A Massively Parallel Sequencing Approach Uncovers
Ancient Origins and High Genetic Variability of
Endangered Przewalski’s Horses
, Oliver A. Ryder
, Allison R. Fisher
, Bryant Schultz
, Sergei L. Kosakovsky Pond
, and Kateryna D. Makova*
Department of Biology, The Pennsylvania State University
San Diego Zoo Institute for Conservation Research, San Diego Zoo Global, California
Division of Infectious Diseases, Division of Biomedical Informatics, School of Medicine, University of California–San Diego
Department of Biochemistry and Molecular Biology, The Pennsylvania State University
*Corresponding author: E-mail: email@example.com.
Accepted: 30 June 2011
The endangered Przewalski’s horse is the closest relative of the domestic horse and is the only true wild horse species
surviving today. The question of whether Przewalski’s horse is the direct progenitor of domestic horse has been hotly
debated. Studies of DNA diversity within Przewalski’s horses have been sparse but are urgently needed to ensure their
successful reintroduction to the wild. In an attempt to resolve the controversy surrounding the phylogenetic position and
genetic diversity of Przewalski’s horses, we used massively parallel sequencing technology to decipher the complete
mitochondrial and partial nuclear genomes for all four surviving maternal lineages of Przewalski’s horses. Unlike single-
nucleotide polymorphism (SNP) typing usually affected by ascertainment bias, the present method is expected to be largely
unbiased. Three mitochondrial haplotypes were discovered—two similar ones, haplotypes I/II, and one substantially divergent
from the other two, haplotype III. Haplotypes I/II versus III did not cluster together on a phylogenetic tree, rejecting the
monophyly of Przewalski’s horse maternal lineages, and were estimated to split 0.117–0.186 Ma, signiﬁcantly preceding
horse domestication. In the phylogeny based on autosomal sequences, Przewalski’s horses formed a monophyletic clade,
separate from the Thoroughbred domestic horse lineage. Our results suggest that Przewalski’s horses have ancient origins
and are not the direct progenitors of domestic horses. The analysis of the vast amount of sequence data presented here
suggests that Przewalski’s and domestic horse lineages diverged at least 0.117 Ma but since then have retained ancestral
genetic polymorphism and/or experienced gene ﬂow.
Key words: wild horse, next-generation sequencing, mitochondrial DNA, nuclear DNA, phylogeny.
Understanding the genetic relationship between domestic
and Przewalski’s horses is critical for unraveling the domes-
tication history of the former and for formulating conserva-
tion and breeding strategies for the latter. Indeed, the
endangered Przewalski’s horse (Equus przewalskii) is the on-
ly wild horse living at present time and is the closest extant
relative of the domestic horse (Equus caballus). Previously
inhabiting an extensive range of steppe in both Asia and
Europe, Przewalski’s horse had become virtually extinct in
the wild due to human activity by the middle of the
1960s; however, it had subsequently been bred in captivity
and reintroduced to the wild (Ryder and Wedemeyer 1982;
Ryder 1993; Bouman and Bouman 1994). The present-day
Przewalski’s horse population, consisting of over 2,000 ani-
mals, originated from 12 founders captured at the turn of
the 19th century, a mare captured in 1957 and her descend-
ents, some of which were hybrids with domestic horses. On-
ly four Przewalski’s horse matrilines, potentially identiﬁable
as discrete mitochondrial lineages, currently survive (Volf
et al. 1991; Oakenfull and Ryder 1998). Przewalski’s horse
is phenotypically distinct from the domestic horse in having
shorter stature and more robust build than the former horse
(Sasaki et al. 1999), although there is substantial variation
ª The Author(s) 2011. Published by Oxford University Press on behalf of the Society for Molecular Biology and Evolution.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/
3.0), which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
1096 Genome Biol. Evol. 3:1096–1106. doi:10.1093/gbe/evr067 Advance Access publication July 29, 2011
among domestic horse breeds. The karyotype of Przewal-
ski’s horse (2n 5 66) differs from that of the domestic horse
(2n 5 64) by a Robertsonian translocation (Benirschke et al.
1965; Bowling and Ruvinsky 2000; Myka et al. 2003; Yang
et al. 2003; Ahrens and Stranzinger 2005). Despite these
differences, interbreeding between Przewalski’s and domes-
tic horses produces fertile offspring (Short et al. 1974).
Previous studies presented contradictory conclusions re-
garding whether domestic and Przewalski’s horses formed
monophyletic genetic clades. Although some protein, mi-
crosatellite, and Y chromosome analyses have supported
phylogenetic separation of the two taxa (Bowling et al.
2003; Wallner et al. 2003, ﬁg. 1A), the latest investigations
of autosomal (Lau et al. 2009; Wade et al. 2009) and X chro-
mosomal DNA (Lau et al. 2009) have not. For instance, our
previous phylogenetic analysis of several autosomal and X
chromosomal introns placed Przewalski’s horses within
the domestic horse clade (Lau et al. 2009). Similarly, phylo-
genetic separation between Przewalski’s versus domestic
horses was not observed after typing ;1,000 autosomal sin-
gle-nucleotide polymorphisms (SNPs) as part of the analysis
of the horse genome (Wade et al. 2009). This led to the
speculation that the E. przewalskii lineage could either be
very recently derived from one or more E. caballus lineages
(ﬁg. 1C) or that the two horses intermixed to a limited de-
gree following divergence from a common ancestor (Wade
et al. 2009). Only relatively short genomic regions (or a small
number of sites) have been analyzed in the reports men-
tioned above. Studies of mitochondrial DNA (mtDNA) of
Przewalski’s horse have so far been limited to sequencing
the control region (Ishida et al. 1995; Oakenfull and Ryder
1998; Oakenfull et al. 2000).
Because Przewalski’s horses are the only truly wild horses
existing today, they have been hypothesized to be the direct
ancestors of domestic horses (Mohr 1959; Ryder 1994, ﬁg.
1B). The history of horse domestication has been investi-
gated largely from mitochondrial and Y chromosome se-
quences, leading to suggestions of a limited number of
patrilines (Lindgren et al. 2004), but numerous matrilines
et al. 2001) incorporated into the genetic pool of do-
mestic horses. Multiple domestication events have been
suggested to occur (Vila
et al. 2001; Jansen et al. 2002). Al-
though genetic studies have yet to identify when and where
horse domestication ﬁrst took place, a recent archaeozoo-
logical report indicated the presence of domesticated horses
;5,500 years ago in Kazakhstan (Outram et al. 2009).
