Abstract
目的
利用两样本孟德尔随机化方法探究血液中的代谢物与冠心病发病风险之间的因果关系。
方法
采用公共数据库中的大规模的代谢组-全基因组关联分析数据和冠心病全基因组关联分析数据,通过逆方差加权法和其他4种两样本孟德尔随机化模型来评估血液代谢物与冠心病之间的因果关系。并进行异质性检验、基因多效性检验和敏感性分析来评估结果的稳定性和可靠性。
结果
486个代谢物中,共有32个代谢物与冠心病的因果关系效应值达到名义上显著(逆方差加权法,P < 0.05),其中包含11个已知代谢物和21个未知代谢物。有3个已知代谢物[N-乙酰鸟氨酸,(Des-Arg9)-缓激肽,丁二酰基肉碱]在至少3种孟德尔随机化模型中达到统计学显著。利用leave-one-out法进行敏感性分析并剔除混杂的工具变量后,3个代谢物对应的因果效应不再具有统计学意义。
结论
在本研究中纳入的486个代谢物,尚未有充足证据表明其与冠心病发生风险之间存在稳健的因果关系。
Keywords: 孟德尔随机化, 血液代谢物, 冠心病, 因果关联
Abstract
Objective
To explore the causal relationship between blood metabolites and the risk of coronary artery disease (CAD) using a two-sample Mendelian randomization (MR) approach.
Methods
Based on the data from a large-scale metabolome-based genome-wide association study (mGWAS) and the GWAS of CAD, we investigated the causality between blood metabolites and CAD using an inverse variance weighted (IVW) method and another 4 two-sample MR models. Heterogeneity, horizontal pleiotropy, and sensitivity tests were performed to evaluate the stability and reliability of the results.
Results
Among the 486 blood metabolites, 32 metabolites showed nominally causative association with CAD with the IVW method (P < 0.05), including 11 known metabolites and 21 unknown metabolites. Three known metabolites [N-acetylornithine, bradykinin-des-arg(9), and succinylcarnitine] were statistically significant in at least 3 MR models, but their causal effects on CAD were no longer significant after sensitivity analysis using leave-one-out method and elimination of the confounding instrumental variables.
Conclusion
There is no strong evidence to support a robust causal relationship between the 486 blood metabolites and the risk of CAD.
Keywords: Mendelian randomization, blood metabolites, coronary artery disease, causal relationship
冠状动脉粥样硬化性心脏病,简称冠心病,是世界范围内最常见的心血管疾病,也是导致患者死亡的主要原因之一[1]。虽然药物和手术治疗在近年来取得了巨大的进步,但是冠心病患者的死亡率仍然很高[2-4]。因此,了解冠心病的发病机制和识别新的预防靶点对冠心病的防治具有重要意义。
血液中的代谢物作为一种环境暴露下的功能中间体,其往往可以反应个体的遗传组成并能预测或影响疾病的发生发展[5]。近年来,血液代谢组学的研究已为预测心血管疾病的发生提供了众多的生物标志物并建立了可靠的预测模型[6-8]。但是,许多代谢物往往只能提供与疾病发生的关联性,其因果关系尚不明确。孟德尔随机化(MR)作为一种流行的遗传流行病学研究设计方法,它通过使用遗传变异作为工具变量,可以探究暴露和结局之间的因果关系[9]。然而到目前为止,在探究与冠心病发生风险相关的MR分析中,其关注的暴露因素大多是广泛表型,如身高[10]、腰围[11]、臀围[12]等,鲜有关注血液代谢物这样的暴露因素集合。基于代谢组学的全基因组关联研究(mGWAS)是一种识别代谢物数量性状位点[13]以了解疾病相关遗传变异的代谢背景的有效途径。Shin等于2014年发表了迄今为止最大规模的mGWAS,绘制了人类血液代谢物的遗传图谱[14],为血液代谢组学的遗传基础提供重要参考价值。Yang等[15]已经利用两样本MR的分析方法评估了这些血液中的代谢物与5种主要的精神疾病之间的因果关系,并成功地发现了两个代谢物与精神分裂症和注意缺陷或多动障碍之间存在稳健的因果关系,为该类疾病的预测和治疗提供了重要的参考。基于此方法来探究这些血液代谢物与冠心病发生风险之间的因果关系。有助于更深入地了解冠心病的发病机制,并可能为冠心病患者的临床诊疗提供新的见解,但目前尚未有相关的报道。
因此,本研究从分子机制角度出发,采用两样本MR分析方法,使用上述大规模的mGWAS数据为暴露文件,以及另一项超大规模的冠心病GWAS数据为结局文件,以探究这些血液中的代谢物与发生冠心病之间的因果关系。本研究具有一定的理论依据和临床转化价值,研究结果可以为指导冠心病的风险预测和治疗工具的开发提供参考。
1. 资料和方法
1.1. 资料
本研究采用2014年Shin等<sup>[<xref ref-type="bibr" rid="b14">14</xref>]</sup>发表于Nature Genetics上的迄今为止最大规模的mGWAS数据作为暴露文件。该数据是一项包含7824例欧洲人的荟萃分析数据,经过严格的质量控制,共210万个SNP位点和486种血液代谢物(其中包含309种已知代谢物和177种未知代谢物)用于全基因组关联分析。这些代谢物可分为8种代谢物大类:碳水化合物、氨基酸、核苷酸、辅因子和维生素、脂类、肽类、能量产物和异源性生物代谢产物。所有关联分析的汇总数据可在数据库网站公开获取:<a href="http://metabolomics.helmholtz-muenchen.de/gwas/" target="_blank">http://metabolomics.helmholtz-muenchen.de/gwas/</a>.
