How sure are we that later Neanderthal uniparentals are introgressed from Africans and not the other way around? My understanding of that initial claim is that earliest Nean and Denisovan uniparentals are a clade, but that later Nean uniparentals are more similar to modern humans. What makes it impossible that those uniparentals originated in Neanderthals and introgressed into Africans?
I think the main argument is in comparison with 430 kY Sima de los Huesos genome, which is close to later Neanderthals in autosomes but much closer to Denisovans rather than N in mtDNA. Later N have mtDNA closer to HS rather than D. The existing models are most compatible with the scenario of transfer of both uniparental lines from HS to N some time between 250 and 350 kY ago.
Personally I don't believe these models are close enough to real history. They operate on "pulses" concept, but if there was a continuous bidirectional gene flow (even if very weak in some periods), it is much harder to determine which allele belongs to which population.
And there are more and more signs that the flow was nearly continuous (with some breaks of course) up to the complete absorption of N into larger HS deme.
But geneticists are uncomfortable with a continuous flow, first because it is harder to model and second (or actually first) because it not quite compatible with Out of Africa generally - which, even in a weaker formulation, states that if there were earlier "exits from Africa" they must have gone completely extinct. There is more and more evidence that it was not so.
Great comment. I agree with you that more frequent contacts and gene flow are a more realistic scenario than the stepwise "big event" models in most genetic work. I had this debate with David Reich now 15 years ago, I thought then that the kind of island-model few-contact models were going down a blind path that was not leading to biological insights.
I would add that one thing that has adjusted my thinking is the Denisova evidence. The genomes from that site demonstrate clearly that frequent, small-scale gene flow was happening across a very long time. But at the same time it shows that these contacts rarely propagated genes back far. Slow and more-or-less continuous gene flow within a source-sink population structure does over time start to yield similar patterns of genetic variation as an island model with discrete contacts. So it does become a matter of scale: the island-tree type model is not necessarily a bad approximation at the largest scale but is worse and worse at lower scales.
Great question — Kirill mostly answered below, it really is the contrast of Sima de los Huesos and later Neanderthal genomes that weighs most on this. I would add the evidence from the nuclear genome (including the X chromosome work in this current study) weighs in favor of Africa -> Neanderthal and not the opposite direction.
The Nature study is paywalled, do you know what percent African/HS they estimate the Neanderthal X to be?
I understand the point about the difference between Sima de los Huesos and later N genomes. I admit, I just find it very odd trying to reconcile a couple of distinct claims relating to N demographic and genetic structure:
1. N's lived in fragmented, isolated groups with minimal gene flow between them, as seemingly demonstrated by very low heterozygosity compared to modern HS.
2. Somehow HS Y and mtDNA (and apparently a good chunk of X ancestry as well) completely swept through the N population, totally replacing their native uniparentals, but only managed to replace 6% of the overall autosomal DNA? Daniel Tabin, a PhD at Reich's lab, just noted this as well during a podcast with Razib Khan (see around the 1:07 mark https://www.youtube.com/watch?v=Fs8baY3le1w) and even touches on the low mathematical probability of something like this occurring, and how incongruous it is from the perspective of a population practicing either patrilocality or matrilocality.
3. Greg Cochran offered a possible explanation once, that because of greater genetic load in N due to their small effective population size, their uniparentals might have been dinged up in particular, making introgression of HS uniparentals more likely. But echoing back to Tabin's point, if N were really so messed up already at 250k ybp that their native uniparentals were so defective relative to HS, why did only 6% of their autosome get replaced? Why didn't the main OOA replacement of N occur right when that initial HS > N happened, as opposed to only 50K ybp?
Tabin made a great comment about how we keep adding new "epicycles" to our standard model, and I think that's exactly right. Again, I don't have access to this current paper, so maybe there's more details explicated that would attenuate my skepticism, but it seems like the instinct in this topic is always to try to amend the standard model instead of stepping back and thinking of newer models that might better fit the data.
This is a series of good questions, and it will take a longer answer than a comment. You center on a key issue: Neanderthal demography seems like it was borderline non-viable, and that led to many possibly bad genetic consequences. Moreover, they seem demographically fragmented with low connectivity across their geographic range.
African introgression seems like a paradox: How did it spread across a fragmented Neanderthal range? And if it was so consequential, why didn't the Africans just roll over the Neanderthals or genetically swamp them sooner? (Maybe equally likely given the large number of documented contacts and range overlap, why didn't the Denisovans ever roll over the Neanderthals?)
