Recently, arrays of DNA based markers have been developed, both to undertake studies of genetic variability, and to investigate genetic relationships between populations (Bradley et al., 1996; Cañon et al., 2000; Solis et al., 2005). Among these, microsatellites are considered by many to be the most suitable marker system for evaluating breed genetic diversity (Takezaki and Nei, 1996). Several genetic studies of equine populations have described the usefulness of microsatellite markers, as well as blood group and biochemical loci, for establishing genetic relationships between populations, and for describing genetic variability among and within breeds (Cothran et al., 1998; Cañon et al., 2000; Aranguren-Mendez et al., 2002; Juras et al., 2003; Aberle et al., 2004; Gupta et al., 2005; Glowatzki-Mullis et al., 2005; Luis et al., 2007; Royo et al., 2007). Nowadays, however, these are being progressively replaced by SNP markers, in, for example, the control of relationships (e.g. IBD matrix) (Flury et al., 2010). Nevertheless, in small breeds, genotyping with high density SNP chips turns out to be very expensive, thereby limiting the availability of such data.

Historically, Greece has had a profound influence on the development of Western civilization. However, little is known, not only as to the relationships of Greek domestic animals to other foreign breeds, but also to their own origins and breeding histories. Greece, due to its geographic location, is a crossroad between three continents. Furthermore, its mountainous terrain and numerous islands have lead to the creation of many small populations of distinct and isolated breeds. On the basis of phenotypes and genotypes, seven native horse breeds have been identified so far (Zafrakas, 1991; Alifakiotis, 2000), namely the Crete, the Elis Mountain (or Pinias), the Elis Valley (or Andravidas), the Skyros, the Thessalias, the Pindos and the Zakynthos. The names attributed to these breeds are those of the various regions where they were originally preponderant (Alifakiotis, 2000). Among these, the phenotype in the Skyros pony is distinct from those of the other breeds (Zafrakas, 1991). This in itself is a small-sized animal, with an average adult height of 109 cm in stallions and 107 cm in mares. They are mainly bay-colored with dark and strong hooves, and very long manes and tails (Alifakiotis, 2000). As shown by Apostolidis et al. (2001), this phenotypic difference seems to be linked to the Skyros pony being genetically less similar to the other Greek horse breeds than these themselves to one another. However, the literature is very poor concerning the description of Greek horse breeds in general. The Skyros pony population, in particular, has never been studied in its entirety, as neither have the genetic relationships of Greek horse breeds as a whole, either mutually or with other domestic breeds elsewhere.

Population sizes in the Greek breeds are estimated to range from 50 to over 1000 individuals (Alifakiotis, 2000; DAD-IS 2007). Thus, most can be considered as small, and, according to criteria established by the FAO (FAO, 1998; DAD-IS, 2007), in a critical- or endangered-maintained risk status. Genetic principles, when applied to a small population, indicate that the genetic variability of such a population will decrease across generations, with the consequential need for increased conservation measures. Genetic variability may be defined as the ‘genetic ability to change’, and therefore, the capacity to respond to environmental variation and future needs (Rochambeau et al., 2000). Thus, the evaluation of genetic variability is one of the first steps in the process of species genetic conservation, in accordance with the hypothesis of correlations between preserving both genetic variability and population viability. The analysis of the information contained in registered pedigrees can also contribute towards knowledge on population structure and the evaluation of genetic variability (Valera et al., 2005).

In comparison to other studbooks, the Skyros pony preliminary studbook proved to be much less complete. It is characterized by a very high percentage of animals with one or both parents unknown (26.33% and 35.45% as against 1.94% and 1.28% for the Andalusian studbook) (Valera et al., 2005). This situation is explained by long generation intervals, births having only recently been recorded, and mares roaming free, to return pregnant, the sire being obviously unknown. In comparison, the percentage of known ancestors in the Lipizzan studbook, for example, was above 90% at the 10^th generation and above 70% at the 14^th (Curik et al., 2003). This value is comparable to the first generation in the Skyros studbook, although the situation is improving, with the value of geq globally increasing according to the birth-year of the animals. Generation intervals computed for the Skyros breed were lower than those reported for other horse breeds with deeper pedigrees, as the 9.7 years in French Arabs and 11.8 in Trotteur Français (Moureaux et al., 1996). Even so, this is very high for an endangered breed. Generation intervals in horses are commonly long (Strom and Philipsson, 1978), this basically depending on its use (leisure or racing) being incompatible with pregnancy and a breeding life. For the Skyros pony, the cause is more linked to management, with 60 ancestors being accountable for about 100% of genetic variability. This value was lower than the 331 reported for Andalusian horses (Valera et al., 2005). Although the values for the number of ancestors explaining 70% (10) and 50% (5) of the genetic variability are quite similar (13 and 6, respectively for Andalusian), the lack of difference between f_e (13.3) and f_a (13.1) showed that, based on pedigree, no significant bottleneck had occurred. N_g was low due to the high probability of gene loss in the last generation, as a result of few descendants (the number of births has been globally declining since 1997), and the repeated use of the same individuals for breeding.

