Neanderthals and humans interbred in Europe for 5,400 years

At the center of the debate on the emergence of modern humans and their spread throughout the globe is the question of whether archaic Homo lineages contributed to the modern human gene pool, and more importantly, whether such contributions impacted the evolutionary adaptation of our species. A major obstacle to answering this question is that low levels of admixture with archaic lineages are not expected to leave extensive traces in the modern human gene pool because of genetic drift. Loci that have undergone strong positive selection, however, offer a unique opportunity to identify low-level admixture with archaic lineages, provided that the introgressed archaic allele has risen to high frequency under positive selection. The gene microcephalin (MCPH1) regulates brain size during development and has experienced positive selection in the lineage leading to Homo sapiens. Within modern humans, a group of closely related haplotypes at this locus, known as haplogroup D, rose from a single copy ≈37,000 years ago and swept to exceptionally high frequency (≈70% worldwide today) because of positive selection. Here, we examine the origin of haplogroup D. By using the interhaplogroup divergence test, we show that haplogroup D likely originated from a lineage separated from modern humans ≈1.1 million years ago and introgressed into humans by ≈37,000 years ago. This finding supports the possibility of admixture between modern humans and archaic Homo populations (Neanderthals being one possibility). Furthermore, it buttresses the important notion that, through such adminture, our species has benefited evolutionarily by gaining new advantageous alleles. The interhaplogroup divergence test developed here may be broadly applicable to the detection of introgression at other loci in the human genome or in genomes of other species.

The extent to which anatomically modern humans admixed with archaic Homo has been the subject of repeated speculation, particularly in regards to Neanderthals (2–22). Thus far, the mainstream view from fossil and genetic studies leans toward a model where anatomically modern humans fully replaced archaic Homo lineages rather than admixed with them (2–8). However, a number of investigators have voiced opposition to this total replacement model on a number of grounds, and the debate has yet to be resolved (9–22). Particularly needed to settle this debate is the identification of genetic loci that show telltale signs of admixture. There have been several reports of loci in the human genome that display unusually deep genealogy (15, 16, 23, 24), and in some cases, admixture between humans and archaic Homo lineages has been invoked as a possible explanation. However, these studies cannot differentiate the admixture model from other possibilities, such as long-standing balancing selection, that also could contribute to deep genealogies (see Discussion). As such, proponents of the admixture scenario have yet to identify a concrete example of a genetic locus for which there is compelling evidence of admixture. Furthermore, most discussions of admixture tend to treat it as a selectively neutral event, one that happened simply as a byproduct of the geographical overlap between modern humans and archaic populations. Such discussions often overlook the possibility that admixture with archaic lineages, if it indeed occurred, might have brought adaptive alleles (along with the traits they determine) into the modern human gene pool, thus profoundly impacting the biological evolution of our species.



A major difficulty in looking for traces of ancient admixture in the modern human gene pool is that, under neutrality, very low levels of admixture are not expected to be readily detectable because of the effects of genetic drift. Indeed, simulations of ancient admixture showed that an archaic genetic contribution of <0.1% is unlikely to be detectable even in a very large set of polymorphism data (25). Thus, the absence at present of conclusive genetic data in support of the introgression scenario should not be taken as evidence against the possibility of any introgression. If introgression of archaic lineages into the modern human gene pool indeed occurred, then genes that have been subject to recent positive selection in humans may be enriched for introgressed alleles. Although selectively neutral alleles introgressed from archaic lineages at low levels are likely lost by drift or swamped by the large influx of modern human DNA, an introgressed allele that is selectively advantageous could escape the effect of genetic drift and rise to high frequency. As such, these alleles might become detectable in the modern human gene pool.

The gene microcephalin is a critical regulator of brain size. In humans, loss-of-function mutations in this gene cause a condition known as primary microcephaly, which is characterized by a severe reduction in brain volume (by 3- to 4-fold) but, remarkably, a retention of overall neuroarchitecture and a lack a overt defects outside of the brain (26). The exact biochemical function of microcephalin has yet to be elucidated, but this gene likely plays an essential role in promoting the proliferation of neural progenitor cells during neurogenesis (26). microcephalin has been shown to be the target of strong positive selection in the evolutionary lineage leading from ancestral primates to humans (27, 28). This observation, coupled with the fact that this gene is a critical regulator of brain size, suggests the possibility that the molecular evolution of microcephalin may have contributed to the phenotypic evolution of the human brain (27, 28). In a recent study, we found that the haplotype structure at the human microcephalin locus is consistent with the action of recent positive selection (29). Specifically, we found that a class of haplotypes at the locus, dubbed haplogroup D, has a remarkably young coalescence age (≈37,000 years) despite an exceptionally high worldwide frequency (≈70%). This observation implies a rapid rise in the frequency of haplogroup D in humans, which is incompatible with genetic drift and instead supports the notion that positive selection has operated on haplogroup D to drive up its frequency. In the present work, we examine the origin of haplogroup D. We provide evidence that haplogroup D may have originated from a lineage separated from modern humans for ≈1.1 million years and introgressed into the human gene pool by ≈37,000 years ago. We discuss the implications of our findings for the understanding of modern human origins and the biological adaptation of our species as it spreads around the globe.


