bottlenecks

Bornean and Sumatran orangutans are the most highly divergent subspecies within any of the living species of great apes. The two farther apart even than chimpanzees and bonobos, which are good biological species. The time of the Bornean-Sumatran orangutan divergence as estimated from mtDNA is around 3.5 million years ago.

This is old enough that many primatologists consider the two populations as separate biological species. The species distinction is supported by some aspects of morphology, but as yet we have no good nuclear DNA information about the extent of divergence. In chimpanzees, nuclear genetic comparisons suggest a relatively recent founding of one subspecies and recurrent gene flow between the others, despite high mtDNA divergence between the subspecies. So information from across the genomes of Bornean and Sumatran orangutans may be necessary to substantiate the hypothesis of long isolation suggested by mtDNA.

Within Borneo, different local populations of orangutans have strong genetic differentiation, with few shared mtDNA haplotypes among them. A new study by Natasha Arora and colleagues [1] has provided further detail about these relationships within Borneo. Based on earlier work, they expected to find high population differentiation within Borneo, and that is what they found:

[O]ur analyses revealed high and significant mitochondrial differentiation, with populations within currently recognized subspecies generally displaying as much differentiation as those between subspecies. Of notable interest is the great extent of subdivision and lack of reciprocal monophyly for the morphologically recognized subspecies P. p. morio and P. p. wurmbii. MtDNA haplotype sharing is uncommon and for populations separated by rivers occurs only in two instances: (i) for SA and GP and (ii) for the northern and southern populations across the Kinabatangan river. In both cases, very recent common ancestry could explain the incomplete mtDNA lineage sorting. For North Kinabatangan (NK) and SK, Jalil et al. (27) proposed an expansion from a recent common refugium further west in Mount Kinabalu, as posited for other Bornean species (46, 47, 49). DV, with its low haplotype diversity, might also be the result of a recent range expansion. GP is located proximally to the Bangka–Belitung–Karimata–Schwaner divide, from where orangutans are presumed to have dispersed to the rest of Borneo (12) and where we might expect a rich haplotype diversity. However, the presence of only one mtDNA haplotype shared with populations further east suggests that the current population in GP is recent and/or underwent a severe recent bottleneck. This and other local bottlenecks make it impossible to reconstruct a colonization of Borneo through the southwestern “choke point” (52).

They were able to confirm the relatively strong differentiation of Bornean populations by examining nuclear microsatellites. These do not give a great indication of the time period over which the populations may have developed their differentiation, but the microsatellites do document the relative lack of allele sharing between the populations, attesting a history of low gene flow in the recent past. The populations they identify as strongly differentiated do not correspond entirely with the subspecies recognized along morphological lines, but there are strongly differentiated populations here.

The "news" aspect of the paper is the one unexpected observation: the mtDNA ancestor of Bornean orangutans lived relatively recently, only around 176,000 years ago (with a range of error stretching from 72,000 to 320,000 years ago. The data in the study do not allow us to distinguish whether this was a time when the Bornean population may have been founded, or whether instead the mtDNA lineage spread through pre-existing populations. The authors pursue the hypothesis that Bornean orangutans were limited to a refugium sometime during the early Late Pleistocene:

Assuming that orangutans arrived in Borneo around the same time as gibbons and macaques, the recent coalescence of Bornean orangutans could be explained by a bottleneck through a severe rainforest contraction. Such a bottleneck would have had a more dramatic impact on the mtDNA structure of orangutans compared with other species as a result of their low densities and slow life histories (18) as well as habitat requirements.

The comparison with gibbons and macaques is necessary because both have substantially deeper mtDNA coalescence times within their Bornean populations. If the forest had been substantially reduced to a small area where orangutans could survive, we might expect the other primates to reflect this event -- and they don't. Nevertheless, a grab-bag of climate change scenarios appear next:

Geomorphological and palynological data indicate the presence of dryer, more open vegetation in southern and western Borneo during the last glaciation (2, 41), and by extrapolation also during other glaciations (but c.f. refs. 42, 43). Climate change was especially severe during an extended cold period within the penultimate glaciation between 130 and 190 ka (44, 45), which occurred approximately at the time of mean coalescence of Bornean mtDNA haplotypes. More recently, the last Toba eruption approximately 74 ka resulted in a short, albeit signi␣cant, decrease in regional temperatures, ensued by a 1,800-y cold stadial (9, 10). Our data do not provide clear signals to make conclusive statements about potential Toba effects. Nonetheless, the coldest period of the penultimate glaciation (44, 45) was more prolonged than the cold period following the last Toba eruption, suggesting more severe effects of the former on the extent of rainforest across Sundaland. In any event, suitable rainforest habitat for orangutans should have existed in certain regions in Borneo where a refugium population survived the dry glacial conditions.

