by David Reich
The identity of the population, however, was unclear. No skeletons or toolmaking styles existed to give a hint, as they had in the case of the Neanderthals. For Neanderthals, archaeological discoveries had motivated the sequencing of a genome. For this new archaic group, the genetic data came first.
A Genome in Search of a Fossil
I first found out about this previously unknown archaic human population in early 2010, while visiting Svante Pääbo’s laboratory in Leipzig, Germany. I was there on one of the thrice-yearly trips I had been making since joining the consortium that Pääbo put together in 2007 to analyze the Neanderthal genome. One evening, Pääbo took me out to a beer garden and told me about the new mitochondrial sequence they had come across. Miraculously, the Denisova finger bone had provided one of the best-preserved samples of ancient DNA ever found. While Pääbo had screened dozens of Neanderthal samples to find a few with up to 4 percent primate DNA, this finger bone had about 70 percent. Pääbo and his team had already been able to obtain more data on the whole genome (not just mitochondrial DNA) from this small bone than they had previously obtained from Neanderthals. He asked if I’d be interested in helping to analyze the data. The invitation to analyze the Denisovan genome was the greatest piece of good fortune I have had in my scientific career.
The mitochondrial genome suggested that the Denisova finger bone came from an individual who was part of a human population that split from the ancestors of modern humans and Neanderthals before they separated from each other. But mitochondrial DNA only records information on the entirely female line, a tiny fraction of the many tens of thousands of lineages that have contributed to any person’s genome. To understand what really happened in an individual’s history, it is incomparably more valuable to examine all ancestral lineages together. For the Denisova finger bone, the whole genome painted a very different picture from what was recorded in the mitochondrial DNA.
The first revelation from the whole genome was that Neanderthals and the new humans from Denisova Cave were more closely related to each other than either was to modern humans—a different pattern from what was observed in mitochondrial DNA.5 We eventually estimated the separation between the Neanderthal and Denisovan ancestral populations to have occurred 470,000 to 380,000 years ago, and the separation between the common ancestral populations of both of these archaic groups and modern humans to have occurred 770,000 to 550,000 years ago.6 The different pattern of relatedness for mitochondrial DNA and the consensus of the rest of the genome were not necessarily a contradiction, as the time in the past when two individuals share a common ancestor at any section of their DNA is always at least as old as the time when their ancestors separated into populations, and can sometimes be far older. However, by studying the whole genome we can learn when the populations split, recognizing that the whole genome encompasses a whole multitude of ancestors so that we can search for short segments of the genome with a relatively low density of mutations reflecting a shared ancestor who lived just before the population separation. Our findings meant that the Denisovans were cousins of Neanderthals, but were also very different, having separated from Neanderthal ancestors before many Neanderthal traits appeared in the fossil record.
We had a heated debate about what to call the new population, and decided to use a generic non-Latin name, “Denisovans,” after the cave where they were first discovered, in the same way that Neanderthals are named after the Neander Valley in Germany. This decision distressed some of our colleagues, who lobbied for a new species name—perhaps Homo altaiensis, after the mountains where Denisova Cave is located. Homo altaiensis is now used in a museum exhibit in Novosibirsk in Russia that describes the discovery at Denisova. We geneticists, however, were reluctant to use a species name. There has long been contention as to whether Neanderthals constitute a species separate from modern humans, with some experts designating Neanderthals as a distinct species of the genus Homo (Homo neanderthalensis), and others as a subgroup of modern humans (Homo sapiens neanderthalensis). The designation of two living groups as distinct species is often based on the supposition that the two do not in practice interbreed.7 But we now know Neanderthals interbred successfully with modern humans and in fact did so on multiple occasions seems to undermine the argument that they are distinct species. Our data showed that Denisovans were cousins of Neanderthals, and thus if we are uncertain about whether Neanderthals are a species, we need to be uncertain about whether Denisovans are a species as well. Decisions about whether extinct populations are distinct enough to merit designation as different species are traditionally made based on the shapes of skeletons, and for Denisovans there are very few physical remains, providing even more reason to be cautious.
The few remains that we do have are intriguing. Derevianko and his colleagues sent Pääbo a couple of molar teeth from Denisova Cave that contained mitochondrial DNA closely related to the finger bone. These teeth were enormous, beyond the range of nearly all teeth previously reported in the genus Homo. Large molars are thought to be biological adaptations to a diet that includes lots of tough uncooked plants. Prior to the Denisovans, the humans closest to us who were known to have had teeth of this size were the primarily plant-eating australopithecenes, like the famous “Lucy,” whose skeleton, dating to more than three million years ago, was found in the Awash Valley of Ethiopia. “Lucy” did not use tools and had a brain only slightly larger than chimpanzees’ after correcting for her smaller body size, but she walked upright. Thus the little skeletal information we had confirmed the idea that Denisovans were very distinctive compared to both Neanderthals and modern humans.
The Hybridization Principle
Armed with a whole-genome sequence, we tested whether the Denisovans were more closely related to some present-day populations than others. This led to a huge surprise.