冠心病的GWAS数据来源于2011年Schunkert等<sup>[<xref ref-type="bibr" rid="b16">16</xref>]</sup>发表于Nature Genetics的一项包含22个独立研究的超大规模的冠心病荟萃分析数据,样本来自欧洲人群的22 233例冠心病患者和64 762例健康人,共有约240万个SNP位点用于关联分析。数据收录在CARDIoGRAMplusC4D,可在数据库网站公开获取:<a href="http://www.cardiogramplusc4d.org/" target="_blank">http://www.cardiogramplusc4d.org/</a>.
1.2. 工具变量的选择
我们对486种代谢物的遗传变异采用了统一的入选标准。选择MR分析中常用的较为宽松的阈值,即P < 1×10-5为显著的关联分析结果入选条件;在提取出每个代谢物对应的显著的SNP后,以千人基因组中欧洲人(EUR)基因型为参考模板,进行连锁不平衡分析,同时满足以下三个条件认为连锁不平衡并保留P值最小的SNP作为独立的遗传变异:(1)位于同个染色体;(2)相互距离在500 kb以内;(3)连锁不平衡参数r2>0.1。
因为处于同个代谢通路的代谢物可能会受到相似的遗传变异调控,即可能存在多个代谢物与同一个SNP显著相关,这将违反MR假设标准。因此,本研究采用限制性选取工具变量的方法[17],排除与两个以上代谢物均显著相关的SNP。同时排除已知的冠心病风险因素相关的SNP(包括身体质量指数[18]、身高[10]、腰围[11]、臀围[12]、腰臀比[19]、血脂[20]相关的SNP)。
1.3. 统计分析
本研究涉及的暴露因素为血液中的代谢物,数量众多,与既往的单一暴露因素的MR研究相比,有更为巨大的工作量。因此,本研究中通过编写Perl代码、R代码和Shell代码进行批量处理,即依次探究每个代谢物与冠心病的因果关系。其中,连锁不平衡分析采用PLINK(version 1.9)软件[21];MR分析、基因多效性检验以及敏感性分析采用R中的TwoSampleMR软件包(version 0.4.22)[22]于Linux系统上进行分析。
1.3.1. MR分析
本研究采用逆方差加权法(IVW)[23]作为首要的因果效应估计。IVW法是一种较为理想状态下的估计,是假设在所有遗传变异都是有效工具变量的基本前提下进行的有效分析,具有较强的因果关系检测能力。但是IVW法特别要求遗传变异仅通过研究中的暴露影响目标结局。尽管此研究已尽可能排除了已知的混杂的SNP,然而仍然有许多未知混杂因素会导致基因多效性并对效应值的估计产生偏倚。因此,我们采用了另外4种方法来检验结果的可靠性和稳定性,即MREgger回归[24]、加权中位数法(WME)[25]、基于众数的简单估计[26]、基于众数的加权估计[26]。依次对每个代谢物进行MR分析,如果以上五种不同的MR模型对因果效应产生了相似的估计值,我们则认为该代谢物与冠心病的因果关系是稳定且可靠的。
IVW法分析的结果中,我们采用严格的多重假设检验的阈值P < 1.03×10-4( P < 0.05/486)来检验显著的因果关系。并且同时关注P值大于等于1.03×10-4但小于0.05值的代谢物来作为冠心病潜在风险预测因子。P值小于0.05的因果关系将进行如下的异质性检验和基因多效性检验。
1.3.2. 异质性检验
由于不同分析平台、实验条件、入选人群的以及SNP的差异,两样本MR分析法可能存在异质性,从而对因果效应的估计产生偏倚。因此,本研究中对主要的IVW分析法和MR-Egger回归采取异质性检验,检验的结果中P值大于0.05则认为纳入的工具变量不存在异质性,可以忽略异质性对因果效应估计产生的影响。
1.3.3. 基因多效性检验
MR分析的假设之一是工具变量只能通过暴露影响结局,若工具变量不通过影响暴露而直接影响结局则违背了MR思想,所以需要检验暴露与结局之间的因果推断是否存在基因多效性。采用MR-Egger回归分析可以来评价基因多效性产生的偏倚,其回归截距可以评估多效性的大小,截距越接近于0,则基因多效性的可能性越小。本研究中通过判断基因多效性检验的P值来衡量分析中是否存在基因多效性,若P>0.05,则认为因果分析中基因多效性的可能性较弱,可以忽略其产生的影响。
1.3.4. 敏感性分析
除了采用上述4种方法(MR-Egger回归法、加权中位数法、基于众数的简单估计法、基于众数的加权估计法)来检验结果的可靠性和稳定性。本研究还采用leave-one-out法来进行敏感性分析。即对IVW法中P值小于0.05,并且通过了异质性检验和基因多效性检验的代谢物,逐一去除各个相关的SNP并计算剩余的SNP的合并效应,以评估各个SNP对于代谢物的影响。
2. 结果
2.1. 工具变量(SNP)信息
在486个代谢物中,与之关联的P < 1×10-5的共有39 142个SNP,连锁不平衡分析后,得到10 905个独立的SNP,其中,有447个SNP至少与两个代谢物显著相关。另外,与冠心病风险因素相关的SNP共有319个,这些SNP均不包含在本研究中。排除混杂的SNP后,共10 458个SNP纳入后续分析,这些SNP在冠心病的GWAS数据中存在9108个,位点覆盖率为87%。工具变量的质控流程图如图 1所示。每个代谢物对应的工具变量的数量的中位数为13,后续分析中排除拥有工具变量的数量小于等于3的5个代谢物和大于等于100的5个代谢物。
1.

