Abstract

The present-day Sable Island horse population, inhabiting an island off the eastern coast of Canada, is believed to have originated mainly from horses confiscated from the early French settlers in Nova Scotia in the latter half of the 18th century. In 1960, the Sable Island horses were given legal protected status and no human interference has since been allowed. The objective of this study was to characterize the current genetic diversity in Sable Island horses in comparison to 15 other horse breeds commonly found in Canada and 5 Spanish breeds. A total of 145 alleles from 12 microsatellite loci were detected in 1093 horses and 40 donkeys. The average number of alleles per locus ranged from 4.67 in the Sable Island horse population to 8.25 in Appaloosas, whereas the mean observed heterozygosity ranged from 0.626 in the Sable Island population to 0.787 in Asturcons. Various genetic distance estimates and clustering methods did not permit to support that the Sable Island horses originated from shipwrecked Spanish horses, according to a popular anecdote, but closely resemble light draft and multipurpose breeds commonly found in eastern Canada. Based on the Weitzman approach, the loss of the Sable Island horse population to the overall diversity in Canada is comparable or higher than any other horse breed. The Sable Island horse population has diverged enough from other breeds to deserve special attention by conservation interest groups.

The Sable Island horses are an isolated and protected population living on a sand bar located approximately 200 km off the east coast of Nova Scotia, Canada (Figure 1). The island is 42 km in length, with a maximum width of less than 1.5 km, with a total surface area of approximately 3400 ha (Catling et al. 1985). The present-day population is believed to descend from horses confiscated from the French settlers and placed on Sable Island after the Acadian Expulsion of 1755 (Christie 1995). The horses were left largely unattended until 1801 when the first permanent life-saving station was established on the island. From 1801 to 1940, various breeding mares and stallions, likely Thoroughbred, Morgan, and Clydesdale, were introduced to the island and selected progeny were brought back to the mainland for sale (Christie 1995). Although popular accounts often suggest that the Sable Island population is descended from shipwrecked Spanish horses in the 1700s, this is not supported by historical research (Campbell 1974; Christie 1995).

Figure 1

Geographic location of the Sable Island off the coast of Nova Scotia, Canada. Representatives of the Sable Island horse population (inset).

Figure 1

Geographic location of the Sable Island off the coast of Nova Scotia, Canada. Representatives of the Sable Island horse population (inset).

In 1960, the horses of Sable Island were given legal protected status in the Sable Island Regulations of the Canada Shipping Act. Since then the horses have been protected from human interference. All observation and research activities are noninvasive, and no veterinary care or supplementary feed is provided.

Long-term studies have been underway since the mid-1980s, and life history has been recorded for most horses. The population generally ranges between 250 and 400 horses with severe weather and harsh winters being significant mortality factors (Lucas et al. 1991). The horses live in breeding groups of 2–10 males and females of varying ages and in all-male groups comprised mostly of immature individuals. Whereas the home ranges of family bands tend to be well defined, those of all-male groups are variable.

With a generation interval of approximately 4 years, the Sable Island horse population has been isolated from possible introgression for about 11–12 generations. The current genetic structure of the Sable Island horse population is likely the result of natural selection and genetic drift. As with any other minor horse breeds with declining registrations in Canada, the Sable Island population could be facing genetic erosion. Although not recognized as a horse breed under the Canadian Animal Pedigree Act, the Sable Island horse may represent a valuable genetic resource. Within the context of horse population and breed conservation, genetic characterization is the first step in developing proper management strategies. Microsatellite markers are widely used to estimate genetic diversity within and among horse breeds (Aberle et al. 2004; Glowatzki-Mullis et al. 2005; Solis et al. 2005), and large databases of genotypes for most Canadian horse breeds are available. Microsatellites are also particularly well suited for the estimation of genetic structure (Pritchard et al. 2000), differentiation among populations (Corander et al. 2003), and individual assignments to predefined groups or clusters (Paetkau et al. 2004).

In this study, we estimated the genetic diversity and divergence in Sable Island horses in relation to other isolated breeds, draft and multipurpose breeds, and Spanish breeds. We applied 3 different approaches to describe the distribution of genetic diversity and to estimate the relative contribution of each breed to the overall genetic variability in Canadian horses.

Material and Methods

Breed and Sampling

For the Sable Island population (Figure 1), 57 samples were selected from a bank of frozen tissues collected from horses that died of natural causes between 1987 and 2000. Those selected were young animals that died before producing offspring and were considered to be the least related. Because it was possible to ensure that none of the sampled horses shared a mother and to reduce the chance of shared paternity, individuals were selected spatially; in addition to this, they were also selected temporally, horses born before 1989 or after 1993. In total, 1093 individuals representing 15 horse breeds found in Canada: Appaloosa, Arabian, Belgian, Canadian, Fjord, Hackney, Hannoverian, Icelandic, Morgan, Newfoundland, Peruvian, Percheron, Saddlebred, Thoroughbred, Standardbred, and the Sable Island population; along with 5 Spanish breeds: Asturcon, Malloquin, Menorquin, Pottok, and Andalusian were analyzed (Table 1). An additional 40 donkeys were also included for comparison and as an outgroup for phylogenetic purposes.

Table 1

Estimates of average and effective number of alleles per locus, and heterozygosity in horse breeds from Canada and Spain, including the Sable Island horse population and the donkey. Identified breed-specific alleles (excluding the donkey)

