, 1687 (1997);
et al. Carles Vilà,
Multiple and Ancient Origins of the Domestic Dog
This copy is for your personal, non-commercial use only.
. clicking herecolleagues, clients, or customers by
, you can order high-quality copies for your
If you wish to distribute this article to others
. herefollowing the guidelines
can be obtained by
Permission to republish or repurpose articles or portions of articles
(this information is current as of May 6, 2010 ):
The following resources related to this article are available online at www.sciencemag.org
version of this article at:
including high-resolution figures, can be found in the online
Updated information and services,
27 articles hosted by HighWire Press; see:
This article has been
This article appears in the following
registered trademark of AAAS.
is aScience1997 by the American Association for the Advancement of Science; all rights reserved. The title
CopyrightAmerican Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005.
(print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by theScience
on May 6, 2010
dependent lysis, because the calcium chelator
EGTA, which blocks the perforin pathway, also in-
hibits up-regulation of FasL, as required in our stud-
ies with primary human T cells.
11. Tumor necrosis factor–? (TNF-?) induces lysis of a
murine fibrosarcoma cell line independently of per-
forin or Fas-FasL (19). This cytokine did not contrib-
ute to lysis of the CTLs used in this study because
addition of blocking antibody to TNF-? did not inhibit
12. G. Berke, Annu. Rev. Immunol. 12, 735 (1994).
13. The MHC class I–restricted CD8?T cell lines that
specifically recognize a defined influenza virus matrix
protein presented by human leukocyte antigen (HLA)
A2 (CD8.FP1, CD8.FP2) were generated from the
blood of healthy HLA A.2?donors by stimulation of
PBMCs with the peptide (10 ?g/ml) (gift of Cytel, San
Diego, CA) (20). Lines were maintained by weekly
stimulation with the peptide, with irradiated autolo-
gous PBMCs as feeder cells. Before the experi-
ments, CD8?cells were enriched by immunomag-
14. S. H. E. Kaufmann, Immunol. Today 9, 168 (1988).
15. K. Benihoud, D. Bonardelle, P. Bobe, N. Kiger, Eur.
J. Immunol. 27, 415 (1997).
16. T. S. Griffith, T. Brunner, S. M. Fletcher, D. R. Green,
T. A. Ferguson, Science 270, 1189 (1995); H. Yagita
et al., Immunol. Rev. 146, 223 (1995).
17. P. Laochumroonvorapong et al., Infect. Immun. 65,
127 (1997); A. Cooper, C. D’Souza, A. A. Frank, I.
Orme, ibid., p. 1317; R. J. Mazzaccaro, J. Flynn, B.
R. Bloom, unpublished results.
18. S. Pena et al., J. Immunol. 158, 2680 (1997); M.
Anderson et al., EMBO J. 14, 1615 (1995); Y. Mi-
yakawa et al., Infect. Immun. 64, 926 (1996).
19. R. K. Lee, J. Spielman, D. Y. Zhao, K. J. Olsen, E. R.
Podack, J. Immunol. 157, 1919 (1996); M. Y. Braun,
B. Lowin, L. French, H. Acha-Orbea, J. Tschopp, J.
Exp. Med. 183, 657 (1996).
20. A. J. McMichael, F. M. Gotch, J. Santos-Aguado, J.
L. Strominger, Proc. Natl. Acad. Sci. U.S.A. 85,
21. N. Kayagaki et al., J. Exp. Med. 182, 1777 (1995).
J. Klebanoff, ibid. 184, 429 (1996).
23. P. L. Coleman and G. D. Green, Methods Enzymol.
80, 408 (1981).
