A slower mutation rate has implications... and here they are

I've talked about the implications of a slower mutation rate in a few posts in August, and now an Opinion by Scally and Durbin has appeared in Nature Reviews Genetics that spell these implications out.

First, a reminder of what this is all about: researchers have long used an autosomal mutation rate for humans that was based on a calibration of the split between humans and chimps (actually, other assumptions regarding effective population size were involved, but, in any case, a 2.5x10-8/bp/gen rate came to be widely used). More recently, a variety of studies using a variety of techniques (latest one from ASHG 2012) all came up with a rate that is about half that value.

A smaller mutation rate implies more time to accumulate an observed level of sequence difference. We obtain age estimates by dividing sequence difference by mutation rate, and halving the latter means doubling age estimates.

Scally and Durbin work out precisely these implications. From their summary:

The four key points may be summarized as follows. First, the divergence between modern humans and both Neanderthals and Denisovans, which was originally estimated to be 272,000–435,000 years ago, is revised to 400,000–600,000 years ago. This is in better agreement with the range of estimated split times from mtDNA and also with the idea that the ancestral population of these groups may have been H. heidelbergensis. Second, for the split between the Khoe–San and other modern humans, revised estimates from nuclear genomic data suggest a divergence 250,000–300,000 years ago, older than single locus estimates for the root of the human tree. Third, revised estimates of the separation time between Africans and non-Africans suggest that this predates the appearance of modern humans in Europe and Asia by up to 60,000 years. We have suggested a scenario of exodus from Africa via an intermediate population in East Africa and the Middle East, which may fit better with growing evidence for modern human occupation of the latter region before the wider colonization of Eurasia and may provide a longer interval for Neanderthal admixture with non-African populations. Finally, revised split times of 40,000–80,000 years ago for Europeans and Asians agree better with the palaeoanthropological record and with estimates from mtDNA. 

The authors compare autosomal with mitocondrial dates in their paper. But, mitochondrial dates are not independent: the state of the art time-dependent mtDNA clock used in the most recent studies has been indirectly calibrated with a young Human-Chimp split, and, subsequently the mtDNA clock may need updating itself. So, both mtDNA Eve and L3 itself may become older, although it is not clear whether this will be "enough":  mtDNA exhibits a non-linear mutation rate (saturation), so the mitochondrial clock needs to be examined anew, but I think that older dates are a certainty. How much older will depend on a potentially older human-chimp divergence, as well as where on the saturation curve an age estimate is.

Second, if divergence is pushed to ~300 thousand years, then we're either (a) underestimating the appearance of modern humans in Africa, or (b) the higher divergence dates may be driven, in part, by archaic admixture, if some African populations with shallow divergence times mixed with other African populations with very deep divergence times (archaic admixture). I am fairly convinced on the basis of population-specific SNPs that some such admixture has taken place. Some of the remarkable excess of polymorphism and rare haplotype clusters in Africans may be a legacy of such events.

Finally, a pre-100ka divergence of Africans and Eurasians agrees well with the first appearance of modern humans in Asia (the Qafzeh/Skhul hominins from the Levant), as well as the appearance of the Nubian complex in south Arabia. I have proposed a theory that seems to match the evidence: the "two deserts" theory that derives Eurasians from an Out-of-Sahara/North Africa migration prior to 100 thousand years when modern humans first settled in the Levant and Arabia. However, they remained "bottled up" there by the Nearndertals to their north before eventually venturing out to populate the rest of  Eurasia.

At c. 70ka, the drying up of the Sahara and Arabia and/or the Toba eruption, became responsible for the Great  Eurasian Bottleneck that is so often detected in genetic data. This event led some modern humans (Y-haplogroup: CF) to expand north Out-of-Arabia, where they intermixed with Neandertals and eventually "broke through" the Neandertal barrier to populate the rest of Eurasia. A second group of humans (Y-haplogroup: DE) expanded southwest (back-migrants into Africa, carrying Y-haplogroup E), or southeast (Y-haplogroup D, now having a relic distribution in Asia).

In conclusion, I think we are at the beginnings of a new synthesis that will see a lot of dates multiplied by two, and a few human evolution textbooks re-written.

UPDATE (12 Sep 2012): Co-author Aylwyn Scally communicated the following with respect to the need to re-calibrate mtDNA ages in view of the new autosomal rate (and hence Human-Chimp speciation):

mtDNA studies have tended to use 6-6.5 Mya as a value for the human-chimp TMRCA at the mtDNA locus. As you know, the TMRCA or sequence divergence time predates the speciation time by an amount which is proportional to the ancestral effective population size. (And in fact the mtDNA calibrations were already at odds with the previous nuclear genomic estimates, which had a low speciation time of 4 Mya, but I guess they didn't believe such low estimates!) Now with a revised nuclear mutation rate we estimate a human-chimp speciation around 6-7 Mya. The ancestral human-chimp effective population size seems to have been much larger than that of modern humans, so whereas two modern humans typically have a common mtDNA ancestor perhaps 150,000 years ago, for two individuals back then it might have been as long as 0.5 Mya. But even so it means we expect a the human-chimp mtDNA TMRCA to be only 0.5 My earler than the human-chimp speciation, i.e. 6.5-7.5 Mya. If the old calibration point was 6 Mya, the new one might be as much as 25% more, but not double. 

In other words, the calibration point for mtDNA isn't changed much, because they were already implicitly using an older human-chimpanzee speciation date. If the mtDNA dates remain discordant with nuclear ones even after we analyse more whole genome data, we should probably consider other possible explanations, some of which we discuss in the paper.

Nature Reviews Genetics, advance online publication, Published online 11 September 2012 | doi:10.1038/nrg3295

Revising the human mutation rate: implications for understanding human evolution

Aylwyn Scally & Richard Durbin

Abstract

It is now possible to make direct measurements of the mutation rate in modern humans using next-generation sequencing. These measurements reveal a value that is approximately half of that previously derived from fossil calibration, and this has implications for our understanding of demographic events in human evolution and other aspects of population genetics. Here, we discuss the implications of a lower-than-expected mutation rate in relation to the timescale of human evolution.

Link