Phylogenetic relationships of Argentinean Creole horses and other South American and Spanish breeds inferred from mitochondrial DNA sequences

Transcription

1 Phylogenetic relationships of Argentinean Creole horses and other South American and Spanish breeds inferred from mitochondrial DNA sequences P. M. Mirol*, P. Peral García*, J. L. Vega-Pla and F. N. Dulout* *CIGEBA, Facultad de Ciencias Veterinarias, Universidad Nacional de La Plata, La Plata, Argentina. Laboratorio de Grupos Sanguíneos, Servicio de Cria Caballar, Córdoba, Spain Summary South American horses constitute a direct remnant of the Iberian horses brought to the New World by the Spanish conquerors. The source of the original horses was Spain, and it is generally assumed that the animals belonged to the Andalusian, Spanish Celtic, Barb or Arabian breeds. In order to establish the relationship between Argentinean and Spanish horses, a portion of the mitochondrial D-loop of 104 animals belonging to nine South American and Spanish breeds was analysed using SSCP and DNA sequencing. The variability found both within and between breeds was very high. There were 61 polymorphic positions, representing 16% of the total sequence obtained. The mean divergence between a pair of sequences was 2.8%. Argentinean Creole horses shared two haplotypes with the Peruvian Paso from Argentina, and the commonest haplotype of the Creole horses is identical to one of the Andalusian horses. Even when there was substantial subdivision between breeds with highly significant Wright s Fixation Index (FST), the parsimony and distance-based phylogenetic analyses failed to show monophyletic groups and there was no clear relationship in the trees between the South American and any of the other horses analysed. Although this result could be interpreted as mixed ancestry of the South American breeds with respect to the Spanish breeds, it is probably indicating the retention of very ancient maternal lineages in the breeds analysed. Keywords D-loop, horse, mtdna, phylogeny. Introduction The Argentinean Creole horse constitutes a direct remnant of the Iberian horses brought to the New World by the Spanish conquerors during the 15th century. The source of the original horses was Spain, and this was at a time when the Spanish horse was being used for improvement of horse breeding throughout Europe. On the basis of historical records, at least 250 horses where shipped to the continent, Address for correspondence P. M. Mirol, School of Biological Sciences, Queen Mary, University of London, Mile End Road, London E1 4NS, UK. p.m.mirol@qmul.ac.uk Accepted for publication 29 March 2002 from the second voyage of Columbus in 1493 to Nuñez de Cabeza s trip in 1540 (Rodero et al. 1992). These horses, as all other domestic animals transported, were quickly dispersed and became very well adapted to the new environment. At that time, Spanish exports arrived directly from the ports of Spain to their final destination, with a stop on the Canary Islands or the Antilles. There were in Spain three main morphological equine types: the Celtic type of tarpanic origin in the north and west region of the peninsula; the Spanish type, descendent from the African Barb horse, in the south and east; and finally, in the central area, the hybrid between both of them. As Seville, Cadiz and other southern ports monopolized the navigation to America, it could be assumed that horses taken to America were mainly of the Spanish type or Andalusian (Rodero et al. 1992).

2 MtDNA variation in South American and Spanish horses 357 However, the possibility of taking to America other equine types cannot be ruled out. During his second trip to the New World, Columbus complained to the Spanish Crown that the excellent horses shown to him before the departure had been changed by cheaper animals (Tudela 1987). In 1508 the Spanish Crown authorized the transport of 40 horses from Castilla in the expedition organized by Alonso Ojeda and Diego de Nicuesa to Panama. The horses were of the Celtic type, small and resistant. Furthermore, many animals died during the 2-month trip, and other animals from the intermediate ports, Canary Islands or Antilles, could have replaced them. The advent of mitochondrial DNA analysis in population genetics during the 1970s produced a revolutionary change regarding historical, biogeographic