24. CD1?macrophages, which were infected with live M.
tuberculosis, were labeled with 100 ?Ci of51Cr (ICN,
Costa Mesa, CA) for 1 hour and plated in a 96-well
V-bottom plate at a final concentration of 4000 targets
per 100 ?l. Appropriate samples were incubated with
before addition of the T cells. After a 9-hour incubation,
target cell lysis was calculated by quantifying51Cr re-
lease in a gamma-counter. The data are given as per-
cent specific lysis, calculated as [(cpm release from ex-
perimental ? cpm spontaneous release)/(maximal re-
lease ? cpm spontaneous release) ? 100]. The spon-
T cells was ?15%.
25. CD1?macrophages were pulsed with soluble M.
tuberculosis extract (5 ?g/ml) overnight, detached
with 1 mM EDTA, and labeled with 100 ?Ci of51Cr
for 1 hour. For inhibition of the interaction between
FasL and Fas, the assay was done in the presence
of blocking antibodies to FasL (21) (Pharmingen,
San Diego, CA) or Fas (22) (Immunotech, West-
brook). Degranulation of the cytotoxic granules
was induced by initial treatment of T cells with 25
mM Sr2?(Aldrich, Milwaukee, WI) for 18 hours.
51Cr release was determined after a 4-hour incu-
bation. Expression of Fas on the target cells was
not affected by infection with M. tuberculosis, as
determined by flow cytometry.
26. CTLs (5 ? 105) were incubated in the presence of 25
mM Sr2?for 10 hours in a final volume of 1.5 ml. The
amount of N?-CBZ-L-lysine thiobenzyl (BLT)–ester-
ase in the supernatant was determined by the BLT-
esterase assay (23). The supernatants (20 ?l) were
coincubated with 35 ?l of 1 mM BLT (Sigma), 35 ?l
of 1 mM 5,5?-dithio-bis-(2-nitrobenzoic acid) (Sig-
ma), and 10 ?l of a 0.1% Triton X-100 (Sigma) solu-
tion. After a 30-min incubation at 37°C, the absor-
bance at 405 nm was determined.
27. P. Chomczynski and N. Sacchi, Anal. Biochem. 162,
28. FasL primer and probe sequences were designed as
follows: FasL 5?: CAGCTCTTCCACCTGCAGAAGG;
FasL 3?: AGATTCCTCAAAATTGATCAGAGAGAG;
FasL probe: GAGCCGAGGAGTGTGGCCCATTTA-
ACAGGG. PCR was performed with AmpliTaq poly-
merase (0.5 U/ml; Perkin-Elmer–Cetus) plus 2.5 mM
MgCl2for 25 s of denaturation at 94°C followed by
45 s of annealing and extension at 65°C for 35 con-
secutive cycles. PCR products were resolved by aga-
rose gel electrophoresis and transferred to nylon
membranes (Amersham, Arlington Heights, IL). The
membranes were then probed with an internal oligo-
nucleotide probe labeled at the 5? end with ?-32P and
polynucleotide kinase (Boehringer Mannheim, Ger-
many) and visualized by autoradiography.
29. E. P. Miskovsky et al., J. Immunol. 153, 2787 (1994).
30. We are indebted to M. Horwitz for use of his P3 labo-
ratory and P. Sieling for continuous support and helpful
sches Krebsforschungszentrum, Heidelberg (S.S.);
NIH (R.L.M., S.A.P., B.R.B., M.B.B.); the Arthritis
Foundation (S.A.P.); the Howard Hughes Medical In-
stitute (B.R.B.); the Swiss Foundation for Grants in
Medicine and Biology (J.-P.R); and the United Na-
tions Development Programme–World Bank–World
Health Organization Special Program for Research and
Training in Tropical Diseases (IMMLEP) and the Derma-
tologic Research Foundation of California.
30 January 1997; accepted 14 April 1997
Multiple and Ancient Origins of the
Carles Vila `, Peter Savolainen, Jesu ´s E. Maldonado,
Isabel R. Amorim, John E. Rice, Rodney L. Honeycutt,
Keith A. Crandall, Joakim Lundeberg, Robert K. Wayne*
Mitochondrial DNA control region sequences were analyzed from 162 wolves at 27
localities worldwide and from 140 domestic dogs representing 67 breeds. Sequences
that wolves were the ancestors of dogs. Most dog sequences belonged to a divergent
monophyletic clade sharing no sequences with wolves. The sequence divergence within
this clade suggested that dogs originated more than 100,000 years before the present.
