Advertisement

The association between mitochondrial DNA abundance and stroke: A combination of multivariable-adjusted survival and Mendelian randomization analyses

  • Leon G. Martens
    Correspondence
    Corresponding author. Department of Internal Medicine, Section of Gerontology and Geriatrics, Leiden University Medical Center, PO Box 9600, 2300 RC, Leiden, the Netherlands.
    Affiliations
    Department of Internal Medicine, Section of Gerontology and Geriatrics, Leiden University Medical Center, Leiden, the Netherlands
    Search for articles by this author
  • Jiao Luo
    Affiliations
    Department of Internal Medicine, Section of Gerontology and Geriatrics, Leiden University Medical Center, Leiden, the Netherlands

    Department of Clinical Epidemiology, Leiden University Medical Center, Leiden, the Netherlands
    Search for articles by this author
  • Marieke J.H. Wermer
    Affiliations
    Department of Neurology, Leiden University Medical Center, Leiden, the Netherlands
    Search for articles by this author
  • Ko Willems van Dijk
    Affiliations
    Department of Human Genetics, Leiden University Medical Center, Leiden, the Netherlands

    Department of Internal Medicine, Division of Endocrinology, Leiden University Medical Center, Leiden, the Netherlands

    Einthoven Laboratory for Experimental Vascular Medicine, Leiden University Medical Center, Leiden, the Netherlands
    Search for articles by this author
  • Sara Hägg
    Affiliations
    Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden
    Search for articles by this author
  • Felix Grassmann
    Affiliations
    Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden

    Institute of Medical Sciences, University of Aberdeen, Foresterhill, Aberdeen AB25 2ZD, UK
    Search for articles by this author
  • Raymond Noordam
    Affiliations
    Department of Internal Medicine, Section of Gerontology and Geriatrics, Leiden University Medical Center, Leiden, the Netherlands
    Search for articles by this author
  • Diana van Heemst
    Affiliations
    Department of Internal Medicine, Section of Gerontology and Geriatrics, Leiden University Medical Center, Leiden, the Netherlands
    Search for articles by this author
Open AccessPublished:June 19, 2022DOI:https://doi.org/10.1016/j.atherosclerosis.2022.06.1012

      Highlights

      • Mitochondrial DNA (mtDNA) abundance might be causally associated with stroke.
      • The association between mtDNA abundance and stroke attenuated after correction.
      • Lower mtDNA abundance showed higher incidence of ischemic stroke.

      Abstract

      Background and aims

      Mitochondrial dysfunction is associated with increased reactive oxygen species (ROS) that are thought to drive disease risk, including stroke. We investigated the association between mtDNA abundance, as a proxy measure of mitochondrial function, and incident stroke, using multivariable-adjusted survival and Mendelian Randomization (MR) analyses.

      Methods

      Cox-proportional hazard model analyses were conducted to assess the association between mtDNA abundance, and incident ischemic and hemorrhagic stroke over a maximum of 14-year follow-up in European-ancestry participants from UK Biobank. MR was conducted using independent (R2 < 0.001) lead variants for mtDNA abundance (p < 5 × 10-8) as instrumental variables. Single-nucleotide polymorphism (SNP)-ischemic stroke associations were derived from three published open source European-ancestry results databases (cases/controls): MEGASTROKE (60,341/454,450), UK Biobank (2404/368,771) and FinnGen (10,551/202,223). MR was performed per study, and results were subsequently meta-analyzed.

      Results

      In total, 288,572 unrelated participants (46% men) with mean (SD) age of 57 (8) years were included in the Cox-proportional hazard analyses. After correction for considered confounders (BMI, hypertension, cholesterol, T2D), no association was found between low versus high mtDNA abundance and ischemic (HR: 1.06 [95% CI: 0.95, 1.18]) or hemorrhagic (HR: 0.97 [95% CI: 0.82, 1.15]) stroke. However, in the MR analyses after removal of platelet count-associated SNPs, we found evidence for an association between genetically-influenced mtDNA abundance and ischemic stroke (odds ratio, 1.17; confidence interval, 1.03, 1.32).

      Conclusions

      Although the results from both multivariable-adjusted prospective and basis MR analyses did not show an association between low mtDNA and increased risk of ischemic stroke, in-depth MR sensitivity analyses may suggest evidence for a causal relationship.

