Genetic variation
Among our sample of 743 Finnhorses, we found 16 of the 18 previously described horse haplogroups, with only the European haplogroup K and the Middle Eastern haplogroup O missing. Interestingly, haplogroup F, which so far was detected only in the Przewalski’s horse, was present in four of the Finnhorses analysed here, of which three were not included in any of the breeding sections and one was registered as a riding horse. To date, haplogroup F has not been found in any modern horse breed, which is unexpected since it was recently suggested that the Przewalski’s horse, the only extant wild horse species, is derived from the early domestic horses of the Botai culture [38]. This result contradicted the findings of previous studies that have placed the Przewalski’s horse as a sister taxon of domestic horses, with some possible gene flow from the domestic horse to the Przewalski’s horse [39, 40]. The existence of the Przewalski haplogroup F in the Finnhorse, if verified by sequencing whole mitogenomes, may reveal surprising events in the domestication history of the horse.
Within the Finnhorse breeding sections, haplogroups L and M were most frequent in riding horses, haplogroups B and Q in trotters, haplogroups B, C and M in draught horses, and haplogroups G and L in pony-sized horses. All these haplogroups were considered as of European origin by [11] except for haplogroup Q, which has a more Asian or Middle Eastern distribution and is present, for example, in the Arabian horse. Mitochondrial DNA diversity in the Finnhorses (\(\uppi\) = 0.022 and \({\hat{h}}\) = 0.979) is relatively similar to that detected in many other horse breeds [41,42,43]. In the domestic horse, it has been suggested that the large number of haplogroups and haplotypes spread over wide geographic regions results from a large number of mares having been incorporated into the domestic horse population [11, 12, 44]. This ancestral polymorphism has seemingly been retained in the Finnhorse as well, which is probably due to the high level of maternal genetic diversity of the founding population of the Finnhorse. Contrary to a previous Bayesian skyline plot that was constructed from a mixture of breeds in [11] and showed a decrease in the female effective population size about 7000 years ago and an increase thereafter, the female effective size in Finnhorses began to decrease approximately 300 years ago. This decline was accentuated approximately 110 years ago, at the time when the breed was founded, and again approximately 50 years ago, when the breeding sections for trotters, riding and pony-sized horses were founded (Fig. 2a, b). This is possibly due to a founder effect connected with the selection of horses that were included in the breed and in the breeding sections.
The present Finnhorse is closely related to the native Scandinavian, Estonian and Mongolian horses [6, 13] and presumably to the Russian heavy Mezen horse and the native Lithuanian Žemaitukas horse [14]. In our study, the Finnhorses also cluster with the native Estonian horse and with the native British and Irish breeds (Fig. 5). These close relationships between eastern, southern and western native breeds might explain the presence of haplotypes in the Finnhorse that are found in breeds from Europe, Central Asia and Middle East. Historically, King Gustav Vasa established stud farms in Finland as early as the 16th century, with horses being imported from the Netherlands and northern Germany, to increase the size of Finnish horses. These horses most likely had the same ancestry as the modern Friesian and Oldenburg breeds, including the Spanish ancestry that was used to create the Friesian breed [45]. During and after the Thirty Years’ War (1618–1648), the Finnish cavalrymen who returned home brought with them horses from Central Europe and the Baltic region that were then bred with the Finnish horses. During the eighteenth and nineteenth centuries, some Arab horses were imported to Finland as well as Warmblood and heavy Ardennes horses from Sweden, and Orlov Trotters and ‘cossack horses’, possibly Don horses, from Russia [14, 46]. By the early twentieth century, all this crossbreeding had resulted in three different types of Finnish horses: heavy draught-type horses, light and long-legged (race) horses, and light and tough pony-sized horses [17]. Although most of the imported horses were stallions because they had more breeding value, occasionally imported mares may have introduced their Central European or eastern mitochondria into the Finnish horse population. Thus, the history of the breed may have resulted in a high level of mitochondrial genetic diversity and in a large female effective population size.