    Heterozygosity  Breed-specific alleles 
Breeds N Average number of alleles per locus, NA (SE) Effective number of alleles per locus, NE (SE) Observed (SE) Expected (SE) FIS Breed Locus Allele Frequency 
Appaloosa (AP) 56 8.25 (0.57) 4.70 (0.33) 0.78 (0.02) 0.78 (0.02) 0.001 AR HMS3 158 0.030 
Arabian (AR) 50 6.41 (0.29) 3.57 (0.33) 0.66 (0.02) 0.72 (0.02) 0.083 FJ HTG10 95 0.051 
Belgian (BE) 37 6.92 (0.53) 4.24 (0.38) 0.70 (0.03) 0.75 (0.02) 0.065 FJ AHT4 168 0.059 
Canadian (CN) 57 6.83 (0.41) 3.99 (0.38) 0.76 (0.03) 0.73 (0.03) −0.034 HA HTG10 117 0.010 
Fjord (FJ) 59 6.17 (0.44) 3.55 (0.26) 0.71 (0.04) 0.70 (0.03) −0.011 IC ASB2 224 0.009 
Hackney (HA) 49 7.17 (0.44) 3.67 (0.25) 0.71 (0.03) 0.72 (0.02) 0.017 NE ASB17 123 0.009 
Hannoverian (HN) 48 7.00 (0.49) 4.44 (0.29) 0.77 (0.02) 0.77 (0.02) −0.007 TH ASB23 192 0.010 
Icelandic (IC) 53 6.75 (0.49) 4.03 (0.27) 0.72 (0.02) 0.74 (0.02) 0.026 ST HMS6 139 0.008 
Morgan (MO) 54 7.92 (0.47) 4.23 (0.31) 0.76 (0.03) 0.76 (0.02) 0.001 AST HMS3 154 0.011 
Newfoundland (NE) 54 8.00 (0.70) 5.11 (0.59) 0.75 (0.03) 0.78 (0.03) 0.033 MAL HMS7 171 0.011 
Peruvian (PE) 55 6.83 (0.32) 3.95 (0.29) 0.74 (0.03) 0.73 (0.03) −0.009 MEN ASB23 213 0.048 
Percheron (PH) 41 7.92 (0.47) 4.45 (0.27) 0.78 (0.02) 0.78 (0.01) −0.009     
Saddlebred (SA) 60 7.08 (0.50) 4.14 (0.30) 0.73 (0.03) 0.75 (0.02) 0.029     
Sable Island (SI) 57 4.67 (0.51) 3.21 (0.36) 0.63 (0.03) 0.65 (0.04) 0.043     
Thoroughbred (TH) 50 5.50 (0.42) 3.87 (0.28) 0.75 (0.03) 0.73 (0.02) −0.023     
Standardbred (ST) 59 6.33 (0.36) 3.69 (0.21) 0.71 (0.03) 0.73 (0.02) 0.023     
Asturcon (AST) 45 7.08 (0.66) 4.19 (0.37) 0.79 (0.04) 0.74 (0.03) −0.058     
Mallorquin (MAL) 44 7.17 (0.44) 4.12 (0.35) 0.78 (0.03) 0.74 (0.02) −0.046     
Menorquin (MEN) 52 7.08 (0.34) 3.94 (0.26) 0.75 (0.02) 0.74 (0.02) −0.018     
Pottok (POT) 45 8.08 (0.54) 4.98 (0.47) 0.78 (0.03) 0.78 (0.03) −0.007     
Andalusian (AND) 68 7.42 (0.48) 4.06 (0.25) 0.74 (0.03) 0.75 (0.02) 0.010     
Donkey (DO) 40 5.17 (0.65) 2.88 (0.38) 0.55 (0.06) 0.59 (0.06) 0.061     
    Heterozygosity  Breed-specific alleles 
Breeds N Average number of alleles per locus, NA (SE) Effective number of alleles per locus, NE (SE) Observed (SE) Expected (SE) FIS Breed Locus Allele Frequency 
Appaloosa (AP) 56 8.25 (0.57) 4.70 (0.33) 0.78 (0.02) 0.78 (0.02) 0.001 AR HMS3 158 0.030 
Arabian (AR) 50 6.41 (0.29) 3.57 (0.33) 0.66 (0.02) 0.72 (0.02) 0.083 FJ HTG10 95 0.051 
Belgian (BE) 37 6.92 (0.53) 4.24 (0.38) 0.70 (0.03) 0.75 (0.02) 0.065 FJ AHT4 168 0.059 
Canadian (CN) 57 6.83 (0.41) 3.99 (0.38) 0.76 (0.03) 0.73 (0.03) −0.034 HA HTG10 117 0.010 
Fjord (FJ) 59 6.17 (0.44) 3.55 (0.26) 0.71 (0.04) 0.70 (0.03) −0.011 IC ASB2 224 0.009 
Hackney (HA) 49 7.17 (0.44) 3.67 (0.25) 0.71 (0.03) 0.72 (0.02) 0.017 NE ASB17 123 0.009 
Hannoverian (HN) 48 7.00 (0.49) 4.44 (0.29) 0.77 (0.02) 0.77 (0.02) −0.007 TH ASB23 192 0.010 
Icelandic (IC) 53 6.75 (0.49) 4.03 (0.27) 0.72 (0.02) 0.74 (0.02) 0.026 ST HMS6 139 0.008 
Morgan (MO) 54 7.92 (0.47) 4.23 (0.31) 0.76 (0.03) 0.76 (0.02) 0.001 AST HMS3 154 0.011 
Newfoundland (NE) 54 8.00 (0.70) 5.11 (0.59) 0.75 (0.03) 0.78 (0.03) 0.033 MAL HMS7 171 0.011 
Peruvian (PE) 55 6.83 (0.32) 3.95 (0.29) 0.74 (0.03) 0.73 (0.03) −0.009 MEN ASB23 213 0.048 
Percheron (PH) 41 7.92 (0.47) 4.45 (0.27) 0.78 (0.02) 0.78 (0.01) −0.009     
Saddlebred (SA) 60 7.08 (0.50) 4.14 (0.30) 0.73 (0.03) 0.75 (0.02) 0.029     
Sable Island (SI) 57 4.67 (0.51) 3.21 (0.36) 0.63 (0.03) 0.65 (0.04) 0.043     
Thoroughbred (TH) 50 5.50 (0.42) 3.87 (0.28) 0.75 (0.03) 0.73 (0.02) −0.023     
Standardbred (ST) 59 6.33 (0.36) 3.69 (0.21) 0.71 (0.03) 0.73 (0.02) 0.023     
Asturcon (AST) 45 7.08 (0.66) 4.19 (0.37) 0.79 (0.04) 0.74 (0.03) −0.058     
Mallorquin (MAL) 44 7.17 (0.44) 4.12 (0.35) 0.78 (0.03) 0.74 (0.02) −0.046     
Menorquin (MEN) 52 7.08 (0.34) 3.94 (0.26) 0.75 (0.02) 0.74 (0.02) −0.018     
Pottok (POT) 45 8.08 (0.54) 4.98 (0.47) 0.78 (0.03) 0.78 (0.03) −0.007     
Andalusian (AND) 68 7.42 (0.48) 4.06 (0.25) 0.74 (0.03) 0.75 (0.02) 0.010     
Donkey (DO) 40 5.17 (0.65) 2.88 (0.38) 0.55 (0.06) 0.59 (0.06) 0.061     

Microsatellite Typing

Twelve fluorescently labeled microsatellites were genotyped in the horse breeds, the donkey, and the Sable Island population: ASB2, ASB23 (Breen et al. 1997), AHT4, AHT5 (Binns et al. 1995), HTG4, HTG7, HTG10 (Marklund et al. 1994), HMS3, HMS6, HMS7 (Guerin et al. 1994), LEX33 (Coogle et al. 1996), and VHL20 (Van Haeringen et al. 1994). Microsatellite alleles were amplified in different multiplexes by the polymerase chain reaction (PCR). All microsatellite allele sizes were adjusted to the reference samples provided by the International Society for Animal Genetics. Estimated allele sizes in nucleotides were used throughout this study.

