DNA甲基化在吸菸與血脂性狀之間的中介作用

Abstract

吸菸是心血管疾病（cardiovascular disease, CVD）已知的重要危險因子，並與脂質特徵相關，例如高密度脂蛋白膽固醇（high-density lipoprotein cholesterol, HDL-C）與三酸甘油酯（triglycerides, TG）。DNA 甲基化可能是解釋這些關聯背後的表觀遺傳途徑之一。據此，本研究旨在探討 DNA 甲基化在吸菸與脂質特徵之間關聯中可能扮演的中介角色。 本研究使用臺灣人體生物資料庫中 2,463 名參與者的全基因組 DNA 甲基化資料進行高維度中介分析。首先，使用調整潛在混雜因子的線性回歸模型，評估吸菸狀態與脂質特徵之間的關聯。接著，針對具有顯著關聯的吸菸－脂質特徵組合，應用高維度中介檢定（high-dimensional mediation testing, HDMT）以辨識可能作為中介因子的 CpG 位點（FDR < 0.05）。為估計整體中介比例，本研究針對辨識出的 CpG 位點進行主成分分析，並透過基於主成分建立的中介模型估計潛在中介因子的中介比例。為進一步篩選出具獨立中介效應的位點，本研究使用 HIMA2建立多中介因子模型以同時考慮多個中介位點。 以目前吸菸作為暴露變項時，在 846,232個 CpG 位點中，分別有 109個與 65個 CpG 位點被辨識為吸菸－HDL-C 與吸菸－log(TG) 關聯中的顯著中介因子。DNA 甲基化估計分別可解釋吸菸與 HDL-C 及 log(TG)關聯中約 65% 與 7% 的中介比例。多重中介因子模型進一步辨識出14個與 HDL-C 相關以及1個與TG相關的獨立中介 CpG 位點。 本研究結果為DNA甲基化參與吸菸與脂質特徵之間的關聯的可能性提供了統計支持。然而，受到橫斷面研究之故有侷限，仍待後續的縱向研究與功能性研究，以闡明其潛在的生物學脈絡。

Cigarette smoking is a well-established risk factor for cardiovascular disease (CVD) and is associated with lipid traits such as high-density lipoprotein cholesterol (HDL-C) and triglycerides (TG). DNA methylation may represent an epigenetic mechanism underlying these associations. This study aimed to explore the potential mediating role of DNA methylation in the association between smoking and lipid traits. High-dimensional mediation analysis was performed using genome-wide DNA methylation data (Illumina EPIC Array) from 2,463 participants in the Taiwan Biobank (TWB). Linear regression models adjusted for potential confounders were used to assess the associations between smoking status and lipid traits. High-dimensional mediation testing (HDMT) was then applied to identify CpG sites that potentially act as mediators (FDR < 0.05) of the significantly associated smoking–lipid pairs. To estimate the overall mediation proportion, principal component analysis (PCA) was performed on the identified CpG sites, and the mediation proportion explained by the top principal components was estimated. In addition, HIMA2 was applied to identify independent mediating CpG sites. Using current smoking as the exposure, among 846,232 CpG sites, 109 and 65 CpG sites were identified by HDMT as potential mediators of the smoking–HDL-C and smoking–log(TG) associations, respectively. The net PCA-based mediation estimates across significant principal components were approximately 65% for HDL-C and 7% for log(TG). For log(TG), six principal components showed significant indirect effects in opposing directions, resulting in substantial cancellation in the net estimate. HIMA2 identified 14 joint-model mediation signals for HDL-C and identified cg00574958 in CPT1A for log(TG). These findings provide statistical evidence supporting a potential role of DNA methylation in the association between smoking and lipid traits. Further longitudinal and functional studies are needed to clarify the underlying biological mechanisms.