Comparative phylogeography of a bathymetrically segregated pair of sister taxa of rockfishes (genus Sebastes): black rockfish, Sebastes melanops, and yellowtail rockfish, Sebastes flavidus

Twelve pairs of sister taxa in the speciose rockfish genus, Sebastes, overlap coastal distributions but are bathymetrically segregated. These pairs are ideal for comparative studies to understand how life-history traits, historical events, and environment interact to produce population genetic structure. Black rockfish, Sebastes melanops, forms one such pair. Its sister species, yellowtail rockfish (Sebastes flavidus), shows a genetic cline likely influenced by a dispersal barrier at Cape Mendocino, CA and northward range expansion. Due to geographic overlap and close systematic relationship, we predicted black rockfish was influenced by similar evolutionary processes and thus would show genetic pattern concordance with yellowtail rockfish. We analyzed ~ 1000 black rockfish from 22 sites spanning the species’ range to test the null hypothesis of no structure, using the same markers that characterized yellowtail rockfish (i.e., 812 bp of the mitochondrial cytochrome b gene and six microsatellite loci). We reject the null hypothesis based on existence of at least three populations and microsatellite genetic divergence that separates the Alaskan and Continental U.S. populations (FCT = 0.021, p << 0.001), and a mitochondrial genetic cline near Cape Mendocino (FCT = 0.132, p < 0.01). We also found single collections genetically divergent from neighboring collections. Like yellowtail rockfish, oceanographic dispersal barriers and northern range expansion were inferred to influence black rockfish, however, unlike yellowtail rockfish, Cape Mendocino did not split the range into two stocks and was, therefore, inferred to be a less severe barrier. We hypothesize a higher frequency of extinction/recolonization events in black rockfish populations may have led to more complex genetic structure.