Fig. 1. Distribution of pairwise sequence divergence between and within D and non-D chromosomes at the microcephalin locus.

Highly Unusual Genealogy of the microcephalin Locus. We have examined (29) microcephalin in a panel of 89 individuals assembled to approximate the worldwide diversity of major human populations. We resequenced the panel for a 29-kb region that spans exons 4–9 of the 14-exon microcephalin gene. This process led to the identification of 220 segregating sites delineating 86 distinct haplotypes (Table 1, which is published as supporting information on the PNAS web site). Of the 178 chromosomes we sampled, 124 (or 70%) belonged to haplogroup D, defined by the derived C residue at the G37995C diagnostic nonsynonymous polymorphic site. (For simplicity, we will refer to haplogroup D as the D allele and the non-D haplotypes as the non-D allele.) Despite the high worldwide frequency of the D allele, its coalescence age is merely ≈37,000 years, far younger than the non-D allele (29). Closer scrutiny of the haplotype data in Table 1 revealed a rather peculiar pattern: the average pairwise divergence between D and non-D chromosomes across the 29-kb region is 3.3 times the divergence seen within non-D chromosomes and a striking 30 times the divergence seen within D chromosomes (Fig. 1). Further examination showed that this is because microcephalin has a highly unusual, lopsided, and deeply divided genealogy, which is schematized in Fig. 2 A.


Fig. 2. Comparison of the microcephalin genealogy with an idealized genealogy. Each filled triangle represents a genealogical clade, with the width of the triangle representing frequency in the population. (A) The genealogy consistent with the haplotype data at the microcephalin locus. The coalescence age of D chromosomes (≈37,000 years), non-D chromosomes (≈990,000 years), and between D and non-D chromosomes (≈1,700,000 years) are indicated. (B) The idealized genealogy of a partial positive selective sweep, wherein the adaptive allele first emerged by a mutational event on a random chromosome in the population.

Three features are prominent in this unusual genealogy. First, the D chromosomes coalesce to its most recent common ancestor (MRCA) at ≈37,000 years before present, whereas the non-D chromosomes coalesce at a far older ≈990,000 years before present. The much younger coalescence age of the D chromosomes, despite their much higher frequency, is consistent with the action of positive selection on the D allele as reported previously (29). Second, and more surprisingly, however, we found that the D and non-D chromosomes belong to two distinct, deeply divided clades connected by a single branch around the root of the tree (except for a few rare recombinants between the two clades, as discussed later). In other words, the D clade is a distant outgroup of the non-D clade, and vice versa. As shown in Fig. 3 and Table 1, the deep division between the D and non-D clades is due to a large number of segregating sites scattered throughout the 29-kb resequenced region that consistently differentiate the two clades, i.e., segregating sites for which the D chromosomes are characterized by one allele, whereas the non-D chromosomes are characterized by the other allele. For most of these segregating sites, the D chromosomes bear the derived allele, whereas the non-D chromosomes bear the ancestral allele (Fig. 3), an observation consistent with the fact that the internal branch from the MRCA of the D clade to the root of the tree is much longer than the internal branch from the MRCA of the non-D clade to the root (Fig. 2 A). Such a tree topology does not resemble the expected genealogy of a recent selective sweep, as schematized in Fig. 2 B, wherein the adaptive allele is introduced by a mutational event in a panmictic population. Fourth, whereas the coalescence age of ≈990,000 years for the non-D clade is similar to the human genome average, the D and non-D clades coalesce with each other at a much older ≈1,700,000 years before present, with virtually no genetic exchange between the two clades except for a few rare recombinants (discussed below). These unusual features of the microcephalin genealogy suggest the possibility that the MRCA of the D clade introgressed into humans from a divergent Homo lineage at or some time before ≈37,000 years ago. In the ensuing sections, we describe stringent statistical tests that support this introgression model.


Fig. 3. Distribution of congruent or near-congruent segregating sites in the 29-kb resequenced region of microcephalin. Congruent sites are defined as showing consistently different alleles between D and non-D haplotypes; near-congruent sites are defined as having no more than four differences from congruent sites. Sites for which the D chromosomes are characterized by the derived allele are indicated by long blue lines, whereas sites for which the D chromosomes are characterized by the ancestral allele are indicated by short red lines (for exact positions of these sites, see Table 1). Also indicated is the G37995C nonsynonymous site used to define the D chromosomes (bearing the derived C allele) and the non-D chromosomes (bearing the ancestral G allele).

