Background
Pulmonary embolism (PE) is a complex and multifactorial disease, together with deep vein thrombosis (DVT), commonly referred as venous thromboembolism (VTE). Twin studies have estimated the heritability of VTE to be approximately 50%, indicating that genetic factors may play a significant role in the pathogenesis of the disease [
1]. Over the past decades, family and population studies have discovered dozens of variants across the genome that contribute to the genetic risk of PE or VTE [
2,
3]. The largest meta-analysis of genome-wide association study (GWAS) for VTE has identified 34 independent genetic signals [
4]. Most of the reported associated loci regulate the coagulation and anticoagulation functions [
4,
5], which are vital disease-causing mechanisms in PE. Additionally, platelet, inflammation and erythrocytes have also been associated with the risk [
4]. However, we still have limited knowledge of the genetic architecture of PE, leaving a large proportion of heritability unexplained [
6,
7].
The global disease burden of PE has been steadily increasing in the past decade, affecting 100–200 per 100,000 individuals in western countries. Nevertheless, the prevalence of PE in EAS is 1/3–1/5 of that in EUR [
8]. Little is known about the genetic and other factors accounting for PE prevalence between East Asians (EAS) and European ancestry (EUR). Studies have suggested that ancestry-specific allele frequencies may explain the differences [
9]. Genetic studies with diverse populations are valuable for identifying more genetic risk factors of PE and maximizing the relevance of findings across populations [
10]. However, there have been few genetic studies for PE among EAS.
Genetic studies of VTE have been performed in subjects with EUR or African American (AA) ancestry. However, due to varying minor allele frequencies (MAF) across different populations, some of the associated variants identified in one population may not be replicated in another population. For example, rs6025 in
F5 (Factor V Leiden, FVL), the well-known leading single nucleotide polymorphism (SNP) in EUR VTE patients, has been rarely reported among EAS, with a MAF reported to be 0.024 in EUR and 0 in EAS [
11‐
13]. Similarly, variants in
THBD have been reported to be associated with VTE in AA but not in EUR [
14‐
16]. Such inconsistency makes it unreliable to generalize the genetic findings from EUR to other populations. Direct application of the PE risk assessment models with genetic factors discovered in the EUR may lead to inaccurate estimates of the actual PE risk among EAS, exacerbating health disparities in diverse populations [
17,
18].
Till now, there is no solid evidence of the PE risk assessment base on genetic study in EAS. To accelerate our understanding of the genetic basis for PE in EAS, we performed a large-scale genome-wide association study in the Han Chinese population and developed a population-specific polygenic risk score (PRS) to identify sub-populations at higher risk of PE.
Discussion
As most large-scale genetic studies of VTE have been conducted in EUR ancestry [
4,
45,
48,
49], we performed the first GWAS in the Han Chinese population to expand the genetic landscape of PE. We identified three genome-wide significant loci, of which two were known to be associated with PE (
FGG,
ABO). Another locus at
FABP2 reached the significance threshold both in the discovery stage and meta-analysis. The risk allele at the
FABP2 locus (rs1799883) is reported as a functional variant to increase the gene expression by qPCR and Western-blot expreriments. The results showed that the carriers of that mutation have higher blood TC and LDL-C [
50]. We further performed MR analysis and found that increased levels of LDC-C and TC were associated with a higher risk of PE, which implied the inhibition of LDL-C and TC to be a potential measure of PE prevention.
Some significant loci for the risk of VTE have been reported in European ancestry in the past but have not been validated in the Asian population, we attempted to replicate the previously known loci identified among the EUR population in the current study. In our study, 10 of the 22 leading variants were replicated, located on FGG, F11, ABO, CSRNP1, and FGA. The lack of replication in the current study might be partially attributed to insufficient power and diverse LD patterns across ancestries rather than different biological effects. These findings further emphasize the importance of including diverse ancestral groups in genomic studies to maximize the power for detecting disease associations.
Fatty acid binding proteins (FABPs) are key proteins in lipid transport. FABP2 can traffic lipids from the intestinal lumen to enterocytes and bind superfluous fatty acids to maintain a steady pool of fatty acids in the epithelium.
FABP2 polymorphism is known to be significantly associated with serum total cholesterol and LDL-C [
47]. Based on the above evidence, we speculate that variants in
FABP2 may contribute to PE through lipid metabolism function. Our findings were consistent with previous studies on the role of metabolic traits in PE [
51], which is the first time to be verified among the EAS population. Lipid-lowering drugs for prevention or even adjunctive therapy of PE have been proposed in many clinical trials [
52]. For example, statins contribute to PE prevention through anti-inflammatory and LDL-lowering effects [
53]. Proprotein convertase subtilisin/kexin type (PCSK9) inhibitor has also been identified to lower the risk of VTE by LDL reduction [
54]. Our study provided a shred of robust evidence that lipid-lowering therapy may also be considered to prevent PE occurrence in the EAS population.
In addition to FABP2, FGG, and ABO, there were also 13 loci reaching genome-wide significance in the meta-analysis. However, these loci did not achieve genome-wide significance in the discovery stage, nor had they been previously identified as PE-associated variants. There was currently insufficient evidence to support the reliability of those association results so we put emphasis on the three loci (FABP2, FGG, and ABO). More East Asian cohorts are needed to verify the associated loci in the future.