Determining the genetic relationship and divergence time
between Przewalski’s versus domestic horses is expected to
inform conservation efforts aimed at preserving the genetic
diversity of Przewalski’s horses. An accurate picture of ge-
netic diversity is particularly important for endangered spe-
cies such as Przewalski’s horse, for which the number of
founding individuals was small, and thus, the effect of in-
breeding has been a constant concern. Inbreeding reduces
genetic diversity, causes high mortality, and short life span
(Ralls and Ballou 1983) and has affected a captive popula-
tion of Przewalski’s horse (Bouman and Bos 1979). Despite
efforts to minimize inbreeding of Przewalski’s horses and to
maintain their current genetic variation (Ryder et al. 1984;
e et al. 1990; Zimmermann 1997), crucial studies of
their genetic diversity have so far been limited. Przewalski’s
horse genetic diversity has been estimated utilizing blood
group and allozyme loci (Bowling and Ryder 1987), mtDNA
(Ryder 1994), and single strand conformation polymor-
phism analysis for major histocompatibility complex genes
(Hedrick et al. 1999), leading to substantially disparate val-
ues. None of the previous investigations assessed genetic
variation of Przewalski’s horses from DNA data on a
In this study, we estimate the genetic diversity of Przewal-
ski’s horses and elucidate their phylogenetic relationship
with domestic horses. Using massively parallel sequencing
technology, we obtained complete mitochondrial and par-
tial nuclear genomes for four Przewalski’s horse individuals
FIG.1.—Hypothetical scenarios of divergence between Equus caballus and Equus przewalskii. Gray box indicates horse domestication. (A) E.
caballus and E. przewalskii form two reciprocally monophyletic lineages. The two sister species diverged from a common ancestor. (B) E. caballus is
derived from E. przewalskii. E. przewalskii is the direct ancestor of E. caballus.(C ) E. przewalskii is derived from E. caballus. E. caballus is the direct
ancestor of E. przewalskii.
Evolutionary History of Przewalski’s Horses GBE
Genome Biol. Evol. 3:1096–1106. doi:10.1093/gbe/evr067 Advance Access publication July 29, 2011 1097
representing all four surviving mitochondrial lineages. We
constructed phylogenetic trees and assessed nucleotide di-
versity of Przewalski’s horses based on mitochondrial, auto-
somal, and X chromosomal data separately. Based on these
results, we discuss the genetic relationship of Przewalski’s
versus domestic horses. In particular, we address whether
Przewalski’s horses and domestic horses represent two dis-
tinct evolutionary genes pools in the diversity of horses. This
study is valuable for guiding Przewalski’s horse conservation
efforts and illuminates the history of horse domestication.
Unlike SNP typing usually affected by ascertainment bias
(reviewed in Nielsen 2004), the present method is expected
to be largely unbiased. Indeed, most previous studies com-
menced with an SNP discovery protocol that by deﬁnition
is limited to a subset of samples or populations. Once discov-
ered, SNPs were typed in a larger sample, whereas variants
unique to individuals not utilized in SNP discovery remained
unassayed. This represents a serious problem that can now be
tackled with the use of massively parallel sequencing technol-
ogy (e.g., Luikart et al. 2003; Fridjonsson et al. 2011).
Materials and Methods
Four Przewalski’s horse individuals representing all four surviving
mitochondrial lineages (Bowling and Ryder 1987) were ana-
lyzed: female Belina or OR383,studbooknumber319,maternal
lineageStarajaII; female Anushka orOR2661,studbooknumber
book number 339, maternal lineage Bijsk B; and male Bars or
KB7674, studbook number 285, maternal lineage Orlica III.
For Somali wild ass, we analyzed a female, sample OR3030.
All DNA samples were kindly provided by the San Diego Zoo
Safari Park; samples OR383, OR1305, and OR3030 came from
animals resident in that park, whereas samples OR2661 and
KB7674 came from animals originally housed in Minnesota
Zoo and Tierpark Hellabrun (Munich), respectively.
DNA Extraction and Next Generation Sequencing
Genomic DNA was isolated using QIAGEN DNeasy Blood & Tis-
sue kit. Paired-end sequences, either 35- or 100-bp reads (for
Przewalski’s horses and for Somali wild ass, respectively) sepa-
Solexa Genome Analyzer System II. The number of reads
obtained per individual is listed in supplementary table S1 (Sup-
plementary Material online). All reads generated in this study
were deposited in sequencing trace archive (submission ID:
DRA000429; study ID: DRP000437; sample IDs: DRS000782-
DRS000786; experiment IDs: DRX000819-DRX000823; run
IDs: DRR001222-DRR001226; analysis ID: DRZ000051).
Sequencing Read Mapp ing and Filtering
We used Burrows-Wheeler Aligner (Li and Durbin 2009)
with default parameters and allowing no more than two
mismatches to map full-length (untrimmed) paired-end
reads to the horse nuclear genome (eca2) and its reference
mtDNA (NC_001640). Reads were mapped against each in-
dividual chromosome separately, and reads mapping in dis-
cordance to their mate pair relationships were discarded
(i.e., when both reads of a pair mapped to the same strand
and/or the distance between the two mapped reads was
greater than 2,000 bp). Only pairs of reads that mapped
to unique positions in the horse genome were retained. Ba-
ses with Illumina sequencing quality score below 20 were
eliminated after mapping.
For mtDNA, a consensus sequence was constructed for
each sequenced individual (GenBank accession numbers
AP012267–AP012271). Only sites supported by at least
three (by at least two for Bars) uniquely mapped reads were
used in the subsequent analysis. Sites with deletions, as
compared with reference, were excluded. Sites possessing
polymorphisms with frequency 0.28 (supported by at least
two reads in Bars) were excluded (a total of 28 sites, see sup-
plementary table S2, Supplementary Material online, for
a summary of heteroplasmic sites). The most frequent base
was taken as the consensus for the other polymorphic sites.
For nuclear DNA, we only used sites supported by at least
two uniquely mapped sequencing reads with an identical
call, and all polymorphic sites were discarded.
Estimation of Divergence Time and Phylogenetic
For the mtDNA alignment, divergence time was estimated
by Bayesian ‘‘relaxed molecular clock’’ approach imple-
mented in BEAST (Drummond and Rambaut 2007). We ﬁt-
ted a Tamura–Nei sequence evolution model (Tamura and
Nei 1993), assuming a relaxed molecular clock with uncor-
related rates sampled from the log-normal distribution and
the Yule process tree prior. For the Bayesian analysis, we
used Markov chain Monte Carlo sampling for 10 million
generations (burn-in 1,000 generations). Because of excess
transitions, unequal nucleotide frequencies, and variation of
substitution rate among different sites for mtDNA, the
Tamura–Nei model is thought to be the appropriate model
for such data (Tamura and Nei 1993). Moreover, because the
evolutionary distances among analyzed samples are very
small, the choice of the model is not expected to inﬂuence
the results substantially. Previously estimated divergence
time between domestic horse and Somali wild ass of 2.0
Ma was used as a calibration point (Forste
For the large-scale pairwise alignments of autosomes and
chromosome X, we calculated maximum likelihood pairwise
distances utilizing the Tamura–Nei sequence evolution model
(Tamura and Nei 1993) and estimated their sampling errors
using nonparametric bootstrap. The Tamura–Nei model
was chosen because it has a simple closed form solution
for estimating the distance from the matrix of pairwise differ-
ence counts (Tamura and Nei 1993). A custom utility in C was
Goto et al. GBE
1098 Genome Biol. Evol. 3:1096–1106. doi:10.1093/gbe/evr067 Advance Access publication July 29, 2011
developed to permit rapid estimation and bootstrapping
from long pairwise alignments; source code is available
for download from the HyPhy subversion code repository. Es-
timated distances were used then to build phylogenetic trees
with the Neighbor-Joining method (Saitou and Nei 1987) as
implemented in HyPhy (Kosakovsky Pond and Muse 2006).
