
Editor's Introduction
Genomic and archaeological evidence suggest a dual origin of domestic dogs
We know that dogs are descended from wolves, but when and where they were domesticated has been up for debate. Using modern genetic techniques, a wealth of information can be gained from the DNA of dogs that existed thousands of years ago. Scientists gathered the genetic information of many dogs, both modern and ancient, and a new fossil from an ancient dog found in Ireland. A large, international team of archaeologists and geneticists collaborated to solve the puzzle of dog domestication. Were dogs domesticated once? Twice? Where did it happen?
Paper Details
Supporting Media Partner

Abstract
The geographic and temporal origins of dogs remain controversial. We generated genetic sequences from 59 ancient dogs and a complete (28x) genome of a late Neolithic dog (dated to ~4800 calendar years before the present) from Ireland. Our analyses revealed a deep split separating modern East Asian and Western Eurasian dogs. Surprisingly, the date of this divergence (~14,000 to 6400 years ago) occurs commensurate with, or several millennia after, the first appearance of dogs in Europe and East Asia. Additional analyses of ancient and modern mitochondrial DNA revealed a sharp discontinuity in haplotype frequencies in Europe. Combined, these results suggest that dogs may have been domesticated independently in Eastern and Western Eurasia from distinct wolf populations. East Eurasian dogs were then possibly transported to Europe with people, where they partially replaced European Paleolithic dogs.
Video. Question: Where were dogs first domesticated? Answer: It's complicated (Courtesy Science Magazine).
Report
Dogs were the first domestic animal and the only animal domesticated before the advent of settled agriculture (1). Despite their importance in human history, no consensus has emerged regarding their geographic and temporal origins, or whether dogs were domesticated just once or independently on more than one occasion. Although several claims have been made for an initial appearance of dogs in the early upper Paleolithic [~30,000 years ago; e.g. (2)], the first remains confidently assigned to dogs appear in Europe ~15,000 years ago and in Far East Asia over 12,500 years ago (1, 3). Although archaeologists remain open to the idea that there was more than one geographic origin for dogs [e.g. (4, 5)], most genetic studies have concluded that dogs were probably domesticated just once (6), disagreeing on whether this occurred in Europe (7), Central Asia (8), or East Asia (9).
Recent paleogenetic studies have transformed our understanding of early human evolution [e.g. (10, 11)]. We applied a similar approach to reconstruct the evolutionary history of dogs. We generated 59 ancient mitochondrial DNA (mtDNA) sequences from European dogs (from 14,000 to 3000 years ago) as well as a high-coverage nuclear genome (28x) of an ancient dog dated to ~4800 calendar years before the present (12) from the Neolithic passage grave complex of Newgrange (Sí an Bhrú) in Ireland. We combined our ancient sample with 80 modern publically available full genome sequences and 605 modern dogs (including village dogs and 48 breeds) genotyped on the CanineHD 170,000 (170 K) single-nucleotide polymorphism (SNP) array (12).
We first assessed characteristics of the Newgrange dog by typing SNPs associated with specific phenotypic traits and by inferring its level of inbreeding, compared to other breed and village dogs (12). Our results suggest that the degree of artificial selection and controlled breeding during the Neolithic was similar to that observed in modern free-living dogs. In addition, the Newgrange dog did not possess variants associated with modern breed-defining traits, including hair length or coat color. And although this dog was likely able to digest starch less efficiently than modern dogs, it was able to do so more efficiently than wolves (12).
A phylogenetic analysis based on 170 K SNPs revealed a deep split separating the modern Sarloos breed from other dogs (Fig. 1A). This breed, created in the 1930s in the Netherlands, involved breeding German Shepherds with captive wolves (13), thus explaining the breed’s topological placement. The second deepest split [evident on the basis of both the 170 K SNP panel (Fig. 1A) and genome-wide SNPs (fig. S4)] separates modern East Asian and Western Eurasian (Europe and the Middle East) dogs. Moreover, the Newgrange dog clusters tightly with Western Eurasian dogs. We used principal components analysis (PCA), D statistics, and the program TreeMix (12) to further test this pattern. Each of these analyses unequivocally placed the Newgrange dog with modern European dogs (figs. S5 to S7). These findings demonstrate that the node separating the East Asian and Western Eurasian clades is older than the Newgrange individual, which was directly radiocarbon dated to ~4800 years ago.