and phylogenetic perspectives on intra- and interspecific genetic structure (Avise 1994). Since then, it has been widely used to infer intra- and interspecific phylogenetic relationships. Mitochondrial DNA studies in horses have proved useful to characterize intrabreed variation (Kavar et al. 1999; Kim et al. 1999; Bowling et al. 2000), although retention of ancestral polymorphism makes phylogenetic inference difficult (Vilà et al. 2001). In this context we analyse the mitochondrial DNA variation of Argentinean Creole horses and some other South American and Spanish horses. The characterization of the Creole breed has been traditionally based on morphology and behaviour, and it was only during the last few years that analysis based on molecular genetic markers have been introduced (Peral García et al. 1996). Knowledge of the South American breeds is also important to conservation genetics of domestic horses, as the New World varieties are probably closer in type to the historical horse of Spain than are the current horses in Iberia, which over the last 500 years have interbred with other breeds. Materials and methods DNA extraction, PCR amplification and SSCP of the D-loop Total DNA was extracted from blood samples using the DNAZOL purification kit (Gibco Life Technologies, Rockville, MD, USA) following the manufacturer instructions. Samples included 45 Argentinean Creole horses (ARC), 30 Peruvian Paso from Argentina (PPA), 18 Arabian (16 from Argentina AR, and two from Spain ARSP) and 11 Spanish horses belonging to the breeds Asturcon (AST), Losino (LO), Potoka (PO), Mallorquina (MA), Menorquina (ME) and Spanish Pure Breed or Andalusian (AND). When pedigree information was available, we selected individuals that have not shared a common ancestor for at least three generations. The polymerase chain reaction (PCR) primers used were: L-strand: 5 -AGGACTATCAAAGGAGAAGCTCTA-3 (P1; Ishida et al. 1994) and H-strand: 5 -CCTGAAGTA- GGAACCAGATG-3 (H16498; Marklund et al. 1994), which amplify a 466-bp region situated between the trna Thr (position 15397; Xu & Árnason 1994) and the central domain of the D-loop (position 15863; Xu & Árnason 1994). The 50 ll reaction mix contained approximately 100 ng of total horse DNA, 0.5 lm of each primer, 0.1 mm of dntps and 2 U of Taq polymerase (Gibco BRL, Rockville, MD, USA) in 20 mm Tris HCl (ph 8.4), 50 mm KCl and 2mMMgCl 2, under mineral oil. The PCR consisted of a first denaturation step at 96 C for 2 min followed by 35 cycles of 1 min at 94 C, 30 s at 55 C and 1 min at 72 C, with an elongation step of 5 min at 72 C in the last cycle. The size of the products was estimated by 1.5% agarose gel electrophoresis with pbr322 MspI Digest as size marker. Argentinean Creole horses, Peruvian Paso from Argentina and Arabian (Argentina) were pre-selected before sequencing by PCR SSCP analysis, to include a range of genetically distinct individuals. Fifteen microliters of each PCR product was added to 40 ll of dye LIS (10% sucrose, 0.01% bromophenol blue and 0.01% xylene cyanol FF; Maruya et al. 1996). The samples were then heated at 96 C for 10 min, cooled on ice for at least 5 min and loaded onto a 10% polyacrylamide gel (49 : 1 acrylamide : bisacrylamide). Electrophoresis was carried out at 4 C, 200 V in 0.5 Tris-Borate-EDTA (TBE) buffer for 18 h. The gels were subsequently fixed in 5% ethanol, stained with 0.2% AgNO 3 and revealed with 2% CaCO 3. Cloning and sequencing of the PCR products The PCR products were purified using QIAquick columns (QIAGEN, Hilden, Germany) and cloned into dt-tailed pgem-t easy vector (Promega, Madison, WI, USA) following manufacturer s recommendations. The DNA sequencing was performed with an Applied Biosystems 377 automated sequencer (BioResource Centre, Cornell University, Ithaca, NY, USA) using T7 and M13 universal primers. At least two independent clones were sequenced from each individual and the unique substitutions were confirmed by sequencing clones from a second PCR product. Data analysis Sequences of the D-loop were aligned using CLUSTAL-V multiple alignment software (Higgins et al. 1992). Sites representing a gap in any of the aligned sequences were excluded from the analysis, and distances between D-loop sequences were estimated using both the absolute number of nucleotide differences and the Kimura two-parameter distance (Kimura 1980) calculated on the basis of an equal

