Molecular Evolution of the Dog Family

Molecular genetic tools have been used to dissect the evolutionary relationships of the dog-like carnivores, revealing their place in the order Carnivora, the relationships of species within the family Canidae, and the genetic exchange that occurs among conspecific populations. High rates of gene flow among populations within some species, such as the coyote and gray wolf, have suppressed genetic divergence, and where these species hybridize, large hybrid zones have been formed. In fact, the phenotype of the endangered American red wolf may be strongly influenced by hybridization with coyotes and gray wolves. Hybridization and habitat fragmentation greatly complicate plans to conserve the genetic diversity of wild canids.
The dog family, Canidae, is a diverse group of 34 species ranging in size and proportion from squat, dachshund-like bushdogs to the long-legged maned wolf, a species sometimes called;a fox-on-stilts. This morphological diversity is matched by the diversity of their natural history: canids inhabit temperate and tropical forests, savanna, tundra and deserts throughout the world. Moreover, canids have a broader appetite than is commonly realized; most include a substantial proportion of vegetable and insect matter in their diet1. In the past, the evolutionary relationships of canids have been studied by morphological approaches, but parallel changes in several evolutionary lineages can make inferences uncertain.
The use of molecular and biochemical techniques to examine genetic differences among species provides an alternative way to investigate phylogenetic relationships. Such methods also have inherent biases, but systematists can search for groupings that are supported by a number of different approaches, and are thus more likely to be genuine phylogenetic relationships. Molecular genetic approaches have provided information about evolutionary divergence at a number of different levels, ranging from the relationships of the Canidae to other carnivore families, to relationships among populations within a single species. This information is discussed in this review.
Relationships of canids to other carnivore families
The order Carnivora includes the cat, hyena, bear, weasel, seal, mongoose, civet and dog families. All have ancient origins some 40-60 million years ago and thus their relationships can be studied by comparing the sequences at single-copy genes that have only a modest rate of sequence evolution2-4. The degree to which two single-copy DNA sequences have diverged can be estimated by the DTm which is the difference between the melting temperature (the point where 50% of DNA is double stranded) for a homologous duplex (i.e. both strands from the same species) and a heterologous duplex (with constituent strands from different species). the value is normalized for the final percentage of hybridization and designated DTmR (Ref. 4). A clustering phylogeny based on the DTmR between carnivore species shpows that extant species are closely related to each other (DTmR <4*C) but are only distantly related to species in other carnivore families[5] (DTmR>18*C). Assuming a constant rate of sequence evolution, the Canidae diverged from other carnivore families approximately 50-60 million years ago, near the time when canids first appeared in the fossil records6,7. Clearly, the Canidae diverged early in the evolution of carnivores, and one should be cautious about attempting to draw conclusions about carnivore gene structure and function from studies on canids alone.
Relationships of canids to each other
Patterns of evolution within the Canidae have been elucidated by use of protein electrophoresis to study allozyme variants and by comparison of G-banded metaphase chromosomes8-10. The differences between allele frequencies for a large number of loci are first used to calculate the genetic distance between pairs of species; from these genetic distances, clusters of species can be discerned8,11. Comparative analysis of chromosomes has also proved very useful because canids have a rich diversity of chromosome morphology ranging from species such as the red fox, which has a low diploid number of chromosomes (2n = 36) and all metacentric autosomes, to the gray wolf, which has a high diploid number (2n = 78) and all acrocentric autosomes (Table 1). The primitive canid karyotype has been reshuffled in different lineages, in a way that reveals the phylogenetic history of the group8-10. The evolutionary sequence of chromosomal rearrangements is deduced by differentially staining chromosomes and matching segments of similar banding patterns in different species9,10. The results of allozyme and chromosome analyses suggest several phylogenetic divisions within the Canidae: (1) the wolf-like canids, including domestic dogs, gray wolves, coyotes, and jackals; (2) the South American canids, including species of diverse morphology but common recent ancestry; (3) the red-fox-like canids of the Old and New World, including red foxes and kit foxes; and (4) monotypic genera -- species such as the bat-eared fox and raccoon dog -- that have a long, separate evolutionary history. The fossil record and genetic distances indicate that these divisions began about 7 -- 10 million years ago.
