A detailed picture of the origin of the Australian dingo, obtained from the study of mitochondrial DNA

Samples. Dingoes (n = 211), dogs (n = 22), and pre-European archaeological dog (n = 19) samples from Polynesia were sequenced in this study and compared with 654 dog and 38 wolf samples used by Savolainen et al. (11). For the geographic origin of dingo samples, see Table 1. The origins of the dog samples [for a detailed list of geographic origin and breeds, see Savolainen et al. (11)] were Europe (n = 207), Africa (n = 35), Southwest Asia (n = 90), India (n = 27), Siberia (n = 24), Arctic America (n = 25), China/Mongolia/Korea (n = 142), Japan (n = 96), Vietnam/Cambodia/Thailand (n = 18), Indonesia (n = 7), Malaysia (n = 2), Philippines (n = 1), and Highland New Guinea (n = 2). Prehistoric archaeological dog samples were from Cook Island (n = 2 dogs from sites dating from 1,000 B.P.), New Zealand (n = 13 from sites dating 700–400 B.P.), and Hawaii (n = 4 from prehistoric sites of unknown age).

Nineteen dingoes were from captivity in zoos, wildlife parks, dingo-conservation groups, and dingo fanciers, and 192 animals were sampled from the wild from 27 regions across the country. The wild animals were chosen based on similarity in appearance to dingoes to exclude as far as possible feral dogs and dog × dingo hybrids. The main areas sampled were the Pilbarra area of Western Australia (25 samples from 10 localities), the north tablelands of New South Wales (20 samples mainly from two national parks), southeastern New South Wales (65 samples from five major localities), and northeastern Victoria (18 samples from two localities). The remaining samples were individual samples from various areas across the country.

PCR and Sequencing. DNA extraction, PCR, and sequence analysis of dingo and modern dog samples were performed as described by Savolainen et al. (11). Five hundred and eighty-two base pairs, positions 15,458–16,039 of the dog mitochondrial genome (12), were analyzed. Archaeological samples were analyzed as described by Matisoo-Smith et al. (13) by using primers L15910 and H16498 [described by DeSalle et al. (14)] and sequenced from at least two independent PCRs to validate polymorphic positions. A total of 263 bp, positions 15,458–15,720, was analyzed.

Phylogenetic Analysis. The tree in Fig. 1A, was calculated as described (11). Briefly, a neighbor-joining starting tree was calculated by using an HKY nucleotide-substitution model with invariable sites and a γ distribution to describe variable rates across sites that were not unvaried (I = 0.7266 and α = 0.6361). The root of the dog/wolf clade was derived by midpoint rooting and a Kishino–Hasegawa maximum-likelihood test of alternative rooting points by using rell bootstrap (1,000 replicates). The tree was searched further by >2.5 million tree bisection and reconnection iterations without additional improvement. The individual dog clades A–D and F were imploded, because the network structure within the dog clades, caused by a large degree of homoplasy, could not be described accurately by tree bifurcations, as depicted in clade A in Fig. 1B. All established dog clades were retained compared with previous reports (10, 11), but note that, because of the homoplasy, the order between some of the clades is different. The order between these clades should be regarded as uncertain at this point. It is important, however, that the order did not affect any of the conclusions regarding dog and dingo origins. The minimum-spanning network shown in Fig. 1B was constructed by using ARLEQUIN software (15).

Calculation of Genetic Distances. Based on the assumption that A29 was the only founder mtDNA type, an estimate of the time for the introduction of dingoes to Australia was calculated from the mean distance to A29 among dingo sequences. The standard deviation of the distance to A29 was calculated by resampling with replacement (1,000 sets) of all dingo sequences (n = 211). The mutation rate of the analyzed region was estimated from the mean genetic distance between the dingo, dog, and wolf mtDNA types and the coyote types in the phylogenetic tree (Fig. 1 A) and the assumption of a divergence time between wolves and coyotes of 1 million years (Myr), which is based on a first appearance of wolves ≈700,000 yr ago and of coyotes ≈1 Myr ago (16, 17).

mtDNA was analyzed in 211 Australian dingoes sampled in all states of Australia (Table 1), 676 dogs from all continents, and 38 Eurasian wolves, and 290 bp were analyzed in 19 pre-European archaeological dog samples from Polynesia. The sequence variation among the dingoes was very restricted compared with that of dogs and wolves (Fig. 2, which is published as supporting information on the PNAS web site). Among the 211 dingoes there were 20 mtDNA types differing by at most two substitutions, whereas among 676 dogs there were 114 mtDNA types with a maximum difference of 16 substitutions between mtDNA types. Two of the dingo mtDNA types were identical to dog mtDNA types (din1 = A29; din19 = A9), whereas the other 18 were unique to dingoes.