Fig. 1. Deep split between East Asian and Western Eurasian dogs. (A) A neighbor-joining tree (with bootstrap values) based on identity by state (12) of 605 dogs. Red and yellow clades represent the East Asian and Western Eurasian core groups, respectively (12). (B) A map showing the location and relative proportion of ancestry [mean D values (12)] of dogs (fig. S10). Negative values (red) indicate that the population shares more derived alleles with the East Asian core, whereas positive values (yellow) indicate a closer association with the Western Eurasian core.
Ancestry as a tree
In figure 1A, DNA sequences were used to graph how dogs—both modern and ancient—evolved from each other. Who came first? Who are the ancestors and who are the descendants?
Check out HHMI BioInteractive's resource to learn how to build a phylogenetic tree and how to make sense of existing trees.
https://www.hhmi.org/biointeractive/creating-phylogenetic-trees-dna-sequences
Dogs around the world
As you can see in figure 1B, most of the dogs in the Western Hemisphere are represented in yellow, which means that they belong in the West Eurasian core group. There are a few exceptions, such as the Greenland Sledge dog and Eurasier, that are represented in varying shades of orange meaning that they have traces of admixture from the East Asian core group—they share small amounts of DNA with this group and therefore have ancestors in both the West Eurasian and East Asian core groups.
Other nodes leading to multiple dog populations and breeds [including the basal breeds (1) such as Greenland sledge dogs or the Siberian husky (Fig. 1A)] are poorly supported, suggesting that these breeds probably possess mixed ancestry from both Western Eurasian and East Asian dog lineages. To further assess the robustness of the deep split and those nodes associated with the potentially admixed lineages, we defined Western Eurasian and East Asian “core” groups (Fig. 1A), supported by the strength of the node leading to each cluster (12). We then used D statistics to assess the affinity of each population to either Western Eurasian or East Asian core groups (12). The results of this analysis again revealed a clear East-West geographic pattern across Eurasia associated with the deep phylogenetic split (Fig. 1B). Breeds such as the Eurasier, Greenland sledge dogs, and Siberian huskies [all basal breeds from northern regions (1)], however, possess strong signatures of admixture with the East Asian core samples (fig. S11), as do populations sampled in East Asia that clustered alongside Western Eurasian dogs (such as a Papua New Guinean village dog; Fig. 1A).
We used the multiple sequentially Markovian coalescent (MSMC) (12, 14) to reconstruct the population history of East Asian and Western Eurasian dogs. An analysis of individual high-coverage genomes demonstrated a long shared population history between the Newgrange dog and modern dogs from both Western Eurasia and East Asia (fig. S15). A reconstruction using two genomes per group improved the resolution for recent time periods (Fig. 2A) and revealed a bottleneck in the Western Eurasian population, after its divergence from the East Asian core. A similar bottleneck observed in non-African human populations has been interpreted as a signature of a migration out of Africa (15). We therefore speculate that the analogous bottleneck observed in our data set could be the result of a divergence and subsequent transportation of dogs from east to west, supporting suggestions drawn from recent analyses of modern dog genomes (8, 9, 16).
Fig. 2. Effective population size, divergence times, and mtDNA. (A) Effective population size through time of East Asian and Western Eurasian dogs and wolves with MSMC. (B) CCR per year for each population pair in Fig. 2A. The CCR represents the ratio of within- and between-population CRs. The ratio measures the age and pace of divergence between two populations. Values close to 1 indicate that both within and between CR are equal, meaning that the two populations have not yet diverged. Values close to 0 indicate that the populations have completely diverged. mu, mutation rate; g, generation time. (C) Bar plot representing the proportion of mtDNA haplogroups at different time periods. (D) Locations of archaeological sites, with haplogroup proportions. (E) Locations of modern samples, with haplogroup proportions.
