Indigenous Greek Horse Breeds: Genetic Structure and the Influence of Foreign Breeds

Abstract

This study aims to examine the genetic structure and diversity levels of seven indigenous Greek horse breeds: Andravida, Pindos, Thessaly, Skyros, Penia, Messara and Rodos, using 15 microsatellites. Phenotypic traits were combined with factorial correspondence analyses to create two datasets: one “Baseline” containing typical samples, and one “Unknown” with non-typical or of disputed origin samples. In the Greek “Baseline” horses, 142 alleles were found. The mean observed and effective number of alleles, the polymorphism information content and the allelic richness were 6.75, 4.14, 0.63 and 5.12, respectively. The expected and observed heterozygosity and inbreeding coefficient varied between 0.81 and 0.29 and 0.79 and 0.24. The above dataset was enriched with data from 41 foreign horse breeds and 40 Przewalski samples to perform a breed assignment. The highest percentage of successfully assigned samples was for Skyros, Messara and Rodos, with rates of 93%, 89% and 100%, respectively, suggesting their considerable homogeneity, while Andravida, Pindos, Thessaly and Penia scored 32.5, 34.1, 44.0 and 45.7%, respectively. Structural analysis confirmed the results of FCA and showed the genetic similarities of the above breeds. The results revealed the influence of foreign breeds (mainly Garrano, Turkoman, Irish Draft, Hanoverian and Belgian Draft). There is an urgent need to implement management measures for the pure homogeneous breeds and selection strategies for the remaining breeds which are genetically similar.

1. Introduction

Indigenous Greek horse breeds have generally been little or poorly studied and are poorly documented in comparison to other European breeds. Greek horse breeds lack worthy phenotypic and genetic standards. Most of them do not have detailed studbooks. Therefore, there are no reliable, objective criteria for the assessment of an individual’s breed, whether based on specific phenotypic characteristics or on the results of a genetic analysis. This fact has a direct impact on the feasibility of developing an effective plan for the management and preservation of the horses in question, as these must rely exclusively on the subjective assessments of the horse’s appearance and undocumented ancestry information provided by the owner. As a result, it is not possible to accurately determine or estimate their population size and their conservation status with precision.

Furthermore, inadequate equine breeding management frequently leads to inbreeding. In other circumstances, the unintentional crossbreeding with other breeds in herds comprising a combination of sexes and breeds, is not uncommon. It is also not unusual for owners to cross indigenous mares with stallions of foreign or unknown origin with the intention of acquiring offspring that are perceived to be of a superior quality. Some common misconceptions further contribute to the problem, such as using a single favorable characteristic like gaitedness as proof of a horse’s origin, although that trait is a simple genetic characteristic—a single base mutation in the DMRT3 gene present in numerous breeds worldwide [1].

Domestic horses started being introduced into the Balkans from the Eurasian steppes around 3000 BCE [2]. During the next millennium, they spread, even reaching the island of Crete [3]. After the 9th century Islamic expansions into Europe and Asia, there was an increased prevalence of horse lineages with Persian roots [4]. During the Ottoman Empire, stallions of Eastern origin were also imported into the Greek peninsula [5]. Since the foundation of the modern Greek state, the indigenous populations have received influences from many modern European and other widespread breeds that have been and continue to be imported to Greece. Recorded horse imports have included the following: German Hannoveraner Schleswig Holstein stallions, during the reign of King Otto and his Bavarian government (mid-19th century); Breton, Anglo-Arabian and Syrian Arabian stallions for military purposes after the Greek–Ottoman wars (1912–1922); and Polish Huzul, Austrian Lipizzaner, French Anglo-Arabian and Hungarian Nonius in the following two decades [5,6]. On the other hand, many autochthonous Greek horses are characterized by their small stature, resilience, and lack of genetic improvement. Their survival is at stake mainly due to abandonment and substitution with popular improved breeds, similar to other endangered isolated primitive European breeds, such as the mountain horse populations of the neighboring Balkans [7,8], and the distant Iberian Portuguese Garrano horses [9], an autochthonous mountain pony from northwestern Portugal and Galicia.

