Cell type-specific purifying selection of synonymous mitochondrial DNA variation

Abstract

While somatic variants are well-characterized drivers of tumor evolution, their influence on cellular fitness in nonmalignant contexts remains understudied. We identified a mosaic synonymous variant (m.7076A > G) in the mitochondrial DNA (mtDNA)-encoded cytochrome c-oxidase subunit 1 (MT-CO1, p.Gly391=), present at homoplasmy in 47% of immune cells from a healthy donor. Single-cell multiomics revealed strong, lineage-specific selection against the m.7076G allele in CD8⁺ effector memory T cells, but not other T cell subsets, mirroring patterns of purifying selection of pathogenic mtDNA alleles. The limited anticodon diversity of mitochondrial tRNAs forces m.7076G translation to rely on wobble pairing, unlike the Watson-Crick-Franklin pairing used for m.7076A. Mitochondrial ribosome profiling confirmed stalled translation of the m.7076G allele. Functional analyses demonstrated that the elevated translational and metabolic demands of short-lived effector T cells (SLECs) amplify dependence on MT-CO1, driving this selective pressure. These findings suggest that synonymous variants can alter codon syntax, impacting mitochondrial physiology in a cell type-specific manner.

Keywords: immunology; mitochondria; selection; single-cell.

Fig. 1.

Identification of a mosaic synonymous mtDNA variant with a CD8⁺ T cell restricted selection bias. (A) Schematic of longitudinal peripheral blood samples obtained from a healthy donor. A total of 5 draws spanning 150 d were taken and processed with the mtscATAC-seq assay. (B) Summary of somatic mtDNA mutations called from aggregated draws. The overall median heteroplasmy is noted, as well as the m.7076A > G allele that is present at a 47.3% pseudobulk heteroplasmy. (C) Distribution of single-cell heteroplasmy across all cells profiled for the m.7076A > G allele. (D) Schematic of the mitochondrial genome with genes contributing to indicated complexes of the respiratory chain (I, III, IV, and V) being color-coded. The asterisk under the MT-CO1 genes denotes the position of the m.7076 allele. Annotation of the mutation, including the protein consequence (p.Gly391=) of the synonymous variant is shown below. (E) Uniform manifold approximation and projection (UMAP) of accessible chromatin profiles of PBMCs assayed via mtscATAC-seq colored by the density of the m.7076A (wildtype) allele. (F) Cluster annotation and cell type labeling of the same cells as in (E). The arrow indicates the CD8⁺ T effector memory (CD8⁺ TEM) population. (G) Ratio of homoplasmic cells with wildtype m.7076A to mutant m.7076G variants across indicated cell state clusters. The arrow highlights the CD8⁺ TEM cell state as the population with the greatest skew (P < 2.2e−16; binomial test).

Fig. 2.

Stable expression of MT-CO1 transcript but altered clone size of CD8⁺ effector memory T cells carrying the mutant m.7076G allele. (A) Heteroplasmy of m.7076A > G in indicated T cell subpopulations based on scRNA-seq. The size of each dot is scaled by the abundance of cells in each cell state. (B) Comparison of MT-CO1 expression across indicated cell states stratified by the m.7076A and G alleles. P-values from a Wilcox test comparing log₂ MT-CO1 UMI counts per cell were not significant at type I error of 0.05. (C) Comparison of TCR clone sizes (number of cells per clone) between CD4⁺ and CD8⁺ T cells with homoplasmic m.7076A or m.7076G alleles. P-values are shown for a Wilcox test comparing clone sizes which were significant at type I error of 0.05. (D) Differential gene expression of all genes between cells with homoplasmic wildtype m.7076A vs. mutant m.7076G within the indicated CD8⁺ T cell compartments. 0 genes were differentially expressed in naive CD8⁺ T cells whereas 32 were differentially expressed in CD8⁺ TEM cells, including the 5 highlighted in the text. No other differentially expressed genes were observed in other T cell subsets. Multiple hypothesis testing: per subtype Bonferroni-adjusted P-value.

Fig. 3.

In vitro activation of T cells refines cell states depleted of the mutant m.7076G allele. (A) Schematic of experimental design. T cells were isolated from Donor 1, in vitro activated, and cultured for 9 d before profiling via flow cytometry and ASAP-seq. (B) UMAP of accessible chromatin profiles and projected ratio of CD8 over CD4 antibody-derived tags from day 9 cells profiled via ASAP-seq. The arrow indicates a population highly enriched for the m.7076A (wildtype) allele. (C) Same as (B) but colored by the density of the m.7076A (wildtype) allele. (D) UMAP embedding colored by KLRG1 (Top) and IL7R (Bottom) antibody tag density. (E) UMAP colored by selected gene activity scores for four indicated gene loci. (F) UMAP colored by indicated cell state cluster. (G) Ratio of wildtype m.7076A to mutant m.7076G cells within indicated cell states. P-value represents the statistical significance of a two-sided binomial test statistic.

Fig. 4.

SLECs exhibit high metabolic activity and dependence on OXPHOS. (A) PBMCs from four healthy donors, including Donor 1 were analyzed using SCENITH. Gating strategy to distinguish CD8⁺CD45RA⁺CD62L⁺ naive, CD8⁺CD45RA⁺CD62L^- effector (SLECs not included), CD8⁺CD45RA^- memory (primarily CD62L^– effector/memory), and CD8⁺KLRG1⁺CD127^low T cells (SLECs) is shown. (B) Representative histograms showing puromycin levels indicative of cell type–specific translational/metabolic activities. (C) SLEC dependencies after glycolysis (DG), OXPHOS (Oligomycin) or both (DGO) are shown. Cells not treated with puromycin (no puromycin) were used as negative control. (D) Quantification of puromycin levels across CD8⁺ T cell subsets. Bulk CD8⁺ T cells without puromycin treatment served as a negative control. (E) T cells from the m.7076A > G donor were treated with OXPHOS (oligo), glycolysis (DG), or both inhibitors (DGO) to quantify absolute differences in puromycin incorporation to probe relative metabolic dependencies. Bar graphs display the reduction in puromycin incorporation of CD8⁺ T cell subsets in comparison to samples without inhibitor treatment. Data of 3 technical replicates are shown.

Fig. 5.

Mitochondrial ribosome profiling reveals translational stalling of the mutant m.7076A > G allele. (A) Schematic of codon, anticodon, tRNA, and codon:anticodon recognition mechanisms for glycine in the human nuclear and mitochondrial genomes. (B) Polysome profile following sucrose gradient and western blots of isolated fractions for mitochondrial ribosome profiling. MRLP11 and RPS6 were blotted to identify enrichment of mitochondrial and cytoplasmic ribosomes, respectively. (C) Summary of heteroplasmy from ribosome profiling libraries [fractions 5 to 9; see panel (B)] showing a relative increase of the mutant m.7076A > G allele in ribosomal bound fractions versus input RNA. Statistical significance was determined using a Fisher’s exact test of 7076A and G read counts summed between replicates. (D) Schematic of the functional effect of the synonymous m.7076A > G variant. Due to decreased codon:anticodon affinity of the m.7076A > G allele, there is an increase in stalling of the MT-CO1 transcript, prohibiting effective translation.