Mitochondrial related genome-wide Mendelian randomization identifies putatively causal genes for multiple cancer types
线粒体相关全基因组孟德尔随机化可识别多种癌症类型的假定致病基因

Summary 总结

Introduction 介绍

Mitochondria are the essential organelles that regulate cellular energy production, metabolism, proliferation and apoptosis. An altered mitochondrial function is a well-known hallmark of cancer, which is commonly characterized by abnormal mitochondrial morphology, deficient mitochondrial copy numbers, aberrant energetic metabolism, accumulation of reactive oxygen species (ROS), imbalanced biogenesis and mitophagy.¹ A mild mitochondrial dysfunction may enhance the amplification and invasion of cancer cells while a severe level of dysfunction may cause cell death to inhibit tumorigenesis. Thus, understanding the roles of mitochondrial dysfunction is essential for cancer research. Mitochondrial dysfunction is a complex cellular process that exhibits a spectrum of pathological conditions although there is no specific biomarker/s to define mitochondrial dysfunction.^2,3 With the exception of the 37 critical bioenergetic genes encoded by the mitochondrion itself, the mitochondrial-related genome encompasses more than 1000 additional nuclear genes, and the genetic predisposition in those genes will potentially cause mitochondrial dysfunction.⁴ Many experimental and epidemiological studies have attempted to infer the causal relationship between mitochondrial dysfunction and cancer by exploring the selective mitochondrial DNA (mtDNA) and mitochondrial-related nuclear DNA mutations that affect mitochondrial function and are associated with the risk of specific cancer types.^5,6 However, the results generated from those studies are inconsistent and one of the reasons is the methodologies used in these studies, which do not consider the effect of confounders to differentiate between cause and consequence. Therefore, a comprehensive analysis of all genes related to mitochondrial dysfunction in multiple cancer types by a robust method is required to determine whether mitochondrial dysfunction per se is a cause or consequence of cancer.
线粒体是调节细胞能量产生、代谢、增殖和凋亡的重要细胞器。线粒体功能改变是癌症的一个众所周知的标志，其通常特征是线粒体形态异常、线粒体拷贝数不足、能量代谢异常、活性氧 （ROS） 积累、生物发生和线粒体自噬失衡。 ¹ 轻度线粒体功能障碍可能会增强癌细胞的扩增和侵袭，而严重的功能障碍可能会导致细胞死亡以抑制肿瘤发生。因此，了解线粒体功能障碍的作用对于癌症研究至关重要。线粒体功能障碍是一个复杂的细胞过程，尽管没有特定的生物标志物来定义线粒体功能障碍，但它表现出一系列病理状况。 ^{2 3} 除了线粒体本身编码的 37 个关键生物能量基因外，线粒体相关基因组包含 1000 多个额外的核基因，这些基因的遗传易感性可能会导致线粒体功能障碍。 ⁴ 许多实验和流行病学研究试图通过探索影响线粒体功能的选择性线粒体 DNA （mtDNA） 和线粒体相关核 DNA 突变来推断线粒体功能障碍与癌症之间的因果关系，这些突变与特定癌症类型的风险相关。 ^{5 6} 然而，这些研究产生的结果不一致，原因之一是这些研究中使用的方法，这些方法没有考虑混杂因素对区分原因和结果的影响。 因此，需要通过稳健的方法对多种癌症类型中与线粒体功能障碍相关的所有基因进行全面分析，以确定线粒体功能障碍本身是癌症的原因还是后果。

Mendelian randomization (MR) is a method that uses genetic variants as instrumental variables (IVs) to explore the potential causal association between lifetime exposure and outcome. In MR, the use of the conceptional random allocation of alleles avoids bias from unobserved confounders such as lifestyle and environmental factors and the problem with reverse causality.⁷ The two-sample MR allows for the assessment of the IVs-exposure association and IVs-outcome association generated from different populations.⁸ Genome-wide association studies (GWAS) exploit the genetic associations with traits based on single nucleotide polymorphisms (SNPs) and integration of the GWAS data with gene expression and methylation GWAS have allowed for the identification of expression or methylation quantitative trait loci (eQTL or mQTL).^9,10 A summary-data-based MR (SMR) has extended and developed the conception of MR that can utilize the independent GWAS summary statistics data and QTL data to prioritize potential causal genes from hits identified in GWAS.⁹ By applying this method followed by a heterogeneity independent instruments (HEIDI) test, the potential causal associations were distinguished from the widespread linkage disequilibrium (LD) in the genome.
孟德尔随机化 （MR） 是一种使用遗传变异作为工具变量 （IV） 来探索终生暴露与结果之间潜在因果关系的方法。在MR中，使用等位基因的概念随机分配可以避免来自未观察到的混杂因素的偏差，例如生活方式和环境因素以及反向因果关系问题。 ⁷ 双样本 MR 允许评估来自不同人群的 IVs 暴露关联和 IV s-结果关联。 ⁸ 全基因组关联研究 （GWAS） 利用基于单核苷酸多态性 （SNP） 的性状的遗传关联，并将 GWAS 数据与基因表达和甲基化 GWAS 整合，从而可以鉴定表达或甲基化数量性状位点（eQTL 或 mQTL）。 ^{9 10} 基于汇总数据的 MR （SMR） 扩展并发展了 MR 的概念，该概念可以利用独立的 GWAS 汇总统计数据和 QTL 数据来确定 GWAS 中确定的命中的潜在致病基因的优先级。 ⁹ 通过应用这种方法，然后进行异质性独立仪器 （HEIDI） 检验，将潜在的因果关联与基因组中广泛的连锁不平衡 （LD） 区分开来。

