David Reich’s lab has a new preprint out, Eight thousand years of natural selection in Europe, which serves as a complement to Massive migration from the steppe is a source for Indo-European languages in Europe. Where the previous work has focused on the relationships of ancient and modern populations, this research puts the spotlight on patterns of natural selection which have shaped ancient and modern populations. The method utilizes the explicit model which is supported by the previous work, that Europeans are best approximated as a three population admixture of a group represented by the hunter-gatherers of Western Europe, the first farmers which brought agriculture to Europe, and the peoples of Central Eurasia which likely brought the Indo-European languages to Europe. In the parlance of these sets of papers, WHG, EFF, and Yamnaya. Basically they have allele frequencies of these ancestral groups, thanks to ancient DNA techniques, and the frequencies in modern populations. By comparing the frequencies one can then infer if the deviations from expectation are large enough to satisfy the conditions you’d expect for a locus subject to a selective sweep of some sort which is changing proportions rapidly as a function of a given selection coefficient.
First, it is very obvious that lactase persistence in Europe has been under strong directional selection over the past 4,000 years. Even in the Bronze Age Central European samples did not exhibit frequencies of the derived variant common across Western and South-Central Eurasia on the LCT locus which is associated with persistence today. A quick survey of the 1000 Genomes data shows that this variant has wide variation in modern European populations which are phylogenetically close. The frequency in the Spanish data set is ~50 percent, but in the Tuscan Italian samples it is ~10 percent for the derived variant. In Denmark and Sweden the derived allele frequency goes up to ~75 percent (the phenotypic expression is dominant, so that means ~95 percent lactase persistence), though in the Finnish sample it is closer to the frequency of the Spanish data set. In South Asia the 1000 Genomes data as well as earlier work shows that frequencies are 25 percent or more in Northwest India, in the Punjab, where dairy culture is most pervasive. It drops as a function of distance from this zone, to 5 percent in the Southern and Eastern South Asia. The haplotype network around this particular mutation implies that it probably originated in Central Eurasia, so the varied frequencies across the Old World is suggestive of both migration and selection. Intriguingly, the lactase persistence allele is not present at appreciably frequencies in the Yamnaya. It begins to appear in cultures such as the
Corded Ware Bell Beaker, though at far lower frequencies than is presently the case in this region.
But the story of lactase persistence is not entirely surprising. Its late evolutionary trajectory in relation to the rise of cattle culture and complex societies in Eurasia points to the reality that evolutionary change in the biological dimension requires a powerful cultural scaffold. That existed in the form of agro-pastoralism in Eurasia. Similar forces are at play across regions of Africa, where signatures of selection are even more evident in groups dependent upon cattle, likely because of the recency of the emergence of the trait, caught in mid-sweep.
A new face in the world?
There are few other signatures evident in these data. Three of them have to do with pigmentation, SLC24A5, SLC45A2, and HERC2. Ewen Callaway reported on the peculiarity last year that Paleolithic European hunter-gatherers may have had dark skin and light eyes. The reasoning here is that a large fraction of the complexion difference between Europeans and Africans is attributable to a derived mutation on SLC24A5, which is nearly fixed in modern Europeans. And yet ancient European hunter-gatherers on the whole were not fixed at this locus, and Western European hunter-gatherers, exhibited the ancestral variants. To get a sense of how peculiar this is the vast majority of the alleles in much of the Middle East are in the derived state, as are about half the alleles in South Asia (I am a homozygote for the derived allele for what it’s worth, and my skin is still notably brown, though obviously not extremely dark). The best available data suggests that the mutant allele emerged recently in the Middle East, and it has expanded out from that point of origin.
SLC45A2 is different in that its distribution is far more constrained to within Europe, though it is found at appreciable frequencies in the Middle East, and at lower frequencies in South Asia. The same for HERC2, though I was surprised to see that the “European” variant associated with blue eye color is actually found at a 0.10 proportion in the 1000 Genomes data in Bangladesh (I am a homozygote for the ancestral variant), the same fraction as the Punjabi sample.*
The results here seem to suggest that all these loci are under selection. The two SLC genes are under positive selection, though SLC24A5 probably got its first boost from EFF with the arrival of agriculture, and was subsequently fixed even when that group fused with the hunter-gatherers who lacked it. Curiously HERC2 is under some negative selection. Remember that all the hunter-gatherers seem to carry the derived variant, so the frequency could only but go down. But in Southern Europe it is likely being driven down in frequency, while it Northern Europe it has been maintained, or rebounded.
Of course one of the major issues we have when evaluating pigmentation loci and their relationship to selection is it’s not always clear if the target of selection is the trait of pigmentation, or something else which the locus modulates, and pigmentation just happens to be a salient side effect. There are many theories about why populations have become depigmented, but none of them are truly well supported in my opinion. Another question is whether we know the genetic architecture of pigmentation well enough to actually infer that these ancient populations are easily predicted in their trait character by modern models which map genotype to phenotype. In other words, were Paleolithic Europeans light skinned because of different alleles? The genetic architecture of skin color is relatively well understood in extant populations. Though it is possible, it so happens that modern Northern Europeans, and to a lesser extent Southern Europeans, harbor a substantial portion of European ancestry which is rooted in the Paleolithic. Studies in admixed African American populations, which are about ~20 percent European, indicate that the primary variants which determine complexion are the ones extant in modern populations, though it may be that there isn’t power to detect the ones from WHG, etc. Of course it could be that the lightening alleles of the Paleolithic Europeans were subject to negative selection, excepting the HERC2/OCA2 locus. But that’s not a particularly parsimonious solution from where I stand (by the way, if selection is targeting something other than pigmentation it is strange that pigmentation associated loci emerge in clusters as positive hits for selection tests).