Bootstrap test (with 1,000 replicates) was performed to ob-
tain the statistical support for each clade of the tree observed.
The joint bootstrapping procedure constructed 1,000 pair-
wise distance matrices by independently sampling each dis-
tance from its corresponding bootstrap distribution. This
procedure is expected to overestimate topological variance,
when compared with the standard multiple sequence align-
ment method, and correspondingly lower conservative values
for phylogenetic clade support.
To test whether the obtained phylogenetic trees have topol-
ogies signiﬁcantly superior to alternative topologies, we
utilized two approaches. For the mtDNA data, we used
the Kishino–Hasegawa, Shimodaira–Hasegawa, weighted
Kishino–Hasegawa, and weighted Shimodaira–Hasegawa
tests (Shimodaira and Hasegawa 1999) as implemented in
CONSEL (Shimodaira and Hasegawa 2001). For the nuclear
DNA data, which did not possess the form of a multiple se-
quence alignment, but rather a series of largely nonoverlap-
ping pairwise alignments, standard tests for phylogenetic
support do not apply; hence, we adopted a simulation ap-
proach. In order to estimate the parameters of the nucleotide
substitution process, we ﬁtted the GTR model with site-to-site
rate variation modeled by a 3-bin general discrete distribution
(which is more ﬂexible than the standard discretized gamma,
Kosakovsky Pond and Frost 2005) to the 5-way alignment of
horse DNA sequences (415,452 bases). We next simulated ge-
nomic alignments of four Przewalski’s horse, the Thorough-
bred horse, and the Somali wild ass sequences (depicted in
ﬁg. 3A) under a collection of alternative topologies and sub-
sampled (N 5 100 times) pairwise alignments of lengths
equal to those obtained from real data and reran the neigh-
bor-joining tree construction (NJ þ TN93) procedure on sim-
ulated data. We asked the following questions using our
simulations: a) assuming that the data are generated using
the inferred tree (e.g., ﬁg. 3A and B) what is the frequency
at which it is recovered from pairwise alignments using NJ þ
TN93? and b) how often is the tree inferred from real data
recovered by NJ þ TN93 if the Thoroughbred lineage is placed
at any alternative location in the Przewalksi horses clade?
To conﬁrm the observed variation in four mtDNA lineages of
Przewalski’s horse, we analyzed eight additional Przewalski’s
horse samples: Basil, Henrietta, Hermonia (descendents of
Staraja II), Bertland, Bonar, Nadiushka, Rolmar (descendents
of Bijsk B), and Kuporovitch (descendent of Bijsk/2). We per-
formed capillary sequencing for two mtDNA regions, corre-
sponding to positions 2981–3647 and 15506–15860 in Xu
and Arnason (Xu and Arnason 1994). These positions were
selected because they contained multiple sites differentiat-
ing the haplotypes. Details are available in the Supplemen-
tary Material online.
Sequencing Complete Mitochondrial Genomes
from All Surviving Przewalski’s Horse Matrilines
Using the paired-end module of an Illumina Genome Analyzer
II, we sequenced genomic DNA (mtDNA and nuclear DNA), of
four Przewalski’s horses (three females: Belina, Anushka, and
Bonnette, and one male: Bars) representing all four surviving
mitochondrial lineages (Staraja II, Bijsk/2, Bijsk B, and Orlica III,
respectively). Genomic DNA from a Somali wild ass (Equus
africanus somaliensis), an outgroup (Beja-Pereira et al.
2004), was also sequenced. For four of the ﬁve sequenced
individuals, DNA was isolated from the heart muscle, resulting
in enrichment for mtDNA. Heart tissue was not available for
Bars, and thus, his blood was utilized instead.
Although we sequenced complete mtDNA in all ﬁve in-
dividuals, our subsequent analysis was limited to the se-
quencing reads that were uniquely mapped to mtDNA, in
order to exclude potential mtDNA insertions in the nuclear
genome (numts; Bensasson et al. 2003). Mutations speciﬁc
to individual lineages of Przewalski’s horse were conﬁrmed
by Sanger sequencing. The ﬁnal data set included 10,840
base pairs (bp) covered in all ﬁve individuals and aligned
to the reference mtDNA sequence of the domestic horse
(supplementary tables S1 and S3, Supplementary Material
online). Mean mtDNA read coverage per site ranged from
60x to 260x for DNA isolated from heart and was 7x for
DNA isolated from blood.
Two mtDNA sequences (heteroplasmic and indel sites
were excluded, see Supplementary Material online) were
identical (haplotype I, observed in Anushka and Belina),
the third sequence (haplotype II, present in Bars) was very
similar to haplotype I, whereas the fourth one (haplotype
III, observed in Bonnette) was substantially different from
both of them (supplementary table S3, Supplementary Ma-
terial online). Partial sequencing of mtDNA from additional
horses belonging to the four original matrilines conﬁrmed
these results (see Supplementary Material online). Haplo-
types I and II differed by only two substitutions in 10,840
bp (both substitutions were located outside of the control
region). In contrast, sequence divergence (corrected for
multiple hits) between haplotype I (or II) and haplotype III
was much higher, 59 ± 6.8 substitutions in 10,840 bp or
0.548 ± 0.064% (outside the control region, there were
47 ± 6.3 substitutions in 10,085 bp or 0.466 ± 0.063%).
Phylogenetic Analysis of mtDNA Sequences
Bayesian phylogenetic analysis of Przewalski’s and publicly
available domestic horse mtDNA sequences (sequences
Evolutionary History of Przewalski’s Horses GBE
Genome Biol. Evol. 3:1096–1106. doi:10.1093/gbe/evr067 Advance Access publication July 29, 2011 1099
located outside the control region were utilized because
they produced the most reliable alignment), with Somali
wild ass as an outgroup, conﬁrmed substantial differences
between the Przewalski’s horse haplotypes I/II versus III (ﬁg.
2). Speciﬁcally, these haplotypes did not form a monophy-
letic clade but instead were separated on the phylogenetic
tree: the mtDNA sequences of haplotypes I/II grouped with
this from the Tibetan breed (from Naqu), whereas haplotype
III was more basal; previous studies have shown that domes-
tic horse mtDNA haplotypes are not breed speciﬁc (Vila
2001; Jansen et al. 2002). The observed topology (ﬁg. 2)
was signiﬁcantly better than alternative topologies assum-
ing that (1) domestic and Przewalski’s horses form two
monophyletic clades, respectively, (2) domestic horse is an
ancestor to Przewalski’s horse, or (3) Przewalski’s horse is
an ancestor to domestic horses (supplementary table S4,
Supplementary Material online). Similar results were ob-
tained when ;600 bp of the mtDNA control region of
37 domestic horse breeds were included in the phylogenetic
analysis (supplementary ﬁg. S1, Supplementary Material on-
line). These observations rely on a substantially longer se-
quence and thus a more informative alignment than in
previous studies (e.g., Jansen et al. 2002), implying that
Przewalski’s horses are not monophyletic for mtDNA.
