Monday, November 28, 2011

Ancient DNA Suggests Multiple Neolithic Population Waves In Iberia

We successfully extracted and amplified mitochondrial DNA from 13 human specimens, found at three archaeological sites dated back to the Cardial culture in the Early Neolithic (Can SadurnĂ­ and Chaves) and to the Late Early Neolithic (Sant Pau del Camp). We found that haplogroups with a low frequency in modern populations—N* and X1—are found at higher frequencies in our Early Neolithic population (∼31%). Genetic differentiation between Early and Middle Neolithic populations was significant (FST∼0.13, P less than 10−5), suggesting that genetic drift played an important role at this time.

To improve our understanding of the Neolithic demographic processes, we used a Bayesian coalescence-based simulation approach to identify the most likely of three demographic scenarios that might explain the genetic data. The three scenarios were chosen to reflect archaeological knowledge and previous genetic studies using similar inferential approaches. We found that models that ignore population structure, as previously used in aDNA studies, are unlikely to explain the data. Our results are compatible with a pioneer colonization of northeastern Iberia at the Early Neolithic characterized by the arrival of small genetically distinctive groups, showing cultural and genetic connections with the Near East.

Via Dienekes.

Based on ancient DNA, this study makes three notable claims:

(1) The Cardial Pottery culture had a strong demic component; it cannot be explained by cultural diffusion in the absence of migration.

(2) There was a demographically distinct wave of migration to Iberia after the initial early Neolithic Cardial Pottery wave and before the Bronze Age, i.e. in the Middle Neolithic.

(3) There is a lack of genetic continuity between the early Cardial Pottery culture and modern Iberians (or for that matter, modern Europeans). A model in which modern Iberians are descendants of the first farmers and herders, or of hunter-gatherers who learned to farm and herd from their neighbors and trade partners via cultural diffusion, is wrong.


Maju reports in the comments to the linked post that the full haplogroup breakdown in this sample was "3H, 2N*, 1U5, 1K and 1X1" (and implictly that five samples didn't produce definitive haplogroup identifications) and that "another paper (published in Spanish language only, from 2009 but unknown) on very early Neolithic DNA from Navarre [found mtDNA haplogroups] . . . 3H*, 2H3, 1U, 1K, 1I and 1 HV."

Without considering subtypes and lumping the two samples one gets from a total sixteen data point sample from two contemporaneous but independent sources:

7H, 2N, 2U, 2K, 1I, 1 HV and 1 X1.

The pre-Neolithic origins of the mtDNA U samples in each group is relatively uncontroversial. Maju argues, and Dienkes disputes, that H is pre-Neolithic.

The notion that one can make a p=0.00001 claim about anything with a simple size of thirteen or less is doubtful. Small samples are inherently prone to contain flukes. The margin of error at the 95% confidence level in a sample of this size for a trait upon which there is an even split in the total population is +/- 27 percentage points (i.e. a 50% results is consistent with an actual percentage in the population of 23%-77%).

Also, unlike an ordinary sample populations, ancient DNA samples from a single region are very likely not independent of each other; there is a good chance that some samples are from members of the same family that were buried together because they were members of the same family and that other burial sites from the same community from other families are absent from the sample because they weren't well preserved.

Thus, the fact that a group of haplogroups that are rare today made up 31% of a non-random sample then, could easily be consistent with a frequency in the overall early Neolithic population of well under 10%.

You can be certain that each of the haplogroups in the sample was present at the time. You can be reasonably confident that groupings of haplogroups that are common, if there is any sound reason for the grouping, were not rare at the time. You can say that it is possible that haplogroups that don't show up at all might not have been present at the time and can be more confident in that conclusion if that haplogroup suddenly appears in reasonable frequency at a more shallow time depth and as the total sample size at the older time depth grows. You can't say a whole lot else.

The overlap of mtDNA haplogroups H, U and K at reasonably similar frequencies in two early Neolithic Iberian ancient DNA samples that probably are independent of each other does suggest that there is a good chance that the percentages of mtDNA H, U and K in these samples are not wildly atypical and that N*, I, HV and X1 were all present in the larger early Neolithic Iberian population, probably at more than vanishingly small frequencies.

While the early Neolithic ancient Iberian mtDNA isn't really consistent with modern distributions, since the odds of having that many quite rare haplogroups in a draw from a modern Iberian population wouldn't be that great, the overlap between modern Iberian mtDNA and early Neolithic Iberian mtDNA isn't huge. An identical sized sample from modern Iberians would probably have an overlap of more than 65% in broad haplogroup type with this early Neolithic sample, and finer distinctions could simple represent mutations within a population in continuity with the modern one. I suspect that in this case, the mtDNA is less volatile than Y-DNA over the last 7,500 years or so.

We think we know that the mtDNA X1 v. X2 split dates to a time prior to the arrival of proto-Native Americans in the Americas, thousands of year before the Neolithic revolution. The Druze of the Levant are only modern population of which I am aware that has significant proportions of both X1 and X2. This suggests that the Druze are either descendants of the ancestral X* population, or are an admixed population that had sources in a population with significant X1 percentages and a population with significant X2 percentages in the place where Druze ethnogenesis took place, which by tradition would be in the mountains of Turkey or Iran.

But, tracing an mtDNA lineage to the Near East doesn't necessarily tell you very much about its time depth. It is undisputed that all modern humans in Europe arrived from a source in common with the Near East in the last 40,000 years or so and that there was a major reduction in modern human populations of Europe at the last global maximum around 20,000 years ago. There is little evidence to tell us how much of a demographic influx, if any, Europe experienced between the last global maximum and the early Neolithic, but it wouldn't be controversial to argue that there was at least some population influx into Europe from the Near East in that time period. Mutation rate based dating also provides very little insight into whether a particular haplogroup had Epipaleolithic or early Neolithic origins.