3 358 Mirol et al. substitution rate per site. Phylogenies were constructed using maximum parsimony with the PAUP 4.0 software (Swofford 1997) and the NEIGHBOR program incorporated in the PHYLIP package (Felsenstein 1991). For this analysis sequences belonging to other horse breeds from GenBank ( were incorporated. In all analyses the sequence of Equus asinus (Xu & Arnason 1996) was used as outgroup. The statistical confidence of each node in the consensus trees was estimated by 1000 bootstrap resampling of the data. Analysis of molecular variance (AMOVA) and pairwise FST distances were calculated using Arlequin (Schneider et al. 2000). Results Single strand conformation polymorphism (SSCP) A total of 91 horses were examined for SSCP, 45 belonging to the Argentinean Creole breed, 16 Arabian and 30 Peruvian Paso. The PCR products were of the same length, approximately 460 base pairs, which is in agreement with the published horse sequences (Ishida et al. 1994). Heteroplasmy was not detected in any of the horses examined. The SSCP analysis revealed 14 variants (Table 1), which were consistently obtained in different runs and also in different SSCP conditions. As it can be seen from Table 1, Argentinean Creole and Peruvian Paso horses have five distinct SSCP patterns, and Arabian horses seven patterns. Arabian and Peruvian Paso present six and three diagnostic patterns, respectively (i.e. patterns not present in any other breed), while there are only two SSCP variants found exclusively in Argentine Creole horses. Three SSCP variants were shared amongst the breeds studied, two of them between Argentinean Creole and Peruvian Paso, and the other between Argentinean Creole and Arabian. The most common variants in each of the breeds analysed are unique to that specific breed. Sequence variation A total of 33 individuals from the Argentinean Creole, Peruvian Paso and Arabian breeds were sequenced, corresponding to the 14 SSCP variants found. At least one representative of each breed was sequenced for each one of the variants. When the resulting sequences were not identical for any particular SSCP pattern, all the individuals showing that particular variant were sequenced. Based on the analysis of 381 nucleotides between positions and 15827, 20 haplotypes were found. The 24 South American horses sequenced (ARC and PPA) showed 11 haplotypes. The number of haplotypes seems low compared with other breeds analysed [13 haplotypes in 16 maternal lines of Lipizzan horses (Kavar et al. 1999), 27 haplotypes in 34 Arabian maternal lines (Bowling et al. 2000)]. Although this result could be indicating a bottleneck during the establishment of the New World breeds, more information is needed to assess this phenomenon. Most haplotypes were restricted to a particular breed, but two of them where shared between Argentinean Creole and Peruvian Paso horses from Argentina. Three of the SSCP variants showed two different haplotypes each, with the extreme case of four different haplotypes showing an indistinguishable SSCP pattern (variant 3), differing in up to 15 nucleotides. Thirteen more individuals were sequenced from the Spanish breeds Asturcon, Losino, Potoka, Mallorquina, Menorquina, Spanish Pure breed or Andalusian and Arabian. All but one of these breeds generated new haplotypes. The only exception corresponded to one Andalusian animal (AND1), which shared its haplotype with Argentinean Creole horses (ARC1). This haplotype was the commonest found in Argentinean horses, corresponding to 26 animals showing SSCP variant 2, from which six horses were sequenced. They were not randomly selected, but rather were chosen by pedigree and also to represent three different breeders in the country. Analysis of the sequences showed 61 polymorphic positions, representing 16% of the total sequence obtained. The mean divergence between sequences was 2.8% (range %). All mutations detected corresponded to transitions, with one position (15534) representing an insertion/deletion of a single base pair located in the stretch of six cytosines. Number of animals showing each SSCP variant Table 1 Number and type of SSCP variants detected in each breed. Breed n nv ARC PPA AR ARC, Argentinean Creole; PPA, Peruvian Paso from Argentina; AR, Arabian, n ¼ number of individuals; nv ¼ number of SSCP variants.