Assuming that the divergence time between horse and
Somali wild ass was 2 Ma (Forste
n 1992), we estimated that
haplotypes I/II and III separated from each other ;0.156 Ma
(conﬁdence interval [CI] is between 0.117 and 0.186 Ma).
Therefore, these two Przewalski’s horse mtDNA lineages
have diverged from a common ancestral sequence anciently.
Diversity and Phylogenetic Analysis of Portions of
the Nuclear Genome of Przewalski’s Horses
Our approach fortuitously provided partial nuclear DNA se-
quences for Przewalski’s horses and Somali wild ass (see Ma-
terials and Methods). The total number of analyzed bases
ranged from 24.5 to 106.2 million for autosomes and from
0.8 to 5.6 million for the X chromosome, depending on in-
dividual (supplementary table S3, Supplementary Material
online), a vast increase over previous studies. Nevertheless,
due to inclusion of the nuclear genome from only one do-
mestic horse breed (the Thoroughbred, the only domestic
horse breed for which large-scale nuclear genome data
are currently available) and due to low sequence coverage
of the nuclear genomes of Przewalski’s horses and Somali
wild ass, caution should be exercised when interpreting
FIG.2.—Phylogenetic tree and divergence time estimates for mtDNA sequences of Przewalski’s and domestic horses. The results of analysis carried
out in BEAST. Based on a fossil record, we assumed that Somali wild ass and domestic horse diverged 2.0 Ma (Forste
n 1992). Numbers below nodes are
posterior probabilities, and bold numbers above nodes are the estimated divergence times. The 95% highest posterior density estimates for each clade
are represented by bars.
Goto et al. GBE
1100 Genome Biol. Evol. 3:1096–1106. doi:10.1093/gbe/evr067 Advance Access publication July 29, 2011
Utilizing these data, we built pairwise alignments for all
possible combinations of six individual nuclear genomes (ﬁve
genomes partially sequenced here plus the reference horse
genome) and estimated the corresponding pairwise nucleo-
tide genetic distances separately for chromosome X and au-
tosomes based on the Tamura and Nei model of sequence
evolution (Tamura and Nei 1993; supplementary table S3,
Supplementary Material online). The distances between
the Somali wild ass sequence and any of the horse sequences
were substantially higher than the other comparisons, reaf-
ﬁrming that Somali wild ass is an adequate outgroup. The
analysis below is based on pairwise alignments because
few bases were expected to align in all ﬁve sequenced ani-
mals, given low nuclear genome coverage (in fact, using the
proportions of genomes covered in the sequenced regions for
each individual from supplementary table S1, Supplementary
Material online, and using 2.5 Gb [Wade et al. 2009]asthe
size of the horse genome, we only expect ;2.5 bases to be
shared among all ﬁve sequenced individuals).
The average pairwise autosomal genetic distance be-
tween the Thoroughbred domestic and Przewalski’s horse
was 0.226% (95% CI 5 0.225–0.227%), slightly higher
than 0.18% reported recently in a study analyzing a smaller
data set (Wade et al. 2009). The average autosomal diver-
gence between Somali wild ass and either the Thorough-
bred domestic or Przewalski’s horse was ;1.1%.
Assuming strict molecular clock and 2 Ma divergence be-
tween horse and Somali wild ass (Forste
n 1992), this corre-
sponds to the rate of 2.75 10
substitutions per site per
year. Using this rate, we computed the coalescence time be-
tween the Thoroughbred domestic horse and sequenced
Przewalski’s horses to be ;0.411 Ma (95% CI 5 0.409–
0.413). The average autosomal pairwise diversity among
the four Przewalski’s horses was 0.195% (95% CI 5
0.189–0.199%; supplementary table S3, Supplementary
Material online), suggesting a coalescence time of 0.353
Ma (95% CI 5 0.344–0.362), an even more ancient origin
than estimated from shorter mtDNA sequences.
To investigate divergence in the nuclear genome between
domestic and Przewalski’s horses, we constructed Neighbor-
Joining (Saitou and Nei 1987) phylogenies from pairwise dis-
tances, separately for autosomes and the X chromosome
(ﬁg. 3). Note that these phylogenetic trees need to be cor-
roborated in future studies including additional domestic
horse breeds. Nevertheless, from our results, in contrast
to the mtDNA tree indicating two distinct Przewalski’s hap-
lotype groups that were intermingled with domestic horse
haplotypes (ﬁg. 2), on the autosomal tree Przewalski’s
horses formed a monophyletic clade (ﬁg. 3A). The paramet-
ric simulation test described in the Materials and Methods
indicated that 1) the data generated under the tree in ﬁgure
3A led to the recovery of the correct tree in 100% of cases
using NJ þ TN93 methodology and 2) the probability of in-
ferring the tree in ﬁgure 3A given any other alternative
placement of Przewalski’s horses (i.e., where these horses
are not monophyletic) was ,0.01 (based on 100 replicates).
These results suggest that the low level of pairwise diver-
gence and very long sequences allow NJ þ TN93 to accu-
rately recover the underlying topology even when only
pairwise alignments are available.
The average X chromosome divergence between Somali
wild ass and either the Thoroughbred domestic horse or
Przewalski’s horses was ;0.817% (supplementary table
S3, Supplementary Material online), lower than for auto-
somes, in agreement with the phenomenon of male muta-
tion bias (Makova and Li 2002) and a smaller effective
population size for the X chromosome. Assuming molecular
clock and 2 Ma divergence between horse and Somali wild
FIG.3.—Neighbor-joining trees of autosomal sequences (A) and chromosome X sequences (B ) based on pairwise genetic distances. Numbers at
the nodes represent bootstrap support values and numbers on the branches indicate genetic distances. Somali wild ass (Equus africanus somaliensis )
was used as an outgroup.
Evolutionary History of Przewalski’s Horses GBE
Genome Biol. Evol. 3:1096–1106. doi:10.1093/gbe/evr067 Advance Access publication July 29, 2011 1101
n 1992), this corresponds to the X chromosomal
rate of 2.04 10
substitutions per site per year. The X
chromosomal distances between horses were also lower
than the autosomal distances in most cases (supplementary
table S3, Supplementary Material online). The X chromo-
somal pairwise distances including data from the only male
used in this analysis, Bars, were unusually high, potentially
due to the mapping of some Y chromosome sequences to
the domestic horse X chromosome. The average nucleotide
diversity among Przewalski’s horse X chromosomes was
0.211% (0.180% excluding Bars). The average pairwise dis-
tance between the X chromosome sequences of Przewal-
ski’s horse Bonnette versus other horses (this is the
deepest root among horse sequences) was 0.186% (95%
CI 5 0.164–0.210%) corresponding to coalescence time
of 0.455 Ma (95% CI 5 0.402–0.514 Ma). Excluding Bars,
this value was 0.179% (95% CI 5 0.157–0.202%), corre-
sponding to coalescence time of 0.439 Ma (95% CI 5
0.385–0.495 Ma), largely in agreement with our autosomal
results. Thus, all three types of data—autosomal, X chromo-
somal, and mitochondrial—point toward ancient genetic
origins of Przewalski’s horses sequenced here.