      Graphical abstract

      Keywords

      1. Introduction

      Stroke is the second leading cause of death and loss of disability-adjusted life years worldwide [
      • Wang H.
      Global age-sex-specific fertility, mortality, healthy life expectancy (HALE), and population estimates in 204 countries and territories, 1950-2019: a comprehensive demographic analysis for the Global Burden of Disease Study 2019.
      ]. Oxidative stress has been hypothesized to play an important role in the pathophysiology of stroke by aggravating secondary damage and increases reperfusion injury after ischemic stroke [
      • Rodrigo R.
      • Fernandez-Gajardo R.
      • Gutierrez R.
      • Matamala J.M.
      • Carrasco R.
      • Miranda-Merchak A.
      • Feuerhake W.
      Oxidative stress and pathophysiology of ischemic stroke: novel therapeutic opportunities.
      ,
      • Allen C.L.
      • Bayraktutan U.
      Oxidative stress and its role in the pathogenesis of ischaemic stroke.
      ,
      • Duan X.
      • Wen Z.
      • Shen H.
      • Shen M.
      • Chen G.
      Intracerebral hemorrhage, oxidative stress, and antioxidant therapy.
      ]. As a result of direct or indirect reactive oxygen species (ROS)-induced damage to the (cerebral) vascular wall, multiple aspects of the vascular system are affected including platelet aggregation, endothelial function, vascular permeability and vasodilation [
      • Allen C.L.
      • Bayraktutan U.
      Oxidative stress and its role in the pathogenesis of ischaemic stroke.
      ]. These local vessel changes induced by oxidative stress can also gradually develop before stroke onset, and therefore may also lead to an increased risk of stroke incidence [
      • Shirley R.
      • Ord E.N.
      • Work L.M.
      Oxidative stress and the use of antioxidants in stroke.
      ].
      Mitochondria are a major source of ROS production [
      • Lackner L.L.
      • Nunnari J.M.
      The molecular mechanism and cellular functions of mitochondrial division.
      ]. Mitochondrial dysfunction leads to an increase in ROS production due to a change in redox homeostasis [
      • Bandy B.
      • Davison A.J.
      Mitochondrial mutations may increase oxidative stress: implications for carcinogenesis and aging?.
      ]. Additionally, impaired mitochondrial dysfunction, frequently proxied by the mitochondrial copy number (mtDNA-CN) [
      • Chung J.K.
      • Lee S.Y.
      • Park M.
      • Joo E.J.
      • Kim S.A.
      Investigation of mitochondrial DNA copy number in patients with major depressive disorder.
      ,
      • Longchamps R.J.
      • Castellani C.A.
      • Yang S.Y.
      • Newcomb C.E.
      • Sumpter J.A.
      • Lane J.
      • Grove M.L.
      • Guallar E.
      • Pankratz N.
      • Taylor K.D.
      • et al.
      Evaluation of mitochondrial DNA copy number estimation techniques.
      ], has been associated with diseases such as diabetes, heart failure, and neurological defects [
      • Yang J.L.
      • Mukda S.
      • Chen S.D.
      Diverse roles of mitochondria in ischemic stroke.
      ]. mtDNA-CN can be assessed relatively easily in large populations by estimating mtDNA abundance from the intensities of genotyping probes representing mitochondrial DNA on genotyping arrays [
      • Longchamps R.J.
      • Castellani C.A.
      • Yang S.Y.
      • Newcomb C.E.
      • Sumpter J.A.
      • Lane J.
      • Grove M.L.
      • Guallar E.
      • Pankratz N.
      • Taylor K.D.
      • et al.
      Evaluation of mitochondrial DNA copy number estimation techniques.
      ,
      • Hagg S.
      • Jylhava J.
      • Wang Y.
      • Czene K.
      • Grassmann F.
      Deciphering the genetic and epidemiological landscape of mitochondrial DNA abundance.
      ,
      • Chong M.
      • Mohammadi-Shemirani P.
      • Perrot N.
      • Nelson W.
      • Morton R.
      • Narula S.
      • Lali R.
      • Khan I.
      • Khan M.
      • Judge C.
      • et al.
      GWAS and ExWAS of blood Mitochondrial DNA copy number identifies 71 loci and highlights a potential causal role in dementia.
      ]. Increased ROS production drives mitochondrial dysfunction causing increased defects in mitochondrial fusion, fission, and mitophagy activation [
      • Liu F.
      • Lu J.
      • Manaenko A.
      • Tang J.
      • Hu Q.
      Mitochondria in ischemic stroke: new insight and implications.
      ], which subsequently lead to subsequent excessive ROS production [
      • Liu F.
      • Lu J.
      • Manaenko A.
      • Tang J.
      • Hu Q.
      Mitochondria in ischemic stroke: new insight and implications.
      ].
      Although a relatively small study was not able to provide evidence of an association between low mtDNA-CN and increased stroke risk [
      • Wachsmuth M.
      • Hubner A.
      • Li M.
      • Madea B.
      • Stoneking M.
      Age-related and heteroplasmy-related variation in human mtDNA copy number.
      ], we hypothesized that leukocyte mtDNA might affect brain pathologies, given the available biological data. Based on the combination of the postulated detrimental biological effect of blood oxidative stress on the (cerebro)vascular endothelial system and its role in secondary damage after stroke occurrence, we investigated the prospective association between mtDNA abundance and incident ischemic and hemorrhagic stroke in a large cohort of European-ancestry participants from the UK Biobank. In addition, we applied Mendelian Randomization (MR) to provide evidence for possible causality [
      • Lawlor D.A.
      • Tilling K.
      • Davey Smith G.
      Triangulation in aetiological epidemiology.
      ,
      • Burgess S.
      • Butterworth A.
      • Thompson S.G.
      Mendelian randomization analysis with multiple genetic variants using summarized data.
      ] as a way to triangulate the results from prospective analyses by obtaining results from two analysis methods, both with different assumptions and limitations [
      • Lawlor D.A.
      • Tilling K.
      • Davey Smith G.
      Triangulation in aetiological epidemiology.
      ].

      2. Materials and methods

      2.1 Population description

      The UK Biobank cohort is a prospective general population cohort with 502,628 participants between the age of 40 and 70 years recruited from the general population between 2006 and 2010 [
      • Sudlow C.
      • Gallacher J.
      • Allen N.
      • Beral V.
      • Burton P.
      • Danesh J.
      • Downey P.
      • Elliott P.
      • Green J.
      • Landray M.
      • et al.
      UK biobank: an open access resource for identifying the causes of a wide range of complex diseases of middle and old age.
      ] (more information can be found online https://www.ukbiobank.ac.uk). Blood samples were collected for genotyping. Access for information to invite participants was approved by the Patient Information Advisory Group (PIAG) from England and Wales. All participants in the UK Biobank provided a written informed consent and local research ethics committees and institutional review boards approved the study. The present study was accepted under project number 56340.
      In the present study, genotyped European-ancestry participants were followed (N = 488,377). Exclusion criteria included: 1) non-European ancestry; 2) participants who failed genotyping quality control and/or with low call rate; 3) related individuals defined by principal components (PCs); 4) participants with high SD of autosomal probes; 5) history of any stroke at baseline; 6) missing covariates. After exclusion, the final analyses were performed in 288,572 participants.

      2.2 Mitochondrial DNA abundance

      We used somatic mtDNA abundance as a proxy measure of mtDNA-CN, as the exposure, which is determined from the intensities of genotyping probes on the mitochondrial chromosome on the Affymetrix Array. The method for computing mtDNA abundance has been described previously [
      • Hagg S.
      • Jylhava J.
      • Wang Y.
      • Czene K.
      • Grassmann F.
      Deciphering the genetic and epidemiological landscape of mitochondrial DNA abundance.
      ]. In brief, the relative amount of mtDNA hybridized to the array at each probe was the log2 transformed ratio (L2R) of the observed genotyping probe intensity divided by the intensity at the same probe observed in a set of reference samples. We used the median L2R values across all 265 variants passing quality control on the MT chromosome as an initial raw measure of mtDNA abundance. To correct for confounding induced by poorly performing probes, we weighted L2R values of each probe by multiplying the weight of the probe generated from a multivariate linear regression model in which those intensities statistically significantly predicted normalized mitochondrial coverage from exome sequencing data, resulting in a single mtDNA abundance estimate for each individual. To eliminate the plate effect, we subsequently normalized the abundance to mean of zero and SD of one within each genotyping plate consisting of 96 wells [
      • Longchamps R.J.
      • Castellani C.A.
      • Yang S.Y.
      • Newcomb C.E.
      • Sumpter J.A.
      • Lane J.
      • Grove M.L.
      • Guallar E.
      • Pankratz N.
      • Taylor K.D.
      • et al.
      Evaluation of mitochondrial DNA copy number estimation techniques.
      ].

      2.3 Covariates

      In addition to age and sex, we took into account data based on self-reported questionnaires (smoking, alcohol consumption, disease status, medication use), blood cell counts (white blood cell counts and platelet counts), body mass index (BMI) in kg/m2, serum lipid levels (total and LDL cholesterol) in mmol/L, and systolic and diastolic blood pressure in mmHg.

