Is multi-regional evolution dead?

A widely-circulated interpretation of the recent work on the genome of Neandertals and the Denisova hominins is that it proves the assimilation model of human origins. First, a reminder of the main models of human evolution from Stoneking et al.:

It is said that modern humans are mostly Out-of-Africa but there has been gene flow into regional Homo sapiens populations: Neandertal-related admixture affecting Eurasians and Denisovan-related admixture affecting Melanesians.

This naive interpretation of the evidence proposes that e.g., Melanesians are about 92% derived from Out-of-Africans and 8% derived from other hominins including both Neandertals and Denisovans.

It's important to state categorically that this only an interpretation of the evidence and in no way a disproof of the multiregional theory of human evolution.

Interbreeding happened

Let's begin by considering what we know to be true: interbreeding happened between widely divergent human populations. Otherwise the pattern of differential affiliation of modern human groups to archaic hominins would be impossible to explain: all humans should have the exact same relationship to the "side-branches" of the human family tree. But, this is clearly not the case.

If interbreeding happened then...

It's important to note how important this insight is. Modern human populations are mutually interfertile, but it is generally assumed that beyond a certain point of genetic divergence interbreeding is impossible for both anatomical and genetic incompatibility reasons.

But, if we accept the possibility of interbreeding even between very divergent populations, then where do we stop? It's possible that H. sapiens himself is the product of such interbreeding.

Admixture with archaics vs. Incomplete fusion

There are two ways of viewing the evidence:

1. Different human populations have admixed with different archaic populations during the Out-of-Africa exodus
2. Humans are descended from multiple archaic populations and have blended together, but the fusion is incomplete

Here is the Reich et al. model:

The alternative model is that Eurasian and African Homo can be envisioned as a mosaic of populations that did not speciate because gene flow between most of them was never interrupted.

Adaptive changes could flow freely in the lattice of populations: they could originate anywhere and spread across the entire species range in a few thousands of years.

• Part of the regional variation spread and became part of the species-wide variation
• Part of the regional variation was lost either because of drift or because it was maladaptive
• Part of the regional variation remained in regional populations

It is this latter part that causes regional populations to show differential affiliation with archaic hominins.

A color analogy

Imagine a canvas painted with an orange color. It is not uniformly orange, however: parts of it are more "red" than average, and parts of it are more "yellow" than average.

One interpretation for this canvas is that the painter used an orange hue to paint it, but small amounts of red and yellow pigment were added in parts of the painting (Assimilation model)

A different interpretation is that the painter used red and yellow hues to paint it, but even though he moved his brush left-right and top-bottom, shuffling pigments around, there was an excess of red pigment at the spot where his brush first hit the canvas. (Multi-regional model)

What about the reduced genetic diversity of our species?

An argument for the recent Out of Africa model is that the reduced genetic diversity of our species necessitates a small effective population size and a recent genealogical time depth.

With respect to the latter, it should be noted that the genetic divergence between modern humans and Neandertals/Denisovans is about 1/3 more than between the most divergent humans (Papuans and San in Reich et al.). Hence, the shallow genealogical time depth is true, but it is not that different from our next-of-kin.

With respect to the reduced genetic diversity, one idea is that it is the result of genetic drift following a bottleneck in a small African population. But, the data can just as well be explained by species-wide selection which culled genetic variation.

The multi-regional hypothesis proposes that the reduced genetic diversity of our species is due to a pattern of selection. The evidence is not inconsistent with this hypothesis, as the human species shows a clear pattern of morphological change in the last 200 thousand years.

Indeed, the fact that at 100 thousand years ago the "chin" makes its appearance in the Levant and China could be interpreted as an Out-of-Africa migration, but it could just as well be interpreted as natural selection producing a chin in different Homo populations, without necessarily any large-scale population movements.

Aren't Africans more genetically diverse?

Another argument for the Out-of-Africa theory is that since Africans are more genetically diverse, that is where humans originated. But, different genes tell different stories; there are loci where Africans sport the greatest variation, and there are those with genealogies better explained by a Eurasian origin. Indeed, recently a paper suggested that some Indian groups may be more genetically diverse than African agriculturalists in a particular genomic region.

But, again, greater African genetic diversity does not necessitate an African origin of mankind, and could be partially explained by a pattern of admixture between divergent populations. Back-migration from Eurasia is rarely considered as a possibility, but there are good reasons to suspect it. For example, Yorubans are genetically closer to Chinese than they are to San, and back-migration to Africa is a possible explanation for this.

The Y-chromosome phylogeny presents strong evidence for back-migration. The basal clades of the phylogeny (A and B) are African, suggesting an African origin of extant human Y-chromosomes, but within the M168 major human lineage, haplogroup DE-YAP has mixed African and Eurasian affiliations, while F-M89 is Eurasian. A Eurasian origin of M168 is a strong possibility.

Another possibility is the larger effective population size in Africa, related perhaps to the fact that until recently our species was well-adapted to an African environment and lacked the know-how to thrive in latitudes further from the equator. In a larger population there are more new mutations and fewer alleles are lost due to drift, hence such a population may appear to be more genetically diverse irrespective of whether or not it is the parental population.

Did archaics contribute at most 8%?

Melanesians emerge as the most "archaic"-admixed population under the assimilation model, having Denisovan/Neandertal-related admixture of ~8%. But, there is reason to doubt this.

