Extensive pathogenicity of mitochondrial heteroplasmy in healthy human individuals

Abstract

A majority of mitochondrial DNA (mtDNA) mutations reported to be implicated in diseases are heteroplasmic, a status with coexisting mtDNA variants in a single cell. Quantifying the prevalence of mitochondrial heteroplasmy and its pathogenic effect in healthy individuals could further our understanding of its possible roles in various diseases. A total of 1,085 human individuals from 14 global populations have been sequenced by the 1000 Genomes Project to a mean coverage of ∼2,000× on mtDNA. Using a combination of stringent thresholds and a maximum-likelihood method to define heteroplasmy, we demonstrated that ∼90% of the individuals carry at least one heteroplasmy. At least 20% of individuals harbor heteroplasmies reported to be implicated in disease. Mitochondrial heteroplasmy tend to show high pathogenicity, and is significantly overrepresented in disease-associated loci. Consistent with their deleterious effect, heteroplasmies with derived allele frequency larger than 60% within an individual show a significant reduction in pathogenicity, indicating the action of purifying selection. Purifying selection on heteroplasmies can also be inferred from nonsynonymous and synonymous heteroplasmy comparison and the unfolded site frequency spectra for different functional sites in mtDNA. Nevertheless, in comparison with population polymorphic mtDNA mutations, the purifying selection is much less efficient in removing heteroplasmic mutations. The prevalence of mitochondrial heteroplasmy with high pathogenic potential in healthy individuals, along with the possibility of these mutations drifting to high frequency inside a subpopulation of cells across lifespan, emphasizes the importance of managing mitochondrial heteroplasmy to prevent disease progression.

Fig. 1.

Distribution of heteroplasmy in the sample. (A) The percentage of individuals carrying a specific number of heteroplasmy. The category of individuals who do not carry any heteroplasmy is highlighted in pink. (B) Histogram for minor allele frequency of heteroplasmy. (C) The genomic distribution of heteroplasmies and their incidences in the sample. The inner layer represents the mitochondrial genome with tRNA genes highlighted in black, rRNA genes in dark gray and protein-coding genes in light gray. The blue layer indicates the number of individuals (in a total of 1,085) carrying heteroplasmy at a specific site. The number of individuals is shown at a common logarithm scale.

Fig. 2.

Mitochondrial heteroplasmy is highly pathogenic. (A) The percentage of loci associated with diseases in all, heteroplasmic, and polymorphic sites. “All diseases” represents all diseases included in MITOMAP. Myopathy and encephalomyopathy are the two disease categories that have the highest number of mitochondrial loci reported to be associated with. (B) The box plot of MutPred pathogenicity scores for all possible NS variants in the mitochondrial genome, NS heteroplasmies, and polymorphisms. Heteroplasmies occurring in multiple individuals were counted only once. (C) The percentage of PolyPhen-2 predicted damaging variants in all possible NS variants, NS heteroplasmies, and polymorphisms. (D) The percentage of predicted deleterious tRNA variants in all possible variants, heteroplasmies, and polymorphisms. tRNA represents all regions of tRNA genes, including loop and stem regions; Loop represents the loop region; Stem-WC refers to the Watson–Crick pairing positions in the stem region; Stem-notWC refers to those that are not Watson–Crick paired. The error bar represents 95% CI from 10,000 bootstraps. **P < 0.01; *P < 0.05.

Fig. 3.

Purifying selection on mitochondrial heteroplasmy. (A) The prevalence of synonymous and NS heteroplasmies, which is defined as the percentage of all possible (synonymous or NS) changes that is observed to be heteroplasmic. **P = 2.01e-10 in χ² test. (B) The distribution of DAF for heteroplasmies in different mtDNA genomic regions. (C) The distribution of DAF for disease-associated and synonymous heteroplasmies. (D) The average pathogenicity score in each bin of DAF. Error bar represents 1 SE. The red line represents model-fitting with a logistic function of $y=0.67/(1+e^{(1-x)/-0.16})$. R² = 0.9794, P < 9.76e-06.

Fig. 4.

Less-efficient purifying selection on mitochondrial heteroplasmy than on polymorphism. (A) The prevalence of synonymous and NS heteroplasmies in comparison with that of synonymous and NS polymorphisms. (B) The selection function for heteroplasmy (or polymorphism) defined by dividing the observed distribution of pathogenicity scores for heteroplasmy by the expected distribution of pathogenicity scores from all possible NS variants. The dashed line represents the expected value, 1, for selection function under neutral evolution. The exponential fit for polymorphism is $y={12e}^{-x/0.23}$. R² = 0.9758, P = 4.71e-06. The exponential function for heteroplasmy is $y={10e}^{-x/0.079}+0.99$. R² = 0.9650, P = 4.65e-06.