Tissue samples were collected according to the guidelines for proper conduct of animal experiments of the Science Council of Japan. Tissue samples from 34 Agu pigs were collected from 26 farms on Okinawa and Ishigaki islands. Because pedigree information is incomplete at many Agu pig farms, we collected samples based on interviews with farmers and using pedigree information to ensure that samples were as unrelated as possible. DNA samples from 31 Landrace, 30 Large White, 31 Duroc, 32 Berkshire, 11 Jinhua, and 14 Meishan pigs were collected from several farms in Japan. Twenty DNA samples were collected from Ryukyu wild boars (Sus scrofa riukiuanus) on Okinawa, Ishigaki and Iriomote islands. Samples from European breeds were collected from unrelated pigs. Jinhua and Meishan pig samples were collected from an inbred strain because the founding population was small. DNA was extracted using conventional phenol/chloroform techniques.

Genotyping was performed using 21 microsatellite markers distributed across 17 chromosomes; these markers were selected based on informativeness and readability (Supplementary Table S1). Microsatellite fragments were amplified using polymerase chain reactions (PCR) with 42 ng samples of pig genomic DNA and fluorophore-labelled primers. The PCR were performed in final volumes of 15 μL containing 1.5 μL of 10× PCR buffer, 0.9 μL of 25 mM MgCl₂, 1.5 μL of 8 mM dNTP, 3 pmol of each primer and 0.4 units of Taq DNA polymerase (Applied Biosystems, Foster City, CA, USA). Thermal cycling was performed using a GeneAmp PCR System 9600 (Applied Biosystems, USA) with an initial denaturation for 9 min at 94°C, followed by 40 cycles of 30 s at 94°C, 30 s at annealing temperature (Supplementary Table S1) and 30 s at 72°C and then a final extension step at 72°C for 5 min. To genotype samples, PCR products were electrophoresed using an ABI 3130xl DNA sequencer (Applied Biosystems, USA). Allele sizes were then determined using Gene-Mapper software (Version 3.7, Applied Biosystems, USA).

Numbers of alleles (N_A), effective numbers of alleles (N_E), observed heterozygosity (H_O), and expected heterozygosity (H_E) were calculated using GenAlEx version 6.5 software [7]. Pairwise F_ST and inbreeding coefficients (F_IS) were calculated using FSTAT version 2.9.3 [8]. Pairwise comparisons of F_ST values were performed using Bonferroni corrections. Deviations from the Hardy–Weinberg equilibrium (HWE) were calculated using GENEPOP program [9]. Polymorphic information contents (PIC) were calculated and allele diversities at each locus were determined using CERVUS version3.0 [10]. Reynolds’ genetic distances [11] were calculated and a phylogenetic tree was constructed via Neighbor-joining tree analysis (NJ) clustering using Populations version 1.2.32 software [12]. Robustness of tree topology was tested using the bootstrap method with 1,000 repetitions. The genetic distances between individuals were determined as −ln(proportions of shared alleles) using a microsatellite analyzer [13] and were used to construct a NJ tree of 203 individuals from 8 pig breeds using MEGA 7.0 [14]. Principal component analysis (PCA) of gene frequencies were performed using the PCAGEN program [15] and Bayesian clustering was performed using STRUCTURE version 2.3.4 [16] to infer population structures and degrees of admixture. Numbers of assumed clusters (K) were estimated by performing 10 independent runs for each K value from 1 to 10 with a burn-in period of 100,000 iterations followed by 100,000 Markov Chain Monte Carlo iterations. The optimal K value was estimated according to the Evanno ΔK method using STRUCTURE HARVESTER 0.6.94 application [17]. The results of STRUCTURE analysis were averaged using CLUMPAK 1.1 [18].