用于MR分析的工具变量质控流程图
Flow chart for quality control of the instrumental variables for MR analyses.
2.2. MR分析结果
本研究采用IVW法作为首要的评估代谢物与冠心病之间因果关系的方法。共有32个代谢物与冠心病的因果关系效应值达到名义上显著(P < 0.05),其中,包含已知代谢物11个,未知代谢物21个;未发现多重假设检验(P < 1.03×10-4)后仍然显著的代谢物(表 1)。
1.
IVW方法达名义上显著的代谢物对应的效应值结果
Nominally significant results of IVW method
| ID | Metabolite | nSNP | Beta | SE | P | OR (95%CI) |
| X: Unknown metabolite; nSNP; Number of the SNP used for testing; SE: Standard error; OR: Odds ratio; 95% CI. 95% confidence interval. | ||||||
| M35187 | X-13429 | 6 | -0.53 | 0.14 | 1.41E-04 | 0.59 (0.45-0.77) |
| M15140 | kynurenine | 39 | 1.31 | 0.39 | 7.55E-04 | 3.70 (1.73-7.92) |
| M35451 | X-13658 | 13 | 0.20 | 0.06 | 1.10E-03 | 1.22(1.08-1.38) |
| M32651 | X-11334 | 26 | -0.60 | 0.21 | 4.66E-03 | 0.55 (0.36-0.83) |
| M34339 | X-12729 | 16 | 0.18 | 0.07 | 7.26E-03 | 1.19 (1.05-1.36) |
| M34912 | X-13215 | 42 | 0.65 | 0.24 | 7.66E-03 | 1.91 (1.19-3.09) |
| M34527 | X-12844 | 14 | -0.71 | 0.27 | 8.18E-03 | 0.49 (0.29-0.83) |
| M34040 | X-12510 | 15 | -0.29 | 0.11 | 8.30E-03 | 0.75 (0.61-0.93) |
| M34389 | 1 -methylxanthine | 15 | -0.41 | 0.16 | 9.56E-03 | 0.67 (0.49-0.91) |
| M34456 | gamma-glutamylisoleucine | 22 | 0.57 | 0.22 | 9.69E-03 | 1.77 (1.15-2.73) |
| M33190 | X-11845 | 17 | 0.16 | 0.06 | 1.08E-02 | 1.17 (1.04-1.33) |
| M22649 | X-09108 | 12 | 1.04 | 0.43 | 1.50E-02 | 2.83 (1.22-6.53) |
| M35331 | X-13553 | 6 | -0.88 | 0.37 | 1.65E-02 | 0.42 (0.20-0.85) |
| M37058 | succinylcarnitine | 46 | 0.45 | 0.19 | 1.99E-02 | 1.57 (1.07-2.29) |
| M34513 | X-12830 | 11 | 0.33 | 0.14 | 2.05E-02 | 1.40(1.05-1.86) |
| M33957 | 1-heptadecanoylglycerophosphocholine | 9 | -0.58 | 0.26 | 2.34E-02 | 0.56 (0.34-0.92) |
| M32802 | X-11485 | 5 | 0.54 | 0.24 | 2.43E-02 | 1.71 (1.07-2.73) |
| M32593 | heme | 14 | 0.41 | 0.18 | 2.48E-02 | 1.51 (1.05-2.16) |
| M16634 | X-04357 | 8 | -0.53 | 0.24 | 2.70E-02 | 0.59 (0.37-0.94) |
| M34420 | bradykinin, des-arg(9) | 24 | -0.14 | 0.06 | 2.82E-02 | 0.87 (0.77-0.99) |
| M33683 | X-12261 | 9 | 0.19 | 0.09 | 2.84E-02 | 1.21 (1.02-1.43) |
| M15630 | N-acetylornithine | 19 | -0.19 | 0.09 | 3.12E-02 | 0.83 (0.70-0.98) |
| M35270 | X-13496 | 41 | 0.79 | 0.37 | 3.14E-02 | 2.20(1.07-4.51) |
| M32855 | X-11538 | 17 | -0.47 | 0.22 | 3.20E-02 | 0.62 (0.40-0.96) |
| M33666 | X-12244 | 29 | 0.42 | 0.20 | 3.54E-02 | 1.52(1.03-2.26) |
| M32418 | myristoleate (14:1n5) | 6 | -0.60 | 0.28 | 3.54E-02 | 0.55 (0.32-0.96) |
| M36515 | X-14588 | 7 | -2.85 | 1.36 | 3.63E-02 | 0.06 (0.00-0.83) |
| M35397 | X-13619 | 28 | -0.79 | 0.38 | 3.70E-02 | 0.45 (0.21-0.95) |
| M32338 | glycine | 29 | -0.48 | 0.24 | 4.71E-02 | 0.62 (0.38-0.99) |
| M34306 | X-12696 | 23 | 0.65 | 0.33 | 4.82E-02 | 1.92(1.01-3.65) |
| M33883 | X-12441 | 14 | 0.15 | 0.08 | 4.94E-02 | 1.16 (1.00-1.35) |
| M15335 | mannitol | 10 | -0.20 | 0.10 | 5.00E-02 | 0.82 (0.66-1.00) |
在11个已知的代谢物中,包括4种可能与增加冠心病发生风险相关的代谢物,即犬尿氨酸,γ-谷氨酰异亮氨酸,丁二酰基肉碱,血红素;7种可能与降低冠心病发生风险相关的代谢物,即1-甲基黄嘌呤,溶血磷脂酰胆碱,(Des-Arg9)-缓激肽,N-乙酰鸟氨酸,肉豆蔻酸酯,甘氨酸,甘露醇。
2.3. 结果的可靠性和稳定性评估
对以上11个已知的代谢物所对应的工具变量进行异质性检验和基因多效性检验,排除2个存在异质性或基因多效性的代谢物(犬尿氨酸和1-甲基黄嘌呤)。五种MR模型、异质性检验和基因多效性检验结果见表 2。剩余的9个代谢物中,有3个代谢物对应的5种MR模型中有至少3种模型的P值小于0.05[N-乙酰鸟氨酸,(Des-Arg9)-缓激肽,丁二酰基肉碱]。尽管五种MR模型都未能全部达到统计学意义上显著,但是它们均具有相似的效应值,其原因可能是因为IVW法比其他四种MR模型拥有更高的检验效能。3个代谢物的MR分析结果散点图见图 2。
2.
已知代谢物的5种MR分析以及异质性检验和基因多效性检验结果
Results of 5 MR models of known metabolites and the heterogeneity and pleiotropy tests
| Metabolite | MR method | P | OR (95%CI) | Heterogeneity | Horizontal pleiotropy |
