A genome-wide association study of mitochondrial DNA copy number in two population-based cohorts

Abstract

Background: Mitochondrial DNA copy number (mtDNA CN) exhibits interindividual and intercellular variation, but few genome-wide association studies (GWAS) of directly assayed mtDNA CN exist. We undertook a GWAS of qPCR-assayed mtDNA CN in the Avon Longitudinal Study of Parents and Children (ALSPAC) and the UK Blood Service (UKBS) cohort. After validating and harmonising data, 5461 ALSPAC mothers (16-43 years at mtDNA CN assay) and 1338 UKBS females (17-69 years) were included in a meta-analysis. Sensitivity analyses restricted to females with white cell-extracted DNA and adjusted for estimated or assayed cell proportions. Associations were also explored in ALSPAC children and UKBS males.

Results: A neutrophil-associated locus approached genome-wide significance (rs709591 [MED24], β (change in SD units of mtDNA CN per allele) [SE] - 0.084 [0.016], p = 1.54e-07) in the main meta-analysis of adult females. This association was concordant in magnitude and direction in UKBS males and ALSPAC neonates. SNPs in and around ABHD8 were associated with mtDNA CN in ALSPAC neonates (rs10424198, β [SE] 0.262 [0.034], p = 1.40e-14), but not other study groups. In a meta-analysis of unrelated individuals (N = 11,253), we replicated a published association in TFAM (β [SE] 0.046 [0.017], p = 0.006), with an effect size much smaller than that observed in the replication analysis of a previous in silico GWAS.

Conclusions: In a hypothesis-generating GWAS, we confirm an association between TFAM and mtDNA CN and present putative loci requiring replication in much larger samples. We discuss the limitations of our work, in terms of measurement error and cellular heterogeneity, and highlight the need for larger studies to better understand nuclear genomic control of mtDNA copy number.

Keywords: ALSPAC; Complex traits; Genetic epidemiology; Genome-wide association study; Mitochondrial DNA.

Fig. 1

a Manhattan (left)/quantile-quantile (QQ) plots (right) for ALSPAC (all mothers). λ = 0.995. ‘Minimally adjusted’ refers to the fact that these results are from the analysis that did not adjust for cell proportions. b Manhattan (left)/quantile-quantile (QQ) plots (right) for UKBS (females). λ = 1.011. ‘Minimally adjusted’ refers to the fact that these results are from the analysis that did not adjust for cell proportions. c Manhattan (left)/quantile-quantile (QQ) plots (right) for random-effects meta-analysis of ALSPAC (all mothers) and UKBS (females). λ = 0.995 and 1.011 for ALSPAC (all mothers) and UKBS females, respectively, and meta-analyses are corrected for these lambdas. ‘Minimally adjusted’ refers to the fact that these results are from the analysis that did not adjust for cell proportions

Fig. 2

a Manhattan (left)/quantile-quantile (QQ) plots (right) for ALSPAC (white cell mothers). λ = 0.992. ‘Minimally adjusted’ refers to the fact that these results are from the analysis that did not adjust for cell proportions. b Manhattan (left)/quantile-quantile (QQ) plots (right) for random-effects meta-analysis of ALSPAC (white cell mothers) and UKBS females. λ = 0.992 and 1.011 for ALSPAC (white cell mothers, see Fig. 2a) and UKBS females (see Fig. 1b), and meta-analyses are corrected for these lambdas. ‘Minimally adjusted’ refers to the fact that these results are from the analysis that did not adjust for cell proportions

Fig. 3

Regional association plots (created with LocusZoom). These three regional association plots (a, b, and c, created with LocusZoom) detail the top loci presented in the meta-analyses of all ALSPAC mothers and UKBS females (a, b) and of the one locus identified after restriction of the meta-analysis to ALSPAC mothers with DNA extracted from white cells, only (c). In each plot, the lead SNP (i.e. the SNP with the lowest p value) is annotated in purple, with other SNPs colour coded according to their values of linkage disequilibrium (in r²) with the lead SNP. Transformed, −log₁₀p values and recombination rate (in centimorgans per megabase, cM/Mb) are shown on the left and right y-axes, respectively. A schematic of the genes in each region, along with coordinates and annotations, is shown at the bottom of each plot, with chromosomal location in megabases (Mb) along the x-axis. See Table 2 for more details of each locus

Fig. 4

Extraction and meta-analysis of two top loci from Cai et al. [17] from study groups in this GWAS. SNPs for two loci identified in a GWAS of in silico estimated mtDNA CN were extracted from each of the study groups in this cohort (NB: a smaller subset of 2833 ALSPAC mothers were used, since there were mother-child duos present between the original study group of 5461 and the two groups of ALSPAC children) total N = 11,253. Columns: SNP = rsID; gene = gene name; group = study group, beta = effect size; LCI/UCI = 95% confidence interval (lower, then upper bound); P = p value; and I² = I² metric for heterogeneity. Meta-analyses were by random-effects and are shown as black diamonds. For reference, the ALSPAC estimate from Cai et al. [17] is shown for each locus. This replication group included ALSPAC 6–9-year-olds (with mtDNA CN assayed from sequence data). Betas had to be harmonised, as those in Cai et al. [17] were given as SD change in mtDNA CN per SD increase in genotype. SD of genotype was estimated from allele frequencies provided for the cohort by Cai et al. [17] given as 0.342 for rs445 and 0.169 for rs11006126 (in the supplement of this paper). SDs were then calculated as √(2 × (1-MAF) × MAF) (evaluating to 0.53 and 0.67 for rs11006126 and rs445, respectively). Betas and standard errors were then transformed from those given in Table 1 of Cai et al. to a ‘per risk allele’ scale, by multiplying the given beta by (1/the estimated SD), i.e. rs11006126 = 0.179 × (1/0.53); rs445 = 0.110 × (1/0.67). CaiALSPAC = result from Cai et al. [17]. UKBSF/UKBSM = females and males in UKBS cohort