- Edited by Francisco Mauro Salzano, Universidade Federal do Rio Grande do Sul, Porto Alegre, Brazil, and approved March 30, 2012 (received for review July 12, 2011)
Abstract
Despite
decades of research across multiple disciplines, the early history of
horse domestication remains poorly understood. On the basis of current
evidence from archaeology, mitochondrial DNA, and Y-chromosomal
sequencing, a number of different domestication scenarios have been
proposed, ranging from the spread of domestic horses out of a restricted
primary area of domestication to the domestication of numerous distinct
wild horse populations. In this paper, we reconstruct both the
population genetic structure of the extinct wild progenitor of domestic
horses, Equus ferus, and the origin and spread of horse
domestication in the Eurasian steppes by fitting a spatially explicit
stepping-stone model to genotype data from >300 horses sampled across
northern Eurasia. We find strong evidence for an expansion of E. ferus
out of eastern Eurasia about 160 kya, likely reflecting the
colonization of Eurasia by this species. Our best-fitting scenario
further suggests that horse domestication originated in the western part
of the Eurasian steppe and that domestic herds were repeatedly
restocked with local wild horses as they spread out of this area. By
showing that horse domestication was initiated in the western Eurasian
steppe and that the spread of domestic herds across Eurasia involved
extensive introgression from the wild, the scenario of horse
domestication proposed here unites evidence from archaeology,
mitochondrial DNA, and Y-chromosomal DNA.
The
origin of horse domestication has been studied intensively for decades,
yet the key question of whether horse domestication originated in a
small number of geographically defined areas or whether numerous wild
populations were domesticated independently remains unanswered. An
increasing body of archaeological evidence points to an origin of horse
domestication in the steppes of modern-day Ukraine and Kazakhstan (1) (see ref. 2
for review). However, there is as yet no corroborating molecular
evidence for a geographically restricted origin of horse domestication
anywhere in the Eurasian steppes, the main distribution area of wild
horses at the time of domestication.
A related question
concerns the spread of horse domestication across the Eurasian steppes:
Did the spread of horse domestication involve actual movement of herds
(“demic spread”), as appears to have been the case in most other
domestic animal species (3)?
Or was it primarily the knowledge of horse domestication techniques
that spread, thus enabling pastoralist societies throughout the steppes
to domesticate locally available wild stock? Whereas a demic spread of
small herds of domestic horses out of a single, geographically
restricted area has been put forward as one possible explanation for the
observed low Y chromosome diversity in modern horses (4),
the multiple-origins scenario is commonly invoked to account for the
large number of female lineages in the domestic horse gene pool (2, 5, 6).
In
this paper, we use a spatially and demographically explicit model,
parameterized with autosomal genotype data from >300 horses, sampled
in 12 localities distributed throughout northern Eurasia (Fig. 1A),
to investigate the origin and spread of horse domestication. The model
presented here allows us to distinguish between horse domestication in a
single geographic area vs. multiple geographic areas, to pinpoint the
geographic origin of domesticated horses in the former case, and to
determine the relative roles of demic spread and recruitment of local
wild stock in the spread of horse domestication.
Fig. 1.
Geographic distribution of sampling locations (country and administrative unit) and the model. (A)
LTA, Lithuania, Vilnius; UKZ, Ukraine, Zakarpattia; UKL, Ukraine, Lviv;
KAL, Russia, Kalmykia; GGA, Georgia, Samegrelo; KSW, Kazakhstan,
Mangystau; KAR, Kazakhstan, Kyzylorda; KKA, Kazakhstan, Karagandy; KYR,
Kyrgyzstan, Naryn; ALT, Russia, Altai Republic; CHA, China, Xinjiang;
MGL, Mongolia, Övörkhangai. (B) Schematic representation of the linear stepping-stone model used in this study (see Materials and Methods for a detailed description of the parameters). Note that for illustrative purposes only five demes are shown. (C) Location of the three putative origins of E. ferus. (D) Locations of the four putative domestication origins.
