Upwelling and eddies affect connectivity among local populations of the goldeye rockfish, Sebastes thompsoni (Pisces, Scorpaenoidei)

Abstrat The goldeye rockfish, Sebastes thompsoni, commercial rockfish catch in the Northwest Pacific Ocean, may influence its population structure. To clarify the population genetic structure of Korean S. thompsoni and its degree of hybridization with the most close species, Sebastes joyneri, we analyzed a mitochondrial (mt) DNA control region and eleven polymorphic microsatellite (ms) loci. S. joyneri individuals were clearly distinguished from S. thompsoni by the mtDNA control region and ms loci results, with single interspecific hybridization between two species suggesting no impact on genetic structure of S. thompsoni. Analysis of mtDNA revealed no population structure within S. thompsoni, suggesting the survival of a single population in southern refugia during the glacial period. The ms loci results, in contrast, showed two genetically distinct clusters within S. thompsoni: One was predominant throughout Korean coasts (from the Yellow Sea, via the Korea Strait to the East Sea); the other was predominant at Dokdo Island in the East Sea; and both occurred in similar ratios at Wangdolcho Reef in the East Sea. A possible factor that restricts gene flow between Korean coastal and offshore populations in the East Sea may be related to the complex oceanic current patterns such as eddies and upwelling, which represent impermeable barriers to population connectivity for this species. Our findings highlight that these two populations might be representative of two separate stock within Korean waters and maintain their geographically related genetic structure.


| INTRODUC TI ON
In marine fishes, high dispersal capability and successful settlement during the early pelagic life stages provide opportunities for habitat expansion, maintenance of sustainable population sizes, and the exchange of genetic material between geographically distant populations (Jones et al., 2009;Kritzer & Sale, 2010;Strathmann et al., 2002). Maintaining relatively strong site fidelity to small home ranges is one of the most important factors in determining genetic breaks and forming population structure in reef fishes (Doherty, Planes, & Mather, 1995;Shulman & Bermingham, 1995).
From the past three decades, many population genetic studies of marine fishes have demonstrated unique genetic structures may be caused by the restricted dispersal of larvae and juveniles (Froukh & Kochzius, 2007;Miller & Shanks, 2004;Roberts, 1997). Restricted dispersal between populations may be due to geographic barriers or oceanographic patterns such as ocean currents, eddies, and upwelling, which cause limited connectivity even among neighboring populations, especially during pelagic life stages (Buonaccorsi et al., 2004;Rocha-Olivares & Vetter, 1999).
The rockfish genus Sebastes Cuvier, 1829 comprises approximately 110 species worldwide, despite their relatively recent divergence (Hyde & Vetter, 2007;Love, Yoklavich, & Thorsteinson, 2002;Nelson, Grande, & Wilson, 2016). Because most adult rockfishes have relatively strong site fidelity (Mitamura et al., 2002;Starr, Heine, Felton, & Cailliet, 2002), it is a taxon with high potential for forming significant within-species population structure. Furthermore, as Sebastes species are ovoviviparous (i.e., the female releases free-swimming larvae), the range and direction of transport during dispersive phases of the early life cycle are potentially important influences on population structure (Love et al., 2002).