Large (12-30 kg)
Canis simensis Simien jackal Ethiopia 78
Canis lupus Gray wolf Holarctic 78
Canis latrans Coyote North America 78
Canis rufus Red wolf Southern US 78
Cuon alpinus Dhole Asia 78
Lycaon pictus African wild dog Subsaharan Africa 78
South American canids
Speothos venaticus Bushdog Northeast S. America 74
Lycalopex uetulus Hoary fox Northeast S. America 74
Cerdocyon thous Crab-eating fox Northeast S. America 74
Chrysocyon brachyurus Maned wolf Northeast S. America 76
Red fox-like canids
Vulpes aelox Kit fox Western US 50
Vulpes vulpes Red fox Old and New World 36
Vulpes chama Cape fox Southern Africa
Alopex lagopus Arctic fox Holarctic 50
Fennecus zerda Fennec fox Sahara 64
Other canids
Otocyon megalotis Bat-eared fox Subsaharan Africa 72
Urocyon cinereoargenteus Gray fox North America 66
Nycteruetes procyonoides Raccoon dog Japan, China 42b
a Diploid chromosome number.
b variable number of B-chromosomes present.
Relationship tree of carnivores based on differences in single-copy DNA sequences5 (DTmR).
Relationships of the wolf-like canids
The wolf-like canids are a closely related group of large carnivores whose chromosomes are stable in morphology and number (2n = 78). Because of the recent common ancestry of the members of this group, genes that have high rates of sequence substitution, such as those found in the vertebrate mitochondrial genome, can be used to resolve their phylogenetic relationships12. A phylogenetic analysis of 736 bp of the mitochondrial cytochrome b gene revealed a close kinship of gray wolves, dogs, coyotes and Simien jackals13-16 (Fig 4). As a group, these three taxa were distinct from the African wild dog and from the golden, side-striped and black-backed jackals. The gray wolf and coyote may have had a recent common North American ancestor about two million years ago17 whereas the Simien jackal, found only in a small area of the Ethiopian highlands, is possibly an evolutionary relic of a past African invasion of gray wolf-like ancestors. The Simien jackal is the most endangered canid18 and should be called a wolf rather than a jackal to reflect its evolutionary heritage.
An unexpected result of this research was the high sequence divergence (about 8%) that was found between two black-backed jackals in the same popuation, or a segment of the mitochondrial cytochrome b gene15. This was the largest divergence in mitochondrial DNA (mtDNA) then recorded within a single population that was Freely interbreeding. (as indicated by analysis of morphology and nuclear genes)19. The mtDNA sequences of these two genotypes evolved at significantly different rates and probably diverged before the speciation event giving rise to black-backed jackals. These findings emphasize the need for caution in the interpretation of phylogenies based on mtDNA; such gene trees are not necessarily species trees and may not accurately reflect phylogenetic affiliations or divergence time20.
The evolution of the domestic dog
The earliest remains of the domestic dog date from 10 to15 thousand years ago21; the diversity of these remains suggests multiple domestication events at different times and places. Dogs may be derived from several different ancestral gray wolf populations, and many dog breeds and wild wolf populations must be analysed in order to tease apart the genetic sources of the domestic dog gene pool. A limited mtDNA restriction fragment analysis of seven dog breeds and 26 gray wolf populations from different locations around the world has shown that the genotypes of dogs and wolves are either identical or differ by the loss or gain of only one or two restriction sites22. The domestic dog is an extremely close relative of the gray wolf, differing from it by at most 0.2% of mtDNA sequence15,22,23.
In comparrison, the gray wolf differs from its closest wild relative, the coyote, by about 4% of mitochondrial DNA sequence14. Therefore, the molecular genetic evidence does not support theories that domestic dogs arose from jackal ancestors24. Dogs are gray wolves, despite their diversity in size and proportion; the wide variation in their adult morphology probably results from simple changes in developmental rate and timing25.
A most parsimonious phylogenetic tree of gray wolf (Wl -- W12) and select coyote (Cl -- C24) genotypes based on data from restriction analyses14. The tree was rooted at the midpoint. Genotypes with asterisks indicate gray wolf genotypes that are identical, or very similar, to those of coyotes. The scale represents percentage sequence divergence.
Relationships of populations within species of wolf-like canids
Wolf-like canids can travel great distances and overcome sizeable topographic obstacles. Gray wolves, for example, have been observed to disperse over a thousand kilometers during their lifetimes26. Because dispersing wolves may establish territories and reproduce, gene flow can occur over much larger distances than is usual for terrestrial vertebrates27. A number of different subspecies of the gray wolf and the coyote have been described28; do molecular genetic analyses support the existence of these subspecies, and if so, how are subspecies related? Because the mitochondrial genome evolves so rapidly, its analysis has been an important source of clues about the differentiation of populations within species. Analysis of mtDNA variation in several hundred coyotes and gray wolves has shown little geographic subdivision of mtDNA genotypes22,29. Within each species, the same genotypes were present at widely spaced locations. There was no significant genetic difference among populations of coyotes, whereas wolves showed only a hint of genetic divergence between Alaskan and southern Canadian populations. Allozyme studies also showed low levels of differentiation among gray wolf populations30.
