Who We Are and How We Got Here: Ancient DNA as a Window into Human History and Biology | David Reich

Thank you, Tamiko, for the introduction, and it's really a great honor, and to everyone who invited me today. So the title of my talk is "Ancient DNA as a Window into Human History and Biology," and I want to begin by talking about this new technology, which I've always viewed as kind of a new scientific instrument with the power that a new scientific instrument can sometimes have.

So 400 years ago, there was the invention of instruments like microscopes and telescopes that made it possible to peer into worlds that had not been looked at before. And in the same way, ancient DNA is able to peer into the past and to understand how people are related to each other and to people living today. And what it shows is worlds that we hadn't imagined before. It's very surprising.

So the measure of a new scientific instrument, the importance of a new scientific instrument, is when you use it to look at something that has not yet been looked at before, how surprising the results are. So I hope that at the end of this presentation, and maybe after the conversation with Carl and the questions, you'll come away feeling that this is really very powerful as a new way of looking at the past.

Ancient DNA begins with a remain. This is an earbone from a human in our laboratory. And then in a clean room where the goal is to get beneath the surface of the bone and not to contaminate it, we take a small part that's hopefully DNA-rich. We take a little bit of powder, maybe the 40th of the weight of a paperclip. We release the DNA in a solution that removes the protein and mineral content.

And then we sequence, often with an enrichment step that improves the efficiency of the sequencer. The talk today is not about the deep integration of ancient DNA into biology, archaeology, and environmental sciences, which is super exciting and is finally now happening. It's really being put to the surface of these other fields. That's another talk that should be given, and I could try to give, but I'm not talking about that today.

This is a talk whose theme is about three examples of how qualitatively new data can falsify assumptions and hopefully change our understanding of the world. So the three examples are one, two, three.

So, one, I'm going to talk about before ancient DNA, we had a fundamentally wrong understanding of how modern humans—people like us who appear in the skeletal record about 300,000 years ago for the first time in Africa—relate to archaic humans like Neanderthals who lived in Europe and Western Eurasia.

So, in 2007, Svante Pääbo, who's in the audience today, started getting genome-scale data from Neanderthals. And this was really a very, very exciting development. And I got a lucky break. So I was a fourth-year assistant professor at Harvard Medical School, and I was studying diseases, especially in African Americans, that had differential risk across populations.

And Svante, because of side work I had been doing, invited me to be part of an international consortium where my role would be to co-analyze the data to learn about modern human population relationships. It was clear to me and to my colleagues, especially Nick Patterson, who's also here and who worked with me on this, that this was the most important data in the world.

I kind of feel this sometimes when reading about gravitational wave data or something like this, that maybe I should start working on that. But it's just such amazing, exciting, like, holy data. It's very special, and I think that that was felt by all of the people in our international collaborations.

Svante had an amazing group already in Leipzig, but I think he decided to put together an international collaboration to bring together people with a variety of skills from all around the world to work on it. And it was really an extraordinary experience with people really not competing at all, but just trying to do the best work because they were in awe of the data that we had.

For this next seven years, I spent most of my own time analyzing this data. I told myself I'm going to do a second postdoc, really. And what happened actually was interesting. Over the weekend, I had lunch with a colleague at Brandeis, and he asked me, "Who did you do your postdoc with?" And he said, "I thought it was Svante Pääbo." So I was very happy to hear about that.

I actually did my postdoc with Eric Lander at the Broad Institute, and that work has also—that influence has also been very, very important to me.

So the background for this work was that geneticists had convinced ourselves—and I had been a minor part of that as a graduate student—that Neanderthals, archaic humans who had lived in Europe and Western Eurasia, and that we had an amazing skeletal record of and amazing archaeology, had not mixed with modern humans in a way that left a trace in people living today, though overlapping in Europe from 45 to 40,000 years ago, and though overlapping in the Middle East for tens of thousands of years before that.

Here's an example of a paper that argued why that seemed to be so. But analyzing the data that was produced by Svante's group, I soon found evidence that they had mixed, and I was really incredulous because of this background, coming very much from this tradition in human DNA that this had not happened. We hadn't seen an ounce of it. We hadn't seen a hint of it.

And there were all sorts of weirdnesses about the genome, all of which could look like Neanderthal admixture. If you're screening through the genome for something weird, then any errors in your sequencing, any errors in your processing might look like Neanderthal ancestry. So we kept trying to make these results go away.

