mtdna N11 - 76,000 years old ?

[TuaMan]

So I recently stumbled upon this very interesting paper, Carriers of Mitochondrial DNA Macrohaplogroup N Lineages Reached Australia around 50,000 years ago following a Northern Asian Route (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4460043/). It's actually a great read, and along with two other related papers covering mtdna M (https://www.ncbi.nlm.nih.gov/pubmed/27832758) and mtdna R (https://www.ncbi.nlm.nih.gov/pubmed/28535779), seems to make a convincing case against the southern coastal route hypothesis for the OOA migration - from the Horn of Africa across the Bab-el-Mandeb, over the southern coast of the Arabian peninsula through India and onto Southeast Asia and Australasia. Instead all three papers argue for a dispersal over the Sinai through the Levant and Central Asia and into Southeast Asia and Australia.

Leaving the issue of OOA via Sinai or OOA via Bab-el-Mandeb aside, I immediately noticed something peculiar in the "Results" section of the mtdna N paper:


Coalescence ages, based on complete mtDNA sequences, for the main branches of macrohaplogroup N(xR), and their present-day geographic distributions are shown in Table 1 and S1–S2 Figs. Haplogroup N11 presents the oldest divergence (around 76 kya) with two main branches, N11a and N11b. N11a is spread in central, western China and Inner Mongolia, and also in southern China and in Makatao from Taiwan [39–42], whereas N11b is found in Philippines [43,44].The second most ancient lineage is N10 (around 66 kya) being mainly detected in southern China, the Tibet and in Lingao from Hainan [39,45].

So according to the paper, N11 and N10 are extremely old lineages, but their estimates seem to contradict the age estimates in the author's own paper for L3 and N as a whole. According to Table 1 in the paper, L3 is around 70,000 years old, so how could they estimate N11 at 76 kya, when N as a whole is supposed to be a daughter of L3? Likewise they estimates N(xR) coalescence at 60,000 kya, much younger than their estimates for either N11 or N10.

To make things even more confusing I tried to go on Phylotree (http://www.phylotree.org/tree/index.htm) to look up the most recent version of the mtdna tree and it seems to have three separate branches under L3 - an apparently African specific L3*, and then M and N branches. If African branches of L3 form their own specific clade to the exclusion of M and N, with all three being nodes under a sort-of macro-L3 parent, then I can understand the seeming contradiction in dates in Table 1 of the paper if they were specifically referring to the coalescence of extant African L3. However, that seems like it would be a paradigm shifting realization in our understanding of the whole OOA event, and I've not seen anyone on this forum or elsewhere make this type of claim, so I'm trying hard to reconcile these apparent contradictions.

I tried searching the forum to see if this paper ever made any kind of splash when it was released and it seems to have flown under the radar. Hopefully someone more knowledgeable than me can try and make sense of some of the questions posed here...

[Megalophias]

L3*(xM, N) do not form a clade, they are just grouped together on the main PhyloTree page for convenience. If you look on the L3* page there are 7 primary L3 branches.

The actual TMRCA of N11 must be younger than that of N and L3, but the date given is only the central estimate. The 95% confidence interval for L3 in that paper is 52.7 to 88.1 thousand years, and of N11 is 48.4 to 104.9 thousand years, so the greater age of N11 is not significant. The central estimate for N11 is almost certainly much too high, but that is expected to happen from time to time. In any given TMRCA estimate the central value is unlikely to be correct. What this study was trying to show is that the northern subclades of N had significantly greater TMRCAs than the southern subclades.

[TuaMan]

Thank you, that makes more sense. I was aware of how the confidence intervals could overlap, but I wasn't sure how to take the high central estimates, since I don't think I've seen such a contradictory estimate before relative the the rest of a haplogroup's tree. But if you can occasionally get weird one-off values like with N10 or N11 in this paper, then that's good to know, so moving forward you know not to read too much into it if you see the same scenario in different papers.

However, one other thing I've noticed when it comes to papers discussing mtdna is that whenever they speak about the "coalescence" of a particular haplogroup or clade, they mean the TMRCA. This is distinct from the Yfull tree (https://www.yfull.com/tree/), which makes a distinction between the formation of a haplogroup and its TMRCA, between which there can be a gulf of tens of thousands of years. Why is this distinction rarely made explicit in discussions with mtdna?

The MRCA of L3 is 70,000 kya, and the MRCA of M and N are 60,000 kya. M and N both derive directly from basal branches of L3, so we can infer that those branches actually formed nearer to the 70,000 kya MRCA of L3, correct? However, how do we know when L3 actually formed, and not just its TMRCA?

[Saetro]

However, one other thing I've noticed when it comes to papers discussing mtdna is that whenever they speak about the "coalescence" of a particular haplogroup or clade, they mean the TMRCA. This is distinct from the Yfull tree (https://www.yfull.com/tree/), which makes a distinction between the formation of a haplogroup and its TMRCA, between which there can be a gulf of tens of thousands of years. Why is this distinction rarely made explicit in discussions with mtdna?

I think this is a function of what you read first.
mtDNA research seemed to take off earlier, and I certainly read papers on it way before YFull existed.
Others have suggested elsewhere in this forum that YFull's usage differs from what they have seen elsewhere.

The important point here is that DNA is an exciting research frontier. Lots of things are happening. People are struggling to make sense of them and to communicate their insights to others.
Some invent, re-invent or modify terms to aid their own understanding, and/or to better communicate an idea to others.
In some cases terms that have established meanings either in the field or in everyday use are used quite differently - even to mean the exact opposite of their usual meaning.
So it can be good to check terms. Often you have to sort this out yourself. YFull actually tells you what they mean.

Your question is a good one, and this forum has some knowledgeable people who can provide good responses.
I also like to see what other scientists are saying about papers like this Fregel 2015 one.
(Although some responses can be just defending their own turf.)
A Google Scholar search on people who cited this paper brought up some interesting comments. (Search the paper, then look for the comment "cited by" and click on that.)
Marrero 2016 provides work on Haplogroup M that may help explain this result for N - for example. https://bmcevolbiol.biomedcentral.com/articles/10.1186/s12862-016-0816-8?utm_medium=email&utm_source=flipboard

[Terminus88]

N comes from Haplogroup O.

[GailT]

However, one other thing I've noticed when it comes to papers discussing mtdna is that whenever they speak about the "coalescence" of a particular haplogroup or clade, they mean the TMRCA. This is distinct from the Yfull tree (https://www.yfull.com/tree/), which makes a distinction between the formation of a haplogroup and its TMRCA, between which there can be a gulf of tens of thousands of years. Why is this distinction rarely made explicit in discussions with mtdna?

mtDNA has a highly variable mutation rate and there is no way to predict how quickly or slowly the mutations occurred that separate a daughter clade from its parent clade. Behar has and age estimate of about 67000 years for L3 and 50000 years for N, and there are 5 mutations that distinguish N from L3 (G8701A C9540T G10398A C10873T A15301G! ). Unless we discover new samples (ancient or modern) that are intermediate between L3 and N, there are no data that can be used to estimate when each of these mutations occurred or in what order they occurred. You can guess that they might have occurred at approximately the average mutation rate, and that N began to diverge from L3 about 64000 years ago, but that would only be a guess and might not be accurate.