Genetic relationships between six eastern Pyrenean sheep breeds assessed using microsatellites

The knowledge of the genetic composition and relationships among livestock breeds is a necessary step for the implementation of management and conservation plans. This study aims to characterise the genetic diversity and relationships among six sheep breeds of meat aptitude that are spread through the eastern Pyrenees: Tarasconnaise, Castillonnaise and Rouge du Roussillon from France, and Aranesa, Xisqueta and Ripollesa from Spain. All but Tarasconnaise are catalogued as endangered. These breeds do not share the same ancestral origin but commercial trades and gene flow between herds are known to have occurred for centuries. Additionally, two outgroup breeds were included: the Guirra, from a different geographical location, and the Lacaune, a highly selected breed of dairy aptitude. A total of 410 individuals were typed using a panel of 12 microsatellite markers. Statistical, phylogenetic and Bayesian analyses showed that eastern Pyrenean breeds retained high levels of genetic diversity and low, but significant, levels of genetic differentiation (FST = 4.1%). While outgroups were clearly differentiated from other breeds, Pyrenean breeds tended to form two clusters. The first encompassed Tarasconnaise and Aranesa, which probably descend from a common meta-population. The second tended to group the other four breeds. However, none reached high mean Q-values of membership to a discrete cluster. This is consistent with the recent past gene flow between breeds, despite different ancestral genetic origins. The genetic characterisation carried out of the eastern Pyrenean sheep populations provides useful information to support decision making on their conservation and focusing efforts and resources to more singular breeds. Additional key words: genetic diversity; genetic resources; endangered breeds; European sheep. * Corresponding author: jordi.jordana@uab.cat Received: 30-04-14. Accepted: 13-11-14. This work has 2 supplementary tables and 1 supplementary figure that do not appear in the printed article but that accompany the paper online. Instituto Nacional de Investigación y Tecnología Agraria y Alimentaria (INIA) Spanish Journal of Agricultural Research 2014 12(4): 1029-1037 http://dx.doi.org/10.5424/sjar/2014124-6173 ISSN: 1695-971X eISSN: 2171-9292 RESEARCH ARTICLE OPEN ACCESS 1030 A. Ferrando et al. / Span J Agric Res (2014) 12(4): 1029-1037 2013), or types of sheep (Santos-Silva et al., 2008; Kusza et al., 2009). Regardless of the approach used, most of the studies carried out are limited to a single country (Álvarez et al., 2004; Glowatzki-Mullis et al., 2009), while historical relationships among neighbouring sheep populations suggest the need of crossborder analyses (Tapio et al., 2005, 2010; Cinkulov et al., 2008; Kusza et al., 2008; Salamon et al., 2014). Six different local sheep breeds are spread in the eastern Pyrenees area (Fig. 1): Tarasconnaise, Castillonnaise and Rouge du Roussillon in France, and Aranesa, Xisqueta and Ripollesa in Spain (Sánchez-Belda & Sánchez-Trujillano, 1986; Babo, 2000; Esteban, 2003; Avellanet, 2006; Parés, 2008). Generally, they show large differences in origin and morphology. The Spanish Aranesa (census estimated < 2,000 individuals) is closely related to the French Tarasconnaise, and probably shares more ancestral origin with the Merino sheep breed. The Ripollesa breed (< 16,000 individuals) is known to have been formed by the intercross between local Pyrenean sheep and transhumant Merinos. The Xisqueta breed (< 15,000 individuals), meanwhile, belongs to the Iberian Trunk, which derived from the primitive sheep that arrived from Central Asia. The Rouge du Roussillon (< 2,500 individuals) is a red-coated breed originated in Northern Africa, which is known in Spain as “Berberina” (Berber). The Castillonnaise breed has a small population in the Central Pyrenees (< 2,000 individuals), and is known locally as the «red head», because of its skin pigmentation. All of them are classified as endangered breeds, with the exception of Tarasconnaise breed (> 120,000 individuals). The aim of this study is to undertake the characterisation of the European sheep genetic resources via the assessment of the genetic relationships between the eastern Pyrenean sheep breeds. Material and methods Sampling and genotyping Blood samples were obtained from 364 individuals chosen at random among different herds and locations, regardless of gender and age. Individuals belonged to the following eastern Pyrenean sheep breeds: Aranesa (94 individuals; from 24 farms, sampled average 4), Tarasconnaise (44 individuals; from 3 farms, sampled average 15), Xisqueta (97 individuals; from 14 farms, sampled average 7), Ripollesa (53 individuals; from 4 farms, sampled average 13), Rouge du Roussillon (52 individuals; from 10 farms, sampled average 5) and Castillonnaise (24 individuals; from 2 farms, sampled average 12). Additionally, the Spanish Guirra (or Levantina Red), a red-coated breed that originated in Northern Africa, and the French Lacaune, a highly selected breed of dairy aptitude with a high census and widely distributed, were used as outgroup populations of reference (33 and 13 individuals, respectively). In this study, the Lacaune breed acted as a “production” outgroup, while the Guirra breed acted as a geographical outgroup. Altogether, the number of individual sheep analysed was 410. Total DNA was isolated from blood samples following Ausubel et al. (1989). A panel of 12 microsatellite markers (see Table 1) was analysed on all the individuals (Avellanet, 2006). Genotyping was performed on an automatic sequencer, ABI PRISM 3730 (Life Technologies, Carlsbad, CA, USA) using GeneMapper v3.7 software (Life Technologies). Statistical analyses Usefulness of the typed microsatellite set was tested using the FSTAT v. 2.9.3.2 program (Goudet, 2001) by computing the following parameters: number of obserFigure 1. Map illustrating the main areas of location of the analysed eastern Pyrenean sheep breeds. Censuses range from roughly 120,000 individuals belonging to the Tarasconnaise breed, to roughly 2,000 belonging to the Castillonnaise and Aranesa breeds. Abbreviations are as follows: Aranesa (Ara), Tarasconnaise (Tar), Xisqueta (Xis), Ripollesa (Rip), Rouge du Roussillon (RoR) and Castillonnaise (Cas). The location of the Guirra (Gui) sheep breed is also given. Genetic relationships of Eastern Pyrenees sheep 1031 ved alleles (k), effective number of alleles (ke), average number of alleles per locus adjusted to 13 individuals (A13; from El Mousadik & Petit, 1996), within-population observed (Ho) and expected heterozygosity (He), and heterozygote def iciency within population (f) computed following Weir & Cockerham (1984). These parameters were computed at the population level, and also at the marker level to assess the usefulness of typed microsatellite set to achieve the goals of this study. Also, using the FSTAT program (Goudet, 2001) the between-populations FST (θ) matrix was computed following Weir & Cockerham (1984). In all cases, conf idence in the computed values was ascertained by jackknifing over loci using 1920 replicates. The significance level of p < 0.05 was adjusted using the Bonferroni correction for multiple independent tests. The possible presence of genetic bottlenecks in our data was tested performing a two-tails Wilcoxon test, as implemented in the program BOTTLENECK (Luikart et al., 1998; Piry et al., 1999), under the conservative stepwise mutation model. The two-tailed test ensures that statistical significance obtained using onetailed tests for heterozygote deficiency or heterozygote excess