But we eventually became convinced after compiling multiple lines of evidence. So here's the simplest of these pieces of evidence. So this is a cartoon of a chromosome, like one of your 23 pairs of chromosomes. And if you take a sequence from a French person, a European, or a Nigerian person, look at a difference between them, there's about one such difference every thousand DNA letters. There's about three billion DNA letters in a genome.

And you ask whether a Neanderthal differs and matches the French or the Nigerian. If the Neanderthal is symmetrically related to these two groups, it should match equally often. But we found much more matching statistically to French people than to Nigerian people. This was true not just for Europeans but for East Asians. It was also true not just for West Africans but for southern Africans and Central African hunter-gatherers.

So it was a difference between sub-Saharan Africans and non-Africans that we saw consistently. Another thing we did is we replaced this European with a second Neanderthal we got data from, and the skew became much, much bigger—about 50 times bigger—because it was 50 times better. We could learn that French were 2% of the way to a second Neanderthal, so it was 2% gene flow into Neanderthals. That was the interpretation.

And an equally big surprise was that there was as much signal of Neanderthal affinity in East Asians and in Papuans from Melanesia as there was in Europeans, despite that Neanderthals never had been in these places. So that was very surprising to us too, because the obvious place that mixture would have occurred was in Europe.

And if you asked physical anthropologists where mixture with Neanderthals happened, if it did happen—some might say it did happen—but they would have argued it was Europe. But we saw no—there was no evidence of a special signal in Europe. In fact, there might even be less Neanderthal ancestry in Europeans today than in East Asians.

So the next thing that we did, among many things, was we built maps of where in each person's genome you can find the Neanderthal ancestry. And it turns out to be much easier than one might expect because it's so distinctive. You move along the genome and there's sharp evidence that there's high rate of matching to Neanderthal, for example, or lineages that are very deeply diverged.

So even though these segments are maybe only 40,000 DNA letters long, you can see them sharply and be confident of many of them in terms of their locations. So in 2014, we and others built a map of where in large numbers of people's DNA these Neanderthal segments occurred, and you could pile them up in large numbers of Europeans and East Asians. The average is 2%, but you see there's spikes where it's more than 2%.

And there's also—which are possibly evidence of natural selection in non-Africans for segments of DNA derived from Neanderthals that maybe even might have been pre-adapted to the Eurasian environments modern humans were moving into from Africa, maybe cold-adapted, for example.

There were also deserts, places where Neanderthal ancestry was absent, that are potentially places where in the mixed individuals it was essential to carry a modern human genetic variant because it was particularly useful for the population.

There also were places in the genome that were systematically depleted in Neanderthal ancestry, especially chromosome X, one of the two sex chromosomes, and sperm-related genes. And this is the characteristic in many species of male hybrid sterility.

So if you take two groups that are at the edge of biological incompatibility and mix them quite rapidly, there is kicking out of the mixed genome of DNA around segments that are related to sperm biology in males and also on the X chromosome, which are associated with male infertility often. So we think this is circumstantial evidence that there is male hybrid sterility associated with this mixture process.

We had a huge piece of luck. Svante's group was working with a group of scientists in Russia, and in 2009, I think, approximately, they shared with the group a pinky bone that size from a—excavated in a cave in Siberia, Denisova Cave. And when DNA was obtained from it, it was high-quality DNA, and it was something else. It was neither Neanderthals, these four here, it was not modern humans. It was something else.

It looked like a distant cousin, but very, very distant—hundreds and hundreds of thousands of years from Neanderthals. So this was extremely interesting because there were no tools found near it. There was not a big skull. It was just floating there, but it was something else—a sort of remnant of a group of people that had been living. And we don't know what—we didn't know what they looked like, but we had their genome.

What we found—we applied the same tools that we had developed for studying mixture with Neanderthals. And immediately we saw that these Denisovans had a high degree of connectedness to one particular group in the world, which are New Guineans. So these are people who live in a very equatorial part of the world.

And you see much more matching to New Guineans than to East Asians, for example, or to any other group in the world. And when you do the same analysis, it's not even 2%, it's more like 3 to 5% Denisovan ancestry in New Guineans. Now, Siberia is very far from New Guinea. So what you're seeing is remnants of a population that was spread over a vast region, probably much bigger actually geographically than the Neanderthals were spread over.

We now have built maps of where the Denisovan ancestry is present, and if 3 to 5% is what you see in New Guinea and Australia, about a twentieth of that is what you see in most of East and South Asia. And there's an especially large amount in different parts of the Pacific who are impacted by Papuan-like ancestry.

