Exploring the Effects of Mitonuclear Interactions on Mitochondrial DNA Gene Expression in Humans

Abstract

Most mitochondrial protein complexes include both nuclear and mitochondrial gene products, which coevolved to work together. This coevolution can be disrupted due to disparity in genetic ancestry between the nuclear and mitochondrial genomes in recently admixed populations. Such mitonuclear DNA discordance might result in phenotypic effects. Several nuclear-encoded proteins regulate expression of mitochondrial DNA (mtDNA) genes. We hypothesized that mitonuclear DNA discordance affects expression of genes encoded by mtDNA. To test this, we utilized the data from the GTEx project, which contains expression levels for ∼100 African Americans and >600 European Americans. The varying proportion of African and European ancestry in recently admixed African Americans provides a range of mitonuclear discordance values, which can be correlated with mtDNA gene expression levels (adjusted for age and ischemic time). In contrast, European Americans did not undergo recent admixture. We demonstrated that, for most mtDNA protein-coding genes, expression levels in energetically-demanding tissues were lower in African Americans than in European Americans. Furthermore, gene expression levels were lower in individuals with higher mitonuclear discordance, independent of population. Moreover, we found a negative correlation between mtDNA gene expression and mitonuclear discordance. In African Americans, the average value of African ancestry was higher for nuclear-encoded mitochondrial than non-mitochondrial genes, facilitating a match in ancestry with the mtDNA and more optimal interactions. These results represent an example of a phenotypic effect of mitonuclear discordance on human admixed populations, and have potential biomedical applications.

Keywords: gene expression; genetic ancestry; mitochondrial DNA; mitonuclear DNA discordance; mitonuclear coevolution; mitonuclear incompatibility.

FIGURE 1

Histogram of mitonuclear DNA discordance values for (A) African Americans, and (B) European Americans. Note that the range of the Y axis is much smaller in (A) than in (B), which reflects the smaller sample size for African Americans than for European Americans among the genotyped individuals in the GTEx dataset.

FIGURE 2

(A) Pairwise boxplot of skeletal muscle mtDNA gene expression in African Americans (n = 85) and European Americans (n = 580), adjusted by age and ischemic time. Stars indicate significant results (p < 0.05) of one-sided, Bonferroni-corrected Mann-Whitney U test of mean ranks (* indicates p < 0.05, ** indicates p < 0.01, *** indicates p < 0.001). (B) Mann-Whitney U test common-language effect sizes comparing adjusted mtDNA gene expression between African Americans and European Americans for skeletal muscle, smooth muscle (esophagus), artery (tibial), nerve (tibial), whole blood, and heart (ventricle and atrium). Bold values indicate significant difference (p < 0.05) using a one-sided, Bonferroni-corrected Mann-Whitney U test of mean ranks. p-values are listed in Supplementary Table S5. Values on the blue background, above 0.5, support lower adjusted expression with higher mitonuclear DNA discordance, consistent with our hypothesis; values on the orange background, below 0.5, support higher adjusted expression with higher mitonuclear DNA discordance. The background is darker with increasing effect. Bonferroni correction accounts for multiple hypothesis testing across the 11 mtDNA protein-coding genes analyzed.

FIGURE 3

(A) Permutation test comparing the mean gene expression (adjusted by age and ischemic time) for skeletal muscle in high vs. low mitonuclear DNA discordance groups, regardless of which population each individual belongs to. The distribution is the expected differences in mean adjusted gene expression between groups (with permuted labels). These groups were determined using a cutoff at the mean mitonuclear DNA discordance. The vertical red line indicates the observed difference in mean TPM between groups in our dataset. (B) Bonferroni-corrected, one-sided permutation test results for skeletal muscle, smooth muscle (esophagus), artery (tibial), nerve (tibial), whole blood, and heart (ventricle and atrium). The mean mitonuclear discordance values were used as the threshold for high and low discordance groups, for each tissue (0.042, 0.047, 0.048, 0.045, 0.045, 0.043, and 0.033, respectively). Bold text indicates significant results (p < 0.05). Values on the blue background indicate lower adjusted expression with higher mitonuclear DNA discordance, consistent with our hypothesis; values on the orange background indicate higher adjusted expression with higher mitonuclear DNA discordance.

FIGURE 4

Testing the relationship between mitonuclear DNA discordance and mtDNA gene expression (adjusted for age and ischemic time) using a one-sided Spearman’s rank-order correlation. Bold values indicate Bonferroni-corrected significant results (p < 0.05). p-values are listed in Supplementary Table S3. Values on the blue background indicate lower adjusted expression with higher mitonuclear DNA discordance, consistent with our hypothesis; values on the orange background indicate higher adjusted expression with higher mitonuclear DNA discordance. The background is darker with increasing effect. Bonferroni correction accounts for multiple hypothesis testing across the 11 mtDNA protein-coding genes analyzed.

FIGURE 5

Local ancestry enrichment for African ancestry in African Americans from the GTEx dataset. The Y axis shows local ancestry deviation and the X axis shows functional categories (high-mt: 167 genes, low-mt: 793, non-mt: 17,456 genes, see text for details). A block bootstrap approach (see Methods) was used to generate the distributions of deviation from global ancestry. Blue horizontal bars show the empirical 95% CI of the mean ancestry deviation. The p-values correspond to a t-test of independent means comparing mean local ancestry deviations at non-mt genes to those at high-mt and low-mt genes.