      2.4 Outcome

      The outcome in the analysis was ischemic and hemorrhagic stroke separately, as well as combined, in the time period August 2006–January 2021. Stroke incidence was obtained via hospital admission data and national health register data and used to identify the date of the first stroke or stroke-related death after baseline assessment. The primary outcomes were any stroke incidence and further specified ischemic and hemorrhagic stroke incidence. Incident disease diagnoses are coded according to the International Classification of Diseases edition 10 (ICD-10). Ischemic stroke was defined as I63 and hemorrhagic stroke as I61. Any stroke was defined as the combination of I63 and I61. Follow-up time is computed from baseline visit to diagnosis of incident disease, loss-to-follow-up or death, or the end of the study period, whichever came first.

      2.5 Data required for Mendelian Randomization analyses

      For MR, genetic variants of mtDNA abundance were used as instrument variables. In a previous study, 129 independent single-nucleotide polymorphisms (SNPs) as genetic variants were found to be independently associated with mtDNA abundance at a genome-wide significance threshold (p < 5 × 10−8); SNPs were additionally pruned to an LD R2 < 0.0001 [
      • Longchamps R.J.
      • Yang S.Y.
      • Castellani C.A.
      • Shi W.
      • Lane J.
      • Grove M.L.
      • Bartz T.M.
      • Sarnowski C.
      • Liu C.
      • Burrows K.
      • et al.
      Genome-wide analysis of mitochondrial DNA copy number reveals loci implicated in nucleotide metabolism, platelet activation, and megakaryocyte proliferation.
      ]. The study was performed in a total of 465,809 individuals using a combined population of the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) consortium and the UK Biobank.

      2.6 Mendelian Randomization outcome datasets

      For the extraction of summary statistics on the associations of the mtDNA abundance related SNPs with ischemic stroke, which was defined as any ischemic stroke (I63), three large studies were used: the MEGASTROKE consortium, the UK Biobank, and the FinnGen study [
      • Sudlow C.
      • Gallacher J.
      • Allen N.
      • Beral V.
      • Burton P.
      • Danesh J.
      • Downey P.
      • Elliott P.
      • Green J.
      • Landray M.
      • et al.
      UK biobank: an open access resource for identifying the causes of a wide range of complex diseases of middle and old age.
      ,
      • Malik R.
      • Chauhan G.
      • Traylor M.
      • Sargurupremraj M.
      • Okada Y.
      • Mishra A.
      • Rutten-Jacobs L.
      • Giese A.K.
      • van der Laan S.W.
      • Gretarsdottir S.
      • et al.
      Multiancestry genome-wide association study of 520,000 subjects identifies 32 loci associated with stroke and stroke subtypes.
      ]. Both UK Biobank and FinnGen were not part of the main analyses of the MEGASTROKE consortium preventing inclusion of overlapping samples in the analyses. In the three studies insufficient data on hemorrhagic stroke was available.
      The trans-ancestry meta-analysis from the MEGASTROKE consortium was used to retrieve the ischemic stroke SNP-outcome data and was based on 60,341 cases and 454,450 controls collected from 29 studies of predominantly European ancestry (86%) [
      • Malik R.
      • Chauhan G.
      • Traylor M.
      • Sargurupremraj M.
      • Okada Y.
      • Mishra A.
      • Rutten-Jacobs L.
      • Giese A.K.
      • van der Laan S.W.
      • Gretarsdottir S.
      • et al.
      Multiancestry genome-wide association study of 520,000 subjects identifies 32 loci associated with stroke and stroke subtypes.
      ].
      For the MR analyses in UK Biobank, cases developed before and after enrolment were considered. Follow-up information that included ischemic stroke occurrence was retrieved through the routinely available NHS database. In the European-ancestry dataset with full genomics data available, we had data on 2404 cases of ischemic stroke and 368,771 controls. We performed new genome-wide association analyses using linear mixed models to assess the associations between genetic instruments and ischemic stroke, adjusted for age, sex and 10 principal components, and corrected for familial relationships using BOLT_LMM (v2.3.2).
      Data from FinnGen (Freeze 5; https://www.finngen.fi/en/), which is an ongoing cohort study launched in 2017, and analyses were based on 10,551 cases of ischemic stroke, and 202,223 controls.
      Although with lower numbers, we additionally performed MR on subtypes of ischemic stroke (Cardio Embolic Stroke: 7193 cases, 406,111 controls, Large Artery Atherosclerosis: 4373 cases, 406,111 controls, Small Vessel Stroke: 5386 cases, 192,662 controls) using data from MEGASTROKE and hemorrhagic stroke (1,687cases, 201,146 controls) from FinnGen.

      2.7 Statistical analysis

      2.7.1 Multivariable-adjusted analyses

      For the analyses and presentation purposes, we divided the study population in 5 equally-sized groups based on mtDNA abundance, with the first quintile containing the group with the lowest levels of mtDNA abundance and the fifth quintile containing the highest levels (used as reference).
      Baseline characteristics of the study population were presented separately per quintile of mtDNA abundance, as mean (SD) for continuous variables if they followed a normal distribution, or as median (Interquartile range) otherwise, and frequency (proportion) for categoric variables.
      The cumulative incidence for competing risk (CICR) was used to plot the cumulative incidence of ischemic and hemorrhagic stroke against follow-up time separately using a Kaplan-Meier survival curve by mtDNA abundance quintiles, where death was accounted for as a competing event. For any ischemic and hemorrhagic stroke, a Cox proportional hazards model was used to estimate the hazard ratio (HR) and 95% confidence interval (CI) presented as stroke incidence, comparing the lowest 20% mtDNA abundance with the highest 20%. Analyses were additionally done stratified by sex. Two multivariate regression models were fitted, where for model 2 covariates were first added individually:
      • -
        Model 1: age, sex, batch, the first 10 genetic principal components, white blood cell counts, platelet count
      • -
        Model 2: Model 1 + BMI, smoking, alcohol consumption, total cholesterol, hypertension, diabetes, cholesterol lowering medication, blood pressure lowering medication
      Covariates were included in regression models given their known relation with both exposure and outcome (age, sex, smoking, alcohol consumption, total cholesterol, disease status, medication status), or were included as a technical correction due to measurement composition (batch, white blood cell count, platelet count). Participants were censored in the event of loss-to-follow-up or death. To check whether the proportional hazards assumption was fulfilled, a Cox proportional hazard assumption test (“cox.zph” from R package “Survival”) was performed. Additionally, mtDNA-CN was assessed continuously as a one-SD lower mtDNA-CN on stroke incidence. Analyses were performed using the “Survival” (cran.r-project.org/web/packages/survival) package in R (v4.1.0)