Just a few months ago, we now have a Denisovan cranial record. Xiaomei Fu and her colleagues obtained DNA and protein sequence from this extremely ancient skull, and it matched as Denisovan. So we're beginning to know what they look like.

So it's been clear that modern and archaic humans mixed everywhere they met. It's not a rare thing for people to mix with people who are quite different from them. It's in fact the rule. So this is work almost entirely from Svante's lab. Here's a paper I was involved in in 2015.

This is the Oase jawbone from a cave in Romania from about 40,000 years ago. This person had a Neanderthal four to six generations back in their family tree—much more Neanderthal than random people that we had data from. Another group study by their group found three individuals in Bulgaria that also were like that.

In Svante's group, they also found, after screening thousands of tiny, uninformative bone fragments in Denisova Cave, most of which were animals, they looked at their protein sequences and saw which of them are human. And of the very small number that were human, they got sequenced from one, and it turned out to be a human, and it was a first-generation hybrid of a Denisovan and a Neanderthal. It's amazing.

So this world was full of people who are mixture of very different humans. Today we're very similar to each other. Even the most different people are at most maybe 200,000 years separated with little gene flow. But 70,000 years ago, there were at least five groups far more different from each other than any groups living today.

A question that I have is: do we still have a fundamentally wrong model of archaic-modern human relationships? This is approximately the picture we now have, without even talking about the recent gene flow between Neanderthals and modern humans in New Guinea, Melanesia, and out of Africa. There's substructure in Africa. There is a deep lineage going into Denisova.

And the standard model has this very interesting event that I hadn't talked about, which is about 5% contribution of African modern humans to Neanderthals about 300,000 years ago. However, the mitochondrial DNA you get from your mother and the Y chromosome you get from your father—these are in Neanderthals something that's related within three or 400,000 years to modern humans.

So if you believe this model, it must come from this event. The probability of that happening is 5% times 5%, so a very, very small number. It's been argued that maybe natural selection for the modern human mitochondrial DNA and Y chromosome, which was better adapted, could explain this, but other work has shown that the rest of the genome for modern humans was strongly selected against. So it's a very strange story.

So it makes one think: are we missing something important, and is there an alternative configuration about the relationships amongst these lineages that might explain the data? Maybe if we think very hard, we'll find it. Or maybe we'll cut some Gordian knot and get more ancient DNA.

Two, so before ancient DNA, we had a fundamentally wrong understanding of how human populations migrate and mix with each other. So I'll tell you a little bit my own story. My own story generating this kind of data started with a frustrating faculty search. I was invited to be on as a guest part of the committee here in 2011.

So as part of this work in Leipzig that I was intensely involved in, I was really convinced that ancient DNA from the last 10,000 years—a topic that Svante wasn't interested in working on—would be transformative for understanding how the world got to be like it is today. There's people speaking 7,000 languages around the world, and that diversity reflects largely things that have happened in the last 10,000 years.

And this technology, which gets DNA out of people 150,000 years ago, maybe it will be really easy for people who lived 3,000 years ago. So let's do that. And I was desperate for someone to be recruited that I could work with. The search committee didn't even invite quite extraordinary applications that we got on the basis that this is, quote, "not relevant to human evolution." And it became clear that no specialist would be invited. It fell between the cracks in the university.

So I retooled my lab very intensively, and we also optimized a new methodology for ancient DNA enrichment, focusing on industrial-scale production of high-quality data, which was what I had done my postdoc in, with working with Eric Lander at what became the Broad Institute. As of a couple of years ago, there were now more than 10,000 individuals as a result of this series of innovations by our group and others.

A prevalent assumption until 2015, which was a reaction to the Nazi abuse of archaeology for politics, was that changes in the material culture record of Europe, which are very dramatic 5,000 to 4,000 years ago, associated with the spread of certain types of ways of burying people and types of goods and pots, were unlikely to be due to migration.

The Nazis had argued that these spreads were associated with the Aryans and were the beginning of the spread of Indo-European languages, and that they spread out of northeastern Europe and central Europe and Germany, and that this what defined the area that was the German homeland and was part of building a claim to territory. That was deconstructed in the subsequent decades.

And it became very difficult in archaeology to argue that migration had occurred as a result of this. And it became a sort of assumption that when cultures change, it should be assumed to be communication of ideas—quote, "pots not people"—that people are communicating with each other, like today people use cell phones all around the world, but they have different backgrounds.