Associations of dog haplotypes with other wolf lineages indicated episodes of admixture
between wolves and dogs. Repeated genetic exchange between dog and wolf popu-
lations may have been an important source of variation for artificial selection.
The archaeological record cannot resolve
whether domestic dogs originated from a
single wolf population or arose from multi-
ple populations at different times (1, 2).
However, circumstantial evidence suggests
that dogs may have diverse origins (3). Dur-
ing most of the late Pleistocene, humans
and wolves coexisted over a wide geograph-
ic area (1), providing ample opportunity for
independent domestication events and
wolves and dogs. The extreme phenotypic
diversity of dogs, even during the early
stages of domestication (1, 3, 4), also sug-
gests a varied genetic heritage. Conse-
quently, the genetic diversity of dogs may
have been enriched by multiple founding
events, possibly followed by occasional in-
terbreeding with wild wolf populations.
We sequenced portions of the mito-
chondrial DNA of wolves and domestic
dogs. Initially, 261 base pairs (bp) of the
left domain of the mitochondrial control
region (5) were sequenced from 140 dogs
representing 67 breeds and five cross-
breeds and 162 wolves representing 27
Asia, and North America (Fig. 1) (6).
Because all wild species of the genus Canis
can interbreed (7) and thus are potential
ancestors of the domestic dog, five coyotes
(Canis latrans) and two golden, two black-
backed, and eight Simien jackals (C. au-
reus, C. mesomelas, and C. simensis, re-
spectively) were also sequenced.
The control region of wolves and dogs
was highly polymorphic (Fig. 1). We
found 27 wolf haplotypes that differed on
average by 5.31 ? 0.11 (?SE) substitu-
tions (2.10 ? 0.04%), with a maximum of
10 substitutions (3.95%). The distribution
of wolf haplotypes demonstrated geo-
graphic specificity, with most localities
containing haplotypes unique to a partic-
ular region (Fig. 1). Four haplotypes (W2,
W7, W14, and W22) had a widespread
distribution. In dogs, 26 haplotypes were
found. Only haplotype D6 also occurred in
C. Vila `, J. E. Maldonado, I. R. Amorim, R. K. Wayne,
Department of Biology, University of California, Los An-
geles, CA 90095–1606, USA.
P. Savolainen and J. Lundeberg, Department of Bio-
J. E. Rice and R. L. Honeycutt, Faculty of Genetics and
Department of Wildlife and Fisheries Sciences, Texas
A&M University, College Station, TX 77843, USA.
K. A. Crandall, Department of Zoology and M. L. Bean
Museum, Brigham Young University, Provo, UT 84602,
*To whom correspondence should be addressed. E-mail:
www.sciencemag.org?SCIENCE?VOL. 276?13 JUNE 1997
on May 6, 2010
some gray wolves from western Russia and
among dogs was similar to that found
among wolves. Dog haplotypes differed by
an average of 5.30 ? 0.17 substitutions
(2.06 ? 0.07%), with a maximum diver-
gence of 12 substitutions (4.67%). Mito-
chondrial haplotype diversity in dogs
could not be partitioned according to
breeds. For example, in eight German
shepherds examined, five distinct se-
quences were found, and in six golden
retrievers, four sequences were detected.