On the X chromosomal tree, the Thoroughbred domestic
horse sequence was intermingled with Przewalski’s horse se-
quences, and no clustering was signiﬁcant (ﬁg. 3B ). Para-
metric simulations of topological signal showed that the
correct tree is recovered with 100% accuracy, and the prob-
ability of inferring the tree in ﬁgure 3B given any other al-
ternative placement of the Thoroughbred lineage is ,0.01
(based on 100 simulations).
Evolution of Przewalski’s and Domestic Horses
According to our data, Przewalski’s horse is not the direct pro-
genitor of the domestic horse. If domestic horses had been
derived from the Przewalski’s horse, then domestic horse se-
quences would have been embedded within the Przewalski’s
horse phylogenetic clade (ﬁg. 1B; Ryder 1994). This expecta-
tion is contradicted by the analysis of mtDNA data. Indeed,
mtDNA haplotypes of Przewalski’s horses are intermingled
with domestic horse sequences (ﬁg. 2, also see below). Ad-
ditionally, Przewalski’s horse autosomal sequences form a sep-
arate monophyletic clade excluding the Thoroughbred
domestic horse (ﬁg. 3A), although this result will have to
be reevaluated when nuclear sequences of additional domes-
tic horses become available.
The hypothesis of a recent origin of Przewalski’s horses
from domestic horses (Oakenfull and Ryder 1998; Wade
et al. 2009), according to which Przewalski’s sequences are
expected to be embedded within the domestic horse phylo-
genetic clade (ﬁg. 1C), is also contradicted by our data be-
cause our analysis placed one of the Przewalski’s horse
haplotypes at the deepest branching point among currently
available complete mtDNA caballine sequences (ﬁg. 2) and at
one of the deepest branching points among currently avail-
able mtDNA control region sequences (supplementary ﬁg. S1,
Supplementary Material online). Autosomal sequences of ad-
ditional domestic horse breeds are needed to test the origin of
Przewalski’s horses from domestic horses more explicitly.
Do Przewalski’s and domestic horses represent two dis-
tinct evolutionary gene pools in the diversity of horses? Sev-
eral recent studies based on the analyses of the mtDNA
control region (Oakenfull and Ryder 1998; Ishida et al.
et al. 2001; Jansen et al. 2002; Kim et al.
1999) and autosomal SNPs (Wade et al. 2009) failed to sep-
arate Przewalski’s and domestic horse sequences in molec-
ular phylogenies. The reports demonstrating genetic
differentiation between Przewalski’s and domestic horses
were based on either nuclear-encoded data sets (e.g.,
Bowling et al. 2003) or Y chromosomal DNA (Wallner
et al. 2003). In the latter study, the split between the two
lineages was estimated to occur 0.123–0.241 Ma, a date sig-
niﬁcantly preceding horse domestication, and the karyotype
present in domestic horses was suggested to represent the
derived condition. Recently, based on a well-resolved phylog-
eny of the Perissodactyla, Steiner and Ryder (Steiner CC, Ryder
OA, submitted) asserted that the ancestral state of the
Robertsonian translocation between E. caballus and E. prze-
walskii was the unfused elements and corresponding higher
diploid number currently present in E. przewalskii,incontrast
with the results of Myka et al. (2003) and of Trifonov et al.
(2008) that relied upon an alternate phylogeny of Equus.
It has been suggested that in Przewalski’s horses, the ge-
nealogy of nuclear DNA might be different from the gene-
alogy of mtDNA (Ishida et al. 1995), as was observed for
African elephants (Roca et al. 2005). Our results indeed in-
dicate the presence of just such differences leading in differ-
ent answers to the question of separation of genetic pool
between domestic and Przewalski’s horses. The mtDNA
analysis resulted in intermingling between Przewalski’s
and Thoroughbred horse sequences, with one of Przewal-
ski’s horse haplotypes located at the most basal position
among the available complete mtDNA caballine sequences.
In contrast, all Przewalski’s horses formed a separate clade
on the autosomal tree with high bootstrap support. The lat-
ter conclusion results from the unbiased analyses of SNP var-
iation using distance methods. Although not deﬁnitive,
because a large number of domestic horses have not been
investigated, the results obtained using nuclear data are
consistent with a scenario that is quite distinct from the view
that emerges from analysis of mtDNA variation.
Different topologies for horse mitochondrial versus nuclear
DNA observed here might be explained by distinct evolution-
ary histories of horse matrilines versus patrilines. Genetic in-
trogression between domestic and Przewalski’s horses may
have been largely female mediated, in agreement with the
hypothesis proposed by Wallner et al. (2003). This explanation
Goto et al. GBE
1102 Genome Biol. Evol. 3:1096–1106. doi:10.1093/gbe/evr067 Advance Access publication July 29, 2011
would be consistent with the nonmonophyletic placement of
Przewalski’s horse sequences in the mtDNA tree (exclusively
maternal) as well as their intermingling in the X chromosomal
tree (predominantly maternal). Male-mediated admixture be-
tween Przewalski’s and domestic horses may have been lim-
ited. Indeed, the male-speciﬁc Y chromosome contains ﬁxed
differences separating Przewalski’s and domestic horses
(Wallner et al. 2003). Also, the present analysis of Przewalski’s
horse autosomal sequences that have a substantial paternal
contribution groups them in a monophyletic clade. The di-
chotomous ﬁndings obtained by us for autosomal versus
mtDNA data can potentially be resolved in the future by
investigations of ancient DNA and greater analysis of se-
quence variation in domestic horses, especially utilizing
methods developed in the present study (that do not intro-
duce ascertainment bias from the individuals used for SNP
discovery; reviewed in Nielsen 2004).
Divergent mtDNA Haplotypes among Przewalski’s
The detailed analysis of complete mitochondrial genomes
from all four surviving maternal lineages of Przewalski’s
horses indicated their ancient nonmonophyletic origins.
We identiﬁed three mtDNA haplotypes; two haplotypes (I
and II) were very similar to each other, whereas the third
one (III) was markedly distinct from the other two haplo-
types. Why are haplotypes I/II so divergent from haplotype
III, and why do not the three haplotypes form a monophyletic
clade, even though the horses harboring them went through
a severe genetic bottleneck (Volf et al. 1991) and do not
exhibit morphological variation?
First, the observed haplotypes could exemplify the ge-
netic polymorphism present in the ancestral horse popula-
tion that existed prior to the divergence of Przewalski’s and
other major modern horse lineages (Jansen et al. 2002). The
deep phylogenetic separation between haplotypes I/II and III
could represent the natural variation within a single species,
driven by an early maternal lineage split. We cannot exclude
the possibility that selection or geographic isolation contrib-
uted to the separation of the haplotypes. Regardless of the
mechanism, Przewalski’s horses (as well as possibly domestic
horses) could have retained such ancestral variation in their
current population. This scenario would interfere with infer-
ring the monophyletic origins of Przewalski’s horses.
