by David Reich
So we now have access to genome-wide data from four highly divergent human populations that all likely had big brains, and that were all still living more recently than seventy thousand years ago. These populations are modern humans, Neanderthals, Siberian Denisovans, and Australo-Denisovans. To these we need to add the tiny humans of Flores island in present-day Indonesia—the “hobbits” who likely descend from early Homo erectus whose descendants arrived at Flores island before seven hundred thousand years ago and became isolated there by deep waters.17 These five groups of humans and probably more groups still undiscovered who lived at that time were each separated by hundreds of thousands of years of evolution. This is greater than the separation times of the most distantly related human lineages today—for example, the one highly represented in San hunter-gatherers from southern Africa and everyone else. Seventy thousand years ago, the world was populated by very diverse human forms, and we have genomes from an increasing number of them, allowing us to peer back to a time when humanity was much more variable than it is today.
How Archaic Encounters Helped Modern Humans
What is the biological legacy of the interbreeding between modern humans and Denisovans? The highest proportion of Denisovan-related ancestry in any present-day population is found in New Guineans and Australians and the people to whom they contributed ancestry.18 However, once we obtained better data and used more sensitive techniques, we found that there is also some Denisovan-related ancestry, albeit far less, even in mainland Asia,19 and it is from the mainland that we have a clue about its biological effects.
The Denisovan-related ancestry in East Asians is about a twenty-fifth of that seen in New Guineans—it comprises about 0.2 percent of East Asians’ genomes, rising to up to 0.3–0.6 percent in parts of South Asia.20 We have not yet been able to determine if the Denisovan-related ancestry in mainland Asia and the islands off Southeast Asia comes from the same archaic population or from different ones. If the ancestry comes from very different sources, we would be detecting yet another instance of archaic human interbreeding with modern humans. But whatever its origin, the Denisovan interbreeding was biologically significant.
One of the most striking genomic discoveries of the past few years is a mutation in a gene that is active in red blood cells and that allows people who live in high-altitude Tibet to thrive in their oxygen-poor environment. Rasmus Nielsen and colleagues have shown that the segment of DNA on which this mutation occurs matches much more closely to the Siberian Denisovan genome than to DNA from Neanderthals or present-day Africans.21 This suggests that some Denisovan relatives in mainland Asia may have harbored an adaptation to high altitude, which the ancestors of Tibetans inherited through Denisovan interbreeding. Archaeological evidence shows that the first inhabitants of the Tibetan high plateau began living there seasonally after eleven thousand years ago, and that permanent occupation based on agriculture began around thirty-six hundred years ago.22 It is likely that the mutation increased rapidly in frequency only after these dates, a prediction that will be possible to test directly through DNA studies of ancient Tibetans.
Interbreeding with Neanderthals helped modern humans to adapt to new environments just as interbreeding with Denisovans did. We and others showed that at genes associated with the biology of keratin proteins, present-day Europeans and East Asians have inherited much more Neanderthal ancestry on average than is the case for most other groups of genes.23 This suggests that versions of keratin biology genes carried by Neanderthals were preserved in non-Africans by the pressures of natural selection, perhaps because keratin is an essential ingredient of skin and hair, which are important for providing protection from the elements in cold environments such as the ones that modern humans were moving into and to which Neanderthals were already adapted.
Superarchaic Humans
Given that Denisovans and Neanderthals are genetically closer to each other than either is to modern humans, it would be reasonable to expect them to be equidistantly related to present-day populations that have not received genetic input from either of these archaic populations—that is, to sub-Saharan Africans. Yet we found sub-Saharan Africans to be slightly more closely related to Neanderthals than to Denisovans.24 This must reflect another example of interbreeding we didn’t know about. The pattern we observed could only be explained by Denisovan interbreeding with a deeply divergent, still unknown archaic population—one from which Africans and Neanderthals have little or no DNA, and which separated from the common ancestors of modern humans, Neanderthals, and Denisovans well before their separation from each other.