These factors, which may hinder gene flow between populations in
Sebastes species, have the potential to promote population differentiation to occur more frequently than in other marine fish (especially pelagic fishes). Conversely, because speciation within Sebastes occurred fairly recently (Briggs, 1995;Hyde & Vetter, 2007), much evidence of hybridization has been reported among closely related species due to incomplete or relaxed reproductive barriers in sympatry (Artamonova et al., 2013;Muto, Kai, Noda, and Nakabo, 2013;Saha et al., 2017). This hybridization is a challenge to delimit species and populations.
The goldeye rockfish, Sebastes thompsoni (Jordan & Hubbs, 1925), is an important component of the commercial rockfish catch in the Northwest Pacific Ocean. In Korea, the range of this species includes the Yellow Sea, the Korea Strait, the East Sea, and around Jejudo Island (Joo, 2006;Kim & Ryu, 2016). The distribution of S. thompsoni in Japan extends southward from southern Hokkaido to Tokyo and Tsushima Island .
Japanese halfbeak (Hyporhamphus sajori) and Korean rockfish (Sebastes schlegelii) are good examples to demonstrate this; both species spend their early life stages in drifting seaweed, showing very high genetic homogeneity despite the geographic distance between populations Uchida & Shojima, 1958;Yu, Kai, & Kim, 2016).
In Japan, a previous study of S. thompsoni around the Japanese Archipelago using seven microsatellite (ms) loci found evidence of genetic homogeneity with the exception of one ms locus (Sekino, Takagi, Hara, & Takahashi, 2001). Sekino et al. (2001) asserted that the Tsushima Warm Current might be a major driver affecting genetic connectivity among local populations of Japanese S. thompsoni. Meanwhile, there are two conflicting viewpoints for the role of Tsushima Warm Current, branches from the Kuroshio Warm Current, in shaping the genetic population structure between the Korean Peninsula and Japanese Archipelago. The Tsushima Warm Current might promote the dispersal of some pelagic larvae and juveniles, and provide the opportunity for gene flow (Kasai, Komatsu, Sassa, & Konishi, 2008;Kim, Bae, Lee, & Yoon, 2017;Song, Gao, Ying, Yanagimoto, & Han, 2017). However, some previous study has shown that the disconnections among the Yellow Sea, Korea Strait, and Japanese Archipelago are related to the influence of different water masses such as the Tsushima Warm Current and Yellow Sea Bottom Cold Water (Han, Kim, Tashiro, Kai, & Yoo, 2017). Therefore, despite the previous study on Japanese S. thompsoni, further investigation of genetic diversity and connectivity among Korean S. thompsoni populations from various oceanographic perspectives is required.
Sebastes joyneri (Günther, 1878) is the species most closely related to S. thompsoni, from the perspectives of morphology and genetics (Hyde & Vetter, 2007;Kai, Nakayama, & Nakabo, 2003;. This species has a greater preference for warmer habitats than S. thompsoni, despite some overlap in their distributions (Jejudo Island, Wangdolcho Reef, and Dokdo Island in Korean waters; Kim et al., 2005;. Sebastes joyneri is often seen forming small groups of a few individuals, mixing with large flocks of S. thompsoni (personal observation).
Speciation event between S. thompsoni and S. joyneri was fairly recently (Hyde & Vetter, 2007), and their present sympatry may be a secondary contact with incomplete or relaxed reproductive barriers after speciation event because their main distributional ranges are somewhat different. Furthermore, the opportunity for introgression will also increase when the spawning periods and areas of two species are in close proximity (Kijewska, Burzyński, & Wenne, 2009;Kim et al., 2017;Montanari, Van Herwerden, Pratchett, Hobbs, & Fugedi, 2012). Therefore, the introgressions occurring between close species in genus Sebastes may affect genetic population structure within species.
The objective of this study was therefore to clarify the connectivity among local populations of Korean S. thompsoni and its degree of hybridization with S. joyneri using the mitochondrial (mt) DNA control region and 11 polymorphic microsatellite loci.

| Sampling
Samples of S. thompsoni were collected from its main habitats throughout Korean waters, from seven locations: Eocheongdo Island Sample identification was conducted according to the morphological characteristics described by Kim et al. (2005)  . Species abbreviations were chosen using the initials of their scientific names (e.g., S. thompsoni: ST; S. joyneri: SJ) and the first letters of the sampling location (e.g., Eocheongdo Island: Eo), joined by an underscore. The tissue samples were preserved in 99% ethanol and stored at −20°C until DNA extraction. The specimens used in this study were deposited at Pukyong National University (PKU).
The cycling conditions were an initial denaturation step for 1 min at 96°C followed by 35 cycles of 1 min at 96°C, 1 min at 56°C, and 2 min at 72°C, with a final step of 10 min at 72°C. The PCR products were purified using a LaboPass PCR Purification Kit (Cosmogenetech Co.; www.cosmogenetech.com). The mtDNA was sequenced on an ABI 3730XL Sequencer (Applied Biosystems; www.appliedbiosystems.com) using an ABI BigDye Terminator Cycle Sequencing Ready Reaction Kit v3.1 (Applied Biosystems).
We also selected 11 microsatellite markers from the loci designed for S. thompsoni: Sth3A, Sth3B, Sth24, Sth37, Sth45, Sth56, Sth86, and Sth91 (Sekino, Takagi, Hara, & Takahashi, 2000) and S. schlegelii: KSs2A, KSs3, and KSs27A (An, Park, Kim, Lee, & Kim, 2009). All markers were labeled with FAM, HEX, and TAMRA fluorescent dyes for the forward primer. Multiplex PCR amplification of nine markers was performed in three sets, and singleplex PCR was also performed for two markers (Sth37 and KSs3) due to limitations of conditional matching. The protocol followed was that outlined by Sekino et al. (2001) and An et al. (2009) for the PCRs, fragment amplification, and scoring the microsatellite loci. The multiplex and single PCR products were combined and analyzed on an ABI 3730XL Sequencer (Applied Biosystems) using the GeneScan 500 LIZ dye size standard (Macrogen Inc.; www.macrogen.com). The alleles were scored with GeneMapper 3.7 (Applied Biosystems).