Second, some haplotypes could have been introduced
from domestic into Przewalski’s horses via interbreeding.
In particular, haplotypes I/II are very similar in sequence to
and cluster together with domestic horse haplotypes and,
thus, might have been acquired by Przewalski’s horse through
introgressive hybridization. Przewalski’s horse haplotype III
might represent the ‘‘true’’ Przewalski’s horse mtDNA. How-
ever, additional sequencing of mtDNA from modern domes-
tic horses from Eurasia may identify mtDNA haplotypes
similar in sequence to Przewalski’s horse haplotype III. A com-
bination of these two scenarios—some haplotypes acquired
via introgression and other haplotypes inherited from the an-
cestral horse population—is also possible.
The presence of highly divergent mtDNA haplotypes is
unexpected for Przewalski’s horses because they have gone
through a genetic bottleneck. However, this is not unprec-
edented because some other mammalian species also ex-
hibit distinct mtDNA haplotypes within their continuous
populations, for example, moose (Hundertmark et al.
2002), reindeer (Flagstad and Røed 2003), elephant
(Fleischer et al. 2001), and mammoth (Gilbert et al. 2008).
A Model of Divergence between Przewalski’s and
The analyses of all three types of genomic data (mtDNA, au-
tosomal, and X chromosomal) indicate that Przewalski’s and
domestic horse lineages diverged signiﬁcantly preceding
horse domestication, thought to have occurred ;5,000–
6,000 years ago (Outram et al. 2009). In fact, mtDNA hap-
lotypes of Przewalski’s horses coalesce 0.117–0.187 Ma,
that is, at least a hundred thousand years prior to horse do-
mestication. Moreover, Przewalski’s horse autosomal se-
quences, as well as X chromosomal sequences, coalesce
several hundred thousand years preceding horse domestica-
tion. These observations are at variance with the hypothesis
that Przewalski’s horse population represents the wild stock
from which the domestic horses were bred, even though our
results suggest a close genetic relationship between mtDNA
haplotypes of some Przewalski’s and domestic horses (see
discussion below). Note that, if both groups of horses re-
tained substantial levels of ancestral polymorphism, this
would interfere with our estimates of the divergence of their
lineages. Nevertheless, the drastic difference between horse
domestication time (;5,000–6,000 years ago) and our co-
alescent estimates (at least 117,000 years ago) is unlikely to
be the result of retained ancestral polymorphism alone.
From our phylogenetic analysis (see above), we con-
cluded that domestic horse is neither derived from Przewal-
ski’s horse nor the opposite. We propose a model according
to which Przewalski’s horse and domestic horse are de-
scendents of two lineages that diverged potentially as early
as ;0.150 Ma. This is consistent with our mtDNA and nu-
clear analyses as well as with the published Y chromosomal
results (Wallner et al. 2003). Indeed, a monophyletic group-
ing of Przewalski’s horses on an autosomal tree is not antic-
ipated in a recently diverged genetic pool containing
substantial shared genetic variation. Nevertheless, the initial
divergence event between the two lineages could have
been followed by the retention of ancestral polymorphism
and/or introgressive hybridization. The signatures of these
events are particularly conspicuous in the mtDNA data
due to the absence of recombination.
The monophyly of Przewalski’s horse nuclear sequences
contradicts the recent ﬁndings of Wade et al. (2009) who
Evolutionary History of Przewalski’s Horses GBE
Genome Biol. Evol. 3:1096–1106. doi:10.1093/gbe/evr067 Advance Access publication July 29, 2011 1103
suggested that only few Przewalski’s horse-speciﬁc muta-
tions are absent from the domestic horse population.
Why do the results of these two studies differ? First, Wade
and colleagues used a smaller data set that was based on
SNPs derived from domestic horses and thus was prone
to ascertainment bias (reviewed in Nielsen 2004). Second,
the monophyly of Przewalski’s horse sequences presented
here might be in part inﬂuenced by sparse sampling of au-
tosomal domestic horse genomes; future studies will have to
evaluate this possibility. Note that three of the Przewalski’s
horses sequenced here (Bars, Belina, and Bonette) are
thought to have no recent (since the bottleneck) domestic
horse genetic contribution to their known pedigrees,
whereas 16% of Anushka’s DNA can be traced back to
a Mongolian domestic mare (Ballou 1994). Nevertheless,
Anushka’s autosomal sequences formed a clade with those
from Bars and did not have a greater genetic distance to
Przewalski’s horses without recent domestic horse contribu-
tion. These results suggest that a more ancient gene ﬂow
might have occurred between ancestral populations of Prze-
walski’s wild horse and Asian domestic horse breeds, espe-
cially since the past distribution of Przewalski’s horse
overlaps with the present-day Mongolian horse distribution
(Ishida et al. 1995) and corroborate our ﬁndings based on
Relatively High Nucleotide Diversity among
The high average nucleotide diversity observed here in the
nuclear and mtDNA of Przewalski’s horses was unexpected,
given that their population had dwindled to a mere dozen
individuals only 40 years ago (Volf et al. 1991) and was sub-
sequently subject to inbreeding. Mean autosomal diversity
in Przewalski’s horses (0.195%; this study) was higher than
that in several breeds of domestic horses studied by us pre-
viously (0.1%; Lau et al. 2009) as well as in the sequenced
Thoroughbred horse (0.05%; Wade et al. 2009). Average X
chromosomal diversity was similarly higher for Przewalski’s
than for domestic horses (0.182% estimated here versus
0.1% in domestic horses as estimated by us; Lau et al.
2009). Note that our estimates of nucleotide diversity from
nuclear DNA data are likely deﬂated because we required
each analyzed site to be supported by two or more identical
sequencing reads. For mtDNA, Przewalski’s horse nucleotide
diversity was also relatively high. When estimated from total
mtDNA, the nucleotide diversity between divergent haplo-
types (0.54%; this study) was comparable to that estimated
for Tibetan horse breeds (0.66%; Xu et al. 2007). When es-
timated from the mtDNA control region, the nucleotide di-
versity between divergent haplotypes (1.6%; this study) was
only slightly lower than that observed in natural populations
of moose (2.5%; Hundertmark et al. 2002) and Asian ele-
phant (1.8%; Roca et al. 2005).
What can explain the relatively high genetic diversity in
the modern population of Przewalski’s horses, despite a re-
cent and severe genetic bottleneck? First, the Przewalski’s
horse population that existed prior to the bottleneck might
have possessed substantial levels of genetic diversity and
some of this diversity may have carried across. Second, in-
terbreeding with domestic horses, known to have occurred
for at least some Przewalski’s horses after the bottleneck
(Volf et al. 1991), and also, potentially, more anciently
(see above), might have elevated genetic diversity in Prze-
walski’s horses. One or both of these factors have likely
counterbalanced the effects of inbreeding.