4 MtDNA variation in South American and Spanish horses 359 Phylogenetic relationships Table 2 shows a summary of nucleotide diversities (p) and pairwise genetic distances among individuals within and between breeds. An AMOVA analysis showed substantial subdivision between breeds (FST ¼ , P < ), but with a large fraction of the variation (81.85%) found within populations. Furthermore, variation within breeds (x ¼ , SE ¼ ) was as high as variation between breeds (x ¼ , SE ¼ ). The data were first evaluated for phylogenetic information. The distribution of randomly generated trees was left-skewed (g 1 ¼ )0.6998, P < 0.01; Hillis & Huelsenbeck 1992), indicating phylogenetic signal in the data set. Maximum parsimony and neighbour-joining trees showed similar patterns. A heuristic search resulted in three equally parsimonious trees of length 123, 76 steps shorter than the shortest randomly generated tree. The consensus tree is shown in Fig. 1a. Two main clusters of haplotypes can be recognized. The first one includes haplotypes from Arabian (8 out of 11), Peruvian Paso (three out of six), Argentinean Creole (two out of seven), and all Menorquina and Potoka. The second group contains most of Argentinean Creole (five out of seven), half of the Peruvian Paso haplotypes, some Arabian (3 out of 11), Asturcon, one Mallorquina and all Andalusian and Losino. The neighbourjoining tree (Fig. 1b) differs in the clustering of AND3, ARC2, ARB3, PPA4, LO1 and LO2, which in the distance analysis appeared as a clade related to the first group of haplotypes. Although the results of a bootstrap analysis with 1000 replications showed significant probabilities for most of the external nodes, only the second group of sequences resulted in a significant bootstrap value (79%), which is mainly because of shared substitutions at positions 15494, 15496, 15534, and The lack of support of most of the nodes is not surprising given the previous bootstrap values found in most of the phylogenetic analysis of horses published to date, where only a few nodes of the reconstructed phylogeny showed significant probabilities (Kavar et al. 1999; Kim et al. 1999; Bowling et al. 2000). Both maximum parsimony and neighbour-joining analysis showed no clear relationship between the South American breeds (Argentinean Creole and Peruvian Paso) and any of the Spanish breeds analysed, apart from the decisive fact of Argentinean Creole and Andalusian horses sharing one haplotype. In order to compare the results obtained using mitochondrial DNA and previous results based on microsatellite data (Cañon et al. 2000), maximum parsimony and neighbour-joining trees were also constructed for the subset of Spanish horses (trees not shown). Parsimony and distance analysis resulted in identical trees, with bootstrap values higher than 50% in most of the nodes. There was a first cluster of sequences including ARSP, Potoka and Menorquina breeds (bootstrap 51%). Each one of the three breeds represented in this group was monophyletic with bootstrap values higher than 90%. A second group consisted of Losino and one Andalusian haplotype (bootstrap 78%), and a third group included AST, AND1 and MA2 (bootstrap 86%). As in the trees in Fig. 1, the MA1 haplotype did not cluster with any other sequence. The results are quite different from the ones obtained using microsatellites (Cañon et al. 2000), where a clear clustering of the Atlantic breeds (including Asturcon, Potoka, and Losino) vs. a cluster of Mediterranean breeds (Mallorquina and Menorquina) was obtained. We also analyse the sequences reported in this work in a wider context. Vilà et al. (2001) have found that a wide variety of mitochondrial haplotypes of horse breeds clustered in seven different clades with low bootstrap support, indicating a high number of ancestral matrilines of ancient origin. We selected 19 haplotypes (accession numbers: AF326677, AF326676, AF326678, AF356672, Table 2 Nucleotide diversity (p) and mean number and range of nucleotide differences within and between breeds. References as in Table 1. Intra-breed values are in bold. ARC PPA AR AND AST LO PO MA ME p ARC 6.92 (1 13) 8.71 (0 15) (3 19) 7.21 (0 11) 8.71 (4 13) 9.57 (7 13) (6 18) 7.71 (1 13) (9 16) PPA 7.77 (1 15) 11 (4 20) 8.5 (1 14) 11 (5 14) 8.75 (6 13) 11.5 (5 19) 9.12 (3 14) 10.5 (5 17) AR (2 20) (2 17) (6 21) (5 15) (1 17) (2 18) (3 17) AND 10 7 (4 10) 6.5 (2 12) 14 (13 15) 7.5 (2 10) 11.5 (10 13) AST 10 (8 12) 18 (17 19) 9 (6 12) 13.5 (12 15) LO 4 12 (11 13) 9.5 (6 12) 10.5 (8 13) PO 2 14 (11 17) 11.5 (9 14) MA (10 15) ME 3

5 360 Mirol et al. (a) Figure 1 (a) Consensus of the three most parsimonious trees found in the sample of South American and Spanish horses analysed. Figures on the internodes are bootstrap probabilities (in percentage) based on 1000 replications. (b) Neighbour-joining tree based on the Kimura two-parameters distances. Bootstrap values on the internodes. Haplotype names as in the text. AF326674, AF326669, AF064628, AF014416, AF014410, AF072996, AF14413, AF072990, AF072988, AF072992, AF072987, AF014409, AF064632, AF072977, D23666) in order to represent the seven clades described. A maximum parsimony and neighbour-joining analysis were conducted with the total data set of 53 sequences. The neighbour-joining tree is shown in Fig. 2. The structure of the tree is the same as the one obtained in Vilà et al. (2001). The Spanish and South American haplotypes are dispersed along the different clades. Discussion In a previous analysis based on five polymorphic protein loci, Peral García et al. (1996) found a close relationship between Andalusian, Barb, Argentinean Creole horses and Peruvian Paso. Bowling (1994) has also suggested a close association between Andalusian and Argentinean Creole based on blood type markers, although in this case the Nei s distances among all breeds and feral populations examined were so similar that the results were not conclusive. Our results indicate that there is a close relationship between Andalusian and Argentinean Creole horses. It is very noteworthy that ARC1, the most common haplotype found in the South American breed with 26 animals showing SSCP variant 2, is identical to one of the haplotypes found in the Andalusian horses examined. A longer sequence of 468 bp (positions ) was compared between both haplotypes and proved to be identical.