| Data analysis
Because the analyzed sample size in the mtDNA control region was relatively small, the analysis included only an examination of the basic molecular characteristics and phylogenetic relationships of both species. The mtDNA sequences were aligned using ClustalW (Thompson, Higgins, & Gibson, 1994) in BioEdit 7 (Hall, 1999).
Significance levels for multiple tests were adjusted by the sequential Bonferroni correction (0.05/11C2 = 0.0009; p < .0009). The model-based Bayesian clustering procedure in STRUCTURE 2.3.4 (Pritchard, Stephens, & Donnelly, 2000) was used to determine how many genetic groups (K) best represented the study areas.
In this analysis, an admixture model without prior sampling location information was used and allowed for correlations between allele frequencies (Falush, Stephens, & Pritchard, 2003). Posterior probabilities were generated for K values 1-9 using 1,000,000 iterations and the MCMC method, with a burn-in of 500,000 iterations, to calculate the probable K value. STRUCTURE HARVESTER 0.6.94 (Earl & vonHoldt, 2012) was used to apply the Evanno method (Evanno, Regnaut, & Goudet, 2005) to determine the value of K that was appropriate for a genetic cluster. Individuals used in STRUCTURE analysis were identified as possible hybrids if at least 10% of their genome (q) originated from other groups (Randi, 2008). Discriminant analysis of principal components (DAPC; Jombart, Devillard, & Balloux, 2010) was used to identify clusters based on the spatial distribution of microsatellite genotypes, using the adegenet 2.0.1 package in R 3.2.2 (www.r-project.org; Jombart, 2008). Initially, genetic data were transformed into uncorrelated components using principal component analysis (PCA), and principal components (PC) were then submitted to discriminant analysis (DA). In this study, 50 PCs were retained, containing 80% of the variation in the data.

Analysis of molecular variance (AMOVA) was performed to
test the best hierarchical groupings based on two alternative hypotheses in sampling areas using Arlequin 3.5.1.2 (Excoffier, Smouse, & Quattro, 1992). The first hypothesis is that Korean S. thompsoni populations in sampling areas form only one cluster.
The second hypothesis is that Korean S. thompsoni populations are divided into two clusters (Eo, Ch, Yo, So, and Wa-1 vs. Wa-2, Do) in sampling areas. Here, we separated the Wangdolcho Reef population of S. thompsoni into two genetic clusters (Wa-1 and Wa-2); this decision followed criteria for the majority cluster (Cluster 2 or 3) in the STRUCTURE analysis (K = 3). The best grouping was defined by the highest differentiation among groups (F CT ) and nonsignificant differentiation within groups (F SC ). F-statistics were obtained in each analysis from 10,000 permutations in Arlequin 3.5.1.2 (Excoffier et al., 1992). Finally, isolation-by-distance (IBD) TA B L E 1 Summary of molecular diversity, mutation neutrality, and statistics across mtDNA control region (765 bp) and 11 microsatellite loci for six S. thompsoni and three S. joyneri populations from each sampling location in Korean waters   Bohonak, 2002;Jensen, Bohonak, & Kelley, 2005), to test for correlation between genetic and geographic distances. Abbreviations for sampling locations are defined in Figure 1 and Table 1 from HWE were confirmed for only five independent loci (KSs3