But it was not possible to test this at all. You just couldn't tell if people were related to each other until the advent of ancient DNA, when you could tell exactly—you can see how people are related to each other.

So one of the things that we've been intensely involved in in our group for the last—it's more than 15 years now, even before I became—same time I think I began approximately about 15 years—is the problem of Indo-European language origins. So Indo-European is one—the language we're speaking now, English, almost all European languages are Indo-European languages.

And it was noticed more than 200 years ago by several people, including William Jones in Kolkata, who knew Greek and Latin from his schoolboy days and began learning Sanskrit, that these languages have many shared features, including words and grammars, and that they might stem—or must stem—from some common source which perhaps no longer exists. And the question was where that might have been.

So until 2015, the large-scale view amongst most archaeologists was that it was farming that spread these languages. We know of an economic revolution that happened in Europe and the Middle East and South Asia beginning 11 or 12,000 years ago in the Fertile Crescent, and after 8,500 years ago that explodes into Europe, and it's transformative economically, and the settlement patterns change.

So that would be a natural way that people could have moved into Europe carrying new languages, and there are also similar impacts to the east. So Colin Renfrew in this book argued that this is the spread of Indo-European languages probably out of Anatolia, what's today Turkey.

So beginning in 2015, that picture completely fell apart as a result of ancient DNA. So here is a picture from a series of papers that we wrote that shows how people today relate to each other.

So this is a scatter plot with an x and a y axis I'm not showing you, and each dot is a person living today labeled by where on the map of Europe and the east people come from, and the data is whole genome data. So it's about 600,000 positions in the genome where people differ from each other.

And we multiply the table of a thousand people by each other to see how closely related each person is to every other. And then we display the variation with principal component analysis, which most efficiently separates the people based on how much variation is explained in that matrix of how closely related people are to each other.

So what you see is this amazing pattern where people splay out, without using any labels in the analysis, in two gradients with relatively few people between them. There's a European gradient on one side, and there's a Near Eastern gradient on the other side, and in between are very few individuals who are plausibly the result of more recent contacts between Europe and the Near East.

So this is mostly people who are Greek and people who are culturally Jewish. On the top you see people who are non-Mediterranean, and on the bottom you see people who are Mediterranean. Now I'm going to gray out the points, and I'm going to show you where the ancient people fall on the same plot just learned from the modern samples.

The hunter-gatherers of Europe, the ones who lived before the farmers arrived, are beyond Europe in the direction of European differentiation from the Near East. The only possible explanation for that mathematically is if people in Europe today are mixtures of people like this and people more Middle Eastern. So there was some mixture occurred.

You don't see people who are like this anymore because their ancestry has been mixed in to people like Europeans. Then the first farmers arrive, and they pile up at the bottom of the European gradient on the Mediterranean side. But you don't see people like most Europeans today. For example, English people are over here.

And then about 5,000 years ago, on the steppes north of the Black and Caspian Sea, these mobile steppe pastoralists appear here. And archaeologically they then explode. They explode to Hungary in the west, and they explode to Mongolia in the east. And after 5,000 years ago, boom, you see Europe form. There's a massive mixture of people like this and people like this, and a little bit like this, to form the gradient that we see today. And ever after it looks like that.

And what this is is the spread of these mobile steppe pastoralists into central and western Europe, mixing with the farmers to produce the gradient that is there today. And the argument that these steppe pastoralists had not made an impact was simply that once farming was established, they could not have made an impact.

How could thinly spread herders make an impact on a densely farmed area, similar to how could the British make an impact on India, which was a dense place? They didn't make a demographic impact, but these people did. The data show it, and it's a large impact. It's about three-quarters in Germany. It's 90% in Britain, and so on.

Here's a few examples. Yamnaya people spreading Indo-European languages west. If you look in Britain, there's no eastern ancestry from the steppe until about 4,500 years ago. The last big stones at Stonehenge go up, built by descendants of the first farmers who arrive in Britain about 6,000 years ago. And then boom, it's a 90% population replacement.

In Iberia, in the same time transect, here's the first farmers 6,000 years ago. They arrived there earlier, but I'm starting the time transect at the same time. 4,500 years ago, you still don't see people from the east. And then suddenly you see them. But here there's a time when they're overlapping, and then they collapse into each other 4,000 years ago.

The dots are colored by the Y chromosome of the people and whether it's a steppe eastern chromosome. And what you see in Iberia is that even though it's a lower mixture proportion, only 40% migration from the east, there's no local Y chromosomes left. It's all turned red.