The ancient DNA evidence for anything other than a couple of subhaplogroups of mtDNA haplogroup U (e.g. U4 and U5) prior to the early Neolithic in Europe is very thin, and as Dienekes notes in the comments to the linked post, there is reason to be skeptical of the only unpublished study that has made that claim (as far as I know, ancient DNA samples from North Africa putting other haplogroups commonly found in Europe there as far back as 12,000 years ago are less controversial). Certainly, non-U mtDNA haplogroups were very rare outside Southern Europe for most of the Upper Paleolithic era. But, it is harder to rule out the possibility that there may have been an influx of people more genetically similar to modern Southern Europeans in Southern Europe probably via a mostly coastal route, perhaps 14,000 years ago or so, a relatively late and geographically confined part of the Upper Paleolithic, but also many thousands of years before farmers or herders from the Near East arrived on the scene.

In an ancient mtDNA sample from pre-Roman Iron Age Iberia, (apparently N=17), "[t]he most frequent haplogroup is H (52.9%), followed by U (17.6%), J (11.8%), and pre-HV, K and T at the same frequency (5.9%). No samples were found to correspond to other haplogroups that are widely present in the Iberian peninsula populations (Table 7), such as V, X, I or W. The North African U6 subhaplogroup and Sub-Saharan African L lineages are also absent from the ancient Iberians analyzed so far[.]" More Iberian ancient DNA links can be found here.

The overlap between this sample and the early Neolithic sample is even stronger. The H, U and K percentages in the Iron Age sample are well within sampling error ranges of either of the early Neolithic samples (or both combined), and those three percentages make up 76.4% of the Iron Age sample. This would suggest that the matriline contribution to the Iberian gene pool in the metal ages was pretty modest, despite suggestive evidence that the Y-DNA contribution was greater. Given the proven antiquity of HV in Iberia, the likelihood that pre-HV is also not a relatively recent arrival also seems pretty strong, suggesting an early Neolithic to Iron Age mtDNA shift in Iberia of less than 20% (subject to a significant margin of error).

There was also more mtDNA continuity from ancient mtDNA of the early Neolithic era to the present in Iberia than in Central Europe where the ancient DNA of LBK Neolithic peoples is in stark contrast to that of both modern Central Europeans and Central European hunter-gatherers.

The mtDNA case for a repopulation of Europe after the Last Glacial Maximum from a Southwest European refugia with population expansion taking hold around 14,000 years ago, which is calibrated to a great extent by mtDNA mutation rate dating (which while subject to serious doubts and error bars is certainly more reliable than Y-DNA mutation rate dating) from the modern distributions of haplogroups H and V and their subtypes can be seen here.

Ultimately, the question of how much of Iberia (and Europe's) mtDNA is Epipaleolithic and how much is Neolithic or later, mostly boils down to your assessment of the time when mtDNA H (and probably HV) arrived. If these haplogroups are Epipaleolithic, then 80% or so of Europe's matrilines are ancient; if these haplogroups are Neolithic, then 80% or so of Europe's matrilines were replaced in the Neolithic revolution or afterwards.

The ancient DNA makes clear that H has been in Iberia at least since the early Neolithic, so if it is Neolithic or later, it was there from the start. But, it doesn't tell us if 5500 BCE or 14000 BCE is closer to the time when it arrived.

The mtDNA mutation dating of mtDNA haplogroup H subhaplogroups, however, in the Iberian refugia link above, however, along with the older Upper Paleolithic ancient mtDNA does, however, make a pretty strong case that H's predominance in Europe is at the very least, a post-LGM event, and probably many thousands of years later than that.

It also bears a mention in passing that no modern humans have ever been found to have Neanderthal mtDNA. (Query if populations with high levels of mtDNA haplogroup U5 have elevated Neanderthal DNA percentages.)


Maju said...

Let's see: I "reported" from what was listed in Jean Manco's page, which is available to all. Later I had access to the paper and saw the full sequences and I realized that what I saw in Jean's site was only the two Early Neolithic sites and that two N* sequences were reported as one because they are identical and the same was the case with two H ones from Chaves (although in this case the likelihood of close relatedness is smaller). Lacking were three individuals from Sant Pau, which is a Late Neolithic site and hence listed later in the chronological sequence at Jean's site.

You can see the whole list here (from this entry at my blog covering the whole matter of Neolithic mtDNA south of the Pyrenees).

Maju said...

"... the early Neolithic ancient Iberian mtDNA isn't really consistent with modern distributions"...

True if you consider only this paper but when you consider the following sites/regions, they are quite consistent with modern distributions:

(1) Chaves (Aragon) from this paper: 2 H, 1 K

(2) Paternabidea (Navarre) from Hervalle 2009: 3 H*, 2 H3, 1 U, 1 K, 1 I, 1 HV

(3) Three pooled South Portuguese sites from Chandler 2005: reportedly 17 H, 1 V, 2 U*, 3 U5

Notice that careful examination of the HVS-I haplotypes may cast doubt on many of the reported haplogroups (and that's why they should use coding region testing - but they do not).

So what we see is that the Catalan sites are out of sync with the rest (too many rare haplogroups, even for West Asia or any other conceivable setting).

My thought is that they may represent indeed some sort of wave from Italy (or who knows where?) but that they were eventually displaced later on in the Chalcolithic (Megaliths? but only on the mountain areas) or maybe the Bronze and Iron Ages (Urnfields from the NE?, Iberians from the South?)

Today Catalans are most similar in Y-DNA to Gascons AFAIK but there are way too many blanks in South France to judge properly. Gascony should not be a post-Neolithic genetic source for Catalonia in any case.