Earlier mixture from modern humans into Neandertal populations

13 minute read

The Neandertal storyline for the period from 70,000 to 40,000 years ago is increasingly a story of unremitting introgression from Neandertals into other populations wherever they met. Partial ancient genomes from Oase and ‘Ust Ishim show evidence for Neandertal introgression between 60,000 and 45,000 years ago in multiple events; the Denisova genome shows evidence for Neandertal introgression, and of course different legacies of ancient introgression from Neandertals can be found in western and eastern Eurasian populations today.

But while Neandertals were the donors of many genetic exchanges, until now no study had clearly shown them to be the recipients of introgression from other populations. That’s what early studies of Neandertal genomes always showed: We have their genes, but they didn’t get ours. I’ve long thought it unlikely that Neandertals could have existed alongside evolving African, Denisovan, and possibly other populations for a half million years without ever receiving any genes from those other people. Every time new findings of Neandertal introgression have come out, the apparent lack of reciprocity has seemed puzzling.

My assumption has been that weak statistics are a more likely problem than modern human virility. A new paper by Martin Kuhlwilm and colleagues (“Ancient gene flow from early modern humans into Eastern Neanderthals” demonstrates that some modern humans really did contribute their genes to at least one Neandertal population. This Neandertal—the Altai high-coverage genome—was found far from Africa, so evidence for modern human introgression in this genome shows the strength of gene dispersal across substantial geographic distances in archaic humans, at a surprisingly early date, probably more than 100,000 years ago.

The details of proportions and times are subject to a population model, and I don’t trust the specifics very far. We should not forget that the statistics are at the edge of their power to resolve these small proportions of mixture. Their power is maximized under the assumption that the number of events is either zero or one. When the evidence rejects zero, that doesn’t mean that one is the answer. I think the real history was much more complex.

Yet despite its simplification of events, what I think is important in this paper is its clear quantification of the asynchronous and asymmetric nature of gene flow among archaic and modern human populations.

Many researchers had suspected from earlier observations that the gene flow was not symmetric, because no Neandertal samples could be shown to have significant recent African ancestry. But by itself this was not a compelling argument. There was always a perfectly reasonable hypothesis to explain the lack of modern genes in Neandertals: the few Neandertal genome samples may all come from individuals who lived before modern humans came into contact with them. So with previous data, it was possible that gene flow did happen synchronously and symmetrically between modern and Neandertal populations when they came into contact, and we fail to observe the consequences in Neandertals because the Neandertal samples do not represent the period subsequent to introgression.

The current study clearly shows that modern humans and Neandertals were in contact long before the date previously inferred for Neandertal introgression into humans. At least one such contact led to long-distance gene flow through territory thought to have been inhabited by Neandertals.

Ghosts lead the way

Kuhlwilm and colleagues tackled the statistics by poking at a mysterious problem: Why does the high-coverage Denisovan genome look more different from living Africans than does the high-coverage Altai Neandertal genome? The explanation for this offered by Kay Prüfer and colleagues (2014) was that Denisovans have some ancestry from an archaic human “ghost population”, more different from modern humans and the Neandertals than either is from the other. This observation has fueled the speculation that this ghost population might be Asian Homo erectus.

Kuhlwilm and colleagues looked at a different scenario: What if the Neandertal looks more like living Africans because it received some gene flow from early modern humans, who had originated in Africa?

Their further comparisons showed that this hypothesis of modern human introgression does not explain the overall similarity of Neandertal and African genomes. Denisovans do indeed have a ghost population in their ancestry.

But still, there is a residual of the Altai Neandertal genome that is much more similar to Africans, even in comparison to other Neandertal individuals. This aspect of the comparisons pointed to more recent genetic exchanges from Africa into the Altai Neandertal’s population:

We find that windows of the Denisovan genome with high divergence to Africans also have a high divergence to the Altai Neanderthal, whereas windows in the Altai Neanderthal genome with high divergence to Africans do not tend to have a high divergence to the Denisovan (Fig. 1a), consistent with gene flow from a deeply diverged hominin into the Denisovan ancestors. On the other hand, we find that windows of the Altai Neanderthal genome with low divergence to Africans have higher divergence to the Denisovan than Denisovan windows with low divergence to Africans (Fig. 1a). These windows in the Altai Neanderthal genome have higher heterozygosity than in the Denisovan genome (Fig. 1b), and 40.7% of their heterozygous sites share a derived allele with Africans, whereas 24.2% do so in the Denisovan. These observations raise the possibility of gene flow from modern humans into Neanderthals.

So it appeared that both instances of ancient introgression are true. Some modern humans contributed to the Altai Neandertal’s genome, and some very divergent archaic population contributed to the Denisovan genome.

To test this hypothesis more generally, the authors carried out a series of simulations of mixture, including assessments of whether a low level of contamination could explain the results (it doesn’t). In the end, their statistics allowed them to estimate the proportion of ancestry from introgression in several archaic and modern samples, assuming their population model. Their figure 3a summarizes:

Kuhlwilm et al figure 3a showing introgression model
Figure 3a from Kuhlwilm and colleagues (2016), showing population model with introgression and approximate fraction of ancestry from gene flow. Black arrows represent the introgression.