Statistical Analysis

The average (NA) and effective (NE) number of alleles, allele frequency per locus, observed (HO) and expected (HE) heterozygosity, fixation index (FIT, FST), detection of breed-specific alleles, and the analysis of molecular variance (AMOVA) were estimated using the GENEALEX 6 program (Peakall and Smouse 2006). The test for deviation from Hardy–Weinberg equilibrium was performed with the software Arlequin version 3.01 (Excoffier et al. 2005).

Genetic divergence among the horse breeds and the Sable Island population was estimated from the actual genotypes and allele frequencies. Different model-specific distance estimators can be used for microsatellite data, and the natural logarithm of the proportion of shared alleles (POSA) (Bowcock et al. 1994) was calculated using the Microsatellite Analyzer (Dieringer and Schlötterer 2002), whereas the Reynold's distances (Reynolds et al. 1983) were estimated using the PHYLIP 3.66 package (Felsenstein 1989–2006). Finally, the Kullback–Leibler (KL) divergence matrix (Corander et al. 2003) between identified clusters was estimated with the Bayesian Analysis of Population Structure software (BAPS) (Corander et al. 2004). The neighbor-joining (NJ) method (Saitou and Nei 1987), as implemented in PHYLIP 3.66, was used to build the phylogenetic trees from the distance matrices, and the results were visualized using Splitstree 4.0 (Huson and Bryant 2006).

Possible admixture between the Sable Island population and the other horse breeds was estimated with the factorial correspondence analysis implemented in GENETIX version 4.05.2 (Belkhir et al. 2004). This method is not sensitive to the different mutation models for microsatellite markers. Marker genotypes were also used to cluster individuals into predefined populations and to inferred clusters with the Bayesian approach implemented in STRUCTURE 2.1 (Pritchard et al. 2000) with 5 independent replicates each using a burn-in period of 100 000 and 1 000 000 iterations. The above approach was used to model 2–25 inferred clusters. Similarly, BAPS was also used to estimate the number of clusters under an admixture model. Individual assignment to predefined populations was tested using the Monte Carlo resampling approach (Paetkau et al. 2004) implemented in GENECLASS 2 (Piry et al. 2004).

The final step in partitioning genetic diversity followed the Weitzman's (1993) diversity function with the 3 different matrices of genetic distances computed above. The Reynold's and POSA distances used 22 predefined groups, identified as breeds, and the KL allelic divergence relied on 20 inferred clusters. The distance matrices were used as input for the WEITZPRO software (Derban et al. 2002).

Results

Microsatellite Loci

A total of 145 different alleles were detected in the horses assayed at 12 microsatellites. Estimates of the mean and expected number of alleles, observed and expected heterozygosities, FIS, and breed-specific alleles are presented in Table 1. The Sable Island population had the lowest diversity (0.63), the smallest number of observed (4.67) and expected (3.21) number of alleles, and a significant heterozygotes deficiency (P = 0.0413).

The AMOVA indicated that 17% of the variation originated among the horse breeds, whereas 83% of the observed variation was coming from within the breeds (ΦPT = 0.174, P = 0.010). Global FST (0.1115, P = 0.0002) over all loci and horse groups indicate that 11.2% of the genetic variability is attributed to significant differences between the horse breeds including the Sable Island horse population. The distribution of the genetic differentiation among breed pairs ranged from 1.3% for the Appaloosa–Hanoverian pair to 19.2% for the Sable Island horse–Thoroughbred pair. All horse group pair differentiations (FST) were highly significant (P < 0.001). Of the 210 possible horse group pair comparisons, the 13 largest estimates of genetic differentiation involved the Sable Island horse population.

Breed Differentiation

Results of the factorial correspondence analysis (Figure 2) clearly separated the Sable Island horse population from the other breeds. The first axis, which accounted for 17.54% of the total inertia, isolated the Sable Island horses and the small multipurpose horses (Icelandic, Fjord, Newfoundland, and Hackney). The second axis accounted for 14.33% of the total inertia and separated the sport horse breeds from the heavier and light draft horses.

Figure 2

Factorial correspondence analysis of the 12 microsatellite loci analyzed in the horse breeds registered in Canada and the Sable Island horse population.

Figure 2

Factorial correspondence analysis of the 12 microsatellite loci analyzed in the horse breeds registered in Canada and the Sable Island horse population.

NJ trees based on the POSA and Reynolds' genetic distance presented similar topologies, comparable to Nei's standard distance and FST (data not presented). Figure 3 presents the split graph using the KL divergence matrix obtained from the Bayesian analysis of genetic differentiation between the horse breeds, the Sable Island horse population, and the donkey (used as an outgroup) under an admixture model (goodness-of-fit of 94.3%). This approach has the advantage of reducing the complexity of the topology and assigning breeds to inferred clusters based on the frequency of multiloci genotypes. The NJ tree clearly clusters the Sable Island population with light draft and multipurpose breeds (Fjord, Icelandic, Newfoundland, Canadian, and Hackney) commonly found in eastern Canada. The admixture model clustered the Appaloosa and Hanoverian together and the Pottok and Peruvian. After the factorial correspondence analysis, the KL divergence matrix also removed the potential relationship of the Sable Island horses to the Spanish breeds (Asturcon, Mallorquin, Menorquin, Pottok, and Andalusian).

Figure 3

Split graphs for the 20 identified clusters based on the KL allele divergence estimates. AP–HN and PE–POT are the groups identified with the Bayesian approach.

Figure 3

Split graphs for the 20 identified clusters based on the KL allele divergence estimates. AP–HN and PE–POT are the groups identified with the Bayesian approach.