| MR Egger | 0.40033 | 2.74 (0.27-27.83) | 4.03E-6 | 0.7887 | |
| wme | 0.10592 | 1.94 (0.87-4.33) | |||
| Kynurenine | ivw | 0.00076 | 3.70 (1.73-7.92) | 0.00001 | |
| Simple mode | 0.33751 | 2.24 (0.44-11.34) | |||
| Weighted mode | 0.55621 | 1.47(0.41-5.31) | |||
| MR Egger | 0.57670 | 0.80 (0.39-1.68) | 0.67876 | 0.9706 | |
| WME | 0.06124 | 0.76 (0.57-1.01) | |||
| Mannitol | ivw | 0.04995 | 0.82 (0.66-1.00) | 0.76761 | |
| Simple mode | 0.16115 | 0.71 (0.46-1.10) | |||
| Weighted mode | 0.15848 | 0.74 (0.50-1.09) | |||
| MR Egger | 0.05559 | 0.79 (0.63-0.99) | 0.05882 | 0.5272 | |
| WME | 0.00979 | 0.82 (0.70-0.95) | |||
| N-aeetylornithine | ivw | 0.03116 | 0.83 (0.70-0.98) | 0.06825 | |
| Simple mode | 0.55030 | 0.84 (0.49-1.46) | |||
| Weighted mode | 0.01738 | 0.82 (0.71-0.95) | |||
| MR Egger | 0.23923 | 0.40 (0.09-1.79) | 0.16543 | 0.5468 | |
| WME | 0.05542 | 0.53 (0.28-1.01) | |||
| Glycine | ivw | 0.04711 | 0.62 (0.38-0.99) | 0.18519 | |
| Simple mode | 0.50131 | 0.68 (0.23-2.04) | |||
| Weighted mode | 0.33285 | 0.65 (0.27-1.54) | |||
| MR Egger | 0.18081 | 0.21 (0.03-1.38) | 0.53684 | 0.3566 | |
| WME | 0.01773 | 0.43 (0.22-0.86) | |||
| Myristoleate (14:ln5) | ivw | 0.03545 | 0.55 (0.32-0.96) | 0.51945 | |
| Simple mode | 0.28016 | 0.47(0.14-1.60) | |||
| Weighted mode | 0.10360 | 0.41 (0.17-0.99) | |||
| MR Egger | 0.63040 | 1.25 (0.52-3.03) | 0.79050 | 0.6550 | |
| WME | 0.11848 | 1.46 (0.91-2.36) | |||
| Heme | ivw | 0.02479 | 1.51 (1.05-2.16) | 0.83430 | |
| Simple mode | 0.33108 | 1.47(0.70-3.09) | |||
| Weighted mode | 0.32179 | 1.47 (0.71-3.05) | |||
| MR Egger | 0.11377 | 0.15 (0.02-1.17) | 0.54925 | 0.2373 | |
| WME | 0.07326 | 0.54(0.27-1.06) | |||
| 1 -Heptadecanoyl glycerophosphocholine |
ivw Simple mode |
0.02337 | 0.56 (0.34-0.92) | 0.47467 | |
| 0.07249 | 0.33 (0.12-0.94) | ||||
| Weighted mode | 0.12721 | 0.44 (0.17-1.13) | |||
| MR Egger | 0.00103 | 0.35 (0.22-0.57) | 0.78504 | 0.0088 | |
| WME | 0.00950 | 0.61 (0.41-0.88) | |||
| 1 -Methylxanthine | ivw | 0.00956 | 0.67(0.49-0.91) | 0.19300 | |
| Simple mode | 0.75693 | 0.87(0.36-2.10) | |||
| Weighted mode | 0.00090 | 0.45 (0.31-0.65) | |||
| MR Egger | 0.40292 | 0.88 (0.65-1.19) | 0.09187 | 0.9872 | |
| WME | 0.00075 | 0.77 (0.66-0.90) | |||
| Bradykinin, des-arg(9) | ivw | 0.02817 | 0.87 (0.77-0.99) | 0.11768 | |
| Simple mode | 0.56018 | 0.90 (0.64-1.27) | |||
| Weighted mode | 0.01041 | 0.78 (0.66-0.93) | |||
| MR Egger | 0.51601 | 1.50 (0.45-5.01) | 0.51997 | 0.7777 | |
| WME | 0.17896 | 1.60 (0.81-3.18) | |||
| gamma-Glutamylisoleucine | ivw | 0.00969 | 1.77 (1.15-2.73) | 0.57806 | |
| Simple mode | 0.02833 | 4.22(1.27-13.98) | |||
| Weighted mode | 0.35725 | 1.51 (0.64-3.54) | |||
| MR Egger | 0.06882 | 2.11 (0.96-4.62) | 0.36369 | 0.4005 | |
| WME | 0.03478 | 1.83 (1.04-3.20) | |||
| Succinylcarnitine | ivw | 0.01994 | 1.57 (1.07-2.29) | 0.37414 | |
| Simple mode | 0.09103 | 2.39 (0.89-6.43) | |||
| Weighted mode | 0.02542 | 1.94(1.11-3.40) |
2.

三个具有潜在因果关系的代谢物的5种MR模型散点图
Scatter plots of the 5 MR models for 3 metabolites with potential causal relationship with CAD. A: Nacetylornithine; B: Bradykinin, des-arg(9); C: Succinylcarnitine.
用leave-one-out法对以上3个代谢物[N-乙酰鸟氨酸,(Des-Arg9)-缓激肽,丁二酰基肉碱]的因果效应进行敏感性分析。3个代谢物中都存在至少1个SNP对结果的效应值产生显著影响,因此,对这些SNP进行剔除后,我们重新对这3个代谢物进行了MR分析。N-乙酰鸟氨酸,(Des-Arg9)-缓激肽和二酰基肉碱的IVW法对应的效应值均不再显著(表 3)。
3.
Leave-one-out法检验剔除混杂SNP后的3种代谢物MR分析结果
