The already famous Lazaridis et al. study on the contribution of various ancient populations to modern European genetics has gone through a revision which does not alter the fundamental conclusions reached in the past but does add some interesting nuances, new graphs and some new data.
Iosif Lazaridis, Nick Patterson, Alissa Mittnik, et al., Ancient human genomes suggest three ancestral populations for present-day Europeans. BioArxiv 2013 (preprint). Freely accessible → LINK (last version) [doi:10.1101/001552]
Most up to date supplementary info → LINK
For background see this previous entry.
Scandinavian hunter-gatherers deviate towards Siberia
Among the new data (or maybe data I skipped in the first read?) is the fact that the ancient Epipaleolithic individuals from Motala (Sweden) deviate towards Mal'ta-1 (Siberia), something that neither Lochsbour (representing Western hunter-gatherers) nor Stuttgart (representing early farmers) do.
This implies that there were already some differences in the Epipaleolithic era among European hunter-gatherers, with those of Magdalenian background lacking the Siberian (ANE) component, which is found however in Scandinavian ones (of Ahrensburgian background?) This may help explaining the extra ANE affinity in Northwest Europe, which is otherwise hard to understand.
It also suggests that Eastern European hunter-gatherers were already in the Epipaleolithic more akin to Siberian ones than those from the West and South of the subcontinent, as well as those from West Asia (otherwise Stuttgart, which has partial West Asian ancestry would show increased affinity). Of course this can only be confirmed by direct analysis of Eastern European Paleolithic remains but seems quite likely in any case.
Principal component analyses (lots of them!)
Ancient samples projected on the global PCA:
|Fig S1-10: projection of ancient samples onto global PCA dimensions 1 & 2|
In this global PCA, EEFs overlap well with the reduced modern European sample (Basques and Sardinians only) and the West Asian one (Georgian, Palestinian Bedouins). However projected hunter-gatherers from Europe and Siberia show a clear "other Asian" deviation. Why? For the very same reason that South Asians and Melanesians do, even if they are clearly distinct populations: because the frame gives them no other choice: they are not quite like modern Europeans and they do not have any African tendency either, so the other populations that are somewhat akin to them are other Asians and there they go.
That's why PCAs must always be taken with a preventive dose of salt: they are very nice visualization tools but they depend too much on the sampling strategy and its intrisical bidimensional limitations.
Modern West Eurasians projected to a PCA of three ancient samples:
|Fig. S10-3 Projection of West Eurasian populations onto the first two principal components|
inferred using Loschbour, Stuttgart, and MA1 (full version).
Quite curious: all West Eurasians cluster tightly in comparison to their ancient "ancestors". It is likely that dimension 2 should be scaled down because the second component is always smaller than the first one (often around half). However I could not find a clear datum to proceed so I retained the original equal scale even if it can be a bit misleading.
While not exactly, Lochsbour and Stuttgart explain the bulk of European (and West Eurasian!) ancestry, at least in comparison to the quite outlying Mal'ta-1 sample, representing ancient Siberians.
Detail (zoom in) of this PCA:
|Fig. S10-4 (I annotated the three ancestral tendencies with arrows for easy of view)|
As expected Eastern/SE populations deviate more towards Stuttgart, Western/NW do towards Lochsbour and in general Northern populations deviate slightly towards Mal'ta, also in West Asia (i.e. Iranians, Turks).
Modern European PCA with ancient samples projected on it:
|Figure S10.5: Projection of ancient samples onto the “European” PCA (annotations in gray are mine)|
Something the authors notice is that their PCA does not approximate a map of Europe, as happens in some cases. They dedicate some time to evaluate this discrepancy, comparing with the PCA of Novembre et al. 2008. The differences are caused because the latter used many more NW and Central European samples and instead had way too little Eastern European ones.
This highlights that sampling strategy is of utmost importance when analyzing autosomal DNA, not just in PCAs because oversampled populations would tend to cope the axes (or components). This should be obvious but is way too often ignored, what may result in spectacular magic hat tricks but hampers serious science.
This is one of the reasons why I do not trust too much autosomal DNA analysis: lack of the fundamental rigor. This is of course not a defect in this particular study but it happens often in many others, scholarly and amateur alike.
The authors believe that the main axis of differentiation in Europe when the subcontinent is considered as a whole may tend to Northeastern Europe rather than SSE/NNW⁸, something that is consistent with their ancient admixture findings elsewhere in the study.
FineStructure PCA output
A key point in this study is that only Sardinians and Basques can be modeled as simple EEF-WHG admixture, all the rest of Europeans needing of the MA1 component to be explained. This is much easier to visualize in the following graph.
We also processed the ChromoPainter/ChromoCombine output with fineSTRUCTURE1 using 250,000 burnin and 2,500,000 runtime MCMC iterations. Fig. S19.2 shows a Principal Components Analysis by fineSTRUCTURE which strongly resembles that of Fig. 1B.
|Fig. S19-2 (annotated in gray by me)|
Here we can see that only Sardinians (and quite insistently Canarians), Basques and some populations related to these (North Iberians, South French) are actually close to the Stuttgart-Lochsbour axis. All the rest need a third ancestry for explanation, which is approximated by MA1 (not plotted but whose tendency I annotated).
Notice that in this case the two ancient samples are not projected, as in the previous graphs but actually computed as part of the wider West Eurasian population.
