It is always a relief to arrive at Denisova Cave in southern Siberia. After a bumpy 11-hour drive southeast from Novosibirsk, across the steppe and through the foothills of the Altai Mountains, the field camp suddenly appears around a bend in the dirt road, and all thought of the long journey evaporates. Steep-sided valleys, swift-running rivers and the traditional wood houses of the local Altai people dominate the landscape; golden eagles soar overhead. A couple of hundred meters away, the limestone cave itself, perched high above the Anui River, beckons with the promise of some of the most exciting research under way in the field of human origins.
Denisova Cave is at the center of a revolution in scientists’ understanding of how our ancestors in the Paleolithic, or Old Stone Age, behaved and interacted. Our species, Homo sapiens, originated in Africa hundreds of thousands of years ago. When it eventually began spreading into Europe and Asia, it encountered other human species, such as the Neandertals, and shared the planet with them for millennia before those archaic species disappeared. Scientists know these groups encountered one another because people today carry DNA from our extinct relatives—the result of interbreeding between early H. sapiens and members of those other groups. What we do not yet know and are eager to ascertain is when and where they crossed paths, how often they interbred and how they might have influenced one another culturally. We actually have quite a few important archaeological sites from this transitional period that contain stone tools and other artifacts. But many of these sites, including Denisova, lack human fossils that are complete enough to attribute to a particular species on the basis of their physical traits. That absence has hindered our ability to establish which species made what—and when.
Now a technique for identifying ancient bone fragments, known as zooarchaeology by mass spectrometry (ZooMS), is finally allowing researchers to start answering these long-standing questions. By analyzing collagen protein preserved in these seemingly uninformative fossil scraps, we can identify the ones that come from the human/great ape family and then attempt to recover DNA from those specimens. Doing so can reveal the species they belong to—be it Neandertal, H. sapiens or something else. What is more, we can carry out tests to determine the ages of the fragments.
Directly dating fossils is a destructive process—one has to sacrifice some of the bone for analysis. Museum curators are thus usually loath to subject more complete bones to these tests. But they have no such reservations with the scraps.
The ability to directly date fossils found in association with artifacts is especially exciting with regard to Denisova and other sites we know sheltered multiple human species in the past. A number of researchers have argued that symbolic and decorative artifacts, which are proxies for modern cognitive abilities, are unique to H. sapiens. Others maintain that Neandertals and other species made such items, too, and may have even passed some of their traditions along to the H. sapiens they met. The ability to identify and date these fossil fragments means researchers can begin to reconstruct the chronology of these sites in far greater detail and elucidate a critical chapter of human prehistory.
Russian archaeologists have been excavating Denisova Cave since the 1980s. But it was an announcement in 2010 that put the site on the map. That year scientists at the Max Planck Institute for Evolutionary Anthropology in Leipzig, Germany, reported on the results of their genetic analysis of a bone found at Denisova in 2008. The DNA they obtained from the fossil—a bit of finger bone—revealed a previously unknown type of hominin, or member of the human family, one that was as closely related to Neandertals as we are. The bone was from a young girl, initially dubbed “X woman,” who belonged to a group of people that scientists now refer to as Denisovans. Since then, a handful of other hominin bones and teeth have been discovered among the excavated remains, both Denisovan and Neandertal.
Those discoveries at Denisova illustrated the powerful information that can be gleaned from fossils using modern genetic approaches, telling us not only about the presence of hitherto unknown species but also about the nature of their interaction with us. We know from genetic analysis, for instance, that Neandertals and modern humans interbred at least three times over the past 100,000 years and that Neandertals and Denisovans, as well as moderns and Denisovans, also mixed. As a result, the long-held view that H. sapiens moved out of Africa and simply wiped out such archaic populations has, in the blink of an eye, given way to a more complex scenario of interbreeding and gene flow between groups—a “leaky replacement” model of modern human origins. Yet most of the fossils at Denisova are so incomplete that we cannot discern which ones might belong to a human species. And the site has been notoriously difficult to date.