Results
The wild progenitor of domestic horses, Equus ferus,
is extinct; we therefore used a stepping-stone dynamic that allowed us
to simultaneously reconstruct both the population genetic structure of E. ferus and that of its domestic descendants (see Fig. 1B for a diagrammatic representation of the model). We considered three population origins of E. ferus—western, central, and eastern Eurasia (Fig. 1C)—and combined each of the three wild horse origins with four putative origins of horse domestication (Fig. 1D),
yielding a total of 12 combined scenarios. Depending on the choice of
parameters, the model can cover a wide range of possibilities, spanning
from populations having undergone range expansions to populations at
migration–drift equilibrium. In the latter case, all three putative
origins of E. ferus in Eurasia would be equally well supported.
In this paper, we define domestication as a process whereby founder
populations of domestic stock are established de novo and distinguish it
from introgression as a process of restocking already domesticated
herds with wild individuals.
For each of the 12 combined
scenarios, we investigated a variety of population dynamics in both
wild and domestic horse populations, as well as the full spectrum of
possible modes of spread of domestication, ranging from a purely demic
spread without wild horse introgression to numerous local domestications
without population movements. In the latter case, all four putative
domestication origins should fit the data equally well. We used an
approximate Bayesian computation framework to determine the likelihood
of each of the 12 scenarios by comparing the model predictions to the
observed data (Fig. 1A and Table S1).
The scenarios positing an origin of E. ferus in eastern Eurasia received the strongest support, regardless of the assumed domestication origin (Fig. 2, red bars). This finding is in agreement with paleontological evidence suggesting that Pleistocene horse species, including E. ferus, originated in North America and migrated to Eurasia via the Bering land bridge (7). Of the four domestication scenarios assuming an origin of E. ferus in eastern Eurasia, the scenario proposing a domestication origin in western central Eurasia received the strongest support (Fig. 2).
Western central Eurasia, as defined here, comprises the steppes of
modern-day Ukraine and northwest Kazakhstan, where some of the earliest
archaeological evidence for managed horse populations has been found (1, 2, 8); here we provide additional evidence for the importance of this region as a primary area of horse domestication.
Fig. 2.
Bayes factors for 12 scenarios describing the origin of the wild progenitor of domestic horses, Equus ferus,
and the origin of horse domestication in Eurasia. Bayes factors are
relative to the best-supported scenario, which posits an origin of E. ferus
in eastern Eurasia and a domestication origin in western central
Eurasia. The scenarios are grouped by domestication origin and colored
by the origin of E. ferus.
It
should be noted that our dataset does not include populations west of
Ukraine. Horses in central and western Europe tend to either belong to a
specific recognized breed or be considerably breed admixed.
Unfortunately, the modeling framework used here is not well suited to
deal with the complex demographic histories experienced by horse
populations in central and western Europe; the role of these areas in
horse domestication can therefore not be confidently established here.
Whereas central Europe is unlikely to have played an important role in
horse domestication due to a lack of suitable wild horse habitat (9), there is increasing evidence that horse domestication may have also occurred in the Iberian Peninsula (9, 10).
Next, we explored the demographic details of the most strongly supported scenario, which posits an expansion of E. ferus
out of eastern Eurasia and a domestication origin in western central
Eurasia. The parameter combinations that best fitted the data strongly
suggest that the expansion of E. ferus occurred ∼160 kya (Fig. 3A). Although the credibility interval (CI) for this date is relatively wide (95% CI: 51–180 kya; Table S2), our estimate is very similar to the estimated age of 130–160 kya for the common maternal ancestor of modern domestic horses (10).
The scenario of an expansion of E. ferus out of eastern Eurasia ∼160 kya is furthermore consistent with paleontological data indicating that E. ferus first appeared in East Asia ∼200 kya (7). Thus, the expansion captured by our model likely reflects the colonization of Eurasia by this species. Our results further suggest that this expansion was characterized by relatively strong founder effects (small cK, Fig. 3B), large effective population sizes (K, Fig. 3C), and a rate of spread in the order of ∼100 km in 240 y (growth rate r = 0.05, Fig. 3D). Due to the early date of this expansion, a wide range of migration rates in wild horses are consistent with the data (mK, Fig. 3E).
