained statistically significant as defined by a conjunction FDR of less than .05, also using different thresholds for pruning SNPs in LD (Supplementary Table 5, available online).

We tested this SNP for replication in 6097 lung cancer case patients and 204 657 control subjects from deCODE, Harvard, Holland, and Spain (Supplementary Table 6, available online). This locus was replicated (P_replication = 2.40 × 10^-4) and attained genome-wide significance for lung cancer risk in the meta-analysis of discovery and replication data (odds ratio = 1.15, 95% confidence interval = 1.10 to 1.21, two-sided P_combined = 5.20x10^-9, P_{heterogeneity} = .91) (Table 1). This SNP was also replicated in the association with TG in 71 113 independent samples from deCODE and Holland (two-sided P = .01, P_combined = 1.34 x 10^-6) (Table 1). The SNP association with lung cancer was mostly driven by the squamous cell carcinoma subtype (P = 2.80x10^-5) and not adenocarcinoma (P = 0.06) (Supplementary Table 7, available online).

Table 1.

Association of rs6904596 at 6p22.1 with both lung cancer risk and blood triglyceride levels

Study	Lung cancer cases/ controls	OR (95% CI)	P*	Study	Triglyceride individuals	$\mathit{\beta }$ (95% CI)	P*
TRICL	12160/16838	1.15 (1.08 to 1.21)	5.50 x 10-6	GLGC	188577	0.0244 (0.013 to 0.035)	1.39 x 10-5
Replication	6097/204657	1.16 (1.07 to 1.25)	2.40 x 10-4	Replication	71113	0.0290 (0.006 to 0.051)	1.14 x 10-2
deCODE	3865/196658	1.17 (1.06 to 1.30)	3.05 x 10-3	deCODE	66027	0.0200 (-0.004 to 0.044)	.10
Harvard	984/ 970	1.18 (0.93 to 1.50)	.17	Holland	5086	0.0891 (0.027 to 0.151)	5.12 x 10-3
Holland	687/ 5158	1.15 (0.96 to 1.37)	.12
Spain	561/ 1871	1.10 (0.88 to 1.37)	.40	Combined	259690	0.0253 (0.015 to 0.035)	1.34 x 10-6
Combined	18257/ 221495	1.15 (1.10 to 1.21)	5.20 x 10-9
PQ-heterogeneity†			.91				.23

*P values were derived from a two-sided Wald test. The reference group for the odds ratio in the lung cancer study was healthy controls without lung cancer. CI = confidence interval; OR = odds ratio.

†Heterogeneity of effect size across studies was evaluated using the Cochran’s Q statistic. The test is defined as one-sided.

rs6904596 localizes to 6p22.1 (27,491,299 bp; hg19) and lies 50 kb 5' of zinc finger protein 184 (ZNF184). It shows expression-QTL in lung tissue (15) with HLA-DRB3 (β = -6.79, P = 1.10x10^-11). Additionally, rs7749305, located on chr6:27,446,566 (r² = 1 with rs6904596 in HapMap 3 of Caucasian populations), shows suggestive regulatory functions. This SNP showed the strongest association with lung cancer but was not genotyped in the Global Lipids Genetics Consortium GWAS. It lies within a DNaseI hypersensitive region in small airway epithelial cells (SAEC) and A549 adenocarcinoma cells (ENCODE) and lies in a region hypomethylated in primary alveolar epithelial cells (AEC) from our laboratory (Supplementary Figure 7, available online). rs7749305 alternate allele C appears to create ATF3 and HIF1A binding sites. Similar findings are evident in adipocytes (ENCODE), extending the pleiotropic association between lung cancer and lipid traits to their function in respective tissue types.

Additional functional analyses on relevant tissue types would be important to study how the genetic risk variants assert their downstream effect on the traits of interest.

Our study emphasizes that pleiotropic analysis of GWAS data of epidemiologically related traits can uncover hitherto unidentified genetic associations. Moreover, some GWAS of quantitative traits may be much larger than disease-specific GWAS (like in the case of CVD risk factors vs lung cancer) and, thus, may improve the likelihood to identify new loci for the disease with the smaller sample size.

Funding

This work was supported by the Intramural Research Program of the Division of Cancer Epidemiology and Genetics, National Cancer Institute, National Institutes of Health, DHHS, Bethesda, MD.

