domestication

I know, what an exciting headline!

I've written quite a bit about the origins of domesticated cattle and introgression among the species of wild cattle giving rise to the current pattern of genetic diversity. I'm keeping track of that area because the process of domestication and subsequent interaction of domesticates with their wild relatives provide one kind of natural model for the interaction of ancient human groups such as the Neanderthals. We used these examples in our 2006 paper [1].

Cattle have been a convenient example because there has been a lot of genetic work on them, and they have multiple wild species that diverged early in the Pleistocene. But other domesticates are also intense targets of sequencing and genome discovery, and as we understand more about their variation, we are beginning to find interesting patterns. Going through my notes today I found an interesting paper on sheep MHC polymorphisms [2]:

Trans-Species Polymorphism and Selection in the MHC Class II DRA Genes of Domestic Sheep

Highly polymorphic genes with central roles in lymphocyte mediated immune surveillance are grouped together in the major histocompatibility complex (MHC) in higher vertebrates. Generally, across vertebrate species the class II MHC DRA gene is highly conserved with only limited allelic variation. Here however, we provide evidence of trans-species polymorphism at the DRA locus in domestic sheep (Ovis aries). We describe variation at the Ovar-DRA locus that is far in excess of anything described in other vertebrate species. The divergent DRA allele (Ovar-DRA*0201) differs from the sheep reference sequences by 20 nucleotides, 12 of which appear non-synonymous. Furthermore, DRA*0201 is paired with an equally divergent DRB1 allele (Ovar-DRB1*0901), which is consistent with an independent evolutionary history for the DR sub-region within this MHC haplotype. No recombination was observed between the divergent DRA and B genes in a range of breeds and typical levels of MHC class II DR protein expression were detected at the surface of leukocyte populations obtained from animals homozygous for the DRA*0201, DRB1*0901 haplotype. Bayesian phylogenetic analysis groups Ovar-DRA*0201 with DRA sequences derived from species within the Oryx and Alcelaphus genera rather than clustering with other ovine and caprine DRA alleles. Tests for Darwinian selection identified 10 positively selected sites on the branch leading to Ovar-DRA*0201, three of which are predicted to be associated with the binding of peptide antigen. As the Ovis, Oryx and Alcelaphus genera have not shared a common ancestor for over 30 million years, the DRA*0201 and DRB1*0901 allelic pair is likely to be of ancient origin and present in the founding population from which all contemporary domestic sheep breeds are derived. The conservation of the integrity of this unusual DR allelic pair suggests some selective advantage which is likely to be associated with the presentation of pathogen antigen to T-cells and the induction of protective immunity.

That probably deserves more thought and explanation that I can give it right now. As the authors point out in the paper, sheep domestication was a complicated process:

The complex origin of domestic sheep is apparent from the presence of at least five distinct mitochondrial lineages [20], some of which cannot be traced to a wild ancestor [24], [25]. This diversity is likely to originate from geographically isolated subspecies of wild sheep that have hybridised as a result of human migrations over the 8–10 millennia since the initial domestication events in the Near East and Asia [26]–[28]. Frequent hybridization events are likely to have occurred between domesticated and local wild populations providing the high levels of MHC diversity evident in present day domestic populations as well as a degree of resistance to endemic disease and adaptation to local environmental conditions [29].

Their interpretation of an ancient selective balance, retained in domesticated sheep from very distant common ancestors with oryx, probably is the most likely scenario. But I think this provides a nice example of how difficult it is to tell ancient balanced polymorphisms apart from relatively recent hybridization. That's a problem that we continue to face with interpreting human and Neandertal genetic variation.

Also the case illustrates how important is the mixture of different wild populations in the origin of domesticates. Even if a wild population makes up a very small fraction of the genetic heritage of the current domesticated species, one or more adaptive loci from that population may nevertheless be very important to the survival and success of the later species.

Genes don't care where they came from, and their function is not irrevocably marked by their origin.

References

I feel like I've been transported into the future to see what science will be like fifteen years from now:

We successfully typed eight mutations in six genes (6) responsible for coat color variation for 89 [out of 152 tested; tables S1 to S5 (6)] ancient samples. To assess coat color variation of predomestic horses, we analyzed the bones of wild horses from the Late Pleistocene and Early Holocene found in Siberia, East and Central Europe, and the Iberian Peninsula. We found no variation in the Siberian and European Pleistocene horses, suggesting that these horses were bay or bay-dun in color.