Moreover, many breeds shared sequences
with other breeds. For instance, dog hap-
lotypes D4, D3, D5, and D1 were found in
14, 14, 9, and 7 breeds, respectively. No
dog sequence differed from any wolf se-
quence by more than 12 substitutions,
whereas dogs differed from coyotes and
jackals by at least 20 substitutions and two
insertions. These results clearly support
wolf ancestry for dogs. However, because
mitochondrial DNA is maternally inherit-
ed, interbreeding between female dogs and
male coyotes or jackals would not be de-
tected. More limited studies of nuclear
markers support the conclusion that the
wolf was the ancestor of the domestic dog
Several methods of phylogenetic anal-
ysis, including maximum likelihood (9),
spanning networks (11), and statistical
parsimony (12), were used to investigate
relationships among sequences. All analy-
ses supported a grouping of dog haplotypes
into four distinct clades, although the to-
pology within and among clades differed
among trees (13). As exemplified by the
neighbor-joining analysis (Fig. 2A), three
of the four monophyletic clades defined a
larger clade containing all but three dog
261 bp of control region sequence from wolves
same sequence as D4 except for an insertion of a
67-bp tandem repeat. The numerals I, II, III, and IV
indicate assignments to the four clades of dog
sequences. Wolf localities: Bulgaria (n ? 1, W7);
Croatia (n ? 5, W2); Estonia (n ? 1, W10); France
(n ? 2, W4); Finland (n ? 2, W10); Greece (n ? 7;
W2, W5, W8, and W9); Italy (n ? 12, W4); Poland
(n ? 1, W3); Portugal (n ? 19; W1 and W2);
Romania (n ? 4; W5 and W6); Russia (n ? 3; W6,
W10, and W26); Spain (n ? 46; W1 and W3);
W18); China (n ? 3; W14, W19, and W27); India
(n ? 1, W12); Iran (n ? 6; W16 and W17); Israel
(n ? 16, W11); Saudi Arabia (n ? 7; W7, W12,
W13, W14, and W15); Turkey (n ? 2, W2); Alaska
(n ? 3, W20); Alberta (n ? 1, W22); Labrador (n ?
3, W22); Mexico (n ? 5, W25); Montana (n ? 1,
W22); Northwest Territories (n ? 3, W22); and
Yukon (n ? 3; W21, W23, and W24). Dog breeds:
basenji (n ? 1, D2); basset (n ? 1, D6); boxer (n ?
1, D4); Norwegian buhund (n ? 1, D1); bulldog
(n ? 1, D6); Chinese crested (n ? 2; D2 and D25);
chow chow (n ? 3; D1, D2, and D3); collie (n ? 1,
D1); border collie (n ? 3; D1 and D5); wirehaired
dachshund (n ? 3; D5 and D10); Australian dingo
(n ? 4, D18); grey Norwegian elkhound (n ? 9; D3
and D8); Eskimo dog (n ? 1, D23); German shep-
herd (n ? 8; D4, D5, D6, D7, and D19); greyhound
(n ? 1, D9); groenendael (n ? 1, D6); Mexican
hairless (n ? 6; D3, D6, D21, and D26); hamilton-
sto ¨vare (n ? 1, D5); Afghanistan hound (n ? 3,
D6); Alaskan husky (n ? 2; D4 and D7); Siberian
husky (n ? 3; D3, D7, and D18); ja ¨mthund (n ? 3;
D7 and D8); keeshond (n ? 1, D5); kuvasz (n ? 1,
D4); Leonberger (n ? 2; D1 and D4); Norwegian lundehund (n ? 1, D16); Mareema (n ? 1, D?6);
Pyrenean mastiff (n ? 1, D11); Newfoundland (n ? 1, D4); otter hound (n ? 1, D6); papillon (n ? 2; D3
and D4); poodle (n ? 1, D3); toy poodle (n ? 1, D6); pug (n ? 1, D26); Chesapeake Bay retriever (n ?
