Polygenic risk score and lung adenocarcinoma risk among never-smokers by EGFR mutation status-a brief report

Abstract

We assessed the association between a genome-wide polygenic risk score (PRS) developed for lung adenocarcinoma (LUAD) risk and mutation on the epidermal growth factor receptor (EGFR) gene in 998 East Asian never-smoking female LUAD cases (518 EGFR-positive; 480 EGFR-negative) and 4,544 never-smoking controls using case-case and multinomial regression analyses. We found that the PRS was more strongly associated with EGFR-positive LUAD compared to EGFR-negative LUAD, where the association between the fourth quartile of the PRS and EGFR-positive LUAD (OR=8.63, 95% CI:5.67, 13.14) was significantly higher than the association between the fourth quartile of the PRS with EGFR-negative LUAD (OR=3.50, 95% CI: 2.44, 5.00) (p-heterogeneity=3.66x10⁻³). Our findings suggest that germline genetic susceptibility may be differentially associated with LUAD in never-smoking female East Asian patients depending on the cancer’s mutation status, which may have important public health and clinical implications.

Introduction

Lung cancer is a major global health problem and the leading cause of cancer death worldwide, representing about 20% of all cancer mortality in 2020¹. While smoking is the most common cause of lung cancer, nearly 25% of lung cancer patients are never-smokers, a proportion that is especially high among East Asian women².

Epidermal growth factor receptor (EGFR) is a transmembrane protein that regulates cellular proliferation and apoptosis³. Mutations in the EGFR gene have been found to be a defining hallmark of lung adenocarcinoma (LUAD), the most common histological subtype among never-smoking lung cancer cases. EGFR mutations are generally more common in never-smokers vs. smokers, LUAD vs. other subtypes, East Asian vs. European populations, and females vs. males^4,5.

We previously reported that two susceptibility loci identified in a genome-wide association study (GWAS) of LUAD among never-smoking East Asian women, namely rs9387478 (ROS1/DCBLD1) and rs2179920 (HLA-DPB1), were more strongly associated with EGFR-positive vs. EGFR-negative LUAD⁶. We recently expanded our GWAS of LUAD in individuals of East Asian ancestry that increased the total number of independent susceptibility loci identified to 28 across 25 regions⁷. Here, we investigated if overall genetic susceptibility to LUAD, defined as a polygenic risk score (PRS) and modelled using a genome-wide Bayesian-based approach, LDpred2^8,9, is differentially associated with LUAD by EGFR mutation status.

Methods

The study population consists of 998 female never-smoking histologically confirmed LUAD cases with data on EGFR mutation status and 4,544 female never-smoking controls from the Female Lung Cancer Consortium in Asia (FLCCA)- a multi-national collaboration of epidemiological studies of lung cancer from Taiwan, Mainland China, Hong Kong, Singapore, South Korea, and Japan. Details on FLCCA and the 11 studies contributing EGFR mutation status data have been previously described in detail⁶. Each study was approved by their local institutional review board and all participants provided informed consent prior to participation.

Genotyping and imputation for the FLCCA study was previously described⁶. Briefly, all samples were genotyped using the Illumina 370K, 610Q, or 660W microarrays either at the US National Cancer Institute’s Cancer Genomics Research Laboratory or at Gene-Square Biotech, Inc. in Beijing. Genotype imputation was conducted using IMPUTE2 software version 2.2.2, and the 1,000 Genomes Project Phase 1 version 3 data was used as the reference panel. The GWAS data are deposited at dbGAP (https://www.ncbi.nlm.nih.gov/gap, study accession: phs000716.v1.p1).

We derived a PRS for all FLCCA study participants using LDpred2^8,9, which is a genome-wide Bayesian-based approach designed to enhance the prediction power and improve the accuracy of risk score estimates by leveraging information on linkage disequilibrium (LD) from reference panels and modeling the correlation between nearby genetic variants. The weights for the 942,592 single nucleotide polymorphisms (SNP) included in the PRS were previously derived using summary statistics from 3,565 never-smoking LUAD cases and 16,238 never-smoking controls of East Asian ancestry from the Nanjing Lung Cancer Study (NJLCS), the National Cancer Center of Japan (NCC) Research Institute, and the Aichi Cancer Center (ACC)⁷.^8,97Tuning and validation were conducted using independent individual level data in FLCCA consisting of 4,438 LUAD cases and 4,544 controls, which generated an area under the receiver operating curve (AUC) value of 0.63. The PRS was then standardized and categorized into quartiles based on the distribution among controls.

Details on methods for DNA extraction of lung tumor tissue used by each study center have been previously described⁶. Briefly, genomic DNA was extracted from fresh, frozen or formalin-fixed, paraffin-embedded (FFPE) samples, and each participating study independently genotyped tumor tissues from lung cancer cases for EGFR mutations on exons 19 and 21. We defined cases with EGFR mutation on either exon 19 or 21 as EGFR-positive (n=518) and cases without such EGFR mutation as EGFR-negative (n=480).

We assessed the association between continuous PRS, PRS quartiles and tumor EGFR mutation status and estimated the odds ratios (OR) and 95% confidence intervals (CI) using logistic regression models adjusted for age, study, and the top 10 principal component scores to control for population stratification. We first conducted a case-case analysis restricting to LUAD cases with data on EGFR mutation status and compared EGFR-positive to EGFR-negative cases. Next, we conducted a multinomial logistic regression with a three-level outcome: controls, EGFR-negative cases and EGFR-positive cases. Differences between EGFR-negative and EGFR-positive LUAD estimates were tested by calculating the z-statistic using the covariance matrix and accounting for shared controls. P-values <0.05 were considered statistically significant and all analyses were performed using the R statistical software (version 4.2.2).