Analysis of the nuclear diversity, measured as the expected heterozygosity based on SNP data, showed that its level in Finnhorses was similar to that of many other breeds, including, for example, the Akhal-Teke, Andalusian, Lusitano, Mongolian and Tuva horses [6], which all have ‘moderate levels’ of nuclear diversity (the highest level i.e. 0.337 in Swiss Warmblood and Paint horses and the lowest level i.e. 0.239 in Clydesdale). The lowest and highest levels of diversity among our data for Finnhorse were in trotters (HE = 0.318) and pony-sized horses (HE = 0.326), respectively. These estimates are slightly higher than the estimates obtained from 27 Finnhorses (0.301) in [6], where the dataset was pruned for linkage equilibrium with an r2 > 0.4 compared to our r2 > 0.5, which might have had some effect on the estimates. Nevertheless, it is evident that the nuclear diversity of each of the Finnhorse breeding sections remains at a good level. Moreover, although ROH were longer in Finnhorses, there does not appear to be any strong inbreeding, since the numbers of ROH and inbreeding coefficients are smaller than in the other breeds that we studied here. Average inbreeding coefficients (\(\hat{F}_{{\text{II}}}\)) estimated in [6] varied from 0.015 in the Mongolian horse to 0.261 in Clydesdale, and was 0.052 for Finnhorses, whereas, based on a larger sample size, we found a lower estimate i.e. 0.032. Although the criteria used for the detection of ROH vary among studies, the length of the genome covered by ROH can be grossly compared between various studies of horse breeds. Using a 50-SNP window and setting the minimum length of ROH to 1000 kb, we found that the mean genome length covered by ROH in the Finnhorse breeding sections ranged from 104.4 to 137.6 Mb, whereas it ranged from 92.0 to 517.0 Mb in the other breeds that we studied. In a recent study [47], also based on a 50-SNP window but using another SNP chip to detect homozygous segments of more than 500 kb, the mean genome length covered by ROH was 305.1 Mb and ranged from 227.5 Mb in the Noriker breed (a heavy Austrian draught horse) to 396.5 Mb in Purebred Arabians. Another study [48], which used the previous version of the Illumina Equine BeadChip and a 50-SNP window to detect homozygous segments of more than 40 kb, found that the mean genome length covered by ROH ranged from 416.5 Mb in a Dülmen horse (a native German pony) to 953.2 Mb in a Thoroughbred.
Although we found no clear evidence of inbreeding, the effective population sizes for each breeding section and for the horses not registered in the studbook were very small (from 43 to 56) when estimated based on the nuclear SNPs. This small effective size can result from only having four founding stallions for the breed but may also stem from recent breeding practices; although such practices were designed to avoid inbreeding, only five stallions have made approximately 50% of the genetic contribution during the 2005–2014 period (estimated from pedigree data based on the expected proportion of alleles in an individual originating from an ancestor [49]). Currently, the number of matings for each stallion is limited to 150 per year [15], which could still be far too many if the aim is to increase the effective population size of the breeding sections and of the whole breed. Moreover, the overall number of stallions in each of the breeding sections is currently relatively small (218 stallions in harness trotters, 103 in riding horses, 64 in pony-sized horses and 30 in draught horses [15]), and the horses share many ancestors in their pedigrees. The effective population size of the entire breed obtained by combining all breeding sections and horses not in the studbook was 161, which is still a reasonably good level and might, in fact, favour the current breeding practice, which is that any registered Finnhorse is allowed to be included in the breeding sections provided that the section-criteria are met. Thus, the non-studbook horses serve as a large gene pool for the breeding sections.
Genetic differentiation and history of breed
Based on the SNP data, we found a clear differentiation between the Finnhorses and the other breeds. All the Finnhorses clustered together in the Bayesian clustering analysis and along the first principal component in the PCA analysis, and were separated from the other breeds that we studied here. However, the differentiation between breeding sections was less clear. In the PCA, some clustering of the trotters and pony-sized horses was observed, but when we estimated FST, FST-estimates were significant only for the comparisons between the draught horses and the others (Table 4). Based on mitochondrial DNA, differentiation was strongest between the pony-sized horses and trotters, and overall, many pairwise comparisons were significant (Table 4). The only non-significant comparisons were between the trotters, draught horses and non-studbook horses and between the riding and pony-sized horses. The higher FST-values obtained based on mitochondrial DNA compared to SNP data are most likely due to the different mode of inheritance of these two marker types: mitochondrial DNA is haploid and maternally inherited and thus has one-fourth of the effective population size of the diploid and biparentally inherited nuclear SNPs. This intensifies the effect of genetic drift on mitochondrial markers, leading to faster differentiation. The mainly non-significant FST-values between the Finnhorse breeding sections suggest very weak, if any, differentiation among trotters, pony-sized horses and riding horses. The traits that are used as criteria for accepting individuals in the breeding sections are quantitative, possibly very complex and likely to have very variable heritabilities. Indeed, estimated heritabilities in the Finnhorse breed vary considerably from height at withers and at croup (0.89 and 0.90, respectively) to movement at walk and trot (0.13 and 0.18, respectively) [50]. The low heritabilities of many of the traits used, the acceptance of new individuals to breeding sections from the ‘common gene pool’ (i.e. non-studbook horses), the fact that several horses are registered in several breeding sections and that the breeding sections have been founded fairly recently, weaken the differentiation among the breeding sections.