In this study, genetic structures of Agu pigs were investigated using analyses of 21 microsatellite markers in 203 animals from 8 pig breeds. Parameters of genetic diversity among European breeds were within previously reported ranges (Table 1) [21–23], although among Chinese breeds, the H_O value of Meishan pigs was lower than that shown in previous studies [23,24]. N_A and N_E values in Jinhua pigs were lower than in previous studies [25,26], whereas the corresponding H_O value was higher. Agu pigs had lower N_E than European breeds and Ryukyu wild boar. Among domestic breeds, Agu pigs had the lowest H_O value and the highest F_IS value, and 10 of 21 markers deviated significantly from the HWE in Agu pigs. These results indicate severe loss of heterozygosity in Agu pig populations, likely resulting from inbreeding. Hence, following replacement with European breeds, low numbers of Agu individuals in a small population likely favored inbreeding. Consequently, Agu pigs had small litter sizes (total number born, 4.8) [27] and low conception rates (20%) [28]. In contrast with Agu pigs, Meishan and Jinhua pigs had low F_IS values, likely reflecting breeding strategies that avoid mating between close relatives.

It is assumed that Japanese Agu pigs originated from the first introduction of pigs from the Asian continent. A previous study of mtDNA control regions by Touma et al [5] revealed that most Agu individuals have East Asian haplotypes, whereas only 14% of Agu pigs carried European haplotypes. These data are evidence of gene flows from European breeds into Agu pig populations. In our studies of nuclear DNA, phylogenetic tree analysis and assessments of Reynolds’ genetic distances revealed shorter genetic distances between Agu pigs and European breeds than between Agu and Chinese breeds (Table 2; Figure 1). Moreover, genetic structure analysis (Figure 4) with K = 2 showed that Agu pigs are a genetic admixture of European and Asian clusters, with predominance of the European cluster. Thus, whereas the nuclear genomes of Agu pigs were influenced by European breeds, Asian mtDNA haplotypes remain predominant. These results suggest that the European breeds that were introduced into Okinawa after 1904 were often used as sires for breeding with Agu females.

Taken with a previous study of mtDNA [5], our analyses of nuclear DNA suggest that Agu pig populations have maternal Asian origins, but that strong nuclear genetic influences were imposed from European breeds after 1904. Nonetheless, PCA analysis showed that Agu pigs are genetically unique among European and Asian breeds (Figure 3). Our optimal structure analysis at K = 7, as determined using the Evanno ΔK method, also showed that Agu pigs form an independent cluster from the other breeds (Figure 4). In agreement, population differentiation values (F_ST) differed greatly between Agu pigs and European and Asian breeds (p<0.001; Table 2), confirming that Agu pigs are genetically distinct. Perhaps the bottleneck effect contributed to this distinctive genetic structure by promoting genetic drift. In addition, Agu pigs were isolated from other pig breeds by the conservation program from the 1980’s, likely enhancing genetic differences. In our NJ tree, which was based on proportions of shared allele distances, 34 Agu individuals were assigned to a single group that was not mixed with other breeds (Figure 2). These data suggest little gene flow between Agu pigs and other breeds in recent decades. However, we have not eliminated the possibility that other breeds contributed to the genetic structure of Agu pigs.

Ryukyu wild boar (Sus scrofa riukiuanus) has inhabited the Okinawa islands since the Pleistocene period [29]. In a previous study of mtDNA control regions, no Ryukyu wild boar haplotypes were found in an Agu pig population [5]. Our nuclear DNA analyses also revealed genetic distinctions between Ryukyu wild boar and Agu pigs (Table 2; Figures 1–4), suggesting little genetic influence from Ryukyu wild boar.

Structure analysis at K = 8 subdivided Ryukyu wild boar into two subpopulations (Figure 4), and one of these included individuals from Okinawa Island, whereas the other subpopulation was from Ishigaki and Iriomote islands. Because these subpopulations are separated by sea, the resulting geographic impossibility of gene flows likely contributed to differentiation. Moreover, Ryukyu wild boar had the highest F_IS value (0.481) in this study (Table 1), and this may be a consequence of the Wahlund effect [30], which is caused by sampling from subpopulations. The low genetic diversity of this small island population may also have contributed to this high F_IS value [24].