Several aspects of Neanderthal variation factor into my thinking about these questions:
1. Neanderthals never seem to have developed the deep regional differences seen in Denisovans or Africans, despite having exactly the same time to do so, and occupying close to the same geographic extent. We do not have direct DNA from southern (southwest Asian) Neanderthals, but these were likely the introgression source population for modern people 45,000 years ago, which genetically was not far from Vindija. It looks like there was substantive replacement of western Neanderthals from an eastern source population sometime after 150,000 years ago. We're talking about clades or lineages of Neanderthals, but all those are very shallow compared to the depth of lineages in both African and Denisovan groups.
2. Climate and environment mattered to Neanderthal evolution in interesting ways. Arguably, their geographic range was more uniform in environments than either Africa or the Denisovan range in East, South, and Southeast Asia. This is the Jared Diamond argument applied to the Pleistocene. Some parts of the Neanderthal range were evidently low-yield and did not sustain long-term populations even though archaeology shows that Neanderthals were intermittently present. Habitat uniformity would have led to a larger long-term size for adaptive evolution. Meanwhile, small-scale demographic fragmentation could not go too far as long as adaptation was continuing in some regions with locally-high carrying capacity that could become demographic sources.
3. The record after 200,000 years ago shows repeated penetration of the Neanderthal geographic range by African and Denisovan groups. You've got possible African-derived groups again and again in southwest Asia and as far as Greece before 150,000 years ago. In the period after 100,000 years ago, you have what seem like short-term dispersals of modern people as far as France. It would be very surprising to me if these dispersals did not lead to genetic impacts on Neanderthal groups, and in the case of the Denisovans we see those local impacts in central Asia. It's interesting that the archaeological record may show more visible effects of these contacts than the genetic record.
4. Uniparental systems are brittle and subject to continual replacement. I don't find it surprising in the least that mtDNA or Y in Neanderthals were replaced from outside. This clearly happened across most African groups also. I did write about this more than 20 years ago, this has been understood for a long time.
My own thinking is that the phenomenon we recognize as "Neanderthals" is a product of the distinctive regional ecology of western Eurasia, and the population cannot be understood separate from its habitats. Resilience was centered in source populations where local circumstances interacted synergistically with Neanderthal genetic and cultural adaptations to enable sustained presence. Resistance to genetic swamping comes from the net out-migration from these sources. It's a lot like the solar wind. It's not that you cannot penetrate it or travel through it. But your impacts will be temporary unless you affect the source, and when you affect the source your impacts will be major.
The introgression pulse (or pulses) at or before 250,000 years ago happened when African introgression did affect a source population. From there, the dispersal throughout the Neanderthal range came for free. Some of the African genes made a big difference to the adaptation of later Neanderthals -- in fact, "Neanderthals" as anthropologists understand them are entirely a product of this African mixture onto the Eurasian genetic background. No "classic Neanderthal" would ever have existed without this African interaction.
The system broke down entirely after 50,000 years ago when modern humans finally tipped across one or more of the Neanderthal source populations. How that happened is the big question of modern human origins. But it's valuable to see that this was a very localized event. Once a source was gone, the rest of Neanderthals were inevitably doomed to swamping. (And once modern humans spread across the former Neanderthal range, most of them, too, would be swamped by later waves of newcomers.)
Thanks for this great discussion. I particularly liked the socio-behavioral part in the main essay!
It would be great to have a major piece on the X, Y, and mtDNA evolutions/dispersions!
One thing I don't have any real grasp on is how robust the evidence is for some major hypotheses (some now often taken as dogma). Are there more than just 2 partial Sima de los Huesos mitogenomes? More than just 3 (and all "late") Neanderthal Y chromosomes? More than just 2 Denisovan Y chromosomes? How much X-chromosome sequence data from SH, N, D, and archaic HS fossils is there?
Some years ago there was some speculation about "archaic introgression" (although that term wasn't in vogue then) of a Cameroonian Y haplotype (00, iirc). What's the current interpretation with that?
I do like reading about ghost populations inferred from analysis of modern populations, but nothing beats sequence data from actual fossils!
Great comment. One thing to keep in mind is that "dogma" in this area flips around pretty fast. It's true that some areas are subject to kind of lockstep skepticism of alternative thinking, that tends to be reinforced with the lab meetings of particular research groups. It's always good to point that out. I tend to watch carefully what people say about mtDNA and Y chromosome because there is a lot of unjustified confidence in these as tracers of population movements in recent people. This often builds from an assumption that they are never influenced by selection, which is a bad assumption.
