Was there a Mitochondrial Eve?

One of the serious problems associated with the idea that there was a Mitochondrial Eve (~200.000 years ago) is the unsubstantiated assumption that mtDNA is only inherited matrilineally. Most researchers are concerned with this issue. For example, Hagelberg (2003) wrote:

If recombination happened, there would be no Mitochondrial Eve […]. Recombination would allow male mtDNA lineages to contribute to offspring, and would mean that our current estimates of the age of Mitochondrial Eve are underestimates, when in fact she might be much older than we think. Alternatively, the assumption of a coalescence model would produce younger ages. Although the corrected dates might not contradict the Out of Africa model, they would make mtDNA a worthless source of evidence for this hypothesis.

A new study seems to confirm that mtDNA does recombine. See Perera et al. (2018). Also, creationists keep claiming that measured mtDNA mutation rates are way higher than initially expected. Does this affect the hypothetical age for Mitochondrial Eve in any way?

1 Like

Great question. What do you think @glipsnort and @Joe_Felsenstein?

Recombination and paternal inheritance would mean that there were multiple mitochondrial Eves, one for each stretch of mtDNA, rather than one. Current estimates of the age of mtEve would then be something like a weighted mean of the ages of the different common ancestors. I suppose this could complicate attempts to trace the history of specific haplogroups, but I don’t see why it would have much effect on our understanding of mtEve herself. What am I missing here?


Well, more that that, right? Because current methods don’t take recombination into account, the current estimates for Mt-eve would be too ancient, right?

If anything I would think they would be too recent. Take the two most diverged mt genomes. In the absence of recombination and ignoring sampling error, the number of mutations separating them corresponds to the tMRCA. Now add recombination, but pretend all parts of the mt genome have the same real tMRCA. Now only part of the two most diverged samples coalesce at the true MRCA; in other parts, they coalesce more recently. So there are fewer mutations separating this pair now than there were without recombination, so we underestimate the time to their coalescence.

Im not sure I follow. If you ignore recombination, there will be more homoplastic mutations, which means they will be counted more than once because they appear at multiple points in the tree. That dialates the estimated time to mt-eve.

If you spuriously create a duplicate mutation, you’ve also spuriously create a duplicate branch that the mutation occurs on. That can affect the estimate of intermediate nodes in the tree, but the total branch length from each tip to the root doesn’t change. That is, if two tips should share an internal branch and the mutations on that branch, but you split that into two parallel branches with parallel mutations, you’ll still go through exactly the same branch length from each as you trace back to the root.

I am not sure that is necessarily true. The key thing is that recombination will introduce homoplasies, that will be distributed across the tree, and they can also be very deep in the tree (think ancestral recombination graphs, ARG).

Can we do an experiment to settle this? Perhaps:

  1. Generation sequences using a coalescent simulation with recombination and without recombination.

  2. Infer TMRCA using a phylogenic tree, not allowing for recombination with an ARG.

Hypothesis 1: TMRCA with recombination > TMRCA no recombination
Hypothesis 2: TMRCA with recombination = TMRCA no recombination
Hypothesis 3: TMRCA with recombination < TMRCA no recombination

I am betting on Hypothesis 1. Seems you are betting on Hypothesis 3?

1 Like

I was thinking ARGs. I just wasn’t thinking very clearly about them. Yes, I now see how neglecting recombination could lead to overestimates of the number of mutations and thus the age. It should easy enough to check whether this is important for human mitochondria, though – how many recurrent mutations (at any significant depth in the tree) are reconstructed in standard mt phylogenies.


If this is correct, then previous estimates for the age of mtEve will certainly turn out to be wrong. Is it actually possible to reduce the age to a few thousands of years? Remember, there’s also the issue of a faster mutation rate:


Yes. I agree.


I agree. Even just argweaver on mitochondrial should give us a reasonable starting point.