For the analysis, 223 Pacific abalones (H. discus hannai) were collected from five coastal locations in Korea in 2003 and 2004 (Figure 1). To cover the main coastal areas in Korea, five hatchery populations were sampled from three different areas (Table 1). The samples were obtained from hatchery-reared populations used as broodstock for artificial reproduction in Ulsan, in the eastern coastal areas (US; 51 individuals); Namhae (NH; 49 individuals) and Wando (WD; 41 individuals) in the southern coastal areas; and Taean (TA; 39 individuals) in the western coastal areas. The samples for one hatchery-reared population were acquired from a population of hatchery-reared offspring in Geoje (GJ; 43 individuals) in the southern coastal areas. The WD hatchery managed a large broodstock composed of wild-caught and hatchery-produced adult Pacific abalones, the GJ hatchery kept no broodstock but has used hatchery-produced adults as parents for reproduction, and the others had small broodstocks composed of hatchery-produced adults. Although the original ancestors of the hatchery abalone were collected locally and the hatchery populations had been reared continuously, no detailed records of their founding and maintenance were available.

Mantle musculature clips were preserved in 99.9% ethanol before being transported to the lab. Total DNA was extracted using the automated DNA extraction system MagExtractor MFX–2100 (Toyobo) with a MagExtractor-Genomic DNA Purification Kit (Toyobo, Osaka, Japan). Extracted genomic DNA was stored at −20 °C prior to polymerase chain reaction (PCR) analysis.

In total, 223 Korean Pacific abalones from the five populations were genotyped using six Pacific abalone microsatellite loci. Hdh1321, Hdh513, Hdh512 and Hdh145 developed for Haliotis discus hannai [15] and Hdd114B and Hdd229 developed for Haliotis discus discus [14] were used to amplify alleles by PCR. The 5’-end of the forward primer of each set of primers was labeled with fluorescent dye (6-FAM, HEX, or NED; Applied Biosystems, Foster City, CA, USA). PCR amplification of the six microsatellite loci was performed in 10-μL volumes containing 0.25 U Taq DNA polymerase, 10× ExTaq buffer, 2 mM dNTP mixture (Takara, Shiga, Japan), 2 μM of each primer set and approximately 10 to 50 ng template DNA using a PTC-0220 DNA Engine Dyad Peltier thermal cycler (MJ Research, Inc., Waltham, MA, USA). PCR conditions included an initial denaturation at 95 °C for 11 min, followed by 35 cycles of denaturation at 94 °C for 1 min, annealing for 1 min at each primer temperature listed in Table 1, and extension at 72 °C for 1 min, with a final extension at 72 °C for 5 min.

For genotyping, 1 μL of PCR product was added to 9 μL of a reaction containing formamide (Hi-Di Formamide, Applied Biosystems, Warrington, UK) and the GeneScan 400HD [ROX] size standard (ABI PRISM, Applied Biosystems, CA, USA), denatured at 95 °C for 2 min, and immediately chilled on ice. Fragment analysis of the reaction products was performed using an ABI 3130 Genetic Analyzer (Applied Biosystems) and GeneMapper software (ver. 4.0; Applied Biosystems). To improve accuracy when determining allele sizes, a control DNA sample was included in each set of samples for each run.

Statistical genetic analyses were conducted for five populations of H. discus hannai. The data were tested for PCR errors due to null alleles, stuttering, and allele dropout using MICRO-CHECKER ver. 2.2.3 (1000 randomizations) [21]. Scoring and human error were estimated by duplicate analyses. The data were also tested whether one or more microsatellites were under selection using LOSITAN software [22,23], which is a selection detection workbench based on a well evaluated F_ST-outlier detection method. Two genetic diversity parameters, the number of alleles per locus (N_A) and the number of unique alleles (U), were determined for each local sample at each locus, using the program Genepop ver. 4.0 (http://kimura.univ-montp2.fr/~rousset/Genepop.htm). Allelic richness (A_R) was corrected for the smallest sample size (n = 39), using the rarefaction method of FSTAT ver. 2.9.3.2 [24]. Using FSTAT, allelic richness can be directly compared among populations, regardless of sample size [25]. For the analysis of molecular variance (AMOVA) [26], components of variance within and between populations based on the infinite allele model (IAM) were estimated using the program Arlequin ver. 3.0 [27]. The significance of AMOVA components was tested using 1,000 permutations. To estimate genetic heterozygosity among the entire set of pairwise population samples, unbiased expected (H_e) and observed (H_o) heterozygosity values [28] were calculated. Overall inbreeding coefficients (F_IS) [29] for each population and locus were also estimated; F_IS can measure deviations from the Hardy–Weinberg equilibrium (HWE) within a population. Deviations from the HWE were tested using probability tests or exact tests by the Markov-chain procedure of Arlequin.

The extent of population subdivision was examined by calculating global multilocus F_ST values [29] and R_ST values (1000 permutations) [30]. The index of pairwise F_ST based on an infinite alleles model (IAM) was estimated and R_ST values based on a stepwise mutation model (SMM) were calculated using Arlequin. The index R_ST incorporates the correlation of the weighted mean allele size, expressed as the number of tandem repeats. Significance levels were adjusted for multiple tests using the sequential Bonferroni correction technique [31].