The DNA data represent only a small number of Neanderthal and Denisovan individuals. There's a very long timespan involved. The mtDNA sample of Neanderthals is around 50 or so, but the number exceeding 130,000 years is only three. Y chromosome data come only from Spy 94a, Mezmaiskaya 2, and El Sidrón 1253, all less than 65,000 years old.
A00 is a very interesting story as an extant representative of a long branch, but its time of common ancestor with other Y variants in humans is not enormously old, basically around 250,000-200,000 years ago. All Y variation in living humans seems to be more recent than the diversification of autosomal genomes of the major African groups, and that's likely the selection dynamic at work on Y chromosomes. There is every possibility that some Denisova-like Y chromosomes might be floating around the Neanderthal population without being noticed by our current samples, and indeed ruling out an A00-like long branch is likely impossible with any conceivable aDNA sample.
This is the main reason why the autosomal comparisons carry a lot more weight for me. An autosomal genome includes thousands of replicates of the same history. One individual gives us hundreds of times better constraint on population history than the largest sample of aDNA is ever going to provide for these populations.
Thanks for the (as always) thorough response! I have a couple of follow-ups:
1) Would the older OoA-archaic Homo sapient introgression that replaced the pre-existing (presumably more Denisovan-like) Neanderthal mtDNAs and Y chromosomes have also significantly flooded the pre-existing Neanderthal X chromosome pool?
2) So is A00 not an introgression but a lucky relic of the African Y population diversity that later (unluckily) didn't make it into the founder OoA-Hs population (or didn't make it through the bottleneck post exit)?
I would like to chat with you about something related to individual genomes and potential information on genomic diversity, but I don't know the appropriate venue. Maybe DM? But the segue would be: Have you ever been to the RSA Golden Gate national park (boundary of Free State and KZN), or to the Fort Nottingham historic site in KZN?
Thanks! I know Golden Gate well, but have not visited Fort Nottingham.
With A00, the clade depth is not as great as initially thought, and so there's not a strong rationale for it to be "ghost" in origin, although it is of course possible. Regarding the Neanderthal X variation, that's the main finding of the study described on the OP, it's a disproportionately high influence of African X chromosomes on the later Neanderthals (represented by the Altai Neanderthal).
How sure are we that later Neanderthal uniparentals are introgressed from Africans and not the other way around? My understanding of that initial claim is that earliest Nean and Denisovan uniparentals are a clade, but that later Nean uniparentals are more similar to modern humans. What makes it impossible that those uniparentals originated in Neanderthals and introgressed into Africans?
I think the main argument is in comparison with 430 kY Sima de los Huesos genome, which is close to later Neanderthals in autosomes but much closer to Denisovans rather than N in mtDNA. Later N have mtDNA closer to HS rather than D. The existing models are most compatible with the scenario of transfer of both uniparental lines from HS to N some time between 250 and 350 kY ago.
Personally I don't believe these models are close enough to real history. They operate on "pulses" concept, but if there was a continuous bidirectional gene flow (even if very weak in some periods), it is much harder to determine which allele belongs to which population.
And there are more and more signs that the flow was nearly continuous (with some breaks of course) up to the complete absorption of N into larger HS deme.
But geneticists are uncomfortable with a continuous flow, first because it is harder to model and second (or actually first) because it not quite compatible with Out of Africa generally - which, even in a weaker formulation, states that if there were earlier "exits from Africa" they must have gone completely extinct. There is more and more evidence that it was not so.
Great comment. I agree with you that more frequent contacts and gene flow are a more realistic scenario than the stepwise "big event" models in most genetic work. I had this debate with David Reich now 15 years ago, I thought then that the kind of island-model few-contact models were going down a blind path that was not leading to biological insights.
I would add that one thing that has adjusted my thinking is the Denisova evidence. The genomes from that site demonstrate clearly that frequent, small-scale gene flow was happening across a very long time. But at the same time it shows that these contacts rarely propagated genes back far. Slow and more-or-less continuous gene flow within a source-sink population structure does over time start to yield similar patterns of genetic variation as an island model with discrete contacts. So it does become a matter of scale: the island-tree type model is not necessarily a bad approximation at the largest scale but is worse and worse at lower scales.
Great question — Kirill mostly answered below, it really is the contrast of Sima de los Huesos and later Neanderthal genomes that weighs most on this. I would add the evidence from the nuclear genome (including the X chromosome work in this current study) weighs in favor of Africa -> Neanderthal and not the opposite direction.