To our knowledge, there has been no MR study investigating the potential causal relationship between mitochondrial dysfunction and the risk of common types of cancer. Therefore, in this study, we aimed to investigate the causal relationship between mitochondrial dysfunction characterized by genetic predisposition in mitochondrial-related genes and multiple cancer types by the comprehensive two-sample MR analysis.
据我们所知，目前还没有MR研究调查线粒体功能障碍与常见类型癌症风险之间的潜在因果关系。因此，本研究旨在通过全面的双样本MR分析，探讨以线粒体相关基因遗传易感性为特征的线粒体功能障碍与多种癌症类型之间的因果关系。

Methods 方法

This study was conducted following the reporting guideline of the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE, Supplementary STROBE-MR checklist table).¹¹
本研究是按照加强流行病学观察性研究报告的报告指南（STROBE，补充 STROBE-MR 清单表）进行。 ¹¹

Study design 研究设计

Fig. 1 summarizes the design of the present study and the workflow of the selection of genetic variants and analytical methods. To determine the mitochondrial dysfunction characterized by the genetic predisposition in the mitochondrial-related genome constituting from both mitochondrion and nuclear, we extracted the inventory of 1136 known mitochondrial-related genes from the human MitoCarta3.0 database.⁴
图1总结了本研究的设计以及遗传变异选择和分析方法的工作流程。为了确定线粒体功能障碍的特征，以线粒体和核构成的线粒体相关基因组的遗传易感性为特征，我们从人类 MitoCarta3.0 数据库中提取了 1136 个已知线粒体相关基因的库存。 ⁴

Open in a separate window

Fig. 1 图1

Flowchart of the analyses performed.
所执行分析的流程图。

To generate eQTL instruments for mitochondrial genes, genetic variants located within 1000 kb on either side of the coding sequence (in cis) that are robustly associated with gene expression were extracted using eQTLs summary statistics obtained from the eQTLGen Consortium (https://www.eqtlgen.org/cis-eqtls.html). The eQTLGen Consortium contains information on 10,317 trait-associated single nucleotide polymorphisms (SNPs) from 31,684 individuals.¹² However, the eQTLGen did not include variants associated with the expression level of genes located on the X and Y chromosomes and mtDNA. From cis-eQTL, 662,968 SNPs associated with the expression of 1013 mitochondrial-related transcripts were selected. MR cis-mQTL instruments for genetic variants robustly associated with mitochondrial gene methylation were extracted using summary data from a meta-analysis of two cohorts (n = 1980).¹⁰ In total, 931,304 SNPs were selected corresponding to 2550 mitochondrial-related DNA methylation CpG sites. MR cis-pQTL instruments for genetic variants associated with the expression of mitochondrial-related proteins were selected from five proteome datasets,13, 14, 15, 16, 17 and 23 SNPs that were robustly associated with 23 mitochondrial-related protein expressions were selected. All SNPs included in the initial analysis had at least a suggestive P_snp-mitodys <5 × 10⁻⁸.
为了生成线粒体基因的 eQTL 工具，使用从 eQTLGen 联盟 （https://www.eqtlgen.org/cis-eqtls.html） 获得的 eQTL 汇总统计提取位于编码序列两侧（顺式）两侧 1000 kb 以内且与基因表达密切相关的遗传变异。eQTLGen 联盟包含来自 31,684 个个体的 10,317 个性状相关单核苷酸多态性 （SNP） 的信息。 ¹² 然而，eQTLGen不包括与位于X和Y染色体和mtDNA上的基因表达水平相关的变异。从cis-eQTL中，筛选出662,968个与1013个线粒体相关转录本表达相关的SNP。使用来自两个队列 （n = 1980） 的荟萃分析的汇总数据提取与线粒体基因甲基化密切相关的遗传变异的 MR cis-mQTL 工具。 ¹⁰ 总共选择了 931,304 个 SNP，对应于 2550 个线粒体相关 DNA 甲基化 CpG 位点。从5个蛋白质组数据集中选取与线粒体相关蛋白表达相关的遗传变异的MR顺式-pQTL仪器，选择13、14、15、16、17和23个SNPs，这些SNPs与23个线粒体相关蛋白表达密切相关。初始分析中包含的所有 SNP 都至少具有提示性 _snp-mitodys P <5 × 10 ⁻⁸ 。

