Neandertal DNA

Blond hair is relatively common in island Melanesia, even though the skin pigmentation of Melanesian peoples is relatively dark. Eimear Kenny and colleagues report in this week's Science that one SNP variant in the gene TYRP1 explains a high proportion of the variance in hair color in this population [1].

Resequencing of TYRP1 exons detected a single previously unknown polymorphism, a C-to-T transition at chr9:12,694,273 (GrCH37/hg19), that corresponds to a predicted arginine-to-cysteine mutation (R93C) in exon 2 of TYRP1 at amino acid position 93 (TT in blond- and CT or CC in dark-haired individuals)...[more on assessing effect in a GWA panel].

We genotyped R93C in 918 Solomon Islanders for whom we had measured hair pigmentation with spectrometry. A recessive model provided the best fit for the data, and R93C genotypes accounted for 46.4% of the variance in hair color (linear regression; P = 2.19 × 10−90; Fig. 1D and table S2). The frequency of the 93C allele in the Solomon Islands is 0.26, and genotyping of R93C in an additional 941 individuals from 52 worldwide populations revealed that the 93C allele is rare or absent outside of Oceania (table S3). Furthermore, we found no evidence for recent gene flow from Europe (i.e., admixture) (figs. S5 and S6) nor a strong signature of recent positive selection for the 93C allele (figs. S9 to S11).

This paper is very short, only a few paragraphs. When I read through it, I got one impression of the results, and that impression changed greatly when I looked into the supplement.

Some underreported facts:

1. The blondness allele is present in all the samples from the Solomon Islands, at a frequency as high as 49% in a large sample from Malaita. In this study, the authors found it at its lowest frequency in "Polynesian outlier" islands near the Solomons.

2. The allele was not found in any of the HGDP samples, even when they were genotyped specifically for this study. That includes the "Melanesian" and "Papuan" samples. These two are relatively small in HGDP (n=14 and n=16 in this study) but even so would probably present this allele were it present at anything like the frequency in the Solomon Islands.

3. The text of the paper reports that a recessive effect model is the best explanation for the relation of hair pigmentation and TYRP1 genotypes. The supplement shows that the recessive model is only very slightly better than a "codominant" model, as it only explains an additional 3 percent of the variance. In the best case considering this allele along with age and geographic origin of the individuals, only 48% of the variation of hair pigmentation can be explained. That leaves 52% to be explained by other genetic and nongenetic causes. There may be a lot of genetic background, which may include more alleles of large effect.

4. Skin pigmentation varies greatly among these Solomon Islands samples, with more than a third of the overall variance in skin pigmentation explained by geography. The tables don't make it clear how pigmentation is patterned by geography. The TYRP1 allele that is the subject of this paper does not explain much variation in skin pigmentation.

5. Sex and age have strong effects on hair pigmentation in this sample, but not on skin pigmentation. Again, these point to background genetic factors. Many populations have sex and age effects on hair pigmentation, so some of the additional causal factors may be widely shared.

I began looking more deeply into TYRP1 R93C for a couple of reasons. The prehistory of human populations in the Solomon Islands goes back more than 30,000 years. Because this allele is not present in mainland Asian populations, as far as we know, but it is present thoughout the Solomons, suggests that it may have become common at or near the initial founding of this population. The LD pattern around the mutation likewise suggests that it has been segregating in this population for a long time. The data are consistent with the idea that blond phenotypes were present in the Solomon Islands as early as the initial colonists who founded the population.

It will be interesting to look further into nearby populations to see if it characterized early colonists more broadly. Blond phenotypes occur very commonly in Aboriginal Australians, also age-dependent in expression, as many children have blond hair that darkens with age. Other Melanesian islands, such as Vanuatu and Fiji, also have a high incidence of blondness. For the islands, I expect that the same allele will be responsible for a similar fraction of the variance. For Australia, I would guess that this allele is also present, but with 40,000 years of evolution, there could well be a more diverse genetic explanation.

Pigmentation variation in Eurasia is clearly a phenotype that has been affected both by recent positive selection and selection on old, standing genetic variants. Europe and East Asia today each have a dozen or more alleles that individually have strong effects on skin, hair, or eye pigmentation. Many of the alleles common in one region are rare in the other. These are well explained by recent selection on pigmentation; if there had been no selection on pigmentation, the populations would not show as extensive a pattern of differences, and new alleles would not have reached high frequencies. But if we had only a single mutation at 30 percent distinguishing one of these populations, which had arisen as early as 30,000 years ago, we would not have a strong case for selection.