The scenario of an expansion of E. ferus out of eastern Eurasia ∼160 kya is furthermore consistent with paleontological data indicating that E. ferus first appeared in East Asia ∼200 kya (7). Thus, the expansion captured by our model likely reflects the colonization of Eurasia by this species. Our results further suggest that this expansion was characterized by relatively strong founder effects (small cK, Fig. 3B), large effective population sizes (K, Fig. 3C), and a rate of spread in the order of ∼100 km in 240 y (growth rate r = 0.05, Fig. 3D). Due to the early date of this expansion, a wide range of migration rates in wild horses are consistent with the data (mK, Fig. 3E).
Fig. 3.
Posterior distributions of demographic parameters for wild and domestic populations estimated through ABC analysis. (A) Timing (t) of the expansion of E. ferus out of eastern Eurasia. (B) Effective founder population sizes of the wild horse (cK, red line), the first domestic population (cd0Kd, blue dashed line), and subsequent domestic populations (cdKd, blue solid line). (C) Carrying capacity (effective population sizes) of the ancestral (K0, dashed red line), wild (K, solid red line), and domestic populations (Kd, blue line). (D) Growth rate, r, of wild horse populations. (E) Migration rates in wild (mK, red line), and domestic populations (mdKd, blue line). (F) The proportion of domestic (q) and wild (1 − q) horses in domestic horse founder populations as a function of their effective size, cdKd.
Both the scenario postulating a purely demic spread (q = 1) and the one postulating a spread based solely on local recruitment (q = 0) are outside the 95% credibility interval for q (the proportion of already domesticated horses in the founding population of subsequent domestic populations; Table S2),
suggesting that the spread of horse domestication out of western
central Eurasia involved both actual population movement and
introgression from the wild.
The best-supported values of q ranged between 0.2 and 0.7 (Fig. 3F), suggesting that both mechanisms were approximately equally important. The demography of horses changed markedly following domestication. Compared with the expansion of E. ferus, the spread of domestic horses was characterized by weaker founder effects (larger cdKd in Fig. 3B) and smaller effective sizes of established populations (Kd, Fig. 3C). We obtained strong support for high migration rates between domestic populations (mdKd, Fig. 3E), consistent with the increased mobility of human societies following horse domestication.
The best-supported values of q ranged between 0.2 and 0.7 (Fig. 3F), suggesting that both mechanisms were approximately equally important. The demography of horses changed markedly following domestication. Compared with the expansion of E. ferus, the spread of domestic horses was characterized by weaker founder effects (larger cdKd in Fig. 3B) and smaller effective sizes of established populations (Kd, Fig. 3C). We obtained strong support for high migration rates between domestic populations (mdKd, Fig. 3E), consistent with the increased mobility of human societies following horse domestication.
Discussion
Previous
research into the origin of horse domestication has led to different,
sometimes contradictory, conclusions regarding the number of
domestication origins and the mode of spread of horse domestication.
Here we provide additional support for an origin of horse domestication
in a geographically defined area in the western Eurasian steppe as
suggested by archaeology (1, 2).
We furthermore show that the spread of horse domestication out of the
western Eurasian steppe was characterized by high levels of
introgression from local wild populations. In light of the exceptionally
high levels of matrilineal diversity in horses (e.g., ref. 11),
we suggest that introgression from the wild was mainly female mediated.
The repeated capture of wild females for the purpose of maintaining or
growing domestic herd sizes may seem counterintuitive, given that in
other domestic animal species, introgression from the wild typically
involved domestic females being impregnated by wild males (12).
However, given the initial difficulties in breeding the most closely related wild relative of domestic horses, Przewalski’s horse, in captivity (13), it can be speculated that, for an indeterminate amount of time, herd sizes could not be maintained solely through breeding existing stock (14). Because stallions are inherently more difficult to handle than mares, the easiest way to maintain or grow herd sizes would have been to restock existing herds with wild females. Whereas the high levels of diversity and limited geographic structure in the horse mitochondrial genome (11) may thus reflect the continued augmentation of domestic herds with wild mares from a wide geographic area, the observed low levels of Y chromosome variability (15) might reflect the strong domestication bottleneck in western central Eurasia (low cd0Kd, Fig. 3B) (4). Additional losses of patrilineal diversity may have occurred as a consequence of successive bottlenecks as domestic herds spread out of the western steppes (4) and of breeding practices reducing the effective size of the male gene pool (16).
