“Out of Africa” bottleneck is what really matters for mutations


At least in relation to mutational load, if you read a new preprint in biorxiv, The demographic history and mutational load of African hunter-gatherers and farmers:

The distribution of deleterious genetic variation across human populations is a key issue in evolutionary biology and medical genetics. However, the impact of different modes of subsistence on recent changes in population size, patterns of gene flow, and deleterious mutational load remains to be fully characterized. We addressed this question, by generating 300 high-coverage exome sequences from various populations of rainforest hunter-gatherers and neighboring farmers from the western and eastern parts of the central African equatorial rainforest. We show here, by model-based demographic inference, that the effective population size of African populations remained fairly constant until recent millennia, during which the populations of rainforest hunter-gatherers have experienced a ~75% collapse and those of farmers a mild expansion, accompanied by a marked increase in gene flow between them. Despite these contrasting demographic patterns, African populations display limited differences in the estimated distribution of fitness effects of new nonsynonymous mutations, consistent with purifying selection against deleterious alleles of similar efficiency in the different populations. This situation contrasts with that we detect in Europeans, which are subject to weaker purifying selection than African populations. Furthermore, the per-individual mutation load of rainforest hunter-gatherers was found to be similar to that of farmers, under both additive and recessive modes of inheritance. Together, our results indicate that differences in the subsistence patterns and demographic regimes of African populations have not resulted in large differences in mutational burden, and highlight the role of gene flow in reshaping the distribution of deleterious genetic variation across human populations.

There’s two major moving parts in this preprint. First, they using phylogenomic methods to explicitly model population history. Second, they integrated their demographic results in generation and interpreting the distribution of mutations within the exomes of these populations. That is, they combined phylogenomics to gain insight into population genomics, as the latter focuses more on the parameters which define variation with a population.

The data they worked with was from the exome. The regions of the genome which translate into genes. That’s ~30 million bases. They get really good precision due to high coverage, hitting site about 70 times. Their sample was about 300 Africans and 100 Europeans, and they got ~500,000 polymorphisms or variants for their trouble.

The populations were labeled by subsistence and provenance. The Europeans were Belgians. For the Africans they had two groups of hunter-gatherer Pymgies, and two groups of Bantu agriculturalists, sampled from western and eastern locations as you see on the map above.

The admixture plots, which separate out individuals into K numbers of populations break out in a way that makes sense. First, Europeans separate, and the eastern agriculturalist populations have a little bit of evidence of European-like ancestry. This is almost certainly Middle Eastern farmer, which has been found in many East African populations, and those populations which have mixed with them. Then the hunter-gathers separate from the agriculturalists. This is in line with expectation and earlier research; the hunter-gatherers of Africa seem very different from the agriculturalists, and are actually more closely related to each other than the agriculturalists in their neighboring regions.

The exception to this pattern is caused by recent gene flow, which is clearly evident above. Due to population size differences it looks like there is more agricultural ancestry in the Pygmies than vice versa. I wish that they had sampled Mbuti Pygmies. I’m told that this group has the least agricultural admixture.

But then they decided to get fancy and explicitly model demographic histories with fastsimcoal2. What does this do? From the website for the software:

While preserving all the simulation flexibility of simcoal2, fastsimcoal is now implemented under a faster continous-time sequential Markovian coalescent approximation, allowing it to efficiently generate genetic diversity for different types of markers along large genomic regions, for both present or ancient samples. It includes a parameter sampler allowing its integration into Bayesian or likelihood parameter estimation procedure.

fastsimcoal can handle very complex evolutionary scenarios including an arbitrary migration matrix between samples, historical events allowing for population resize, population fusion and fission, admixture events, changes in migration matrix, or changes in population growth rates. The time of sampling can be specified independently for each sample, allowing for serial sampling in the same or in different populations.

The models you see that were tested are pretty simple, and they all seem plausible I suppose. Their simulations suggested that the three above scenarios, with alternative branching patterns and various gene flows, were all of equal likelihood. That is, given the models and the data that they had (4-fold synonymous sites which are likely to be neutral) you can’t distinguish which is right.

In all the models hunter-gatherers diverged relatively recently and so did the agriculturalists. Europeans, who are stand-ins for all non-Africans in this scenario, diverged pretty early from the Africans. But how the Africans relate to each other and Europeans is not totally clear. Why? Because ancient population structure. It is becoming rather obvious now that ~100,000 years ago, and earlier, there were many different modern human lineages which had already diversified. The Khoisan seem to have diverged from other human lineages closer to 200,000 thousand than 100,000 years ago. What this means is that for most of the history of anatomically modern humans population structure  existed between distinct lineages. And some of that persists down to today within Africa.