In Melanesia, we have just the opening sketch of pigmentation variation. We know that there is substantial variation in skin and hair pigmentation, and that one mutation unique to this part of the world explains a large fraction (but still a minority) of the variance in hair pigment. The other genes that contribute to variation in hair and skin pigmentation are not known. Possibly, skin pigmentation variation among the geographic regions in this study may reflect late prehistoric migration of people through this region, as agriculture moved into the area and Polynesia was settled. But the genetic part of this story remains to be demonstrated.

Both Asia and Europe have a similar pattern of selection which has favored new alleles along with some old, standing alleles. Across the temperate regions of Europe, East Asia, and the Americas, it is plausible that the disadvantages of dark pigment for vitamin D production manifested themselves. It is also plausible across these regions that the advantages of dark pigment as protection from UV radiation would have been relaxed, allowing sexual selection on pigmentation to play an important role.

The evidence here suggests that this allele in Melanesia has not been recently selected from a new mutation. Additionally weighing against recent selection is the observation that the mutation acts recessively on hair pigmentation -- recent selection is much more likely for mutations with dominant or additive effects.

Together, these observations suggest that variation in human pigmentation emerged in stages. Some genes, such as ASIP, have old alleles that explain some of the variation in pigmentation today and are geographically ubiquitous, in Africa, Eurasia, and the Americas. This genetic variation was older than the Late Pleistocene. Such genes (ASIP is probably an example) today have alleles associated with darker pigment that are common in sub-Saharan Africa. Probably many other genes have variation within Africa that are part of the ancestral pigment variation of humanity. As people dispersed throughout the world, mixing with archaic humans, they carried some of these pigmentation variants along with them.

What's interesting is that even though some of these ancient alleles lighten skin pigmentation, they remain segregating in today's light-pigmented populations. They were not selected to fixation, even though there was plenty of time for them to increase toward fixation, and even though strong selection on pigmentation appears to have been present in many high-latitude populations. Later mutations that lighten pigmentation were strongly selected in these same populations, some reaching very high frequencies, while the old mutations still were not selected to fixation.

The story is of course more complex than a simple count of standing and new mutations. Some genetic changes that lighten pigmentation may have countervailing negative effects. Solving the problem of becoming light pigmented in just the right way may really be a different problem in different populations. Founder effects may have shifted the genetic background of early Eurasian populations just enough to create strong path-dependence for later mutations, allowing some to proceed rapidly and blocking the rise of others.

The story of TYRP1 gives a new perspective on the early evolution of pigmentation outside Africa. Here is a novel allele that originated within the earliest colonists to Oceania, which affects hair pigmentation strongly, in a population that was always low-latitude. It did not come from earlier archaic humans as far as we know so far (not in the Denisova genome). It may have become common by a founder effect. We cannot rule out selection, such as social or sexual selection, as a cause of its initial spread or current geographic distribution, but we have no genetic evidence in favor of such selection. We know from the data that there must be many other loci that affect pigmentation in this population.

This may have been much like the original pigmentation genetics of early modern human populations. It may also be much like the pattern that accounts for pigmentation variation within Africa today. It is not a simple story in which a few loci of large effect explain the evolutionary pattern. It is a story in which a substantial store of segregating variation persists within populations for tens of thousands of years.

Why does that matter? Here's one reason: We're looking at possible pigmentation variants in archaic humans, and we have counted many of them. Anyone might begin this project with the presumption that Neandertals and Denisovans had pigmentation variants that were fixed relative to living people. In that context, it would be surprising to find that they had not introgressed.

But if all these ancient populations had a large store of small-effect variants affecting pigmentation, a mutation that we find in one individual might have been rare in the population. The TYRP1 R93C allele varies from 5 to 50 percent in the Solomon Islands samples. We already know that the MC1R coding variant in some Neandertals is not found in the Vindija genomes. Variation in pigmentation loci may have been ubiquitous in human populations, with few fixed alleles separating populations. The ancient landscape was more like ASIP than SLC24A5.

References

Neandertals have strikingly limited genetic variation. They once lived across a range from Spain to Siberia. Yet when we compare sequences across their whole genomes, we find them to be much less different across this geographic range than people living in the same regions today.

I think this is one of the most fascinating findings of ancient genomics. It may tell us something about Neandertal populations that we did not begin to suspect without their DNA.