Introduction
Phylogeographic breaks have been observed across biogeographic breaks in both terrestrial and marine environments. In marine environments, the causes for these phylogeographic breaks have been linked to oceanographic currents and other factors that can limit exchange of migrants, such as availability of suitable habitats. Although some species show concordant phylogeographic patterns (Goldstien et al. 2006;Martins et al. 2022), there can be discordant genetic structure across closely related taxa (e.g., DiBattista et al. 2012). The reasons for this variance are challenging to explain but are likely due to multiple processes that influence population structure including dispersal abilities (e.g., non-planktonic dispersers tend to share concordant phylogeographic breaks, Pelc et al. 2009). Studies have shown correlation of degrees of population structure with variation in life history characteristics (shallower depth preferences of adults leading to higher levels of population structure, Hickey et al. 2009;Sivasundar and Palumbi 2010). A population genetic study of a taxonomic group harboring high diversity of life history characteristics, such as the Pacific rockfishes (genus Sebastes) would be ideal to advance our understanding of Responsible Editor: O. Puebla . how evolutionary processes interact with life history to produce population structure.
The rockfishes are the most species-rich and ecologically diverse group of marine fishes inhabiting the temperate rocky reefs of the west coast of North America (Love et al. 2002;Gunderson and Vetter 2006). Most species occupy discrete rocky habitats separated by soft sediments, thus fitting the definition of marine metapopulations which are networks of populations that are affected by extinction and colonization events (Hanski and Gilpin 1997;Gunderson and Vetter 2006;Kritzer and Sale 2006). They are live-bearers with generally low adult movement and high site fidelity. They instead depend on pelagic dispersal of larvae and juveniles as a primary means of colonization and population connectivity (Love et al. 2002;Starr et al. 2002;Parker et al. 2007).
The subgenus Sebastosomus, as genetically redefined by Hyde and Vetter (2007), contains five species that form schools or aggregations associated with high-relief habitats. This assemblage is sometimes referred to as the "structure schooling ecological group" (Lenarz et al. 1995). In addition to the adult schooling behavior, the five species share pelagic larval and juvenile characteristics and inter-annual recruitment success tends to co-vary with oceanographic conditions, such as the Pacific Decadal Oscillation (PDO) and El Nino-La Nina events (Lenarz et al. 1995). The subgenus contains two divergent evolutionary lineages. The first includes the shallow-dwelling pair of incipient species blue rockfish (Sebastes mystinus, (Burford and Bernardi 2008) and deacon rockfish (S. diaconus, Frable et al. 2015), and the deeper living widow rockfish (S. entomelas). The second lineage contains olive rockfish (S. serranoides), black rockfish (S. melanops), and yellowtail rockfish (S. flavidus). Of the three species, black and yellowtail rockfishes are most closely related (Hyde and Vetter 2007;Hess et al. 2011) but differ ecologically in bathymetric distribution. As adults, black rockfish remain in near-shore rocky habitats, whereas yellowtail rockfish migrate to deeper waters and are most common over rocky reefs at depths from 90 to 180 m (Love et al. 2002). Both species are highly resident as adults (Stanley et al. 1994;Lea et al. 1999;Parker et al. 2007), in fact the average home range size of black rockfish was estimated to be 55 ± 9 ha (Parker et al. 2007). Black and yellowtail rockfishes share a broadly overlapping latitudinal range from northern Baja California, Mexico to southeastern Alaska but black rockfish continue west to the tip of the Aleutian Archipelago. Both species have peak parturition of larvae in winter (Moser 1996), co-occur as pelagic juveniles for 3-4 months (Love et al. 2002), and share settlement timing and macro-and micro settlement habitat preferences for the kelp canopy and for artificial settlement collectors (Ammann 2004).
Studies on the phylogenetic relationships within the genus (Rocha-Olivares et al. 1999;Hyde and Vetter 2007) provide an evolutionary context for examining the processes of dispersal, gene-flow, and speciation by a comparative genetic approach that examines closely related sister taxa within a subgenus. Examining species that share a common lineage and most but not all key life history characteristics can provide insights into the role of bathymetry, oceanography, and climate in determining dispersal patterns, gene-flow, population structure, and speciation. These insights require taxonomic replication with the groups that are being compared. Although some rockfish sister taxa appear to show expected allopatric signals of speciation (e.g., north and south species on either side of a biogeographic boundary), it is a recurrent theme among Sebastes that many sister taxa overlap across their latitudinal distributions but are bathymetrically segregated into a shallow and deep living species pair. Even more interesting, these species pairs often share a similar time of parturition, a similar pelagic phase, followed by shared juvenile settlement preference for near-shore rocky reef habitat, but subsequent bathymetric segregation of the adults as one species migrates to deeper water, while the other remains a lifetime resident in shallow water. At least twelve bathymetrically segregating sister taxa have been identified within the genus Sebastes (Table 5, Hyde et al. 2008). The study of bathymetrically segregating sister taxa may provide comparative insights into life-stage specific dispersal and gene-flow, as well as the role of bathymetry in the speciation process (Hyde et al. 2008).
Our goal was to compare the population genetic structure of black rockfish (Sebastes melanops) to the published population genetic data set of its bathymetrically segregating sister species, Sebastes flavidus (Hess et al. 2011). Specifically, we tested whether black rockfish exhibited the following two main patterns found in the yellowtail rockfish study, (1) a genetic barrier to dispersal at Cape Mendocino, CA and (2) northern range expansion. We expect this concordance due to the close phylogenetic relationship, overlapping geographic distributions, and similar life-histories that result in similar evolutionary processes, such as physical barriers to dispersal and climate fluctuations. However, these species have different depth distributions, where adult black rockfish are found in shallower waters relative to yellowtail rockfish. Past studies hypothesized that the different depth distributions may result in greater population structure of the shallower dwelling species (Hickey et al. 2009, Sivasundar and Palumbi 2010, Von der Heyden et al. 2013). Thus, we expect higher levels of population structure and steeper isolationby-distance gradients in the shallower-dwelling black rockfish compared to the deeper-dwelling yellowtail rockfish.
Although there have been multiple population genetic studies conducted on black rockfish showing significant genetic differences among collections (e.g., Miller et al. 2005, Seeb and Seeb 2005, Lotterhos et al. 2014, the genetic markers and geographic study areas were not ideal for comparing results with the yellowtail rockfish study (Hess et al. 2011). Differences in study designs make it difficult to conclude whether population genetic structure characterized by these black rockfish studies are due to marker mutation rates, unique geography, demographic fluctuations between different time periods, or variance due to sample size. Here, we used the same DNA markers, six nuclear microsatellite loci and mitochondrial cytochrome b, from the yellowtail rockfish study to characterize population structure of black rockfish. These two marker types are informative, because they address different temporal scales of genetic structure; microsatellites represent recent time scales compared to the older events resolved by mitochondrial DNA (Papura et al. 2012). By comparing our black rockfish study to a previous study on its sister taxa yellowtail rockfish (Hess et al. 2011) using similar markers, spatial scales of sampling, and sampling time frames; we have the ideal set of conditions to examine how shared environment affects species in similar ways and whether bathymetric preferences shape population structure in predictable ways.

Sampling
The tissue samples used for DNA extraction were muscle and fin clips stored in 95% ethanol. All collections were sampled during the 2005 fishing season (June-September) using the catches from various recreational charter boats. Collections from Alaska were obtained by Alaska Department of Fish and Game through coordination with their port sampling program (Fig. 1, Table 1). Seventeen of the 22 total collections had body size information (lengths in cm), which were summarized by averages and ranges in values for each collection. DNA was extracted using DNeasy 96 Tissue Kits with a high throughput liquid-handling BioRobot 8000 (QIAGEN). Samples from each port were pooled into collections and GPS coordinates were calculated by averaging the coordinates of the individuals. Fig. 1 Map of black rockfish sample sites, haplotype frequency charts, and nested clade. Circles with notations represent observed haplotypes. The notation indicates the name of the haplotype. Each line connecting haplotypes represents a single mutation. Small solid circles represent unobserved haplotypes. Groups of haplotypes were nested and assigned a clade number (indicated within the outlined nested group). The square represents the mtDNA haplotypes from a sister species (yellowtail rockfish) used as an outgroup, and indicates clade 1-1 as ancestral. Colors in the pie charts correspond to those colors used in the haplotype network