Currently, seven indigenous breeds are officially recognized and protected by the state: Andravida or Elia or Pedini Elia, Pindos, Thessaly, Skyros or Skyros Small Horse, Penia or Orini Elia, Messara or Cretan and Rodope [10]. The Skyros breed has a detailed breed standard and studbook, in which all breed phenotypic characteristics and traits are documented. Pedigree information is acquired via DNA microsatellite (Msat) parental testing performed by the Skyros Horse Breed Association. The environmental restrictions of Skyros Island combined with the breeds’ distinct phenotype facilitate keeping the herds isolated and purebred. The Ministry of Rural Development and Food, through the animal genetic resource centers by region, keeps rudimentary studbooks for the remaining state-protected breeds, consisting mainly of information about horse identity and pedigree information according to the breeder’s claims. The absence of precise breed standards for indigenous breeds gives rise to ambiguous registration criteria, which are based on subjective data. Ultimately, this does not facilitate the management of the breeds. Furthermore, in the Greek mainland, in many instances, animals of mixed breeds are frequently observed. Thus, in the absence of genetic testing, parentage information is of questionable credibility. The Rodope horse was classified as a distinct breed in 2024, having previously been considered a subpopulation of the Pindos breed. Until recently, there were six officially recognized breeds: Andravida, Pindos, Thessaly, Skyros, Penia and Messara, while bibliographical and research data suggest the existence of more breeds or subpopulations, like Zakynthos [11], a non-descript subpopulation of the Andravida breed and Lesvos gaiter [12]. There is also strong evidence for the existence of an isolated rare breed of miniature horses in Rhodes Island, which had a bottleneck of only six horses (three males and three females) at the time of their discovery and rescue from a feral state 25 years ago [13]. The state has not recognized the Rodos small horse as a breed. The population is currently being managed in captivity and has since doubled in size due to the efforts of the local club for the protection of the Rodos horse, which keeps a well-documented studbook [13]. Overall, except for Pindos, indigenous Greek horse breeds are classified as endangered, while a few small, isolated populations remain unrecognized and are on the verge of extinction [14].

The application of contemporary molecular and technological tools is necessary to facilitate the study and survival of indigenous Greek horse breeds [15,16]. A species-specific selection of highly variable molecular markers, like DNA Msats, is needed. For equine species, the loci set commonly used for Msat analysis includes most or all of the following 17 loci: AHT4, AHT5, ASB2, ASB17, ASB23, CA425, HMS1, HMS2, HMS3, HMS6, HMS7, HTG4, HTG6, HTG7, HTG10, LEX3 and VHL20 [16]. Combined with adequate sample sizes, these can reveal essential information about an individual’s genetic makeup and its origin population, as well as about the level of structure and differentiation within and among populations [16,17,18]. If a genetic database is available, the WhichRun 4.1 breed assignment method can be used to identify the most probable origin population of a sampled individual. Specifically, this method is highly accurate when genotypic data are available from an appropriate set of loci with enough samples per population (≥20 individuals) [19,20].

Subsequently, it is essential to investigate the genetic identity of the Greek horses and build a database of the individual genetic profiles of each breed in order to provide reliable and objective means of identification, assisting in their management and protection. This study is part of a PhD dissertation aiming to achieve the genetic characterization of Greek breeds through the investigation of their genetic structure, genetic relationships with foreign breeds and the potential of individual breed allocation using Msats, for preservation purposes.

2. Materials and Methods

2.1. Ethical Consideration

All experimental procedures were approved by the Bioethical Committee of the Agricultural University of Athens under the guidelines of Council Directive 86/609/EEC regarding the protection of animals used for experimental and other scientific purposes.

2.2. Sample Collection

We collected rooted hair samples from 616 indigenous horses (10 to 20 rooted hairs from each horse) from seven indigenous Greek horse breeds/populations: Andravida (n = 74), Pindos (n = 172), Thessaly (n = 35), Skyros (n = 48), Penia (n = 79), Messara (n = 198) and Rodos (n = 10). The sampling began in September 2021 and ended in December 2023. Sampling was carried out according to previous breed descriptions and the scarce available parentage information [21,22]. Sampling closely related individuals was avoided. The sampled horses can be divided into two main distinct phenotypes according to their withers height (WH). Five breeds consist of horses with a WH of 130 to 165 cm: Andravida, Pindos, Thessaly, Penia and Messara, from which two are gaited (Penia, Messara) and three are non-gaited (Andravida, Thessaly, Pindos). The remaining two breeds, Skyros and Rodos, are miniature horses, with a WH of 105–115 cm. The sampling locations for the studied breeds can be seen in Figure 1.