But there is one explanation for this fact that I and others pointed out long before DNA evidence: The Neandertal population was surely much, much smaller than Holocene population of Europe. Small population size over a long time can restrict genetic diversity. So maybe the Neandertals preserved little genetic variation simply because there were so few of them.

Neandertal mtDNA dynamism

Love Dalén and colleagues [1] add some perspective to this question. The paper adds one novel mtDNA sequence, of the Neandertal from Valdegoba, Spain, to the record of Neandertals. This builds on earlier work by Ludovic Orlando and colleagues, who performed some analysis of Neandertal variation over time when they reported the sequence of the 100,000-year-old Scladina mtDNA sequence [2]. The main contribution of the current paper is its separation of Neandertals into earlier and later subsamples, showing that the Neandertals after 48,000 years ago in Western Europe have greatly restricted mtDNA diversity compared to the earlier sample of Neandertals.

That's a tricky comparison. The paper illustrates it with this figure:

Figure 1 from Dalén et al. 2012. Original caption: "Figure 1. Phylogenetic relationships and geographic distribution of Neandertals. Recent (<48 kyr) western Neandertals are placed within a well defined monophyletic group (blue box), whereas specimens older than 48 kyr constitute a paraphyletic group together with eastern Neandertals (red box). The sampling locations for the specimens are shown with corresponding colour coding."

The blue clade includes all Neandertals after 48,000 years ago from Western Europe; the red clade includes earlier Neandertals from the west as well as both earlier and later Neandertals from the east.

The meat in this phylogram is not only that the later western Neandertals are close relatives, but that they share an ancestor only around 60,000 years ago. That's a mere 20,000 to 25,000 years before the later western Neandertals lived. The variation within these Neandertals is roughly the same as that within a single mtDNA clade within Europe today, such as clade H1.

Comparing the later Neandertal diversity to the variation of present-day Europeans helps to clarify the meaning of low diversity. Low mtDNA diversity doesn't necessarily imply that the later Neandertals in western Europe were few in number. Certainly there are millions of Europeans today who carry clade H1, for example. Low mtDNA diversity tells us something more limited about the ancestors of these Neandertals. Sometime after 60,000 years ago, a pulse of mitochondria came from the east and were remarkably successful in the west.

Looking at the red clade in the figure is also illustrative. Eastern Neandertals and earlier western Neandertals had a lot more diversity than the later western Neandertals. We have to remember that the Scladina individual lived 40,000 years before the common ancestor of the blue clade, so that the greater ages of these specimens matters. Still, when we look at the diversity in that red clade, it is greater than the mtDNA diversity today in the most widespread basal clade outside Africa, the M clade. Taking the mtDNA phylogeny alone, we would say that the 13 Neandertals had a greater sequence difference than all the people who with ancestry outside Africa today. Only when we look at the predominantly African clades today (the L clades) do we start to see sequence differences as great as among these Neandertals.

I began the post by pointing out that small population size alone might explain the low mtDNA diversity of Neandertals. Dalén and colleagues provide a key comparison that helps to reject that hypothesis. Small population size alone cannot explain the discrepancy of mtDNA diversity of these Neandertals across space and time.

The whole-genome perspective

Now, the question is whether this pattern holds true only for mtDNA, or whether the rest of the genome also shows some dynamic within Neandertal populations.

We have quite a lot of information on this point, because the initial sequencing of the Vindija Neandertals was accompanied by a smattering of sequencing of the nuclear genomes of one individual from El Sidrón, the original Feldhofer specimen and the Mezmaiskaya Neandertal specimen. The inclusion of Mezmaiskaya is important, because it alone is not included in the "low mtDNA diversity" red clade pictured above. If the pattern observed for mtDNA is reflected by the rest of the genome, the comparison between Mezmaiskaya and western Neandertal genomes should show substantially more diversity.

When they published the draft Denisova genome, Reich and colleagues [3] used it as an outgroup to investigate variation among the Neandertals, and they focused initially on Mezmaiskaya:

Using the 56 Mb of autosomal DNA sequences determined from [the Mezmaiskaya specimen], we estimate that the DNA sequence divergence between the Vindija and Mezmaiskaya Neanderthals corresponds to a date of 140,000 +/- 33,000 years ago (Supplementary Information section 6) (Fig. 1). This remarkably low divergence—which is about one-third of the closest pair of present-day humans that we analysed—is in agreement with the observation that diversity among Neanderthal mtDNAs is low relative to present-day humans and indicates that the Vindija and Mezmaiskaya Neanderthals descend from a common ancestral population that experienced a drastic bottleneck since separating from the ancestors of the Denisova individual.