Molecular methods and data quality
Polymerase chain reaction (PCR) amplification was used to amplify an 833 bp region that includes the first 812 bp of the cytochrome b mitochondrial region using the primers described in Hess et al (2011). PCR conditions and extraction procedure are the same as in Hess et al. (2011). Six microsatellite loci were genotyped using the methods of Hess et al. (2011).
The genotypic data were organized in MICROSOFT EXCEL and formatted for various population genetic statistical programs using the EXCEL MICROSATELLITE TOOLKIT version 3.1 (Park 2001). MICROCHECKER 2.2.3 (Van Oosterhout et al. 2004) was used to examine all collections and loci for microsatellite null alleles and scoring errors due to stuttering and large allele dropout. Estimates of F IS were calculated according to Weir and Cockerham (1984). All collections were tested for deviations of Hardy-Weinberg equilibrium and all loci were tested for linkage disequilibrium using GENEPOP version 3.4 on the web [1000 dememorizations, 2000 batches, 1000 iterations per batch; (Raymond and Rousset 1995)]. Fisher's method was used to combine probabilities across all loci for each collection and probabilities across all collections for each locus.

Putative genetic barriers to dispersal
We performed a principal component analysis (PCA) on the microsatellite allele-frequency data with the program PCAGEN 1.2.1 (Goudet 1999), for heuristic identification of major genetic differentiation among collections (i.e., genetic discontinuity along the linear coastal distribution of collections). A multidimensional scaling (MDS) method was performed on the mtDNA pairwise F ST matrix using the software package SPSS 11.0.2 (SPSS for MAC OSX 2003).
STRU CTU RE v. 2.3.2 (Pritchard et al. 2000) was used to determine how many populations might be represented among our collections. An initial burn-in of 50,000 iterations followed by 500,000 iterations of the Markov Chain Monte-Carlo method was used to generate posterior probabilities. We used the feature in STRU CTU RE that takes into account spatial information (i.e., from which collection an individual was sampled) to modify the prior distribution for each individual's population assignment to increase the ability to detect structure at low levels of divergence (Hubisz et al. 2009). We ran 10 iterations of K values, 1 through 9, and averaged the estimated log-likelihoods [ln(PrK)] to determine the most likely value of K. We calculated Delta K using the method described by Evanno et al. (2005). After selection of K = 3, 40 iterations were run and the 10 iterations with the highest ln(PrK) were averaged using the FullSearch algorithm in CLUMPP (Jakobsson and Rosenberg 2007), which finds the best configuration of those 10 iterations by optimizing pairwise similarity (H'). To display the STRU CTU RE results on geography we used only collections with sample sizes greater than 10 and plotted the average probability of ancestry to each cluster for each collection according to their latitudinal coordinate (for continental U.S. collections) and for Alaska collections we added the Euclidean distance between longitudinal coordinates in a cumulative way to approximate a scaled coastline distance in units of degrees. This latitudinal scale allowed easier comparison of population structure patterns between black rockfish and yellowtail rockfish (Hess et al. 2011) studies. FSTAT 2.9.3.2 (Goudet 2001) and ARLEQUIN 3.1 (Excoffier et al. 2006) were used to measure F-statistics for the genotype and sequence data, respectively. Where the heuristic methods above detected a regional genetic discontinuity, we divided collections into regions separated by the center of the genetic cline and contrasted the regions using an analysis of molecular variance (AMOVA, Excoffier et al. 1992), a feature in ARLEQUIN 3.1 (Excoffier et al. 2006). ARLEQUIN 3.1 was used to estimate allelic richness, genediversity, and nucleotide diversity. The parameters of allelic richness and gene diversity were used to characterize levels of diversity of the microsatellite variation and plotted versus latitude to test for any significant correlations.
F ST was calculated for all pairwise comparisons of collections. Isolation-by-distance (IBD) regressions were examined at the following two spatial scales: (1) range-wide, and (2) regional (dividing the ranges at points, where a genetic discontinuity is centered). Collections were considered linear and geographical distances were calculated along the coastline rather than as the crow flies. Only collections with a sample size of ten or greater individuals were used to test for isolation by distance by regressing pairwise F ST comparisons with pairwise geographic distance. The regional IBD tests were useful to understand whether an isolated regional break may be confounding a true signal of IBD. To remove these potentially confounding effects from our IBD signal, we tested IBD separately for the collections located on either side of the center of genetic breaks. PASSAGE 1.0 (Rosenberg 2001) was used to perform a series of Mantel and partial Mantel tests for comparing the correlation coefficients of a model of IBD versus a model of fragmentation. Two Mantel tests and two partial Mantel tests were performed for each species. The Mantel tests compared a matrix of genetic distance F ST /(1-F ST ) (Rousset 1997) to either a matrix of pairwise geographical distances (used to assess the overall fit of an IBD pattern) or a matrix of ones and zeros used to categorize collections on the same side or opposite sides of the center of a genetic cline (used to assess the overall fit of a genetic regional break). The partial Mantel tests compared the matrix of genetic distance with one of the other two matrices while holding the remaining matrix constant. Significance for these Mantel tests was tested using 9,999 permutations.