I’ll bullet point some of their inferences from these models (verbatim quotes below):

  1. Our results suggest that the ancestors of the contemporary RHG, AGR and EUR populations diverged between 85 and 140 thousand years ago (kya), from an ancestral population that underwent demographic expansion between 173 and 191 kya
  2. After the initial population splits, the Ne of AGR and RHG (NaAGR and NaRHG) remained within a range extending from 0.55 to 2.2 times the ancestral African Ne (NHUM), whereas EUR (NaEUR) experienced a decrease in Ne by a factor of three to seven.
  3. The ancestors of the wRHG and eRHG populations diverged 18 to 20 kya (TRHG), and underwent a decreased in Ne by a factor of 3.8 to 5.7 for the wRHG (NwRHG) and 7.1 to 11 for the eRHG (NeRHG), regardless of the branching model considered.
  4. The ancestors of the AGR (NaAGR) split into western and eastern populations 6.7 to 11 kya (TAGR), and underwent a mild expansion, by a factor of 2.3 to 3.1 for the wAGR (NwAGR) and 1.2 to 2.2 for the eAGR (NeAGR).
  5. The EUR population experienced a 7.1- to 8.3-fold expansion (NEUR) 12 to 22 kya (TEUR).

No results are perfect. But some of these dates do not make sense. There’s a lot of circumstantial evidence that the ancestors of European populations began to expand over the last 10,000 years. The dates above suggest there was a Pleistocene expansion. Basically you can divide that value by half, and then you get a reasonable range.

Second, both the agriculturalists sampled here are Bantu speaking, and there’s a good amount of cultural and genetic data for recent shared ancestry of the Bantu over the last 3,000 years. I understand that admixture with a very diverged lineage (e.g., eastern Bantu agriculturalist samples mixing with Nilotic populations, which is how they got some non-African ancestry, as well as local Pygmy groups) can inflate these divergence dates. If that’s the case, they should note that in the text.

We don’t have much historical or archaeological clarity from what I know about divergences between Pygmy groups. This particular group has studied the topic and published on it before, so I’m inclined to trust them more than anyone else. But, the above dates for groups we do know make me a bit more skeptical of a simple divergence around the Last Glacial Maximum.

Then there are the earliest divergences. And 85 to 140,000 year interval is huge for when non-Africans split off from Africans. If closer to 140 than 85, then that means that non-African divergence from Africans preserves ancient African diversity. That is, non-Africans descend from an African group that no longer exists (or has not been sampled in this study at least!). I’ve poked around this question, and when you take into account recent gene flow, it is hard to find the specific African group that non-Africans descend from, though there is some consensus that they branched off from the non-Khoisan Africans later than from the Khoisan.

But there is also a lot of archaeological and some ancient genetic DNA now that indicates that the vast majority of non-African ancestry began to expand rapidly around 50-60,000 years ago. This is tens of thousands of years after the lowest value given above. Therefore, again we have to make recourse to a long period of separation before the expansion. This is not implausible on the face of it, but we could do something else: just assume there’s an artifact with their methods and the inferred date of divergence is too old. That would solve many of the issues.

I really don’t know if the above quibbles have any ramification for the site frequency spectrum of deleterious mutations. My own hunch is that no, it doesn’t impact the qualitative results at all.

Figure 3 clearly shows that Europeans are enriched for weak and moderately deleterious mutations (the last category produces weird results, and I wish they’d talked about this more, but they observe that strong deleterious mutations have issues getting detected). Ne is just the effective population size and s is the selection coefficient (bigger number, stronger selection).

Why are the middle two values enriched? Presumably it’s the non-African bottleneck. This is where another non-African population would have been a nice check to make sure that it was the “Out of Africa” bottleneck…but it’s probably asking a bit much to sequence more individuals to 70x coverage.

The lack of difference between the African populations is an indication that recent demography is not shaping the distribution much. Additionally, they note that gene flow between the African groups probably increased diversity in some ways, so that as long as a group is connected with other populations it will probably be rescued (note that none of these in their data were particular inbred as judging by runs of homozygosity).

Finally, they found that the number of homozygote mutations that were deleterious is higher in their model results for Europeans than the African groups. This is not surprising, and what one expects. But, they found that this is a function likely of continuous gene flow between the African groups. Without gene flow homozygosity would have been much higher. This gets back to the fact that gene flow is a powerful homogenizing tool, and the lack of gene flow has to be pretty extreme for divergence to occur.

Which brings us back to the “Out of Africa” event. The next ten years are going to see a lot of investigation of African phyologenomics and population genomics. Basically, the relationships, and selection pressures. It is totally implausible that Bantu groups in Kenya and Tanzania did not absorb local non-Nilotic populations. We’ll figure that out. Additionally, selection pressures are probably different between different groups. We’ll know more about that. But, ancient DNA will probably give us some understanding of why non-Africans went through such a massive demographic sieve. We know in broad sketches. But most people want to fill in the details.