GWAS summary statistics for cancer outcomes were obtained from publicly available databases. A total of 18 types of cancers were included. The details of all QTL and GWAS datasets for this study are presented in Supplementary Table S1 and Supplementary methods.
GWAS癌症结局的汇总统计数据来自公开可用的数据库。共纳入18种癌症。本研究的所有QTL和GWAS数据集的详细信息在补充表S1和补充方法中列出。

Results 结果

MR analysis of mitochondrial genome-wide cis-eQTLs and cancer outcomes
线粒体全基因组顺式-eQTLs和癌症结局的MR分析

After SMR testing, the associations of 662,968 SNPs from blood representing mitochondrial-related gene expression and cancer outcomes were obtained (Fig. 1). To control the genome-wide type I error, we performed multiple testing corrections, with the results showing strong evidence of an association (P_SMR<4.936 × 10⁻⁵ [Bonferroni correction, P < 0.05/1013]) followed by the HEIDI test (P_HEIDI>0.01) implemented in SMR software to investigate if the association was due to a shared causal variant and not pleiotropy. We thus identified 7 association signals across 7 unique genetic loci for breast cancer, 4 association signals across 4 unique genetic loci for prostate cancer and one association signal for gastric cancer. We found no significant genetic correlation for the other cancer types. Sensitivity analysis using additional MR methods relying on similar assumptions was conducted and shown to support our findings (Supplementary Table S2, Supplementary Figure S1 and S2). We further performed colocalization analysis to rule out confounding by LD; strong evidence of colocalization between cancer GWAS and eQTL exists if the posterior probability of shared causal variant across gene expression and cancer (PP.H4) is >0.80. The causal estimates are expressed as β coefficients, and the odds ratios (OR) for 1 standard deviation (SD) change in mitochondrial gene expression level was calculated by the expectation of the β coefficient, as presented in Fig. 2 and Supplementary Table S3.
经过SMR测试，获得了代表线粒体相关基因表达和癌症结果的血液中662,968个SNP的关联（图1）。为了控制全基因组的I型错误，我们进行了多次测试校正，结果显示了关联的有力证据（P _SMR <4.936 × 10 ⁻⁵ [Bonferroni校正，P < 0.05/1013]），然后在SMR软件中实施HEIDI检验（P _HEIDI >0.01），以调查这种关联是否是由于共同的因果变异而不是多效性。因此，我们在乳腺癌的 7 个独特遗传位点上鉴定了 7 个关联信号，在前列腺癌的 4 个独特遗传位点上鉴定了 4 个关联信号，在胃癌的 1 个关联信号中鉴定了 1 个关联信号。我们发现其他癌症类型没有显着的遗传相关性。使用依赖于类似假设的其他 MR 方法进行了敏感性分析，并显示支持我们的发现（补充表 S2、补充图 S1 和 S2）。我们进一步进行了共定位分析，以排除LD的混杂因素;如果基因表达和癌症之间共享因果变异的后验概率 （PP.H4）为>0.80。因果估计值表示为β系数，线粒体基因表达水平 1 个标准差 （SD） 变化的比值比 （OR） 通过β系数的期望值计算，如图 2 和补充表 S3 所示。

Open in a separate window

Fig. 2 图2

Mendelian randomization results for the association between the expression of mitochondrial genes and cancer risk.^aRepresents the effect size (β) of a variant on mRNA expressions. β > 0 means positive association, and β < 0 means negative association. ^bOdd ratios were calculated by the expectation of causal estimate (β coefficient). ^c‘Colocalization’ indicates PP.H4 between eQTLs and cancer outcomes. PP.H4 > 0.8 is the well-applied cut-off for the evidence of colocalization.
孟德尔随机化结果表明线粒体基因表达与癌症风险之间的关联。 ^a 表示变体对 mRNA 表达的影响大小 （β）。β > 0 表示正关联，β < 0 表示负关联。 ^b 奇数比是通过因果估计的期望值（β系数）计算的。 ^c “共定位”表示 PP。eQTL 与癌症结果之间的 H4。PP的。H4 > 0.8 是共定位证据的良好应用临界值。