Results

In case-case comparisons treating EGFR-negative LUAD as the reference category, we found an association between continuous PRS and risk of EGFR-positive LUAD (p-value=0.02), as well as a dose-response relationship between quartiles of the PRS and EGFR-positive LUAD (p-trend=0.003) (Table 1). In the multinomial analysis treating the controls as the reference category, we found that the continuous PRS was associated with EGFR-negative LUAD and EGFR-positive LUAD with OR of 1.82 (95% CI: 1.62, 2.05) and 2.16 (95% CI: 1.92, 2.42) per standard deviation (SD), respectively. We further found that the association between the fourth quartile of the PRS with EGFR-positive LUAD (OR=8.63, 95% CI:5.67, 13.14) was significantly higher than the association between the fourth quartile of the PRS with EGFR-negative LUAD (OR=3.50, 95% CI: 2.44, 5.00) (p-heterogeneity=3.66x10⁻³) (Table1/Figure1).

Table 1.

Association between polygenic risk score (PRS) and lung adenocarcinoma by EGFR status

				Case-Case analysis		Multinomial analysis
	Sample count					EGFR-negative		EGFR-positive
PRS	EGFR-negative	EGFR-positive	Controls	OR (95% CI)a	P-value	OR (95% CI)b	P-value	OR (95% CI)c	P-value
Quartile 1	57	33	1138	REF		REF		REF
Quartile 2	75	80	1139	1.98 (1.14, 3.47)	0.02	1.29 (0.86, 1.92)	0.02	2.52 (1.60, 3.96)	6.67x10−5
Quartile 3	104	115	1131	2.10 (1.24, 3.60)	6.0x10−3	1.81 (1.24, 2.64)	2.13x10−3	3.85 (2.49, 5.95)	1.39x10−9
Quartile 4	244	290	1140	2.44 (1.46, 4.13)	8x10−4	3.50 (2.44, 5.00)	7.39x10−12	8.63 (5.67, 13.14)	7.97x10−22
PRS continuous	480	518	4548	1.17 (1.02, 1.34)	0.02	1.82 (1.62, 2.05)	1.18x10−8	2.16 (1.92, 2.42)	1.00x10−32

Abbreviations: OR, odds ratio; CI, confidence interval; PRS, polygenic risk score

^aEstimates for associations between PRS and EGFR-positive compared to EGFR-negative, adjusting for age, study and 10 principal component scores

^bEstimates for the associations between PRS and LUAD among EGFR-negative cases compared to controls, adjusting for age, study and 10 principal component scores

^cEstimates for the associations between PRS and LUAD among EGFR-positive cases compared to controls, adjusting for age, study and 10 principal component scores

^dTest for heterogeneity comparing estimates from the multinomial regression for EGFR-positive and EGFR-negative fourth quartile using the covariance matrix and accounting for shared controls resulted in P-value=3.66x10⁻³

^eEstimates are per 1 standard deviation (SD)

Figure 1. Association between quartiles of polygenic risk score (PRS) and lung adenocarcinoma by EGFR mutation status.

Odds ratios (OR) and 95% confidence intervals (CI) from a multinomial regression are plotted for each PRS quartile, with the first quartile as the reference category. Test for heterogeneity comparing estimates from the multinomial regression for EGFR-positive and EGFR-negative fourth quartile using the covariance matrix and accounting for shared controls resulted in P-value=3.66x10⁻³.

Discussion

We found that the genome-wide PRS developed for LUAD in East Asian individuals was more strongly associated with EGFR-positive LUAD compared to EGFR-negative LUAD, suggesting that germline genetic susceptibility may be differentially associated with LUAD in never-smoking female Asian patients depending on the cancer’s mutation status.

There have been limited efforts to assess whether genetic susceptibility is differentially associated with the risk of an EGFR mutation in lung cancer patients. A meta-analysis conducted in 2019 found that an EGFR mutation is significantly more likely to be found in lung cancer patients with first-degree family history of lung cancer than their counterparts without family history¹⁰. This finding was especially pronounced in Asian individuals diagnosed with LUAD. Further, several studies found individual germline loci to be associated with the likelihood of EGFR mutation in lung cancer., in genes including TERT, HLA-DPB1, ROS1/DCBLD1, FOXP4, BTNL2, BPTF, and TP63^11–13. To our knowledge, our study is one of the first to extend on previous work by observing differential association between a genome-wide summary of genetic susceptibility to LUAD and the risk of EGFR mutation.

These findings have potentially important implications for etiologic and translational lung cancer research by informing our understanding of how germline genetic susceptibility may impact the type of acquired mutations in LUAD. In never-smoking women, exposure to an exogenous risk factor such as environmental tobacco smoke and air pollution, may result in genotoxic damage that can be enhanced when the DNA repair system is compromised. As such, there may be a subset of individuals with germline mutations involved in the dysfunction of DNA repair for which genomic instability acts as a modifier for the risk of an EGFR mutation in lung tumor¹⁰.

Further, given that patients with LUAD respond differently to treatments that are used as targeted therapy depending on their EGFR mutation status¹⁴, our findings may have important public health and clinical implications, which may guide risk stratification, screening, and treatment. Larger future studies are needed to expand on our findings and investigate potential pathways through which germline genetic susceptibility may impact the acquisition of EGFR mutations.