The Andravida horse is the largest Greek breed, exhibiting typical warm-blooded traits. It is an improved, athletic horse, with a WH of 150–165 cm. It can be found mainly in the Peloponnese and Zakynthos Island, and its conservational status is “Critically endangered” [21,22]. The Pindos breed (here including the Rodope subpopulation) is the most abundant Greek breed, rated as “Vulnerable” and is a characteristically primitive, cow-hocked, hardy mountain pony with a WH of 120–140 cm that can be found in most of the mountains of central and northern Greece, and even in some islands [22,23]. Generally, it is considered a non-gaited breed [22,23]; however, there are a few individuals that gait, a characteristic recently introduced by some breeders due to the rising popularity of gaited horses in Greece. It is worth noting that Pindos, due to its wide distribution in varied, remote regions, is expected to consist of an unknown number of geographically isolated subpopulations, and thus, their proportionate representation in the dataset is challenging. The Thessaly breed is the second largest Greek horse breed, from the homonymous region, with a WH of around 140–150 cm [22,23] and a low percentage of the population is gaited, possibly due to crossbreeding with Penia. Its conservation status is unknown (“Data deficient”) [22]. The Skyros is the most famous and protected Greek horse breed, a miniature cow-hocked horse from Skyros Island with WH of 102–115 cm, and its status is “Critically endangered” [22,23,24]. The Penia is a gaited breed originating from the mountains of the Peloponnese, with a WH of 125–145 cm and is considered as “Endangered” [22,23]. The Messara is also gaited, originates from Crete (plain of Messara), has a WH of 132–142 cm, is post-legged and is “Critically endangered” [22,23]. Lastly, the Rodos population, a well-proportioned, cow-hocked miniature horse with a WH of around 110 cm consists of only 14 horses on Rhodes Island and its conservational status is “Critically endangered” [22]. Individual examples of each of the studied breeds can be seen in Figure 2. Additional information on the Greek breeds’ phenotypes and status is available in Supplementary Table S1.

Samples were stored in plastic bags in dry conditions at room temperature, and then submitted to Texas A&M University Animal Genetics Laboratory. DNA isolation from the hair follicles was achieved using PUREGENE DNA purification kit (QIAGEN, Germantown, MD, USA) according to the manufacturer’s protocol and analyzed for 15 species-specific horse Msat loci (VHL20, HTG4, AHT4, HMS7, HTG6, AHT5, HMS6, ASB2, HTG10, HTG7, HMS3, HMS2, ASB17, ASB23 and LEX33) using multiplex PCR.

Table 1. Microsatellite loci, size and location.

Locus	Amplicon Length (bp)	Chromosome Location
VHL20	89–109	30
HTG4	127–141	9
AHT4	148–164	24
HMS7	173–187	1
HTG6	84–106	15
AHT5	130–146	8
HMS6	159–171	4
ASB2	222–254	15
HTG10	93–113	21
HTG7	120–130	4
HMS3	150–172	9
HMS2	284–304	10
ASB17	93–121	2
ASB23	183–217	3
LEX33	195–221	4

shows the analyzed DNA Msats, their chromosomal location and base pair (bp) length. Msat genotyping was performed using an ABI PRISM 377 (Applied Biosystems 3730 xl, Foster City, CA, USA) according to the procedures described by Juras et al. [25], and we used the nomenclature of the International Society for Animal Genetics.

2.3. Statistical Analysis

The genotypic data were analyzed using several available free software packages: Create 1.33 [26] was used for data editing and input file preparation or conversion. Prior to the statistic and breed assignment tests, we used factorial correspondence analysis (FCA) based upon allelic frequencies (GENETIX 4.05.2) [27] to investigate the genetic structure within the sampled populations. Then, the Greek horse samples were divided into two subgroups: the first group consisted of the most typical representatives of each breed (“Baseline” n = 385 samples from 7 Greek breeds), while the second one included the less typical samples, namely those of uncertain phenotype or origin, or that showing marked deviation from the rest of their breed on the FCA diagrams (“Unknown” n = 231 samples from 6 Greek breeds). The detailed composition of each dataset is shown in

Table 2. Greek horse samples.

	Greek Samples
Breed	Baseline	Unknown	All
Andravida	40	34	74
Pindos	129	43	172
Thessaly	25	10	35
Skyros	41	7	48
Penia	46	33	79
Messara	94	104	198
Rodos	10	0	10
Total	385	231	616

. Only the “Baseline” set was used for statistical analysis, while both datasets were assessed in breed assignment.