For breast cancer, one SD decrease of FDPS expression was associated with 34% lower risk (OR: 0.66, 95% CI: 0.49–0.83, P_SMR = 9.77 × 10⁻⁷) while 1 SD increase of NSUN4 expression was associated with 5% higher risk (OR: 1.05, 95% CI: 1.03–1.07, P_SMR = 5.24 × 10⁻⁶). When sub-grouping breast cancer according to the intrinsic molecular subtypes, the causal associations showed a similar trend only with luminal A-like breast cancer. Interestingly, we found a robust causal association (OR per SD, 1.17; 95% CI: 1.12–1.23, P_SMR = 1.85 × 10⁻⁸) between MTX1 expression and luminal A-like breast cancer, specifically. We also found a strong causal association (OR per SD, 1.26; 95% CI: 1.15–1.37, P_SMR = 2.94 × 10⁻⁵) between COX11 expression and luminal B-like/HER2-negative cancer (Supplementary Table S4, Supplementary Fig. S3). For prostate cancer, one SD increase of NSUN4 expression was associated with 6% higher risk of cancer (OR: 1.06, 95% CI: 1.03–1.09, P_SMR = 1.01 × 10⁻⁵).
对于乳腺癌，FDPS 表达降低 1 SD 与风险降低 34% 相关 （OR： 0.66， 95% CI： 0.49–0.83， P _SMR = 9.77 × 10 ⁻⁷ ），而 NSUN4 表达增加 1 SD 与风险增加 5% 相关 （OR： 1.05， 95% CI： 1.03–1.07， P _SMR = 5.24 × 10 ⁻⁶ ）。当根据内在分子亚型对乳腺癌进行亚组时，因果关系仅与管腔 A 样乳腺癌表现出相似的趋势。有趣的是，我们发现MTX1表达与管腔A样乳腺癌之间存在强大的因果关系（OR标准差，1.17;95%CI：1.12–1.23，P _SMR =1.85×10 ⁻⁸ ）。我们还发现COX11表达与管腔B样/HER2阴性癌症之间存在很强的因果关系（OR per SD，1.26;95%CI：1.15–1.37，P _SMR = 2.94 × 10 ⁻⁵ ）（补充表S4，补充图S3）。对于前列腺癌，NSUN4 表达的 SD 增加与癌症风险增加 6% 相关 （OR： 1.06， 95% CI： 1.03–1.09， P _SMR = 1.01 × 10 ⁻⁵ ）。

Most importantly, our results show that the expression level of NSUN4 increased by rs41293273 is associated with higher risk of both breast cancer and prostate cancer.
最重要的是，我们的研究结果表明，NSUN4 的表达水平增加 rs41293273 与乳腺癌和前列腺癌的风险增加有关。

MR analysis of mitochondrial genome-wide cis-mQTLs and cancer outcomes
线粒体全基因组顺式-mQTLs和癌症结局的MR分析

For the causal association between the DNA methylation of the mitochondrial-related genome and cancer outcomes, Bonferroni correction (P_SMR < 1.961 × 10⁻⁵) and HEIDI test were performed. We identified a total of 15 association signals across 14 unique genetic loci for breast cancer, 11 association signals across 10 unique genetic loci for prostate cancer, one association signal for gastric cancer, 4 association signals across 3 unique genetic loci for lung cancer and 2 association signals across one unique genetic locus for melanoma (Fig. 3 and Supplementary Table S5). The sensitivity analysis supported the same associations (Supplementary Table S6, Supplementary Fig. S4 and S5).
对于线粒体相关基因组的DNA甲基化与癌症结局之间的因果关系，进行了Bonferroni校正（P _SMR < 1.961 × 10 ⁻⁵ ）和HEIDI检验。我们在乳腺癌的 14 个独特遗传位点中共鉴定了 15 个关联信号，在前列腺癌的 10 个独特遗传位点中鉴定了 11 个关联信号，在胃癌的 1 个关联位点中鉴定了 1 个关联信号，在肺癌的 3 个独特遗传位点中鉴定了 4 个关联信号，在黑色素瘤的一个独特遗传位点上鉴定了 2 个关联信号（图 3 和补充表 S5）。敏感性分析支持相同的关联（补充表S6，补充图S4和S5）。

Open in a separate window

Fig. 3 图3

Mendelian randomization results for the association between mitochondrial gene methylations and cancer risk.^aRepresents the effect size (β) of a variant on DNA methylation. β > 0 means positive association, and β < 0 means negative association. ^bOdd ratios were calculated by the expectation of causal estimate (β coefficient). ^c‘Colocalization’ indicates PP.H4 between mQTLs and cancer outcomes. PP.H4 > 0.8 is the well-applied cut-off for the evidence of colocalization.
线粒体基因甲基化与癌症风险之间关联的孟德尔随机化结果。 ^a 表示变体对 DNA 甲基化的影响大小 （β）。β > 0 表示正关联，β < 0 表示负关联。 ^b 奇数比是通过因果估计的期望值（β系数）计算的。 ^c “共定位”表示 PP。mQTL 之间的 H4 与癌症结果。PP的。H4 > 0.8 是共定位证据的良好应用临界值。