is not due to chance (Type I error). Additionally, for short-term evolution and divergence due to genetic drift only (Takezaki & Nei, 1996), both the Nei et al. (1983) DA and the Reynolds et al. (1983) between-populations genetic distance matrices were computed using the POPULATIONS program v.1.2.31 (Langella, 1999). For descriptive purposes, Principal Component Analysis (PCA) was performed on each computed between-populations distance maTable 1. Description of the microsatellite markers used in the current study. Additionally, number of observed alleles (n), effective number of alleles (ke), observed (Ho) and expected (He) heterozygosity are given for each locus type, and as overall mean and standard deviation (s.d.) Locus Chromosome Primers (forwards and reverse) GenBank Reference n ke Ho He accession no. HSC Oar20 F: CTGCCAATGCAGAGACACAAGA M90759 Scott et al. (1991) 17,00 9.94 0.85 0.90 R: GTCTGTCTCCTGTCTTGTCATC INRA49 BTA1 F: GTTTGTATTAGTTTGTGTTCTTTGGC X71588 Vainman et al. (1994) 9,00 2.72 0.57 0.63 R:TTG GCT TCCACAATCACACA MAF65 Oar15 F:AAAGGCCAGAGTATGCAATTAGGAG M67437 Buchanan et al. (1992) 11,00 3.98 0.72 0.74 R: CCACTCCTCCTGAGAATATAACATG MAF214 Oar16 F:AATGCAGGAGATCTGAGGCAGGGACG L38982 Buchanan & 15,00 2.38 0.50 0.58 R:GGGTGATCTTAGGGAGGTTTTGGAGG Crawford (1993) MCM42 Oar9 F:CATCTTTCAAAAGAACTCCGAAAGTG L34281 Hulme et al. (1994) 11,00 3.60 0.65 0.72 R: CTTGGAATCCTTCCTAACTITCGG MCM527 Oar5 F: GTCCATTGCCTCAAATCAATTC L34277 Hulme et al. (1994) 10,00 4.52 0.72 0.77 R: AAACCACTTGACTACTCCCCAA MCM218 Oar4 F: CACTAAAAGCTTATGAAAGTTCCAGC L39828 Hulme et al. (1996) 13,00 3.50 0.67 0.71 R:GATCCTAGCATCAGTCTCCAGATG OARCP49 Oar17 F:CAGACACGGCTTAGCAACTAAACGC U15702 Ede et al. (1995) 24,00 10.04 0.82 0.90 R: GTGGGGATGAATATTCCTTCATAAGG OARCP20 Oar21 F: GATCCCCTGGAGGAGGAAACGG U15695 Ede et al. (1995) 12,00 4.40 0.75 0.77 R: GGCATTTCATGGCTTTAGCAGG OARCP34 Oar3 F:GCTGAACAATGTGATATGTTCAGG U15699 Ede et al. (1995) 10,00 4.69 0.75 0.78 R:GGGACAATACTGTCTTAGATGCTGC OARFCB11 Oar2 F: GCAAGCAGGTTCTTTACCACTAGCACC L01531 Buchanan & 10,00 4.73 0.78 0.78 R:GGCCTGAACTCACAAGTTGATATATCTATCAC Crawford (1993) TGLA53 Oar16 F: CAGCAGACAGCTGCAAGAGTTAGC AC_000173 Zimin et al. (1999) 11,00 7.84 0.80 0.87 R: CTTTCAGAAATAGTTTGCATTCATGCAG GPC_000000185 Mean 12.75 5.19 0.72 0.76 s.d. 4.22 2.61 0.10 0.09 1032 A. Ferrando et al. / Span J Agric


Introduction
Genetic documentation of existing livestock breeds enables the sustainable management and conservation of domestic animal diversity (FAO, 2009).The characterisation of genetic diversity is one of the key points to match future demands of the agri-food markets (Ajmone-Marsan & Globaldiv Consortium, 2010).Molecular techniques (e.g. based on microsatellite markers) can be used to assess within-population genetic diver-sity and between-population differentiation and, therefore, make inferences on the degree of uniqueness of the analysed populations (Groeneveld et al., 2010).Together with the overviews of the sheep genetics scenario in the whole continent (Handley et al., 2007;Peter et al., 2007), during the last decade, such studies have become frequent in European sheep, both considering those sheep breeds spread in a given geographical area (Arranz et al., 2001;Baumung et al., 2006;Ligda et al., 2009;Calvo et al., 2011;Ciani et al.,

Genetic relationships between six eastern Pyrenean sheep breeds assessed using microsatellites