MR analysis results of the 3 metabolites after removing mixed SNP by leave-one-out method
| Metabolite | MR method | nSNP | P | OR (95%CI) |
| MR Egger | 18 | 0.37779 | 0.72 (0.36-1.45) | |
| WME | 18 | 0.31579 | 0.80 (0.52-1.23) | |
| N-acetylornithine | ivw | 18 | 0.46514 | 0.87(0.60-1.26) |
| Simple mode | 18 | 0.72676 | 0.88(0.44-1.77) | |
| Weighted mode | 18 | 0.49999 | 0.85 (0.53-1.36) | |
| MR Egger | 20 | 0.53604 | 1.12(0.78-1.62) | |
| WME | 20 | 0.66920 | 1.04 (0.87-1.25) | |
| Bradykinin, des-arg(9) | ivw | 20 | 0.67259 | 1.03 (0.90-1.18) |
| Simple mode | 20 | 0.87332 | 0.97(0.69-1.38) | |
| Weighted mode | 20 | 0.86910 | 0.97(0.69-1.36) | |
| MR Egger | 45 | 0.15953 | 2.24 (0.74-6.74) | |
| WME | 45 | 0.04549 | 1.95 (1.01-3.77) | |
| Succinylcarnitine | ivw | 45 | 0.06500 | 1.51 (0.97-2.32) |
| Simple mode | 45 | 0.13908 | 2.44 (0.76-7.79) | |
| Weighted mode | 45 | 0.07326 | 2.08 (0.95-4.57) |
3. 讨论
本研究运用了公共数据库中大规模的mGWAS和GWAS数据,采用无偏倚的两样本孟德尔随机化分析方法探究了486种血液代谢物与冠心病发生风险之间的因果关系。然而经过严格的质量控制,尚未找到非常有力的证据表明这些血液代谢物与冠心病的发生之间存在直接的因果关联。
此研究中涉及的486种血液中的代谢物,尽管均未能通过多重假设检验的阈值,但仍然为我们提供了11种潜在的冠心病风险预测因子。包括4种可能与增加冠心病发生风险相关的代谢物(犬尿氨酸,γ-谷氨酰异亮氨酸,丁二酰基肉碱,血红素)。其中包含的3个代谢物[即,N-乙酰鸟氨酸,(Des-Arg9)-缓激肽,丁二酰基肉碱]在至少3种孟德尔随机化模型中都达到了统计学显著。本研究发现N-乙酰鸟氨酸和(Des-Arg9)-缓激肽可能作为潜在的保护性物质从而降低冠心病的发生风险。N-乙酰鸟氨酸是人血去蛋白血浆中的次要成分,暂未发现其关于心血管研究的报道。缓激肽作为一种血管活性激肽,具有改善心功能并可以降低远期心脏事件的作用[27]。实验研究表明,缺乏缓激肽B2受体基因的小鼠更容易出现高血压,心脏肥大和心肌损伤[28]。由此可见,(Des-Arg9)-缓激肽在保护心血管疾病和降低冠心病发生风险之间可能具有重要作用。丁二酰基肉碱是来自血液和肝脏中能量代谢的中间体,已有文献报道其是心血管疾病的危险因素之一[29],然而其与冠心病发生之间的因果关系仍然需要后续的进一步研究来证实。
本研究具有如下创新性:(1)本研究从分子机制角度出发,以血液中的代谢物为暴露因素,探究其与冠心病发生风险之间的因果关系,具有较强理论依据和重要临床研究价值;(2)本研究采用严格的质控条件和分析方法,运用多种模型来评估因果效应,研究结果具有可靠性和稳定性;(3)与既往的单一暴露因素的孟德尔随机化研究相比,本研究中涉及的暴露因素为血液中的代谢物,数量众多,具有十分巨大的工作量和分析挑战性。本研究中也存在一定的局限性:(1)mGWAS数据和冠心病的GWAS数据均来源于欧洲人群,后续更为全面的研究仍需在不同人种之间展开;(2)初步分析得到的冠心病风险预测因子大多为未知代谢物,其功能结构存在不确定性;(3)尽管我们采用迄今为止最大规模的mGWAS数据,后续研究仍然需要进一步扩大的样本量来为代谢物的遗传影响提供更准确的评估。
综上,我们采用了两样本孟德尔随机化的方法探究了486种血液代谢物与冠心病的因果关系。尽管没有发现这些血液代谢物与冠心病发生风险之间存在稳健的因果关系,但本研究中发现的潜在的冠心病风险预测因子仍为揭示遗传-暴露相互作用在冠心病发病机制中的作用提供了新的见解。
Biography
王子贤,在读硕士研究生,E-mail: zxwanghelloworld@foxmail.com
Funding Statement
国家重点研发计划(2017YFC0909301);国家自然科学基金(81872934,81673514);广东省重点领域研发计划(2019B020229003)
Supported by National Natural Science Foundation of China (81872934, 81673514)
Contributor Information
王 子贤 (Zixian WANG), Email: zxwanghelloworld@foxmail.com.
钟 诗龙 (Shilong ZHONG), Email: gdph_zhongsl@gd.gov.cn.
References
- 1.Virani SS, Alonso A, Benjamin EJ, et al. Heart disease and stroke statistics-2020 update: A report from the american heart association. Circulation. 2020;141(9):139–596. doi: 10.1161/CIR.0000000000000757. [Virani SS, Alonso A, Benjamin EJ, et al. Heart disease and stroke statistics-2020 update: A report from the american heart association [J]. Circulation, 2020, 141(9):139-596.] [DOI] [PubMed] [Google Scholar]
- 2.Blaha MJ, Whelton SP, Al Rifai M, et al. Comparing risk scores in the prediction of coronary and cardiovascular deaths: Coronary artery calcium consortium. JACC Cardiovasc Imaging. 2020;S1936-878X(19):31180–5. doi: 10.1016/j.jcmg.2019.12.010. [Blaha MJ, Whelton SP, Al Rifai M, et al. Comparing risk scores in the prediction of coronary and cardiovascular deaths: Coronary artery calcium consortium[J]. JACC Cardiovasc Imaging, 2020, S1936-878X(19): 31180-5.] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Hilvo M, Meikle PJ, Pedersen ER, et al. Development and validation of a ceramide-and phospholipid-based cardiovascular risk estimation score for coronary artery disease patients. Eur Heart J. 2020;41(3):371–80. doi: 10.1093/eurheartj/ehz387. [Hilvo M, Meikle PJ, Pedersen ER, et al. Development and validation of a ceramide-and phospholipid-based cardiovascular risk estimation score for coronary artery disease patients[J]. Eur Heart J, 2020, 41(3): 371-80.] [DOI] [PubMed] [Google Scholar]