The Nature study is paywalled, do you know what percent African/HS they estimate the Neanderthal X to be?
I understand the point about the difference between Sima de los Huesos and later N genomes. I admit, I just find it very odd trying to reconcile a couple of distinct claims relating to N demographic and genetic structure:
1. N's lived in fragmented, isolated groups with minimal gene flow between them, as seemingly demonstrated by very low heterozygosity compared to modern HS.
2. Somehow HS Y and mtDNA (and apparently a good chunk of X ancestry as well) completely swept through the N population, totally replacing their native uniparentals, but only managed to replace 6% of the overall autosomal DNA? Daniel Tabin, a PhD at Reich's lab, just noted this as well during a podcast with Razib Khan (see around the 1:07 mark https://www.youtube.com/watch?v=Fs8baY3le1w) and even touches on the low mathematical probability of something like this occurring, and how incongruous it is from the perspective of a population practicing either patrilocality or matrilocality.
3. Greg Cochran offered a possible explanation once, that because of greater genetic load in N due to their small effective population size, their uniparentals might have been dinged up in particular, making introgression of HS uniparentals more likely. But echoing back to Tabin's point, if N were really so messed up already at 250k ybp that their native uniparentals were so defective relative to HS, why did only 6% of their autosome get replaced? Why didn't the main OOA replacement of N occur right when that initial HS > N happened, as opposed to only 50K ybp?
Tabin made a great comment about how we keep adding new "epicycles" to our standard model, and I think that's exactly right. Again, I don't have access to this current paper, so maybe there's more details explicated that would attenuate my skepticism, but it seems like the instinct in this topic is always to try to amend the standard model instead of stepping back and thinking of newer models that might better fit the data.
This is a series of good questions, and it will take a longer answer than a comment. You center on a key issue: Neanderthal demography seems like it was borderline non-viable, and that led to many possibly bad genetic consequences. Moreover, they seem demographically fragmented with low connectivity across their geographic range.
African introgression seems like a paradox: How did it spread across a fragmented Neanderthal range? And if it was so consequential, why didn't the Africans just roll over the Neanderthals or genetically swamp them sooner? (Maybe equally likely given the large number of documented contacts and range overlap, why didn't the Denisovans ever roll over the Neanderthals?)
Several aspects of Neanderthal variation factor into my thinking about these questions:
1. Neanderthals never seem to have developed the deep regional differences seen in Denisovans or Africans, despite having exactly the same time to do so, and occupying close to the same geographic extent. We do not have direct DNA from southern (southwest Asian) Neanderthals, but these were likely the introgression source population for modern people 45,000 years ago, which genetically was not far from Vindija. It looks like there was substantive replacement of western Neanderthals from an eastern source population sometime after 150,000 years ago. We're talking about clades or lineages of Neanderthals, but all those are very shallow compared to the depth of lineages in both African and Denisovan groups.
2. Climate and environment mattered to Neanderthal evolution in interesting ways. Arguably, their geographic range was more uniform in environments than either Africa or the Denisovan range in East, South, and Southeast Asia. This is the Jared Diamond argument applied to the Pleistocene. Some parts of the Neanderthal range were evidently low-yield and did not sustain long-term populations even though archaeology shows that Neanderthals were intermittently present. Habitat uniformity would have led to a larger long-term size for adaptive evolution. Meanwhile, small-scale demographic fragmentation could not go too far as long as adaptation was continuing in some regions with locally-high carrying capacity that could become demographic sources.
3. The record after 200,000 years ago shows repeated penetration of the Neanderthal geographic range by African and Denisovan groups. You've got possible African-derived groups again and again in southwest Asia and as far as Greece before 150,000 years ago. In the period after 100,000 years ago, you have what seem like short-term dispersals of modern people as far as France. It would be very surprising to me if these dispersals did not lead to genetic impacts on Neanderthal groups, and in the case of the Denisovans we see those local impacts in central Asia. It's interesting that the archaeological record may show more visible effects of these contacts than the genetic record.
4. Uniparental systems are brittle and subject to continual replacement. I don't find it surprising in the least that mtDNA or Y in Neanderthals were replaced from outside. This clearly happened across most African groups also. I did write about this more than 20 years ago, this has been understood for a long time.
My own thinking is that the phenomenon we recognize as "Neanderthals" is a product of the distinctive regional ecology of western Eurasia, and the population cannot be understood separate from its habitats. Resilience was centered in source populations where local circumstances interacted synergistically with Neanderthal genetic and cultural adaptations to enable sustained presence. Resistance to genetic swamping comes from the net out-migration from these sources. It's a lot like the solar wind. It's not that you cannot penetrate it or travel through it. But your impacts will be temporary unless you affect the source, and when you affect the source your impacts will be major.