And what that means is that the local male population completely failed to transmit its Y chromosomes to the subsequent population. And so how that happened, we don't know. But several thousand years later, the descendants of these Iberians came to the Americas, and exactly this happened. Today, people in Colombia have almost no local Y chromosomes. They're almost all European, and they have almost no European mitochondrial DNA. They're almost all Native American. And that's the result of exploitation and social inequality. And perhaps that's what occurred here.

These Yamnaya people, these steppe pastoralists, you can see their DNA spreading throughout. We can see and track it all across Europe and into South Asia in time and place, often matching the reconstructed branches of these languages as learned from linguistics. It's very hard to believe that the spread of Yamnaya is not what spread these languages that are spoken today.

But it's super complicated. We don't always see spreads of genes matching culture. Here's a paper we just published this year on the Phoenicians. So Phoenicians are these traders of the Iron Age who lived in what's now Lebanon, and they then were succeeded by the Carthaginians, which are Phoenician colonies in different parts of the eastern Mediterranean, western and central Mediterranean, marked here especially in North Africa, but also in Sardinia, Iberia, and Sicily.

And a question had been like, how are they related to each other? Certainly they spoke the same language. They shared the same religion. Can we get DNA from them and see? And so we got DNA from a couple hundred people and looked, and they're completely unrelated.

So the Phoenicians cluster here on top of earlier populations from the Middle East. And the yellow people are the Phoenician or the Punic people, the Carthaginians. They're just almost completely non-overlapping. There's almost zero contribution. Greek people are over here. North Africans are over here. Carthage is a North African part of the Mediterranean.

And people in Punic sites across the central and western Mediterranean are a mixture of these two groups, with mostly Greek-like, Sicilian-like ancestry and some North African-type ancestry. And it's not more in North Africa than it is in Sicily or Iberia. It's like a cosmopolitan population that's swirling around this part of the world.

The Carthaginians were largely genetically derived, it seems, from the same source as their Greek commercial rivals, which is kind of amazing. And it forces even authors of our own paper to change their assumption. There was this long-standing debate: should you even use this word? The Romans used Punic to refer to the Carthaginians of the central and western Mediterranean. They're really Phoenicians, but genetically they were completely different by the time we were able to sample them.

Another example from this disruption that you get from ancient DNA is about what white people are. So greater than 8,000 years ago, in the place where white people live today, we actually can sample DNA, and there's four groups—two hunter-gatherer groups, two farmer groups—that we can get data from, and they're each as different from each other as East Asians today.

So how did that happen, and how is that related to what you see today? Well, nobody disappeared. They just mixed with each other. That's people living after 4,000 years. And today the low level of differentiation is just due to mixture of these four groups.

Here's another example. There's so many examples, but the Southwest Pacific, so Vanuatu, there's a whole series of papers we've been involved in where the first people of the Southwest Pacific are almost entirely East Asian in ancestry. So no Papuan ancestry, but today people are about 90% Papuan ancestry.

That's because the first people are almost entirely East Asian. They didn't pick up almost any Papuan ancestry on the way to getting to the Pacific in these outrigger canoes and navigating long distances, even though they passed Papuans. But within 500 years, there's a flood of Papuan ancestry, primarily mediated by male Papuans mixing with the local population, and it results in this rapid and long-standing established change.

So the big perspective change from ancient DNA study is that people living today are almost never the descendants of the people in the same place thousands of years before. There are exceptions, but the rule is that everyone's mixed, and no one is or ever really could be pure in any sense.

When we collect DNA from previously unsampled times and places—the Denisovans and many other places, even within the last 10,000 years—major human lineages are found that simply don't exist anymore in unmixed form, or sometimes haven't left very many descendants, but they were widespread at the time.

And large-scale human movements have occurred at multiple time scales, often disruptive to the populations that experience them. And these patterns are not possible to predict and anticipate without direct data.

So three, before ancient DNA, we had a fundamentally wrong understanding that natural selection has been quiescent in the last 10,000 years. So here's a paper from 2011 that says classic selective sweeps were rare in recent human evolution. So this is a genetic variant that arises due to a new mutation and is advantageous and goes up to 100% frequency because it's advantageous.

And there's several lines of evidence for this. The one in this paper is very interesting. I'm not going to explain it. Another is that if you compare Europeans and East Asians, there's almost no genetic differences that are completely different in frequency across them. If there had been strong natural selection, you would expect a lot such events, but there are not.