Denisovans were genetically a little closer to New Guineans than they were to any population from mainland Eurasia, suggesting that New Guinean ancestors had interbred with Denisovans. Yet the distance from Denisova Cave to New Guinea is around nine thousand kilometers, and New Guinea is, of course, separated by sea from the Asian mainland. The climate in New Guinea is also largely tropical, which could not be more different from Siberia’s bitter winters, and this makes it unlikely that archaic humans adapted to one environment would have flourished in the other.
Skeptical of our findings, we cast around for alternative explanations. Had the ancestors of modern humans been divided into several populations hundreds of thousands of years ago, one of which was more closely related to Denisovans and contributed more to New Guineans’ ancestry than it contributed to the ancestry of most other present-day populations? However, this scenario would suggest that the genetic affinity to Denisovans in present-day New Guineans would be due to segments of DNA that entered the New Guinean lineage many hundreds of thousands of years ago. In New Guinean genomes today we were able to measure the size of intact archaic ancestry segments, and found that the ones related to Denisovans were about 12 percent longer than the ones related to Neanderthals, implying that the Denisovan-related segments had been introduced that much more recently on average.8
As soon as archaic populations mix with modern ones, the DNA segments contributed by archaic humans are chopped up by the process of recombination, spliced together with modern human segments at the rate of one or two splices per chromosome per generation. As discussed in chapter two, the length of Neanderthal ancestry segments corresponds to mixture between fifty-four and forty-nine thousand years ago.9 Based on how much longer the Denisovan segments were than the Neanderthal segments in New Guineans, we could conclude that the interbreeding between Denisovan and New Guinean ancestors occurred fifty-nine to forty-four thousand years ago.10
Figure 10. Approximate proportions of Neanderthal (left) and Denisovan ancestry (right) in representative present-day human populations as a fraction of the maximum detected in any group today. Today, Denisovan ancestry is concentrated east of Huxley’s Line, a
deep-sea trench that has always divided mainland Asia from Australia and New Guinea even in the ice ages when sea levels were lower.
What percentage of New Guinean genomes today derives from Denisovans? By measuring how much stronger the genetic evidence of archaic ancestry is in New Guineans compared to other non-Africans, we estimated that about 3 to 6 percent of New Guinean ancestry derives from Denisovans. That is above and beyond the approximately 2 percent from Neanderthals. Thus in total, 5 to 8 percent of New Guinean ancestry comes from archaic humans. This is the largest known contribution of archaic humans to any present-day human population.
The Denisova discovery proved that interbreeding between archaic and modern humans during the migration of modern humans from Africa and the Near East was not a freak event. So far, DNA from two archaic human populations—Neanderthals and Denisovans—has been sequenced, and in both cases, the data made it possible to detect hybridization between modern and archaic humans that had been previously unknown. I would not be surprised if DNA sequenced from the next newly discovered archaic population will also point to a previously unknown hybridization event.
Breaching Huxley’s Line
Where, given the vast distance between Siberia and New Guinea, did interbreeding between Denisovans and the ancestors of New Guineans occur?
Our first guess was mainland Asia, perhaps India or central Asia, on a plausible human migratory path from Africa to New Guinea. If this had been the case, the lack of much Denisovan-related ancestry in mainland East or South Asia could be explained by later waves of expansion on the part of modern humans without Denisovan-related ancestry, who replaced populations having Denisovan-related ancestry. That these later migrations did not contribute much to the DNA of present-day New Guineans might account for the relatively high proportion of Denisovan-related ancestry in New Guinean populations today.
A first glance at the geographic distribution of Denisovan-related ancestry in present-day people seemed to support this idea. We collected DNA from present-day humans from the islands of the Southwest Pacific and from East Asia, South Asia, and Australia, and estimated how much Denisovan-related ancestry each of them had. We found the largest amounts of ancestry in indigenous populations in the islands off Southeast Asia and especially in the Philippines and the very large islands of New Guinea and Australia (by the word “indigenous” I refer to people who were established prior to the population movements associated with the spread of farming).11 The populations in question are largely east of Huxley’s Line, a natural boundary that separates New Guinea, Australia, and the Philippines from the western parts of Indonesia and the Asian mainland. This line was described by the nineteenth-century British naturalist Alfred Russel Wallace, and adapted by his contemporary the biologist Thomas Henry Huxley to highlight differences in the animals living on either side, for example, it roughly forms the boundary between placental mammals to the west and marsupials to the east. It corresponds to deep ocean trenches that have formed geographical barriers to the crossing of animals and plants, even in ice ages when sea levels were up to one hundred meters lower. It is remarkable that modern humans after fifty thousand years ago made it across this barrier. These pioneers did manage to cross, but it must have been difficult. Modern humans with Denisovan-related ancestry living east of Huxley’s Line—the ancestors of New Guineans, Australians, and Philippine populations who we found are the groups with the largest proportions of Denisovan ancestry today—are likely to have been protected by the same barrier from further migrations from Asia, just like the animals with whom they share their landscape.