It has been shown that in a phylogenetic tree of dog and wolf sequences, domestic dog mtDNA types form several separate branches distributed among the wolf mtDNA types, showing that the domestic dog originates from several maternal wolf lines (10, 11). In a phylogenetic tree of dogs, wolves, and dingoes, all the dingo sequences fall into the main clade (A) of dog sequences containing ≈70% of domestic dog mtDNA types (Fig. 1 A). A minimum-spanning network of clade A shows the dingo mtDNA types forming a single internal cluster around a central type, A29, found in both dingoes and dogs (Fig. 1B). A29, which was found in 53% of all dingoes, is surrounded by the other 12 (discounting indels) less frequent dingo types, differing by a single substitution, as well as by a number of dog mtDNA types. Among domestic dogs, A29 was found only in East Asia (East Siberia, in 2 of 18 individuals; Japan, 4 of 96; Indonesia, 2 of 7), New Guinea (1 of 2), and Arctic America (6 of 25) (11). Among dogs from the islands surrounding Australia, several mtDNA types other than A29 were found. In Island Southeast Asia (Indonesia, n = 7 dogs; Malaysia n = 2 dogs; the Philippines n = 1 dog), there were seven mtDNA types in addition to A29 (Fig. 1B). Among 19 samples of dogs from Polynesian pre-European archaeological sites, there were two mtDNA types, Arc 1 and Arc 2 (Fig. 2), both different from A29, on Cook Islands (Arc1, n = 1; Arc2, n = 1), New Zealand (Arc1, n = 3; Arc2, n = 10), and Hawaii (Arc1, n = 2; Arc2, n = 2). Interestingly, for the region sequenced for both ancient and contemporary samples, Arc2 is identical to mtDNA type A75, which was found only in two Indonesian samples, whereas Arc1 has a sequence found in a number of widely spread types (11).

The mutation rate for the analyzed region was estimated from the mean genetic distance between the dingo, dog, and wolf mtDNA types and the coyote types in the phylogenetic tree (Fig. 1 A) and the assumption of a divergence time between wolves and coyotes of 1 Myr, to 0.0651 substitutions site^–1 Myr^–1 (SD = 0.0027). The divergence time is based on paleontological evidence of a first appearance of wolves ≈700,000 yr ago and of coyotes ≈1 Myr ago (16, 17) and seems to be the most probable estimate; however, it has not been definitely established, and an earlier date of up to ≈2 Myr cannot be ruled out (17–19).

The star-like formation of the dingo mtDNA sequences in the minimum-spanning network (Fig. 1B) and the facts that more than half of the dingoes had the central type A29 and all dingo mtDNA sequences but A29 and A9 were unique to dingoes indicate that all dingo mtDNA types originate from A29. A9 was found in a single dingo individual (Table 1), and considering the relatively high degree of homoplasy in the data set, shown by a large homoplasy index for the phylogenetic tree (HI = 0.56), it is probable that this mtDNA type is the result of a parallel mutation. Assuming that A29 was the only founder type, the probable time for the introduction of dingoes to Australia was estimated at 4,600–5,400 yr ago from the mean distance to A29 among dingo sequences (0.190 substitutions; SD = 0.007) and the mutation rate, based on a wolf–coyote split 1 Myr ago. Taking into account the possibility of a split up to 2 Myr ago, the introduction would have happened some time 4,600–10,800 yr ago. The mean distance to A29 among dingo sequences differs considerably between Western Australia and the other parts of Australia (Table 1). This discrepancy is likely caused by random genetic drift in the Pilbara District population from which most Western Australian samples (25 of 29) were taken, and therefore the mean value for all dingo samples was used.

The position of the dingo mtDNA types in the middle of the dog mtDNA variation, A29 being found among both dogs and dingoes and forming an internal node in clade A surrounded by the other dingo mtDNA types as well as by dog mtDNA types, provides strong evidence that dingoes originated from domesticated dogs. It was shown before that domestic dogs in Eurasia, America, and Africa have a common origin (11, 20), and with this study it is shown that this applies to domesticated dogs on all continents.