1, D13); flat-coated retriever (n ? 3; D4 and D10); golden retriever (n ? 6; D4, D6, D15, and D24);
labrador retriever (n ? 6; D4 and D12); Rhodesian ridgeback (n ? 1, D26); rottweiler (n ? 2, D3);
3; D4 and D7); miniature schnauzer (n ? 1, D9); English setter (n ? 4; D3 and D5); Irish setter (n ? 3; D1
and D9); New Guinea singing dog (n ? 2, D18); shar (n ? 1, D26); Icelandic sheepdog (n ? 1, D3); Old
English sheepdog (n ? 1, D5); shiba inu (n ? 1, D20); Cavalier King Charles spaniel (n ? 1, D17); Irish
water spaniel (n ? 1, D6); springer spaniel (n ? 1, D3); Tibetan spaniel (n ? 1, D6); spitz (n ? 1, D22);
Japanese spitz (n ? 1, D3); airedale terrier (n ? 1, D7); border terrier (n ? 2, D3); fox terrier (n ? 2; D3
and D14); Norfolk terrier (n ? 2, D4); West Highland terrier (n ? 2, D7); Tibetan terrier (n ? 2; D2 and D9);
wachtelhund (n ? 1, D5); whippet (n ? 1, D3); Irish wolfhound (n ? 2, D11); and crossbreeds (n ? 5; D1,
D3, D4, D5, and D18).
Fig. 2. (A) Neighbor-joining tree of wolf and dog
haplotypes (D13 excluded; see Fig. 1) based on
261 bp of control region sequence (17). (B)
Neighbor-joining tree of 8 wolf and 15 dog ge-
notypes based on 1030 bp of control region
sequence. The suffixes a, b, and c after the hap-
lotype labels were used to distinguish identical
261-bp sequences that have different 1030-bp
sequences. Bootstrap support is indicated at
nodes if found in more than 50% of 10,000 boot-
SCIENCE?VOL. 276?13 JUNE 1997?www.sciencemag.org
on May 6, 2010
haplotypes and a subset of wolf haplotypes
(W4 and W5). Clade I included 19 of the
26 dog haplotypes. This group contained
representatives of many common breeds as
well as ancient breeds such as the dingo,
New Guinea singing dog, African basenji,
and greyhound (14). Clade II included dog
haplotype D8, from two Scandinavian
breeds (elkhound and ja ¨mthund), and was
closely related to two wolf haplotypes
found in Italy, France, Romania, and
Greece (W4 and W5). Clade III contained
three dog haplotypes (D7, D19, and D21)
found in a variety of breeds such as the
German shepherd, Siberian husky, and
Mexican hairless. Finally, clade IV con-
tained three haplotypes (D6, D10, and
D24) that were identical or very similar to
a wolf haplotype (W6) found in Romania
and western Russia, which suggests recent
hybridization between dogs and wolves.
Many breeds contained representatives of
more than one dog haplotype grouping
Because the overall bootstrap support
for many of the internodes in Fig. 2A was
low, 1030 bp of the control region were
sequenced for 24 canids, including repre-
sentatives of the four dog clades (15).
Although the association of clades was
different, the analyses of the longer se-
quences provided stronger support for the
four monophyletic groupings of dog hap-
lotypes (Fig. 2B) (13). A Wilcoxon
signed-rank test was used to assess the
monophyly of dog clades (16). Monophyly
of all dog haplotypes can be rejected, and
monophyly of clades I, II, and III is mar-
ginally rejected (P ? 0.0004 and P ?