So for example, just a year ago, Sasha Gusev wrote a—you should read his blog because it's extremely interesting and thoughtful and valuable. This blog, which is: "Where are the recent selective sweeps? How do we make sense of the surprisingly few loci under natural selection?"

So ancient DNA studies—a lot of the interest in ancient DNA was to discover biology, not to discover history. The biggest interest was what makes humans different from Neanderthals biologically. Can we see what's become different from other groups? But the first discoveries were really about population mixture and migrations.

And the main reason for that, perhaps, is just that you can learn so much from a single genome, because a single genome is not just one person. It's a multitude of

people. It's your parents and your grandparents and your eight great-grandparents and your 16 great-grandparents. Each bit of DNA reflects one of those other people. And so you get tens of thousands of independent pieces of information showing how people are related to each other. You get exquisite information about relatedness of people. And so one single genome you can position each person exquisitely relative to others. But to study frequency changes of genetic variants over time to understand biology, you need a large sample size. And we didn't have that at the beginning.

So we started to look. So, we published a paper in 2015 and we looked for places that had changed unusually over the last 8,000 years. We found 12. A study just last year found 21 in a bigger sample size with higher quality data, but it felt like it was maxing out. So, we worked in the last six years to try to improve this. So, we developed several innovations to increase power and to reduce false signals. One was to develop a new statistical method that didn't involve modeling of demographic history but just used the data to see how everybody was related to everybody else and control for all these mixtures and migrations. Another was a very much increased sample size. This study is releasing data from more than 10,000 people in a single study, and to greatly work on data quality through all sorts of data cleaning.

When you now apply this to the data, we found almost 500 places that are genome-scale significant. So this was extremely weird and surprising to us in light of the previous literature and we spent years trying to make this go away, but it kept getting stronger. It sort of felt a little bit like what we had done with the Neanderthal work and the evidence of mixture. So how do we reconcile this observation of this vast load of genome-scale significant findings, about 500 of them, and we think that actually there's at least 5,000, but many of them are sub-significant, with the lack of evidence for selective sweeps? Well, one answer is we're definitely in a period, at least in Europe, of intensified evolution.

So what we can see is that there is enrichment for evolution, for genetic variants that we know from genome-wide association analysis affect immune traits and metabolic traits beyond background, and that there's an acceleration in that enrichment statistically after 5,000 years ago. So we're in a period of accelerating biological change in this period, provably based on the ancient DNA data. And the second answer is that their selection is fluctuating over time. So if you plot the true ages of the mutations that we find under selection here against their expected age if they had been under positive natural selection the whole time, you would expect it to be on an x equals y line, and about a third of them are, but about two-thirds of them are not, and in fact the red ones have reversed direction, or it's the ancient allele that's being positively selected.

So what's happening is these are genetic variants that were existing in the population before that have come under a different type of selection and reversed. And so the reason they haven't become 100% different in frequency between Europeans and East Asians is that there are variants that go up and then they go down and they go up and they go down in different environments. Here's two examples for celiac disease. This is related to gluten insensitivity, which both shot up in the last 4,000 years. These are risk factors for celiac disease. They probably do something valuable in addition to causing risk for celiac disease. Here's a major risk for tuberculosis, for clinically severe tuberculosis. It's now at a few percent frequency, but 3,000 years ago it was 9% and it rocketed downward under the pressure of selection, but before that it rocketed upward. You can see the change in both of these are significant. And I don't know, maybe tuberculosis becomes endemic and that's why it decreases in frequency.

You can also look at combinations of genetic variants that shape complex traits. So for studies of complex traits like diabetes, you can look at tens of thousands of people with type two diabetes and tens of thousands of people without and look to see which genetic variants affect this trait. And you can do a linear combination of the variants that predict risk for type two diabetes to make an even better predictor, a sum of about a thousand variants, for example. And you can watch how that changes over time. And you can ask, does that combination of variants that's a good predictor for type two diabetes risk change over time more than expected by chance? And what we see is that that's pushed down by about a standard deviation on a scale of current variation over about 10,000 years, clearly under the pressure of selection as well as type two diabetes related traits.

You also see this for some psychotic related traits like schizophrenia where there's clear selection against what today manifests as schizophrenia. It does not mean schizophrenia was more common in the past. It could be that environments today are more likely to cause schizophrenia than environments in the past, but it does mean that people with the genetic inheritance of people from 10,000 years ago put in today's environment would be more likely to get this. And there's also traits like that are traits of like that are years of schooling which believe it or not have strong genetic predictability that have increased over time, about in this case by about six standard deviations.