2013), or types of sheep (Santos-Silva et al., 2008;Kusza et al., 2009).Regardless of the approach used, most of the studies carried out are limited to a single country (Álvarez et al., 2004;Glowatzki-Mullis et al., 2009), while historical relationships among neighbouring sheep populations suggest the need of crossborder analyses (Tapio et al., 2005(Tapio et al., , 2010;;Cinkulov et al., 2008;Kusza et al., 2008;Salamon et al., 2014).Six different local sheep breeds are spread in the eastern Pyrenees area (Fig. 1): Tarasconnaise, Castillonnaise and Rouge du Roussillon in France, and Aranesa, Xisqueta and Ripollesa in Spain (Sánchez-Belda & Sánchez-Trujillano, 1986;Babo, 2000;Esteban, 2003;Avellanet, 2006;Parés, 2008).Generally, they show large differences in origin and morphology.The Spanish Aranesa (census estimated < 2,000 individuals) is closely related to the French Tarasconnaise, and probably shares more ancestral origin with the Merino sheep breed.The Ripollesa breed (< 16,000 individuals) is known to have been formed by the intercross between local Pyrenean sheep and transhumant Merinos.The Xisqueta breed (< 15,000 individuals), meanwhile, belongs to the Iberian Trunk, which derived from the primitive sheep that arrived from Central Asia.The Rouge du Roussillon (< 2,500 individuals) is a red-coated breed originated in Northern Africa, which is known in Spain as "Berberina" (Berber).The Castillonnaise breed has a small population in the Central Pyrenees (< 2,000 individuals), and is known locally as the «red head», because of its skin pigmentation.All of them are classified as endangered breeds, with the exception of Tarasconnaise breed (> 120,000 individuals).
The aim of this study is to undertake the characterisation of the European sheep genetic resources via the assessment of the genetic relationships between the eastern Pyrenean sheep breeds.

Sampling and genotyping
Blood samples were obtained from 364 individuals chosen at random among different herds and locations, regardless of gender and age.Individuals belonged to the following eastern Pyrenean sheep breeds: Aranesa (94 individuals; from 24 farms, sampled average 4), Tarasconnaise (44 individuals; from 3 farms, sampled average 15), Xisqueta (97 individuals; from 14 farms, sampled average 7), Ripollesa (53 individuals; from 4 farms, sampled average 13), Rouge du Roussillon (52 individuals; from 10 farms, sampled average 5) and Castillonnaise (24 individuals; from 2 farms, sampled average 12).Additionally, the Spanish Guirra (or Levantina Red), a red-coated breed that originated in Northern Africa, and the French Lacaune, a highly selected breed of dairy aptitude with a high census and widely distributed, were used as outgroup populations of reference (33 and 13 individuals, respectively).In this study, the Lacaune breed acted as a "production" outgroup, while the Guirra breed acted as a geographical outgroup.Altogether, the number of individual sheep analysed was 410.
Total DNA was isolated from blood samples following Ausubel et al. (1989).A panel of 12 microsatellite markers (see Table 1) was analysed on all the individuals (Avellanet, 2006).Genotyping was performed on an automatic sequencer, ABI PRISM 3730 (Life Technologies, Carlsbad, CA, USA) using GeneMapper v3.7 software (Life Technologies).

Statistical analyses
Usefulness of the typed microsatellite set was tested using the FSTAT v. 2.9.3.2 program (Goudet, 2001) by computing the following parameters: number of obser- ved alleles (k), effective number of alleles (k e ), average number of alleles per locus adjusted to 13 individuals (A 13 ; from El Mousadik & Petit, 1996), within-population observed (H o ) and expected heterozygosity (H e ), and heterozygote def iciency within population (f) computed following Weir & Cockerham (1984).These parameters were computed at the population level, and also at the marker level to assess the usefulness of typed microsatellite set to achieve the goals of this study.Also, using the FSTAT program (Goudet, 2001) the between-populations F ST (θ) matrix was computed following Weir & Cockerham (1984).In all cases, confidence in the computed values was ascertained by jackknifing over loci using 1920 replicates.The significance level of p < 0.05 was adjusted using the Bonferroni correction for multiple independent tests.