0.053, respectively). In both trees, dog
haplotype clades II and IV are most closely
related to wolf sequences from eastern Eu-
rope (Greece, Italy, Romania, and western
The coyote and wolf have a sequence
divergence of 0.075 ? 0.002 (17) and
diverged about one million years ago, as
estimated from the fossil record (18). Con-
sequently, because the sequence diver-
gence between the most different geno-
types in clade I (the most diverse group of
dog sequences) is no more than 0.010, this
implies that dogs could have originated as
much as 135,000 years ago (19). Although
such estimates may be inflated by unob-
served multiple substitutions at hypervari-
able sites (20), the sequence divergence
within clade I clearly implies an origin
more ancient than the 14,000 years before
the present suggested by the archaeologi-
cal record (21). Nevertheless, bones of
wolves have been found in association
with those of hominids from as early as the
middle Pleistocene, up to 400,000 years
ago (1, 22). The ancient dates for domes-
tication based on the control region se-
quences cannot be explained by the reten-
tion of ancestral wolf lineages, because
clade I is exclusively monophyletic with
respect to dog sequences and thus the
separation between dogs and wolves has
been long enough for coalescence to have
occurred. To explain the discrepancy in
dates, we hypothesize that early domestic
dogs may not have been morphologically
distinct from their wild relatives. Con-
ceivably, the change around 10,000 to
15,000 years ago from nomadic hunter-
gatherer societies to more sedentary agri-
cultural population centers may have im-
posed new selective regimes on dogs that
resulted in marked phenotypic divergence
from wild wolves (23).
Although individual breeds show uni-
formity with respect to behavior and mor-
phology, most breeds show evidence of a
genetically diverse heritage because they
contain different haplotypes. Moreover,
dog sequences cluster with different groups
of wolf haplotypes. Therefore, after the
origin of dogs from a wolf ancestor, dogs
and wolves may have continued to ex-
change genes. Backcrossing events could
have provided part of the raw material for
artificial selection and for the extraordi-
nary degree of phenotypic diversity in the
domestic dog. Domestic species of plants
and animals whose wild progenitors are
extinct cannot be enriched through peri-
odic interbreeding, and change under ar-
tificial selection may be more limited.
Consequently, the preservation of wild
progenitors may be a critical issue in the
continued evolution of domestic plants
REFERENCES AND NOTES
1. J. Clutton-Brock, in The Domestic Dog, Its Evolution,
Behaviour and Interactions with People, J. Serpell,
Ed. (Cambridge Univ. Press, Cambridge, 1995), pp.
2. S. J. Olsen and J. W. Olsen, Science 197, 533
3. J. Clutton-Brock, A Natural History of Domesticated
Mammals (Cambridge Univ. Press, British Museum
of Natural History, Cambridge, 1987).
4. F. E. Zeuner, A History of Domesticated Animals
(Harper & Row, New York, 1963); H. Epstein, The
Origin of the Domestic Animals of Africa (Africana
Publishing, New York, 1971); I. L. Brisbin Jr., Am.
Kennel Gaz. January 1976, p. 23.
5. C.Saccone,M.Attimonelli,E.Sbisa `,J.Mol.Evol.26,
6. DNA from blood, tissue, or hair was isolated by cell
lysis followed by organic solvent purification. A
261-bp control region fragment was sequenced
from a slightly larger (394 bp) fragment amplified
with primers L15910 and H16498 [R. DeSalle, A. K.
Williams, M. George, Methods Enzymol. 224, 176
(1993)]. See polymerase chain reaction and se-
quencing conditions and protocols in J. E. Maldo-
nado, F. O. Davila, B. S. Stewart, E. Geffen, R. K.
Wayne, Mar. Mamm. Sci. 11, 46 (1995).
7. A. P. Gray, Mammalian Hybrids, a Check-list with
Bibliography (Farnham Royal, Bucks, Common-
wealth Agricultural Bureaux, UK, 1954).
8. D. Gottelli et al., Mol. Ecol. 3, 301 (1994); J. Garcı´a-
Moreno, M. D. Matocq, M. S. Roy, E. Geffen, R. K.
Wayne, Conserv. Biol. 10, 376 (1996); P.W. Hedrick,
P. S. Miller, E. Geffen, R. K. Wayne, Zoo Biol. 16, 47
9. K. Strimmer and A. von Haessler, Mol. Biol. Evol. 13,
10. D. L. Swofford, PAUP: Phylogenetic Analysis Using
Parsimony, Version 3.0 (Illinois Natural History Sur-
vey, Champaign, IL, 1990).