Citation: The demographic history and mutational load of African hunter-gatherers and farmers, Marie Lopez, Athanasios Kousathanas, Helene Quach, Christine Harmant, Patrick Mouguiama-Daouda, Jean-Marie Hombert, Alain Froment, George H Perry, Luis B Barreiro, Paul Verdu, Etienne Patin, Lluis Quintana-Murci, doi: https://doi.org/10.1101/131219

11 thoughts on ““Out of Africa” bottleneck is what really matters for mutations

  1. Do you think that absorbing very divergent groups could be important to the reality of the dating here?

    The Non-African population from Belgium long ago absorbed a significant amount of Neanderthal genes. It is likely that this also brought in a lot of nonsynonymous mutations that could make the bottleneck appear to have been even more severe, and the divergence time further back.

    At the same time, Africa likely contained some highly divergent populations that are now extinct, but contributed to some modern populations more than to others.

    Averaging divergence times over the entire genome will never make sense in these cases, because there were not tree-like splits.

    1. Do you think that absorbing very divergent groups could be important to the reality of the dating here?

      that’s my suspicion. that being said, 2% of the european genome is neanderthal. quick back of envelope suggests that might add 10,000 years to 50,000? still not enough.

      1. Right, but if Africans separately had admixture from a different divergent population, then it could easily push the apparent Africa/Non-Africa split back to 85-140,000 years ago.

      2. Would 0.4% from something that diverged 3 million years ago + 2% from something that diverged 900 kya (Neanderthals and others) work? That’d be about 30k no? Those are the values from the paper I shared the other day on archaic admixture in Eurasians.

  2. Much more relevant here. Hot item on human evolution in today’s (2017/04/27) newspapers:

    “Scientists Say People Arrived Earlier Than Thought in North America: Early humans first inhabited North America more than 130,000 years ago, new report says”
    By Robert Lee Hotz
    April 26, 2017
    https://www.wsj.com/articles/scientists-say-people-arrived-earlier-than-thought-in-north-america-1493226000

    Fractured mastodon bones and stone tools unearthed in southern California offer evidence that people first inhabited North America more than 130,000 years ago, substantially earlier than archaeologists have long believed, according to a new report in Nature. …

    “The controversy centers on discoveries during freeway construction in the heart of San Diego, where paleontologists from the San Diego Natural History Museum found splintered mastodon bones and the heavy stones likely used to crack them open. The site was first excavated in 1992, but until now scientists had been unable to date the age of the bones accurately. …

    “By their calculations, the timing of that first migration reaches so far back in evolutionary history that any one of several different early human species, including Neanderthals, could have been the trailblazer. …

    “With no direct evidence of humans to be found there, the scientists analyzed impact marks on bones belonging to a single mastodon and five distinctive “hammer stones” uncovered with them. …

    “Finally, they turned to a technique that measures the natural decay of uranium. They singled out three mastodon bones for testing and arrived at an age of 130,000 years old, give or take 9,000 years.”

    1. Razib, would really appreciate if you could please post yours as well as your family’s FTDNA 2.0 ancestry results as I am curious to see how they have changed from FTDNA 1.0

  3. But there is also a lot of archaeological and some ancient genetic DNA now that indicates that the vast majority of non-African ancestry began to expand rapidly around 50-60,000 years ago.

    Hasn’t that archaeology already been disproven by the early AMH admixture into Altai Neanderthals? I’m not sure one would be able to tell the difference between AMH tools and those used by archaics. There could have been an Arabia standstill of some sort maybe. I’m not saying I necessarily disagree with your conclusions here, but I don’t think the argument is bulletproof. The bigger problem to me than the timeline is that they have the split between Eurasians, Bantu farmers and Paleoafricans contemporaneous, which seems to contradict a lot of what we know, no?

    No results are perfect. But some of these dates do not make sense. There’s a lot of circumstantial evidence that the ancestors of European populations began to expand over the last 10,000 years. The dates above suggest there was a Pleistocene expansion. Basically you can divide that value by half, and then you get a reasonable range.

    I wouldn’t discard this out of hand. Fu et al showed a pretty big population turnover in Europe around that period (with increasing affinity to the ME), and I don’t think it’s implausible that something similar could have happened in the ME around the same time.

    That is, non-Africans descend from an African group that no longer exists (or has not been sampled in this study at least!).

    I’d think the simplest explanation here would be that we branched off from East Africa, and that some of that Eurasian affinity is actually ancient rather than recent, no?

    1. Hasn’t that archaeology already been disproven by the early AMH admixture into Altai Neanderthals?

      i think it likely that proto-eurasian AMH were around for a while. but i don’t think these contribute anything to later pops.

      I wouldn’t discard this out of hand. Fu et al showed a pretty big population turnover in Europe around that period (with increasing affinity to the ME), and I don’t think it’s implausible that something similar could have happened in the ME around the same time.

      yeah, but they should pick up the most recent/largest expansion. if the time scale is right, they don’t. my understanding is most of these methods pick up the most recent pop change best, and older ones less well.

      I’d think the simplest explanation here would be that we branched off from East Africa, and that some of that Eurasian affinity is actually ancient rather than recent, no?

      the drift is share with EEF more than anything else. so that suggests it’s back migration during the holocene, not shared drift with all non-africans, as would be the case in the scenario you’re talking about.

Comments are closed.