6 MtDNA variation in South American and Spanish horses 361 (b) Figure 1b Continued. The mean number of nucleotide differences between both breeds was 7.21 (range 0 11), the lowest of all interbreed comparisons, and very similar to the mean number of nucleotide differences within the Argentinean Creole horses (6.92, range 1 13). The Nei s genetic distances were higher in all comparisons with Celtic and Arabian horses. With the Peruvian Paso from Argentina, the mean number of nucleotide differences was 8.71 (range 0 15). There were two haplotypes shared between both breeds (CP1 and CP2). The origin of the Peruvian Paso can be traced to the 16th century, from Barb and Andalusian horses brought by Spanish conquerors to Peru, so their close association with the Argentinean breed is not surprising. Furthermore, the Argentinean Peruvian Paso is a young breed, and so, even when stallions were brought from Peruvian Paso in Peru, the mares were frequently not pure Peruvian Paso, and their mtdna was probably of mixed ancestry. The mean number of nucleotide differences between Peruvian Paso and Andalusian is 8.50 (range 1 14), again the lowest of all comparisons and similar to the comparison Argentinean Creole Peruvian Paso and the Peruvian Paso intrabreed differences (7.78, range 1 15). Even when the distances between Argentinean Creole, Peruvian Paso and Andalusian are low, haplotypes representing both Argentinean Creole and Peruvian Paso appeared in the two main clades found in the parsimony and neighbour-joining trees (Fig. 1a,b). This could be interpreted as mixed ancestry or multiple origin of the South American breeds, as was suggested when similar results were obtained in studies of mtdna of Lipizzan horses (Kavar et al. 1999) and Cheju horses (Kim et al. 1999). An alternative explanation could be that the phylogenetic

7 362 Mirol et al. Figure 2 Parsimony tree obtained when some of the sequences reported in Vilà et al. (2001) were included in the analysis. Pleist: remains of horses dated years ago, Anc: remains of horses dated years ago (Vilà et al. 2001). Figures in the internodes represent bootstrap values, and letters indicate the different clades described by Vilà et al. In bold and underlined, sequences from the horse breeds analysed in the present work. reconstruction is reflecting very ancient maternal lineages, present in the Spanish ancestors and so in the South American descendants. This hypothesis is in agreement with the tree depicted in Fig. 2. Haplotypes of Argentinean Creole horses appeared in four different clades (A, C, D and F, nomenclature of Vilà et al. 2001), and Argentinean Peruvian Paso in three clades (A, C and D). Andalusian horses are clustered in clades C and D, which also contain Argentinean Creole and Peruvian Paso. Vilà et al. (2001) propose that modern horse sequences do not define monophyletic groups with respect to wild progenitors, as would be expected if they were founded from a limited wild stock. They assume that the high diversity of matrilines observed suggests the use of a large number of wild populations in the origin of the domestic horse. In this context, the lack of strongly supported phylogenetic relationships among the breeds analysed in this work, indicates the retention of very ancient mitochondrial diversity. The phylogenetic pattern is rather different if microsatellites are used as molecular markers. As the development of distinct breeds usually follows a pattern of very restricted selection based in a few males serving many females, individuals from the same breed are generally clustered together in a phylogenetic tree when microsatellites are used (Vilà et al. 2001). In a previous work, Cañon et al. (2000) studied the genetic structure of Spanish Celtic horses, including Asturcon, Losino, Potoka, Mallorquina and Menorquina, using microsatellites. The genetic distances between breeds