We estimated the allelic frequencies, number of alleles (Na) and effective number of alleles (Ne), we tested for the Hardy–Weinberg equilibrium, and calculated the observed and expected heterozygosity (Ho and He), the allelic patterns (including No. of different alleles with a frequency ≥5% (Na (freq ≥ 5%), No. of private alleles, No. of locally common alleles (freq. ≥ 5%) found in 25% or fewer populations and No. of locally common alleles (freq. ≥ 5%) found in 50% or fewer populations) and F-statistics per locus and pairwise differentiation indices (F_ST) with GeneAlEx 6.5 [28]. Fixation index (Fis) values and allelic richness (A_R), according to El Mousadik and Petit [29] based on the minimum sample size of 10 diploid individuals and were calculated using Fstat 2.9.4 [30]. The polymorphism information content (PIC) for each locus was calculated according to the formula of Botstein et al. [31] with Cervus [32]. To further investigate the genetic homogeneity of the sampled Greek populations and the influence of indigenous or foreign breeds on their gene pools, we used a breed/population assignment method based upon the genotypic data of the sampled individuals using WhichRun 4.1 [33] (jackknife = on and off, 1st most likely, options: critical pop = off, locus coefficient = 10).

An additional dataset of 41 foreign horse breeds plus one set of Przewalski samples as an outgroup (n = 40) (total of 42 populations and n = 2362 samples), provided by Texas A&M University Animal Genetics Laboratory, was added to the “Baseline” dataset, containing foreign breeds that have possibly contributed to the Greek horses’ gene pool, as well as many other widespread horse breeds. The foreign breeds’ dataset contained 25 to 129 samples per breed. More information about the foreign and Greek datasets can be found in Supplementary Table S2.

Additionally, a structural analysis was realized using STRUCTURE 2.3.4 [34] and STRUCTURE SELECTOR [35] to select the best K to analyze the genetic structure of Greek horse populations and explore the genetic differences or similarities among and within them. pophelper [36] was used for the visualization of the results. Finally, a neighbor-joining phylogenetic tree was constructed using PHYLIP 3.65 [37] and visualized using Iroki online tree viewer [38] to graphically illustrate the genetic relationships among breeds.

3. Results

3.1. Genetic Diversity

All loci were amplified successfully for the Greek samples, with no missing alleles. The allele frequencies for each locus and population of the “Baseline” dataset (populations = 49, n = 2747) are available in Supplementary Table S3. Summary results of the Chi-Square tests for the Hardy–Weinberg equilibrium can be found in Supplementary Table S4. The H-W tests per locus and population revealed that only the Pindos breed exhibited notable significant deviations from the H-W equilibrium at six of the fifteen loci. Other Greek populations had up to three loci showing H-W deviations. Across the Greek populations, 142 different alleles were found, the mean number of alleles $\overline{\mathrm{N}\mathrm{a}}$ was 6.75, the mean effective number of alleles $\overline{\mathrm{N}\mathrm{e}}$ was 4.14, the mean allelic richness $\overline{{\mathrm{A}}_{\mathrm{R}}}$ was 5.12 and the mean polymorphism information content $\overline{\mathrm{P}\mathrm{I}\mathrm{C}}$ was 0.63.

The mean genetic variability estimations per population for the total of 49 populations are shown in detail in Supplementary Table S5. A graphical representation of the main variability estimators and allelic patterns is shown in Figure 3. For the Greek dataset samples, the observed number of alleles (Na) ranged from 1.73 (Rodos) to 9.2 (Pindos), the effective number of alleles (Ne) from 1.44 (Rodos) to 4.99 (Pindos), polymorphism information content (PIC) from 0.19 (Rodos) to 0.75 (Pindos, Andravida) and allelic richness (A_R) from 1.73 (Rodos) to 6.26 (Thessaly). A comparison of variability measures and allelic patterns across populations depicts the genetic potential of the Greek group, which tended to have high values, apart from the Rodos population. The Pindos breed had the highest A_R, Ne, Na and PIC across the 49 populations. Penia had the second highest values within the Greek group. The highest values of mean He and mean Ho were observed in the Andravida breed (0.81 and 0.78, respectively), closely followed by the Thessaly, the Penia and the Pindos, while the lowest values were found in Rodos (0.29 and 0.23, respectively) (see Supplementary Table S5). The highest Fis value was +0.07 in the Pindos and Thessaly breeds. For the Pindos breed with the largest sample size (n = 129), the excess of homozygotes may have been due to the Walhund effect. That is, the existence of subpopulations within the Pindos breed that do not communicate with each other and are differentiated in terms of gene frequencies. Possibly, the same reason could explain the high Fis values in the Thessaly breed. On the other hand, the lowest value was found to be −0.15 in the Rodos population, but this should be interpreted with caution due to the very small population and sampling size (10 samples). The Fis values for the Messara, Penia, Skyros and Andravida breeds were found to be near 0 (−0.03 to +0.01), suggesting neither significant inbreeding nor outbreeding.