The phylogenetic tree of mtDNA genotypes can also reveal evidence of geographic subdivision (Fig. 5). In small vertebrates that have poor dispersal ability, the phylogenetic relationships of mitochondrial DNA genotypes from different populations often correspond to the physical distance between the populations or to the presence of geographic barriers31,32. The greater the geographic distance, the larger the genetic divergence. In gray wolves and coyotes, the relationship between genotypes did not reflect the geographic distance between localities. Closely-related coyote genotypes were found in regions as distant as California and distantly related genotypes were found at a single locality in southern California. This result supports the idea that gene flow is a force that homogenizes genetic variation, perhaps across large parts of the continent, but these findings also cast doubt on the validity of the dozen or more subspecies described for both species. The subspecies differences, which are based on pelage or skeletal morphology, may reflect inadequate sampling, rapid evolution of specific ecotypes through selection, or differences in food supply33. The molecular genetic evidence suggests that these phenotypic differences do not signify a long history of genetic isolation.
The population structure of Old World wolves differs from that of their relatives in North America. In crowded Europe, wolf populations are highly fragmented and small in size. Analysis of mtDNA in European wolves showed that, with one exception, each population had a single genotype not found elsewhere22. The genetic differences among the seven observed genotypes were small: just one or two restriction sites among the 95 that were sampled. However, the structured distribution of these genotypes suggested geographic subdivision and thus led to the concern that each population should be conserved and bred separately22. Hundreds of years ago, gray wolves ranged throughout Europe, as they do now in northern Canada, and probably showed little geographic subdivision. As available habitats for wolves decreased and populations became small, genotypes were fixed at random in the remaining populations, leaving a fractured genetic landscape. Because this landscape reflects the recent activities of humans, preserving each population separately through captive breeding amounts to a continuation of artificial selection on a grand scale and is not justified.
Gene flow within other canid species
Do other wolf-like canids show more geographic structure in their distribution of genotypes than wolves and coyotes? The African wild dog, a large wolf-like canid found in subsaharan Africa, is a good candidate, since the Rift Valley lakes may effectively interrupt gene flow between the eastern and southern populations16,18. Indeed, there seems to be no gene flow across this barrier, since eastern and southern African wild dogs do not share any mtDNA genotypes16. Moreover, the sequence divergence between the genotypes is substantial: about 1% of the sequence of the mitochondrial cytochrome b gene differs between the two genotype groups, a figure that is nearly an order of magnitude greater than the divergence between the most different genotypes within a population. Because the difference between populations was so much greater than that within each population, it was recommended that to preserve genetic diversity, east and south African wild dogs should not be interbred in captivity16.
Do the genotypes of small, less mobile canids have a geographic structure more like other small vertebrates, such as rodents, than that of their larger canid brethren? The diminutive kit fox, a species that lives in the arid lands of the American west, has a distribution that encircles the Rocky Mountains. Analysis of the mtDNA of this species showed two distinct genetic gradients. One was precipitous, and had developed between populations on either side of the Rocky Mountains34; the difference between these populations was nearly as great as between either population and the arctic fox, a species classified in a separate genus. The other gradient was among populations on the same side of the Rockies, and was more gradual. Neighbouring populations shared a greater number of genotypes, and these were more similar to each other than to those of distantly separated populations. Thus, the kit fox showed the two common patterns characteristic of smaller, genetically well-partitioned vertebrates: isolation by topographic barriers, and genetic differentiation increasing with distance.
