Neanderthals and humans interbred in Europe for 5,400 years

In 2006, when the first two papers analyzing fragments of Neanderthal nuclear DNA sequence were published (Green et al., 2006, Noonan et al., 2006), the discordance between their conclusions suggested that contamination of the samples with modern human DNA remained a serious concern (Wall and Kim, 2007). The current paper soundly demonstrates that the effects of contamination can be brought under control, at least for mtDNA, and that, provided the sequencing can go deep enough in a DNA library of sufficient complexity, the prospects for convincingly determining the true Neanderthal sequence are excellent. The only caveat is that if errors or contaminants were more common than the true Neanderthal sequence, deep sequencing would only find the erroneous consensus. By sequencing to a 35× depth, the current paper not only obtains the desired sequence but also quantifies many aspects of the errors induced by DNA degradation.



Although the mtDNA data suggest that there was no admixing between Neanderthal and modern humans, this conclusion remains controversial. The estimated time of common ancestry between this Neanderthal and modern human mtDNA was 660,000 years ago (Figure 1). Yet there are many nuclear genes in the modern human genome where the best estimates of the time to most recent common ancestry are in the range of 1 to 4 million years ago, leaving open the possibility of admixture with even more ancient hominids (Garrigan and Hammer, 2006). The long period of coexistence of modern humans and Neanderthals, as well as the great depth of common ancestry of modern human nuclear genes, make it quite plausible that there was opportunity for interbreeding (Figure 1). If there had been admixture, say 100,000 years ago, giving modern humans small segregating pieces of our genome with Neanderthal ancestry, it would be nearly impossible to identify them as such, even with full genome sequences.



Whether the sum of the data lies in favor of some admixture or zero admixture remains unclear (Plagnol and Wall, 2006, Fagundes et al., 2007). The observation that mtDNA sequence from Neanderthals lies outside the range of modern human variation was made some years ago (Krings et al., 1997), and the conclusion then was that this was evidence against admixture. But Nordborg (1998)) pointed out that mtDNA follows clonal haploid transmission, and so the genealogy inferred from mtDNA is only one sampling among millions of possible genealogies. Admixture could have easily occurred without leaving any trace in current mtDNA sequences. Moreover, because the mtDNA data are from a single Neanderthal, it remains possible that other samples may produce a different conclusion, especially regarding the time of common ancestry. Also, perhaps the interbreeding was strictly unidirectional; for example, only human female by Neanderthal male matings occurred and never the reverse. This would yield modern humans with admixed nuclear genes but a complete absence of Neanderthal mtDNA. Although the data do not yet allow one to conclude that there was absolutely zero interbreeding, they nevertheless suggest that it was not rampant.



Natural selection is expected to perturb the background pattern of DNA sequence changes differently from the random sampling process of neutral drift, and we now have a plethora of statistical methods for inferring the signature of selection. The Neanderthal mtDNA sequence shows a clear elevation of nonsynonymous (amino-acid altering) changes relative to modern human polymorphisms, a pattern of change that is consistent with some degree of relaxed efficacy of purifying selection in Neanderthals. Green et al. argue that the most likely cause for this pattern is a smaller population size of Neanderthals than that of modern humans, making selection less efficient at removing slightly deleterious polymorphisms from the Neanderthal population. Positive adaptive selection would also accelerate nonsynonymous changes, but the data are strongly consistent with an overall pattern of purifying selection instead. The mitochondrial gene COX2 encoding subunit 2 of cytochrome c oxidase has four amino acid differences between modern human and Neanderthal. But even this substantial change did not attain significance for positive selection. Given the mere 0.5% sequence divergence between modern human and Neanderthal nuclear genes (Noonan et al., 2006), adaptive changes would have to generate a large sequence discrepancy to be detectable.

The production of the complete mtDNA genome sequence of a Neanderthal sample is a grand success, and it whets the appetite for continuing the endeavor to produce a full nuclear genome sequence. Several attributes of the nuclear genome will make this a much less tractable challenge, including its sheer size, its redundancy and heterozygosity, an unknown error model (which will likely differ from that of mtDNA), and the challenge of identifying contaminants when the true divergence is likely only to be 0.5%. But one clear lesson from the new Green et al. study is that brute-force sequence depth can go far to resolve ambiguities and to distinguish the error from the true sequence. Perhaps the most informative attribute of the genetic data for determining whether or not there was human-Neanderthal admixture will come from quantifying patterns of shared polymorphisms between modern humans and Neanderthals. Such population genetic inference would require analysis of multiple Neanderthal samples, so it is fortunate that fossil remains of more than 400 of these beings have been found.