3.2. F-Statistics

Table 3. Genetic differentiation among Greek Baseline horse breeds (n = 385) per locus (F-statistics, inbreeding coefficient F_IS, overall fixation index F_IT, differentiation index F_ST), s.e.m = standard error of the mean (s.e.m = s.d./√nloci). Background colors’ scale specification: the darker the color, the higher the values.

Locus	FIS	FIT	FST
VHL20	−0.01	0.10	0.11
HTG4	−0.01	0.23	0.23
AHT4	−0.01	0.15	0.16
HMS7	0.03	0.15	0.13
HTG6	−0.06	0.00	0.05
AHT5	−0.04	0.07	0.10
HMS6	0.03	0.16	0.13
ASB2	0.00	0.12	0.12
HTG10	−0.01	0.09	0.10
HTG7	−0.01	0.18	0.19
HMS3	−0.03	0.14	0.16
HMS2	−0.03	0.09	0.11
ASB17	−0.01	0.14	0.15
ASB23	0.01	0.16	0.16
LEX33	0.01	0.17	0.17
mean ± s.e.m	−0.01 ± 0.01	0.13 ± 0.01	0.14 ± 0.01

presents, in the form of a heat map, the F-statistic analysis results per locus over the seven Greek Baseline horse breeds (n = 385). Across all loci, the mean F_IS value was −0.01, the mean F_IT was 0.13 and the mean F_ST was 0.14. In Greek populations, F_IS ranged from −0.06 in HTG6 to +0.03 in HMS6 and HMS7, showing that HTG6 was the most heterozygous locus while HMS6 and HMS7 were the least. F_IT ranged from 0.00 (HTG6) to 0.23 (HTG4) and F_ST ranged from 0.05 (HTG6) to 0.23 (HTG4), both showing HTG4 as the most differentiated of the 15 loci and HTG6 as the least differentiated among the studied Greek horse populations. Across all loci, the degree of genetic differentiation among the sampled Greek horse breeds was moderate (F_ST = 0.14 ± 0.01) in relation to other horse breeds [39].

Pairwise differentiation F_ST values can be seen in

Table 4. Heat map of pairwise estimations of genetic differentiation F_ST between seven indigenous Greek horse breeds using 15 microsatellite markers. Background colors’ scale specification: the darker the color, the higher the F_ST values.

	Andravida	Pindos	Thessaly	Skyros	Penia	Messara	Rodos
Andravida	0.000
Pindos	0.014	0.000
Thessaly	0.016	0.013	0.000
Skyros	0.057	0.054	0.059	0.000
Penia	0.018	0.015	0.016	0.070	0.000
Messara	0.042	0.040	0.035	0.101	0.034	0.000
Rodos	0.229	0.212	0.228	0.289	0.220	0.248	0.000

(heat map table) among the Greek Baseline populations. Among all 49 populations, the pairwise F_ST values are presented in more detail in Supplementary Table S6. Among the Greek populations, the highest differentiation was between Rodos and Skyros (F_ST = 0.289) and the lowest between Pindos and Thessaly (F_ST = 0.013). Overall, Rodos had high F_ST values (0.212–0.289) in all combinations with the other breeds, followed by Skyros (F_ST from 0.054 to 0.101) and then Messara (F_ST from 0.034 to 0.101). All other breed pairs (among Andravida, Pindos, Thessaly and Penia) showed very low differentiation values (F_ST from 0.013 to 0.018) [39].

3.3. Factorial Correspondence Analysis

The FCA diagram in Figure 4 shows a clear differentiation of the Skyros, Messara and Rodos populations, while the Andravida, Pindos, Thessally and Penia seem to overlap and occupy common space.

