by David Reich
The researchers then transformed the resulting DNA fragments into a form that could be sequenced. First, they chemically removed the ragged ends of the DNA fragments that had been degraded after tens of thousands of years buried under the ground. In an extra measure to remove contamination beyond what had been done in the 2006 study, Pääbo and his team attached an artificially synthesized sequence of letters, a chemical “barcode,” to the ends of the DNA fragments. Any contaminating sequences that entered the experiment after the attachment of the barcode could thus be distinguished from the DNA of the ancient sample. The final step was to attach molecular adapters at either end that allowed the DNA fragment to be sequenced in one of the new machines that had made sequencing tens of thousands of times cheaper than the previous technology.
The best-preserved Neanderthal samples turned out to be three approximately forty-thousand-year-old arm and leg bones from Vindija Cave in the highlands of Croatia. After sequencing from these bones, Pääbo’s team found that the great majority of DNA fragments they obtained were from bacteria and fungi that had colonized the bones. But by comparing the millions of fragments to the present-day human and chimpanzee genome sequences, they found gold amidst the dross. These reference genomes were like the picture on a jigsaw puzzle box, providing the key to aligning the tiny fragments of DNA they had sequenced. The bones contained as much as 4 percent archaic human DNA.
Once Pääbo realized in 2007 that he would be able to sequence almost the entire Neanderthal genome, he assembled an international team of experts with the goal of ensuring that the analysis would do justice to the data. This is how I got involved, together with my chief scientific partner, the applied mathematician Nick Patterson. Pääbo reached out to us because over the previous five years we had established ourselves as innovators in the area of studying population mixture. Over the course of many trips to Germany, I played an important role in the analyses that proved interbreeding between Neanderthals and some modern humans.
Affinities Between Neanderthals and Non-Africans
The Neanderthal genome sequences we were working with were unfortunately full of errors. We could see as much because the data suggested that several times more mutations had occurred on the Neanderthal lineage than on the modern human lineage after the two sequences separated from their common ancestors. Most of these apparent mutations could not be real, since mutations occur at an approximately constant rate over time, and as the Neanderthal bones were ancient, they were actually closer in time to the common ancestor than are present-day human genomes, and so should have accumulated fewer mutations. Based on the degree of excess mutations on the Neanderthal lineage, we estimated that the Neanderthal sequences we were working with had a mistake approximately every two hundred DNA letters. While this might sound small, it is actually much higher than the rate of true differences between Neanderthals and present-day humans, so most of the differences we found between the Neanderthal sequence and present-day human sequences were errors created by the measurement process and not genuine differences between the Neanderthal and present-day human genomes. To deal with the problem, we restricted our study to positions in the genome that are known to be variable among present-day humans. At these positions, an error rate of about 0.5 percent was too low to confuse the interpretation. Based on these positions, we designed a mathematical test for measuring whether Neanderthals were more closely related to some present-day humans than to others.
The test we developed is now called the “Four Population Test,” and it has become a workhorse for comparing populations. The test takes as its input the DNA letters seen at the same position in four genomes: for example, two modern human genomes, the Neanderthal, and a chimpanzee. It examines whether, at positions where there is a mutation distinguishing the two modern human genomes that is also observed in the Neanderthal genome—which must reflect a mutation that occurred prior to the final separation of Neanderthals and modern humans—the Neanderthal matches the second human population at a different rate from the first. If the two modern humans descend from a common ancestral population that separated earlier from the ancestors of Neanderthals, there is no reason why the mutation is more likely to have been passed down one modern human line than another, and thus the rate of matching of each of the two modern human genomes to Neanderthal is expected to be equal. In contrast, if Neanderthals and some modern humans interbred, the modern human population descended from the interbreeding will share more mutations with Neanderthals.
Figure 7. We can evaluate whether two populations are consistent with descending from a common ancestral population through the “Four Population Test.” For example, consider a mutation that occurred in the ancestors of the Neanderthal (letter T, above) that is not seen in chimpanzee DNA. There are about 9 percent more of these mutations shared with Europeans than with African genomes, reflecting a history of Neanderthal interbreeding into the ancestors of Europeans.
When we tested diverse present-day human populations, we found Neanderthals to be about equally close to Europeans, East Asians, and New Guineans, but closer to all non-Africans than to all sub-Saharan Africans, including populations as different as West Africans and San hunter-gatherers from southern Africa. The difference was slight, but the probability of these findings happening by chance was less than one in a quadrillion. We reached this conclusion however we analyzed the data. This was the pattern that would be expected if Neanderthals had interbred with the ancestors of non-Africans but not Africans.
Trying to Make the Evidence Go Away
We were skeptical about this conclusion because it went against the scientific consensus of the time—a consensus that had been strongly impressed on many members of our team. Pääbo had done his postdoctoral training in the laboratory that in 1987 had discovered that the most deeply splitting human mitochondrial DNA lineages are found today in Africa, providing strong evidence in favor of an African origin for all modern humans. Pääbo’s own 1997 work strengthened the evidence for a purely African origin by showing that Neanderthal mitochondrial DNA fell far outside all modern human variation.19
I too came into the Neanderthal genome project with a strong bias against the possibility of Neanderthal interbreeding with modern humans. My Ph.D. supervisor, David Goldstein, was a student of Luca Cavalli-Sforza, who had made a fully out-of-Africa model a centerpiece of his models of human evolution, and I was steeped in this paradigm. The genetic data I knew about supported the out-of-Africa picture so consistently that from my perspective the strictest possible version of the out-of-Africa hypothesis, in which there was no interbreeding between the ancestors of present-day humans and Neanderthals, seemed like a good bet.