That is a complicated figure, which includes two major new components described in this paper. First, the authors infer that the gene flow from Neandertals into living humans was probably from an ancestral population more similar to the European Neandertal genomes than to the Altai Neandertal. Second, they infer that a population similar to the majority ancestral population of all modern humans contributed between 0.1 and 2.1% of the ancestry of the Altai Neandertal.

The source of Neandertal introgression in living humans

I’ll take the two observations in order. If the gene flow from the modern human population into the Altai Neandertal population was symmetric, then we would expect that the Altai population should have contributed back into human populations.

In their new paper, Kuhlwilm and colleagues base their argument in part upon comparison with the low-coverage genome data from Vindija and El Sidrón:

When we refine our estimates of gene flow by adding the chromosome 21 sequences of the European Neanderthals to our genome-wide data, G-PhoCS infers significant rates of gene flow from Neanderthals into modern humans outside Africa only for El Sidrón and Vindija Neanderthals (0.3–2.6%) (Fig. 3a), suggesting that Neanderthals from Europe are more closely related than the Altai Neanderthal to the population that interbred with modern humans outside Africa 47,000–65,000 years ago.

In other words, Neandertals belonged to at least two relatively divergent populations, one represented by the Altai high-coverage genome, and the other represented by low-coverage genomes from El Sidrón and Vindija. With our present estimations of mutation rates, these populations were as different from each other as the most different living African populations are from each other. This is in turn much more different than the living people of the Altai and Western Europe. Kuhlwilm and colleagues infer that these Neandertal populations had been differentiating from each other for tens of thousands, possibly more than a hundred thousand years.

The genetic heritage that today’s humans have from Neandertals appears more similar to the European Neandertals than to the Altai Neandertal.

Let me unpack that briefly. The statistical procedure here is making the assumption that there is a single source of introgression for modern humans. This is an assumption that I strongly doubt. But if that assumption were true, it should be possible to identify which Neandertal sample was most similar to the source population for this single pulse of interbreeding.

Prüfer and colleagues (2014) included genome sequence from the Mezmaiskaya Neandertal infant in their comparisons, and compared it to both the Altai and Vindija Neandertal sequences. They found that the Mezmaiskaya genome is significantly closer to the inferred source of introgression from Neandertals into humans than the other Neandertals.

The Mezmaiskaya observation seems relatively consistent with the geography and timing; Mezmaiskaya is a relatively late Neandertal specimen from the Caucasus, and lies geographically closer to the Africa-West Asia connection than any other Neandertal with genetic data. In the analysis by Prüfer and colleagues, the Vindija sequence is closer to Mezmaiskaya than either is to the Altai Neandertal, so the results in that study are consistent with the comparisons here by Kuhlwilm and colleagues.

A similar analysis has not yet been carried out for the Oase 2 genomic data, which shows evidence for a very recent introgression from Neandertals only a few generations before the individual lived, around 40,000 years ago (Fu et al. 2015). So although it seems likely that this ancestry might come from a European branch of Neandertals, we do not yet know.

Whenever introgression from Neandertals into living humans has been modeled as a single pulse, the timing of that pulse has appeared to be on the order of 60,000 years ago, with a large confidence region. The genome of the 45,000-year-old ‘Ust-Ishim modern human specimen enabled a similar estimate with higher precision, putting introgression between 7000 and 13,000 years before this individual lived (Fu et al. 2014). Together, the evidence now shows that introgression from Neandertals into modern human populations happened multiple times, and likely in multiple places.

The source and timing of the modern human introgression into the Altai

Immediately following the previous quote, Kuhlwilm and colleagues discuss their estimation of the amount and timing of introgression from modern humans into the Altai Neandertal population.

Conversely, significant rates of gene flow from modern humans into Neanderthals are inferred only into the ancestors of the Altai Neanderthal (0.1–2.1%) (Extended Data Figs 6 and 7). This suggests that modern human introgression into Neanderthals occurred mainly after the divergence of the Altai Neanderthal from El Sidrón and Vindija lineages 110,000 (68,000–167,000) years ago (Fig. 3b). However, it is possible that the lack of complete genomes from the European Neanderthals currently precludes the identification of modern human gene flow into them.

In their comparisons, only the Altai Neandertal gives a significant result for African introgression; no significant introgression from modern humans is observed in the Vindija or El Sidrón genomes. The central Asian population represented by the Altai genome sequence may have been more connected with West Asia in its history.