The evidence for an unknown archaic contribution to Denisovans is that at locations in the genome where all Africans share a mutation, the mutation is more often seen in Neanderthals than in Denisovans. Because these are mutations that all Africans carry, we know that they occurred long ago, as it typically takes around a million years or more in humans for a new mutation not under natural selection to spread throughout a population and achieve 100 percent frequency. The only way to explain the fact that Denisovans do not also share these mutations is if the ancestors of the Denisovans interbred with a population that diverged from Denisovans, Neanderthals, and modern humans so long ago that nearly all modern humans carry the new mutation.
By examining mutations that occur at 100 percent frequency in present-day Africans, and measuring the excess rate at which they matched the Neanderthal over the Denisovan genome, we estimated that the unknown archaic population that interbred into Denisovans first split off from the lineage leading to modern humans 1.4 to 0.9 million years ago and that this unknown archaic population contributed at least 3 to 6 percent of Denisovan-related ancestry. The date is shaky, as knowledge of the human mutation rate is poor. However, even with the uncertainty about the mutation rate, we can estimate relative dates reasonably well, and we can be confident that this previously unsampled human population split off at about twice the separation time of Denisovans, Neanderthals, and modern humans. I think of this group as “superarchaic” humans, as they represent a more deeply splitting lineage than Denisovans. They are what I call a “ghost” population, a population we do not have data from in unmixed form, but whose past existence can be detected from its genetic contributions to later people.
Eurasia as a Hothouse of Human Evolution
From a combination of archaeological and genetic data, we can be confident of at least four major population separations involving modern and archaic human lineages over the last two million years.
The skeletal evidence shows that the first important spread of humans to Eurasia occurred at least 1.8 million years ago, bringing Homo erectus from Africa. The genetic evidence suggests that a second lineage split from the one leading to modern humans around 1.4 to 0.9 million years ago, giving rise to the superarchaic group that we have evidence of through its mixture with the ancestors of Denisovans and that plausibly contributed the highly divergent Denisovan mitochondrial DNA sequence that shares a common ancestor with both Neanderthals and modern humans in this time frame. Genetics also suggests a third major split 770,000 to 550,000 years ago when the ancestors of modern humans separated from Denisovans and Neanderthals, followed by Denisovans and Neanderthals from each other 470,000 to 380,000 years ago.
These genetic dates depend on estimates of the mutation rate and will change as those estimates become more exact. It is easy to get ensnared in trying to establish neat correlations between genetic dates and the archaeological record, only to have dates shift when a new genetic estimate of the rate of occurrence of new mutations comes along, causing the whole intellectual edifice to come tumbling down. However, the order of these splits and the distinctness of the populations can be determined well from genetics.
The usual assumption is that all four of these splits correspond to ancestral populations in Africa expanding into Eurasia. But does this really have to be the case?
The argument that modern humans radiated from Africa comes from the observation t
hat the most deeply divergent branches among present-day humans are most strongly represented in African hunter-gatherers (such as San from southern Africa and central African Pygmies). The oldest remains of humans with anatomically modern features are also found in Africa and date to up to around three hundred thousand years ago. However, the genetic comparisons of present-day populations that point to an origin in Africa can only probe the population structure that has arisen in the last couple of hundred thousand years, the time frame of the diversification of the ancestors of present-day populations. With ancient DNA data in hand, we are confronted with the observation that of the four deepest human lineages from which we have DNA data, the three most deeply branching ones are represented only in human specimens excavated from Eurasia: the Neanderthals, the Denisovans, and the “superarchaic” population that left traces among the Siberian Denisovans.
Part of the reason we detect the oldest splitting lineages in Eurasians may be what scientists call “ascertainment bias”: the fact that almost all ancient DNA work has been done in Eurasia rather than in Africa, and so naturally that is where new lineages have been discovered. Perhaps if we had as many archaic ancient DNA sequences from Africa as we do from Eurasia, we would find lineages there that split from modern humans and Neanderthals even more deeply in time than the superarchaic.