Breed and Individual Assignment

Using the complete multiloci dataset, including the donkey, 2 different Bayesian approaches were used to infer the number of likely clusters (2 ≤ K ≤ 25) under an admixture model (Figure 4). Both approaches identified 20 possible clusters (Table 2). Individual genotype membership to inferred clusters varied from 0.95 for the donkeys (Cluster XIV) and just more than 0.90 for the Sable Island horses (Cluster XV). The individual memberships for the Appaloosa and Hanoverian indicated that these horses are highly variable and difficult to assign to a unique cluster. The probabilities of individual assignment to predefined breeds or population showed similar trends. Within the Appaloosa breed, 43 (77%) horses were properly assigned and the remaining 13 could be assigned to other breeds including the Hanoverian, Thoroughbred, Standardbred, Saddlebred, Menorquin, Mallorquin, and Arabian. One hundred percent of the Fjord, Sable Island, Icelandic, and Standardbred horses, and as expected of the donkeys, were properly assigned to the predefined population. For the remaining breeds, more than 88% of the horses could be assigned to the alleged breed of origin.

Table 2

Estimated membership to inferred clusters obtained by STRUCTURE and individual assignment results of each individual horse to its own predefined breed or to other breeds

  Inferred Clusters Assignment 
Breed code Number of individuals II III IV VI VII VIII IX XI XII XIII XIV XV XVI XVII XVIII XIX XX To self group To other groups 
AP 56 0.047 0.030 0.126 0.029 0.099 0.047 0.019 0.038 0.024 0.230 0.022 0.029 0.044 0.003 0.019 0.046 0.063 0.032 0.020 0.033 43 13 76.79 
AR 50 0.010 0.011 0.032 0.007 0.018 0.011 0.010 0.020 0.010 0.019 0.010 0.014 0.731 0.002 0.021 0.021 0.011 0.015 0.010 0.016 47 94.00 
BE 37 0.013 0.048 0.007 0.015 0.009 0.010 0.023 0.009 0.068 0.007 0.011 0.017 0.012 0.003 0.007 0.012 0.014 0.009 0.686 0.021 36 97.30 
CN 57 0.011 0.737 0.018 0.011 0.015 0.012 0.020 0.021 0.024 0.012 0.012 0.012 0.013 0.003 0.008 0.015 0.007 0.007 0.024 0.018 56 98.25 
FJ 59 0.009 0.007 0.007 0.842 0.008 0.008 0.008 0.007 0.011 0.005 0.015 0.007 0.008 0.002 0.017 0.007 0.009 0.007 0.008 0.009 59  100.00 
HA 49 0.022 0.025 0.010 0.021 0.011 0.070 0.013 0.019 0.016 0.010 0.019 0.012 0.015 0.003 0.012 0.015 0.014 0.010 0.022 0.662 47 95.92 
HN 48 0.027 0.017 0.199 0.036 0.064 0.036 0.027 0.050 0.025 0.243 0.017 0.021 0.055 0.002 0.024 0.051 0.026 0.036 0.018 0.028 42 87.50 
IC 53 0.008 0.008 0.007 0.022 0.008 0.012 0.019 0.011 0.009 0.007 0.791 0.008 0.011 0.004 0.018 0.009 0.008 0.010 0.009 0.020 53  100.00 
MO 54 0.546 0.016 0.019 0.012 0.064 0.033 0.022 0.033 0.023 0.041 0.015 0.016 0.015 0.003 0.020 0.016 0.032 0.020 0.019 0.034 48 88.89 
NE 54 0.020 0.021 0.017 0.022 0.017 0.038 0.577 0.041 0.027 0.013 0.038 0.016 0.014 0.002 0.027 0.024 0.013 0.013 0.042 0.016 49 90.74 
PE 55 0.010 0.035 0.009 0.015 0.007 0.015 0.011 0.017 0.730 0.008 0.019 0.011 0.009 0.003 0.008 0.012 0.009 0.017 0.039 0.017 53 96.36 
PH 41 0.041 0.014 0.030 0.016 0.019 0.024 0.033 0.545 0.013 0.020 0.013 0.023 0.027 0.004 0.018 0.020 0.032 0.021 0.056 0.029 38 92.68 
SA 60 0.035 0.011 0.029 0.016 0.027 0.027 0.016 0.024 0.012 0.029 0.018 0.009 0.014 0.003 0.011 0.017 0.657 0.023 0.010 0.012 56 93.33 
SI 57 0.006 0.005 0.006 0.005 0.005 0.005 0.006 0.007 0.005 0.004 0.007 0.005 0.005 0.002 0.903 0.005 0.006 0.004 0.005 0.005 57  100.00 
TH 50 0.010 0.008 0.406 0.008 0.015 0.012 0.011 0.008 0.006 0.431 0.006 0.010 0.013 0.002 0.010 0.011 0.009 0.013 0.006 0.006 48 96.00 
ST 59 0.017 0.007 0.012 0.008 0.765 0.017 0.008 0.013 0.011 0.038 0.008 0.008 0.013 0.003 0.008 0.012 0.013 0.009 0.015 0.012 59  100.00 
AST 45 0.021 0.012 0.055 0.010 0.011 0.013 0.009 0.022 0.017 0.018 0.013 0.690 0.017 0.004 0.009 0.014 0.016 0.023 0.014 0.011 42 93.33 
MAL 44 0.017 0.007 0.012 0.025 0.025 0.689 0.015 0.018 0.010 0.016 0.015 0.016 0.031 0.003 0.008 0.043 0.015 0.010 0.012 0.013 40 90.91 
MEN 52 0.035 0.037 0.015 0.010 0.023 0.037 0.015 0.033 0.025 0.029 0.014 0.015 0.022 0.002 0.012 0.589 0.019 0.026 0.025 0.019 46 88.46 
POT 45 0.030 0.060 0.030 0.024 0.021 0.060 0.020 0.187 0.077 0.030 0.013 0.092 0.041 0.004 0.013 0.053 0.054 0.066 0.083 0.043 40 88.89 
AND 68 0.017 0.017 0.018 0.012 0.014 0.015 0.021 0.019 0.019 0.024 0.014 0.013 0.017 0.003 0.009 0.017 0.010 0.705 0.016 0.020 67 98.53 