11. L. Excoffier, P. E. Smouse, J. M. Quattro, Genetics
131, 479 (1992).
12. A. R. Templeton, K. A. Crandall, C. F. Sing, ibid. 132,
13. The coyote was used as an outgroup on the basis
of morphologic and mitochondrial DNA sequence
analyses [R. M. Nowak, North American Quaterna-
ry Canis (Monograph of the Museum of Natural
History, University of Kansas, N 6, Lawrence, KS,
1979); R. K. Wayne, Trends Genet. 6, 218 (1993);
C. Vila ` et al., unpublished data]. Trees retaining the
four indicated clades resulted from neighbor-join-
ing searches with Tamura-Nei, Kimura 2–parame-
ter, Kimura 3–parameter, and HKY85 models of
evolution; parsimony searches with equal, 2:1, and
10:1 weights of transversions to transitions; and
maximum likelihood analyses with the HKY85
model of evolution, taking into account differences
in nucleotide frequencies and transition and trans-
14. R. Coppinger and R. Schneider, in The Domestic
Dog, Its Evolution, Behaviour and Interactions with
People, J. Serpell, Ed. (Cambridge Univ. Press,
Cambridge, 1995), pp. 21–47.
15. We amplified the entire canid control region with
primers in (6) and the following primers: L16462
(5’-CATACTAACGTGGGGGTTAC-3’), H222 (5?-A-
CAAACCCCCCTTACC-3?), and H652 (5?-AAG-
GCTAGGACCAAACCT-3?). The primer numbering
refers to their approximate location in the human
mitochondrial genome [S. Anderson et al., Nature
290, 457 (1981)]. The 1030-bp sequence that we
report includes the entire control region except for a
hypervariable region close to the 3? end that encom-
passes an imperfect 20-bp repeat having a variable
number of units [A. R. Hoelzel, J. V. Lopez, G. A.
Dover, S. J. O’Brien, J. Mol. Evol. 39, 191 (1994)].
16. A. R. Templeton, Evolution 37, 221 (1983).
17. Sequence divergence values and trees were based
on a gamma correction [K. Tamura and M. Nei, Mol.
Biol. Evol. 10, 513 (1993)] with the empirically deter-
mined parameter a ? 0.18 for the 261-bp sequenc-
es and a ? 0.22 for the 1030-bp sequences [S.
Kumar, K. Tamura, M. Nei, MEGA: Molecular Evolu-
tionary Genetic Analysis, Version 1.01 (Pennsylvania
State University, University Park, PA, 1993)].
Biol. Evol. 8, 297 (1991).
19. A relative rate test indicated that substitution rates
were similar for all wolf and dog sequences [F. Ta-
jima, Genetics 135, 599 (1993)].
20. T. J. Parsons et al., Nature Genet. 15, 363 (1997).
21. G. Nobis, Umshau 19, 610 (1979); S. J. Olsen, Ori-
gins of the Domestic Dog (University of Arizona
Press, Tucson, AZ, 1985).
22. M. E. Thurston, The Lost History of the Canine Race.
Our 15,000-Year Love Affair with Dogs (Andrews &
McMeel, Kansas City, KS, 1996).
23. D. F. Morey, Am. Sci. 82, 336 (1994).
24. We thank P. Taberlet for providing wolf sequences
from France and Poland and J. Castroviejo, Grupo
Lobo-Portugal, E. Geffen, and many others for their
help in obtaining samples. C.V. was supported by a
postdoctoral fellowship from the Spanish Ministerio
de Educacio ´n y Ciencia, and I.R.A. was supported
by a fellowship from Junta Nacional de Investigas˜ ao
supported in part by NSF grants to R.L.H. (DEB-
9208022) and R.K.W. (BSR-9020282) and by a
Sloan Young Investigator Award to K.A.C.
10 February 1997; accepted 28 April 1997
www.sciencemag.org?SCIENCE?VOL. 276?13 JUNE 1997
on May 6, 2010