However, given the initial difficulties in breeding the most closely related wild relative of domestic horses, Przewalski’s horse, in captivity (13), it can be speculated that, for an indeterminate amount of time, herd sizes could not be maintained solely through breeding existing stock (14). Because stallions are inherently more difficult to handle than mares, the easiest way to maintain or grow herd sizes would have been to restock existing herds with wild females. Whereas the high levels of diversity and limited geographic structure in the horse mitochondrial genome (11) may thus reflect the continued augmentation of domestic herds with wild mares from a wide geographic area, the observed low levels of Y chromosome variability (15) might reflect the strong domestication bottleneck in western central Eurasia (low cd0Kd, Fig. 3B) (4). Additional losses of patrilineal diversity may have occurred as a consequence of successive bottlenecks as domestic herds spread out of the western steppes (4) and of breeding practices reducing the effective size of the male gene pool (16).
The
geographic pattern of within- and between-population heterozygosity
observed in contemporary horses from the Eurasian steppes reflects the
combined effects of an east-to-west expansion of E. ferus and repeated introgression of local wild animals into spreading domestic herds. The expansion of E. ferus out of eastern Eurasia set up an isolation-by-distance pattern (off-diagonal elements in Fig. S1A) and a weak east-to-west decline in within-population genetic diversity (on-diagonal elements in Fig. S1A). The pattern of isolation-by-distance has been preserved in modern Eurasian steppe horses (off-diagonal elements in Fig. S1B)
due to extensive backcrossing of domestic herds to wild populations
from across the steppes. The demic spread of domestic horses out of
western central Eurasia, combined with the continued genetic input from
local wild populations, has resulted in an accentuation of the original
east-to-west decline in within-population genetic diversity (on-diagonal
elements in Fig. S1B)
because hybridization involved increasingly genetically differentiated
populations. Whereas migration rates in domestic horses were found to be
high (Fig. 3E), the relatively recent occurrence of horse domestication (from ∼5.5 kya) (1, 17)
means that there has not been enough time for increased population
movements to obscure the genetic signature of the initial colonization
process on large geographic scales, an observation that also holds true
for other species (18⇓–20), including humans (21, 22).
In
the past decade, spatially explicit models have successfully been used
to unravel complex demographic scenarios in the context of the spread of
agricultural practices. Using a modified wave-of-advance model, Ackland
et al. (23)
identified the factors influencing the hitchhiking of cultural traits
alongside beneficial subsistence technologies in heterogeneous
landscapes. Itan et al. (24)
used a demic computer model to trace the spread of lactase persistence
and therefore dairying in Europe. More recently, Van Etten and Hijmans (25)
used a landscape model, parameterized with archaeobotanical and genetic
data, to trace the origin and routes of spread of maize in the
Americas. In this paper, the combined use of autosomal markers and
spatially explicit modeling has allowed us to reconstruct the population
genetic structure of the extinct wild ancestor of domestic horses and
to unravel the complex processes involved in horse domestication. The
scenario of horse domestication proposed here unites evidence from
archaeology, mtDNA, and Y-chromosomal sequence data that has previously
given rise to conflicting scenarios.
Materials and Methods
Sample Collection and DNA Analysis.
Our
dataset consists of 322 nonbreed horses sampled in 12 geographically
localized (within a radius of 100 km) sampling areas spanning eight
countries (Fig. 1A and Table S1).
Populations of nonbreed horses (local village horses that are mainly
used for work and do not belong to any particular breed) have not been
subjected to the same degree of intense artificial selection and closed
breeding as many modern breeds and are thus expected to be more
informative regarding horse population history.
Genomic DNA was extracted from 15–20 hair roots per individual according to a protocol adapted from ref. 26.