DO 40 0.003 0.002 0.002 0.002 0.003 0.003 0.003 0.003 0.003 0.003 0.003 0.002 0.003 0.949 0.003 0.002 0.003 0.003 0.003 0.002 40  100.00 
  Inferred Clusters Assignment 
Breed code Number of individuals II III IV VI VII VIII IX XI XII XIII XIV XV XVI XVII XVIII XIX XX To self group To other groups 
AP 56 0.047 0.030 0.126 0.029 0.099 0.047 0.019 0.038 0.024 0.230 0.022 0.029 0.044 0.003 0.019 0.046 0.063 0.032 0.020 0.033 43 13 76.79 
AR 50 0.010 0.011 0.032 0.007 0.018 0.011 0.010 0.020 0.010 0.019 0.010 0.014 0.731 0.002 0.021 0.021 0.011 0.015 0.010 0.016 47 94.00 
BE 37 0.013 0.048 0.007 0.015 0.009 0.010 0.023 0.009 0.068 0.007 0.011 0.017 0.012 0.003 0.007 0.012 0.014 0.009 0.686 0.021 36 97.30 
CN 57 0.011 0.737 0.018 0.011 0.015 0.012 0.020 0.021 0.024 0.012 0.012 0.012 0.013 0.003 0.008 0.015 0.007 0.007 0.024 0.018 56 98.25 
FJ 59 0.009 0.007 0.007 0.842 0.008 0.008 0.008 0.007 0.011 0.005 0.015 0.007 0.008 0.002 0.017 0.007 0.009 0.007 0.008 0.009 59  100.00 
HA 49 0.022 0.025 0.010 0.021 0.011 0.070 0.013 0.019 0.016 0.010 0.019 0.012 0.015 0.003 0.012 0.015 0.014 0.010 0.022 0.662 47 95.92 
HN 48 0.027 0.017 0.199 0.036 0.064 0.036 0.027 0.050 0.025 0.243 0.017 0.021 0.055 0.002 0.024 0.051 0.026 0.036 0.018 0.028 42 87.50 
IC 53 0.008 0.008 0.007 0.022 0.008 0.012 0.019 0.011 0.009 0.007 0.791 0.008 0.011 0.004 0.018 0.009 0.008 0.010 0.009 0.020 53  100.00 
MO 54 0.546 0.016 0.019 0.012 0.064 0.033 0.022 0.033 0.023 0.041 0.015 0.016 0.015 0.003 0.020 0.016 0.032 0.020 0.019 0.034 48 88.89 
NE 54 0.020 0.021 0.017 0.022 0.017 0.038 0.577 0.041 0.027 0.013 0.038 0.016 0.014 0.002 0.027 0.024 0.013 0.013 0.042 0.016 49 90.74 
PE 55 0.010 0.035 0.009 0.015 0.007 0.015 0.011 0.017 0.730 0.008 0.019 0.011 0.009 0.003 0.008 0.012 0.009 0.017 0.039 0.017 53 96.36 
PH 41 0.041 0.014 0.030 0.016 0.019 0.024 0.033 0.545 0.013 0.020 0.013 0.023 0.027 0.004 0.018 0.020 0.032 0.021 0.056 0.029 38 92.68 
SA 60 0.035 0.011 0.029 0.016 0.027 0.027 0.016 0.024 0.012 0.029 0.018 0.009 0.014 0.003 0.011 0.017 0.657 0.023 0.010 0.012 56 93.33 
SI 57 0.006 0.005 0.006 0.005 0.005 0.005 0.006 0.007 0.005 0.004 0.007 0.005 0.005 0.002 0.903 0.005 0.006 0.004 0.005 0.005 57  100.00 
TH 50 0.010 0.008 0.406 0.008 0.015 0.012 0.011 0.008 0.006 0.431 0.006 0.010 0.013 0.002 0.010 0.011 0.009 0.013 0.006 0.006 48 96.00 
ST 59 0.017 0.007 0.012 0.008 0.765 0.017 0.008 0.013 0.011 0.038 0.008 0.008 0.013 0.003 0.008 0.012 0.013 0.009 0.015 0.012 59  100.00 
AST 45 0.021 0.012 0.055 0.010 0.011 0.013 0.009 0.022 0.017 0.018 0.013 0.690 0.017 0.004 0.009 0.014 0.016 0.023 0.014 0.011 42 93.33 
MAL 44 0.017 0.007 0.012 0.025 0.025 0.689 0.015 0.018 0.010 0.016 0.015 0.016 0.031 0.003 0.008 0.043 0.015 0.010 0.012 0.013 40 90.91 
MEN 52 0.035 0.037 0.015 0.010 0.023 0.037 0.015 0.033 0.025 0.029 0.014 0.015 0.022 0.002 0.012 0.589 0.019 0.026 0.025 0.019 46 88.46 
POT 45 0.030 0.060 0.030 0.024 0.021 0.060 0.020 0.187 0.077 0.030 0.013 0.092 0.041 0.004 0.013 0.053 0.054 0.066 0.083 0.043 40 88.89 
AND 68 0.017 0.017 0.018 0.012 0.014 0.015 0.021 0.019 0.019 0.024 0.014 0.013 0.017 0.003 0.009 0.017 0.010 0.705 0.016 0.020 67 98.53 
DO 40 0.003 0.002 0.002 0.002 0.003 0.003 0.003 0.003 0.003 0.003 0.003 0.002 0.003 0.949 0.003 0.002 0.003 0.003 0.003 0.002 40  100.00 
Figure 4

Proportion of membership 1133 individuals from Canadian horse breeds, the Sable island population, the Spanish breeds, and the donkey for K = 2, 5, 10, and 15–20.

Figure 4

Proportion of membership 1133 individuals from Canadian horse breeds, the Sable island population, the Spanish breeds, and the donkey for K = 2, 5, 10, and 15–20.

Breed Contribution to Diversity

Using different measures of isolation and genetic distances, and following the Weitzman's diversity function, Table 3 summarizes the marginal contribution and loss of genetic diversity at the breed level. Regardless of the distance estimate used, the biggest decrease in diversity would occur from the hypothetical loss of the Sable Island horse population. Considering the potential loss of the minor light draft and multipurpose breeds registered in Canada (Canadian, Morgan, Hackney, Newfoundland, Icelandic, Fjord, Saddlebred, and Sable Island population), the marginal loss of diversity from this group would be 40% compared with 18% for the Spanish breeds and 28% for the other more common horse breeds with higher numbers of registrations.