| Introgressive hybridization between Sebastes species
In the NJ tree based on the mtDNA control region sequences, S. thompsoni and S. joyneri were clearly distinguished, except for one individual (PKU_5153, collected from Dokdo Island) that identified morphologically as S. thompsoni but was genetically similar to S. joyneri. The average genetic distance (d) between S. thompsoni and S. joyneri was 0.047 ( Figure S2).
According to the STRUCTURE analysis based on 11 ms loci, when K = 2, all S. joyneri individuals were clearly separated from those of S. thompsoni, which comprised two genetically different compositions (Figure 2a) Purebred genotypes = q ≥ .90; hybrid genotypes = 0.1 < q < .9. Intraspecific hybrid: hybrid genotypes between Clusters 2 and 3; interspecific hybrid: hybrid genotypes between Clusters 1 and 2 or 3. n: number of specimens (mtDNA control region); and N: number of haplotypes. Abbreviations for geographic locations are defined in Figure 1 and Bold values showed significant differences (p < .05). *Differences were significant after Bonferroni correction (p < .0009).
(.1 < q < .9) (Figure 2b, Table 2). However, except for single individual PKU_5153, the remaining three putative hybrid individuals cannot be recognized as hybridization because they did not share a rare allele derived from S. joyneri.

| Genetic population structure
The NJ tree of S. thompsoni populations based on the mtDNA control region sequences showed a single panmictic group ( Figure S2), supported by no significant differences of pairwise F ST values (p > .05).
Hence, we found no evidence of genetic population structure even when using hypervariable mtDNA control region sequences. In contrast to the mtDNA analysis results, two clusters were found in the   (Table 4). Finally, there was no indication of genetic isolation by distance between the geographic ranges of individuals collected ( Figure S3).
Additionally, for S. joyneri individuals collected from three locations in Korean waters, no significant difference was found in F ST values of mtDNA (p > .05) and msDNA (p > .0009) data (Table 3), and there was no evidence of genetic population structure identified with the STRUCTURE or DAPC analysis using msDNA data (Figures 2 and 3). null alleles produced very similar results of genetic differentiation as when they were included (Dick, Shurin, & Taylor, 2014;Lane, Symonds, & Ritchie, 2016;Underwood, Travers, & Gilmour, 2012).

| D ISCUSS I ON
Furthermore, it is known that null alleles could not change the overall patterns of genetic population revealed in ms loci analysis (Carlsson, 2008).  (Grant & Bowen, 1998). However, reduced levels of genetic diversity in mtDNA and heterozygous deficiency in ms loci were re- It is also unlikely that effect of selection or immigration occurred at Chujado and Yokjido Island in Korea Strait because of their main stable habitat. Thus, genetic drift within the both sites is the most likely explanation for relative low genetic diversity, implying a reduction in effective population size due to local overfishing (Frankham, Ballou, & Briscoe, 2010). Actually, most fishery for this species is concentrated in Korea Strait, including Chujado and Yokjido Island, for decades due to the relatively low depth of water and the ease of accessibility, so it has been exposed to the risk of local overfishing. The effects of local overfishing on genetic diversity have been demonstrated at the population study of New Zealand snapper (Pagrus auratus) that showed the significant decrease in both heterozygosity and the mean number of alleles in one population (Tasman Bay) during its exploitation history (Hauser, Adcock, Smith, Ramírez, & Carvalho, 2002). Similarly, in this study, these declines were also observed in the Chujado and Yokjido Island populations (Table 1).
Therefore, Korean S. thompsoni showed a high genetic diversity, but the difference of diversity among sampling locations has been identified due to human activity.
According to Muto et al. (2013), interspecific hybridization between Sebastes vulpes and Sebastes zonatus, occurring in the Northwest Pacific Ocean, was detected in a total of 63 (35.6%) individuals using a combination of amplified fragment length polymorphisms (AFLP), mtDNA, and morphometric characters. Interestingly, this study showed that rates of hybridization varied considerably among the locations, which is associated with variations in habitat segregation, implying that the extent of habitat segregation of the two species dependent on vertical water temperature regimes determined the opportunity for hybridization. Four Sebastes species in the North Atlantic Ocean have also shown evidence of interspecific hybridization when in sympatry (Artamonova et al., 2013;Pampoulie & Daníelsdóttir, 2008;Roques, Sevigny, & Bernatchez, 2001).  (Randi, 2008) but also shares the pure both parental genotypes (q ≥ .90). The occurrence ratio of hybridization between the two species was considerably lower than expected, despite the high likelihood of hybridization due to extensive overlap in the mating and release seasons of the two species (Kokita & Omori, 1998;Nagasawa & Kobayashi, 1995;Yang et al., 2016). Consequently, in this study, the observed low hybridization rates between Korean S. thompsoni and S. joyneri suggest that there was no effect on the determination of population structure of Korean S. thompsoni.