      2.7.2 Mendelian Randomization

      All the analyses were done using R (v4.1.0) statistical software (The R Foundation for Statistical Computing, Vienna, Austria). MR analyses were performed using the R-based package “TwoSampleMR” (https://mrcieu.github.io/TwoSampleMR/) [
      • Hemani G.
      • Zheng J.
      • Elsworth B.
      • Wade K.H.
      • Haberland V.
      • Baird D.
      • Laurin C.
      • Burgess S.
      • Bowden J.
      • Langdon R.
      • et al.
      ].
      For our primary MR analysis, Inverse-Variance weighted (IVW) regression analyses were performed [
      • Burgess S.
      • Butterworth A.
      • Thompson S.G.
      Mendelian randomization analysis with multiple genetic variants using summarized data.
      ]. Estimates were calculated for each genetic instrument using the Wald ratio (SNP – outcome association divided by the SNP – exposure association) and subsequently meta-analyzed using the inverse-weighted meta-analyses weighted on the standard error of the SNP-outcome association (assuming no measurement error [NOME] in the exposure) [
      • Burgess S.
      • Thompson S.G.
      Interpreting findings from Mendelian randomization using the MR-Egger method.
      ]. The calculated estimates were expressed as odds ratios (OR) on ischemic stroke per SD (obtained from the exposure data) difference in mtDNA abundance.
      To ensure that the results obtained from the IVW analyses were not biased due to directional pleiotropy, we performed MR-Egger regression analysis and Weighted-Median Estimator [
      • Burgess S.
      • Thompson S.G.
      Interpreting findings from Mendelian randomization using the MR-Egger method.
      ]. Although MR-Egger is considered as a relatively inefficient approach (e.g., large confidence intervals), this method does not force the regression line to go through the intercept. The intercept depicts the estimated average pleiotropic effect across the genetic variants, and a value that differs from zero indicates that the IVW estimate is biased [
      • Bowden J.
      • Davey Smith G.
      • Burgess S.
      Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression.
      ]. The Weighted-Median estimator analysis can provide a consistent valid estimate if at least half of the instrumental variables are valid [
      • Bowden J.
      • Davey Smith G.
      • Haycock P.C.
      • Burgess S.
      Consistent estimation in mendelian randomization with some invalid instruments using a weighted median estimator.
      ]. In addition, MR-PRESSO (MR Pleiotropy RESidual Sum and Outlier) was applied to detect and correct for horizontal pleiotropy through removing outlying causal estimates based on individual instruments [
      • Verbanck M.
      • Chen C.Y.
      • Neale B.
      • Do R.
      Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases.
      ], as implemented in the R-based package “MR-PRESSO” (https://github.com/rondolab/MR-PRESSO). The Cochran's Q statistic was performed in order to test the heterogeneity between the estimated Wald ratios from different genetic variants [
      • Greco M.F.
      • Minelli C.
      • Sheehan N.A.
      • Thompson J.R.
      Detecting pleiotropy in Mendelian randomisation studies with summary data and a continuous outcome.
      ]. Additionally, a Steiger directionality test was performed to ensure consistent causal direction-of-effect. A power calculation was performed with mRnd (https://shiny.cnsgenomics.com/mRnd/) [
      • Brion M.J.
      • Shakhbazov K.
      • Visscher P.M.
      Calculating statistical power in Mendelian randomization studies.
      ]. With power = 0.80, minimal effect size (OR) was 1.076.
      Recent research has proven that two-sample MR methods can safely be used for one-sample MR in large databases [
      • Minelli C.
      • Del Greco M.F.
      • van der Plaat D.A.
      • Bowden J.
      • Sheehan N.A.
      • Thompson J.
      The use of two-sample methods for Mendelian randomization analyses on single large datasets.
      ]. This allows us to use the UK Biobank database in our sample set despite also being used as our exposure dataset. As a limitation to this method, results of MR-Egger analyses are to be interpreted with caution when used to check for pleiotropy [
      • Minelli C.
      • Del Greco M.F.
      • van der Plaat D.A.
      • Bowden J.
      • Sheehan N.A.
      • Thompson J.
      The use of two-sample methods for Mendelian randomization analyses on single large datasets.
      ].
      The main MR analyses were performed in the individual datasets, and subsequently meta-analyzed to derive the pooled estimates for the exposure on the risk of ischemic stroke using a fixed-effect model. Heterogeneity testing of the estimates across three datasets was performed by I2, and corresponding p-value was obtained from the Cochran's Q test. All meta-analyses were performed in the R-based “meta” package (https://cran.r-project.org/web/packages/meta/index.html).

      2.7.3 Sensitivity analysis after stratification of genetic instruments

      SNPs identified in relation to mtDNA-CN have been found in relation to platelet activation and megakaryocyte proliferation [
      • Longchamps R.J.
      • Yang S.Y.
      • Castellani C.A.
      • Shi W.
      • Lane J.
      • Grove M.L.
      • Bartz T.M.
      • Sarnowski C.
      • Liu C.
      • Burrows K.
      • et al.
      Genome-wide analysis of mitochondrial DNA copy number reveals loci implicated in nucleotide metabolism, platelet activation, and megakaryocyte proliferation.
      ], which both could affect stroke risk and could potentially lead to biased results, we first examined the associations between the SNPs and platelet count in our study sample (adjusted for age, sex, and the first 10 genetic principal components); all SNPs with p<(0.05/123) in its association with platelet count were excluded from further MR sensitivity analyses.