A coalescence time of 176,000 years ago does not point to a short-duration bottleneck that began 74,000 years ago. If orangutans in the Middle Pleistocene of Borneo had high genetic differentiation, a crash would have to have been very severe -- eliminating all but one small regional population -- to have effected the present distribution. Still, the great uncertainty in the actual coalescence time leaves open many possibilities, and the refugium hypothesis in the general case is worth testing, even if the Toba eruption in particular cannot explain the data.

Given the uncertainty about the habitat structure of the now-submerged areas of Sunda, we may also want to consider the hypothesis that the present orangutans arrived recently on Borneo from mainland Southeast Asia. Even if orangutans had lived on Borneo during the Middle Pleistocene, they may not have been the current orangutans. Or even better, they may have been Neanderorangs -- an initial population that was genetically swamped by migrants arriving from elsewhere. The deep Sumatra-Borneo divergence means that the Bornean population was probably not recently derived from Sumatra, but that's a very restricted source compared to the Late Pleistocene distribution of orangutans across mainland and island East and Southeast Asia.

Some other animals walked from Sumatra to Borneo repeatedly during the Pleistocene, including humans. In the human case, we know that a large fraction of the genetic ancestry of Bornean and Javan people was derived from Asia within the last 100,000 years -- in other words, Late Pleistocene gene flow. The movement of genes may have happened in the context of a dispersal of Asian (or ultimately, African-derived) populations into island Southeast Asia. The paper includes some discussion of other primate species:

For instance, the south Bornean gibbon Hylobates albibarbis and the Sumatran–Malaysian gibbon Hylobates agilis have a TMRCA of 1.56 Ma (36), and Bornean and Sumatran pig-tailed macaques have one of 3 to 4 Ma (37). By contrast, the Bornean–Sumatran common ancestor of both the silvered langur(39) and clouded leopard (40) is much more recent than that of orangutans, gibbons, and pig-tailed macaques, probably because of a higher ␣exibility in habitat use.

The pig-tailed macaque divergence time is more or less the same as the orangutan divergence; the others are more like the time range for human dispersals into island Southeast Asia. We can add to the primates a few other medium-sized mammals; for example, clouded leopards are highly differentiated between Sumatran and Bornean populations, and their mtDNA divergence occurred sometime after 3 million years ago.

There may be no contradiction between the recent mtDNA common ancestor and the high degree of population structure in Bornean orangutans; the mtDNA could have been selected. We really would want resequencing of a lot more loci in these orangtuan populations, for which we may not have to wait too long. Mitochondrial DNA is convenient in many ways, including its greater sensitivity to restricted population size and higher mutation rate. But the intrinsic variance of a single gene system under genetic drift is so high that this disadvantage probably outweighs all advantages for reconstructing population sizes.

At any rate, the orangutans now provide an additional case where the subspecies-level history of hominoids is more complex than depicted five or six years ago. Uncovering these kinds of dynamics highlights the need for better modeling of demography and dispersal within a geographically widespread species. Isolation-by-distance and long-lasting subspecies are well-defined models, but when they are refuted, we have a lack of well-defined alternatives.

References

Bray and colleagues [1] report on genotyping of 471 people of Ashkenazi Jewish descent. This is one of the largest samples of a single human population, and is therefore very interesting for studies of population history and recent natural selection.

There's a lot in the paper. One of the key findings in the paper is that the Ashkenazi population doesn't look bottlenecked -- in fact, it looks outbred compared to Europeans generally. The paper also documents a high amount of admixture with non-Ashkenazi Europeans, ranging from 35% to 55%. Figuring out the actual history of the population -- when and where its ancestors lived and how they interacted with other people -- is beyond the scope of this kind of analysis. But I expect that somebody can put together a really compelling historical account using these data.