Evolutionary clock
In figure 2B, four different groups—the Newgrange dog, wolves, East Asian, and Western Eurasian—were paired. With a total of six comparisons, scientists estimated the timing of when each group split from each other. The Y-axis represents the cross coalescence rate (CCR). At a relative CCR of one, we learn that the Newgrange dog and Western Eurasian dogs split most recently (less than 10,000 years ago), whereas East Asian dogs and wolves split a very long time ago (more than 70,000 years ago).
Haplogroups
Haplogroups are populations that share a common ancestor.
Descendants and their ancestors will share more of their DNA. Haplogroups can be identified by studying the variation that occurs in certain parts of mitochondrial DNA. Certain pieces of DNA are inherited together from a common ancestor, so by studying the combinations of genes that are inherited we can assign the ancestor and its descendants to a haplogroup. Haplogroups are assigned a letter (A, B, C, or D in this case).
We then used MSMC to compute divergence times as a means to assess the time frame of the shared population history among dogs, and between dogs and wolves. To obtain reliable time estimates, we used the radiocarbon age of the Newgrange dog to calibrate the mutation rate for dogs (12) (fig. S16). This resulted in a mutation rate estimate of between 0.3 × 10−8 and 0.45 × 10−8 per generation, which is similar to that obtained with an ancient grey wolf genome (17). Using this mutation rate, we calculated the divergence time between the two modern Russian wolves (18) used in this study and the modern dogs to be 60,000 to 20,000 years ago (fig. S17 and Fig. 2B). This date should not be interpreted as a time frame for domestication, because the wolves we examined may not have been closely related to the population(s)/s that gave rise to dogs (6).
These analyses also suggested that the divergence between the East Asian and Western Eurasian core groups (~14,000 to 6400 years ago) occurred commensurate with, or several millennia after, the earliest known appearance of domestic dogs in both Europe (>15,000 years) and East Asia (>12,500 years) (1) (Fig. 2B and fig. S17). In addition, admixture signatures from wolves into Western Eurasian dogs most likely pushed this estimated time of divergence deeper into the past (12), meaning that the expected time of divergence between the Eastern and Western cores is probably later than our estimate. These results imply that indigenous populations of dogs were already present in Europe and East Asia during the Paleolithic (before this genomic divergence). Under this hypothesis, this early indigenous dog population in Europe was replaced (at least partially) by the arrival of East Eurasian dogs.
To investigate this potential replacement, we sequenced and analyzed 59 hypervariable mtDNA fragments from ancient dogs spread across Europe, and we combined those with 167 modern sequences (12). Each sequence was then assigned to one of four major well-supported haplogroups (groups A to D) (19). Although the majority of ancient European dogs belonged to either haplogroup C or D (63 and 20%, respectively), most modern European dogs possess sequences within haplogroups A and B (64 and 22%, respectively) (Fig. 2, C to E). Using simulations, we showed that this finding cannot be explained by drift alone (12). Instead, this pattern arose from clear turnover in the mitochondrial ancestry of European dogs, most likely as a result of the arrival of East Asian dogs. This migration led to a partial replacement of ancient dog lineages in Europe that were present by at least 15,000 years ago (1).
Although the mtDNA turnover is obvious, the nuclear signature reveals an apparent long-term continuity. Assessments of ancestry in humans have demonstrated that major (nuclear) turnovers can be difficult to detect without samples from the admixing population (11). A genome-wide PCA analysis revealed that PC2 clearly discriminates the Newgrange dog from other modern dogs (fig. S8), suggesting that this individual could possess ancestry from an unsampled population.