      3. Results

      3.1 Baseline characteristics of the study population

      A total of 288,572 participants were included in the final study sample (see full procedure in Supplementary Fig. 1) for multivariable-adjusted survival analyses. Participants with unavailable genetic data (N = 14,251), not used to compute the genetic principal components (N = 81,623), or having unrealistic SD of autosomal probes (N = 9440) were excluded according to standard UK Biobank quality control recommendations. Subsequently, we excluded related participants (N = 38,642), and those with a non-white British ancestry (N = 65,498). Finally, 4602 participants were excluded due to a history of stroke before study enrollment. Participants in the lower mtDNA abundance quintile (Table 1) had a mean age of 57.5 versus 56.1 year in the highest quintile, a mean BMI of 27.7 versus 27.0 kg/m2, T2D prevalence of 2.8% versus 2.0%, and 11.3% were current smokers compared with 8.4% in the highest quintile.
      Table 1Baseline characteristics of the study participants stratified by quintiles of mtDNA abundance.
      Q1Q2Q3Q4Q5
      N57,71557,71457,71457,71457,715
      mtDNA abundance (normalized)−1.4 (0.5)−0.5 (0.2)0.0 (0.1)0.5 (0.2)1.4 (0.5)
      Age (years)57.5 (8.0)57.1 (8.0)56.8 (8.0)56.5 (8.0)56.1 (8.0)
      Sex (female %)52.053.254.054.355.7
      BMI (kg/m2)27.7 (5.0)27.5 (4.8)27.4 (4.7)27.2 (4.6)27.0 (4.5)
      Diastolic blood pressure (mmHg)82.6 (10.2)82.4 (10.0)82.3 (10.0)82.1 (10.1)81.7 (10.1)
      Systolic blood pressure (mmHg)139.5 (18.8)138.6 (18.7)138.2 (18.6)137.6 (18.3)136.7 (18.3)
      White Blood Cell count (109 cells/L)7.4 (1.8)7.1 (1.7)6.9 (1.7)6.6 (1.7)6.4 (2.7)
      Platelet count (109 cells/L)245.5 (58.0)250.8 (57.7)253.4 (58.4)256.3 (59.2)259.5 (63.8)
      Blood pressure-lowering medication %
       Yes19.818.417.516.615.6
       No80.281.6(82.583.484.4
      Cholesterol (mmol/L)5.8 (1.2)5.7 (1.1)5.7 (1.1)5.7 (1.1)5.7 (1.1)
      Cholesterol lowering medication %
       Yes13.913.513.112.512.2
       No86.186.586.987.587.8
      Alcohol consumption %
       Less than once per week29.128.227.827.326.6
       Once or twice per week25.626.426.226.426.6
       More than four times per week45.345.245.946.346.7
      Smoking %
       Never53.354.155.055.656.5
       Past35.035.234.834.934.9
       Current11.310.49.99.48.4
      Type 2 diabetes %
       Yes2.82.52.32.22.0
       No97.297.597.797.898.0
      Data are mean (SD) for continuous variables or percentages for dichotomous variables. mtDNA abundance is presented as normalized in unit standard deviations. BMI, Body Mass Index.

      3.2 Multivariable-adjusted survival analyses mtDNA abundance and stroke

      A total of 6218 of the 288,572 participants (2.15%) had a stroke incidence, of which 3994 (1.38%) were ischemic and 1883 (0.65%) hemorrhagic over a median (IQR) follow-up of 11.8 (11.1–12.5) years. The incidence of ischemic stroke was higher in the lower mtDNA-CN quintiles than in the higher quintiles (Fig. 1A), while hemorrhagic stroke incidence was similar in all mtDNA-CN quintiles (Fig. 1B); in both cases the analyses fulfilled the proportional hazard assumption (p-value: 0.84 and 0.88).
      Fig. 1
      Fig. 1Cumulative incidence of ischemic (A) and hemorrhagic (B) stroke by quintiles of mtDNA abundance.
      We calculated the cumulative incidence for ischemic and hemorrhagic stroke, accounting for death as a competing event. Differences in cumulative incidence between groups were assessed using Gray's test.
      After stratification based on mtDNA-CN (Table 2), in model 1, mtDNA abundance was associated with any stroke and ischemic stroke incidence, when comparing the first quintile with the highest 20% mtDNA abundance (any stroke: hazard ratio (HR), 1.11; 95% confidence interval (CI): 1.02 to 1.20; ischemic stroke: HR, 1.15; 95% CI: 1.04 to 1.27). Similarly, a one-SD increase in mtDNA abundance was associated with lower risk of incident ischemic stroke (HR, 0.96; 95% CI: 0.93 to 0.99). No association was found between mtDNA abundance and incident hemorrhagic stroke.
      Table 2The multivariable-adjusted association between mtDNA abundance and incident stroke in European-ancestry participants from UK Biobank.
      ContinuousQ1Q2Q3Q4Q5
      Stroke incidenceHR (95%CI)HR (95%CI)HR (95%CI)HR (95%CI)HR (95%CI)HR (95%CI)
       AnyModel 10.97 (0.95, 1.00)1.11 (1.02, 1.20)1.07 (0.98, 1.16)1.03 (0.95, 1.12)1.03 (0.95, 1.12)1.00 (ref)
      Model 20.99 (0.96, 1.02)1.06 (0.97, 1.16)1.05 (0.96, 1.15)1.02 (0.94, 1.12)1.01 (0.92, 1.10)1.00 (ref)
       IschemicModel 10.96 (0.93, 0.99)1.15 (1.04, 1.27)1.11 (1.00, 1.23)1.07 (0.96, 1.18)1.05 (0.95, 1.16)1.00 (ref)
      Model 20.98 (0.94, 1.01)1.07 (0.96, 1.19)1.08 (0.97, 1.21)1.03 (0.92, 1.15)1.00 (0.89, 1.12)1.00 (ref)
       HemorrhagicModel 11.02 (0.97, 1.07)0.97 (0.84, 1.13)0.98 (0.84, 1.13)0.93 (0.80, 1.09)0.92 (0.79, 1.07)1.00 (ref)
      Model 21.02 (0.97, 1.07)0.98 (0.83, 1.15)0.99 (0.84, 1.17)0.97 (0.82, 1.14)0.93 (0.79, 1.10)1.00 (ref)
      Estimated hazard ratios per-SD increase in mtDNA abundance (continuous), or for the 1st to the 4th quintile compared with the 5th (reference) quintile (categorical) on any, ischemic, and hemorrhagic stroke. Model 1 includes age, sex, batch, PCs, white blood cell count, platelet. Model 2 includes model 1, BMI, smoking, total cholesterol, hypertension, diabetes, cholesterol lowering medication, blood pressure lowering medication. CI, confidence interval; HR, hazard ratio.
      After correcting for other confounders, the associations with stroke and ischemic stroke attenuated (any stroke: HR, 1.06; 95% CI: 0.97 to 1.16; ischemic stroke: HR, 1.07; 95% CI: 0.95 to 1.19), as did the continuous model on ischemic stroke (HR, 0.98; 95% CI: 0.94 to 1.01).