I turned quickly to the issue of selection. They are able to substantiate evidence of positive selection on several disease-causing alleles in the Ashkenazi population, including the Tay-Sachs allele. The lack of evidence for bottlenecks or founder effects pretty much takes away the alternative explanation. Yet they were unable to show statistical evidence of selection on some other disease-causing alleles in Ashkenazi populations:

To explore whether regions of selection in the AJ population included any loci of known Ashkenazi diseases, we examined 21 disease- and cancer-susceptibility loci with known mutations found at higher frequency in the Ashkenazi population. Only 6 of the 21 genes fell in or near (within 500 kb) the top 5% of the AJ iHS windows (Table 2). Among these is the Tay-Sachs disease gene, HEXA, whose selection has been widely debated (4, 5, 14–16) and was found ~400 kb downstream of a window on chromosome 15 identified in the top 1% of the AJ iHS hits. Although none of the SNPs interrogated immediately adjacent to the HEXA locus showed elevated iHS signals, it is possible that the nearby region may contain regulatory elements under selection that affect HEXA expression. Cochran et al. (14) speculated that selection of many of the AJ- prevalent disease loci, especially the lysosomal diseases, conferred an increase in intelligence that was necessary historically for the AJ economic survival. Our data shows evidence of strong selection at or near only six disease loci, including only one out of the four AJ- prevalent lysosomal storage diseases, thus arguing that most AJ disease loci are not under strong positive selection, but rather rose to their current frequency through genetic drift after a bottleneck. However, we cannot exclude the possibility that selection of some AJ disease loci are outside the limits of detection by the extended haplotype tests, which are known to have less power to detect se- lection of lower frequency alleles (38, 41).

It seems to me that this passage probably wasn't written by the same author who showed the lack of evidence for founder effects a few pages before. In this case, the confusion probably comes from the fact that the "detection of positive selection" is actually a refutation of the hypothesis of genetic drift. With a larger sample it will be possible to test the hypothesis with greater power.

Ddisease-causing alleles are at low frequencies currently, making them unlikely to rise to the top percentages of the statistics. It would be interesting to control for current frequency, but I haven't seen a test that uses frequency information in this way.

It's quite remarkable to reflect on the idea that positive selection has now been demonstrated on six disease-causing alleles in the Ashkenazi population. Every one of these is a case of overdominance -- where the heterozygote carrying an allele has some selective advantage, while the homozygote carrying two copies has a disorder. I was having a conversation with a very prominent geneticist a few months ago, who claimed that no case of overdominance in humans had ever been demonstrated except sickle cell. Now, that was obviously false even at the time -- as I pointed out, the many hemoglobinopathies are fairly clear examples. But we've come an awfully long way.

From data like these, we're going to learn a huge amount about low-frequency selected alleles. The Tay-Sachs-causing allele is one of the most common recessive lethal genes in any human population, but like all genes subject to strong selection in homozygotes, it remains rare. Finding selection on these kinds of alleles is very hard unless sample sizes increase to several hundred individuals. Here we are seeing evidence of selection in historic populations -- within the last 2000 years. More will be coming.

References

Amos and Hoffman (2009) describe a study of microsatellite (STR) data taken from 53 populations -- the HGDP dataset. They suggest that the worldwide diversity of STR loci is consistent with a double-bottleneck population history: An initial bottleneck accompanying a dispersal of humans from Africa some 50,000 years ago, followed by a second bottleneck as people moved into Beringia much later. Despite the cline of diversity across Eurasia leading to lower diversity farther from Africa, they do not see any evidence for successive (sequential) bottlenecks across this region.

In general, I think that people are going farther than the data on this question of human migrations. Even considering only 53 population samples, there are thousands of ways that they might have been connected to each other over time. Rarely does anybody test simple null hypotheses, like isolation by distance. Often they mix two or more distinct scenarios in an attempt to find a closer fit to data. The problem is that two or more distinct scenarios will inevitably provide a closer fit, merely by virtue of adding parameters. The question is whether the fit is significantly better.

I find some things to like in this paper. They treat STR data in a better way than many other studies, and I think they've done the right thing in examining the question of heterozygosity versus allele number. That statistic is worth more consideration.