3.4. Breed Assignment

The results of the breed assignment of the Greek “Baseline” subgroup (typical samples) can be seen in Figure 5 and Supplementary Table S7, while the results of the “Unknown” subgroup (non-typical, of doubted origin or shown as outliers in previous FCA diagrams) are in Figure 6 and Supplementary Table S8. The selected “Baseline” samples had higher success rates than the “Unknown” samples. The self-assignment scores for individuals being successfully assigned to their suggested breed of origin (marked with gray in the tables) were 32.5%, 34.1%, 44%, 92.7%, 45.7%, 89.4% and 100% for Andravida, Pindos, Thessaly, Skyros, Penia, Messara and Rodos, respectively, for the “Baseline” dataset (Figure 5 and Supplementary Table S7). The corresponding assignment scores in the “Unknown” dataset were 26.5%, 9%, 0%, 85.7%, 42.4% and 70.2% for the Andravida, Pindos, Thessaly, Skyros, Penia and Messara, respectively (Figure 6 and Supplementary Table S8). In Figure 5 and Supplementary Table S7, among the available foreign breeds, the Turkoman, the Irish Draft, the Hanoverian, the Belgian Draft and the Garrano seem to have contributed the most to the Greek genetic pool, with total percentages of 2.1% (eight Greek samples assigned), 2.1% (eight samples), 2.6% (ten samples), 2.6% (ten samples) and 3.1% (twelve samples), respectively. A total of 106 Greek “Baseline” horses, or 27.5%, were assigned to the foreign breeds. The Andravida, the Thessaly, the Pindos and the Penia breeds had the highest percentage of foreign breed assignments, with rates of 48%, 57.5% and 39.5%, respectively (Figure 5 and Supplementary Table S7). The Penia had 30% assignment to foreign breeds, whereas for the Skyros, Messara and Rodos, foreign assignments were 7.2%, 3.3% and 0%, respectively, implying they were the least outbred. As mentioned earlier, the Andravida horses were self-assigned in only 32.5% of cases, and 10% were classified to other Greek breeds (e.g., Pindos (2.5%), Thessaly (5%) and Penia (2.5%)). Similarly, 26.5% of the Pindos individuals, 8% of Thessaly, 2% of Skyros, 23.9% of Penia and 7.5% of Messara were assigned to other Greek breeds (Figure 5 and Supplementary Table S7).

In Figure 6 and Supplementary Table S8, the dataset with the less typical “Unknown” samples scored lower successful rates of assignment. There were no Rodos samples to be assigned in this dataset, as all were of certain origin. The foreign breeds that stood out here were the Garrano (five samples or 2.2%), the Hanoverian (twelve samples or 5.2%), the Holstein (six samples or 2.6%), the Missouri Fox Trotter (seven samples or 3%) and the Turkoman (nine samples or 3.9%). Overall, more samples were assigned to foreign breeds in comparison to the typical samples, as the “Unknown” dataset had 78 samples or 33.8% assigned to foreign breeds, whereas the “Baseline” had 106 samples or 27.5%. It is worth noting that none of the non-typical Thessaly samples succeeded in being assigned to their own breed; they were instead classified as foreign breeds (e.g., Hanoverian (30%), Turkoman (20%), Welsh Pony (20%), Missouri Fox Trotter (10%) and Andravida (20%).

3.5. Structural Analysis

A structural analysis diagram of the forty-one foreign and seven Greek populations is shown in Figure 7 (20,000 Burn-in period, 100,000 Reps). STRUCTURE 2.3.4 analysis was performed for K 34 to 49, with five runs per K. The best K was determined to be 46 (Supplementary Figure S1a). The same analysis using only the seven Greek horse populations found the best K to be four (Supplementary Figure S1b). The majority of foreign populations represented more or less breeds with pedigrees with separate colors; however, there are several populations with admixed horses. The situation of Greek horse breeds (from code number 42 to 48; Figure 7a,b) shows that the populations of Skyros, Messara and Rodos consist of homogeneous breeds. Conversely, the Andravida, Thessaly and a large part of Pindos and Penia seem to consist of admixed animals. An interesting finding of this structural analysis is the presence of two subpopulations in the Pindos breed: one of them homogeneous, the other consisting of individuals with many different origins. The presence of the subpopulation shown with a uniform yellow color in Figure 7 was traced back to a batch of samples from a relatively isolated horse population from the southwestern Pindos mountains. We suggest that this subpopulation represents the purer and more authentic Pindos breed.