      3.3 Mendelian Randomization on mtDNA abundance and ischemic stroke

      3.3.1 Main analyses

      We did not observe evidence favoring an association between genetically-influenced lower mtDNA-CN and ischemic stroke (Fig. 2). The odds ratios per 1 SD less mtDNA-CN were 1.07 (95%CI: 0.95, 1.20) in MEGASTROKE, 1.04 (95%CI: 0.79, 1.37) in the UK Biobank, and 0.99 (95%CI: 0.82, 1.20) in FinnGen. After meta-analysis, in a combined sample size of 1,098,740 (of which 73,296 cases), the pooled odds ratio was 1.04 (95%CI: 0.95 to 1.15) per 1-SD decrease in genetically-influenced mtDNA abundance.
      Fig. 2
      Fig. 2Causal association between mtDNA abundance and ischemic stroke occurrence.
      The exact set of variants, their corresponding coefficients, standard errors, and p-values are presented in Supplementary Table 1. Variance explained (R2) was 2.0% and calculated based on the derived summary statistics. The MR-Egger intercept indicated no pleiotropy (p > 0.05). Although several outliers were identified with MR-PRESSO in MEGASTROKE and FinnGen, results remained similar after removal of these outlying SNPs. The Steiger test of directionality showed a correct causal direction, indicating that there is no evidence for reverse causation, and no different results were observed with MR-sensitivity analyses, MR-Egger and weighted median (Supplementary Table 2).
      Sub-analyses performed with separate outcomes, cardioembolic, large artery atherosclerosis, small-vessel, and hemorrhagic stroke (Supplementary Fig. 2 and 3), showed no evidence favoring a different result.

      3.3.2 Additional sensitivity analyses

      A total of 61 SNPS were associated with platelet count, which were subsequently excluded from additional sensitivity analyses. In the full sample, a 1-SD genetically-determined lower mtDNA abundance was associated with a higher risk of ischemic stroke (OR: 1.165; 95% CI: 1.026 to 1.323), although results from FinnGen did not align with those obtained in UK Biobank and MEGASTROKE (Supplementary Fig. 4).

      4. Discussion

      In the UK Biobank cohort, consisting of 288,572 participants after exclusion, an initial association was found between mtDNA abundance and incident ischemic stroke, which attenuated after adjustment for confounders. Consistent with the prospective analyses, MR analyses, using a total sample size of 73,296 cases and 1,025,444 controls, showed no evidence for an association between genetically-predicted mtDNA abundance and ischemic stroke. However, some in-depth sensitivity analyses in which SNPs associated with platelet count were excluded, did provide some preliminary evidence for low mtDNA-CN as possible causal driver for ischemic stroke.
      Although of specific interest, caution in these results is warranted given that they were mainly driven by results derived from MEGASTROKE, and to a lesser extent by UK Biobank. Furthermore, results from these additional MR sensitivity analyses deviated significantly from those observed in the prospective multivariable-adjusted analyses, and therefore do not meet the requirements for triangulation [
      • Lawlor D.A.
      • Tilling K.
      • Davey Smith G.
      Triangulation in aetiological epidemiology.
      ]. Collectively, our results indicate that there is only weak evidence for a causal association between mtDNA abundance and ischemic stroke, and more studies are required to elucidate the nature of the pleiotropy identified in our study, which goes beyond the current scope.
      Previously, an association was observed between low mtDNA-CN and increased risk of incident stroke in 20,162 participants, followed over a 13.5-year period, during which 1584 stroke events occurred [
      • Ashar F.N.
      • Zhang Y.
      • Longchamps R.J.
      • Lane J.
      • Moes A.
      • Grove M.L.
      • Mychaleckyj J.C.
      • Taylor K.D.
      • Coresh J.
      • Rotter J.I.
      • et al.
      Association of mitochondrial DNA copy number with cardiovascular disease.
      ], and therefore deviate from our study done in a larger sample of 288,752 participants with 6218 stroke cases. Difference in baseline health characteristics are possible reasons explaining the observed differences in results.
      Recent studies showed that mtDNA-CN could be a marker of stroke prognosis after hospitalization [
      • Song L.
      • Liu T.
      • Song Y.
      • Sun Y.
      • Li H.
      • Xiao N.
      • Xu H.
      • Ge J.
      • Bai C.
      • Wen H.
      • et al.
      mtDNA copy number contributes to all-cause mortality of lacunar infarct in a Chinese prospective stroke population.
      ,
      • Chong M.R.
      • Narula S.
      • Morton R.
      • Judge C.
      • Akhabir L.
      • Cawte N.
      • Pathan N.
      • Lali R.
      • Mohammadi-Shemirani P.
      • Shoamanesh A.
      • et al.
      Mitochondrial DNA copy number as a marker and mediator of stroke prognosis: observational and mendelian randomization analyses.
      ]. By analyzing mtDNA-CN, and consequently oxidative stress, our findings did provide some, albeit circumstantial, evidence for a relationship between oxidative stress and stroke occurrence, although this association attenuated after adjustment for confounders. In the Mendelian Randomization analysis, after excluding SNPs associated with platelet count, we also found an association between genetically determined mtDNA abundance and ischemic stroke risk. In contrast to ischemic stroke, we did not find an association between mtDNA abundance and hemorrhagic stroke in univariate or MR analyses. This difference might be explained because hemorrhagic stroke, in contrast to ischemic stroke, is also often caused by non-classic cardiovascular mechanisms such as vascular amyloid deposition in cerebral amyloid angiopathy [
      • Grysiewicz R.A.
      • Thomas K.
      • Pandey D.K.
      Epidemiology of ischemic and hemorrhagic stroke: incidence, prevalence, mortality, and risk factors.
      ].
      Our data on mtDNA abundance was obtained from leukocytes. Although some of the leukocytes may be directly involved in the pathology of stroke, additional cell types such as endothelial and smooth muscle cells, that we did not query for mitochondrial abundance, are clearly more directly involved. This could potentially explain our overall null findings. Studies on the differences in mitochondrial function within an individual between cell groups are largely non-existent. However, mtDNA-CN measured in blood has been associated with gene expression in other tissues, which suggests mtDNA-CN derived from leukocytes can reflect metabolic health across multiple tissues [
      • Yang S.Y.
      • Castellani C.A.
      • Longchamps R.J.
      • Pillalamarri V.K.
      • O'Rourke B.
      • Guallar E.
      • Arking D.E.
      Blood-derived mitochondrial DNA copy number is associated with gene expression across multiple tissues and is predictive for incident neurodegenerative disease.
      ]. Thus, the evidence so far indicates that mitochondrial dysfunction, as measured with leukocyte mtDNA-CN, is systemic. Of interest, using similar methodology as in our study, low genetically-influenced mtDNA has recently been associated with increased dementia risk [
      • Chong M.
      • Mohammadi-Shemirani P.
      • Perrot N.
      • Nelson W.
      • Morton R.
      • Narula S.
      • Lali R.
      • Khan I.
      • Khan M.
      • Judge C.
      • et al.
      GWAS and ExWAS of blood Mitochondrial DNA copy number identifies 71 loci and highlights a potential causal role in dementia.
      ]. This would further indicate that lower mtDNA-CN, although measured in leukocytes, can reflect processes of a systemic increase in disease risk.
      A key strength of this study is the statistical power of the analyses of the association between stroke and mitochondrial abundance (288,572 participants for the multivariable survival analysis and 1,098,740 for the MR, respectively). Additionally, we adopted the triangulation of causal inference [
      • Lawlor D.A.
      • Tilling K.
      • Davey Smith G.
      Triangulation in aetiological epidemiology.
      ]. By using two different approaches in observational research to study the association between low mtDNA abundance and (ischemic) stroke risk, we increased the credibility of our results. Although results from both our used approaches were not exactly similar, they were directionally consistent.
      Some limitations are to be considered. First, mtDNA abundance was determined from intensities of genotyping probes on mitochondrial DNA, whereas the assessment with whole-exome sequencing is generally considered to result in more reliable mtDNA abundance estimates [
      • Zhang P.
      • Lehmann B.D.
      • Samuels D.C.
      • Zhao S.
      • Zhao Y.Y.
      • Shyr Y.
      • Guo Y.
      Estimating relative mitochondrial DNA copy number using high throughput sequencing data.
      ,
      • Picardi E.
      • Pesole G.
      Mitochondrial genomes gleaned from human whole-exome sequencing.
      ]. Although Hägg et al. showed a moderate correlation between mtDNA based on SNP array intensities and exome sequencing of 0.33 [
      • Hagg S.
      • Jylhava J.
      • Wang Y.
      • Czene K.
      • Grassmann F.
      Deciphering the genetic and epidemiological landscape of mitochondrial DNA abundance.
      ], analyses still indicated the measurements of SNP array intensities reflect underlying biology of mtDNA abundance. For this reason, the increased variance is most likely the result of nondifferential measurement error, and therefore considered to mainly cause a reduction in statistical power. As a consequence, the true associations, particularly those from the multivariable-adjusted prospective analyses, are most likely larger than observed. Second, our study population consists of predominantly Caucasian participants, limiting the generalizability of the results to other ancestry groups. Third, Mendelian Randomization functions on several assumptions. However, using several sensitivity analyses such as MR-Egger and MR-PRESSO, we can establish with some conviction that these are fulfilled. In addition, although in a one-sample MR (as conducted in the UK Biobank) the assumption of independence does not hold up, previous studies have shown that two-sample MR methods can be used reliably with large enough biobanks [
      • Minelli C.
      • Del Greco M.F.
      • van der Plaat D.A.
      • Bowden J.
      • Sheehan N.A.
      • Thompson J.
      The use of two-sample methods for Mendelian randomization analyses on single large datasets.
      ]. Last, despite a large sample size in the multivariable adjusted analysis, stroke, especially hemorrhagic stroke incidence, was relatively few. However, as an association was found before correction, we think our analyses had enough power to detect a difference between groups. In addition, we used one of the larger data sets available.
      In conclusion, despite a large sample size, our prospective study did not find evidence for an association between mtDNA abundance and ischemic or hemorrhagic stroke. After exclusion of pleiotropic SNPs associated with platelet count, we found some preliminary evidence for an association between genetically determined lower mtDNA-CN and ischemic stroke risk using MR. However, further studies are required for validation and to examine the nature of this type of pleiotropy.