The lack of evidence for the modern human introgression into the European Neandertals is one piece of evidence for the timing of this gene flow. It must have happened after those Neandertal populations differentiated from each other. A second piece of evidence about the timing comes from comparing African genomes. The authors describe their results:

We applied this method to six African genomes from three different populations (San, Mbuti, and Yoruba) and the two archaic genomes, and estimated the ages of haplotypes for which one archaic genome coalesces within the subtree of the African genomes more recently than it coalesces with the other archaic genome (Fig. 2a, inset). When we compare the age distribution of such ‘African’ haplotypes (≥50 kb), we find that the Altai Neanderthal genome has more young ‘African’ haplotypes (Fig. 2a, left) than the Denisovan genome (P < 0.01; fraction of MCMC replicates). The majority of these young haplotypes are estimated to coalesce with the African genomes 100,000–230,000 years ago, suggesting that they entered into the ancestors of the Altai Neanderthal well before the reported gene flow from Neanderthals into modern humans outside Africa 47,000–65,000 years ago. Both the cumulative and average length of the young ‘African’ haplotypes is longer in the Altai Neanderthal genome than in the Denisovan genome.

Both the Altai Neandertal and Denisovan genomes share many haplotypes with the genomes of living Africans, including haplotypes that are polymorphic within Africa, and many that are rare elsewhere in the world. This kind of sharing is the inevitable result of incomplete lineage sorting from the common ancestors of Africans, Neandertals, and Denisovans.

But haplotype length decreases as time passes since these common ancestors were shared. If there had been no subsequent gene flow from Africans, the Neandertal and Denisovan genomes would have the same distribution of haplotype lengths shared with African genomes, even for that set of haplotypes that are specifically African today, shared with one African genome and not others.

The fact that the Neandertal shares more long haplotype blocks with living Africans is evidence for gene flow between them. The lengths give an estimate of the timing, which seems to have been around 100,000 years ago.

Qafzeh 11 and Qafzeh 9
Qafzeh 11 and Qafzeh 9 skulls and mandibles. Credit: John Hawks CC-BY-NC-ND

A center-of-the-road hypothesis

Let me briefly propose a hypothesis that I’m not sure I fully support, but is the sort of thing many anthropologists might arrive at:

  1. Many anthropologists think that the human remains from Skhul and Qafzeh in Israel, both dating to the interval between 90,000 and 110,000 years, represent a modern human population with substantially African ancestry.

  2. Many anthropologists think that some later skeletal samples in the same region, including skeletal remains from Amud, Dediriyeh, and Kebara, all dating after 90,000 and before 40,000 years ago, represent a population movement of Neandertals into West Asia, that may have supplanted the Skhul-Qafzeh population.

  3. If these assumptions about population affinities were correct, then maybe the modern humans who are found at Skhul and Qafzeh were actually part of a broader dispersal into parts of western Asia, and maybe South Asia, for which we have no skeletal evidence. This population continued to exist in other parts of West Asia even as a more Neandertal-influenced population occupied the Levant. This population was the ancestor of the modern human component of the Altai Neandertal genome.

Now, I do not fully accept propositions 1 and 2. I am not convinced that Skhul and Qafzeh did not represent a substantially admixed population already, and I am not convinced that the Levantine Neandertals are necessarily very Neandertal at all. But even a substantially admixed early population might explain the introgression of genetic material similar to Africans into the Altai.

Functional importance of introgression

Kuhlwilm and colleagues looked at areas within the Altai Neandertal genome that are most similar to African genomes, and are therefore candidates for gene regions that came into the Neandertal population from modern human introgression. They find some interesting hits:

Seven segments exceed 200 kb (Table 2) and the longest one (309 kb) overlaps with a region suspected to have been under positive selection in modern humans3. This region has a transcription factor gene (NR5A2) involved in liver development16. One segment of 150 kb is located within the FOXP2 gene (Table 2), which encodes a transcription factor that may be relevant for language acquisition.

This hit with FOXP2 could be chance, but who knows? Maybe this and other genes important to language today were spreading. If these were not already language-using hominins, such a gene would be unlikely to have any utility in their population.

At the very least, functional analysis may help to think about how the genes of modern humans were dispersing through thousands of miles of Neandertal territory, for tens of thousands of years.


Fu, Q., Li, H., Moorjani, P., Jay, F., Slepchenko, S. M., Bondarev, A. A., ... & Meyer, M. (2014). Genome sequence of a 45,000-year-old modern human from western Siberia. Nature, 514(7523), 445-449. doi:10.1038/nature13810

Fu, Q., Hajdinjak, M., Moldovan, O. T., Constantin, S., Mallick, S., Skoglund, P., ... & Viola, B. (2015). An early modern human from Romania with a recent Neanderthal ancestor. Nature, 524, 216–219. doi:10.1038/nature14558

Kuhlwilm, M et al. 2016. Ancient gene flow from early modern humans into Eastern Neanderthals. Nature doi:10.1038/nature16544

Hershkovitz, I., Smith, P., Sarig, R., Quam, R., Rodríguez, L., García, R., ... & Gopher, A. (2011). Middle pleistocene dental remains from Qesem Cave (Israel). American Journal of Physical Anthropology, 144(4), 575-592. doi:10.1002/ajpa.21446