First of all, consider that between Green et al. and Reich et al. we went from 1-4% archaic contribution in Eurasia to 1-8%. It only took one additional hominin to find extra archaic admixture. What will happen when we sample another one?

But, the most important thing is to understand how these admixture estimates are arrived at: they are derived from sequence shared by archaics and non-Africans but not with Africans:

• Sequence in which Africans are polymorphic and non-Africans are monomorphic may be due to archaic admixture in Africa
• Sequence where everyone is polymorphic may derive from anywhere, not necessarily from Africa; the same is true for sequence where everyone is polymorphic and derived with respect to chimpanzee
• Sequence where Eurasians are polymorphic and Aficans are monomorphic and in which Neandertals are monomorphic may derive from non-Neandertal non-African archaic hominins

In short, we should take the admixture ceiling of 8% with a huge grain of salt.

Conclusion

At the beginning of 2010 I would describe myself as fairly agnostic on the issue of human origins, but leaning towards the Recent Common Origin hypothesis due to:

1. The Y-chromosome and mtDNA phylogeny
2. The 100ky gap between Omo 1 and Qafzeh and the 50ky gap between Qafzeh and the later anatomically modern humans from Eurasia
3. The greater African genetic diversity, and its clinal diminution from east Africa
4. The lack of evidence for interbreeding between humans and archaic hominins

2010 has destroyed factors 2-4:

• Zhirendong Cave gave us a chin in China, 100ky, and hence Qafzeh could no longer be interpreted as the early-Out-of-Africa that failed. The gap between anatomical modernity in Africa and in Eurasia narrowed.
• Xing et al. narrowed the presumed gap in genetic diversity between Africans and Eurasians, and showed that the diversity decrease cline is spurious.
• Reich et al., Green et al., and Krause et al. showed us that interbreeding between Homo sapiens and divergent Homo populations must've taken place.

As the beginning of 2011 approaches I am much less convinced of Out-of-Africa with Assimilation. Hopefully, in the coming years both the 1000 genomes project and ancient DNA work on early Homo sapiens or even other archaic hominins will add more data to the discussion.

UPDATE:

Razib links to my post with some comments of his own. A few observations:

• The importance of the Xing et al. paper is not so much in "disproving" the greater African genetic diversity. That conclusion cannot be reached from a single genetic region and without including African hunter-gatherers. Its importance is, however, in showing that the presumed smooth cline of decrease of genetic diversity from east Africa and across Eurasia is spurious; hence it weakens a crucial Out-of-Africa argument
• Razib points out that recent selection tends to be regional (e.g., different genetic causes of light skin color in Europeans vs. Asians). That is true, but note that recent selection signatures in the last 10 thousand years should not be a guide to what happens in our species over 200-300 thousand years. Our ability to detect directional selection is time-limited: at its beginning it's lost in the noise, at its end, the selected allele is quasi-fixed. In a small and mobile population of Pleistocene hunter-gatherers, even one geographically as dispersed as Pleistocene Homo the entire process may take a few tens of thousands of years, depending on strength of selection. Hence, we are only detecting mostly recent signatures of selection that arose in dense and sedentary agricultural populations partly because older episodes of selection have already run their natural course.
• The deep rooting of the Y-chromosome phylogeny in Africa is not in doubt, but the rooting of the major M168 clade, that accounts for surely over 90% of our species, and perhaps much more is in doubt. The point of the back-migration argument is not that Y-chromosomes did not originate in Africa (they did), but that African genetic diversity could be inflated by M168 back-migrants. I have yet to see a clear separation of African genetic diversity into what is indigenous and what can be accounted for by admixture. Hence, while the Xin et al. paper shows that the Eurasian cline of reduced diversity is not resilient to the addition of new populations, it is not at all clear to me that the greater genetic diversity of east Africans is not (at least in part) due to the trilateral admixed status of east Africans (natives, people from West Eurasia, and from deeper Africa).
• Razib argues that we are unlikely to increase the archaic admixture percentage by much because there are only so many archaic hominins around. First of all, the multiregional model could be true even if we detected 0% archaic admixture, as long as fusion between different (not necessarily all) archaic populations was complete. But, also, we should remember that the genomic coverage of Pleistocene Homo is so geographically limited! Both Homo erectus soloensis and Homo erectus pekinensis are absent, as are the more recent hominins that might be directly relevant (e.g., Mungo Man, Dali, Zhiren). The same is true for Africa where recovery of ancient DNA seems extremely unlikely given present technology.
• Razib points out that Fst distances between humans would be much higher under multiregional evolution, and that humans share a lot of common recent ancestry. First of all, genetic divergence dates for modern humans are about 400-600ky according to Reich et al. but we should remember that different genes tell different stories: there are both very old and very shallow coalescence times in the human genome, and a model of multiple archaic populations in Africa is a good fit for the data, as is an Out-of-Africa bottleneck 150ky, much earlier than previously thought. Add selection to the equation, and even greater time depths become plausible.

In short, I am not advocating either OoA with Assimilation or MRE at this point, but I would strongly argue that while Assimilation has become more plausible than Replacement after 2010, it's not the case that Assimilation has become more plausible than MRE.

The Human Evolution Wars have just begun...