台灣族群泛癌遺傳研究： 整合大規模關聯分析與功能基因體圖譜繪製

Abstract

癌症長期位居臺灣主要死因之一，然而多種癌症共同易感性的遺傳架構仍尚未完全闡明。過去大多數研究主要聚焦於單一癌症類型，本研究則旨在探討臺灣族群中泛癌症易感性的共同及性別特異性遺傳決定因子，並進一步提供其潛在生物學機制的線索。 本研究納入中國醫藥大學附設醫院精準醫療計畫共 344,202 名參與者。經品質控制與傾向分數配對後，分別於男性、女性及全體樣本，進行泛癌症全基因體關聯分析。接著針對顯著位點於 25 種癌症類別中執行跨癌症表型關聯分析。此外，利用功能性定位與註釋分析平台 (FUMA)，包括 MAGMA 基因層級分析、表現數量性狀位點分析、染色質交互作用分析、組織富集分析以及基因集富集分析，以評估所找出的特定遺傳位點的生物學意義。 全基因體關聯分析分別於男性、女性及全體族群中鑑定出 19、30 及 50 個達建議顯著門檻（P<10^(-6)）的獨立易感位點。其中，10、7 及 25 個位點進一步達到全基因體顯著水準（P<〖5×10〗^(-8)），其餘位點則達建議顯著門檻。值得注意的是，位於 HLA 區域附近的 rs111984068 (P=9.46×10^(-25))、位於 KIR2DL4 基因內的 rs76869409 (P=5.05×10^(-24)) 以及位於 PRKCE 基因內的 rs12712969 (P=4.58×10^(-12))，在所有分析族群中皆達全基因體顯著水準。進一步的性別分層分析顯示不同的遺傳易感架構，在男性族群中，位於 POU5F1B 附近的 rs4582524 (P=1.16×10^(-9)) 以及位於 TERT 基因內的 rs13167280 (P=4.56×10^(-8)) 呈現顯著的男性特異性關聯；而在女性族群中，位於 CCDC170/ESR1 附近的 rs12665607 (P=1.81×10^(-17))及位於 CCDC170 基因內的 rs9383589 (P=8.95×10^(-17)) 則展現明顯的女性特異性效應。跨癌症表型關聯分析顯示，男性相關位點主要與前列腺癌、頭頸癌及肝癌相關；女性相關位點則主要與乳癌及子宮頸癌相關。在全體族群分析中，乳癌、頭頸癌及肝癌為最常觀察到關聯的癌症類型。基因層級分析發現，TERT、HLA-B 及 PSORS1C2 為男性族群中最具代表性的易感基因，支持端粒維持與免疫監視途徑可能參與泛癌症遺傳易感性；而 CCDC170 則為女性族群中最顯著的候選基因，與荷爾蒙相關癌症易感性一致。在全體族群分析中，額外鑑定出 RCCD1、PRC1 及 VPS33B 等易感基因，這些基因皆曾被報導與腫瘤發生及癌症進展相關。功能註釋分析顯示，多數易感變異位於非編碼調控區域，而基因集富集分析則顯示其與免疫相關途徑及 APOBEC3 介導之胞嘧啶脫胺相關生物過程具有顯著關聯。最後貝氏共定位分析顯示，全體族群分析中的 RCCD1 在多種組織中呈現較高的共定位後驗機率，而 MHC 區域亦反覆觀察到較高的共定位訊號。然而，fastENLOC 分析未發現任何基因達到區域共定位的證據（RCP > 0.50）。 總而言之，本研究闡明了臺灣族群泛癌症共同與性別特異性的遺傳易感架構，並提供相關潛在生物學機制的線索。研究結果支持免疫監視途徑、APOBEC3 介導之胞嘧啶脫胺作用以及基因遠端調控機制可能參與泛癌症遺傳易感性。其中，RCCD1 及 KIR2DL4 基因內的 rs76869409，因其在不同分析中的一致性關聯訊號，未來值得進一步評估其作為癌症風險分層及精準預防策略候選標記的潛力。

Cancer has consistently ranked among the leading causes of mortality in Taiwan, yet the genetic architecture underlying susceptibility across multiple cancer types remains incompletely understood. While most previous genetic studies have focused on individual malignancies, the present study aimed to identify shared and sex-specific genetic determinants of pan-cancer susceptibility in the Taiwanese population and to provide biological insights into the identified susceptibility loci. A total of 344,202 participants from the China Medical University Hospital Precision Medicine Project were included. After quality control and propensity score matching, genome-wide association studies (GWAS) were performed separately in male, female, and whole-population cohorts. Significant loci were subsequently evaluated through cross-cancer phenome-wide association studies (PheWAS) across 25 cancer categories. Functional Mapping and Annotation of GWAS (FUMA), including MAGMA gene-based analysis, expression quantitative trait locus (eQTL) mapping, chromatin interaction analysis, tissue enrichment analysis, and gene-set enrichment analysis, was conducted to investigate the biological relevance of the identified loci. GWAS identified 19, 30, and 50 independent suggestive susceptibility loci (P<10^(-6)) in the male, female, and whole-population analyses, respectively. Of these, 10, 7, and 25 loci exceeded the genome-wide significance threshold (P<〖5×10〗^(-8)), while the remaining loci met the suggestive significance threshold. Three single-nucleotide polymorphisms (SNPs), rs111984068 near the HLA region (P=9.46×10^(-25)), rs76869409 within KIR2DL4 (P=5.05×10^(-24)), and rs12712969 within PRKCE (P=4.58×10^(-12)), achieved genome-wide significance across all analytical cohorts. Sex-stratified analyses further revealed distinct susceptibility architectures. In males, rs4582524 near POU5F1B (P=1.16×10^(-9)) and rs13167280 within TERT (P=4.56×10^(-8)) exhibited strong male-specific associations, whereas rs12665607 near CCDC170/ESR1 (P=1.81×10^(-17)) and rs9383589 within CCDC170 (P=8.95×10^(-17)) showed prominent female-specific effects. Cross-cancer PheWAS analyses indicated that male-associated loci wer mainly linked to prostate, head and neck, and liver cancers, whereas female-associated loci were primarily related to breast and cervical cancers. In the whole-population analysis, breast, head and neck, and liver cancers represented the most frequently associated malignancies. Gene-based analyses identified TERT, HLA-B, and PSORS1C2 as the top susceptibility genes in males, supporting a potential role for telomere maintenance and immune surveillance pathways, whereas CCDC170 emerged as the dominant signal in females, consistent with hormone-related cancer susceptibility. In the combined analysis, additional susceptibility genes including RCCD1, PRC1, and VPS33B were identified, all of which have previously been implicated in tumor development and progression. Functional annotation revealed that most susceptibility SNPs were located within non-coding regulatory regions, whereas gene-set enrichment analysis implicated immune-related pathways and APOBEC3-mediated cytidine deamination processes. Bayesian colocalization analyses identified high-PP4 colocalization signals for RCCD1 across multiple tissues in the whole-population analysis, whereas recurrent high-PP4 signals were also observed within the MHC region. No loci achieved an RCP greater than 0.50 in the fastENLOC analyses. In conclusion, this study identified shared and sex-specific components of inherited pan-cancer susceptibility in the Taiwanese population. The findings suggest that immune surveillance pathways, APOBEC3-mediated cytidine deamination processes, and regulatory genomic mechanisms may contribute to shared genetic susceptibility across multiple cancer types. RCCD1 together with several highly reproducible susceptibility SNPs, particularly rs76869409 within KIR2DL4, warrant further evaluation as candidate markers for future cancer risk stratification and precision prevention strategies.