      Financial support

      This work was supported by the VELUX Stiftung [grant number 1156 ] to DvH and RN, and JL was supported by the China Scholarship Counsel [No. 201808500155 ]. RN was supported by an innovation grant from the Dutch Heart Foundation [grant number 2019T103 to R.N.]. Parts of this work were funded by the Åke Wibergs Foundation (grant number M19-0294 to F.G).

      Author contributions

      LGM, JL, RN and DvH designed research; LGM and JL conducted research; LGM and JL performed statistical analysis; LGM, JL, RN, and MJHW wrote paper; LGM had primary responsibility for final content. DvH, KWvD, SH, FG and MJHW contributed to the data interpretation and commented on initial versions of the manuscript; All authors read and approved the final manuscript.

      Declaration of interests

      The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

      Acknowledgements

      The authors are grateful to the UK Biobank for allowing us the use of their data. The analyses done in UK Biobank were done under project number 56340. Furthermore, the authors acknowledge the participants and investigators of the MEGASTROKE consortium and the FinnGen Biobank who contributed to the summary statistics data which are made available for further studies.

      Appendix A. Supplementary data

      The following are the Supplementary data to this article:

      References

        • Wang H.
        Global age-sex-specific fertility, mortality, healthy life expectancy (HALE), and population estimates in 204 countries and territories, 1950-2019: a comprehensive demographic analysis for the Global Burden of Disease Study 2019.
        Lancet. 2020; 396: 1160-1203
        • Rodrigo R.
        • Fernandez-Gajardo R.
        • Gutierrez R.
        • Matamala J.M.
        • Carrasco R.
        • Miranda-Merchak A.
        • Feuerhake W.
        Oxidative stress and pathophysiology of ischemic stroke: novel therapeutic opportunities.
        CNS Neurol. Disord.: Drug Targets. 2013; 12: 698-714
        • Allen C.L.
        • Bayraktutan U.
        Oxidative stress and its role in the pathogenesis of ischaemic stroke.
        Int. J. Stroke. 2009; 4: 461-470
        • Duan X.
        • Wen Z.
        • Shen H.
        • Shen M.
        • Chen G.
        Intracerebral hemorrhage, oxidative stress, and antioxidant therapy.
        Oxid. Med. Cell. Longev. 2016; 20161203285
        • Shirley R.
        • Ord E.N.
        • Work L.M.
        Oxidative stress and the use of antioxidants in stroke.
        Antioxidants. 2014; 3: 472-501
        • Lackner L.L.
        • Nunnari J.M.
        The molecular mechanism and cellular functions of mitochondrial division.
        Biochim. Biophys. Acta. 2009; 1792: 1138-1144
        • Bandy B.
        • Davison A.J.
        Mitochondrial mutations may increase oxidative stress: implications for carcinogenesis and aging?.
        Free Radic. Biol. Med. 1990; 8: 523-539
        • Chung J.K.
        • Lee S.Y.
        • Park M.
        • Joo E.J.
        • Kim S.A.
        Investigation of mitochondrial DNA copy number in patients with major depressive disorder.
        Psychiatr. Res. 2019; 282112616
        • Longchamps R.J.
        • Castellani C.A.
        • Yang S.Y.
        • Newcomb C.E.
        • Sumpter J.A.
        • Lane J.
        • Grove M.L.
        • Guallar E.
        • Pankratz N.
        • Taylor K.D.
        • et al.
        Evaluation of mitochondrial DNA copy number estimation techniques.
        PLoS One. 2020; 15: e0228166
        • Yang J.L.
        • Mukda S.
        • Chen S.D.
        Diverse roles of mitochondria in ischemic stroke.
        Redox Biol. 2018; 16: 263-275
        • Hagg S.
        • Jylhava J.
        • Wang Y.
        • Czene K.
        • Grassmann F.
        Deciphering the genetic and epidemiological landscape of mitochondrial DNA abundance.
        Hum. Genet. 2021; 140: 849-861
        • Chong M.
        • Mohammadi-Shemirani P.
        • Perrot N.
        • Nelson W.
        • Morton R.
        • Narula S.
        • Lali R.
        • Khan I.
        • Khan M.
        • Judge C.
        • et al.
        GWAS and ExWAS of blood Mitochondrial DNA copy number identifies 71 loci and highlights a potential causal role in dementia.
        Elife. 2022; : 11
        • Liu F.
        • Lu J.
        • Manaenko A.
        • Tang J.
        • Hu Q.
        Mitochondria in ischemic stroke: new insight and implications.
        Aging Dis. 2018; 9: 924-937
        • Wachsmuth M.
        • Hubner A.
        • Li M.
        • Madea B.
        • Stoneking M.
        Age-related and heteroplasmy-related variation in human mtDNA copy number.
        PLoS Genet. 2016; 12: e1005939
        • Lawlor D.A.
        • Tilling K.
        • Davey Smith G.
        Triangulation in aetiological epidemiology.
        Int. J. Epidemiol. 2016; 45: 1866-1886
        • Burgess S.
        • Butterworth A.
        • Thompson S.G.
        Mendelian randomization analysis with multiple genetic variants using summarized data.
        Genet. Epidemiol. 2013; 37: 658-665