We tried to make results like these go away. And we wondered if they were a statistical artifact perhaps of inadequate correction for migrations and mixtures in our study but they replicate, for example, the years of schooling one, when we measure the effects of the variants on years of schooling in East Asians and correlate them to the trajectory in West Eurasians, two groups that have almost completely uncorrelated population structure. It's very hard to see that that could be a result by chance and they replicate when we measure the effects of these in families where it's very hard or almost impossible to see how that could be the result of population structure.

So the summary of the talk is that this is three examples of how ancient DNA has really forced our hand, challenging understanding of the world in deep ways. Before ancient DNA we had a fundamentally wrong understanding of how modern humans relate to archaic humans. Before ancient DNA, we had a fundamentally wrong understanding of how human populations migrate and mix with each other. And before ancient DNA, we had a fundamentally wrong understanding that natural selection has been quiescent in the last 10,000 years. Thanks, this is a photograph of people currently in our research group here and there are many past lab members who contributed to this work and many collaborators and funders. And now I'd like to invite Carl Zimmer to join me for a conversation.

I've never sequenced my own genome and I've never sent my DNA to any direct-to-consumer ancestry testing company. I don't know if it's stubbornness at this point. But I've never done it.

That's an interesting question. I don't know. It just seems very important, this stuff. It just seems fundamentally important. So I don't really think about it that way. It just seems like an important set of facts about the world that we're related to archaic humans in these ways. And so I don't think about that. That's not a very helpful answer.

No, it doesn't really make a difference and I actually think a lot of those numbers are exaggerated. So, as far as I can tell, I don't know what the current estimates are. They may just be counting informative positions at some point, but the actual estimates we obtain have statistical errors about 0.5% on a per individual basis. And the actual theoretical variation is about 0.1 or 0.05%. So, I think people are just quoting noise mostly if they're non-Africans. If they're African-American, it will be really reduced because they'll have mostly West African ancestry, for example. But I think people are really to a large extent just interested in if they say, oh, I have 3% Neanderthal ancestry, it's probably just a two standard deviation variation just from the measurement, not from the actual real inheritance.

About 2% of your DNA is from Neanderthals. 2% is a lot. The Neanderthal DNA is actually especially different. So it's loaded with variation compared to random variation in the genome. It's also been selected against. So that depletes its biological significance on average. So it's not clear whether on average that 2% is more or less important than average. There's, I think if I remember right, about 2,000 segments of Neanderthal DNA estimated in, if I remember right, I may get this number off, in each non-African person. And so the error around that number is very, very small. So almost everybody will have just within, you know, 3% of that number just by chance.

I think it gives us a lot of clues. I mean, I don't know if we know what happened when Neanderthals and modern humans met and mixed. The Neanderthals as groups disappear and they get mixed into modern humans in Eurasia. Whether that was violent or whether that was just environmental degradation, we don't know. But I think documenting these events as they occur through ancient DNA allows us to obtain a much richer picture of what happened. In principle, one could analyze the genomic distribution of segments of Neanderthal DNA in modern humans to learn something about the nature of the event. Kind of like we study the in Iberia and you see that there's very little local Y chromosome contribution from the local population. Maybe you could hope to do something like that. But there's so much natural selection that's shaped the bits of DNA from Neanderthals after it that it distorts the picture and it makes it hard to read that off.

One of the things I wanted to say in response to your sort of comment about that very complicated tree is that that tree is the result of an accretion of findings. So one finding after the other established one of those lineages or mixture events and they kind of got grafted onto each other. And it sometimes reminds me a little bit of the sort of old models in like 2,000 years ago, like these Ptolemaic epicycles adding things together to make the solar system seem to work still. And one wonders whether there's a reconfiguration that might make things work differently.

Just another one. When the Denisovan genome was first obtained in Svante's group, the mitochondrial DNA is like 800,000 years diverged or 700,000 years or 900,000 years diverged. And it seemed to be an outgroup, a more distant relative of both modern humans and Neanderthals. Modern humans and Neanderthals were sisters and Denisovan was much more distant. But then the whole genome was sequenced and it was very clear that Neanderthals and Denisovans were sisters and I think we were in an age at the time of the superiority of the genome. So if you have the whole genome it tells you they're most closely related then that's the main story. But actually there's a diversity of stories. Each part of the genome tells its own story and they're all valid stories. And if the Y chromosome and the mitochondrial DNA tell you a profoundly different story, why should we ignore them? They may be numerically less important, but that might be because of some other biological or demographic process that if we think hard enough, we can figure out. So, it's not really at all clear to me what the right solution is. It's just that the current picture feels so wrong to me.