Table 3

Weitzman's diversity and marginal loss (%) estimates

 POSA Reynold's distance Fixation index FST KL divergence 
Total diversity 7.50 1.47 0.78 15.94 
Breed Diversity Marginal loss Diversity Marginal loss Diversity Marginal loss Cluster Diversity Marginal loss 
AP 7.23 3.43 1.44 2.15 0.77 2.06 AP–HN 15.55 2.40 
AR 6.94 7.41 1.36 7.67 0.72 7.42 TH 14.94 6.23 
BE 7.08 5.55 1.40 4.75 0.74 5.00 BE 15.32 3.87 
CN 7.07 5.59 1.39 5.22 0.74 5.16 PE 15.30 4.00 
FJ 6.96 7.12 1.37 6.62 0.73 6.41 MO 15.29 4.05 
HA 6.99 6.77 1.37 7.07 0.73 6.75 PH 15.25 4.29 
HN 7.28 2.91 1.44 1.65 0.77 1.56 NE 15.31 5.06 
IC 7.00 6.68 1.35 7.84 0.72 7.56 CN 14.93 6.30 
MO 7.10 5.33 1.41 4.19 0.75 3.95 HA 14.89 6.58 
NE 7.10 5.29 1.40 4.36 0.75 4.24 AR 14.83 6.93 
PE 7.06 5.85 1.40 4.49 0.75 4.38 SA 14.91 6.41 
PH 7.06 5.81 1.40 4.81 0.74 4.77 ST 14.80 7.15 
SA 7.04 6.08 1.38 5.98 0.74 5.90 FJ 14.77 7.30 
SI 6.67 10.98 1.24 15.67 0.64 17.29 IC 14.69 7.82 
TH 7.05 5.95 1.36 7.47 0.72 7.55 SI 13.32 16.41 
ST 7.05 5.90 1.38 6.02 0.74 5.79    
 POSA Reynold's distance Fixation index FST KL divergence 
Total diversity 7.50 1.47 0.78 15.94 
Breed Diversity Marginal loss Diversity Marginal loss Diversity Marginal loss Cluster Diversity Marginal loss 
AP 7.23 3.43 1.44 2.15 0.77 2.06 AP–HN 15.55 2.40 
AR 6.94 7.41 1.36 7.67 0.72 7.42 TH 14.94 6.23 
BE 7.08 5.55 1.40 4.75 0.74 5.00 BE 15.32 3.87 
CN 7.07 5.59 1.39 5.22 0.74 5.16 PE 15.30 4.00 
FJ 6.96 7.12 1.37 6.62 0.73 6.41 MO 15.29 4.05 
HA 6.99 6.77 1.37 7.07 0.73 6.75 PH 15.25 4.29 
HN 7.28 2.91 1.44 1.65 0.77 1.56 NE 15.31 5.06 
IC 7.00 6.68 1.35 7.84 0.72 7.56 CN 14.93 6.30 
MO 7.10 5.33 1.41 4.19 0.75 3.95 HA 14.89 6.58 
NE 7.10 5.29 1.40 4.36 0.75 4.24 AR 14.83 6.93 
PE 7.06 5.85 1.40 4.49 0.75 4.38 SA 14.91 6.41 
PH 7.06 5.81 1.40 4.81 0.74 4.77 ST 14.80 7.15 
SA 7.04 6.08 1.38 5.98 0.74 5.90 FJ 14.77 7.30 
SI 6.67 10.98 1.24 15.67 0.64 17.29 IC 14.69 7.82 
TH 7.05 5.95 1.36 7.47 0.72 7.55 SI 13.32 16.41 
ST 7.05 5.90 1.38 6.02 0.74 5.79    

Discussion

Genetic Variation

This study presents an analysis of genetic diversity in a set of 15 Canadian-registered horse breeds, 5 Spanish breeds, and a feral horse population from Sable Island. The number of alleles per locus (Table 1) is a simple and common measure of genetic diversity and, in some cases, may be more informative than genic heterozygosity, especially when populations have gone through recent bottlenecks (Maruyama and Fuerst 1985; Luikart et al. 1998). Because of their recent breeding history and some introgression from other breeds, the Appaloosa, Morgan, Newfoundland, Hackney, and Saddlebred showed the largest average number of alleles. For the other breeds commonly found in Canada, the number of alleles and heterozygosities are comparable to those previously reported for the Arabian, Icelandic, Hanoverian, and Thoroughbred (Aberle et al. 2004; Glowatzki-Mullis et al. 2005). Our estimates for the Spanish breeds are similar to those reported by Marletta et al. (2006) and Vega-Pla et al. (2006) but slightly higher than those presented by Cañon et al. (2000) and Solis et al. (2005). These latter 2 studies collected relatively large samples from different populations, whereas we collected breed genotypes from an available database, irrespective of localities.

We have also observed a relatively large number of breed-specific alleles (Table 1). Except for the ABS23 192 allele found in one Thoroughbred, which could be a single nucleotide insertion/deletion or a PCR artifact, all other rare alleles appear valid, albeit at very low frequencies. It is surprising to find that such alleles are still segregating in minor breeds such as the Fjord (with 2 rare alleles), Icelandic, Newfoundland, and the Hackney. Drift alone may not be severe enough to cause the loss of these particular alleles in these breeds.

Our results for the donkey reference population sample revealed less genetic variation (Table 1) than previously reported for other breeds and populations (Jordana et al. 2001; Aranguren-Méndez et al. 2002). Our random sampling of genotypes from an established database probably underestimated the average number of alleles and heterozygosities, and the latter 2 published studies better reflect genetic variation in Spanish donkeys.

As previously reported by Behara, Colling, Cothran, Gibson (1998) and Behara, Colling, Gibson (1998) and confirmed in this study, rare and endangered horse breeds in Canada such as the Canadian, Hackney, Morgan, Newfoundland, and Saddlebred show levels of genetic variability comparable or even higher than any other breeds analyzed. The status of these breeds for conservation efforts is based on the small or declining number of breeding individuals and annual registrations. However, our results show that genetic diversity measured in these breeds has been maintained to high levels and supports continued conservation efforts. On the other hand, the Sable Island horse population sample analyzed in this study may not represent a truly random collection of animals. As such, our measures of genetic diversity in this particular group may be biased and better reflect the upper limit of estimated genetic diversity.

With a reduced effective number of alleles per locus and a significant heterozygous deficiency, the Sable Island horse population exhibits lower genetic diversity than other feral horse populations (Ashley 2004; Morais et al. 2005; Vega-Pla et al. 2006).