Transdisciplinary Research for Cancer of Lung (TRICL): National Institute of Health U19 CA148127-01 (PI: Amos), Canadian Cancer Society Research Institute (No. 020214, PI: Hung). The Environment and Genetics in Lung Cancer Etiology (EAGLE), Prostate, Lung, Colon, Ovary Screening Trial (PLCO), and Alpha-Tocopherol, Beta-Carotene Cancer Prevention (ATBC) studies were supported by the Intramural Research Program of the National Institutes of Health, National Cancer Institute (NCI), Division of Cancer Epidemiology and Genetics. ATBC was also supported by US Public Health Service contracts (N01-CN-45165, N01-RC-45035, and N01-RC-37004) from the NCI.

PLCO was also supported by individual contracts from the NCI to the University of Colorado Denver (NO1-CN-25514), Georgetown University (NO1-CN-25522), the Pacific Health Research Institute (NO1-CN-25515), the Henry Ford Health System (NO1-CN-25512), the University of Minnesota (NO1-CN-25513), Washington University (NO1-CN-25516), the University of Pittsburgh (NO1-CN-25511), the University of Utah (NO1-CN-25524), the Marshfield Clinic Research Foundation (NO1-CN-25518), the University of Alabama at Birmingham (NO1-CN-75022), Westat, Inc. (NO1-CN-25476), and the University of California, Los Angeles (NO1-CN-25404). The Cancer Prevention Study-II (CPS-II) Nutrition Cohort was supported by the American Cancer Society. Funding for the Lung Cancer and Smoking study was provided by the National Institutes of Health (NIH), Genes, Environment and Health Initiative (GEI) Z01 CP 010200, NIH U01 HG004446, and NIH GEI U01 HG 004438. For the lung study, the GENEVA Coordinating Center provided assistance with genotype cleaning and general study coordination, and the Johns Hopkins University Center for Inherited Disease Research conducted genotyping. Harvard Lung Study: The Harvard Lung Cancer Study is supported by the National Institutes of Health (R01 CA092824, P50 CA090578, and R01 CA074386).

Global Lipids Genetics Consortium (GLGC): Data on the lipid traits were provided by the Global Lipids Consortium through their access portal (http://csg.sph.umich.edu/abecasis/public/lipids2013/). The full Consortium acknowledgements are available in the supplementary information of Willer et al.

Tobacco, Alcohol and Genetics (TAG) consortium: Data on the smoking traits were provided by the Tobacco, Alcohol and Genetics through their access portal (http://www.broadinstitute.org/mpg/ricopili/). The full Consortium acknowledgements are available in the supplementary information of The Tobacco, Alcohol and Genetics.

Lung eQTL study: The lung eQTL study at Laval University was supported by the Chaire de pneumologie de la Fondation JD Bégin de l’Université Laval, the Fondation de l’Institut universitaire de cardiologie et de pneumologie de Québec, the Respiratory Health Network of the FRQS, the Canadian Institutes of Health Research (MOP - 123369), and the Cancer Research Society and Read for the Cure. Y. Bossé is the recipient of a Junior 2 Research Scholar award from the Fonds de recherche Québec – Santé (FRQS).

Lung meQTL study: The meQTL study based on the environment and Genetics in Lung Cancer Etiology (EAGLE) study was supported by the Intramural Research Program of the Division of Cancer Epidemiology and Genetics, National Cancer Institute, NIH, Department of Health and Human Services.

Epigenetics: This study was supported by NIH grant 1 R01 HL114094 to IAL-O and ZB, NIH grants 1 P30 H101258 and R37HL062569-13 to ZB. ZB is supported by the Ralph Edgington Chair in Medicine. CNM was supported in part by the department of Surgery, University of Southern California. Generation of epigenetic data was supported in part by a Norris Comprehensive Cancer Center core grant, award number P30CA014089, from the National Cancer Institute.

This work utilized the computational resources of the NIH High-Performance Computing (HPC) Biowulf cluster (http://hpc.nih.gov).

Notes

The study funders had no role in the design of the study; the collection, analysis, or interpretation of the data; the writing of the manuscript; or the decision to submit the manuscript for publication. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Cancer Institute or the National Institutes of Health. The authors have no conflicts of interest to disclose.

The authors thank deCODE genetics for contributing GWAS data.