Return of the Neanderchimps

Back in 2005, I reviewed the first description of fossil chimpanzee teeth, from the Middle Pleistocene of the Kapthurin Formation, Kenya, dating to around 500,000 years ago. At the time, I noted that no chimpanzees have lived in the area in historic times, and that mtDNA evidence then suggested that East African chimpanzees (Pan troglodytes schweinfurthii) may have been recently derived from Central Africa. Together, those observations raised a mystery -- if today's chimps had no ancestors anywhere near Kenya 500,000 years ago, to what group did these fossil chimpanzee teeth belong? I suggested an answer: a cryptic population of chimpanzees partially or completely replaced by the dispersal of Eastern chimpanzees. In other words, Neanderchimps.

Well, now that we know for sure that Neandertals are human, too... it's a good time to revisit the Neanderchimps. What can we say today about the population structure of chimpanzees in the past, and is it still possible that these chimpanzee fossil teeth are out of kilter with the population genetics of today's chimpanzees?

A few weeks ago, we had Jody Hey visiting here on campus, and he gave a talk about his recent work on chimpanzee population genetics. Together with Rasmus Nielsen and others, Hey has been developing Bayesian methods for estimating the times of divergence, migration rates, and effective population sizes of species.

The basic idea is that present-day samples of a species like chimpanzees reflect a branching process from an ancestral population. Each branch may exchange migrants with other branches, each branch has an effective population size, and each may begin with some kind of population bottleneck. That makes for a very complicated model -- for example, with only two populations, there are six parameters, not counting bottlenecks. With each additional population, the number of parameters is compounded by additional effective size, time of splitting, and migration rate to and from all other populations. The number of parameters increases faster than a factorial of the number of populations.

Hey began this work several years ago, initially limited to the two-population case. Together with Yong-Jin Won, he showed that West African chimpanzees (P. troglodytes verus) have a substantially smaller effective size than central African chimpanzees (P. troglodytes troglodytes). These two subspecies appeared to have diverged within the last 300,000-400,000 years. And while there was little evidence for gene flow from central into west African chimpanzees, there was clear evidence for gene flow the other direction, from west into central Africa.

Sound familiar?

In a series of two-way analyses, Won and Hey showed that bonobos diverged from chimpanzees approximately 400,000-800,000 years ago, that there was no substantial evidence of gene flow into or out of bonobos after their speciation, and that the efective size of bonobos was around the same as that of west African chimpanzees, a bit under 10,000 effective individuals.

Now, in 2010, Hey has extended both the data and method to encompass more than a single divergence between two populations. In the case of Pan, Hey has included three extant subspecies of common chimpanzees (P. t. troglodytes, P. t. verus, and P. t. schweinfurthii), together with bonobos (P. paniscus). Among those, in a bifurcating model of population divergence, there are three speciation times, ten effective sizes, and lots of asymmetrical migration rates, all scaled in one way or another to mutation rate. It takes a lot of data to estimate these parameters simultaneously. The study uses 73 loci from an average of 78 individuals split among the populations, which is apparently not quite enough data to get good parameter estimates for the migration rates, as the probability surfaces for these are shallow and relatively unresolved with a few exceptions.

The parameters describing divergence times and effective sizes under the model have tighter posterior probability distributions, so that they are reasonably well estimated using these data. Here are the highlights:

  1. Bonobos split from chimpanzees around 930,000 years ago (680,000-1.54 million).

  2. The effective sizes of most populations were small (around 10,000 or less). The Pan ancestral population was moderately larger (around 17,000 effective individuals).

  3. Only central African chimpanzees were substantially larger in effective size, upward of 25,000-30,000 effective individuals during the last 460,000 years.

  4. All common chimpanzees (Pan troglodytes) descend from an ancestral population that existed 460,000 years ago (350,000-650,000).

  5. East African chimpanzees split very recently, only around 93,000 years ago (41,000-157,000) from central African chimpanzees.

All these estimates result from a fairly restrictive model. Each population is described by two parameters, their interactions by an additional two parameters per population pair. The ideas of pulses of population mixture or founder effects are simply not possible in the model. I don't see this as a weakness -- I'd much rather begin with even simpler models. But it does mean that we cannot generalize the results past the model. In particular, we shouldn't compare these times and migration rates directly with those obtained under the model that Green and colleagues (2010) applied to the Neandertal genome.