From their introduction:

The question of how many bottlenecks account for the distribution of modern human diversity has been relatively little studied (Rogers & Harpending 1992) and yields conflicting results. First, simulations indicate that the observed pattern is consistent with a linear stepping-stone model featuring a long series of founder events (Ramachandran et al. 2005; Liu et al. 2006). However, this does not preclude equally good fits based on other models. Equally, at the other extreme, large steps in single nucleotide polymorphism diversity between adjacent populations have been used to argue for two dominant bottlenecks, one ‘out of Africa’ and one around the Bering land bridge where humans crossed into the Americas (Hellenthal et al. 2008). The latter event is supported by both mitochondrial data (Wallace et al. 1985; Fagundes et al. 2008) and data from a few nuclear markers (Hey 2005). However, mitochondrial sequences only inform on female lineages, while the adjacent population approach is least reliable in regions like the Bering Strait where population samples are extremely sparse.

They make a good point here: that "other models" may produce "equally good fits." That is a routine problem in "modern human origins" research -- alternative models are rarely evaluated, and people almost never take a null hypothesis testing approach.

I've always been hesitant to give much credence to demographic studies based on microsatellites. The evidence for mutation-drift disequilibrium in a stepwise mutation model is an unusual pattern of variance among the individual length variances of STR loci. That's a complicated statistic, and it responds poorly to deviations from the pure stepwise mutation model. In particular, any constraints on allele length will eliminate outliers in allele length variance, making the population look like it went through a bottleneck of some kind.

The current study looks for disequilibrium in the relation of heterozygosity to allele number -- the logic being that a population crash will eliminate rare alleles but not common ones, leaving heterozygosity nearly the same but cutting allele number substantially. They also are explicit about the problems of the stepwise mutation model:

One problem with the Bottleneck test is that microsatellites do not follow a strict SMM. Known deviations include mutation biases favouring expansion or contraction (Xu et al. 2000), interruption mutations within the repeat tract that slow the rate of slippage (Jin et al. 1996; Kruglyak et al. 1998), occasional larger ‘jump’ mutations of several repeat units (Di Rienzo et al. 1994; Schlötterer et al. 1998) and some form of upper length boundary that prevents indefinite expansion (Amos & Clarke 2008).

They go on to argue that their test is less susceptible to deviations in the mutation model. That claim deserves further evaluation. The information they provide about the pattern of variation of different classes of STR loci is useful, as it points to ways that the mutation model may have influenced the appearance of a bottleneck. But since they find "strong and consistent evidence of a bottleneck at the lowest variability loci." Given a correlation between diversity and strength of evidence of a bottleneck, and remembering that the signature of a bottleneck in their test is high diversity per allele, I would want to look for some mechanical explanation for the correlation.

A (possibly additional) problem: The number of rare alleles within any single subpopulation is rapidly increased by migration from other subpopulations. All you need is a single migrant to bring in a new allele from somewhere else, and you've got another allele. Now, obviously this allele may not be sampled in any given dataset, so you have to account for sampling. But the point remains: it doesn't take much migration to increase allele numbers.

Migration after any bottlenecks should, then, hide the evidence for them. It's not hard to imagine scenarios equally consistent with the data. For example, if every different population underwent a single bottleneck simultaneously, they would be unlikely to lose the same rare alleles, but would retain nearly the same heterozygosity. As migration resumed between them after the bottleneck, the rare alleles would be replenished in a way the reflects the subsequent migration rate and population size (the same number of migrants has a faster effect on allele frequencies in a smaller population).

Or, if there were any interaction between a single bottlenecked ("founder") population and other pre-existing populations, it would tend to reduce the sign of a bottleneck. It seems plausible that the data in this paper might be explained by such interactions in South or East Asia, providing them with a store of rare alleles that didn't make it to the Near East.

This is why it's useful to start simple. The simplest model in this case would include migration and expansion (which we know from non-genetic evidence happened) and no bottlenecks. Plumb these parameters to see if any acceptable fits turn up. They do with the question of the heterozygosity cline alone -- a simple trend of directionally biased migration is sufficient for that. Looking at the allele number together with the heterozygosity may reject that model, in which case you'd want to pick out the next simplest. In that sense, mtDNA may actually be giving more information than the hundreds of STR loci, because with mtDNA clades there is a way to estimate haplotype ages -- meaning that the demographic hypothesis must give rise to those haplotype ages in addition to the geographic dispersion of haplotypes and within-subpopulation diversity.