Parameters computed from the results of DNA analysis proved to be similar to those calculated for other breeds, especially for bottlenecked and small-sized populations. On a whole, these parameters showed higher or similar values than those obtained by Avdi and Banos (2008), consistent with the fact that we studied the entire Skyros pony population, instead of just one herd. MNA (5.93) was lower than that presented by Rognon et al. (2005) for seven French horse breeds, where the lowest value was for the Comtois (6.36). Nevertheless, Ae (3.22) was within the range of reported values, i.e., from 3.05 for the Breton to 3.94 for the Selle Français. As with the Skyros pony, both the Comtois and Breton are local breeds. Thus, the value obtained for Skyros ponies was among the lowest observed, which is easily explained by the very limited size of the population (genetic drift effect). A number of studies have reported similar values for He and Ho in horse breeds, based upon microsatellite loci (Cañon et al., 2000; Curik et al., 2003; Aberle et al., 2004; Juras and Cothran, 2004; Gupta et al., 2005; Rognon et al., 2005; Luis et al., 2007). The number of loci tested ranged from 11 (Rognon et al., 2005) to 30 (Aberle et al., 2004). As there was no instance of exactly the same set of loci as ours being employed, a direct comparison becomes impossible, although these results are nevertheless useful for a better understanding of variation in the Skyros. The value of He (0.621) for the Skyros horse was well within the range for domestic horses, as a whole, although it was at the lower end of the range. The lowest values, viz. 0.442 for the Friesian (Juras and Cothran, 2004), 0.506 for the Sorraia and 0.609 for the Exmoor (Luis et al., 2007), were all from breeds with either small population-size or recent bottlenecks. The same pattern was seen for Ho. Thus, levels of heterozygosity in the Skyros breed are most like those observed in horse breeds with small population size that have undergone bottlenecks and inbreeding in recent times, which is consistent with the recent history of the Skyros horse. Actually, it is known that the population size has decreased, as confirmed by bottleneck-analysis of two of the three sub-populations. Nevertheless, no bottleneck signature was detected by testing in the population present on Skyros, even though the probability is high that this population has undergone outstanding reduction of late. There are five possible explanations (Luikart et al., 1998): 1) although a bottleneck occurred in the past, possibly more than 12 to 15 generations ago, it did not constitute an immediate and permanent bottleneck in population size (Luikart and Cornuet, 1998). 2) the bottleneck was too small to be detectable (Luikart et al., 1998). 3) either insufficient polymorphic loci were sampled to acquire the required statistical power for detecting the bottleneck, or the individuals sampled were not representative of the bottlenecked population itself. 4) a demographic, and not a genetic bottleneck, occurred. 5) the bottlenecked population was incompletely isolated, and so, genes obtained from immigrants (e.g., rare alleles) obscured subsequent genetic effects. In this case, these immigrant genes could have originated from the white horses present on the island, and which had been introduced more recently than the original pony. Hypothesis 1, 3 and 5 were, in this case, the most plausible explanations, with preference for the first, since 12–15 generations ago falls into the time of the foundation of modern horse breeds. However, no sufficient information was available to choose or definitely exclude either one or the other of these assumptions.

However, the relatively high level of heterozygosity and PIC values was comparable to those found in the Marwari horse population. This reflected high residual genetic variability that could be exploited for planning breeding strategies and giving precedence to this breed for conservation measures (Gupta et al., 2005).

As to the relationship between the Skyros pony and other Greek horse breeds, this study confirmed the conclusions by Apostolidis et al. (2001), regarding the former. Levels of genetic variability among Greek horse breeds, in general, were all within the range seen for other domestic horses. Values for Ho of biochemical loci varied widely, with the lowest (0.307) found in the Pinias breed, a farm horse encountered in mountainous regions. This breed is relatively numerous, with a census population in Greece of about 5,000. The highest Ho was found in the Skyros pony from the island of the same name. As the census numbers of this small horse are less than 200, this high Ho was unexpected. Nevertheless, the Ho for microsatellites in this breed was the lowest in the four Greek breeds studied, and was even lower than the mean value for domestic breeds, as a whole. In general, there was no clear pattern of genetic variation associated with population size or degree of geographic isolation. This is most likely due to historical factors, such as how recently changes in population numbers took place or undocumented cross-breeding. Furthermore, individual genetic variation at biochemical loci does not correlate well with that at microsatellite loci. Variability at microsatellite loci is largely affected by the number of alleles, and, based upon demography, may change more rapidly than that at protein loci (Luis et al., 2007). In the Crete horse, another island population, the opposite pattern of variation is the case, with relatively low values for protein loci but relatively high variation at microsatellite loci.