- 4.Maron DJ, Hochman JS, Reynolds HR, et al. Initial invasive or conservative strategy for stable coronary disease. N Engl J Med. 2020;382(15):1395–407. doi: 10.1056/NEJMoa1915922. [Maron DJ, Hochman JS, Reynolds HR, et al. Initial invasive or conservative strategy for stable coronary disease[J]. N Engl J Med, 2020, 382(15): 1395-407.] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Tremblay BL, Guénard F, Lamarche B, et al. Familial resemblances in human plasma metabolites are attributable to both genetic and common environmental effects. Nutr Res. 2019;61:22–30. doi: 10.1016/j.nutres.2018.10.003. [Tremblay BL, Guénard F, Lamarche B, et al. Familial resemblances in human plasma metabolites are attributable to both genetic and common environmental effects[J]. Nutr Res, 2019, 61: 22-30.] [DOI] [PubMed] [Google Scholar]
- 6.Wishart DS. Metabolomics for investigating physiological and pathophysiological processes. Physiol Rev. 2019;99(4):1819–75. doi: 10.1152/physrev.00035.2018. [Wishart DS. Metabolomics for investigating physiological and pathophysiological processes[J]. Physiol Rev, 2019, 99(4): 1819-75.] [DOI] [PubMed] [Google Scholar]
- 7.Zhang L, Wei TT, Li Y, et al. Functional metabolomics characterizes a key role for N-acetylneuraminic acid in coronary artery diseases. Circulation. 2018;137(13):1374–90. doi: 10.1161/CIRCULATIONAHA.117.031139. [Zhang L, Wei TT, Li Y, et al. Functional metabolomics characterizes a key role for N-acetylneuraminic acid in coronary artery diseases [J]. Circulation, 2018, 137(13): 1374-90.] [DOI] [PubMed] [Google Scholar]
- 8.Marklund M, Wu JHY, Imamura F, et al. Biomarkers of dietary omega-6 fatty acids and incident cardiovascular disease and mortality. Circulation. 2019;139(21):2422–36. doi: 10.1161/CIRCULATIONAHA.118.038908. [Marklund M, Wu JHY, Imamura F, et al. Biomarkers of dietary omega-6 fatty acids and incident cardiovascular disease and mortality[J]. Circulation, 2019, 139(21):2422-36.] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Davies NM, Holmes MV, Davey Smith G. Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians. BMJ Clin Res Ed. 2018;362:k601. doi: 10.1136/bmj.k601. [Davies NM, Holmes MV, Davey Smith G. Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians [J]. BMJ Clin Res Ed, 2018, 362: k601.] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.李 云霞, 李 洪凯, 马 韫韬, et al. 基于两样本孟德尔随机化的身高和冠心病风险之间因果关系. https://www.cnki.com.cn/Article/CJFDTOTAL-SDYB202005017.htm. 山东大学学报:医学版. 2020;58(5):107–14. [李云霞, 李洪凯, 马韫韬, 等.基于两样本孟德尔随机化的身高和冠心病风险之间因果关系[J].山东大学学报:医学版, 2020, 58(5): 107-14.] [Google Scholar]
- 11.刘 新辉, 李 洪凯, 李 明卓, et al. 腰围和冠心病因果关系的孟德尔随机化研究. https://www.cnki.com.cn/Article/CJFDTOTAL-SDYB201911018.htm. 山东大学学报:医学版. 2019;57(11):103–9. [刘新辉, 李洪凯, 李明卓, 等.腰围和冠心病因果关系的孟德尔随机化研究[J].山东大学学报:医学版, 2019, 57(11): 103-9.] [Google Scholar]
- 12.袁 同慧, 李 洪凯, 马 韫韬, et al. 臀围与冠心病的因果推断:孟德尔随机化分析. https://www.cnki.com.cn/Article/CJFDTOTAL-WSYJ202003005.htm. 卫生研究. 2020;49(3):362–7. doi: 10.19813/j.cnki.weishengyanjiu.2020.03.003. [袁同慧, 李洪凯, 马韫韬, 等.臀围与冠心病的因果推断:孟德尔随机化分析[J].卫生研究, 2020, 49(3): 362-7.] [DOI] [PubMed] [Google Scholar]