The possible presence of genetic bottlenecks in our data was tested performing a two-tails Wilcoxon test, as implemented in the program BOTTLENECK (Luikart et al., 1998;Piry et al., 1999), under the conservative stepwise mutation model.The two-tailed test ensures that statistical significance obtained using onetailed tests for heterozygote deficiency or heterozygote excess is not due to chance (Type I error).
Additionally, for short-term evolution and divergence due to genetic drift only (Takezaki & Nei, 1996), both the Nei et al. (1983) D A and the Reynolds et al. (1983) between-populations genetic distance matrices were computed using the POPULATIONS program v.1.2.31 (Langella, 1999).For descriptive purposes, Principal Component Analysis (PCA) was performed on each computed between-populations distance ma-  (Pritchard et al., 2000) was used to ascertain cryptic genetic structure in the analysed dataset.The program gives Bayesian estimates of the natural logarithm of the probability that a given genotype X is part of a given population K (ln Pr(X|K)).As the implemented algorithm uncovers 'hidden structure' without using a priori knowledge about the number of clusters (populations or breeds) present in a dataset, the most likely K-value in the data set was identified according to Evanno et al. (2005) using the Structure Harvester v.0.6.8 website (Earl & vonHoldt, 2012).K was set to vary between 1 and 6, and 20 simulations with different starting points for each K-value.All runs used burn-in and data collection periods of 100,000 iterations.The program was run under the admixture model, considering correlated allele frequencies.
Finally, population structuring was further assessed by computing an Analysis of Molecular Variance (AMOVA), using the program Arlequin 1.1 (Schneider et al., 1997).The statistical significance of the values was estimated by permutation analysis using 1,000 permutations.

Results
The microsatellite set typed allowed a total of 153 alleles to be identified in the whole of the analysed po-pulation.Altogether, the markers set used was highly polymorphic (H e = 0.76 ± 0.09; Table 1).Furthermore, statistically significant deviations of Hardy-Weinberg proportions per locus and breed were few and not consistently found across populations (Table 2).Therefore, these deviations could be due to chance (Type I error).Overall, the marker set typed is sufficient enough to obtain the goals of this research.
F-statistics computed for the whole population were significantly different from 0 (p < 0.01).An average heterozygote deficiency (F IS ) of 2.7% was observed at the studied loci.For F ST values, there was a 4.1% genetic differentiation among the breeds studied, and the overall average heterozygote deficiency (F IT ) was 6.7% Table 3 shows the parameters for the genetic variability of each analysed sheep breed.In the case of the Lacaune breed, values are affected by a very low sample size.The Xisqueta breed had the higher raw, (11.25 ± 3.44) and adjusted for sample size (7.12 ± 2.10), average number of alleles per locus.The observed number of alleles for this breed was about twice that observed for the Castillonnaise and Guirra breeds.Regarding expected heterozygosity, this pattern can still be observed: despite most breeds having H e values of 0.73 or higher, the Xisqueta breed had an expected heterozygosity of 0.76 ± 0.10, while the Castillonnaise and Guirra breeds had values of 0.70 ± 0.10 and 0.68 ± 0.11, respectively.Only the Aranesa, Xisqueta and Guirra breeds showed statistically signif icant heterozygote def iciency, characterised by positive f values (roughly 0.05).
All the computed between-breeds F ST values were statistically signif icant at the signif icance level of p = 0.05 (Table 4).The six breeds located in the eastern Pyrenees showed lowerpaired F ST values, but significant with each other, ranging from 0.011 to 0.053.The Guirra and Lacaune breeds, used as reference populations, showed the highest values among them (F ST = 0.146), and with respect to Pyrenean breeds (ranging between 0.084 and 0.114).In any case, the pairs formed by any combinations of the Aranesa, Tarasconnaise, Xisqueta and Ripollesa breeds showed very low differentiation values.
Under the conservative stepwise mutation model, the two-tailed Wilcoxon signed-rank test gave statistical support to the presence of a genetic bottleneck due heterozygote deficiency in the Guirra (p = 0.013), Ripollesa (p = 0.006) and Xisqueta (p = 0.008) sheep breeds.