Implications for Breeding Strategies
Our results have direct implications for the strategies of
breeding Przewalski’s horses, hundreds of which are kept
in captivity and are being released to the wild via reintroduc-
tion programs. The major goal of managing the Przewalski’s
horse population is to maintain it in sufﬁcient size and genetic
diversity to protect the species from extinction. To achieve this
goal, a careful analysis of past, current, and future genetic
characteristics of the population is required (Ballou 1994).
From this perspective, ﬁrst, our results demonstrate the exis-
tence of two highly divergent mitochondrial haplogroups in
Przewalski’s horses. It is imperative to ensure the survival of
both these haplotypes in growing populations of Przewalski’s
horses because all reintroduction projects should include rep-
resentation of the entire species gene pool (Ryder 1993).Sec-
ond, our analysis points toward substantial genetic diversity
persisting in the current population and likely present in the
founders of the surviving population. Nevertheless, inbreed-
ing should be kept at a minimum to preserve this genetic di-
versity. Third, and albeit indirectly, our results suggest ancient
introgression of domestic horse genes into Przewalski’s horse
genes. This questions the need to separate ‘‘pure’’ Przewal-
ski’s horses (e.g., the Munich line) from ‘‘non-pure’’ Przewal-
ski’s horses (i.e. with known domestic horse contributions)
because even the former lineage likely experienced some,
perhaps more ancient, admixture with domestic horses.
Moreover, interbreeding with domestic horses might have el-
evated the nucleotide diversity of Przewalski’s horses.
To further illuminate the natural history of Przewalski’s
horses, it will be necessary to investigate mtDNA haplotypes
as well as nuclear DNA from preserved skin specimens pres-
ent in museum collections around the world. This would al-
low one to evaluate the genetic diversity of Przewalski’s
horses prior to the bottleneck and, based on collection sites,
to correlate this diversity with the geographic distribution.
Additionally, the complete sequencing of the Przewalski’s
horse genome (or exome) and its detailed comparison with
the domestic horse genome is expected to facilitate the dis-
covery of genotypic differences of phenotypic consequen-
ces distinguishing these two closely related species.
Goto et al. GBE
1104 Genome Biol. Evol. 3:1096–1106. doi:10.1093/gbe/evr067 Advance Access publication July 29, 2011
Supplementary tables S1–S4, ﬁgure S1 and other
supplementary materials are available at Genome Biology
and Evolution online.
We thank Wen-Yu Chung, Benjamin Dickins, and Gurupra-
sad Ananda for assistance with computational analyses; Cyn-
thia Steiner, Melissa Wilson Sayres, Hie Lim Kim, Masafumi
Nozawa, Chungoo Park, and Lydia Krasilnikova for helpful
comments; and Leona Chemnick for her assistance in sample
handling and preparation. This work is supported by start-up
funds from the Eberly College of Science at The Pennsylvania
State University to K.D.M., by National Science Foundation
and National Institutes of Health grants to A.N., and by Na-
tional Institutes of Health grants to S.K.P. Additional funding
is provided, in part, under a grant with the Pennsylvania De-
partment of Health using Tobacco Settlement Funds. The De-
partment speciﬁcally disclaims responsibility for any analyses,
interpretations, or conclusions. Additional support from the
Sonny Foundation and the John and Beverly Stauffer Foun-
dation is gratefully acknowledged.
Ahrens E, Stranzinger G. 2005. Comparative chromosomal studies of E.
caballus (ECA) and E. przewalskii (EPR) in a female F1 hybrid. J Anim
Breed Genet. 122(Suppl 1):97–102.
Ballou JD. 1994. Population biology. In: Houpt KA, editor. Przewalski’s
horse: the history and biology of an endangered species. Albany
(NY): The State University of New York Press. p. 93–113.
Beja-Pereira A, et al. 2004. African origins of the domestic donkey.
Benirschke K, Malouf N, Low RJ, Heck H. 1965. Chromosome
complement: differences between Equus caballus and Equus
przewalskii, poliakoff. Science 148:382–383.
Bensasson D, Feldman MW, Petrov DA. 2003. Rates of DNA duplication
and mitochondrial DNA insertion in the human genome. J Mol Evol.
Bouman I, Bouman J. 1994. The history of the Przewalski’s horse. In:
Boyd L, Houpt KA, editors. Przewalski’s horse: the history and
biology of an endangered species. Albany (NY): The State University
of New York Press. p. 5–38.
Bouman JG, Bos H. 1979. Two symptoms of inbreeding depression in
Przewalski horses living in captivity. In: de Boer LEM, Bouman J,
Bouman I, editors. Genetics and hereditary diseases of the
Przewalski horse. Rotterdam (The Netherlands): Foundation for the
Preservation and Protection of the Przewalski horse. p. 165–168.
Bowling AT, Ruvinsky A. 2000. Genetic aspects of domestication,
breeds, and their origins. In: Bowling AT, Ruvinsky A, editors. The
genetics of the horse. Wallingford (UK): CABI Publishing. p. 25–51.
Bowling AT, Ryder OA. 1987. Genetic studies of blood markers in
Przewalski’s horses. J Hered. 78:75–80.
Bowling AT, et al. 2003. Genetic variation in Przewalski’s horses, with
special focus on the last wild caught mare, 231 Orlitza III. Cytogenet
Genome Res. 101:226–234.
Drummond AJ, Rambaut A. 2007. BEAST: Bayesian evolutionary analysis
by sampling trees. BMC Evol Biol. 7:214.
Flagstad O, Røed KH. 2003. Refugial origins of reindeer (Rangifer
tarandus L.) inferred from mitochondrial DNA sequences. Evolution
Fleischer RC, Perry EA, Muralidharan K, Stevens EE, Wemmer CM. 2001.
Phylogeography of the asian elephant (Elephas maximus) based on
mitochondrial DNA. Evolution 55:1882–1892.
n A. 1992. Mitochondrial-DNA time-table and the evolution of
Equus: comparison of molecular and paleontological evidence. Ann
Zool Fennici. 28:301–309.
Fridjonsson O, et al. 2011. Detection and mapping of mtDNA SNPs in
Atlantic salmon using high throughput DNA sequencing. BMC
Gilbert MT, et al. 2008. Whole-genome shotgun sequencing of
mitochondria from ancient hair shafts. Science 317:1927–1930.
Hedrick PW, Parker KM, Miller EL, Miller PS. 1999. Major histocompat-
ibility complex variation in the endangered Przewalski’s horse.
Hundertmark KJ, et al. 2002. Mitochondrial phylogeography of moose
(Alces alces): late pleistocene divergence and population expansion.
Mol Phylogenet Evol. 22:375–387.
Ishida N, Oyunsuren T, Mashima S, Mukoyama H, Saitou N. 1995.
Mitochondrial DNA sequences of various species of the genus Equus
with special reference to the phylogenetic relationship between
Przewalskii’s wild horse and domestic horse. J Mol Evol.