- 13.Adamski J. Genome-wide association studies with metabolomics. Genome Med. 2012;4(4):34. doi: 10.1186/gm333. [Adamski J. Genome-wide association studies with metabolomics[J]. Genome Med, 2012, 4(4): 34.] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Shin SY, Fauman EB, Petersen AK, et al. An atlas of genetic influences on human blood metabolites. Nat Genet. 2014;46(6):543–50. doi: 10.1038/ng.2982. [Shin SY, Fauman EB, Petersen AK, et al. An atlas of genetic influences on human blood metabolites[J]. Nat Genet, 2014, 46(6): 543-50.] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Yang J, Yan B, Zhao B, et al. Assessing the causal effects of human serum metabolites on 5 major psychiatric disorders. Schizophr Bull: Bp. 2020;46(4):804–13. doi: 10.1093/schbul/sbz138. [Yang J, Yan B, Zhao B, et al. Assessing the causal effects of human serum metabolites on 5 major psychiatric disorders[J]. Schizophr Bull: Bp, 2020, 46(4): 804-13.] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Mehta NN. Large-scale association analysis identifies 13 new susceptibility loci for coronary artery disease. Circ Cardiovasc Genet. 2011;4(3):327–9. doi: 10.1161/CIRCGENETICS.111.960443. [Mehta NN. Large-scale association analysis identifies 13 new susceptibility loci for coronary artery disease[J]. Circ Cardiovasc Genet, 2011, 4(3): 327-9.] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Hwang LD, Lawlor DA, Freathy RM, et al. Using a two-sample Mendelian randomization design to investigate a possible causal effect of maternal lipid concentrations on offspring birth weight. Int J Epidemiol. 2019;48(5):1457–67. doi: 10.1093/ije/dyz160. [Hwang LD, Lawlor DA, Freathy RM, et al. Using a two-sample Mendelian randomization design to investigate a possible causal effect of maternal lipid concentrations on offspring birth weight[J]. Int J Epidemiol, 2019, 48(5): 1457-67.] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Geng T, Ce S, Li C, et al. Childhood BMI and adult type 2 diabetes, coronary artery diseases, chronic kidney disease, and cardiometabolic traits: a Mendelian randomization analysis. Yearb Paediatr Endocrinol. 2018;41(5):1089–96. doi: 10.2337/dc17-2141. [Geng T, Ce S, Li C, et al. Childhood BMI and adult type 2 diabetes, coronary artery diseases, chronic kidney disease, and cardiometabolic traits: a Mendelian randomization analysis[J]. Yearb Paediatr Endocrinol, 2018, 41(5): 1089-96.] [DOI] [PubMed] [Google Scholar]
- 19.Emdin CA, Khera AV, Natarajan P, et al. Genetic association of waistto-hip ratio with cardiometabolic traits, type 2 diabetes, and coronary heart disease. JAMA. 2017;317(6):626–34. doi: 10.1001/jama.2016.21042. [Emdin CA, Khera AV, Natarajan P, et al. Genetic association of waistto-hip ratio with cardiometabolic traits, type 2 diabetes, and coronary heart disease[J]. JAMA, 2017, 317(6): 626-34.] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Holmes MV, Asselbergs FW, Palmer TM, et al. Mendelian randomization of blood lipids for coronary heart disease. Eur Heart J. 2015;36(9):539–50. doi: 10.1093/eurheartj/eht571. [Holmes MV, Asselbergs FW, Palmer TM, et al. Mendelian randomization of blood lipids for coronary heart disease[J]. Eur Heart J, 2015, 36(9): 539-50.] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Chang CC, Chow CC, Tellier LC, et al. Second-generation PLINK: rising to the challenge of larger and richer datasets. Gigascience. 