Fig. 2 summarises the between-breeds genetic relationships assessed using F ST (Fig. 2A), D A (Fig. 2B) and Reynolds' (Fig. 2C) distance matrices (see also Tables S1 and S2 [pdf online]).The information provided is consistently the same: the eastern Pyrenean sheep breeds and the two outgroups (the Guirra and the Lacaune breeds) were differentiated on the X-axis, while differentiation between outgroups could be found on the Y-axis.However, while the information provided by using F ST and Reynolds' distance matrices was fully consistent, D A allowed the Castillonnaise and the Rouge du Roussillon breeds to be differentiated from the rest of the eastern Pyrenean sheep.
Structuring Bayesian analysis (STRUCTURE) enabled three genetic clusters (K = 3) to be identified in our dataset (Table 5; Fig. S1 [pdf online]).Cluster 1 includes the Guirra (outgroup) individuals; Cluster 2 includes most of the Aranesa and Tarasconnaise genotypes, and roughly a third of the Xisqueta and the Ripollesa and Castillonnaise individuals; and, finally, Cluster 3 includes two thirds or more of the Xisqueta, Ripollesa and Rouge du Roussillon individuals.The Castillonnaise breed individuals are equally assigned to Clusters 2 and 3.
Finally, AMOVA analysis showed that most of the genetic variance is explained by between-individuals variation (95.85%), while the remaining 4.15% is explained by between-breeds differences.These estimates were statistically significant for p < 0.0001.
Table 3. Number of individuals analysed (N), number of observed alleles (k), effective number of alleles (k e ), average number of alleles per locus adjusted to 13 individuals (A 13 ), within-population observed (H o ) and expected heterozygosity (H e ), and heterozygote deficiency within population (f) computed following Weir & Cockerham (1984).Values are given as a mean ± standard deviation.Estimates (f) were obtained using jackknifing over the loci

Discussion
Even though the analysed sheep breeds have limited census and are classified as "at risk" in their respective countries (Babo, 2000;Esteban, 2003), the current populations illustrate that they still retain a noticeable genetic variability.Although they are not directly comparable with the current research due to the different microsatellite set tested, the expected heterozygosity assessed in eastern Pyrenean sheep is comparable with that previously reported in different Iberian sheep sets, ranging from 0.66 (Alvarez et al., 2004) to 0.77 (Arranz et al., 2001).Handley et al. (2007) argue that, when compared with Northern European populations, Southern European sheep breeds keep higher genetic diversity due to their usually higher census sizes and the strong trading movement that Southern sheep populations have experienced during centuries.
In any case, similar scenarios like that of eastern Pyrenean sheep have been previously reported in limited geographical areas gathering sheep populations under low selective pressure and high between-populations gene flow (Tolone et al., 2012).The eastern Pyrenean sheep breeds are found in valleys, historically sharing summer pastures, livestock markets, and having similar external influences due to transhumance.In this respect, it is necessary to highlight the well-documented influence of the Merino breed in the formation of the Aranesa and Ripollesa breeds, and the fact that the Aranesa is considered the Spanish representative of the Tarasconnaise breed, and that the Xisqueta and the Aranesa breeds have actually interchanged individuals for centuries (Babo, 2000;Esteban, 2003;Parés, 2008;Parés et al., 2011).
Even though both the Guirra and the Xisqueta breeds have recently experienced a strong population bottleneck that has given a traceable signal using molecular markers, the genetic scenario of the Ripollesa breed is less clear.The Ripollesa breed is undergoing a selection programme probably leading to the existen-  S1 and S2 [pdf online] for further details).Factor 1 is on the X-axis while Factor 2 is on the Y-axis.
ce of multiple local bottlenecks, at a farm level, that have not been solved via between-farms gene flow (i.e.sharing artificial insemination rams).In any case, since between-breeds differentiation follows, as expected, a geographical pattern (see Fig. 2), the genetic bottlenecks identified are not likely to affect our results.