But another possibility suggests itself, which is that the ancestral population of modern humans, Neanderthals, and Denisovans actually lived in Eurasia, descending from the original Homo erectus spread out of Africa. In this scenario, there was later migration back from Eurasia to Africa, providing the primary founders of the population that later evolved into modern humans. The attraction of this theory is its economy: it requires one less major population movement between Africa and Eurasia to explain the data. The superarchaic population and the ancestral population of modern humans, Denisovans, and Neanderthals could both have arisen within Eurasia, without requiring two further out-of-Africa migrations, as long as there was just one later migration back into Africa to establish shared ancestry with modern humans there.
Figure 11. Can the modern human lineage have sojourned for hundreds of thousands of years outside Africa? Conventional models have the human lineage evolving in Africa at all times. To explain current skeletal and genetic data, a minimum of four out-of-Africa migrations are required. However, if our ancestors lived outside Africa from before 1.8 million years ago until up to three hundred thousand years ago, as few as three major migrations would be required.
An argument from economy is not a proof. But the bigger point is that the evidence for many lineages and admixtures should have the effect of shaking our confidence in what to many people is now an unquestioned assumption that Africa has been the epicenter of all major events in human evolution. Based on the skeletal record, it is certain that Africa played a central role in the evolution of our lineage prior to two million years ago, as we have known ever since the discovery of the upright walking apes who lived in Africa millions of years before Homo. We know too that Africa has played a central role in the origin of anatomically modern humans, based on the skeletons of humans with anatomically modern features there up to around three hundred thousand years ago, and the genetic evidence for a dispersal in the last fifty thousand years out of Africa and the Near East. But what of the intervening period between two million years ago and about three hundred thousand years ago? In a large part of this time, the human skeletons we have from Africa are not obviously more closely related to modern humans than are the human skeletons of Eurasia.25 Over the last couple of decades, there has been a pendulum swing toward the view that because our lineage was in Africa before two million years ago and after three hundred thousand years ago, our ancestors must always have been there. But Eurasia is a rich and varied supercontinent, and there is no fundamental reason that the lineage leading to modern humans cannot have sojourned there for an important period before returning to Africa.
The genetic evidence that the ancestors of modern humans may have spent a substantial part of their evolutionary history in Eurasia is in fact consistent with a theory advanced by María Martinón-Torres and Robin Dennell.26 Theirs is a minority viewpoint within the fields of archaeology and anthropology, but a respected one. They argue that humans they call Homo antecessor, found in Atapuerca, Spain, and dating to around one million years ago, show a mix of traits indicating that they are from a population ancestral to modern humans and Neanderthals. This is a very ancient date for a modern human/Neanderthal ancestral population to exist in Eurasia. Many who think that Neanderthals in Europe descend from an out-of-Africa radiation of an ancestral population would assume that the ancestors of both populations were still in Africa at that time. Combining this evidence with archaeological analysis of stone tool types, Martinón-Torres and Dennell argue for the possibility of continuous Eurasian habitation from at least 1.4 million years ago until the most recent common ancestor of humans and Neanderthals after eight hundred thousand years ago, at which point one lineage migrated back to Africa to become the lineage that evolved into modern humans.27 The Martinón-Torres and Dennell theory becomes more plausible in light of the new genetic evidence.
Part of the “out of Africa” allure is the simplifying idea that Africa—and especially East Africa—has always been the cradle of human diversity and the place where innovation occurred, and that the rest of the world is an evolutionarily inert receptacle. But is there really such a strong case that all the key events in human evolution happened in the same region of the world? The genetic data show that many groups of archaic humans populated Eurasia and that some of these interbred with modern humans. This forces us to question why the direction of migration would have always been out of Africa and into Eurasia, and whether it could sometimes have been the other way around.
