Fig. 1: Cryptic mtDNA mutation is a predominant form of mutation and reaches physiologically relevant levels in later life.

a Cells in tissues carry two mutant types—those common to multiple cells in the tissue and those unique to a single cell. b We exploit scRNA-seq and scATAC-seq on tissues taken from individuals of varying ages and pool variants from every cell allowing us to not only construct an expression matrix, but to examine the site frequency spectrum of a tissue and link changes in expression of single cells to inferred mtDNA mutant load. c Modelling sub-cellular population genetics with an out-of-equilibrium infinite-sites Moran model (left) we can predict how the normalised site frequency spectrum of cryptic mtDNA mutations (right) evolves over a lifetime and predict an accumulation of high heteroplasmy mutations in aged individuals. d Single-cell sequencing is necessary to reveal true mutant load of a tissue. In total, 90.8 % of mtDNA mutations are at a pseudobulk heteroplasmy h < 0.5 % (marked in blue) and so would not be reliably detected in most bulk experiments (data taken from all mutations found in ref. ³⁰). e, f The cryptic site frequency spectrum (cSFS) evolves with time. We see that for 19 individuals across 3 human datasets taken from different tissues and differing sequencing techniques, the rank biserial correlation distance (RBC-difference, a measure of how likely a mutation sampled from one spectrum will have a higher heteroplasmy than a mutation sampled from another) between two spectra increases with the age difference between the individuals (two-sided Spearman correlation r ≈ 0.70 and p < 10⁻²⁶). By looking directly at an example pair of spectra (f) we see that, as predicted by our theory, there is an accumulation of mutants at high heteroplasmies (for a breakdown of the number of cells and mutations represented in the normalise cSFS see Table S5, for the cSFS of all donors see Supplementary Fig. S9). g The mitochondrial load μ^10% of potentially pathogenic, cryptic mutations increases with age in human pancreas cells. We show the mitochondrial load μ^10% for all eight donors in ref. ³⁰ and observe an increase with age (two-sided Spearman correlation r≈0.04 and p < 0.049) and observe an even stronger effect for larger heteroplasmy thresholds (see Supplementary Discussion S8.5) The standard deviation σ(μ^10%) of the mitochondrial load also increases with age (blue asterisks; two-sided Spearman correlation r≈0.88 and p < 0.005).