Breed Differentiation and Assignment

Different classical estimates of genetic distances were used to illustrate genetic divergence between the horse breeds and the Sable Island feral population and all resulted in very similar topologies. Our results for the Spanish horse breeds closely resemble those of Vega-Pla et al. (2006). Marletta et al. (2006) elegantly used kinship distances and molecular coancestry (Caballero and Toro 2002) along with a distance estimate based on the average proportion qk(i) of the genome of breed i that comes from an ancestral population k (with k equal to the number of breeds), following the Bayesian approach of Pritchard et al. (2000), to illustrate the genetic relationships among western Mediterranean native horse breeds. Here, we also used a Bayesian approach to estimate the KL divergence matrix between underlying clusters of multilocus genotypes as a measure of the relative genetic distance between these clusters (Corander et al. 2003). Results from the analyses of horse breed differentiation and individual assignment to unique clusters clearly define the Sable Island feral horse population and its close phylogenetic relationships to sport and light draft horse breeds commonly found in eastern Canada. Consequently, the popular anecdotal influence of Spanish breeds in the Sable Island horse population could not be supported. Within the cluster of eastern horse breeds found in Canada, it remains difficult to assess the origin of the Sable Island horse population; however, these horses share a common ancestor with the Newfoundland, Icelandic, and Fjord breeds.

Several Canadian horse breeds analyzed in this study are considered as critical to endangered by Rare Breeds Canada and other conservation interest groups. All these recognized breeds exhibit levels of genetic diversity comparable to and in some cases higher than more common and major horse breeds. As such, in situ conservation efforts can still rely on accessible genetic variability and thus should be able to maintain genetic diversity through well-designed breeding programs. Furthermore, the Weitzman's diversity function as applied to a different measure of differentiation clearly illustrates the relevance of these breeds to the overall genetic diversity of the Canadian horse population.

The Sable Island feral population represents a unique group of individuals. Legally protected since 1960, this horse population has significantly diverged from any other ancestral breeds. No introgression has occurred for at least 11 to 12 generations, and genetic drift and inbreeding likely played a role in the observed loss of genetic diversity as measured by the number of alleles per loci, heterozygous deficiency, and the large genetic distances to other horse breeds. All the measured genetic statistics, including the marginal contribution to diversity under a hypothetical extinction, probably underestimate the true value of the Sable Island horse as a genetic resource. Field observations on body condition and reproduction (Lucas et al. 1991) indicate that the Sable Island horse is well adapted to the local conditions of a seasonally demanding environment with limited resources and may be characterized by higher frequencies of alleles associated with life-history traits.

References

Aberle
KS
Hamann
H
Drogemuller
C
Distl
O
Genetic diversity in German draught horse breeds compared with a group of primitive, riding and wild horses by means of microsatellite DNA markers
Anim Genet.
 , 
2004
, vol. 
35
 (pg. 
270
-
277
)
Aranguren-Méndez
J
Gomez
M
Jordana
J
Hierarchical analysis of genetic structure in Spanish donkey breeds using microsatellite markers
Heredity
 , 
2002
, vol. 
89
 (pg. 
207
-
211
)
Ashley
MC
Population genetics of feral horses: implications of behavioral isolation
J Mammal
 , 
2004
, vol. 
85
 (pg. 
611
-
617
)
Behara
AMP
Colling
DT
Cothran
EG
Gibson
JP
Genetic relationships between horse breeds based on microsatellite data: applications for livestock conservation
In: Organizing Committee, editors. Proceedings of the 6th World Congress on Genetics Applied to Livestock Production
 , 
1998
, vol. 
Vol. 28
 
Armidale, NSW (Australia)
University of New England
(pg. 
119
-
122
Jan. 11–16
Behara
AMP
Colling
DT
Gibson
JP
Genetic diversity in endangered and common Canadian horse breeds. Anim Genet
 , 
1998
, vol. 
29
 
Suppl 1
pg. 
16
 
Belkhir
K
Borsa
P
Chikhi
L
Raufaste
N
Bonhomme
F
GENETIX 4.05, logiciel sous Windows TM pour la génétique des populations
 , 
1996
Montpellier (France)
Laboratoire Génome, Populations, Interactions, CNRS UMR 5000, Université de Montpellier II
Binns
MM
Holmes
NG
Holliman
A
Scott
AM
The identification of polymorphic microsatellite loci in the horse and their use in thoroughbred parentage testing
Br Vet J
 , 
1995
, vol. 
151
 (pg. 
9
-
15
)
Bowcock
AM
Ruiz-Linares
A
Tomfohrde
J
Inch
E
Kidd
JR
Cavalli-Sforza
LL
High resolution of human evolutionary trees with polymorphic microsatellites
Nature
 , 
1994
, vol. 
368
 (pg. 
455
-
457
)
Breen
M
Lindgren
G
Binns
MM
Norman
J
Irvin
Z
Bell
K
Sandberg
K
Ellegren
H
Genetical and physical assignments of equine microsatellites-first integration of anchored markers in horse genome mapping
Mamm Genome
 , 
1997
, vol. 
8
 (pg. 
267
-
273
)
Caballero
A
Toro
MA
Analysis of genetic diversity for the management of conserved subdivided populations
Conserv Genet.
 , 
2002
, vol. 
3
 (pg. 
289
-
299
)
Campbell
L
Sable Island, fatal and fertile Crescent
Hantsport (Nova Scotia)
 , 
1974
Lancelot Press Limited
pg. 
104
 
Cañon
J
Checa
ML
Carleos
C
Vega-Pla
JL
Vallejo
M
Dunner
S
The genetic structure of Spanish Celtic horse breeds inferred from microsatellite data
Anim Genet.
 , 
2000
, vol. 
31
 (pg. 
39
-
48
)
Catling
PM
Freedman
B
Lucas
Z
The vegetation and phytogeography of Sable Island, Nova Scotia. Proc N S Inst Sci
 , 
1985
, vol. 
34
 (pg. 
181
-
247
)
Christie
BJ
The horses of Sable Island
RR 2 Porters Lake (Nova Scotia)
 , 
1995
2nd ed.
Pottersfield Press
pg. 
111
 
