The media is blowing up with a new story about the phylogeography and phylogenetics of the domestic dog.* The New York Times has a good write up, and I like its title: Central Asia Could Be Birthplace of the Modern Dog (the headline was changed to “15,000 Years Ago, Probably in Asia, the Dog Was Born” while I was writing this post, which strikes me as even more tentative). The conditional clause is pretty important, because this is not definitive, but suggestive. The paper is titled Genetic structure in village dogs reveals a Central Asian domestication origin, but the text itself is a lot more qualified than the title might let on. There are many more chapters to this story to be told.
The issue of the where dogs came from is kind of a big deal. There have been claims for the origin of dogs in East Asia, Middle East, and Europe. If you look at these papers there are major issues in terms of the underlying data used to make these inferences (e.g., the East Asian argument relies on Y lineages, while the arguments favoring Europe are based on mtDNA, with a particular comparison to ancient genomes from Europe, and the Middle Eastern result turns out to have had issues of sample bias). To be fair, you can only make arguments based on the data you have on hand. But, the track record for humans should make us careful about that in relation to strong general inferences.
A big upside of this particular paper is that the massive geographic coverage. This took a lot of work, as I discussed this with Ryan Boyko (second author) a few years ago (dogs can be ornery and village dogs in particular are not hygienic creatures). Additionally, as they note they use a “semicustom” SNP-chip, with nearly 200,000 markers. This is sufficient density in mammals for most inference, and is probably overkill for breed analysis which used to be the bread & butter of dog genetics. As you can see from the PCA plot, a lot of the genetic variation is partitioned by geography.
Basically, dogs at some point diverged from a common ancestor, and accrued variation which distinguished lineages by region. Second, the spread of European origin dogs is a major dynamic which affects much of the world. In the paper itself the authors mention that there’s going to be some ascertainment bias in their sample set toward this admixture in non-European lineages because of the concentration of post-colonial lineages in cities. You see similar patterns with domestic cats, which is an organism I know better. A rule of thumb seems to be that if Christianity was successful a region, the local domestic lineages may also have been swamped (perhaps this is correlated with cultural complexity, which are correlated with dogs well adapted toward resisting invasive European dogs?). Nevertheless, there are interesting populations like the Carolina dog which preserve pre-European lineages.
The major finding, allowed by their geographic finding, is that Central Asian dogs exhibit evidence of a long ago bottleneck in their genome due to their pattern of linkage disequilibrium (LD). LD can be thought of as a violation of Mendel’s law of assortment on a biomolecular scale. Genetic variants at different loci are co-inherited together at a higher fraction than would be the expected case if they were segregating in the population independently (e.g., if allele A and a are both at 0.5, and B and b are 0.5, then A should be found in equal proportion associated with B and b, but in a case of LD it wouldn’t be). In a physical sense one can imagine these as haplotypes, sequences of variation across the genome, where particular alleles are often found on the same segment.
All this is important as LD is a signature of different demographic events, including the bottleneck above. When you have a very recent admixture between two populations you get elevated LD. Adjoining ancestrally informative alleles take generations to disassociate from each other through of recombination. But bottlenecks can also induce increased LD, as particular haplotypes increase rapidly in frequency through sampling processes. Populations which have gone through sharp recent reductions in population will exhibit long range LD. LD between alleles which are posited relatively far apart on a genomic scale. Over time recombination breaks apart these associations, and LD gets shorter and shorter. This “decay” in LD over time has been used to peg admixture dates, but in this case the authors note that:
LD is lowest in Afghanistan and Central Asia at short inter-SNP distances (< 0.0005 cM) and lowest in Vietnam at intermediate distances (0.01–0.05 cM), with rates increasing in other populations depending on their isolation and distance from Asia. These patterns of LD decay strongly suggest a Central Asian origin for domestic dogs with a subsequent population expansion (larger contemporary Ne) in East Asia and elsewhere. These patterns are consistent if physical, rather than genetic, inter-SNP distance is measured, or if different subsets of dogs are used for each population
To my mind there are two other major reasons that it is likely that Central Asia is a highly likely candidate region for the origin of modern domesticates. Simply, and alluded to in the paper, is that Central Asia is in a central position in relation to the full pre-Columbian range of the species. The likelihood from all the genetic evidence is that dogs are descended from a population similar to Eurasian wolves; so their locus of origin has to be in the Palearctic ecozone. Second, there is archaeological signs that canids with a dog-like morphology arose very early here. Ancient DNA also lends some support to this supposition.
Is this is a slam dunk case? Not at all. As some commentators have pointed out, at a certain level of granularity inferences from extant variation have limited utility. But the question is the level of granularity…modern human origin in Africa inferred from archaeology, mtDNA, and microsatellites, have been broadly supported. This seems to be the scale that this sort of paper is targeting.
The second major issue that this paper tries to pin indirectly is the period of domestication. They converge upon the date of ~15,000 years before the present, which seems to be the new mainstream/conservative estimate. An ancient DNA paper which was published this spring though pushes the possible date as far back as ~40,000 years before the present. Additionally, dogs moved with humans, and it strikes me that if the Amerindians brought dogs with them, then a date of ~15,000 years is close to the lower bound (the distribution of dates isn’t symmetrical).
In a few years, with lots of whole-genome analyses with this level of geographic coverage, as well as ancient DNA, I expect that some of the story above will be confirmed, but the overall picture is likely to be complex and more bizarre than we would expect. I say that because it seems very unlikely to me that dogs are going to be any simpler a story than that of humans. If the vast majority of Northern Europeans were replaced ~5,000 years ago, was there a concomitant replacement of dogs? Some ancient DNA indicates replacement, just as has been the case with humans. It seems likely that large animals during the Pleistocene underwent complex meta-population dynamics of extinction and recolonization several times. One can imagine a scenario where the vast majority of modern dog ancestry derives from Central Eurasian ~15,000 years before the present, after the Last Glacial Maximum, but where threads of more diverged ancestry persist in some populations, not to mention acknowledged gene flow from local wolves.
At the end of the day, if someone made me bet where the ancestral lineage which diverged from the ancestors of Eurasian wolves flourished, I would say Central Asia. But, my confidence in this assertion is only moderate, and it seems likely that even this simple answer might mislead us as to the complexity of the bigger picture. I suspect we’ll known the answer within the next five years.
Addendum: This paper has some of the first large scale analysis of Y chromosomal SNP variation. Intriguingly they found “We also see indigenous Mt haplotypes segregating in Carolina dogs and Xoloitzcuintlis, but no unique Y haplotypes indicative of indigenous ancestry were found in American dogs outside of the Arctic.” This strangely recapitulates sex-biased gene flow in the New World for humans, indicating that the story of man and his best friend exhibit more similarities than we might have thought.
* I should add that several of the authors on this paper are friends.