We got involved in the Denisova project six years ago through our expertise in chronology, particularly the use of radiocarbon dating to establish time frames for archaeological sites. For material from the Middle and Upper Paleolithic periods (broadly the time spanning 250,000 to 40,000 and 40,000 to 10,000 years ago, respectively), dating is hugely important because the sites themselves often lack distinctive tool types that are associated with tightly defined periods. We are working to provide a robust chronology at Denisova and other Paleolithic locations in Eurasia.
We were both at the site in 2014 attending a meeting of the Denisova team when we came up with an idea that we thought might help us build a more nuanced picture of the interactions that occurred among our species, Neandertals and Denisovans. One thing that was apparent at Denisova was that all the known hominin remains were absolutely tiny, just three to five centimeters long. The X woman finger bone, for instance, was about the size of a lentil and weighed less than 40 milligrams. A great proportion of bone material at the site was broken, principally because of the activity of predators such as hyenas, which den in caves to have their young and chew up bones while feeding. Since 2008 more than 135,000 bones have been excavated at Denisova, but 95 percent of them cannot be identified taxonomically, because they are too fragmentary. In contrast, the preservation of biomolecules in these fragments—including those molecules that make up DNA—is amazing: the two most complete ancient hominin genomes ever recovered come from Denisova fossils. What if, we wondered, there were a way of screening these many thousands of bone fragments at the site to find more human bones? If we could do this, perhaps we could generate more genetic and chronometric data or even find a new type of hominin lurking in the cave. It was then that we realized that we might be able to carry out exactly this kind of screening using ZooMS.
ZooMS, also called collagen peptide mass fingerprinting, allows investigators to assign fragments of bone to the proper taxonomic group by analyzing the proteins in bones. Bone collagen protein is made up of hundreds of small compounds called peptides that vary slightly among different types of animals. By comparing the peptide signatures of mystery bones against a library of such signatures from known animals, it is possible to assign the unidentified bones to the correct family, genus and sometimes species. First developed by Michael Buckley, now at the University of Manchester, and Matthew Collins of the University of York, both in England, ZooMS has been employed for more than a decade to identify the animal bones at archaeological sites. It is relatively cheap, costing around $5 to $10 per sample, and minimally destructive—it requires only around 10 to 20 milligrams of bone for analysis. It is also rapid; one person can screen hundreds of bones a week.
To our knowledge, no one had used ZooMS to search for human bones before. But we figured we had a shot. Even small fragments should be potentially useful, we reasoned, because the bone collagen and DNA preservation in Denisova is unsurpassed, given its stable and very low average annual temperature of below zero degrees Celsius. We knew we would not be able to get species-level identification with ZooMS. The collagen peptide signatures of human species and the great apes are too similar to discriminate. But no great apes are known to have roamed this part of the world during the Paleolithic. So if we could identify a piece of bone as belonging to a member of the group that comprises great apes and humans—together known as Hominidae—we could be fairly certain that it belonged to a human of some kind and subject it to genetic analysis that could provide the species identification.
Ancient DNA expert Svante Pääbo of the Max Planck Institute for Evolutionary Anthropology, who leads the Neandertal Genome Project and whose group published the Denisovan genome in 2010, was at Denisova for the 2014 meeting. We had not formally met Pääbo before, and we were keen to see what he thought of our idea for screening bone fragments and whether he would be interested in collaborating on that effort. He jumped at the chance and gave his immediate support. We then discussed our plan with Anatoly Derevianko of the Russian Academy of Sciences, who oversees work done at Denisova, and Michael Shunkov, director of the excavations. Both were interested. And so later that year we began the process of sampling a few thousand little fragments of, for all intents and purposes, “worthless” bones that had been recently excavated from the site.
