September 13, 2009

On the mtDNA molecular clock (Endicott et al. 2009)

This is an excellent paper to read on the technicalities and controversies surrounding the human mtDNA clock. The authors argue against over-reliance on the human chimpanzee splitting as a calibration point, both because of uncertainty about the age of that event, and also because of limitations of the ρ statistic which is widely used to associate the accumulation of genetic diversity with time (e.g., its performance under selection). They further argue that existing mtDNA age estimates are largely over-estimates for such events such as sapiens-Neanderthal divergence, and the arrival of modern humans in various parts of the world. Their concluding remarks:
Further research is needed to improve our confidence in
molecular estimates of human evolutionary timescales.
First, the most reliable calibrations within the human tree
need to be identified. For mitochondrial DNA, this depends
on finding well-defined haplogroups that can be precisely
associated with dated palaeoanthropological evidence [17].
Second, the variation in observed rates across different
timescales needs to be accurately quantified [16–18].
Third, these patterns of rate variation need to be investigated
for nuclear data, including the Y-chromosome and
short tandem repeats.
The chief recommendation arising from the current
state of knowledge in the field is for a movement away
from reliance on the human-chimpanzee calibration;
instead, calibrations within the human tree are preferred
(but see [14]). There are several recent examples of estimates
made using archaeological calibrations [15–17,35],
extending the efforts of earlier authors [3,60]. Considering
recent advances in phylogenetic methodology, there is now
a compelling motivation to employ statistical models that
take into account rate heterogeneity among sites and
among lineages, that correct for multiple substitutions
(saturation), and that incorporate directly the uncertainty
in the ages of calibrations used. Some methods also allow
the statistical evaluation of competing demographic
models, which can have an important influence on estimates
of rates and timescales [17,23].
I have my own misgivings about the possibility of archaeologial calibration of the mtDNA clock. Archaeology provides us with evidence that the arrival of modern humans in a part of the world could not have been later than X years ago, the age of the earliest archaeological or osteological finds.

But, it does not really tell us how many of them arrived, or what their own mtDNA time depth was: if a small number of migrants arrive, it is possible that either their own common ancestor lived fairly close to the time of migration, or that only one of them -due to genetic drift in the small population- will leave matrilineal descendants. In other words, if a small number of migrants participates in the early colonization of a region, chances increase that their mtDNA time depth will be close to the time of their arrival; conversely, known time of arrival -from archaeology- calibrates the molecular clock. Indeed, if the population stays small for a long time after arrival, the common matrilineal ancestor may "reset" several times, and the population's antiquity (based on mtDNA diversity) will appear to be much younger than it really is.

However, if the number of migrants is not small, then in all likelihood the common ancestor precedes the migration substantially, and calibration of the molecular clock by the visible migration would lead to an overestimate in the rate in which mtDNA diversity accumulates, and a molecular clock that produces more recent ages than the true ones.

Many existing works make the assumption of neutrality about mtDNA evolution in humans. This means that no mtDNA lineage has an advantage over any other; and subsequently, the fact that we are all descended from a relatively small number of "mothers" (like the "Daughters of Eve") becomes difficult to explain. Massive disappearance of other lineages (besides the few surviving mothers) is only possible under conditions of strong genetic drift in small populations. Hence, the conclusion, reiterated time and again in the literature about humans being reduced to a few hundred or a few thousand individuals, which has sparked the new mythos of a "small band of humans surviving to colonize the entire world".

In reality, our descent from a small number of "mothers" can be reconciled with a large human population under the assumption that mtDNA is under substantial natural selection. If that is the case, the limited number of surviving lineages is not due to drift in a small population, but to selection in a large one.

My personal guess is that the molecular clock won't be calibrated by reliance to archaeology, but by improvements in the affordability of sequencing. At present it is not really affordable to do full mtDNA genome scans in a few thousand mother-daughter pairs to obtain reliable mutation rate estimates, but this is likely to eventually change, leading to better estimates of the splitting times of various mtDNA lineages.

Some previous topics on the question:
Trends in Ecology and Evolution doi:10.1016/j.tree.2009.04.006

Evaluating the mitochondrial timescale of human evolution

Phillip Endicott, Simon Y.W. Ho, Mait Metspalu and Chris Stringer


Different methodologies and modes of calibration have produced disparate, sometimes irreconcilable, reconstructions of the evolutionary and demographic history of our species. We discuss how date estimates are affected by the choice of molecular data and methodology, and evaluate various mitochondrial estimates of the timescale of human evolution in the context of the contemporary palaeontological and archaeological evidence for key stages in human prehistory. We contend that some of the most widely-cited mitochondrial rate estimates have several significant shortcomings, including a reliance on a human-chimpanzee calibration, and highlight the pressing need for revised rate estimates.



Maju said...

I appreciate this meditation, both by the researchers and yourself.

One of the items (from you) that caught my eye is: ... if the population stays small for a long time after arrival, the common matrilineal ancestor may "reset" several times, and the population's antiquity (based on mtDNA diversity) will appear to be much younger than it really is. What surprised me favorably considering that you are an enthusiast of recentist timescales.

From the researchers, I'd emphasize: The chief recommendation arising from the current state of knowledge in the field is for a movement away from reliance on the human-chimpanzee calibration...

I have dealt once on this issue (comment on Caswell 2008) and the conclusions seem that the genetic divergence of Pan and Homo is at least 15% older. This considering the largest usually managed date (7 million years - you can also read 5 million!) and the minimal extension needed to make it fit with the actual Chimp/Bonobo split and the formation of the river Congo (the physical barrier separating both species). With the occasional 5 million y.a. false reference and with a less conservative assunption for the bonobo-chimp split, the estimates could be even 100% larger.

All that assuming that the MC is really worth anything.

Ponto said...

The dates must comply with other data no matter how flimsy. For Europe, dates of occupation of AMH has to be tempered by actual archeological and fossil evidence with correct date estimations.

The populations of hominins, Neanderthal or AMH was always very small and subject to many bottlenecks, and inbreeding events. Naturally using the existing mtDNA or Y chromosome STRs is subject to many assumptions and errors, some age estimation has to be used. Finding more ancient dna for testing, easier with mtDNA, will help correlate dates.