8 MtDNA variation in South American and Spanish horses 363 were in all cases higher than the ones we obtained here using mtdna (Table 2), and many of the breeds were defined as monophyletic groups. They also found a clear clustering among all Atlantic breeds (Asturcon, Losino, Potoka), different from the clade containing the Mediterranean breeds (Mallorquina and Menorquina). This tree is different from the one based on mtdna, where only two of the Spanish breeds (Potoka and Menorquina) constitute monophyletic groups, and there is no association between Atlantic breeds on the one hand, and Mediterranean breeds on the other. This difference is reflecting the maternally dominated genetic flow between breeds and the male-biased selection in the breeds development. In conclusion, we provide the first mitochondrial characterization of South American and Spanish horse breeds. In addition, our results support the very ancient origin of the matrilines in horses, from the perspective of modern New World breeds. Accession numbers GenBank accession numbers for the sequences presented here are AF to AF Acknowledgements This paper was supported by National Research Council and Universidad Nacional de La Plata grants to PPG and FND, and SECyT and Fundación Antorchas grants to PMM. The authors wish to thank Mariana Kienast for the collection of the samples and Jeremy B. Searle and Chris Faulkes for very useful comments. References Avise J.C. (1994) Molecular Markers, Natural History and Evolution. Chapman & Hall, New York. Bowling A.T. (1994) Population genetics of Great Basin feral horses. Animal Genetics 25, Bowling A.T., Del Valle A. & Bowling M. (2000) A pedigree-based study of mitochondrial D-loop sequence variation among Arabian horses. Animal Genetics 31, 1 7. Cañon J., Checa M.L., Carleos C., Vega Pla J.L. & Dunner S. (2000) The genetic structure of Spanish Celtic horse breeds from microsatellite data. Animal Genetics 31, Felsenstein (1991) PHYLIP (Phylogeny Inference Package), Version 3.4. University of Washington, Seattle, WA. Higgins D.G., Bleasby A.J. & Funchs R. (1992) CLUSTAL V: improved software for multiple sequence alignment. Computer Applications in the Biosciences 8, Hillis D.M. & Huelsenbeck J.P. (1992) Signal, noise and reliability in molecular phylogenetic analysis. Journal of Heredity 83, Ishida N., Hasegawa T., Takeda K., Sakagami M., Onishi A., Inumaru S., Kamtsu M. & Mukoyama H. (1994) Polymorphic sequence in the D-loop region of equine mitochondrial DNA. Animal Genetics 25, Kavar T., Habe F., Brem G. & Dovc, P. (1999) Mitochondrial D-loop sequence variation among the 16 maternal lines of the Lipizzan horse breed. Animal Genetics 30, Kim K.-I., Yang Y.-H., Lee S.-S., Park C., Ma R., Bouzat J.L. & Lewin H.A. (1999) Phylogenetic relationships of Cheju horses to other horses breeds as determined by mtdna D-loop sequence polymorphism. Animal Genetics 30, Kimura M. (1980) A simple method for estimating evolutionary rate of base substitutions through comparative studies of nucleotide sequences. Journal of Molecular Evolution 16, Marklund S., Chaudhary R., Marklund L., Sandberg K. & Andersson L. (1994) Extensive mtdna diversity in horses revealed by PCR SSCP analysis. Animal Genetics 26, Maruya E., Saji H. & Yokoyama S. (1996) PCR LIS SSCP (Low ionic strenght stranded conformatioon polymorphism) a simple method for high resolution allele typing of HLA-DQB1, and DPB1. Genome Research 6, Peral García P., Kienast M., Villegas E., Díaz S. & Dulout F. (1996) Estudio de relaciones genéticas entre razas equinas mediante el análisis multivariado. Agro Sur 24 (1), Rodero, A., Delgado, J.V. & Rodero, E. (1992) Primitive Andalusian livestock and their implications in the discovery of America. Archivos de Zootecnia 41, Schneider S., Roessli D. & Excoffier L. (2000) Arlequin: A Software for Population Genetics Data Analysis, Ver Genetics and Biometry Laboratory, Department of Anthropology, University of Geneva, Geneva. Swofford D.L. (1997) PAUP: Phylogenetic Analysis Using Parsimony. Smithsonian Institution, Washington, DC. Tudela J. (1987) El Legado de España a America. Ediciones Pegaso, Madrid. Vilà C., Leonard J.A., Götherström A., Marklund S., Sandberg K., Lidén K., Wayne R.K. & Ellegren H. (2001) Widespread origins of domestic horse lineages. Science 291, Xu X. & Árnason U. (1994) The complete mitochondrial DNA sequence of the horse, Equus caballus: extensive heteropplasmy of the control region. Gene 148, Xu X. & Árnason U. (1996) The complete mitochondrial DNA (mtdna) of the donkey and mtdna comparison among four closely related mammalian species-pairs. Journal of Molecular Evolution 43,