In the abstract, it seemed like it would be quick work. In reality, we were faced with the massive job of painstakingly removing a minuscule sample of bone from each fragment for analysis, taking care not to touch the potentially valuable specimens with anything that might contaminate them. One of our students, Samantha Brown, who took on the project for her master’s research dissertation, carried out much of this work, logging countless hours at our lab at the University of Oxford.
Buckley collaborates with us on this project. Once we had 700 to 800 bone samples ready, Brown went to his lab to prepare and analyze them. The results were interesting: we had mammoths, hyenas, horses, reindeer, woolly rhinos—the full panoply of Ice Age beasts—but sadly, none of the peptide signatures corresponded with Hominidae. It was disappointing, but we decided to try a second batch to see whether we could locate even one human bone from the mass of fragments. Although we did not fancy our chances, we were hoping to be proved wrong.
Then one evening in the summer of 2015 we received an e-mail from Buckley. He had noticed that one of our samples, DC1227, had the characteristic peptide markers coding for Hominidae. We had a fragment of human bone—the proverbial needle in the haystack! We were ecstatic; our crazy idea seemed to have actually panned out.
Early the next day we went to our lab at Oxford to find the bone among the archived samples. We were somewhat deflated when we saw that the bone we had found was tiny even for a Denisova specimen—only 25 millimeters long—which did not leave much for further studies. But given the exceptional biomolecular preservation of the Denisova remains, we believed that it would be enough to allow us to apply the techniques we wanted to use to find out as much about the bone as possible. We photographed it at high resolution, put it through a CT scanner, and drilled additional samples for dating and isotope analysis before Brown took the bone to Leipzig for DNA analysis in Pääbo’s lab.
Several weeks later the dating results came back. The absence of any traceable radioactive carbon in the sample implied that our little bone was more than 50,000 years old. And before long, we learned from Pääbo that its mitochondrial DNA—which resides in the energy-producing organelles of cells and is passed down from mother to child—indicated that the bone came from an individual who had a Neandertal mother. We had found a hominin bone fragment hidden among thousands of “junk” bones and proved that the concept could work. Pääbo’s team was planning to extract the much more informative nuclear genome from the bone, which now went by the site fossil I.D. “Denisova 11,” or “Denny,” as we nicknamed it. In the meantime, we decided to test our approach at another site.
US VS. Them
Vindija Cave in Croatia is a key site for understanding late Neandertals in Europe. For many years radiocarbon dates indicated that Neandertals there might have survived until 30,000 years ago, providing evidence for a potential overlap phase with anatomically modern humans, who arrived in the region by 42,000 to 45,000 years ago. Such a lengthy coexistence hinted that rather than being driven to extinction by modern humans, the Neandertals had been assimilated into their population. While reassessing the Vindija chronology, we decided it might be interesting to use ZooMS to assess the unidentified bones from the site. Previous work on the more complete bones from Vindija had shown that cave bears dominate the remains, accounting for some 80 percent of the bones, so we were not expecting to find the variety and breadth of fauna that we had detected at Denisova. Cara Kubiak, then another of our students, took on the project.
Surprisingly, the 28th sample out of the 383 we analyzed yielded a peptide sequence consistent with Hominidae. Later, Pääbo’s team confirmed it genetically as a Neandertal. This bone was around seven centimeters long and, intriguingly, exhibited cut marks and other signs of human modification. Neandertal bones sometimes bear these markings, which may well be evidence of butchery and cannibalism.
The specimen, known as Vi-*28, turned out to be pivotal for our chronology work. Historically, archaeologists and preparators treated the bones from Vindija with conservation products to protect them. That practice makes radiocarbon dating very difficult because these products add carbon to the bone. Unlike other human bones from the site, Vi-*28 was not conserved; misidentified as an animal bone, it had eluded treatment—a boon for us. Radiocarbon dating of Vi-*28 revealed that it belonged to a Neandertal from more than 47,000 years ago. This finding, published in 2017, along with dates we obtained from other Neandertals, showed that they disappeared from Vindija more than 40,000 years ago, before modern humans arrived at the site. The earlier dating results suggesting that they had persisted until at least 30,000 years ago were a fiction, influenced by contaminating carbon that had not been effectively removed. ZooMS again had proved its worth.