But after those words of caution, what can we make of this proposed population history for chimpanzees? Here are some possible conclusions relevant to human evolution:

  1. Eastern and central chimpanzee subspecies share a more recent history than would have been true of humans and Neandertal populations at the time the latter existed. Western chimpanzees are more distant from other chimps than the Neandertals and humans were from each other.

  2. For that matter, population differences between MSA humans within Africa may have been nearly as great as those between eastern and central African chimpanzee subspecies.

  3. Bonobos and chimpanzees split roughly a million years ago with little if any subsequent interbreeding. At least in the west (Africa, Europe and West Asia), Pleistocene human populations did not experience this kind of allopatric speciation. At the moment, I enter that as an assertion, which I'll follow up later by some discussion of the pre-Neandertal problem.

  4. The effective sizes estimated for ancient human populations are not especially low.

  5. Range expansions and partial or complete replacements were part of the population history of chimpanzees. They managed these dynamic events without handaxes, fire, projectile weapons, language, or any of the other proposed trappings of Pleistocene humans.

I want to follow up on a couple of these. First, effective size: You often hear people claiming that humans have much lower genetic diversity than chimpanzees. It is true only in a limited sense. Bonobos, west African and east African chimpanzees are populations with lower genetic variation than humans. The estimate for the effective size of the common chimpanzee ancestral population, 7100, is substantially lower than estimated for the human ancestral population during the same time period, a period stretching from roughly a million to 460,000 years ago. The common ancestral population of chimpanzees and bonobos is inferred to have had an effective size close to that of ancestral humans at the same time, around 17,000 effective individuals prior to a million years ago.

One may object that chimpanzees cover a much smaller area than Pleistocene humans, so we should expect their effective size to be much lower. But genetic variation can be related to population size only by assuming a population model, and Hey's analysis gives us a model quite starkly different from the usual. That doesn't mean it's correct, or that it is a better estimator of the census size of the ancient populations. But it reminds us that comparing the genetic variation of humans and chimpanzees is too simplistic; that the gene trees within each populations are very sensitive to the relative contributions of different parts of each species' range during the last 500,000 years. In chimpanzees, the high genetic variation mostly can be attributed to the central African subspecies; in humans, the extant genetic variation can mostly be attributed to Africa.

Let's ponder chimpanzee range expansions for a moment longer. We know that in the early Middle Pleistocene, chimpanzee-like apes lived in western Kenya. The only chimpanzees who live anywhere near that area today seem to have been much more strongly connected to chimpanzees in western Congo prior to 93,000 years ago, and that central African population still has much more variation than the eastern ones. That suggests a recent range expansion, Late Pleistocene in age, into East Africa.

We don't know that the earlier chimpanzees became extinct. They may have contributed genes into later P. schweinfurthii, just as Neandertals did into living humans. We can tell stories about climate change and the former East African chimpanzees, just as people have done about human origins, megadroughts and volcanoes. But one thing is clear about the chimpanzees: there was no modern chimpanzee revolution. The other chimpanzee subspecies, P. t. verus, is still here.

UPDATE (2010-05-20): "More on chimpanzee population structure" discusses a subsequent paper on the same topic.


Gagneux P, Gonder MK, Goldberg TL, Morin PA. 2001. Gene flow in wild chimpanzee populations: what genetic data tell us about chimpanzee movement over time and space. Phil Trans R Soc Lond B 356:889-897.

Goldberg TL, Ruvolo M. 1997. Molecular phylogenetics and historical biogeography of east African chimpanzees. Biol J Linn Soc 61:301-324.

Hey J. 2010. The divergence of chimpanzee species and subspecies as revealed in multipopulation isolation-with-migration analyses. Mol Biol Evol 27:921-933. doi:10.1093/molbev/msp298

McBrearty S, Jablonski NG. 2005. First fossil chimpanzee. Nature 437:105-108. doi:10.1038/nature04008

Won Y-J. Hey J. 2005. Divergence population genetics of chimpanzees. Mol Biol Evol 22:297-307. doi:10.1093/molbev/msi017