This scenario of poor genetic differentiation is clearly illustrated by the between-breeds genetic distances computed (Table 4; Table S1 [pdf online]; Fig. 2).While all the eastern Pyrenean sheep breeds are certainly separated from the outgroups, the group formed by the Aranesa, Tarasconnaise, Xisqueta and Ripollesa breeds clearly shows a noticeable genetic identity.While one could expect that the Rouge du Roussillon breed would show the highest differentiation within the eastern Pyrenean group, this was the situation detected for Castillonnaise breed.Even though the Castillonnaise breed could be classified as belonging to the Pyrenean "Entrefino" sheep group by its fleece characteristics (Parés, 2008) and, therefore, it would be expected to be genetically closer to the Aranesa, Tarasconnaise, Xisqueta and Ripollesa breeds, probably the recent history of the breed, in which its inter-crossing with British 'Southdown' individuals is documented (Babo, 2000), has probably lead to some differentiation.Interestingly enough, in spite of their common geographical North-African origin, the Rouge du Roussillon sheep has a genetic differentiation with the Guirra breed similar to those sheep assessed for the other eastern Pyrenean group.This suggests that the genetic background of the North-African red-coat sheep that reached the Mediterranean coasts of Southern Europe was diluted through inter-crossing with local sheep in each area.Similar results, with respect to Rouge du Roussillon breed, were obtained by Parés et al. (2011) in a study of fleece characteristics in Mediterranean red-coat sheep breeds.
This overall scenario was confirmed via STRUCTU-RE analysis.Even though the Aranesa and Tarasconnaise individuals were grouped together into Cluster 2, a significant proportion of the other eastern Pyrenean sheep individuals were in this cluster, and a nonnegligible proportion of Aranesa and Tarasconnaise individuals were assigned to Cluster 3.
In conclusion, the genetic characterisation of the eastern Pyrenean sheep populations carried out provides useful information to support decision making on their conservation.While the partition of the eastern Pyrenean sheep into different breeds explains a significant part of their genetic variation, as suggested by the AMOVA analyses, the analysed populations form quite uniform clusters that could be considered, as a whole, for further trans-boundary initiatives for livestock diversity conservation (Hoffmann, 2011).Our results also confirm the historical information suggesting that both the Tarasconnaise and the Aranesa breeds belong to a common meta-population.However, the current analysis also shows that there has been a signif icant gene flow among neighbouring sheep breeds and, therefore, a comprehensive understanding of the eastern Pyrenean sheep is necessary before the implementation of conservation strategies in such populations.

Figure 1 .
Figure 1.Map illustrating the main areas of location of the analysed eastern Pyrenean sheep breeds.Censuses range from roughly 120,000 individuals belonging to the Tarasconnaise breed, to roughly 2,000 belonging to the Castillonnaise and Aranesa breeds.Abbreviations are as follows: Aranesa (Ara), Tarasconnaise (Tar), Xisqueta (Xis), Ripollesa (Rip), Rouge du Roussillon (RoR) and Castillonnaise (Cas).The location of the Guirra (Gui) sheep breed is also given.

Figure 2 .
Figure 2. Distribution of the analysed sheep populations according to the results of the Principal Component Analysis performed on the between-populations F ST (Plot A), D A (Plot B) and Reynolds' (Plot C) distance matrices (see TablesS1 and S2[pdf online] for further details).Factor 1 is on the X-axis while Factor 2 is on the Y-axis.

Table 1 .
Description of the microsatellite markers used in the current study.Additionally, number of observed alleles (n), effective number of alleles (k e ), observed (H o ) and expected (H e ) heterozygosity are given for each locus type, and as overall mean and standard deviation(s.d.) trix, using the SAS/STAT TM program (SAS Inst.Inc., Cary, NC, USA), to summarise the contents of the information.Only factors with an eigenvalue >1 were retained, and their scores were used to construct a multidimensional plot using Microsoft Excel TM .The Structure program

Table 4 .
Between-breeds F ST (θ) distance values.All the values are significant at p < 0.05

Table 5 .
Proportion of membership of each sampled population in each of the three clusters inferred in the most likely run using the STRUCTURE program.Proportions of membership higher than 0.5 are in bold