DNA extracts were purified (QIAquick purification kit; Qiagen) and
standardized to a concentration of 10 ng of DNA/μL. A total of 26
microsatellite markers were amplified in two multiplex reactions (Table S3),
using the Type-it Microsatellite PCR kit from Qiagen. PCR
amplifications were carried out according to the manufacturer’s
instructions using 10 ng of template DNA and a total volume of 12.5 μL.
Samples were run on an ABI 3730 Genetic Analyzer (Applied Biosystems).
Alleles were assigned with GeneMapper Software v.3.7 (Applied
Biosystems).
Modeling Framework.
We
used a general stepping-stone framework that can represent a wide
variety of demographic scenarios, including populations at
migration–drift equilibrium and populations having undergone range
expansions, depending on the choice of parameters. Eurasian populations
of E. ferus were represented by a linear chain of 80 demes
(numbered 0–79, from east to west), each 100 km apart, stretching from
east to west and starting from latitude 50° N and longitude 125° E (deme
0). Sampling locations were assigned to demes on the basis of their
shortest distance on land from deme 0, such that, for example, a
location 1,200 km from deme 0 would be assigned to deme 12. The shortest
distances on land between populations were calculated as in ref. 22. To avoid boundary artifacts, we appended 20 demes on both sides of the chain (see Fig. 1D
for a schematic representation of the model). To test the sensitivity
of our results to deme spacing, we repeated the full analysis using 160
demes with a spacing of 50 km as well as 40 demes with a spacing of 200
km, respectively; sampling locations within the chain of demes were
recalculated on the basis of the new deme spacings.
We considered three putative population origins of E. ferus, one in eastern Eurasia (deme 0), one in central Eurasia (deme 40), and one in western Eurasia (deme 80) (Fig. 1C). In each scenario, the deme corresponding to the population origin of E. ferus in Eurasia was populated by randomly sampling cK0 diploid individuals from a hypothetical ancestral population of size K0. The initial founder population grows linearly at rate rK horses per generation until it reaches carrying capacity K. Demes at carrying capacity send out a fraction c of colonizers to neighboring empty demes. In addition, occupied neighboring demes exchange mNmin migrants per generation, where Nmin represents the smaller of the two population sizes.
The domestication process was initiated 450 generations before present (t
− 450), the number of generations that have elapsed since domestication
started, assuming an average generation time of 12 y for wild horses (27) and a start date for horse domestication around 6,000 y ago (1, 16).
Populations of domestic horses were represented by a linear chain of
demes parallel to the one representing the wild horse populations and
with the same spatial structuring. The sampled horse populations (Fig. 1A)
were placed on this chain according to their geographic distance from
the easternmost deme as described above. We considered four putative
origins of horse domestication, one in eastern Eurasia (deme 0), one in
eastern central Eurasia (deme 25), one in western central Eurasia (deme
50), and one in western Eurasia (deme 75) (Fig. 1D). In each run, domestication was initiated by randomly sampling cd0Kd
individuals from the deme representing the wild population closest to
the putative domestication origin. The initial founder population grows
within one generation to size Kd, the carrying capacity of the domestic demes. Subsequently, empty demes (i.e., domestic herds) are established from cdKd
effective individuals representing a mixture of already domesticated
horses from the previous deme and local wild horses in proportions q and 1 − q, respectively. The parameter q
thus describes the proportion of already domesticated horses in the
founding population of subsequent domestic populations, with q = 1 corresponding to a pure demic spread of domestication and q
= 0 corresponding to multiple independent domestications without a
demic component. Once established, neighboring domestic populations
exchange migrants at rate md.
Given
the large number of parameters and scenarios, running a suitable number
of stochastic simulations is challenging. We optimized computing time by
finding the analytical solution predicting the expected values of
within- and between-population heterozygosity for any given combination
of parameter values for each scenario (SI Materials and Methods).
Given the expected values, we generated approximate stochastic samples
from the model by adding noise generated with the same correlation
structure between the elements of the matrix of heterozygosity values as
in the observed data (estimated using the bootstrap method and 10,000
replicates). A validation of this approach is given in SI Materials and Methods.