Coming from this background, we were deeply suspicious of the evidence we were finding for interbreeding with Neanderthals, and so we applied a particularly stringent series of tests in order to find some problem with our evidence. We tested whether the result was dependent on the genome sequencing technology that we used, but we obtained the same result from two very different technologies. We considered the possibility that the finding might be an artifact of a high rate of error in ancient DNA, which is known to affect particular DNA letters much more than others. However, we obtained the same result regardless of the type of mutation we analyzed. We wondered if our finding resulted from contamination of the Neanderthal sample by present-day humans. This could perhaps have tainted the data despite the measures that Pääbo’s team had taken to guard against it in the lab, and despite the tests we had performed on the data to measure the degree of modern human contamination, which had suggested that any contamination that was present was too small to produce the patterns we observed. However, even if there had been contamination from present-day humans, the patterns we observed looked nothing like what would be expected from it. If there had been contamination, it would most likely have come from a European, since almost all the Neanderthal bones we analyzed were excavated and handled by Europeans. Yet the Neanderthal sequence we had
was no closer to Europeans than to East Asians or to New Guineans—three very different populations.
We remained skeptical, wondering if something we had not thought of could explain the patterns. Then, in June 2009, I attended a conference at the University of Michigan where I met Rasmus Nielsen, who had been scanning through the genomes of diverse humans from around the world. In most parts of the genome, Africans are more genetically diverse than non-Africans and carry the most deeply diverging lineages, as is the case with mitochondrial DNA. But Nielsen was identifying rare places in the genome where the genetic diversity among non-Africans was greater than in Africans because of lineages that split off the tree of present-day human sequences early and were present only in non-Africans. These sequences just might be derived from archaic humans who had interbred with non-Africans. Nielsen joined our collaboration and compared the regions he and his colleagues identified to the data. When he compared twelve of his special regions to the Neanderthal genome sequence, he found that in ten of them there was a close match to the Neanderthal. This was far too high a fraction to happen by chance. Most of Nielsen’s highly divergent bits of DNA had to be Neanderthal in origin.
Next, we obtained a date for when the Neanderthal-related genetic material entered the ancestors of non-Africans. To do this, we took advantage of recombination—the process that occurs during the production of a person’s sperm or eggs that swaps large segments of parental DNA to produce novel spliced chromosomes that are passed to the offspring. For example, consider a woman who is a first-generation mixture of a Neanderthal mother and a modern human father. In her cells, each pair of her chromosomes consists of one unbroken Neanderthal chromosome and one unbroken modern human chromosome. However, her eggs contain twenty-three mixed chromosomes. One chromosome in an egg of hers might have its first half of Neanderthal origin and its other half of modern human origin. Suppose she mates with a modern human, and mixture continues down the generations with more modern humans. Over the generations, the segments of Neanderthal DNA get chopped into smaller and smaller bits, with recombination operating like the whirring blade of a food processor, splicing the parental DNA at random positions along the chromosome in each generation. By measuring the typical sizes of the stretches of Neanderthal-related DNA in present humans, evident from the size of sequences that match the Neanderthal genome more than they do sub-Saharan African genomes, we can learn how many generations have passed since the Neanderthal DNA entered a modern person’s ancestors.
Figure 8. When a person produces a sperm or an egg, he or she passes down to the next generation only one chromosome from each of the twenty-three pairs he or she carries. The transmitted chromosomes are spliced-together versions of the ones inherited from the mother and father (facing page). This means that the sizes of the bits of Neanderthal DNA in modern human genomes became smaller as the time since mixture increased (above, real data from chromosome 12).
With this approach, we found that at least some Neanderthal-related genetic material came into the ancestors of present-day non-Africans eighty-six thousand to thirty-seven thousand years ago.20 We have since refined this date by analyzing ancient DNA from a modern human from Siberia who, radiocarbon dating studies show, lived around forty-five thousand years ago. The stretches of Neanderthal-derived DNA in this individual are on average seven times larger than the stretches of Neanderthal-derived DNA in modern humans today, confirming that he lived much closer to the time of Neanderthal mixture. His proximity in time to the mixing event makes it possible to obtain a more accurate date of fifty-four thousand to forty-nine thousand years ago.21
But in 2012 we hadn’t yet proven that the interbreeding we had detected was with Neanderthals themselves. The most serious questioning came from Graham Coop, who was convinced that we had detected interbreeding with archaic humans, but pointed out that it was possible that the interbreeding hadn’t actually been with Neanderthals.22 Instead, the patterns could be the result of interbreeding with an as yet unknown archaic human in turn distantly related to Neanderthals.