The colocalization analysis showed that different genetic variants regulating NSUN4 had different effects on methylation levels, hence the outcome. For example, one SD decrease of NSUN4 methylation by rs6682266 was associated with 9% lower risk of breast cancer (OR: 0.91, 95% CI: 0.87–0.96, P_SMR = 1.50 × 10⁻⁵), and conversely, one SD increase of NSUN4 methylation by rs6681857 was associated with 7% higher risk of breast cancer (OR: 1.07, 95% CI: 1.04–1.10, P_SMR = 5.81 × 10⁻⁶). Here, in total, we found 6 unique loci that regulated the methylation level of 7 different CpG sites in NUSU4, and were positively associated with the risk of breast cancer (Fig. 3). Analysis on breast cancer molecular subtypes with NSUN4 methylation showed a similar causal association but only for luminal A-like breast cancer (Supplementary Table S7 and Supplementary Fig. S6). Out of these 6 loci, two were also positively associated with risk of prostate cancer. Furthermore, an increase of 1 SD of NUDT5 methylation was associated with 4% lower risk of prostate cancer (OR: 0.96, 95% CI: 0.95–0.98, P_SMR = 1.15 × 10⁻⁵) (Fig. 3). For lung cancer, one SD increase of VARS2 methylation was associated with more than 20% higher risk of cancer depending on different methylation CpG sites (Fig. 3). We found no significant causal associations between mitochondrial genome methylation and other cancer risks.
共定位分析表明，调节NSUN4的不同遗传变异对甲基化水平的影响不同，因此结果不同。例如，rs6682266 将 NSUN4 甲基化降低 1 个 SD 与乳腺癌风险降低 9% 相关 （OR： 0.91， 95% CI： 0.87–0.96， P _SMR = 1.50 × 10 ⁻⁵ ），相反，rs6681857 使 NSUN4 甲基化 SD 增加一个 SD 与乳腺癌风险增加 7% 相关 （OR： 1.07， 95% CI： 1.04–1.10， P _SMR = 5.81 × 10 ⁻⁶ ).在这里，我们总共发现了 6 个独特的位点，它们调节 NUSU4 中 7 个不同 CpG 位点的甲基化水平，并且与乳腺癌风险呈正相关（图 3）。对具有 NSUN4 甲基化的乳腺癌分子亚型的分析显示出类似的因果关系，但仅适用于管腔 A 样乳腺癌（补充表 S7 和补充图 S6）。在这 6 个位点中，有 2 个位点也与前列腺癌风险呈正相关。此外，NUDT5 甲基化增加 1 SD 与前列腺癌风险降低 4% 相关 （OR： 0.96， 95% CI： 0.95–0.98， P _SMR = 1.15 × 10 ⁻⁵ ） （图 3）。对于肺癌，根据不同的甲基化 CpG 位点，VARS2 甲基化的一个 SD 增加与癌症风险增加 20% 以上相关（图 3）。我们发现线粒体基因组甲基化与其他癌症风险之间没有显着的因果关系。

Furthermore, gene methylation is known to influence gene expression. Here, we also performed SMR analysis on the causal association between mitochondrial-related gene methylation and expression by mapping the gene methylation to expression through shared genetic variants. After multiple testing corrections and the HEIDI test, we obtained the gene list for the mitochondrial gene expression regulated by DNA methylation CpG sites (Supplementary Table S8). For the putative causal genes that we identified above, SMR results showed that the methylation of NSUN4, which was regulated by rs6682266, rs5013329, rs56063031 and rs6681857, was associated with NSUN4 expression, and VARS2 methylation by rs2596495 was also associated with VARS2 expression (Supplementary Table S8).
此外，已知基因甲基化会影响基因表达。在这里，我们还通过将基因甲基化映射到通过共享遗传变异的表达，对线粒体相关基因甲基化和表达之间的因果关系进行了 SMR 分析。经过多次测试校正和HEIDI测试，我们获得了由DNA甲基化CpG位点调控的线粒体基因表达的基因列表（补充表S8）。对于我们上面确定的推定致病基因，SMR结果显示，受rs6682266、rs5013329、rs56063031和rs6681857调控的NSUN4的甲基化与NSUN4表达相关，rs2596495的VARS2甲基化也与VARS2表达相关（补充表S8）。

