Epigenomic partitioning of an polygenic risk score for asthma reveals distinct genetically driven disease pathways

Background: Asthma is common chronic inflammatory disease of the airways with a heterogenous clinical presentation. Individual differences in asthma susceptibility remain poorly understood, although genetics is thought to play a major role. Aim: To build a polygenic risk score (PRS) for asthma and determine whether predictive genetic variants can be epigenomically linked to specific pathophysiological mechanisms. Methods: PRSs were constructed using data from genome-wide association studies and performance was validated using data generated in the Rotterdam Study, a Dutch prospective cohort of 14,926 individuals. Outcomes used were asthma, childhood-onset asthma, adulthood-onset asthma, eosinophilic asthma and exacerbations. Chromatin immunoprecipitation followed by high-throughput sequencing (ChIP-Seq) data from 14 primary cell types, including lung epithelial cells and T lymphocytes was used for epigenomic PRS partitioning. Results: All PRSs successfully predicted risk to develop asthma and related outcomes, with the strongest predictive power (2.42 odds ratios per PRS standard deviation, area under the curve of 0.736) achieved for childhood-onset asthma. PRSs allowed for stratification of the Rotterdam Study cohort into groups at low or high risk to develop asthma. PRS partitioning using genome-wide epigenomic profiles identified 5 clusters of variants within gene regulatory regions linked to specific asthma-relevant cells, genes and biological pathways. Conclusions: PRSs can predict whether individuals in a Dutch cohort developed asthma and asthma-related phenotypes, which is most effective for childhood-onset asthma. Importantly, we show that PRS partitioning based on epigenomics data dissects a genetic risk score into blocks of regulatory variants with differential predictive power, which likely represent distinct genetically driven disease pathways. These findings have potential implications for personalized risk mitigation and treatment strategies. Chromatin immunoprecipitation DNA-sequencing (ChIP-seq) data for histone 3 lysine 4 dimethylation (H3K4me2) modifications for the following cell types: naive CD8+ T cells (CD3+CD8+CD45RA+), memory CD8+ T cells (CD3+CD8+CD45RO+), naive B cells (CD3-CD19+CD27-IgD+), memory B cells (CD3-CD19+CD27+IgD-) resting eosinophils (CD33-CD66b+SIGLEC8+CD16-) resting neutrophils (CD33-CD66b+SIGLEC8-CD16+), lung epithelium (Bronchial brush in air-liquid interface culture), resting mast cells (Folkerts et al. Allergy, 2020).

背景：哮喘是一种常见的气道慢性炎症性疾病，临床表现具有异质性。尽管遗传因素被认为发挥关键作用，但哮喘易感性的个体差异仍未得到充分阐明。
研究目的：构建哮喘的多基因风险评分（polygenic risk score, PRS），并明确具有预测价值的遗传变异能否通过表观基因组学手段与特定病理生理机制建立关联。
研究方法：基于全基因组关联研究数据构建PRS，并采用荷兰鹿特丹研究（一项纳入14926名受试者的前瞻性队列研究）产生的数据对PRS的预测效能进行验证。本研究的结局指标包括哮喘、儿童起病型哮喘、成人起病型哮喘、嗜酸性粒细胞性哮喘以及哮喘急性加重。采用包括肺上皮细胞、T淋巴细胞在内的14种原代细胞类型的染色质免疫共沉淀联合高通量测序（chromatin immunoprecipitation followed by high-throughput sequencing, ChIP-Seq）数据开展表观基因组学PRS分区分析。
研究结果：所有PRS均能有效预测哮喘及相关结局的发病风险，其中儿童起病型哮喘的预测效能最优（每PRS标准差对应的优势比为2.42，曲线下面积为0.736）。PRS可将鹿特丹研究队列划分为哮喘发病低风险组与高风险组。基于全基因组表观基因组谱的PRS分区分析，在基因调控区域内鉴定出5组变异，这些变异与哮喘相关的特定细胞、基因及生物学通路存在关联。
研究结论：PRS能够预测荷兰队列个体的哮喘及哮喘相关表型发病风险，其中对儿童起病型哮喘的预测效果最佳。值得注意的是，本研究证实基于表观基因组学数据的PRS分区分析，可将遗传风险评分拆解为具有不同预测效能的调控变异模块，这些模块或代表不同的遗传驱动性疾病通路。上述发现可为个性化风险防控与治疗策略提供潜在参考价值。
本研究使用的表观基因组数据为以下细胞类型的组蛋白3赖氨酸4二甲基化（histone 3 lysine 4 dimethylation, H3K4me2）修饰的染色质免疫共沉淀测序（chromatin immunoprecipitation DNA-sequencing, ChIP-seq）数据：初始CD8+T细胞（CD3+CD8+CD45RA+）、记忆CD8+T细胞（CD3+CD8+CD45RO+）、初始B细胞（CD3-CD19+CD27-IgD+）、记忆B细胞（CD3-CD19+CD27+IgD-）、静息嗜酸性粒细胞（CD33-CD66b+SIGLEC8+CD16-）、静息中性粒细胞（CD33-CD66b+SIGLEC8-CD16+）、肺上皮细胞（气液界面培养的支气管刷检样本）以及静息肥大细胞（Folkerts等，《Allergy》，2020年）。