The introgression pulse (or pulses) at or before 250,000 years ago happened when African introgression did affect a source population. From there, the dispersal throughout the Neanderthal range came for free. Some of the African genes made a big difference to the adaptation of later Neanderthals -- in fact, "Neanderthals" as anthropologists understand them are entirely a product of this African mixture onto the Eurasian genetic background. No "classic Neanderthal" would ever have existed without this African interaction.
The system broke down entirely after 50,000 years ago when modern humans finally tipped across one or more of the Neanderthal source populations. How that happened is the big question of modern human origins. But it's valuable to see that this was a very localized event. Once a source was gone, the rest of Neanderthals were inevitably doomed to swamping. (And once modern humans spread across the former Neanderthal range, most of them, too, would be swamped by later waves of newcomers.)
Thanks for this great discussion. I particularly liked the socio-behavioral part in the main essay!
It would be great to have a major piece on the X, Y, and mtDNA evolutions/dispersions!
One thing I don't have any real grasp on is how robust the evidence is for some major hypotheses (some now often taken as dogma). Are there more than just 2 partial Sima de los Huesos mitogenomes? More than just 3 (and all "late") Neanderthal Y chromosomes? More than just 2 Denisovan Y chromosomes? How much X-chromosome sequence data from SH, N, D, and archaic HS fossils is there?
Some years ago there was some speculation about "archaic introgression" (although that term wasn't in vogue then) of a Cameroonian Y haplotype (00, iirc). What's the current interpretation with that?
I do like reading about ghost populations inferred from analysis of modern populations, but nothing beats sequence data from actual fossils!
Great comment. One thing to keep in mind is that "dogma" in this area flips around pretty fast. It's true that some areas are subject to kind of lockstep skepticism of alternative thinking, that tends to be reinforced with the lab meetings of particular research groups. It's always good to point that out. I tend to watch carefully what people say about mtDNA and Y chromosome because there is a lot of unjustified confidence in these as tracers of population movements in recent people. This often builds from an assumption that they are never influenced by selection, which is a bad assumption.
The DNA data represent only a small number of Neanderthal and Denisovan individuals. There's a very long timespan involved. The mtDNA sample of Neanderthals is around 50 or so, but the number exceeding 130,000 years is only three. Y chromosome data come only from Spy 94a, Mezmaiskaya 2, and El Sidrón 1253, all less than 65,000 years old.
A00 is a very interesting story as an extant representative of a long branch, but its time of common ancestor with other Y variants in humans is not enormously old, basically around 250,000-200,000 years ago. All Y variation in living humans seems to be more recent than the diversification of autosomal genomes of the major African groups, and that's likely the selection dynamic at work on Y chromosomes. There is every possibility that some Denisova-like Y chromosomes might be floating around the Neanderthal population without being noticed by our current samples, and indeed ruling out an A00-like long branch is likely impossible with any conceivable aDNA sample.
This is the main reason why the autosomal comparisons carry a lot more weight for me. An autosomal genome includes thousands of replicates of the same history. One individual gives us hundreds of times better constraint on population history than the largest sample of aDNA is ever going to provide for these populations.
Thanks for the (as always) thorough response! I have a couple of follow-ups:
1) Would the older OoA-archaic Homo sapient introgression that replaced the pre-existing (presumably more Denisovan-like) Neanderthal mtDNAs and Y chromosomes have also significantly flooded the pre-existing Neanderthal X chromosome pool?
2) So is A00 not an introgression but a lucky relic of the African Y population diversity that later (unluckily) didn't make it into the founder OoA-Hs population (or didn't make it through the bottleneck post exit)?
I would like to chat with you about something related to individual genomes and potential information on genomic diversity, but I don't know the appropriate venue. Maybe DM? But the segue would be: Have you ever been to the RSA Golden Gate national park (boundary of Free State and KZN), or to the Fort Nottingham historic site in KZN?
Thanks! I know Golden Gate well, but have not visited Fort Nottingham.
With A00, the clade depth is not as great as initially thought, and so there's not a strong rationale for it to be "ghost" in origin, although it is of course possible. Regarding the Neanderthal X variation, that's the main finding of the study described on the OP, it's a disproportionately high influence of African X chromosomes on the later Neanderthals (represented by the Altai Neanderthal).