Range expansion and other demographic patterns
BOTTLENECK 1.2.02 (Cornuet and Luikart 1996) was used to test all collections for recent effective population size reduction or expansion. The stepwise mutation model (S.M.M.) was used for reporting results, because it is likely the best fit for microsatellite data FSTAT version 2.9.3.2 was used to calculate allelic richness, gene-diversity (Hs), and pairwise F ST . P values for population differentiation were calculated by permutation.
Effective population size based on linkage disequilibrium at all six microsatellite markers was calculated with LDNE which uses Burrows method to calculate r 2 , a measure of linkage disequilibrium (Waples and Do 2008;Waples 2006). LDNE was run using the three default p critical values 0.05, 0.02, and 0.01 and with confidence intervals using jackknifing over loci.
We constructed haplotype trees using statistical parsimony (Templeton et al. 1992) using the program TCS v. 1.21 (Clement et al. 2000) to directly compare results across the black rockfish and yellowtail rockfish (Hess et al. 2011) studies using the same markers and methodology. The haplotype trees were used to define a series of hierarchically nested clades following the nesting rules given in Templeton et al. (1987) and Templeton and Sing (1993).
The haplotype network was rooted using the outgroup species (yellowtail rockfish). We used the one-step clade frequencies for each collection to display these mtDNA clade patterns on geography. Similar to the methods described above for displaying the patterns in microsatellite STRU CTU RE on geography, we used a latitudinal scale to plot collections by coastline distances in latitude for continental U.S. collections and distances in longitude for Alaska collections.

Basic statistics
Average body sizes generally increased with latitude (Table 1, Supplemental Fig. 1). For the southernmost collections with size information (collections 2, 3, and 5), average body size was 30.8 cm; the intermediate collections (8-13) averaged 42.0 cm; and the northernmost collections (15-22) averaged 49.0 cm. In general, each of these groups of collections had similar variation in ranges of lengths across collections within each group; however, collection 8 had highest standard deviation in lengths within the group of collections from "intermediate" latitude, and collection 15 had highest standard deviation in lengths within the group of collections from the northern latitudes (Supplemental Fig. 1).
The total haplotype network that connects all haplotypes sequenced in black rockfish and includes a connection to a sister species outgroup (yellowtail rockfish) is displayed (Fig. 1). There were 56 haplotypes found in black rockfish for a length of 833 bp of aligned sequences for 997 individuals (Genbank accession nos. FJ664347; OP718400-OP718454). There were 54 polymorphic sites which were comprised of 48 transitions (TS) and 6 transversions (TV) for a TS TV ratio of 8. There were 45 synonymous changes and 9 nonsynonymous changes. The haplotype diversity was 0.3154 ± 0.177 and the nucleotide diversity was 0.000475 ± 0.000490.
The six microsatellite loci did not show any significant linkage disequilibrium using Fisher's method to combine p values for each locus pair across all collections. However, using sequential Bonferroni correction to examine pairs of loci within collections, showed a single locus pair (Spi4 and Sal3) within one black rockfish collection (collection 3, San Francisco n = 55) as being significant at the 0.05 alpha level.
Two black rockfish collections, collection 7 (Crescent City n = 11) and collection 13 (La Push n = 63) had significant positive F IS values after sequential Bonferroni correction (0.448 and 0.147 at Sra16-5 and Sal3, respectively; Supplemental Table 1). However, Fisher's method indicated only collection 7 as having an overall trend in departure from HWE across loci (p < 0.01). Analysis with Fisher's method across all collections, Spi10 was found with significant departure from HWE (p < 0.01).
MICROCHECKER version 2.2.3 showed evidence of null alleles at locus Spi10 for three collections (Black rockfish collections 3, 15, and 19). In addition, one collection (13) showed evidence of null alleles at locus Sal 3.