PRS has been widely used in the prediction of common diseases, and the PRS model for VTE had been validated in European ancestry. The early genetic risk model of VTE mainly focused on two loci, rs6025 and rs1799963.The Caprini risk assessment model, primarily relying on these two loci, is extensively utilized to predict VTE risk [
17]. However, these two variants are almost absent in the EAS population. With the entry into the GWAS era, more loci were used for risk stratification, Crous-Bou et al. established a new risk model based on the 16 SNPs and found that the risk of VTE in patients with high PRS score was 2.02 times that of patients with low PRS score, and achieved better results [
55]. Klarin generated a 297 variant polygenic risk score to predict VTE events among patients [
45]. Previous research has indicated a reduced accuracy of PRS models when transferred across ancestries [
18]. We, therefore, constructed a 288-variant PRS obtained from the EAS population for PE risk prediction. The PRS incorporated population-specific variants and outperformed in EAS population with an AUC of 0.765. Individuals in the top 10% group of PRS had a 5.08-fold of PE risk relative to the general population (30th–70th quantile). However, the model needs further validation in independent datasets with larger sample sizes.
As GWAS have uncovered hundreds of common genetic variants involved in PE susceptibility, our study shed new light on the genetic architecture of PE among Han Chinese population. Nevertheless, like most complex diseases, the common variants discovered in GWAS only explain a fraction of total disease heritability. The rare variants across the whole genome could also play an important role in disease development[
56]. Therefore, large-scale sequencing studies of PE in the EAS population are needed to measure the relationship between rare variants and PE risk.
There are several limitations in the current study. Since associations do not imply causation, further research is required to clarify the functional consequences of these novel signals in PE development. We acknowledge the imbalance of sample size and age and sex differences between cases and controls. However, we employed PLINK firth logistic regression with age and sex as covariates in GWAS analysis to control for type-I error issues. Considering the potential risk of inducing biased and spurious associations, we opted not to perform whole genome imputation. Instead, we restricted our imputation to the genomic regions within + / − 500kbp of the FABP2, FGG, and ABO loci, which constituted the main findings of our study. While we acknowledge that this approach may have led to the neglection of genetic signals, we held that these loci exhibited sound reliability. We look forward to expanding our analysis by incorporating more extensive genetic data in future studies.
Nevertheless, our study represents the first multicenter PE genetic study in diverse areas across China, which is a good representative of the Han Chinese population. We revealed through extensive genetic analyses that FABP2 polymorphism is associated with PE risk and the lipid-metabolic pathways are crucial in the PE development. Although more studies are needed to confirm the value of FABP2, the inhibition of FABP2 is promising to benefit from early intervention in reducing the risk of thrombosis. Our study also demonstrated the utility of applying a population-specific PRS model for PE risk prediction. The clinical use of PRS has the potential to recognize high-risk patients and improve health outcomes through eventual routine implementation as clinical biomarkers.
Acknowledgements
We appreciate the continuous support and contributions from Hongyu Zhao (Yale University), Xiaohui Wang (The First Affiliated Hospital of Chongqing Medical University), Guohua Yu (Weifang No.2 People's Hospital), Ruhong Xu (Dongguan People’s Hospital), Weijia Liu (Guizhou Provincial People’s Hospital), Zhonghe Zhang, Jun An (The First Affiliated Hospital of Dalian Medical University), Guofeng Ma, Chao Yan (Sir Run Run Shaw Hospital, Zhejiang University School of Medicine), Lijun Suo (Linzi District People’s Hospital), Xiaoqing Li, Yingying Pang, Beilei Gong (The First Affiliated Hospital of Bengbu Medical College), Wei Yang (Xiangya Hospital Central South University), Wenmei Zhang (Beijing Anzhen Hospital, Capital Medical University), Qin Luo (Fuwai Hospital, Chinese Academy of Medical Science; National Center for Cardiovascular Diseases), Hui Jia, Yunxia Liu (Central Hospital Affiliated to Shenyang Medical College), Ying Chen (People’s Hospital of Xinjiang Uygur Autonomous Region), Wei Zhou (Tianjin Medical University General Hospital), Ling Zhu, Yi Liu (Shandong Provincial Hospital), Xia Li (The First Affiliated Hospital, Sun Yat-Sen University), Xiaowei Gong (The Second Hospital of Hebei Medical University), Jifeng Li (Beijing Chao-Yang Hospital, Capital Medical University), Linli Duan, Simin Qing, Chunli Liu (The First Affiliated Hospital of Guangzhou Medical University), Baomin Fang, He Yang (Beijing Hospital), Chaosheng Deng, Minxia Yang, Dawen Wu (The First Affiliated Hospital of Fujian Medical University), Songping Huang, Qinghua Xu (Quanzhou First Hospital), Faguang Jin, Ning Wang, Yanli Chen, Yanyan Li (Tangdu Hospital), Jingping Yang, Xiyuan Xu, Baoying Bu (The Third Affiliated Hospital of Inner Mongolia Medical University), Chunxiao Yu, Zhenfang Lu, Jing Hua (Beijing Jingmei Group General Hospital), Chaobo Cui, Ning Wang (Harrison International Peace Hospital), Zhenyang Xu, Hongxia Zhang, Jinxiang Wang (Beijing Luhe Hospital, Capital Medical University), Shudong Zhang, Lijun Kang (Yantaishan Hospital), Lu Guo, Jing Zhang (Sichuan Academy of Medical Sciences & Sichuan Provincial People’s Hospital), Gang Chen, Yunxia Zhao, Zixiao Liu (The Third Hospital of Hebei Medical University), Jinming Liu, Qinhua Zhao (Shanghai Pulmonary Hospital), Xiaoyun Hu, Fangfang Fan (The First Hospital of Shanxi Medical University). Also, we would like to thank the research participants of WeGene and the members of WeGene Research Team for making this work possible. We would like to thank the patient participants of CURES and centers.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.