2015;4(1) doi: 10.1186/s13742-015-0047-8. [Chang CC, Chow CC, Tellier LC, et al. Second-generation PLINK: rising to the challenge of larger and richer datasets[J]. Gigascience, 2015, 4(1):.] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Hemani G, Zheng J, Elsworth B, et al. The MR-Base platform supports systematic causal inference across the human phenome. Elife. 2018;7:e34408. doi: 10.7554/eLife.34408. [Hemani G, Zheng J, Elsworth B, et al. The MR-Base platform supports systematic causal inference across the human phenome[J]. Elife, 2018, 7:e34408.] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Burgess S, Butterworth A, Thompson SG. Mendelian randomization analysis with multiple genetic variants using summarized data. Genet Epidemiol. 2013;37(7):658–65. doi: 10.1002/gepi.21758. [Burgess S, Butterworth A, Thompson SG. Mendelian randomization analysis with multiple genetic variants using summarized data[J]. Genet Epidemiol, 2013, 37(7): 658-65.] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Bowden J, Davey Smith G, Burgess S. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. Int J Epidemiol. 2015;44(2):512–25. doi: 10.1093/ije/dyv080. [Bowden J, Davey Smith G, Burgess S. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression[J]. Int J Epidemiol, 2015, 44(2): 512-25.] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Bowden J, Davey Smith G, Haycock PC, et al. Consistent estimation in Mendelian randomization with some invalid instruments using a weighted Median estimator. Genet Epidemiol. 2016;40(4):304–14. doi: 10.1002/gepi.21965. [Bowden J, Davey Smith G, Haycock PC, et al. Consistent estimation in Mendelian randomization with some invalid instruments using a weighted Median estimator[J]. Genet Epidemiol, 2016, 40(4): 304-14.] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Hartwig FP, Davey Smith G, Bowden J. Robust inference in summary data Mendelian randomization via the zero modal pleiotropy assumption. Int J Epidemiol. 2017;46(6):1985–98. doi: 10.1093/ije/dyx102. [Hartwig FP, Davey Smith G, Bowden J. Robust inference in summary data Mendelian randomization via the zero modal pleiotropy assumption[J]. Int J Epidemiol, 2017, 46(6): 1985-98.] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Gunaruwan P, Maher A, Williams L, et al. Effects of bradykinin on venous capacitance in health and treated chronic heart failure. Clin Sci: Lond. 2009;116(5):443–50. doi: 10.1042/CS20080096. [Gunaruwan P, Maher A, Williams L, et al. Effects of bradykinin on venous capacitance in health and treated chronic heart failure[J]. Clin Sci: Lond, 2009, 116(5): 443-50.] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Palkhiwala SA, Frishman WH, Warshafsky S. Bradykinin for the treatment of cardiovascular disease. Heart Dis. 2001:333–9. doi: 10.1097/00132580-200109000-00010. [Palkhiwala SA, Frishman WH, Warshafsky S. Bradykinin for the treatment of cardiovascular disease[J]. Heart Dis, 2001: 333-9.] [DOI] [PubMed] [Google Scholar]
- 29.Ruiz-Canela M, Hruby A, Clish CB, et al. Comprehensive metabolomic profiling and incident cardiovascular disease: a systematic review. JAm HeartAssoc. 2017;6(10):e005705. doi: 10.1161/JAHA.117.005705. [Ruiz-Canela M, Hruby A, Clish CB, et al. Comprehensive metabolomic profiling and incident cardiovascular disease: a systematic review[J]. JAm HeartAssoc, 2017, 6(10): e005705.] [DOI] [PMC free article] [PubMed] [Google Scholar]
