Genome-wide association study of lung adenocarcinoma in East Asia and comparison with a European population

Abstract

Lung adenocarcinoma is the most common type of lung cancer. Known risk variants explain only a small fraction of lung adenocarcinoma heritability. Here, we conducted a two-stage genome-wide association study of lung adenocarcinoma of East Asian ancestry (21,658 cases and 150,676 controls; 54.5% never-smokers) and identified 12 novel susceptibility variants, bringing the total number to 28 at 25 independent loci. Transcriptome-wide association analyses together with colocalization studies using a Taiwanese lung expression quantitative trait loci dataset (n = 115) identified novel candidate genes, including FADS1 at 11q12 and ELF5 at 11p13. In a multi-ancestry meta-analysis of East Asian and European studies, four loci were identified at 2p11, 4q32, 16q23, and 18q12. At the same time, most of our findings in East Asian populations showed no evidence of association in European populations. In our studies drawn from East Asian populations, a polygenic risk score based on the 25 loci had a stronger association in never-smokers vs. individuals with a history of smoking (P_interaction = 0.0058). These findings provide new insights into the etiology of lung adenocarcinoma in individuals from East Asian populations, which could be important in developing translational applications.

Fig. 1. Manhattan plot for GWAS meta-analysis of lung adenocarcinoma in East Asians.

The x-axis represents chromosomal location, and the y-axis represents -log₁₀(p-value). All p values were two-sided and not adjusted for multiple testing. The red horizontal line denotes the p value threshold for declaring genome-wide significance at $5\times {10}^{-8}$. For each box, red text represents a novel variant (12 novel variants, including the lead variants from 10 novel loci, rs12664490 by conditional analysis at 6p21.1, a locus previously reported in East Asians, and rs71467682 at 15q21.2, a locus preciously reported in Europeans); black text represents a previously reported association (16 variants in total, including three independently associated variants in 5p15.33 locus). For each locus, a green circle represents the top p value from the discovery samples, a red diamond represents the p value combining the discovery and the replication data, a black square represents the p value combining our discovery data and Chinese samples in Dai et al. (for three variants identified in a cross-ancestry analysis of East Asians and Europeans in Dai et al., see Supplementary Table 3). In summary, 28 variants at 25 loci achieved genome-wide significance, including 16 previously reported variants and 12 novel variants.

Fig. 2. Colocalization of lung adenocarcinoma GWAS signal from the new locus on Chr11 with FADS1 eQTL signal.

Colocalization analysis was performed using HyPrColoc with summary statistics from Taiwanese lung eQTL data (for FADS1 gene, A) and those of EA GWAS discovery set (B). LD R² (1000 Genomes, EA) of each SNP with the GWAS lead SNP, rs174559 (red circle), is color-coded as shown in the top band. Colocalization posterior probability (PP) is shown next to the candidate SNP, rs174559. Note that the p value of rs174559 in GWAS was based on the discovery data and did not include the Japanese replication data. All eQTL p values were two-sided and not adjusted for multiple testing.

Fig. 3. Comparing odds ratios (ORs) of lung adenocarcinoma susceptibility variants between East Asian (EA) and European (EUR) populations.

Here, the effect allele was defined as the minor allele in EA. Each error bar represents the 95% confidence interval of the OR (the center). A Susceptibility variants previously discovered (at genome-wide significance) in both EA and EUR populations. B Variants previously identified by multiple-ancestry meta-analysis of Chinese and EUR populations; C Variants were identified by multiple-ancestry meta-analysis combining EA samples in our study and EUR samples in ILCCO. D Variants identified only in EA populations. E Novel variants identified in the current study; F Variants identified only in EUR populations. Variants are labeled with *, **, *** and **** corresponding to 0.01 ≤ p_het < 0.05, 0.001 ≤ p_het < 0.01, 0.0001 ≤ p_het < 0.001 and p_het < 0.0001, respectively; here, p_het (t-statistic, two-sided) is the p value for testing the heterogeneity of effect sizes between EA and EUR populations. Sample sizes for EUR populations in all panels: 11,273 cases and 55,483 controls. Sample sizes for EA populations: 11,753 cases and 30,562 controls for (A, B, C, D, and F); 21,658 cases and 150,676 controls for (E).

Fig. 4. A polygenic risk score (PRS) is more strongly associated with risk of lung adenocarcinoma in never-smokers than in individuals with a history of smoking (P = 0.0058).

The PRS was defined based on 25 independent variants that achieved genome-wide significance in EA with weights derived from the meta-analysis of the current study (Supplementary Table 4). The odds ratios (ORs) and the standard errors of the 12 novel variants were based on 21,658 cases and 150,676 controls. The ORs and the standard errors of the other 13 variants were based on 11,753 cases and 30,562 controls. The figure shows the ORs and their 95% confidence intervals comparing each quintile group to the middle quintile for individuals with a history of smoking (blue) and never-smokers (red).

Fig. 5. The expected area under the receiver operating characteristic curve (AUC) of a polygenic risk score (PRS) built based on a GWAS of specified sample sizes for lung adenocarcinoma in never-smoking East Asians.

For “1 million controls”, the x-coordinate represents the number of cases, assuming the study has 1 million controls. For “Equal number of cases and controls”, the x-coordinate represents the numbers of cases, assuming the same number of cases and controls.