Model Fitting.
We
fitted our model by using an approximate Bayesian computation
framework, using the ABC-GLM algorithm implemented in the ABCtoolbox
software (28).
We used six summary statistics to describe our dataset: After assigning
the sampled populations to three groups (west, central, and east, Fig. S2),
we computed average within-population heterozygosities within each of
the groups (three estimates) and average between-group heterozygosities
(west vs. central, west vs. east, and central vs. east, three
estimates).
We started by randomly sampling 55 million combinations of parameter values within the following ranges: t ∈ [500, 15,000] generations (corresponding to 6–180 kya), m ∈ [10−6, 10−3], cK ∈ [1, 104], r ∈ [0.005, 1], K ∈ [4,000, 105], K0 ∈ [1, 105], md ∈ [10−4, 0.5], Kd ∈ [500, 104], cdKd ∈ [1, 103], q ∈ [0, 1], and cdKd0 ∈ [1, 103]. Whereas t and q
were sampled according to the uniform distribution of their
untransformed values, all other parameters were sampled from the uniform
distribution of their log-transformed values. For each parameter value
combination, we then generated our six summary statistics for each of
the 12 scenarios, combining all possible origins for wild horses and
domestication events in our model. In these calculations we took the
mutation rate (μ) to be 1.5 × 10−4 per generation (the average of mutation rate estimates for two microsatellite markers, AHT4 and HTG10) (29).
Because horse domestication occurred relatively recently, the only
parameters that are considerably affected by the mutation rate are the
time of the initial expansion of E. ferus in Eurasia (t) and the ancestral population size of E. ferus (K0).
For each scenario, we ran ABC-GLM on the accepted parameter
combinations (on the basis of the 0.1 percentile of Euclidean distances
between simulated and observed summary statistics) to estimate posterior
distributions of the model parameters and the likelihood of the summary
statistics as estimated from the genetics data (Table S2).
We then used Bayes factors, given by the ratio of estimated likelihoods
for each pair of scenarios, for model comparison. Because both Bayes
factors and posterior distributions for a deme spacing of 100 km (Figs. 2 and 3) were in close agreement with those obtained for deme spacings of 50 km and 200 km, respectively (Figs. S3–S5), we refer to the former only in the main text.
Acknowledgments
Two
anonymous referees provided several useful comments on the manuscript.
We thank Marsha Levine, Ian Colvin, and Gulnara Aitpaeva for sharing
contacts of local collaborators and Rebecca Cassidy, Vicky Collard,
Jaqueline Ripart, Bryan Miller, Natalia Vibla, and Giedre Keen for help
with sampling. V.W. thanks the National Institute of Agricultural
Botany, Cambridge, for use of their laboratory facilities. This work was
partially supported by a research studentship from the Biotechnology
and Biological Sciences Research Council and a doctoral studentship from
the German Academic Exchange Service (to V.W.), a Biotechnology and
Biological Sciences Research Council grant and a Philip Leverhulme Award
(to A.M.), and a Leverhulme Trust project grant (to G.B. and M.A.B.).
Footnotes
- ↵1V.W. and A.E. contributed equally to this work.
- ↵2To whom correspondence may be addressed. E-mail: vmw23@cam.ac.uk, aje44@cam.ac.uk, or am315@cam.ac.uk.
- Author contributions: V.W., A.E., and A.M. designed research; V.W. performed research; V.W., M.A.B., G.B., E.B., B.K.H., S.L., D.L., M.O.-G., G.V.S., and V.S. contributed new reagents/analytic tools; V.W. and A.E. analyzed data; and V.W., A.E., and A.M. wrote the paper.
- The authors declare no conflict of interest.
- This article is a PNAS Direct Submission.
- Data deposition: The microsatellite genotyping data reported in this paper have been deposited in the Dryad repository, http://dx.doi.org/10.5061/dryad.c598v (doi:10.5061/dryad.c598v).
- This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1111122109/-/DCSupplemental.
Nenhum comentário:
Postar um comentário
Observação: somente um membro deste blog pode postar um comentário.