Moreover
Korean S. thompsoni was found to be genetically indistinguishable between sampling locations using the mtDNA control region sequence. Previous study in the Japanese Archipelago, which was performed preliminarily for 10 individuals in only two regions (East Sea and Pacific coast), respectively, suggest that genetic population structure is likely a single panmictic group like this study, but showed a higher genetic diversity (overall h = 1.0000, π = 0.0345) rather than Korean S. thompsoni (overall h = 0.9998, π = 0.0231; Higuchi & Kato, 2002). This would suggest that stable population levels and extensive gene flow in Japan have fostered the generation and persistence of very high genetic diversity (Grant & Bowen, 1998 Im, Jo, Ji, Myoung, & Kim, 2017 (Yu et al., 2016). Yu et al. (2016) suggested that the decline of surface seawater temperatures in the East Sea might have been caused by the blockage of the Tsushima Warm Current during the last glacial maximum, resulting in the local extinction of East Sea species and a subsequent substantial drop in genetic diversity, which is currently observed. S. joyneri would also have suffered similar demographic events, but probably more severely due to its preference for warmer habitats than S. thompsoni (Konishi & Nakabo, 2007). Therefore, Korean S. thompsoni comprised a single panmictic mitochondrial group, perhaps due to the survival of a single population in southern refugia, and a subsequent rapid recent expansion and migration to the East and Yellow Seas.
Unlike the results of mtDNA control region sequences (present study) and those of seven ms loci (Sekino et al., 2001) (Kokita & Omori, 1998, 1999Nagasawa & Kobayashi, 1995 Figure 5). These eddies flow between the Wangdolcho Reef and Dokdo Island, and may facilitate genetic exchange between them, while at the same time possibly precluding genetic exchange between Korean coastal populations and offshore populations ( Figure 5).  (Palof et al., 2011;Siegle, Taylor, Miller, Withler, & Yamanaka, 2013). According to Palof et al. (2011), it is likely that such eddies may contribute to genetic discontinuities by entraining or preventing pelagic larvae and juveniles from moving.
Therefore, the strong eddies around the Ulleungdo and Dokdo Islands in the East Sea might cause self-replenishment, suggesting a potential for allopatric differentiation over time.
Wangdolcho Reef, located 23 km offshore from the western margin of the East Sea, is influenced by two major currents (North Korean Cold current and East Korea Warm Current) with upwelling and small eddies, which occurs mainly along the western Reef (Lee & Myoung, 2003). Coastal upwelling acts as a physical barrier to larval dispersal for marine fishes (Graham, Field, & Potts, 1992), and also strongly divides water masses, enhancing the larval fish assemblage dichotomy (Tiedemann & Brehmer, 2017 (Shim et al., 2008), related to upwelling or eddies as mentioned above.
Although Sokcho is closer to Dokdo Island and Wangdolcho Reef than the other sampling locations in this study, the Sokcho population differed from these populations in the multiplex tests (e.g., STRUCTURE, DAPC, and AMOVA; Figures 2 and 3, Table 4). This is a good evidence of connectivity between the Korea Strait populations (i.e., the Yokjido and Chujado Islands) and the East Sea coastal population (Sokcho), which may be explained by a combination of (1) the physical characteristics of warm currents (such as Tsushima Warm Current and East Korea Warm Current) prevailing in the surface layer (Choi et al., 2012;Lee & Niiler, 2005) and (2) the biological characteristics of pelagic larvae and juveniles attaching to drifting seaweed (Kokita & Omori, 1998, 1999Nagasawa & Kobayashi, 1995).
Most S. thompsoni populations in the Yellow Sea exhibit patchy distribution due to lack of suitable habitat and are exposed to the unique marine environments (semi-enclosed, strong tidal current, shallow water, and high-turbidity) and paleoclimatic change (Kim, 2009;Xu & Oda, 1999). These conditions may facilitate population subdivision due to limited gene flow and adaptive radiation (Gunderson, Vetter, Kritzer, & Sale, 2006;Mayer, Schiegg, & Pasinelli, 2009). However, this study did not detect any evidence of genetic variation between the Eocheongdo Island population