Discussion 讨论

In this study, we demonstrate that mitochondrial dysfunction characterized by genetic predisposition has causal effect on cancers, and identified important putative causal mitochondrial-related genes as follows: 1) FDPS for breast cancer; 2) NUDT5 for prostate cancer; 3) VARS2 for lung cancer and 4) NSUN4 for both breast and prostate cancers. Our results show that genetic determinants of mitochondrial dysfunction were associated with the risk of cancer in a cancer type-specific manner, which provides robust evidence for underlying mechanisms linking the genetic loci, gene expression, and methylation with multiple cancers.
在这项研究中，我们证明了以遗传易感性为特征的线粒体功能障碍对癌症具有因果作用，并确定了重要的推定因果线粒体相关基因如下：1）乳腺癌的FDPS;2）NUDT5用于前列腺癌;3） VARS2 用于肺癌，4） NSUN4 用于乳腺癌和前列腺癌。我们的结果表明，线粒体功能障碍的遗传决定因素以癌症类型特异性的方式与癌症风险相关，这为将遗传位点、基因表达和甲基化与多种癌症联系起来的潜在机制提供了强有力的证据。

The FDPS is a key enzyme that is involved in the mevalonate pathway to catalyze the biosynthesis of cholesterol and sterol, and to isoprenylate cellular metabolites such as Ras, Rac, Rab and Rho for membrane anchorage and cellular signaling.²² FDPS has been investigated over decades for its physiological function and was found to be associated with leukemia growth,²³ the progression of prostate cancer,²⁴ the poor breast cancer prognosis,²⁵ and directly involved in glioblastoma drug resistance²⁶ and pancreatic cancer radioresistance.²⁷ However, its causal relationship with cancer is unclear. In this study, we show that gene expression of FDPS has a causal relationship with breast cancer. Studies have shown that knockdown FDPS enhanced apoptosis and ectopic overexpression of FDPS promoted cancer colony growth and proliferation by affecting STAT3, AKT and ERK pathways.²⁴ Prenylation is important for exerting the activity of oncogenic proteins, thus prenylation inhibitors have been widely applied in clinical trials for cancer treatment.²⁸ FDPS was shown to be a key target of nitrogen-containing bisphosphonates and it is already a clinical drug target by Zoledronic acid.²⁹ According to the DRUGBANK database (https://go.drugbank.com/), more drugs have been investigated such as Ibandronate, Minodronic acid and Incadronic acid that target FDPS.
FDPS 是一种关键酶，参与甲羟戊酸途径，催化胆固醇和甾醇的生物合成，并异异戊二烯化细胞代谢物（如 Ras、Rac、Rab 和 Rho）以进行膜锚定和细胞信号传导。 ²² FDPS的生理功能已经研究了几十年，发现它与白血病的生长、 ²³ 前列腺癌的进展、 ²⁴ 乳腺癌预后不良有关， ²⁵ 并直接参与胶质母细胞瘤耐药 ²⁶ 性和胰腺癌放射耐药性。 ²⁷ 然而，它与癌症的因果关系尚不清楚。在这项研究中，我们发现FDPS的基因表达与乳腺癌有因果关系。研究表明，敲低FDPS通过影响STAT3、AKT和ERK通路来增强FDPS的细胞凋亡和异位过表达，促进癌症集落的生长和增殖。 ²⁴ 异戊二烯化对于发挥致癌蛋白的活性很重要，因此异戊二烯化抑制剂已广泛应用于癌症治疗的临床试验中。 ²⁸ FDPS被证明是含氮双膦酸盐的关键靶点，并且已经是唑来膦酸的临床药物靶点。 ²⁹ 根据DRUGBANK数据库（https://go.drugbank.com/），已经研究了更多的药物，如针对FDPS的伊班膦酸、米诺膦酸和耷加膦酸。

NSUN4 is an rRNA m⁵C methyltransferase that can induce the methylation of the 12S rRNA of the small ribosomal subunit joining in mitochondria and promote rRNA rearrangements to form peptidyl transferase center.^30,31 A previous study showed that breast cancer and prostate cancer shared a common risk locus (rs5013329) and indicated that NSUN4 is the strongest shared functional candidate at 1p34.³² However, our results showed that rs5013329 related to NSUN4 methylation was associated with decreased risk of breast cancer only. Importantly, we identified additional 1 genetic locus related to NSUN4 expression and 2 loci related to NSUN4 methylation that were causally associated with both breast and prostate cancer (Supplementary Fig. S7). Furthermore, our phenome-wide scan analysis showed that the causal relationship between NSUN4 and both breast and prostate cancers was not caused by horizontal pleiotropy. Together, these results emphasize the potentially important role of NSUN4 in carcinogenesis.
NSUN4 是一种 rRNA m ⁵ C 甲基转移酶，可诱导连接线粒体的小核糖体亚基的 12S rRNA 甲基化，并促进 rRNA 重排形成肽基转移酶中心。 ^{30 31} 之前的一项研究表明，乳腺癌和前列腺癌具有共同的风险位点 （rs5013329），并表明 NSUN4 是 1p34 最强的共享功能候选者。 ³² 然而，我们的结果表明，与 NSUN4 甲基化相关的 rs5013329 仅与乳腺癌风险降低相关。重要的是，我们确定了另外 1 个与 NSUN4 表达相关的遗传位点和 2 个与 NSUN4 甲基化相关的基因位点，它们与乳腺癌和前列腺癌有因果关系（补充图 S7）。此外，我们的全表型扫描分析表明，NSUN4与乳腺癌和前列腺癌之间的因果关系不是由水平多效性引起的。总之，这些结果强调了NSUN4在致癌作用中的潜在重要作用。