Interspecific hybridization and the origin of the red wolf
Species, such as wolves and coyotes, that are highly mobile and can interbreed under some conditions, may form large hybrid zones. Several hundred years ago, coyotes were numerous only in the southern United States and wolves were common toward the north. Where wolves are abundant, they will exclude the much smaller coyote from their territories35. After the arrival of European settlers, agriculture and predator control programs caused wolf populations to dwindle, while the coyote, a remarkably flexible and opportunistic species, expanded its geographic range to areas north and east17. Today the coyote is found throughout most of North America. In eastern Canada, an area invaded b coyotes in the last 100 years, several genotypes identical or very similar to those found in coyotes were discovered in individuals phenotypically identified as gray wolves14 (genotypes with asterisks in Fig. 5). Wolves with these "coyote" genotypes increased in frequency toward the east, from 50% in Minnesota to 100% in Quebec. The hypothesis advanced to explain this pattern was that coyotes and wolves had hybridized in areas of eastern Canada where wolves were rare and coyotes common. The interspecific transfer of mtDNA was asymmetric; none of the coyotes sampled had wolf-like genotypes although coyote genotypes were common in gray wolves. Because mtDNA is maternally inherited without recombination, this result reflects a mating asymmetry: male wolves mate with female coyotes, and their offspring backcross to wolves. Either the reverse cross is rare, or the offspring of such backcrosses to coyotes do not reproduce. This mating asymmetry may indicate that the smaller male coyotes cannot inspire the larger female gray wolves to mate with them.
Theory predicts that older hybrid zones between wolves and coyotes may be much larger than that in eastern Canada, and may be up to several thousand kilometers in width15,36. Accordingly, attention has been focused on a potentially older and more extensive hybrid zone in the southern United States. The zone includes populations of three wolf-like canids: the red wolf, gray wolf and coyote. The red wolf is intermediate in size between coyote and gray wolves and can potentially hybridize with both species. It is also an endangered species that became extinct in the wild about 1975, and descendants of the last populations were used to found a successful captive breeding and reintroduction program. If the red wolf were a distinct species ancestral to wolves and coyotes37, there should be unique mtDNA genotypes that define a separate species clade15, a pattern previously found in wolf-like canids13-16.
However, captive red wolves had a genotype that was indistinguishable by restriction site analysis from those found in coyotes from Louisiana. Because hybridization was thought to occur between the two species as the red wolf became rare, the presence of the coyote-derived genotypes in captive red wolves could represent an accident of sampling and not be representative of the ancestral population. Subsequently, an additional mtDNA analysis of 77 samples obtained in about 1975 from areas inhabited by the last wild red wolves showed that all had either a coyote or gray wolf genotype15.
Conceivably, hybridization between gray wolves and coyotes alone could explain the intermediate morphology of red wolves. To test this hypothesis, DNA was isolated from six museum skins of red wolves obtained from Five states in about 1910, a time before hybridization of red wolves and coyotes was thought to be common. Phylogenetic analysis of 398 bp of the cytochrome b gene showed that red wolves at that time did not have a distinct genotype; all six had genotypes classified with gray wolves or coyotes, a result consistent with a hybrid origin for the species15. Although more research needs to be done, the implication of this result is troubling for the US Endangered Species Act because a policy on hybrids has not been formulated. In some situations we may wish to protect hybrids, such as the red wolf, because they are unique. Elsewhere, in Minnesota for example, hybridization may be undesirable because it jeopardizes the genetic integrity of the gray wolf, a threatened species. Similarly, in Italy, hybridization with domestic dogs may be changing the character of gray wolves that enter small towns to feed because their natural prey has been depleted. Even the highly endangered Simien jackal is threatened with hybridization by feral domestic dogs. Molecular genetic analyses offer a powerful means to determine if hybridization is changing the composition of these endangered populations.
Future research on the population genetics of canids should focus on the analysis of polymorphic nuclear genes to complement the mtDNA data. However, nuclear DNA domains that evolve as fast as highly variable mtDNA regions have yet to be identified, and may not exist. Hypervariable simple sequence repeat loci38 may prove useful; these loci are abundant in the nuclear genome and evolve through loss or gain of repeat units rather than sequence substitutions. Analysis of simple sequence repeats will not provide the detailed picture of the succession of historical changes revealed by sequence data but may furnish estimates of gene flow and hybridization among closely related canid populations.

Acknowledgements
I appreciate comments on the manuscript by D. Girman, K. Koepfli, P. Sunnucks and B. van Valkenburgh, and the support of the US Fish and Wildlife Service and the NSF.