3.6. Phylogenetic Tree

A phylogenetic neighbor-joining tree (Figure 8) was constructed with PHYLIP 3.65 according to Nei distances. The dataset contained the seven Greek populations, and the forty-one foreign and Przewalski samples as an outgroup. The two Greek gaiters, Messara and Penia, appear to group together on their own branch. Their close presence has also been reported in other studies [24,40]. The Thessaly breed is set near this branch but closer to the common center of the dendrogram and it is not grouped with any foreign breed. This is reasonable, as this breed was formed with the contribution of many breeds, confirmed from the results of FCA and structural analysis (Figure 4 and Figure 7). Pindos stands close to the Garrano, and their proximity in the dendrogram does not imply a crossbreeding between them; Garrano was never imported into Greece. However, their position together could be explained by a common ancestor as both are old, primitive, indigenous breeds. Skyros is placed near the Welsh Pony and Rodos near the Icelandic horse. These two small horses would be expected to be seen in close branches to form a group; however, they unexpectedly fall apart, which may be due to a possible influence on the Rodos population by horses of Italian origin due to the Italian influence on the island of Rhodes, something that remains to be examined in the future. Rodos is not a real pony; its body size is very small. However, its position is justified near the Icelandic and Shetland Pony. In another study [41], based on MSY HT topology, Rodos and Skyros horses grouped relatively close to each other and near Russian, East and South-East Asian breeds (e.g., Mezen and Japanese horses) together with the Icelandic, Shetland Pony and Norwegian Fjord that are also present in our dataset. The Skyros and Rodos horses seem to have remained relatively unaffected by crossbreeding with other modern horse breeds due to their geographical isolation. As is indicated in this and the above study [41], they appear to have retained similarities with other, now distant, isolated populations.

Andravida appears not to cluster with other breeds and stands near the common root of the dendrogram, possibly due to its admixed constitution.

4. Discussion

This was the first study to investigate the indigenous Greek horse breeds with regard to their genetic structure, breed assignments and genetic relationships with foreign breeds. Overall, based on the estimated variability measures, and in comparison with the foreign horse breeds, the Greek group had high genetic variability levels for the Pindos, Penia, Andravida and Thessaly breeds, moderate for the Skyros and Messara regarding their population size, whereas Rodos showed very low variability levels [42,43]. The Pindos breed was the only population showing deviations from the HW equilibrium, due to the presence of subpopulations within the breed.

The results of FCA indicated that the Skyros, Messara and Rodos are clearly distinct and more homogeneous than the remaining populations. Particularly for Skyros, which has been the subject of extensive research, our results are consistent with previous studies [12,24,40,44]. The Rodos population is suffering from the loss of genetic diversity, as expected, which calls for an urgent and well-planned conservation intervention. On the FCA diagram, the Andravida, the Pindos, the Thessaly and the Penia breeds showed overlapping distributions, suggesting crossbreeding, something that has also been observed in a previous study using different markers and analyses [40]. Particularly, the Thessaly and Pindos breeds overlapped, a finding that agrees with a recent genetic analysis based on STRs [45], suggesting a shared genetic background.