That adds substantially to the mtDNA picture. The mtDNA variation of western Neandertals may reflect population turnover after 50,000 years ago. But the nuclear genome comparison cannot be explained by this single event. The variation of nuclear genomes between Mezmaiskaya and El Sidrón spans across more than half the Neandertal geographic range and requires mechanisms that restricted genetic variation across at least the period after 140,000 years ago.

I think we can do quite a bit better using the nuclear genetic information already available, keeping an explicit phylogeographic model in mind. My view is that Neandertal populations were dynamic throughout their existence, with repeated waves of population turnover across broad geographic scales. The mtDNA of later western Neandertals may reflect a large, recent event. But there must also have been earlier ones to limit variation of the nuclear genome. The implication is that early Neandertals like Krapina may have had relatively little genetic connection to later Neandertals in the same region, like Vindija.

That picture matches what we are beginning to understand about the population history of Europe during the last 30,000 years. I think that's how human populations have always behaved.

Revisiting Neandertal races

I wrote extensively about Neandertal mtDNA in 2009, noting the work of Virginie Fabre and colleagues [4], which showed the geographic structure of Neandertal mtDNA variation ("Neandertal races?"). Fabre and colleagues showed that Neandertal mtDNA variation is apportioned unequally across space, and made sense of the variation using a phylogeographic model with three broad geographic groups. I pointed out then that an alternative explanation might be that the specimens represent different times:

Many have pointed out, going back to McCown and Keith (1939), that time is another possible cause of morphological differentiation of Neandertals. The mtDNA sequences cover a wide range of times -- the Scladina sequence comes from roughly 100,000 years ago, the others cover the span from 50,000 down to 29,000 years ago. Why not test temporal groups instead of geographic groups? Temporal clusters might reflect interglacial colonizations, differential gene flow, or natural selection. There is a good precedent -- last year a report of complete mtDNA sequences from woolly mammoths found evidence for geographic structure among mtDNA lineages, one of which apparently replaced the other (Gilbert et al. 2008).

Time is just one example of an alternative model for variation. But I think it helps to clarify the basic problem of the a priori models -- you have to draw boundaries between the specimens somewhere.

The problem still remains even in the current paper. Why should we divide time arbitrarily at 48,000 years ago? Why divide time in western Europe but not across the eastern part of the Neandertals' range?

Combining space and time into a single phylogeographic picture is complicated. We end up using a null model to generate millions of pseudosamples to represent the exact time and place we found specimens, hoping to show the null model wrong. Refuting a null model doesn't necessarily tell us much about the behavior of ancient populations that flowed across space and interacted at different times. I think that life was more complicated rather than less, and look to models from more recent populations to understand it.

How not to publicize your work

The paper by Dalén and colleagues is such a neat piece of work, I think it's a shame that Uppsala University had to go and spoil it with this silly press release: "European Neanderthals Were On the Verge of Extinction Even Before the Arrival of Modern Humans".

The paper pointedly does not show that Neandertals were on the "verge of extinction". Neandertals in the eastern part of their range show no sign of any demographic collapse, and the western part of the range arguably only shows signs of recovery and expansion.

What the paper actually tells us is about the dynamism of Neandertal populations, which is very comparable to that of the Europeans of the last 10,000 years. Keeping this comparison in mind helps remind us that very large groups of people may still have low mtDNA diversity, reflecting the history of population movements and interactions in the past. Comparing the mtDNA with nuclear genetic evidence is also essential to this picture. Neither of these tell us that Neandertals were near extinction.

Please, if you're putting together a press package about Neandertals, stop framing it around the concept of Neandertal extinction. You aren't going to say anything novel about this, and it just encourages lazy science writing. And it's a false concept. The Neandertals didn't become extinct.

UPDATE (2012-03-06): A reader points out that several of the dates for specimens in the paper are different than reported in the literature. I noticed that too, and don't know quite what to make of it. I don't think that the differences in dates affect the general result, that later specimens in Western and Central Europe are relatively invariant compared to the Eastern European and Asian sample. But it is a reminder that the results do depend on a certain ordering and geographic sampling of specimens and may change if we fill in the gaps.

References