背景:白血病的临床治疗靶点尚待确定,白血病与端粒长度之间的因果关系尚不清楚。
方法:这项工作采用了来自eQTLGen联盟的2,200个可药用基因的顺式表达数量性状基因座(eQTL)和全基因组关联研究(GWAS)对来自英国生物库的7种血细胞类型中端粒长度的汇总数据,荷兰队列作为风险敞口。来自FinnGen和LeeLab的淋巴样白血病(LL)和髓样白血病(ML)的GWAS数据被用作发现和复制队列的结果。分别。从七个MR模型和一系列敏感性分析中得出了稳健的孟德尔随机化(MR)结果。进一步实施了基于汇总数据的MR(SMR)分析和全转录组关联研究(TWAS),以验证已鉴定的可药用基因与白血病之间的关联。采用单细胞类型表达分析来鉴定白血病偶然基因在人骨髓和外周血免疫细胞上的特异性表达。多变量MR分析,连锁不平衡评分回归(LDSC),和贝叶斯共定位分析进一步验证端粒长度与白血病的关系。中介分析用于评估鉴定的可药物基因通过端粒长度影响白血病的作用。全表型MR(Phe-MR)分析,用于评估白血病致病基因和端粒长度对1,403种疾病表型的影响。
结果:结合来自两个队列的MR估计的meta分析结果,SMR和TWAS分析,我们确定了五个LL因果基因(TYMP,DSTYK,PPIF,GDF15、FAM20A)和三个ML因果基因(LY75、ADA、ABCA2)作为白血病的有希望的药物靶标。单变量MR分析显示遗传预测的较高白细胞端粒长度增加LL的风险(比值比[OR]=2.33,95%置信区间[95%CI]1.70-3.18;P=1.33E-07),没有异质性和水平多效性。来自两个队列的荟萃分析的证据加强了这一发现(OR=1.88,95%CI1.06-3.05;P=0.01)。多变量MR分析显示白细胞端粒长度与LL之间存在因果关系,其他6种血细胞端粒长度没有干扰(OR=2.72,95%CI1.88-3.93;P=1.23E-07)。LDSC的证据支持白细胞端粒长度与LL之间的遗传正相关(rg=0.309,P=0.0001)。共定位分析显示,白细胞端粒长度在LL上的因果关系是由TERT区域的遗传变异rs770526驱动的。通过两步MR的介导分析表明,TYMP对LL的因果效应部分由白细胞端粒长度介导,介导比例为12%。
结论:我们的研究确定了几个与白血病风险相关的药物基因,并为白血病的病因和药物开发提供了新的见解。我们还发现,遗传预测的较高白细胞端粒长度增加了LL风险及其潜在的作用机制。
BACKGROUND: Clinical therapeutic targets for leukemia remain to be identified and the causality between leukemia and
telomere length is unclear.
METHODS: This work employed cis expression quantitative trait locus (eQTL) for 2,200 druggable genes from the eQTLGen Consortium and genome-wide association studies (GWAS) summary data for telomere length in seven blood cell types from the UK Biobank, Netherlands Cohort as exposures. GWAS data for lymphoid leukemia (LL) and myeloid leukemia (ML) from FinnGen and Lee Lab were used as outcomes for discovery and replication cohorts, respectively. Robust Mendelian randomization (MR) findings were generated from seven MR models and a series of sensitivity analyses. Summary-data-based MR (SMR) analysis and transcriptome-wide association studies (TWAS) were further implemented to verify the association between identified druggable genes and leukemia. Single-cell type expression analysis was employed to identify the specific expression of leukemia casual genes on human bone marrow and peripheral blood immune cells. Multivariable MR analysis, linkage disequilibrium score regression (LDSC), and Bayesian colocalization analysis were performed to further validate the relationship between telomere length and leukemia. Mediation analysis was used to assess the effects of identified druggable genes affecting leukemia via telomere length. Phenome-wide MR (Phe-MR) analysis for assessing the effect of leukemia causal genes and
telomere length on 1,403 disease phenotypes.
RESULTS: Combining the results of the meta-analysis for MR estimates from two cohorts, SMR and TWAS analysis, we identified five LL causal genes (TYMP, DSTYK, PPIF, GDF15, FAM20A) and three ML causal genes (LY75, ADA, ABCA2) as promising drug targets for leukemia. Univariable MR analysis showed genetically predicted higher leukocyte telomere length increased the risk of LL (odds ratio [OR] = 2.33, 95 % confidence interval [95 % CI] 1.70-3.18; P = 1.33E-07), and there was no heterogeneity and horizontal pleiotropy. Evidence from the meta-analysis of two cohorts strengthened this finding (OR = 1.88, 95 % CI 1.06-3.05; P = 0.01). Multivariable MR analysis showed the causality between leukocyte telomere length and LL without interference from the other six blood cell
telomere length (OR = 2.72, 95 % CI 1.88-3.93; P = 1.23E-07). Evidence from LDSC supported the positive genetic correlation between leukocyte
telomere length and LL (rg = 0.309, P = 0.0001). Colocalization analysis revealed that the causality from leukocyte
telomere length on LL was driven by the genetic variant rs770526 in the TERT region. The mediation analysis via two-step MR showed that the causal effect from TYMP on LL was partly mediated by leukocyte telomere length, with a mediated proportion of 12 %.
CONCLUSIONS: Our study identified several druggable genes associated with leukemia risk and provided new insights into the etiology and drug development of leukemia. We also found that genetically predicted higher leukocyte telomere length increased LL risk and its potential mechanism of action.