...

In contrast, a rapid and substantial increase in the number of coat colorations is found in both Siberia and East Europe beginning in the fifth millennium B.P. (Fig. 1 and figs. S1 and S2). Although the earliest chestnut allele (MC1R gene) was identified in a Romanian sample from the late seventh millennium B.P., chestnut horses were first observed in Siberia (fifth millenium B.P.). Their prevalence increased rapidly, reaching 28% during the Bronze Age.

The whole paper is only a page long. Like, "Oh yeah, we just genotyped 89 horse skeletons from the Neolithic up to the Bronze Age, and here's how the process of selection on color patterns worked out."

Couple of things --

1. The pigment-altering mutations at these genes do not all show statistical signs of selection in contemporary samples of horses. But they aren't there in the ancient horses. That's the best evidence of selection you could possibly have. Message: tests of selection on contemporary samples are weak, particularly for loci with rare alleles or more than two alleles.

2. The study shows how much you can learn with a moderate number of samples. The advantages working in this case: the genes responsible for pigmentation phenotypes are well-characterized, the number of loci is well-matched by the sample size; the study focuses on recovering SNPs instead of sequencing.

UPDATE (2009-04-25): I looked at the statistical test of selection used in the paper with a little more detail. It comes from a paper by Jonathan Bollback, Thomas York and Rasmus Nielsen, published last year in Genetics. It's a very clever test. Assume a sample of genes taken from a population at some discrete set of times, t₁, t₂, t₃, and so on. Under genetic drift, this series of frequencies approximates a random walk, with the size of deviations depending on the effective population size. Under constant directional selection, the random walk will be biased in a way that can be approximated by a diffusion model of selection.

So if you have a big enough sample, you can estimate the relative contribution of stochastic (drift) effects and deterministic (directional selection) effects on the frequencies over time.

In the current case, two alleles have sample sizes that are large enough, frequency changes that are large enough and constant enough in direction to show that drift is minor and selection is strong. For example, the chestnut variant goes from zero in the Neolithic to 65 percent in the Medieval time period, with every change a between time increments an increase. That's an allele with 44 copies (by my quick count) in all samples.

Four alleles do not have such large changes (they are initially absent, but present in only four or fewer individuals in the later time intervals). There's only one copy of the SILV9 color variant mutation in the entire sample of archaeological horses. With a very small sample, the sampling variance will be larger than the stochastic effects of drift. So even if those few copies are exclusively in the latest time increments, the time series won't look unusual compared to drift.

But what the test doesn't consider is the current frequency. Since this is known with much less sampling variance, we can compare the present frequency with the frequency summed across the older time intervals to get a more powerful test of neutrality. Also, the test assumes that selection is constant -- if selection had a stochastic element, varying over time, that would have the same impact on the sample as drift or sampling variance. That's why the absence of an allele in an ancient sample can be stronger evidence of selection than the fine-scaled record of change over time.

Complicating matters is population structure. The coat color variants are not distributed evenly across the modern horse population; they are distributed into different breeds. The strong association of some color variants with breeds is, of course, evidence of its own that the color variants have been correlated with fitness under domestication. Now, whether this fitness association is a deliberate result of people liking colors (because they differentiate breeds, or look pretty, or whatever) or whether they arise incidentally (by linkage with other traits) isn't tested by these data. These functional aspects of selection provide another possible test of neutrality -- for example, in this case all the non-bay-black alleles increased over time, a bias that isn't consistent with chance.

References:

Bollback JP, York TL, Nielsen R. 2008. Estimation of 2N_es from temporal allele frequency data. Genetics 179:497-502. doi:10.1534/genetics.107.085019

Ludwig A, Pruvost M, Reissmann M, Benecke N, Brockmann GA, Castaños P, Cieslak M, Lippold S, Llorente L, Malaspinas A-S, Slatkin M, Hofreiter M. 2009. Coat color variation at the beginning of horse domestication. Science 324:485. doi:10.1126/science.1172750