        • Sudlow C.
        • Gallacher J.
        • Allen N.
        • Beral V.
        • Burton P.
        • Danesh J.
        • Downey P.
        • Elliott P.
        • Green J.
        • Landray M.
        • et al.
        UK biobank: an open access resource for identifying the causes of a wide range of complex diseases of middle and old age.
        PLoS Med. 2015; 12: e1001779
        • Longchamps R.J.
        • Yang S.Y.
        • Castellani C.A.
        • Shi W.
        • Lane J.
        • Grove M.L.
        • Bartz T.M.
        • Sarnowski C.
        • Liu C.
        • Burrows K.
        • et al.
        Genome-wide analysis of mitochondrial DNA copy number reveals loci implicated in nucleotide metabolism, platelet activation, and megakaryocyte proliferation.
        Hum. Genet. 2022; 141: 127-146
        • Malik R.
        • Chauhan G.
        • Traylor M.
        • Sargurupremraj M.
        • Okada Y.
        • Mishra A.
        • Rutten-Jacobs L.
        • Giese A.K.
        • van der Laan S.W.
        • Gretarsdottir S.
        • et al.
        Multiancestry genome-wide association study of 520,000 subjects identifies 32 loci associated with stroke and stroke subtypes.
        Nat. Genet. 2018; 50: 524-537
        • Hemani G.
        • Zheng J.
        • Elsworth B.
        • Wade K.H.
        • Haberland V.
        • Baird D.
        • Laurin C.
        • Burgess S.
        • Bowden J.
        • Langdon R.
        • et al.
        The MR-Base Platform Supports Systematic Causal Inference across the Human Phenome. vol. 7. Elife, 2018
        • Burgess S.
        • Thompson S.G.
        Interpreting findings from Mendelian randomization using the MR-Egger method.
        Eur. J. Epidemiol. 2017; 32: 377-389
        • 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: 512-525
        • Bowden J.
        • Davey Smith G.
        • Haycock P.C.
        • Burgess S.
        Consistent estimation in mendelian randomization with some invalid instruments using a weighted median estimator.
        Genet. Epidemiol. 2016; 40: 304-314
        • Verbanck M.
        • Chen C.Y.
        • Neale B.
        • Do R.
        Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases.
        Nat. Genet. 2018; 50: 693-698
        • Greco M.F.
        • Minelli C.
        • Sheehan N.A.
        • Thompson J.R.
        Detecting pleiotropy in Mendelian randomisation studies with summary data and a continuous outcome.
        Stat. Med. 2015; 34: 2926-2940
        • Brion M.J.
        • Shakhbazov K.
        • Visscher P.M.
        Calculating statistical power in Mendelian randomization studies.
        Int. J. Epidemiol. 2013; 42: 1497-1501
        • Minelli C.
        • Del Greco M.F.
        • van der Plaat D.A.
        • Bowden J.
        • Sheehan N.A.
        • Thompson J.
        The use of two-sample methods for Mendelian randomization analyses on single large datasets.
        Int. J. Epidemiol. 2021;
        • Ashar F.N.
        • Zhang Y.
        • Longchamps R.J.
        • Lane J.
        • Moes A.
        • Grove M.L.
        • Mychaleckyj J.C.
        • Taylor K.D.
        • Coresh J.
        • Rotter J.I.
        • et al.
        Association of mitochondrial DNA copy number with cardiovascular disease.
        JAMA Cardiol. 2017; 2: 1247-1255
        • Song L.
        • Liu T.
        • Song Y.
        • Sun Y.
        • Li H.
        • Xiao N.
        • Xu H.
        • Ge J.
        • Bai C.
        • Wen H.
        • et al.
        mtDNA copy number contributes to all-cause mortality of lacunar infarct in a Chinese prospective stroke population.
        J Cardiovasc Transl Res. 2020; 13: 783-789
        • Chong M.R.
        • Narula S.
        • Morton R.
        • Judge C.
        • Akhabir L.
        • Cawte N.
        • Pathan N.
        • Lali R.
        • Mohammadi-Shemirani P.
        • Shoamanesh A.
        • et al.
        Mitochondrial DNA copy number as a marker and mediator of stroke prognosis: observational and mendelian randomization analyses.
        Neurology. 2022; 98: e470-e482
        • Grysiewicz R.A.
        • Thomas K.
        • Pandey D.K.
        Epidemiology of ischemic and hemorrhagic stroke: incidence, prevalence, mortality, and risk factors.
        Neurol. Clin. 2008; 26 (vii): 871-895
        • Yang S.Y.
        • Castellani C.A.
        • Longchamps R.J.
        • Pillalamarri V.K.
        • O'Rourke B.
        • Guallar E.
        • Arking D.E.
        Blood-derived mitochondrial DNA copy number is associated with gene expression across multiple tissues and is predictive for incident neurodegenerative disease.
        Genome Res. 2021; 31: 349-358
        • Zhang P.
        • Lehmann B.D.
        • Samuels D.C.
        • Zhao S.
        • Zhao Y.Y.
        • Shyr Y.
        • Guo Y.
        Estimating relative mitochondrial DNA copy number using high throughput sequencing data.
        Genomics. 2017; 109: 457-462
        • Picardi E.
        • Pesole G.
        Mitochondrial genomes gleaned from human whole-exome sequencing.
        Nat. Methods. 2012; 9: 523-524