Putative genetic barriers to dispersal
A multidimensional scaling (MDS) plot was generated from pairwise F ST comparisons based on mtDNA gene frequencies in collections of n ≥ 10 (Supplemental Fig. 2). This plot Species were analyzed separately, and the frequencies of the 1-step mitochondrial clades were calculated for each collection with greater than 10 individuals and then these average frequencies were interpolated between collections to create a continuous 100% stacked chart that was scaled to latitudinal position of each collection. Yellowtail rockfish data came from Hess et al. (2011) helped to visualize the trends in the pairwise F ST matrix (Table 2). We found that collections 2, 3, and 4 diverged from most of the other collections except for collection 9 which clustered with collections 2, 3, and 4. The haplotypic compositions of the collections revealed a large increase in clade 1-6 haplotypes (southern distributed haplotypes shown in red, Fig. 2) at these collections compared to relatively higher percentages of the ancestral clade 1-1 haplotypes (dark blue haplotypes, Fig. 2) at neighboring sites. Collections 2, 3, 4, and 9 formed most of the significant pairwise F ST comparisons with the full data set (all of these four collections had high frequencies of clade 1-6 haplotypes). The mtDNA data set showed an overall F ST of 0.11. AMOVA results showed most of the genetic variation (89.3%) is captured within collections. When the black rockfish range is divided into two regions at the center of the genetic cline between collections 4 and 5, these regions represent 13.15% of the genetic variation and yield an F CT = 0.132, that is significant (p < 0.01). Making a third region that consists of the single divergent collection 9 increased the percent of genetic variation (16.9%) and increased F CT (0.169, p << 0.001).
We performed a principal components analysis (PCA) on the black rockfish microsatellite allele frequency data of all collections with sample size n ≥ 10 (Supplemental Fig. 3) and identified two regional groups (Alaska collections 15-22 versus continental U.S. collections 2-14) along the primary axis. Furthermore, there was a single collection 8 that appeared to be separated from all other collections on the y-axis. The patterns in this PCA plot were consistent with the significant F ST values that were observed in the genetic distance matrix (Table 2). For example, there were significant F ST values between Alaskan and U.S. continental collections as well as between the single collection 8 with most other collections. Analysis of the six microsatellites across all collections yields an overall F ST value of 0.017 with 95% confidence intervals (CI) from bootstrapping over loci (0.011-0.022). These values were the same even when only collections with ten or greater individuals were considered. Exclusions of the locus and collection shown to deviate from HWE did not significantly alter these values for either species. For example, when Spi10 is excluded, the F ST value was 0.018 (95% CI 0.012-0.023).
Most of the microsatellite genetic variation (97.3%) was captured within collections; however, there was also a significant portion of variation that was captured by dividing the sampled ranges according to the genetic cline indicated by the heuristic examinations with PCA. Separating the range into two regions between the Alaska and the continental U.S. range captured a significant amount of variation (2.1%; F CT = 0.021, p << 0.001). Making a third region that consists of a single collection (collection 8 at Brookings, OR) increased this percent genetic variation (2.4%) and F CT value (0.024, p << 0.001).
There are clearly two different regions that appeared to be important in the microsatellite data versus the mitochondrial data for black rockfish. The relatively large genetic divergence between neighboring collections (i.e., genetic break) detected in the heuristic analyses indicated a location in California centered between collections 4 and 5 for the mitochondrial data set but the only genetic break in the continental U.S. coast observed in the microsatellite data was centered on collection 8. Furthermore, a larger genetic regional break appeared in the microsatellite data set between the continental U.S. and Alaskan collections. In contrast, for the black rockfish mtDNA data, this continental U.S. versus Alaska division yielded a lower percentage of the genetic variation (4.4%, F CT = 0.044, p < 0.05) than was found when we divided by the Californian cline (13.15%, F CT = 0.132, p < 0.01).

Range-wide level
At the range-wide level, we detected a significant IBD regression for both DNA marker data sets (Table 3). The black rockfish mtDNA data set showed a significant pattern but the R 2 and slope were low: 0.024 linearized F ST per 1000 km (R 2 = 0.076, p < 0.001, Fig. 3a). These values for the black rockfish mtDNA IBD regression were only slightly improved by omitting collection 9, which was genetically divergent from its neighbors. For the microsatellite data set the range wide IBD equates to 0.007 linearized F ST per 1000 km (R 2 = 0.270, p < 0.001; Fig. 3b). When we excluded the black rockfish collection 8 at Brookings, OR (highly genetically divergent from all other collections) we observed an IBD regression with a higher R 2 0.008 linearized F ST per 1000 km (R 2 = 0.648, p < 0.001).

Regional level
We next examined IBD regressions on the regional level by examining IBD patterns on either side of the steep genetic clines that were identified in the heuristic analyses. The two regions examined for the mtDNA data set included collections north and south of the break between collections 4 and 5. Neither region showed a significant IBD regression, unless collection 9 was excluded from the northern region, in which case the IBD regression was significant and yielded the following values: 0.007 linearized F ST per 1000 km (R 2 = 0.105, p = 0.001; Table 3).
The microsatellite data were split into regions separating the continental U.S. and Alaska. The continental U.S. was not significant, and the Alaskan range showed a significant negative slope: − 0.004 F ST per 1000 km (R 2 = 0.365, p = 0.004). When collection 8 (Brookings, OR)

Mantel correlations are higher for regional divisions versus isolation by distance
We found that using both mtDNA and microsatellite data sets, the Mantel test comparing regional affiliation versus genetic distance matrix yielded a higher correlation, which is the result expected if a regional genetic break were a better fit to the data compared to a model of IBD. For example, the black rockfish mtDNA Mantel test of pairwise geographical distance versus pairwise linearized F ST yields a significant correlation of 0.28 (right-tailed and two-tailed p < 0.05). In contrast, the Mantel test of the two regions north and south of collections 4 and 5, yields a correlation of 0.51 (righttailed and two-tailed p < 0.01). The black rockfish microsatellite Mantel test of pairwise geographical distance versus pairwise linearized F ST yields a correlation of 0.52 (righttailed and two-tailed p < 0.001). Whereas the Mantel test of Alaska and the continental U.S. regions versus pairwise linearized F ST yields an improved correlation of 0.55 (righttailed and two-tailed p < 0.001).