The Most Ancient DNA Yet
At the beginning of 2014, Matthias Meyer, Svante Pääbo, and their colleagues in Leipzig extended by a factor of around four the record for the oldest human DNA obtained, sequencing mitochondrial DNA from a more than four-hundred-thousand-year-old Homo heidelbergensis individual from the Sima de los Huesos cave system in Spain where twenty-eight ancient humans were found at the bottom of a thirteen-meter shaft.28 The Sima skeletons have early Neanderthal-like traits, and the archaeologists who excavated them have interpreted them as being on the lineage leading to Neanderthals after the separation from the ancestors of modern humans. Two years after Meyer and Pääbo published mitochondrial DNA data from Sima de los Huesos, they published genome-wide data.29 Their analysis not only confirmed that the Sima humans were on the Neanderthal lineage, but went further in showing that the Sima humans were more closely related to Neanderthals than they are to Denisovans. These results provided direct evidence that Neanderthal ancestors were already evolving in Europe at least four hundred thousand years ago, and that the separation of the Neanderthal and Denisovan lineages had already begun by that time.
But the Sima data were also perplexing: Sima’s mitochondrial genome was more closely related to Denisovans than to Neanderthals, at odds with the genome-wide pattern of it being most closely related to Neanderthals.30 If there were only one discrepancy between the average relationship measured by the whole genome and the relationship seen in mitochondrial DNA, it might just be possible to believe that this was a statistical fluctuation. But there are two discrepancies in the genetic relationships: the fact that the Sima de los Huesos individual has Denisovan-type mitochondrial DNA despite being closer to Neanderthals in the rest of the genome, and the fact that the Siberian Denisovan individual has mitochondrial DNA twice as divergent from modern humans and Neanderthals as they were from each other despite being closer to Neanderthals in the rest of the genome.31 The coincidence of these two observations is so improbable that it seems more likely that there is a deeper story to unravel.
Perhaps the superarchaic humans—the ones who interbred with Denisovans—were a much more important part of Eurasian human populati
on history than we initially imagined. Maybe, after separating from the lineage leading to modern humans around 1.4 to 0.9 million years ago, these superarchaic humans spread across Eurasia and began to evolve the ancient mitochondrial lineage found in the Denisovans and Sima humans. At roughly half this time, another group may have split off the lineage leading to modern humans and then spread throughout Eurasia. This group may have mixed into the superarchaic population, contributing the largest proportion of ancestry to populations in the west that evolved into Neanderthals, and a smaller but still substantial proportion of ancestry to populations in the east that became the ancestors of Denisovans. This scenario would explain the findings of two anciently divergent mitochondrial DNA types in the different groups. It could also explain an odd unpublished observation I have: that in studying the variation in the time since the common genetic ancestor of modern human genomes with both Denisovan and Neanderthal genomes, I have not been able to find evidence for a superarchaic population that contributed to Denisovans but not to Neanderthals. Instead the patterns suggest that Denisovans and Neanderthals both had ancestry from the same superarchaic population, with just a larger proportion present in the Denisovans.
Johannes Krause and colleagues have suggested an alternative theory. Krause’s idea is that several hundred thousand years ago, an early modern human population migrated out of Africa and mixed with groups like the one that lived in Sima de los Huesos, replacing their mitochondrial DNA along with a bit of the rest of their genomes and creating a mixed population that evolved into true Neanderthals.32 The idea might seem complicated, but in fact it could explain multiple disparate observations beyond the fact that Neanderthals had a mitochondrial sequence much more similar to modern humans than it did to either the Sima de los Huesos individual or the Siberian Denisovan. It could account for the fact that the estimated date of the common ancestor of humans and Neanderthals in mitochondrial DNA (470,000 to 360,000 years ago)33 is paradoxically more recent than the estimated date of separation of the ancestors of these two populations based on the analysis of the whole genome (770,000 to 550,000 years ago).34 It could also explain how it was that Neanderthals and modern humans both used complex Middle Stone Age methods of manufacturing stone tools, even though the earliest evidence for this tool type is hundreds of thousands of years after the genetically estimated separation of the Neanderthals and modern human lineages.35 The theory finally becomes more plausible in light of a study led by Sergi Castellano and Adam Siepel that suggested up to 2 percent interbreeding into the ancestors of Neanderthals from an early modern human lineage.36 If Krause’s theory is right, this could have been the lineage that spread the mitochondrial DNA found in all Neanderthals.