References:

Amos W, Hoffman JI. 2009. Evidence that two main bottleneck events shaped modern human genetic diversity. Proc R Soc Lond B (online) doi:10.1098/rspb.2009.1473

My series on mutual information and tests of selection (which began with "Information theory: a short introduction") is at a branching point. One of the critical factors determining the power of such tests is the ancient rate of genetic drift. So it's important to come to some understanding of the archaeological record and our best estimates of ancient demography, so that we can independently test the hypothesis that genetic drift was very strong in recent human evolution. That's a long project, potentially the topic of several review papers. Since nobody else has put together these data in useful way for population genetics, I'm going to do it in one place. What you see in this series are my notes about this project. Being notes, they are not complete, but they may occasionally be better than any other sources. Where it's appropriate, I'll spin off the results for review and publication, and point to them here.

Many geneticists believe that there were massive population bottlenecks within the last 30,000 years, citing both genetic and archaeological evidence in support of this proposition. Some claim that there have been significant population bottlenecks in the last 5000 years.

Some archaeologists agree. However, I think this is one of those Inigo Montoya cases: "That word, I do not think it means what you think it means." Archaeology and genetics have completely different interpretations of the words, "bottleneck," "contraction," and "expansion." The result has been a lot of confusion about the relation of archaeological and genetic estimates of population size.

A population bottleneck impacts genetics by increasing the rate of inbreeding. This takes time to change gene frequencies, and does so in inverse proportion to population size. It may seem surprising that a truly massive die-off, on the scale of the Black Death, will have no measurable genetic impact. But cutting a population of millions down by half just doesn't impact gene frequencies. That is, unless you are looking at genes that helped people to survive the plague, in which case you're looking at natural selection, not a bottleneck.

A significant genetic bottleneck is not just any population contraction -- it's an event in which the population is cut by a large fraction for a long time. In paleontological terms, we're usually considering cases where the ratio of the number of individuals and the number of generations is near one. In other words, if you cut the population down to a thousand individuals, and keep it there for a thousand generations, you're going to have a large genetic impact. Likewise, you can have a significant bottleneck that's ten generations long, but you need to cut the population down to around ten people.

You can do a bit better measuring inbreeding by looking at lots and lots of people to study very rare alleles, like a rare genetic disease in a founder population. There, you may spot changes that unfolded in ten generations, even in a relatively large population of a hundred people. Increasingly, as we develop larger and larger datasets of gene variations, we will add power to detect such events in human prehistory.

In archaeology, a significant event is one in which fewer sites were occupied by ancient people in a well-studied region. The length of such a contraction depends on the sampling intensity and dating methods available -- it might be a hundred years or many thousands. Likewise, the magnitude of population contraction will be uncertain -- you can get an accurate estimate, but with substantial sampling error. As in genetics, there are other possible explanations for an apparent contraction. We might lack geological exposures of the right age, or people may simply have moved from formerly favored locations to new ones. Worse, it might just be that archaeologists haven't looked hard enough at a given time interval.

Archaeology is necessarily imprecise about the census population that existed at any given time. So is genetics. Both have their strengths and weaknesses. We want these different areas of evidence to bear on the same prehistoric events.

Too much, instead of testing hypotheses, people just line up chronologies and look for matches. A geologist may claim that African paleoclimate is important because it may explain ``modern human origins.'' An archaeologist may claim that a hiatus at a site is consistent with ``genetic bottlenecks.'' And the geneticist may claim that inbreeding in a modern-day genetic sample dates to a period of time corresponding to the replacement of one tool industry by another.

Any might be a valid hypothesis, but we need to take it further, to actually provide some tests. I believe we can do better now, because of the growing amount of genetic information. But we're going to have to do away with the facile idea that we're looking for massive bottlenecks, we need to introduce a recognition of the role of selection in human genetic variation, and we need to start addressing the archaeological record as it really exists.

That's a forward to what follows. I'm going through regions of the world at different time intervals, to discuss what we know about population size from the archaeological record.

Next: No Late Pleistocene bottleneck in southern Africa