Structure
The STRU CTU RE results were consistent with other microsatellite analyses. For example, in black rockfish no more than three populations could be interpreted (Supplemental Fig. 4). Moreover, it appeared that high proportions of membership in the three population clusters identified by STRU CTU RE were geographically restricted to Alaska, continental U.S., and collection 8 at Brookings, OR, respectively (Fig. 4). This result corroborates our previous findings that the regions of Alaska and the continental U.S. are significantly genetically divergent from each other and our finding that collection 8 is significantly genetically divergent from all other collections. Finally, when we removed collection 8 from the STRU CTU RE analysis, the results supported no more than two populations (Supplemental Fig. 4), indicating that without collection 8 only the regional break between Alaska and the continental U.S. would be present.

Range expansion and other demographic patterns
Black rockfish had low mitochondrial gene diversity in the northern portion of its range. Gene diversity was consistently low from the coast of Washington at collection 12 and continuing north through the Alaskan range (Supplemental Fig. 5). The significant pairwise F ST values (Table 2) correspond with the collections that have a large discrepancy in mtDNA gene diversity. The low gene diversity was not exclusive of the Alaskan collections and was variable across the continental U.S. collections including 6, 7, 10, 12, 13, and 14 (Supplemental Fig. 5). There was no correlation between microsatellite diversity and latitude (Supplemental Table 1). We also split the collections into two groups according to whether they were in the continental U.S. or Alaska and estimated whether these allelic richness or gene diversity parameters were significantly different between these regions. Over all loci, neither allelic richness nor gene diversity was significantly different between these regions. The BOTTLENECK results suggest collection 8 is different from the others indicated by a deficit of heterozygosity (i.e., lower observed heterozygosity compared to expected heterozygosity) that corresponds across all six loci (Supplemental Fig. 6). Four loci (Spi6, Spi10, Spi4, and Sal3) showed significant (p < 0.01) deficits of heterozygosity under the SMM model. There were no range wide trends in the difference in observed versus expected heterozygosity across all loci which is consistent with the lack of correlation of the other diversity measures with latitude reported above.

MtDNA haplotypic distributions
The nested haplotype network of black rockfish (Fig. 1) and each 1-step group is color coordinated with the chart showing the 1-step clade frequencies across all collections shown (Fig. 2). The details of specific numbers of haplotypes found in each collection are given in Supplemental Table 2. Two noteworthy patterns were resolved by viewing the 1-step clade frequencies across geography. The first pattern involves haplotype H, which was nested within clade 1-1. Clade 1-1 is dominated by one haplotype (A) which represents 812 individuals; haplotype A is likely the ancestral haplotype with an outgroup probability greater than 95%. Haplotype H (n = 21) is the only other haplotype within this clade that had greater than n = 5. Haplotype H is restricted geographically to Alaskan collections west of collection 15 in Yakutat, AK (Fig. 1). The geographical center of haplotype H is located significantly far from the geographic center of its nesting clade which may indicate range expansion.
The second major pattern involves clades 2-1 and 2-2. The tip clade 2-2 was restricted to the continental U.S. coast, primarily south of Washington State. Clade 2-1 is well-supported as the most ancestral clade, due to it being the closest connection to other species.  Table 1).