Discussion
In this study, we investigate the origin of the microcephalin D allele in modern humans. We show that the D allele is unlikely to have arisen within a panmictic population. Instead, our data are consistent with a model of population subdivision followed by introgression to account for the origin of the D allele. By this model, schematized in Fig. 4 B, the lineage leading to modern humans was split from another Homo lineage, and the two lineages remained in reproductive isolation for ≈1,100,000 years. During this period of reproductive isolation, the modern human lineage was fixed for the non-D allele at the microcephalin locus, whereas the other Homo lineage was fixed for the D allele. These two alleles are differentiated by a large number of sequence differences accumulated during the prolonged isolation of the two populations. At or sometime before ≈37,000 years ago, a (possibly rare) interbreeding event occurred between the two lineages, bringing a copy of the D allele into anatomically modern humans. Whereas the original D-bearing Homo population had since gone extinct, this introgressed copy of the D allele in humans had subsequently spread to exceptionally high frequency throughout much of world because of positive selection.

Several studies have reported loci in the human genome that are associated with unusually deep genealogies containing highly divergent clades (15, 16, 23, 24). Although this type of observation can result from the admixture of reproductively isolated populations, the observation is itself insufficient evidence for admixture. In particular, deep divergence can occur if two or more allele classes at a locus are maintained by balancing selection for a prolonged period. So, even for a locus whose genealogy is too deep to be statistically compatible with neutral evolution, it is essential to rule out balancing selection before the admixture model can be adopted. One notable example is the MAPT locus, which has two distinct haplogroups, H1 and H2, that diverged from each other ≈3 million years ago, whereas the coalescence age of H2 is much younger (24). One explanation for this unusual observation is that H2 introgressed into modern humans from an archaic Homo lineage (17). However, because H1 and H2 are inverted relative to each other and thus cannot recombine, deep divergence between them across an extend genomic region and the young age of H2 are also compatible with balancing selection that maintained H2 at low frequencies for extended periods (along with H1) followed by a recent rise in the frequency of H2 (24). For the microcephalin locus, in contrast, we were able to rule out the possibility of balancing selection with a high degree of confidence. As such, microcephalin shows by far the most compelling evidence of admixture among the human loci examined thus far.

Speculation about the identity of the archaic Homo population from which the microcephalin D allele introgressed into the modern human gene pool points to the Neanderthal lineage as a potential (although by no means only) candidate. Anatomically modern humans and Neanderthals shared a long period of coexistence, from as early as 130,000 years ago in the Middle East (39) to as late as 35,000 years ago in Europe (40), consistent with the estimated introgression time of the microcephalin D allele at or sometime before ≈37,000 years ago. Furthermore, the worldwide frequency distribution of the D allele, exceptionally high outside of Africa but low in sub-Saharan Africa (29), suggests, but does not necessitate, admixture with an archaic Eurasian population. Finally, our estimate of the separation time between D and non-D alleles (i.e., ≈1,100,000 years with a lower-bound confidence interval of ≈530,000 years) is largely consistent with the divergence time between modern humans and Neanderthals based on mitochondrial DNA (mtDNA) sequence difference (320,000–740,000 years; refs. 41 and 42) and with the earliest appearance of Neanderthals in the fossil record ≈500,000 years ago (43). It would be of great interest to sequence the microcephalin locus in Neanderthals or other archaic Homo lineages, should it become technically feasible to retrieve and analyze nuclear DNA from ancient hominid remains.

Our results not only provide genetic evidence in support of the possibility of admixture between modern humans and an archaic Homo lineage but also support the notion that the biological evolution of modern humans might have benefited from the contribution of adaptive alleles from our archaic relatives. In the case of microcephalin, it is all the more intriguing given the fact that the adaptive allele is associated with an important brain development gene. As anatomically modern humans emerged from Africa and spread across the globe, the “indigenous” Homo populations they encountered had already inhabited their respective regions for long periods of time and were, in all likelihood, better adapted to the local environments than the colonizing humans, at least in some biological domains. It is perhaps not surprising then that modern humans, although likely superior in their own way, could in theory benefit from adopting some adaptive alleles from the populations they replaced. That this might indeed be the case for the brain size-determining gene microcephalin should add an important new perspective to the discussion of human origins and the recent evolution of our species. Furthermore, any admixture between modern humans and archaic populations is likely to affect more than one locus in the genome. Our study thus provides a methodological template for identifying additional loci in the human genome that might harbor alleles from archaic populations through introgression and subsequent positive selection.

Evans, Patrick D., et al. "Evidence that the adaptive allele of the brain size gene microcephalin introgressed into Homo sapiens from an archaic Homo lineage." Proceedings of the National Academy of Sciences 103.48 (2006): 18178-18183.