Jansen T, et al. 2002. Mitochondrial DNA and the origins of the
domestic horse. Proc Natl Acad Sci U S A. 99:10905–10910.
Kim KI, et al. 1999. Phylogenetic relationships of Cheju horses to other
horse breeds as determined by mtDNA D-loop sequence poly-
morphism. Anim Genet. 30:102–108.
Kosakovsky Pond SL, Frost SD. 2005. A simple hierarchical approach to
modeling distributions of substitution rates. Mol Biol Evol. 22:223–34.
Kosakovsky Pond SL, Muse SV. 2006. HyPhy: hypothesis testing using
phylogenies. Bioinformatics 21:676–679.
Lau AN, et al. 2009. Horse domestication and conservation genetics of
Przewalski’s horse inferred from sex chromosomal and autosomal
sequences. Mol Biol Evol. 26:199–208.
Li H, Durbin R. 2009. Fast and accurate short read alignment with
Burrows-Wheeler transform. Bioinformatics 25:1754–1760.
Lindgren G, et al. 2004. Limited number of patrilines in horse
domestication. Nat Genet. 36:335–336.
Luikart G, England PR, Tallmon D, Jordan S, Taberlet P. 2003. The power
and promise of population genomics: from genotyping to genome
typing. Nat Rev Genet. 4(12):981–94.
Makova KD, Li WH. 2002. Strong male-driven evolution of DNA
sequences in humans and apes. Nature 416:624–626.
Mohr E, 1959. Das Urwildpferd. Die Neue Brehm-Bu¨ cherei. Wittenberg
(Lutherstadt): A. Ziemsen Verlag, p. 144.
Myka JL, Lear TL, Houck ML, Ryder OA, Bailey E. 2003. FISH analysis
comparing genome organization in the domestic horse (Equus
caballus) to that of the Mongolian wild horse (E. przewalskii).
Cytogenet Genome Res. 102:222–225.
Nielsen R. 2004. Population genetic analysis of ascertained SNP data.
Hum Genomics. 1(3):218–224.
Oakenfull EA, Lim HN, Ryder OA. 2000. A survey of equid mitochondrial
DNA: implications for the evolution, genetic diversity, and conser-
vation of Equus. Conserv Genet. 1:341–355.
Oakenfull EA, Ryder OA. 1998. Mitochondrial control region and 12S
rRNA variation in Przewalski’s horse (Equus przewalskii). Anim
Evolutionary History of Przewalski’s Horses GBE
Genome Biol. Evol. 3:1096–1106. doi:10.1093/gbe/evr067 Advance Access publication July 29, 2011 1105
Outram AK, et al. 2009. The earliest horse harnessing and milking.
e FPG, Zimmerman W, Ryder OA, Dolan JM. 1990. The phenotypic
approach I genetic management of Przewalski’s horse. In: Seal US,
Foose TJ, Lacy RC, Zimmerman W, Ryder OA, Prince
e FPG, editors.
Przewalski’s horse draft global conservation plan. Apple Valley (MN):
Ralls K, Ballou J. 1983. Extinction: lessons from zoos. In: Schonewald-
Cox CM, Chambers SM, MacBryde B, Thomas WL, editors. Genetics
and conservation. San Francisco (CA): The Benjamin/Cummings
Publishing Company, Inc. p. 164–184.
Roca AL, Georgiadis N, O’Brien SJ. 2005. Cytonuclear genomic
dissociation in African elephant species. Nat Genet. 37:96–100.
Ryder OA. 1993. Przewalski’s horse: prospects for reintroduction into
the wild. Conserv Biol. 7:13–15.
Ryder OA. 1994. Genetic studies of Przewalski’s horses and their impact
on conservation. In: Boyd L, Houpt KA, editors. Przewalski’s horse:
the history and biology of an endangered species. Albany (NY): The
State University of New York Press. p. 75–92.
Ryder OA, et al. 1984. Genetics of Equus przewalskii Poliakov 1881:
analysis of genetic variability in breeding lines, comparison of equid
DNAs and a brief description of a cooperative breeding program in
North America. Equus 2:207–227.
Ryder OA, Wedemeyer EA. 1982. A cooperative breeding program for
the Mongolian wild horse, Equus przewalskii, in the United States.
Biol Conserv. 22:259–271.
Saitou N, Nei M. 1987. The neighbor-joining method: a new method for
reconstructing phylogenetic trees. Mol Biol Evol. 4:406–425.
Sasaki M, Endo H, Yamagiwa D, Yamamoto M, Arishima K, Hayashi Y.
1999. Morphological character of the shoulder and leg skeleton in
Przewalski’s horse (Equus przewalskii). Ann Anat. 181:403–407.
Shimodaira H, Hasegawa M. 1999. Multiple comparisons of Log-
likelihoods with applications to phylogenetic inference. Mol Biol
Shimodaira H, Hasegawa M. 2001. CONSEL: for assessing the conﬁdence
of phylogenetic tree selection. Bioinformatics 17:1246–1247.
Short RV, Chandley AC, Jones RC, Allen WR. 1974. Meiosis in interspeciﬁc
equine hybrids II. The Przewalski horse/domestic horse hybrid (Equus
przewalskii X E. caballus). Cytogenet Cell Genet. 13:465–478.
Tamura K, Nei M. 1993. Estimation of the number of nucleotide
substitutions in the control region of mitochondrial DNA in humans
and chimpanzees. Mol Biol Evol. 10:512–526.
Trifonov VA, et al. 2008. Multidirectional cross-species painting
illuminates the history of karyotypic evolution in Perissodactyla.
Chromosome Res. 16:89–107.
C, et al. 2001. Widespread origins of domestic horse lineages.
Volf J, Kus E, Prokopova
L. 1991. General studbook of the Przewalski
horse (Zoological Garden Prague, Prague, Czech Republic)
[cited 2011 Jul 1]. Available from: http://przwhorse.pikeelectronic.
Wade CM, et al. 2009. Genome sequence, comparative analysis,
and population genetics of the domestic horse. Science 326:
WallnerB,BremG,Mu¨ ller M, Achmann R. 2003. Fixed nucleotide
differences on the Y chromosome indicate clear divergence
between Equus przewalskii and Equus caballus.AnimGenet.
Xu S, et al. 2007. High altitude adaptation and phylogenetic analysis of
Tibetan horse based on the mitochondrial genome. J Genet
Xu X, Arnason U. 1994. The complete mitochondrial DNA sequence of
the horse, Equus caballus: extensive heteroplasmy of the control
region. Gene 148:357–362.
Yang F, et al. 2003. Karyotypic relationships of horses and zebras: results
of cross-species chromosome painting. Cytogenet Genome Res.
Zimmermann W. 1997. Die Bedeutung von Semire-servaten fu¨ r das EEP
Przewalskipferd. Zoo Magazin Nordrhein-Westfalen. 3:70–75.
Associate editor: Michael Purugganan
Goto et al. GBE
1106 Genome Biol. Evol. 3:1096–1106. doi:10.1093/gbe/evr067 Advance Access publication July 29, 2011