A year later we were able to rule out Coop’s scenario after Pääbo’s laboratory sequenced a high-quality Neanderthal genome from a toe bone found in southern Siberia dating to at least fifty thousand years ago (if a sample is older than about fifty thousand years, radiocarbon dating can only provide a minimum date, so it actually could be substantially older).23 For this genome, we were able to gather about forty times more data than from the Croatian Neanderthal. With so much data, we could cross-check the sequence and edit away the errors. The resulting sequence was freer of errors than most genomes that are generated from living humans. The high-quality sequence allowed us to determine how closely related modern humans and Neanderthals are to each other based on the number of mutations that have occurred on the lineages since they separated. We found few or no segments where the Siberian Neanderthal shared common ancestors with present-day sub-Saharan Africans within the last half million years. However, there were shared segments with non-Africans roughly within the past one hundred thousand years. These dates fell within the time frame when Neanderthals were fully established in West Eurasia. This meant that the interbreeding was with true Neanderthals, not some distantly related groups.
Mixing in the Near East
So how much Neanderthal ancestry do people outside of Africa carry today? We found that non-African genomes today are around 1.5 to 2.1 percent Neanderthal in origin,24 with the higher numbers in East Asians and the lower numbers in Europeans, despite the fact that Europe was the homeland of the Neanderthals.25 We now know that at least part of the explanation is dilution. Ancient DNA from Europeans who lived before nine thousand years ago shows that pre-farming Europeans had just as much Neanderthal ancestry as East Asians do today.26 The reduction in Neanderthal ancestry in present-day Europeans is due to the fact that they harbor some of their ancestry from a group of people who separated from all other non-Africans prior to the mixture with Neanderthals (the story of this early-splitting group revealed by ancient DNA is told in part II of this book). The spread of farmers with this inheritance diluted the Neanderthal ancestry in Europe, but not in East Asia.27
Based on archaeological evidence alone, it would seem a natural guess that Neanderthals interbred with modern humans in Europe, the place where Neanderthals originated. But is that the place where the main interbreeding that left its mark in people today occurred? The genetic data cannot tell us for sure. Genetic data can show how people are related, but humans are capable of migrating thousands of kilometers in a lifetime even on foot, so genetic patterns need not reflect events that occurred near the locations where the people who carry the DNA live. If the ancient DNA studies of the last few years have shown anything clearly, it is that the geographic distribution of people living today is often misleading about the dwelling places of their ancestors.
However, we can make plausible conjectures about geographic origin. Evidence of interbreeding is detected today not just in Europeans but also in East Asians and New Guineans. Europe is a cul-de-sac of sorts within Eurasia, and would not have been a likely detour for modern humans expanding eastward. So where could Neanderthals and modern humans have met and mixed to give rise to a population that expanded not only to Europe but also to East Asia and New Guinea? Archaeologists have shown how in the Near East, Neanderthals and modern humans traded places as the dominant human population at least twice between 130,000 and 50,000 years ago, and it is reasonable to guess that they might have met during this period. So interbreeding in the Near East provides a plausible explanation for the Neanderthal ancestry that is shared by Europeans and East Asians.
Did interbreeding happen in Europe at all? In 2014, Pääbo’s group sequenced DNA from a skeleton from Oase Cave in Romania, the same skeleton that Erik Trinkaus had interpreted as a hybrid of Neanderthals with modern humans, based on features of its skull that were similar to both.28 Our analysis of the data showed that the Oase individual, who radiocarbon dating studies had sh
own lived about forty thousand years ago, had around 6 to 9 percent Neanderthal ancestry, far more than the approximately 2 percent that we measure in present-day non-Africans.29 Some stretches of Neanderthal DNA extend a third of the length of his chromosomes—a span so large and unbroken by recombination that we can be sure that the Oase individual had an actual Neanderthal no more than six generations back in his family tree. Contamination cannot explain these findings, as it would dilute the Neanderthal ancestry in the Oase individual, not increase it. It would also generate random matching to Neanderthals throughout the genome, not large stretches of Neanderthal DNA that could be readily identified by eye when we simply plotted along the genome the positions of mutations that match the Neanderthal genome sequence more closely than they match modern humans. This evidence of Neanderthal interbreeding didn’t need statistics. The proof was in the picture.
The discoveries about the interbreeding in the recent family tree of the Oase individual suggested that modern humans and Neanderthals also hybridized in Europe, the homeland of the Neanderthals. But the population of which Oase was a part—and which carried this clear imprint of interbreeding with European Neanderthals—may not have left any descendants among people living today. When we analyzed the genome of Oase, we found no evidence that he was more closely related to Europeans than to East Asians. This means that he had to have been part of a population that was an evolutionary dead end—a pioneer modern human population that arrived early in Europe, flourished there briefly and interbred with local Neanderthals, and then went extinct. Thus, while the Oase individual provides powerful evidence that interbreeding between Neanderthals and modern humans occurred in Europe, he does not provide any evidence that Neanderthal ancestry in non-Africans today is derived from European Neanderthals. It remains the case that the most likely source of Neanderthal ancestry in non-Africans is Near Eastern Neanderthals.