Can Detrimental Mtdna Mutations Be under Positive Selection in the Germline?

It is conventionally assumed that detrimental mitochondrial DNA (mtDNA) mutations generally are under negative purifying selection in the germline, i.e., that children on average carry a lower burden than mothers. We asked whether, at least for some detrimental mutations, this may not be the case. We chose the infamous 3243G>A mutation, which disrupts mitochondrial function and causes a variety of serious neuromuscular disorders including a severe mitochondrial disease, MELAS. This mutation occurs frequently (detectable in as many as 1 in 500 individuals), and is well studied. To determine whether this mutation is under selection, the mutation level in kids is compared to that in mothers. However, it has been shown that the level of the 3243G>A mutation appears to exponentially decrease with age in blood, where it is commonly measured (~2% per year). This estimation is based on a biologically sensible numerical model that simulates random drift in blood stem cells and removes cells that reach a lethal threshold (Rajasimha, 2008). This model asymptotically converges to an exponential function that describes a decline of 2% heteroplasmy per year. Thus, for fair comparison, kid/mom heteroplasmy ratios need to be corrected for the age difference. With the conventional 2% per year correction, no significant difference between moms and kids could be detected.There are indications, however, that the change of the 3243G>A level with age may be more complex than simple exponent. It may depend on initial heteroplasmy levels and the age. Indeed, we demonstrate that the 2% model predictions are systematically under‐estimates, and that several predictive models that include heteroplasmy and age as variables ensure better fit of the data and decrease this negative bias. One such model is an adaptation of the original numerical model (Rajasimha, 2008). If its asymptotic behavior is not unconditionally assumed, this model explicitly includes heteroplasmy and age as variables. This non‐asymptotic model produces a better fit and corrects the negative bias.We reasoned that a negative bias in prediction might have obscured positive selection on 3243G>A in the germline if it existed. To clarify the issue, we used the non‐asymptotic numerical model to predict the children’s heteroplasmy levels at their mothers’ age. Indeed, in preliminary simulations with this improved correction, a statistically significant positive germline selection of 3243G>A was demonstrated. More analysis is needed to confirm this initial result.We note that, interestingly, if such positive selection of pathogenic mtDNA mutations existed, it might have had an adaptive purpose. Our simulations demonstrate that under realistic conditions, positive selection in the germline may promote removal of detrimental mtDNA mutations at the population level and thus help to evade the dangers of the infamous Muller’s ratchet.