NUDT5 is differentially expressed in different types of cancer and positively correlated with aggressive cancer disease phenotype, knockdown of which can suppress the proliferation of cancer cells without inducing DNA oxidative lesion.³³ In our study, we show that the methylation level of NUDT5 has a strong causal effect on prostate cancer. Those findings highlight that NUDT5 may represent a promising drug target for cancer prevention and treatment. More studies now focus on the identification of NUDT5 inhibitors from approved drugs and small molecules, and a potent TH5427 was tested and shown to block hormone signaling and disrupt the proliferation of breast cancer cells.³⁴
NUDT5 在不同类型的癌症中差异表达，并与侵袭性癌症疾病表型呈正相关，敲低该表型可以抑制癌细胞的增殖而不诱导 DNA 氧化损伤。 ³³ 在我们的研究中，我们发现NUDT5的甲基化水平对前列腺癌有很强的因果关系。这些发现强调，NUDT5可能代表了癌症预防和治疗的有前途的药物靶点。现在更多的研究集中在从批准的药物和小分子中鉴定NUDT5抑制剂，并且测试了一种有效的TH5427，并显示可以阻断激素信号传导并破坏乳腺癌细胞的增殖。 ³⁴

Several mutations in VARS2 have been associated with mitochondrial diseases such as complex I defect, early onset of mitochondrial encephalomyopathies and encephalocardiomyopathies,³⁵ and cancer risks including breast cancer, colon and lung cancer.36, 37, 38 Here, we propose VARS2 as a causal gene only for lung cancer. The mechanisms of VARS2 in lung cancer carcinogenesis need further evaluation by experimental studies.
VARS2 中的几种突变与线粒体疾病有关，例如复合体 I 缺陷、线粒体脑肌病和脑心肌病的早期发作， ³⁵ 以及包括乳腺癌、结肠癌和肺癌在内的癌症风险。36， 37， 38 在这里，我们提出 VARS2 仅作为肺癌的致病基因。VARS2在肺癌发生中的机制需要通过实验研究进一步评估。

The main strength of the present study is that we performed a comprehensive MR analysis between mitochondrial dysfunction, characterized by genetic predisposition in all known mitochondrial-related genes, and their causal relationship with cancers. The inclusion of all genes related to mitochondria eliminates the selection bias in previous studies and might be able to address mitochondrial dysfunction directly. Secondly, we have included a very large sample size and 18 different cancer outcomes from GWAS summarized statistics, which allowed us to gain sufficient power to elucidate causal relationships and make conclusive estimations for several cancer types. Thirdly, we used SMR as the primary analysis and performed a sensitivity analysis using 5 additional MR approaches and colocalization analysis, which shows the robustness of our findings. Finally, we only included samples of European ancestry, thus, we minimized the biases caused by different genetic backgrounds.
本研究的主要优势在于，我们对线粒体功能障碍（以所有已知线粒体相关基因的遗传易感性为特征）及其与癌症的因果关系进行了全面的 MR 分析。包含所有与线粒体相关的基因消除了先前研究中的选择偏差，并且可能能够直接解决线粒体功能障碍。其次，我们从GWAS汇总统计数据中纳入了非常大的样本量和18种不同的癌症结果，这使我们能够获得足够的力量来阐明因果关系，并对几种癌症类型做出结论性估计。第三，我们使用SMR作为主要分析，并使用5种额外的MR方法和共定位分析进行了敏感性分析，这表明了我们研究结果的稳健性。最后，我们只纳入了欧洲血统的样本，因此，我们最大限度地减少了不同遗传背景引起的偏差。