Using the WhichRun 4.1 method, the “Baseline” dataset scored higher than the “Unknown” dataset, confirming our choice to divide the samples. The WhichRun 4.1 method proved efficient for scoring individuals from the Skyros, the Messara and the Rodos with high success rates (89–100%). The above horse breeds appear to be the most purebred. The relatively low correct assignment rates for the Penia (46%), the Thessaly (44%), the Pindos (34%) and lastly the Andravida (33%) breeds, combined with the FCA, enhance the suggestion of the exchange of genetic material through crosses among Greek breeds carried out by breeders without programming and control. Another issue which could explain the relatively low scores of the assignments is related to the low degree of genetic differentiation F_ST among the Greek sampled horse breeds (F_ST < 0.2). Τhe integration of foreign breeds in the dataset has also highlighted the effects of certain breeds on the gene pool of the Greek indigenous horse breeds. From these findings it can be argued that the Andravida breed was most influenced by foreign breeds (58% of its samples were assigned to the Irish Draft, Garrano and Holstein). The next breed was Thessaly with a percentage of 48% being influenced by the Turkoman, the Garrano and the Connemara Pony. For the Pindos, 40% was ranked to the Belgian Draft, Hanoverian and Garrano. Lastly, for the Penia, 30% was assigned to the Akhal Teke, Turkoman, Irish Draft and Thoroughbred. Notably, Messara, besides its satisfactory high score (89%), showed some influence from Penia. Both breeds were the only ones who were gaiters. For this reason, exchanges took place that were traced back to specific horses of the more widely distributed and less prized Penia breed that possibly had been falsely considered as Messara. Overall, the foreign breeds that appear to have affected the Greek “Baseline” horses the most were the Belgian Draft, Garrano, Hanoverian, Turkoman and the Irish Draft, while for the Unknown Greek horses, they were the Garrano, Hanoverian, Holstein, Missouri Fox Trotter and the Turkoman. The influences of the Turkoman, presumably from the times of the Ottoman empire or earlier [5], as well as from the possible affinity of the primitive Greek horses with the Garrano, an autochthonous mountain pony from northwestern Portugal and Galicia, can be of interest for horse phylogeny and the domestication history of European breeds. However, the WhichRun 4.1 results must be interpreted with caution. The breed assignments can only compare the breeds used in the analysis which may or may not include breeds which have actually interacted with the target breeds, and also breeds which never could have interacted with the target breed. In particular, the Garrano probably has not interbred with any of the native Greek breeds. It is a breed with high genetic diversity which may show a “primitive” overall genotype which causes it frequently to match up with target breeds in WhichRun 4.1 analyses.

Structural analysis confirmed the uniformity of the Skyros, Messara and Rodos breeds, a finding emerging from the FCA and breed assignment methods. The most striking result was the depiction of a possibly more purebred subpopulation within Pindos breed. This subpopulation will be proven to be of value for the future conservation planning of the breed.

In the constructed phylogenetic tree, Penia and Messara clustered with Thessaly. The positioning of Pindos close to Garrano, Skyros next to Welsh Pony and Rodos next to Icelandic could be explained by their common ancestors all being old, primitive, indigenous breeds, which are not known to have been imported into Greece. Andravida stands among foreign breeds showing no special affinity to the other Greek populations. Notably, the occurrence of crossbreeding among Greek and foreign populations complicates further the observed positions in the phylogenetic tree. Further sampling and the inclusion of more breeds from neighboring countries could improve the topology of the Greek horse breeds.

It is evident that Greece is not a country with a significant tradition of horse breeding. Of the seven recognized equine breeds, only the Skyros has a breeders’ association. From the rest, three have breeders’ clubs (Messara has six, Andravida two, Thessaly one) while the rest (Penia, Pindos, Rodope) lack any form of organization. Finally, the Rodos is not an officially recognized breed. Breed associations are paramount for breed conservation, maintaining a detailed studbook to ensure a unified strategy and implementation plan for the specific needs of each Greek horse breed. In addition to the above, the Rodos, with a very small population, requires urgent measures to increase the population size in order to avoid further genetic erosion. For Rodos, a controlled upgrading breeding system should be considered [46], using a compatible and preferably native breed (possibly Skyros). The Skyros population has been increasing steadily in recent years and should continue to increase to obtain adequate levels of genetic diversity. The Messara breed needs to exclude the crossbred animals from its population. The Messara and Penia breeds need to be differentiated using genetic testing due to their misleading phenotypic similarities. The Pindos population needs further analysis to identify and separate the core population from the crossbred one. For the Penia and Thessaly breeds, further sampling is needed for better assignment results. As for the Andravida, strict management practices are needed to avoid the exchange of genetic material with animals of similar characteristics from foreign breeds. Finally, for the recently recognized breed of Rodope, a thorough analysis is needed since it has never been studied.

5. Conclusions

All Greek indigenous horse breeds need improvements in their management and documentation methods. Microsatellite genotyping is a genetic tool which, combined with dependable studbooks, can prove crucial for indigenous horse breed conservation through reliable tracking, breeding stock selection as well as proper management and control of subsidization programs. Most vulnerable breeds, with low genetic variability and limited population sizes, must be urgently prioritized. The Rodos population requires urgent measures to survive genetic erosion and extinction. The Andravida, Pindos, Thessaly and Penia breeds need better management to kept them separate from other Greek and foreign breeds. The addition of supplementary samples, ensuring an adequate sample size and even representation of regional subpopulations, should clarify if their assignment rates would improve or if they have formed a uniform, crossbred population, and thus determine whether they can still be separated and preserved with organized breeding.