Comparison of genetic barriers for black and yellowtail rockfishes
Our population genetic study of black rockfish found evidence of more than one stock. In fact, based on microsatellites, there may be at least three populations along the species range; one concentrated in the south (continental U.S. West Coast), one that is concentrated at a single collection (Brookings, OR), and one that is concentrated in the north (Western Alaska). Our analyses of both mitochondrial sequences and nuclear microsatellite genotypic data from black rockfish collections reveal the presence of abrupt genetic clines (i.e., short isolation-by-distance Two genetic clines occurred on relatively small geographic scales within the continental U.S. West Coast of our study (within 2-3 degrees latitude). A relatively abrupt genetic discontinuity revealed by our mtDNA data set was centered between latitudes 38-40 (collections 4 and 5, south of Cape Mendocino near Point Arena). This is similar to the pattern in yellowtail rockfish which exhibited a steep cline that included those same latitudes but was centered at Cape Mendocino (Latitude 40.5; Hess et al. 2011). However, unlike yellowtail rockfish, this genetic discontinuity was not as severe, it did not divide the entire species into two separate regional stocks, and it may be centered slightly further south (between collections 4 and 5, Point Arena). The composition of mitochondrial haplotypes that comprise this cline showed relatively high frequency in southern haplotypes, which then decrease rapidly going northward (collection 7, Crescent City, CA); and then the southern haplotypes again increase to a high frequency (collection 9, Charleston, OR). These two transitions represent the two clines exhibited in the mitochondrial structure. The microsatellite genetic structure showed discontinuity within the continental U.S. West Coast centered around Brookings, OR (collection 8). However, this black rockfish Fig. 4 Spatially scaled frequencies of ancestry to STRU CTU RE clusters for a black rockfish and b yellowtail rockfish collections using microsatellite DNA analysis. Species were analyzed separately and three clusters (black, white, and gray) were resolved for black rockfish and two clusters (shown in black versus white shading) were resolved for yellowtail rockfish (data from Hess et al. 2011). Only collections with N ≥ 10 samples are included in this interpolation of individual probability of ancestry to each cluster which was averaged across individuals within each collection microsatellite discontinuity contrasted with the patterns in yellowtail rockfish in a manner similar to the characteristics of the black rockfish mitochondrial clines, because it did not cleanly split the continental U.S. West Coast range into two stocks. Finally, a second significant microsatellite genetic discontinuity was centered on the region between the Alaskan and continental U.S. collections. This discontinuity may be less abrupt than the U.S. West Coast genetic discontinuity, but it is unclear whether this is simply due to a lack of samples between Neah Bay, WA and Yakutat, Alaska. The entire Canadian range is unsampled and may reveal more abrupt transitions if data were available.

Northward range expansion
The microsatellite data from the yellowtail rockfish study (Hess et al. 2011) had provided some indication that northern range expansion may have been part of the historical processes affecting population structure in the species. We would generally expect low genetic diversity in the northern collections of a species that has experienced rapid range expansion at least for cases in which genetic drift is high and population sizes are small (Excoffier et al. 2009). However, highly dispersive species like marine fishes may be capable of maintaining genetic diversity at the extents of their range (Ray and Excoffier 2010). Although black rockfish is a marine species, it may not exhibit high dispersive ability due to the fact that it displays some degree of substructure; therefore, it is not clear whether the process of northern range expansion would result in low genetic diversity. Characterization of mtDNA haplotype distributions provided some evidence of a recent expansion of black rockfish into Alaska and at least some indication that mtDNA diversity may have been affected by this event. The cytochrome gene diversity in this study was consistently lower in Alaskan collections versus continental U.S. collections. In fact, mtDNA genediversity appears significantly lower in the northern portion of the species range starting north of the Columbia River at the border of Oregon and Washington and continuing into Alaska. Furthermore, in a previous microsatellite study by Seeb and Seeb (2005) they find a decreasing allelic richness going west into Alaska. In contrast, we did not find a decreasing trend over all loci with regard to allelic richness in our microsatellite data (we used a different set of loci than Seeb). The fact that low genetic diversity was not apparent across both marker types (mtDNA and microsatellites) used in this study may owe to the intermediate nature of the dispersal of this marine species. Specifically, black rockfish may exhibit high enough dispersal to prevent drift from resulting in low diversity across all marker types despite a putative range expansion.
There are patterns in the geographic distributions of mtDNA haplotypes that could indicate that a recent range expansion has occurred. Specifically, one indicative pattern of range expansion is when a relatively new haplotype (tip clade) exhibits a relatively geographic widespread distribution and can spread relatively far from the geographic center of the distribution of the ancestral haplotype (Templeton 1998). We observed this type of pattern which involved haplotype H (tip clade) that was shown to be distributed relatively far from the center of the geographic distribution of its ancestral haplotype (A) at the extreme end of the range in Western Alaska and it had a relative large distribution observed across seven collections (collections 16-22). Similar to our microsatellite results, an abrupt genetic cline was described in black rockfish within Alaska. Seeb and Seeb (2005) found support for a genetic discontinuity at the Alaska Gyre and suggested the direction of ocean currents in the Gulf of Alaska (GOA) restricts dispersal. The main results were: (1) significant pairwise F ST values between all western GOA collections and every southeastern Alaska and Pacific Northwest collection, and (2) a trend toward decreasing allelic richness starting from the Pacific Northwest to the western GOA collections. However, ocean currents acting as gene flow barriers would not fully explain the second result. An alternative explanation is the western GOA collections were founded by a relatively small group of individuals (lacking in genetic variation), thus resulting in the observed decrease in allelic richness. One reason for expecting this founder scenario in the Alaskan range of this species is due to the last glacial maximum (LGM) that occurred just 20,000 years ago, which may have made this northern coastal habitat unsuitable to nearshore species. Glaciers and their associated ice shelves are hypothesized to have covered much of the continental shelf between the Alaska Peninsula and British Columbia (Mann and Hamilton 1995).