Our MSMC analysis revealed that the population split between the Newgrange dog and the East Asian core [as measured by cross coalescence rate (CCR)] is older (on average) than the split between modern Western Eurasian and East Asian lineages (Fig. 2B). Simulations suggest that this pattern could be explained by a partial replacement model in which the Newgrange dog retained a degree of ancestry from an outgroup population (fig. S20, A and B) that was different from modern wolves (12). Alternatively, this pattern could also be explained by secondary gene flow from Asian dogs into modern European dogs (fig. S20C). Nevertheless, our simulations show that secondary gene flow has a smaller effect on CCR than the partial replacement model (fig. S20, B and D). Moreover, secondary gene flow cannot explain the placement of the Newgrange dog on our genome-wide PCA (fig. S8). Overall, these observations are consistent with a scenario in which the Newgrange dog retained a degree of ancestry from an ancient canid population that falls outside of the variation of modern dogs, but that is also different from modern wolves. This pattern also suggests that the replacement of European indigenous Paleolithic dogs may not have been complete.
To assess the consilience between our results and the archaeological record, we compiled evidence for the earliest dog remains across Eurasia (Fig. 3A). We found that although dogs are present at sites as old as 12,500 years in Eastern Eurasia (China, Kamchatka, and East Siberia) and 15,000 years in Western Eurasia (Europe and the Near East), dog remains older than 8000 years have yet to be recovered in Central Eurasia (Fig. 3A and table S7). Combined with our DNA analyses, this observation suggests that two distinct populations of dogs were present in Eastern and Western Eurasia during the Paleolithic.
Fig. 3. Archaeological evidence for the first appearance of dogs across Eurasia and a model of dog domestication. (A) Map representing the geographic origin and age of the oldest archaeological dog remains in Eurasia (12). (B) A suggested model of dog domestication under the dual-origin hypothesis. An initial wolf population splits into East and West Eurasian wolves that were then domesticated independently before becoming extinct (as indicated by the † symbol). The Western Eurasian dog population (European) was then partially replaced by a human-mediated translocation of Asian dogs at least 6400 years ago, a process that took place gradually after the arrival of the eastern dog population.
Follow the dots
In figure 2A we see that there are only blue dots in Central Eurasia, while there are both red and blue dots in Western Eurasia. Why? In the scale, we see that blue represents the oldest samples whereas red represents the youngest samples. The best explanation is that dogs were transported by humans from East to West.
Dual origin theory
Challenging the pre-existing theory, these authors propose that dogs were domesticated twice—once in Europe and again in Asia. The first dogs in Europe didn't survive. They died off (mostly, if not completely) and were replaced by dogs that were transported by humans from East Asia.
The establishment of these populations is consistent with two scenarios: a single origin of Eurasian dogs, followed by early transportation, founder effects, isolation, and drift; or two independent domestication processes on either side of Eurasia. In the first scenario, the archaeological record should reveal a temporal cline of the first appearance of dogs across Eurasia stemming from a single source. Given the current lack of dog remains prior to 8000 years ago in Central Eurasia, a scenario involving a single origin followed by an early dispersal seems less likely.
Given our combined results, we suggest the following hypothesis: Two genetically differentiated and potentially extinct wolf populations in Eastern (8, 9) and Western (7) Eurasia may have been independently domesticated before the advent of settled agriculture (Fig. 3A). The eastern dog population then dispersed westward alongside humans at some point between 6400 and 14,000 years ago, into Western Europe (10, 11, 20), where they partially replaced an indigenous Paleolithic dog population. Our hypothesis reconciles previous studies that have suggested that domestic dogs originated either in East Asia (9, 19) or in Europe (7). For numerous reasons, the null hypothesis should be that individual animal species were domesticated just once (21). The combined genetic and archaeological results presented here, however, suggest that dogs, like pigs (22), may have been independently domesticated twice. Nevertheless, given the complexity of the evolutionary history of dogs and uncertainties related to mutation rates, generation times, and the incomplete nature of the archaeological record, our scenario remains hypothetical. Genome sequences derived from ancient Eurasian dogs and wolves, combined with detailed morphological and contextual studies of the archaeological remains, will provide the necessary means to assess whether dog domestication occurred more than once.