This study has several limitations as well. Although we drew on the large available GWAS data sources, no genetic variants were obtained that represent the mitochondrial protein expression and the available eQTL and mQTL datasets did not have information on genetic variants that were associated with gene expression or methylation level in the X chromosome, Y chromosome and mitochondrial genome; the mitochondrial genome-wide associated genetic variants in this study mainly laid on the mitochondrial-related nuclear genome rather than the mitochondrial genome itself because a mitochondrial genome-specific QTL dataset has not yet been developed. Moreover, GWAS dataset that directly reflected on mitochondrial dysfunction is not available, hence we cannot assess the direction of causal relationship by using bi-directional MR based on current software resources. In this study, we showed that causal effects of mtDNA copy number variations and cancers were bi-directional in a cancer-specific manner. However, GWAS summary statistics of mtDNA copy number variation are potentially underpowered to detect the direction of causal association between mitochondrial dysfunction and cancers. Furthermore, univariable MR estimates the total effect of exposure on the outcome. As an extension, multivariable MR simultaneously estimates several potentially related exposures with a shared set of SNPs on the outcome using GWAS summary statistics, allowing for the assessment of the direct causal effect of a single exposure on the outcome. In this study, the exposure was mitochondrial dysfunction characterized by predisposition in the mitochondrial-related gene, which only can be retrieved from the QTL datasets other than the GWAS dataset. Thus, we are unable to perform multivariable MR to estimate the direct causal effects of mitochondrial dysfunction on cancers. Further studies should be conducted on the question of whether mitochondrial dysfunction is causally associated with cancer when GWAS or more advanced methods are available.
这项研究也有一些局限性。尽管我们利用了大量可用的GWAS数据源，但没有获得代表线粒体蛋白表达的遗传变异，并且可用的eQTL和mQTL数据集没有与X染色体、Y染色体和线粒体基因组中的基因表达或甲基化水平相关的遗传变异的信息;由于尚未开发线粒体基因组特异性QTL数据集，本研究中线粒体全基因组相关遗传变异主要基于线粒体相关核基因组，而不是线粒体基因组本身。此外，没有直接反映线粒体功能障碍的GWAS数据集，因此我们无法基于当前软件资源使用双向MR来评估因果关系的方向。在这项研究中，我们发现mtDNA拷贝数变异和癌症的因果效应是双向的。然而，mtDNA拷贝数变异的GWAS汇总统计可能不足以检测线粒体功能障碍与癌症之间的因果关系。此外，单变量 MR 估计暴露对结果的总体影响。作为扩展，多变量 MR 使用 GWAS 汇总统计同时估计具有一组共享 SNP 的多个潜在相关暴露对结果的影响，从而可以评估单次暴露对结果的直接因果影响。在这项研究中，暴露是线粒体功能障碍，其特征是线粒体相关基因的易感性，只能从GWAS数据集以外的QTL数据集中检索。 因此，我们无法进行多变量 MR 来估计线粒体功能障碍对癌症的直接因果影响。当GWAS或更先进的方法可用时，应进一步研究线粒体功能障碍是否与癌症有因果关系的问题。

This study leverages MR to examine the potential causal relationship between mitochondrial dysfunction characterized by genetic predisposition in mitochondrial-related genes and cancer, and demonstrates the importance of mitochondrial dysfunction in the pathogenesis of multiple cancer types. The identified putative genes can function as potential pharmacological targets for cancer treatment and prevention, further research could explore details of the underlying biological mechanisms.
本研究利用MR研究了以线粒体相关基因遗传易感性为特征的线粒体功能障碍与癌症之间的潜在因果关系，并证明了线粒体功能障碍在多种癌症类型发病机制中的重要性。鉴定的推定基因可以作为癌症治疗和预防的潜在药理学靶点，进一步的研究可以探索潜在生物学机制的细节。

Contributors 贡献

YL designed the study and performed the statistical analysis; YL, KS, NZ, XW and AAM participated in data interpretation; YL and AAM wrote the first draft, and KS, NZ, XW and JS revised the article. YL and NZ had direct access and responsibility for verifying all data reported in the manuscript. All authors read and approved the submitted version of the manuscript.
YL设计了这项研究并进行了统计分析;YL、KS、NZ、XW 和 AAM 参与了数据解释;YL 和 AAM 撰写了初稿，KS、NZ、XW 和 JS 修改了文章。YL和NZ可以直接访问并负责验证手稿中报告的所有数据。所有作者都阅读并批准了提交的稿件版本。

Data sharing statement 数据共享声明

Data used in this study are available from the referenced peer-reviewed studies and listed in Supplementary Table S1. Summary statistics for GWAS are publicly available for download. The statistical code needed to reproduce the results in the article is available upon request.
本研究中使用的数据来自参考的同行评审研究，并列在补充表S1中。GWAS的汇总统计数据可公开下载。可根据要求提供在文章中重现结果所需的统计代码。

Declaration of interests 利益申报

The authors declare that there is no conflict of interest associated with this manuscript.
作者声明，本手稿不存在任何利益冲突。

Acknowledgments 确认

Footnotes 脚注

References 引用