A secondary issue in relation to pigmentation is that the Yamnaya population does not seem to have been particular fair of hair or azure of eye. The frequency of the derived HERC2 SNP is in the range of North Indian populations, while the SLC45A2 SNP is in the same frequency range as Middle Eastern groups. One might suggest that the Yamnaya are not representative of the population which was intrusive to Europe, but note that the frequencies of the alleles in question during the Late Neolithic and Bronze Age are intermediate between it and modern groups. These results imply in situ evolution within Europe over the Holocene, and down into historical times, toward the phenotype which we ascribe uniquely to Europeans. This is strange especially in light of the fact that a later eastern branch of Indo-Europeans seem to have been quite light. I don’t think we can make final inferences, but to me it is starting to look like the “Proto-Indo-European” complex of peoples was highly cosmopolitan and heterogeneous. Should we expect anything other? As the Mongols expanded in all directions their divergent tendrils were embedded in different ethnic substrate (e.g., Tatars, Khitai and Jurchen in China, Kipchak Turks in Russia, etc.).
The other major locus that showed up was one related to fatty acid metabolism, FADS1. Many tests for selection in humans and domestic animals show changes in the ability to process nutritive inputs. It seems an eminently plausible candidate phenotype to target for selection since the relationship to fitness is straightforward. Using polygenic score methods they also find that there was selection for shorter stature in early Neolithic populations in places like Spain. I think in the future one area of investigation is going to be in the domain of biological adaptations on the margin of farming populations which are put into a Malthusian pressure cooker. Humans, on average, were getting smaller until recently in comparison to their average stature during the Last Glacial Maximum. The Yamnaya people, in contrast to the Neolithic Iberians, seem to have been rather tall. Perhaps it had something to do with the nature of agro-pastoralism? (though do note that without lactase persistence they’d miss out on about 1/3 of the calories in the form of lactose sugar, though not the protein and fat)
But there’s a twist which I haven’t gotten to, and that’s the one in regards to the hunter-gatherers from the Scandinavian region. Unlike the WHG samples you can see that they exhibit mixed frequencies of derived and ancestral alleles at the SLC loci. That’s peculiar, since geographically they are more distant from the core region from which EFF issued. We do know that their ancestry is somewhat exotic, as paper on Indo-European migrations pointed out that they seem to carry the same ancestral component which the Indo-Europeans brought to most of Europe, that of the Ancestral North Eurasians (albeit at far lower fractions than the EHG group which was a partial precursor of the Yamnaya population).
The past is complex and doesn’t fit into a solid narrative. And yet the weirdest aspect of the Scandinavian samples is that they carry the East Asian/Native American variant of EDAR at appreciable frequencies! The figure to the right illustrates this. In blue you have the focal SNP (dark is homozygote, light is heterozygote, dark circle means only one allele was retrieved). In the Chinese from Beijing population (CHB) the derived variant is at high frequency. In the sample of Northwest Europeans from Utah (CEU) it is not present. You can confirm these findings in the 1000 Genomes and elsewhere. In European EDAR of the East Asian form seems only to be found in Finland and associated populations. Using ALDER the authors conclude that admixture occurred on the order of 1 to 2 thousand years before the present, from an East Asian-like group (in the Indo-European paper they found this source best matched the Nganasans of North Central Siberia). An interesting fact which also comes out of this finding is that the haplotype that the derived SNP arose against is relatively common in Northern Europe. The arrows in the figure point to individuals who carry the ancestral SNP, but exhibit the same haplotype which is dominant in East Asia (and also among the Scandinavian hunter-gatherers with the derived variant). The authors state that “The statistic f4(Yoruba, Scandinavian hunter-gatherers, Han, Onge Andaman Islanders) is significantly negative (Z=-3.9) implying gene flow between the ancestors of Scandinavian hunter-gatherers and Han so this shared haplotype is likely the result of ancient gene flow between groups ancestral to these two populations.” Though in earlier work on these data sets they left open the possibility of gene flow between Eastern and Western Eurasia during the Paleolithic as a way to explain some results, it was not offered as a result for the Scandinavian hunter-gatherers. I do not know what to think of the fact that the haplotype that the derived East Asian SNP arose in is common in Northern Europe (though without the derived SNP, which is likely only present in a few populations due to recent Siberian admixture). Could it be that ancient gene flow from Western Eurasian Paleolithic people occurred into East Asian populations, and that then this haplotype accrued the mutation which later swept to near fixation? If that is the case I’m curious about haplotype networks, as Northern Europeans should be more diverse when it comes to the haplotype in question.
In the near future we’ll probably have better and more numerous whole genome sequences of ancient samples. Some of the confusions engendered by this work will be cleared up, as better data renders paradox crisply coherent. The preprint is free to anyone, and I invite readers to dig deeply into it. Though the results yielded only a few positive signals of selection, they’re subtle and complex in their implications. I certainly haven’t thought through everything….
* The fraction of blue eyes is MUCH higher among Punjabis than Bengalis in my experience. It goes to the point that blue eyes likely expresses against the genetic background found in Europeans, where there are other depigmenting alleles near fixation.