SUPPLEMENTARY MATERIALS
www.sciencemag.org/content/352/6290/1228/suppl/DC1
Materials and Methods
Figs. S1 to S29
Tables S1 to S7
REFERENCES AND NOTES
1. G. Larson et al., Proc. Natl. Acad. Sci. U.S.A. 109, 8878–8883 (2012).
2. M. Germonpré, M. Lázničková-Galetová, M. V. Sablin, J. Archaeol. Sci. 39, 184–202 (2012).
3. M. Pionnier-Capitan et al., J. Archaeol. Sci. 38, 2123–2140 (2011).
4. T. Dayan, J. Archaeol. Sci. 21, 633–640 (1994).
5. M. Ollivier et al., PLOS One 8, e75110 (2013).
6. A. H. Freedman et al., PLOS Genet. 10, e1004016 (2014).
7. O. Thalmann et al., Science 342, 871–874 (2013).
8. L. M. Shannon et al., Proc. Natl. Acad. Sci. U.S.A. 112, 13639–13644 (2015).
9. G.-D. Wang et al., Cell Res. 26, 21–33 (2016).
10. M. E. Allentoft et al., Nature 522, 167–172 (2015).
11. W. Haak et al., Nature 522, 207–211 (2015).
12. See the supplementary materials.
13. D. Morris, Dogs: The Ultimate Dictionary of Over 1,000 Dog Breeds (Trafalgar Square, North Pomfret, VT, 2008).
14. S. Schiffels, R. Durbin, Nat. Genet. 46, 919–925 (2014).
15. H. Li, R. Durbin, Nature 475, 493–496 (2011).
16. M. Pilot et al., Proc. R. Soc. B Biol. Sci. 282, 20152189 (2015).
17. P. Skoglund, E. Ersmark, E. Palkopoulou, L. Dalén, Curr. Biol. 25, 1515–1519 (2015).
18. G. D. Wang et al., Nat. Commun. 4, 1860 (2013).
19. P. Savolainen, Y. P. Zhang, J. Luo, J. Lundeberg, T. Leitner, Science 298, 1610–1613 (2002).
20. L. M. Cassidy et al., Proc. Natl. Acad. Sci. U.S.A. 113, 368–373 (2016).
21. G. Larson, J. Burger, Trends Genet. 29, 197–205 (2013).
22. L. A. F. Frantz et al., Nat. Genet. 47, 1141–1148 (2015).
ACKNOWLEDGMENTS
Raw reads of the Newgrange dog have been deposited at the European Nucleotide Archive (ENA) with project number PRJEB13070. Mitochondrial sequences as well as genotype files (in plink format) were deposited on Dryad (doi:10.5061/dryad. 8gp06). We thank G. Wang, J. Schraiber, L. Orlando, L. Dalén, R. E. Green, P. Savolainen, and E. Loftus for their valuable input; and A. Osztás, I. Zalai-Gaál, R.-M. Arbogast, A. Beeching, A. Boroneant, O. Lecomte, S. Madeleine, C. and D. Mordant, M. Patou-Mathis, P. Pétrequin, L. Salanova, J. Schibler, A. Tsuneki, and F. Valla for providing archaeological material. L.A.F.F., O.L., A.L., and G.L. were supported by a European Research Council grant (ERC-2013-StG-337574-UNDEAD) and Natural Environmental Research Council grants (NE/K005243/1 and NE/K003259/1). L.A.F.F. was supported by a Junior Research Fellowship (Wolfson College, University of Oxford). M.P.-C. was supported by a BDI CNRS grant. V.E.M, V.M, M.D.T., and the sequencing of the Newgrange dog genome were funded by ERC Investigator grant 295729-CodeX awarded to D.G.B. We also acknowledge the National Museum of Ireland for providing the petrous bone of the Newgrange dog and the Science Foundation Ireland Award 12/ERC/B2227 and Trinseq. A.B. was supported by a Romanian National Authority for Scientific Research (PN-II-ID-PCE-2011-3-1015). The work at ENS Lyon and at Muséum National d'Histoire Naturelle Paris was also supported by Nestlé Purina. The authors declare no conflict of interest.