References
1 van Valkenburgh, B. (1988) Paleobiology 14, 156-173
2 Kohne, D.E., Chiscon, S.A. and Hoyer, B.H. (1972) J. Hum. Evol. 1, 627 -- 644
3 Benveniste, R.E. (1976) Nature 261, 101 -- 108
4 Benveniste, R.E. (1985) in Molecular Evolutionary Genetics (MacIntyre, R.E., ed.), pp. 359-417, Plenum Press
5 Wayne, R.K., Benveniste, R.E. and O'Brien, S.J. (1989) in Carnivore Behavior, Ecology and Evolution (Gittleman, J.L., ed.), pp. 465-494, Cornell University Press
6 Flynn, J.M. and Galiano, H. (1982) Amer. Mus. Novit. 2632, 1 -- 16
7 Wayne, R.K., van Valkenburgh, B. and O'Brien, S.J. (1991) Mol. Biol. Evol. 8, 297 -- 319
8 Wayne, R.K, and O'Brien, S.J. (1987) Syst. Zool. 36, 339-355
9 Wayne, R.K., Nash, W.G. and O'Brien, S.J. (1987) Cytogenet. Cell Genet. 44, 123-133
10 Wayne, R.K., Nash, W.G. and O'Brien, S.J. (1987) Cytogenet. Cell Genet. 44, 134 -- 141
11 Nei, M. (1987) Molecular Evolutionary Genetics, Columbia University Press
12 Brown, W.M. (1986) in Molecular Evolutionary Biology (MacIntyre, R.E., ed.), pp. 95-128, Cornell University Press
13 Wayne, R.K. el a(. (1990) Proc. Natl Bead. Sci. USA 87, 1772 -- 1776
14 Lehman, N. el al. (1991) Evolution 45, 104 -- 119
15 Wayne, R.K. and Jenks, S.M. (1991) Nature 351, 565 -- 568
16 Girman, D.J. et al. J. Heredity (in press)
17 Nowak, R.M. (1979) North American Quaternary Canis, Monogr. Mus. Nat. Hist. Univ. Kansas
18 Ginsberg, J.R. and Macdonald, D.W. (1990) Faxes, Wolves, Jackals and Dogs: An Action Plan for the Conservation of Canids, International Union for Conservation of Nature and Natural Resources, Gland, Switzerland
19 Wayne, R.K., van Valkenburgh, B., Fuller, T.K. and Kat, P.W. (1990) in Molecular Evolution (UCLA
Symposium on Cellular Biology, Nehru Series) (Vol. 122) (Clegg, M. and O'Brien, S.J., eds), pp. 161-170
20 Pamilo, P, and Nei, M. (1988) Mol. Biol. Ecol. 5, 568-583
21 Olsen, S.J. (1985) Origins of the Domestic Dog, The Fossil --Record, University of Arizona Press
22 Wayne, R.K., Lehman, N., Allard, M.W. an Honeycutt, R.L. (1992) Conserv. Biol. 6, 559 -- 56923
Templeton, A.R. (1989) in Speciation and its Consequences (Otte, D. and Endler, J.A., eds), pp. 3-27,
Sinauer Associates
24 Lorentz, K.Z. (1954) Man Meets Dog, Methuen Press
25 Wayne, R.K. (1986) Evolution 40, 243-261
26 Mech, D.L. (1987) in Mammalian Dispersal Patterns (Chepko-Sade, D.B. and Tang Halpin, Z., eds), pp. 55 -- 74, University of Chicago Press
27 Chepko-Sade, D.B. et al. (1987) in Mammalian Dispersal Patterns (Chepko-Sade, D.B. and Tang Halpin, Z., eds), pp. 287-322, University of Chicago Press
28 Hall, E.R. (1981) The Mammals of North America, Wiley
29 Lehman, N. and Wayne, R.K. (1991) Genetics 128, 405-416
30 Kennedy, P.K., Kennedy, M.L., Clarkson, P.L. and Liepins, I.S. (1991) Can.J. Zool. 69, 1183 -- 1188
31 Avise, J.C. and Ball, R.M. (l990) in Oxford Surveys in Evolutionary Biology (Vol. 7) (Futuyma, D. and
Antonovics, J., eds), pp. 45-67, Oxford University Press 32 Avise, J.C. et al. (1987) Annu. Rev. Ecol. Syst. 18, 489-522
33 Thurber, J.M. and Peterson, R.O. (1991) J. Mammal.72, 750 -- 755
34 Mecure, A., Rails, K., Koepfli, K.P. and Wayne, R.K. Evolution (in press)
35 Carbyn, L.N. (1982) Can. Field Nat. 96, 339-341
36 Barton, N.H. and Hewitt, G.M. (1989) Nature 141, 497 -- 503
37 Nowak, R.M. (1992) Co38O strander, E.A., Jong, P.M., Rine, J. and Duyk, G. (1992) Proc. Natl Acad. Sci. USrf 8 3419-3423
R.K. WAYNE IS IN THE INSTITUTE OF ZOOLOGY, ZOOLOGICAL SOCIETY OF LONDON, REGENTS PARK LONDON, UK NW1 4RY.

Edited by CCT for The Guild