That would be, I mean, that is often how it gets solved when a unique specimen is obtained and resolves it. There's a very exciting specimen that's been sequenced in various forms in Leipzig and Svante's group from Spain, Sima de los Huesos, and I didn't talk about it very much but it's surprising, it constrains the story and makes it difficult to come up with solutions that are alternative.

Yeah, that's a great question. I think archaeologists thinking about this, there is burial traditions that are associated with the Yamnaya that spread very dramatically and archaeologists, even the strongest proponents of the Yamnaya being involved in spreading Indo-European languages, really couldn't bring themselves to argue that this was really a truly large number of people. It seemed unbelievable. So they would argue, well, this is maybe an elite cultural tradition that gets learned and adopted by neighbors and sweeps through bringing people's languages. It's like the same reason, for example, Hungarian maybe gets adopted even though there's not a large-scale demographic replacement by people from the east. Maybe that's what's going on. Nobody dared to say there were a large number of people. But the DNA screams at you, there's a huge number of people associated with this. There's populations collapsing and being displaced or admixed with these groups that are expanding. People make arguments about maybe very early use of the horse by the Yamnaya might have given them an advantage. There's arguments about whether they used the horse uniquely or in a truly domesticated way. Their use of wagons, the use of this steppe as a food source. So people make these arguments, but the actual observation that they made, that expansion, is just an unbelievable thing.

Right, so we are five to seven years behind depending on where we are in terms of what's happening with European ancient DNA in some of the places you mentioned. But it's exponentially growing also in sub-Saharan Africa with working data in places that are very hot and you might not expect preservation to exist. And in the places in sub-Saharan Africa we've gotten data from many of the same types of patterns that are observed. Example one, you get DNA from Malawi in South Central Africa 3,000 years ago to 8,000 years ago. There's a population there that's separated for 150,000 years from everybody living today that barely contributed to anybody today. It's established there for a long period of time and then there's an expansion of people ultimately probably from areas around Cameroon associated with Bantu languages. That's almost a complete displacement of this group.

You look in Cameroon itself, you think this should be where these people come from 3,000 to 8,000 years ago in the rock shelter in the grasslands area of Cameroon, supposed to be the homeland of the Bantu expansion. They're completely different from the Bantu. So everywhere you look where you get data from places that have not had data from before, people are different in the past from how they are today. Either completely different or mixed with the present-day populations. So certainly in different parts of sub-Saharan Africa, we're seeing these patterns. Similarly to Europe and the Middle East and East Asia, the same, not very good data from South Asia but it seems likely for multiple reasons and the little data we do have that it's the same. Patterns are qualitatively somewhat different in the islands of the Pacific, even with a short time, often the same as I mentioned in the native in the Americas, it may be qualitatively different with less frequent turnover.

Well, this is speculative of course, but certainly the case in Iberia I told you about can't have been a happy occasion for the men involved, the local men involved. They did not succeed in transmitting their Y chromosomes to subsequent generations. Whether that's due to the local women preferring foreign men or whether that's due to the local men being, you know, dying in some way, you know, we don't know. But the sex-biased admixture is sort of a canary in the coal mine sign of inequality or unequal interactions amongst communities.

We could also think, and actually it's a useful thing to think about, about the Neanderthal case in Europe. Neanderthals disappear pretty rapidly over a few thousand-year period across Europe and actually it's a wave of loss. So in certain parts it's earlier, in other parts it's later, but in any one part it might be relatively quickly. And so you might think a simple story like is dramatized in some of the novels about this, like that the modern humans are outcompeting the Neanderthals because they're more violent, their culture is more aggressive or maybe they have some technology, even the Neanderthals are very sophisticated and who knows, but actually what's probably happening is like lots of extinctions locally of populations for various reasons and most of the modern human early groups in Europe also disappear.

So the individuals from Bacho Kiro, the individuals from Oase I talked about and other individuals, they're from lineages that didn't contribute substantially to later Europeans. And there's 39,000 years ago there's a climatic fluctuation that results in almost everybody who was living there before in Europe, including the Neanderthals, but also including all the modern humans seeming to disappear and they're replaced by more humans from probably the Near East or maybe one isolated group in Europe. So it's not clear that it's a systematic thing but rather many, many groups going one way or the other.