But a deeper look suggests that population mixture in the heart of Asia is not as easy an explanation as it might at first seem. Although some populations east of Huxley’s Line have large amounts of Denisovan-related ancestry, the situation is very different to the west. Most notably, the indigenous hunter-gatherers of the Andaman Island chain off the coasts of India and Sumatra, and also the indigenous hunter-gatherers of the Malay Peninsula of mainland Southeast Asia, descend from lineages just as divergent as those in indigenous New Guineans and Australians, and yet they do not have much Denisovan-related ancestry. There is also no evidence of elevated Denisovan-related ancestry in genome-wide data from the approximately forty-thousand-year-old human of Tianyuan Cave near Beijing in China, which was sequenced several years later by Pääbo and his laboratory.12 Had the interbreeding occurred in mainland Asia, and modern humans carrying Denisovan-related ancestry then spread all over, multiple populations of the region as well as ancient humans from East Asia would be expected to carry Denisovan-related ancestry in amounts comparable to what is seen in New Guineans. But this is not what we observe.
The simplest explanation for the large fractions of Denisovan-related ancestry on the islands off the southeastern tip of Asia and in New Guinea and Australia would be the occurrence of interbreeding near the islands—on the islands themselves or in mainland Southeast Asia—but in either case in a tropical region very far from Denisova Cave. However, the anthropologist Yousuke Kaifu pointed out in a talk I attended in 2011 that the hypothesis of interbreeding near the islands is difficult to square with an absence of archaeological artifacts in the region that could plausibly reflect the presence of a big-brained cousin of Neanderthals and modern humans. Kaifu also pointed out that no big-skulled skeletons from this time in this region have so far been found. This makes me think that it is more likely that interbreeding occurred in southern China or mainland Southeast Asia. There are archaic human remains from Dali in Shaanxi province in north-central China, from Jinniushan in Liaoning in northeastern China, and from Maba in Guangdong in southeastern China, all dating to around two hundred thousand years ago, all of which are more plausible skeletal matches for the Denisovans. An archaic human from Narmada in central India may date to around seventy-five thousand years ago. Chinese and Indian government rules complicate the export of skeletal material, but world-class ancient DNA labs have now been established in China and are beginning to be built in India. DNA from these samples could lead to extraordinary insights.
Meet the Australo-Denisovans
While the interbreeding Neanderthals were close relatives of those we obtained samples from and sequenced, the archaic people who interbred with the ancestors of New Guineans were not close relatives of the Siberian Denisovans. When we examined the genomes of present-day New Guineans and Australians, and counted the number of DNA letter differences between them and the Siberian Denisovans to estimate when their ancestors separated from a common parent population, we discovered that everywhere in the genome, the number of differences was at least what would be expected for a population split that occurred 400,000 to 280,000 years ago.13 This meant that the ancestors of the Siberian Denisovans separated from the Denisovan lineage that contributed ancestry to New Guineans two-thirds of the way back to the separation of the ancestors of Denisovans from Neanderthals.
In light of the remote relationship, the two groups probably had different adaptations, which would explain how they were able to thrive in such different climates. Given the extraordinary diversity of Denisovans—with much more time separation among their populations than exists among present-day groups—it makes sense to think of them as a broad category of humans, one branch of which became the ancestors of the archaic population that interbred with New Guineans and another that became Siberian Denisovans. Most likely there are other Denisovan populations as well that we haven’t sampled at all. Maybe we should even consider Neanderthals as part of this broad Denisovan family.
We never assigned a special name to the Denisovan-related population that interbred with modern humans who migrated to the islands off Southeast Asia, but I like to call them “Australo-Denisovans” to highlight their likely southern geographical distribution. Anthropologist Chris Stringer prefers “Sunda Denisovans” after the landmass that joined most of the Indonesian islands to the Southeast Asian mainland.14 But this would not be an accurate name if the interbreeding
occurred in what is now mainland Southeast Asia, China, or India.
It is tempting to think that the Australo-Denisovans, Denisovans, and Neanderthals descend from the first Homo erectus populations that expanded out of Africa, and that modern humans descend from the Homo erectus populations that stayed in Africa, but that would be wrong. The oldest Homo erectus skeletons outside of Africa have been found at the site of Dmanisi in Georgia dating to around 1.8 million years ago, and on the island of Java in Indonesia dating to almost the same time. If Homo erectus from the first radiation out of Africa was ancestral to the Denisovans and Neanderthals, then the split of these populations from modern humans would be at least as old as the dispersal to Eurasia—far too old to be consistent with the genetic observations. The genetic data give a split date of 770,000 to 550,000 years ago, too recent to be consistent with a 1.8-million-year-old population separation.
There is, however, a candidate in the fossil record for an ancestor in the right period, dating to long after the Homo erectus out-of-Africa migration but after the Homo sapiens one. A big-skulled skeleton found near Heidelberg in Germany in 1907 and dated to around six hundred thousand years ago15 was plausibly from a species that was ancestral to modern humans and Neanderthals,16 and by implication, Denisovans too. Homo heidelbergensis is often viewed as both a West Eurasian and an African species, but not an East Eurasian species. However, the genetic evidence from the Australo-Denisovans shows that the Homo heidelbergensis lineage may have been established very anciently in East Eurasia too. One of the profound implications of the Denisovan discovery was that East Eurasia is a central stage of human evolution and not a sideshow as westerners often assume.