The minimum-spanning network of phylogenetic clade A has a complicated pattern of dog haplotypes, caused by a large degree of homoplasy, which shows that the analyzed region is not sufficient for obtaining a phylogenetic resolution of dog mtDNA types. However, the dingo types are well grouped, and the very restricted sequence variation among dingoes, with more than half of the dingoes having mtDNA type A29 and all other mtDNA types forming a star-like formation around A29, and the fact that all other dingo types but A9 were unique to dingoes suggest that all dingo mtDNA types originate from A29. This indicates that the dingo population was formed either from very few dogs, theoretically as few as a single pregnant female, or from a group of dogs that had radically lost genetic variation through one or several severe bottlenecks on their way from the Asian mainland through Island Southeast Asia. However, the presence of a number of mtDNA types other than A29 on the islands surrounding Australia, both in modern samples and in samples from pre-European archaeological sites, indicates that there were several mtDNA types other than A29 present in the region and points at a single founding event. It also indicates that there have not been any substantial later introductions of dogs to Australia before the arrival of Europeans.

Today there is a large amount of hybridization between dingoes and domestic dogs in eastern Australia, but for this study the wild dingoes were sampled based on similarity in appearance to dingoes to exclude as far as possible feral dogs and dog × dingo hybrids, and the absence among the dingoes of any mtDNA types found in European dog breeds (11) indicates a low degree of hybrids among the sampled dingoes. However, because of the maternal mode of inheritance of mtDNA, hybridization between male dogs and female dingoes would not be identified by these mtDNA analyses. It can be noted that, similar to dingoes, the New Guinean dogs had the mtDNA type A29 and a unique type differing from A29 by one substitution. These so-called New Guinea singing dogs are feral and show some morphological and behavioral similarities to dingoes. A common origin and some gene flow between the two populations is therefore possible.

Among domestic dogs, A29 was found only among East Asian, Island Southeast Asian, and American dogs, and the mtDNA types radiating from A29 in the minimum-spanning network were found almost exclusively in East Asia (11), strongly indicating an East Asian rather than Indian origin for the dingo ancestor. The estimated time for the founding of the dingo population, ≈5,000 yr ago, fits relatively well with the archaeological record of the region, with the oldest finds of dingo being 3,500 years old and the earliest finds of dogs on nearby islands being 3,500-year-old remains on Timor (7). An East Asian ancestry ≈5,000 yr ago suggests that the dingoes may have arrived in connection with the expansion from south China into Island Southeast Asia of the Austronesian culture, which involved domestic dogs, pigs, and chicken. According to the current theories, the expansion started ≈6,000 yr ago from Taiwan via the Philippines to Indonesia, where it was split into a westward and an eastward direction and had by 4,000 yr ago reached Timor (7, 21).

In conclusion, this study of mtDNA sequence variation among dingoes provides a number of clues from which a detailed picture of the origin and history of the Australian dingo can be derived. The dingo originated from a population of East Asian dogs. Type A29 was one of several domestic dog mtDNA types brought into Island Southeast Asia, but only A29 reached Australia. The dingo population was probably founded from a small number of animals, as the last trickle of domestic dogs through a series of bottlenecks, or even by a single chance event and has since remained effectively isolated from other dog populations. The dingoes may have arrived in connection with the expansion, starting ≈6,000 yr ago, from south China into Island Southeast Asia of the Austronesian culture. By this time, domestic dogs had existed for several thousand years (4, 11), and the present semidomestic state of the dingo can probably be attributed to a long existence as a feral animal. After >3,500 years of isolation, the dingoes represent a unique isolate of early undifferentiated dogs.

This paper was submitted directly (Track II) to the PNAS office.

Abbreviation: Myr, million years.

Data deposition: The sequences reported in this paper have been deposited in the GenBank database (accession nos. AY660633–AY660656).

We thank David Jenkins (Australian Hydatid Control and Epidemiology Program, Canberra) and the Australian Dingo Conservation Association for assistance in obtaining dingo samples; Dr. Michael W. Fox (Humane Society of the United States, Washington, DC) for Indian dog samples; Dr. Melinda Allen (University of Auckland, Auckland, New Zealand), Dr. Foss Leach (Museum of New Zealand Te Papa Tongarewa, Wellington), the Archaeozoology Laboratory of the Museum of New Zealand, the Canterbury Museum, and the Otago Museum for access to archaeological samples; and Sonia Townsend and Ann Horsburgh for laboratory assistance. This work was supported by the Swedish Research Council, the Australian Research Council, the Native Dog Conservation Society, the New Zealand Foundation for Research Science and Technology, and the University of Auckland Research Committee. This article was approved for public release by the Los Alamos National Laboratory as technical information release LA-UR 03-5890.