Coogle
L
Bailey
E
Reid
R
Russ
M
Equine dinucleotide repeat loci from LEX025 to LEX033
Anim Genet.
 , 
1996
, vol. 
27
 (pg. 
289
-
290
)
Corander
J
Waldmann
P
Marttinen
P
Sillanpää
MJ
BAPS 2: enhanced possibilities for the analysis of genetic population structure
Bioinformatics
 , 
2004
, vol. 
20
 (pg. 
2363
-
2369
)
Corander
J
Waldmann
P
Sillanpää
MJ
Bayesian analysis of genetic differentiation between population
Genetics
 , 
2003
, vol. 
163
 (pg. 
367
-
374
)
Derban
S
Foulley
JL
Ollivier
L
WEITZPRO: a software for analyzing genetic diversity
2002
Paris (France)
INRA
Dieringer
D
Schlötterer
C
Microsatellite analyser (MSA): a platform independent analysis tool for large microsatellite data sets
Mol Ecol Notes
 , 
2002
, vol. 
3
 (pg. 
167
-
169
)
Excoffier
L
Laval
G
Schneider
S
Arlequin ver 3.0: an integrated software package for population genetics data analysis
Evol Bioinform Online
 , 
2005
, vol. 
1
 (pg. 
47
-
50
)
Felsenstein
J
PHYLIP (phylogeny inference package). Version 3.66
 , 
1989
Seattle (WA)
Genomes Sciences, Department of Genetics, University of Washington
 
Glowatzki-Mullis
ML
Muntwyler
J
Pfister
W
Marti
E
Rieder
E
Poncet
PA
Gaillard
C
Genetic diversity among horse populations with a special focus on the Franches-Montagnes breed
Anim Genet.
 , 
2005
, vol. 
37
 (pg. 
33
-
39
)
Guerin
G
Bertaud
M
Amigues
Y
Characterization of seven new horse microsatellites: HMS1, HMS2, HMS3, HMS5, HMS6, HMS7 and HMS8
Anim Genet
 , 
1994
, vol. 
25
 pg. 
62
 
Huson
DH
Bryant
D
Application of phylogenetic networks in evolutionary studies
Mol Biol Evol.
 , 
2006
, vol. 
23
 (pg. 
254
-
267
)
Jordana
J
Folch
P
Aranguren
JA
Microsatellite analysis of genetic diversity in the Catalonian donkey breed
J Anim Breed Genet.
 , 
2001
, vol. 
118
 (pg. 
57
-
63
)
Lucas
Z
Raeside
JR
Betteridge
KJ
Non-invasive assessment of the incidences of pregnancy and pregnancy loss in the feral horses of Sable Island
J Reprod Fertil
 , 
1991
Suppl. 44
(pg. 
479
-
488
)
Luikart
G
Allendorf
FW
Cornuet
J-M
Sherwin
WB
Distortion of allele frequency distributions provides a test for recent population bottlenecks
J Hered
 , 
1998
, vol. 
89
 (pg. 
238
-
247
)
Marklund
S
Ellegren
H
Ericksson
S
Sanberg
K
Andersson
L
Parentage testing and linkage analysis in the horse using a set of highly polymorphic microsatellites
Anim Genet.
 , 
1994
, vol. 
25
 (pg. 
19
-
23
)
Marletta
D
Tupac-Yupanqui
I
Bordonaro
S
Garcia
D
Guastella
AM
Crisione
A
Cañon
J
Dunner
S
Analysis of genetic diversity and determination of relationships among western Mediterranean horse breeds using microsatellite markers
J Anim Breed Genet.
 , 
2006
, vol. 
123
 (pg. 
315
-
325
)
Maruyama
T
Fuerst
PA
Population bottlenecks and non equilibrium models in population genetics. II. Number of alleles in a small population that was formed by a recent bottleneck
Genetics
 , 
1985
, vol. 
111
 (pg. 
675
-
689
)
Morais
J
Oom
MM
Malta-Vacas
J
Luis
C
Genetic structure of an endangered Portuguese semiferal pony breed, the Garrano
Biochem Genet
 , 
2005
, vol. 
43
 (pg. 
347
-
364
)
Paetkau
D
Slade
R
Burden
M
Estoup
A
Genetic assignment methods for the direct, real-time estimation of migration rate: a simulation-based exploration of accuracy and power
Mol Ecol
 , 
2004
, vol. 
13
 (pg. 
55
-
65
)
Peakall
R
Smouse
PE
GENALEX 6: genetic analysis in Excel. Population genetic software for teaching and research
Mol Ecol Notes
 , 
2006
, vol. 
6
 (pg. 
288
-
295
)
Piry
S
Alapetite
A
Cornuet
JM
Paetkau
D
Baudouin
L
Estoup
A
GeneClass2: a software for genetic assignment and first-generation migrant detection
J Hered
 , 
2004
, vol. 
95
 (pg. 
536
-
539
)
Pritchard
JK
Stephens
M
Donnelly
P
Inference of population structure using multilocus genotype data
Genetics
 , 
2000
, vol. 
155
 (pg. 
945
-
959
)
Reynolds
J
Weir
BS
Cokerham
C
Estimation of the coancestry coefficients: basis for a short-term genetic distance
Genetics
 , 
1983
, vol. 
105
 (pg. 
767
-
779
)
Saitou
N
Nei
M
The neighbor-joining method: a new method for reconstructing phylogenetic trees
Mol Biol Evol.
 , 
1987
, vol. 
4
 (pg. 
406
-
425
)
Solis
A
Jugo
BM
Meriaux
JC
Iriondo
M
Mazon
LI
Aguirre
AI
Vicario
A
Estomba
A
Genetic diversity within and among four South European native horse breeds based on microsatellite DNA analysis: implications for conservation
J Hered
 , 
2005
, vol. 
96
 (pg. 
670
-
678
)
Van Haeringen
H
Bowling
AT
Stott
ML
Lenstra
JA
Zwaagstra
KA
A highly polymorphic horse microsatellite locus; VHL20
Anim Genet.
 , 
1994
, vol. 
25
 pg. 
207
 
Vega-Pla
JL
Calderón
J
Rodriguez-Gallardo
PP
Martinez
AM
Rico
C
Saving feral horse populations: does it really matter? A case study of wild horses from Doñana national park in southern Spain
Anim Genet.
 , 
2006
, vol. 
37
 (pg. 
571
-
578
)
Weitzman
ML
What to preserve? An application of diversity theory to crane conservation
Q J Econ
 , 
1993
, vol. 
108
 (pg. 
157
-
183
)

Author notes

Corresponding Editor: Ernest Bailey