Other teams have had great success with the technique, too. In 2016 Frido Welker, now at the Natural History Museum of Denmark, and his colleagues reported that they had used ZooMS to identify 28 previously unrecognized hominin fossils among the unidentified bone fragments from the famed site of Grotte du Renne in the Burgundy region of France. Decades ago researchers working there found Neandertal bones in association with an array of surprisingly sophisticated artifacts, including bone tools, as well as pendants and other body ornaments—elements of a so-called Châtelperronian culture that is said to be transitional between the Middle Paleolithic and the Upper Paleolithic. The discovery ran counter to the long-held idea that H. sapiens alone was capable of such ingenuity. In so doing, it touched off an enduring debate over whether the Neandertals were truly associated with the advanced artifacts or whether the archaeological levels at the site had been disturbed somehow, mixing Neandertal bones with later artifacts left behind by H. sapiens.
The 28 bone fragments Welker and his colleagues identified as human using ZooMS all clearly came from the same layer as the advanced tools and ornaments. When they had the bones sequenced, the results were unequivocal: the specimens were Neandertal, not H. sapiens. The work lends considerable support to the notion that Neandertals did indeed make the Châtelperronian and other transitional industries and that they were cleverer than they have often been given credit for.
A Hybrid Child
Throughout our work at Vindija, we continued to analyze samples from Denisova in the hope that we could add more human fossils to our collection. Our efforts yielded two more Hominidae hits: DC3573, which turned out to belong to a Neandertal from more than 50,000 years ago, and DC3758, a 46,000-year-old bone that unfortunately does not preserve any ancient DNA. More than 5,000 bone fragments have now given us a total of five hominin bones that might have languished in obscurity forever if not for ZooMS.
But the most exciting development was yet to come. In May 2017 we were at the Max Planck Institute for Evolutionary Anthropology and met with senior members of Pääbo’s lab, including Matthias Meyer and Janet Kelso. We wanted to know about the status of Denisova 11 and whether they had managed to retrieve nuclear DNA, which would give us a much more detailed picture of who Denisova 11 was.
It is not often in science that one receives completely jaw-dropping news, but Meyer and Kelso delivered exactly that. The nuclear DNA, they said, was curiously split: half was consistent with a Neandertal, and the other half appeared to be derived from a Denisovan. They thought Denisova 11 was a 50–50 hybrid. To exclude all possibility of error, the team was running the samples again to verify this astonishing result. Several months later the final data confirmed this initial finding. The mitochondrial DNA had given us only half of the picture. What we had found was not a Neandertal but an individual with a Neandertal mother and a Denisovan father—a first-generation hybrid, in the parlance of geneticists. The Denisova team announced this astounding discovery in the September 6 Nature, in a paper led by Viviane Slon of the Max Planck Institute for Evolutionary Anthropology.
We know now from the DNA that Denisova 11 was a female who probably lived around 90,000 to 100,000 years ago. And bone-density analysis generated from the CT scan we performed allowed our colleague Bence Viola of the University of Toronto to tentatively estimate her age at death at a minimum of 13 years old. Her Denisovan father himself had a distant Neandertal relative several hundred generations back. Of course, we can never know how these unions came about in prehistory, only that they did. Neither can we establish how Denisova 11 died, just that her remains were probably deposited in the cave sediment by a predator, possibly a hyena.
We will never know whether she died and was ceremonially buried by her loved ones, only to be scavenged by the hyena later, or lost her life to a predator. For tens of millennia this minute piece of her body lay undisturbed in the cave and might well have remained there for many more years, had it not been for the cutting-edge science that has allowed us to breathe life into her story. We are hopeful that ZooMS will help us unlock many more such secrets archived in bone.