This does not deny that other NE European populations have greater affinity to Lochsbour (fig S19-3) but it seems clear that this affinity must be mediated by another branch of ancient European hunter-gatherers, one that existed in Eastern Europe, rather than in the West, and that it had more ancient Siberian (MA1) affinity. It is also likely that this Eastern European aboriginal population was the one which brought the extra MA1 affinity to the rest of Europe, most likely in the context of Indoeuropean (Kurgan) migrations. Along with it they probably also brought extra WHG-like admixture (but actually from an Eastern European source).
The extra MA1 tendency is also present in West Asia. This may have two alternative or complementary explanations:
- There have been also significant Siberian-like intrusions in West Asia after the Neolithic.
- This extra ancient Siberian affinity is in fact (largely?) pre-Neolithic but the founder population of West Asian roots which triggered the European Neolithic in Thessaly was particularly removed from this admixture and more akin to Palestinians or peninsular Arabs than to other West Asians.
Sicilians, Maltese and Ashkenazi Jews are different
These three are the only European populations which have a poor fit with the triple admixture model (EFF+WHG+ANE), suggesting that they have fourth party inputs, most likely extra admixture from West Asia.
This is apparent in the previous graph too (among several).
Tree modeling for the origins of the ancestral populations
|From fig. S16-2 (allowing for five admixture edges)|
|From fig. 16-4 (full-genome coverage, allowing for 5 admixture edges and using Dai instead of Onge)|
The basic topology of the tree is consistent (excepted the partial change of the location of Karitiana Native Americans, which depends on the greater affinity of the East Asian sample used and is essentially neutralized by the admixture edge with Onge or Ma1 respectively). The main admixture events are:
- The Karitiana (and Amerindians by extension) are clearly a mix of East Asians plus Ancient Siberians of Western affinity (MA1), which is represented differently in both trees.
- Early European Farmers (Stuttgart, Iceman) have clear "Basal Eurasian" admixture (which can be interpreted as North or East African input and/or a residual ancient Arabian element, probably both)
- La Braña also has "Basal Eurasian" admixture (surely from NW Africa, what implies that the North African component in Western Iberia is pre-Neolithic)
- Motala has Ancient Siberian (MA1-like) admixture
- Mal'ta 1 probably has some East Asian admixture
- Ötzi the Iceman might have some Western Hunter-Gatherer admixture (~3%)
Of the three Western branches, MA1 is the more distant one. That implies that West Asia and Europe were also exchanging genetics in the Upper Paleolithic, while Siberia remained more isolated in comparison. That stands even when East Asian admixture into MA1 is accounted for.
A Lochsbour's cousin in West Asia
The authors compare and analyze many models of possible admixture leading to the known ancient and modern populations. They seem to favor this one in the end:
This scheme suggests that ancient West Asian ("Near East") populations were closer to ancient Europeans (Lochsbour) than to ancient Siberians (MA1). It also suggests an unknown relative population of Lochsbour (UHG) as partial ancestor of early European farmers (Stuttgart). This population is speculated to have lived in the Balcans.
An issue here is that most modern West Asians and all Caucasian peoples actually have too much MA1 affinity to be a good fit for the (ancient) Near East proto-population concept. About 12-13% among West Asians (Cypriot, Druze) and as much as 29% among Caucasians. They are actually more like "Test" than like "Near East".
The authors conclude that they don't really know if this extra MA1-like ancestry is old or recent. If old, it would imply either two different populations of West Asians or, as they say, the expansion of a West Asian population with extra "Basal Eurasian" ancestry.
This brings us to a key question: what is actually "Basal Eurasian"?
What is "Basal Eurasian"?
Notice that "Basal Eurasian" is defined as phylogenetically intermediate between the Mbuti and Eurasian-plus populations. Quite misleadingly the node is described as "Non-Africans" but that does not need to be true at all. It is just downstream of one of the most ancient African sub-branches, that of Pygmies, so it can still represent African populations which are or were closer to the out-of-Africa branch.
There is no formal ascertainment whatsoever of what is "Basal Eurasian", no comparison with other African populations and no even formal consideration of the (very likely) possibility of various isolate ancient populations existing in NW Africa or Arabia. This is clearly a flaw.
A key piece of information here is that La Braña (ancient NW Iberian hunter-gatherer) consistently shows "Basal Eurasian" admixture. This admixture is much more likely to have arrived from North Africa than anywhere else. NW African genetic markers are still apparent in Western Iberia in fact and there is strong archaeological support for Iberia-NW Africa interaction in Solutrean/Oranian times.
We can only consider in fact this "Basal Eurasian" idea as a mere indicator of African-like affinity, even if it's not Mbuti but something else. This something else can be in fact several things:
- NW African Aterian residual in the case of La Braña
- Arabian OoA residual influence in the case of EEF
- Direct NE African admixture in the proto-EEF West Asian population, strongly indicated by the NE African E1b-M78 (notably its subclade E1b-V13) in ancient Neolithic and modern European Y-DNA.
I am particularly inclined to suspect an almost direct migration from Palestine to Thessaly at the origins of European Neolithic. After all both non-European lineages found in early European farmers (E1b-V13 and G) are common in that area. But of course an Anatolian intermediate station cannot be excluded.
In any case I'd suggest to change the terms "Basal Eurasian" and "non-Africans" by something more neutral, maybe "Ultra-Mediterranean" and "Proto-Eurasian" respectively, where both concepts are allowed to be African, at least partly so.
Update (Apr 23): see also here for some curious aspects of Lochsbour's IBD ancestry.
Update (Apr 23): see also here for some curious aspects of Lochsbour's IBD ancestry.