Comprehensive association testing of common mitochondrial DNA variation in metabolic disease

Abstract

Many lines of evidence implicate mitochondria in phenotypic variation: (a) rare mutations in mitochondrial proteins cause metabolic, neurological, and muscular disorders; (b) alterations in oxidative phosphorylation are characteristic of type 2 diabetes, Parkinson disease, Huntington disease, and other diseases; and (c) common missense variants in the mitochondrial genome (mtDNA) have been implicated as having been subject to natural selection for adaptation to cold climates and contributing to "energy deficiency" diseases today. To test the hypothesis that common mtDNA variation influences human physiology and disease, we identified all 144 variants with frequency >1% in Europeans from >900 publicly available European mtDNA sequences and selected 64 tagging single-nucleotide polymorphisms that efficiently capture all common variation (except the hypervariable D-loop). Next, we evaluated the complete set of common mtDNA variants for association with type 2 diabetes in a sample of 3,304 diabetics and 3,304 matched nondiabetic individuals. Association of mtDNA variants with other metabolic traits (body mass index, measures of insulin secretion and action, blood pressure, and cholesterol) was also tested in subsets of this sample. We did not find a significant association of common mtDNA variants with these metabolic phenotypes. Moreover, we failed to identify any physiological effect of alleles that were previously proposed to have been adaptive for energy metabolism in human evolution. More generally, this comprehensive association-testing framework can readily be applied to other diseases for which mitochondrial dysfunction has been implicated.

Figure 1.

Procedure for identification and comprehensive disease-association testing of all common variants in the mtDNA coding region. 1 and 2, A total of 928 mtDNA sequences of European origin were aligned to identify 144 variants with frequency >1%. 3, tSNPs and multimarker haplotypes of tSNPs were selected to capture all 144 variant sites and haplogroups with r²⩾0.8 (a haplotype of tSNPs [shaded box] captures an untyped SNP [unshaded box]). 4, tSNPs were genotyped in a diabetic case-control panel with available metabolic phenotypes. 5, All hypotheses to be tested were enumerated. 6, Association tests were performed. 7, Studywide significance of results was assessed by permutation (multiple rounds of association testing with randomization of case-control labels or, for quantitative measures, shuffling within a population).