The genetic distance between populations was estimated on the basis of the chord distance, D_CE [32]. D_CE is one of the most efficient distance measures in obtaining correct tree topology from allele frequency data [33]. To examine genetic relationships between populations, a phylogenetic tree was constructed on the basis of pairwise genetic distances for all samples using the UPGMA method of the program POPULATIONS ver. 1.2.30 (http://bioinformatics.org/~tryphon/populations/). Bootstrap values were calculated using 1,000 replicates. The UPGMA tree was visualized using the Tree Explorer program (http://evolgen.biol.metro-u.ac.jp.TE/).

Relationships between geographical populations were assessed using principal component analysis (PCA) of GenAlEx 6.3 (http://www.anu.edu.au/BoZo/GenAlEx/), which was based on the covariance matrix of gene frequencies.

Because hatchery populations are often subjected to founder effects and bottlenecks that result in lower genetic diversity, Bottleneck software ver. 1.2.02 [34] was employed to test the bottleneck hypothesis under a two-phased model of mutation (TPM). This method can be used for testing the departure from mutation-drift equilibrium based on heterozygosity excess or deficiency.

Micro-Checker analysis did not detect any allele scoring errors caused by stuttering or large allele dropout, but it indicated that five of six loci could carry null alleles (Hdd114B was not affected; P < 0.05). However, all six loci were used in this study because no null alleles affected all populations.

Lositan analysis did not detect any outlier loci that had excessively high or low F_ST compared to neutral expectations, indicating all microsatellites used were not candidates for being subject to selection.

Estimates of genetic variability for the five Pacific abalone populations are summarized in Table 2. These results suggest that all of the microsatellite loci were polymorphic, with some level of polymorphism, including large differences in the number of alleles, in all of the studied populations. N_A and A_R ranged from 20.17 in the TA population to 26.67 in NH and from 20.17 in TA to 24.78 in NH, respectively, with the degree of variability differing among the six loci. The mean expected and observed heterozygosities per locus ranged from 0.885 in WD to 0.919 in NH and from 0.752 in NH to 0.838 in WD, respectively. However, despite these differences in genetic diversity, no significant decrease in genetic variability was found among the population samples (Kruskal-Wallis test, P > 0.05). In total, 56 alleles were found to be unique to a single population. The greatest number of unique alleles was detected in NH (20), while TA had only seven unique alleles. The inbreeding coefficient F_IS of the hatchery populations ranged from 0.054 to 0.184 (mean F_IS = 0.121 ± 0.056).

The observed genotype frequencies were tested for agreement with HWE (Table 2). Among the 30 population-locus cases, in 18 cases the observed genotype distribution was generally in accordance with Hardy-Weinberg proportions; however, 12 (40%) departed significantly from HWE, even after adjustment of P values using the sequential Bonferroni method (P < 0.01), and these disequilibriums (DHWE) were detected at all loci except Hdh114B. Ten DHWE cases resulted from a deficiency of heterozygotes and two from an excess of heterozygotes (Table 2). Eleven of the 30 F_IS values at the six loci estimated for the five populations were significantly different from zero (P < 0.01). Significant deviations were not evenly distributed among samples or loci, nor were they associated with a particular locus or sample. DHWE was observed at one locus in WD, but at four loci in GJ.

In total, 233 different alleles were observed across all loci and all samples, ranging from 15 for the Hdh154 locus to 64 for the Hdh513 locus. No population had a diagnostic allele. In the five populations, 64.0 to 76.6% of the alleles were rare alleles with a frequency <5%. They were detected at most loci and were not associated with a particular locus in either population.

In testing the departure from mutation-drift equilibrium based on heterozygosity excess or deficiency for five populations, bottleneck analysis was conducted using Bottleneck software under the TPM of microsatellites. No population displayed significant heterozygosity excess (P > 0.05) through the sign test, standardized differences test and Wilcoxon sign rank test, suggesting that five hatchery populations have not experienced a recent bottleneck.

Low but significant genetic differentiation (overall F_ST = 0.009, P < 0.01) was observed among populations, indicating that the level of genetic heterogeneity among them was low. Table 3 shows the pattern of genetic differentiation among populations observed by comparing D_CE and mean pairwise F_ST and R_ST values using composite allele frequency data. Values of pairwise F_ST among populations were significantly different from zero in all pairwise comparisons except one (all P < 0.01 after sequential Bonferroni correction, except for P = 0.086 between NH and US). On the other hand, the pairwise R_ST among populations was significantly different from zero in only five pairwise comparisons. Estimates of F_ST (range 0.0040 to 0.0176) were generally lower than R_ST values (range −0.0082 to 0.1016). The D_CE values were similar to the F_ST and R_ST values, in the range of 0.018 to 0.026. The AMOVA from the distance matrices for 233 individuals permitted partitioning of the overall variation. The amount of variation attributable to differences between individuals within populations and within-individual differences were large (12.09 and 87.06, respectively; P < 0.05), while only a low but significant proportion of variation (0.85%; P < 0.05) was detected among populations. The UPGMA tree and the PCA scatter plot constructed on the basis of the D_CE values showed that no population formed a separate cluster (Figures 2 and 3). The US population from the eastern coast and the NH population from the southern coast were the most closely related populations, while GJ from the southern coast was the most isolated from the others, the same groupings found with the PCA scatter plot analysis. Together, pairwise F_ST and R_ST tests, the UPGMA tree, and PCA on the five cultured populations of H. discus hannai did not show a consistent relationship between geographic and genetic distances.