Extinction and recolonization events
We have emphasized how isolation-by-distance can be one mechanism to explain genetic patterns of black rockfish collections along the U.S. West Coast. However, two collections were genetically divergent from neighboring collections in mtDNA (collection 9) and microsatellite (collection 8) data sets. These collections had relatively large sample numbers (n ≥ 50); yielded consistently high F ST levels across a majority of collections in the data set, including their nearest neighboring collections; and were both located in southern Oregon on either side of Cape Blanco.
The collections 8 and 9 do not appear to be part of any regional isolation-by-distance trends but appear to be isolated events at small spatial scales. These single divergent collections that lacked isolation-by-distance trends with surrounding collections are unlikely to be explained by sampling error (i.e., samples sizes of these collections were Page 13 of 17 62 not small relative to the rest of the collections in the data set), rather they were likely the result of local extinction and recolonization events. Collection 8 at Brookings, OR, was one of two collections estimated with low effective population size, which lends support for a local extinction and recolonization event. In addition, collection 8 shows a deficit in heterozygosity across several loci that may be indicative of a recent increase in population growth from low effective population size (i.e., results from BOTTLENECK). Collection 8 was genetically differentiated from all other collections in the species range according to the microsatellite data. Interestingly, a previous study of black rockfish identified Brookings, OR as a location with aberrant collections found to deviate from HWE (Miller et al. 2005). We suggest that this area near Brookings, OR may have undergone a recent increase in population size from a small group of founders that could account for collection 8 being highly genetically differentiated from all other collections in this study and the fact that there are deficits of heterozygosity across multiple loci.
Despite these extinction and recolonization events, the overall population structure of black rockfish compared to yellowtail rockfish was not substantially higher as measured by overall F ST . For example, based on microsatellites black rockfish had an overall F ST of 0.017 (95% CI 0.011-0.022) compared to yellowtail 0.011 (95% CI 0.008-0.015). Based on mitochondrial DNA black rockfish data showed that splitting the data set between collections 4 and 5 and making collection 9 into a third region yielded an F CT (0.169, p << 0.001) that was lower than the F CT represented by splitting the yellowtail rockfish range at Cape Mendocino (F CT = 0.32, p << 0.001; Hess et al. 2011). Whether the depth preference differences between these species led to an overall greater number of extinction and recolonization events in black rockfish is an interesting question worth further consideration. However, regardless of possible higher frequency of these extinction and recolonization events, these events did not lead to greater overall genetic differentiation across the range but rather greater localized differentiation of neighboring sites. For example, the concentration of southern haplotypes at the southern end of the range appears suddenly absent near Point Arena (between collection 4 and 5) then continues to be absent until further north around collection 8 and 9 at Cape Blanco, before disappearing again at northern neighboring collection at Newport, OR (10). Stochasticity from recurrent extinction and recolonization events would tend to create a disrupted pattern of haplotypic composition along the coast.
One possible cause of extinction and recolonization may be related to levels of anoxia observed in the Oregon range of black rockfish (Chan et al. 2008). Dissolved oxygen measured from 0 to 800 m depth during upwelling season (mid-April to mid-October) between 42 and 46° N latitude (approximately the area between collection 7 and 11, Fig. 1) crossed the severe hypoxia threshold (below 0.5 ml L −1 ) in 2006, which was a rare event from 1950 to 1999 but has increased from 2000 to 2005 (Chan et al. 2008). Chan et al. (2008) reported that this severe hypoxic event affected all their cross-shelf transect lines between 44.25°N and 45.00°N (approximately the area between collection 9 and 11, Fig. 1), extending from the shelf break to the inner shelf (under 100 m depth) and encompassing at least 3000 km 2 . Furthermore, Chan et al. (2008) showed there was a complete absence of all fish from the rocky reefs normally occupied by rockfishes when they conducted submersible surveys in August 2006. This particular event would have been most likely to affect collection 10 that was obtained in the summer of 2005 within the hypoxic region and before that area was found in summer 2006 to be vacant of rockfishes (Chan et al. 2008). Black rockfish would have been one of the species impacted by this phenomenon as adults mostly occur in less than 55 m depth (Love et al. 2002). Interestingly, collection 10 had significant pairwise F ST based on mtDNA, and lower mtDNA gene diversity compared to its two neighboring collections 9 and 11. Fish may not suffer direct lethal effects from these hypoxic conditions, since they can move to avoid them, but the effects can be similar to extinction and recolonization events as large areas are displaced and later replaced with different compositions of fishes.
It is possible that aspects of the reproductive biology of black rockfish, particularly bet-hedging strategies (e.g., Sogard et al. 2008) combined with sweepstakes type recruitment mechanisms (e.g., Markel et al. 2017), could be an alternative explanation for the genetic disruptions we have observed. However, we hypothesize that extinction/colonization is more likely regarding the Brookings, OR event because of the patterns described above as well as the fact that sweepstakes events would lead to a narrow range of age classes (or size classes as a proxy of age classes) which we did not observe in the case of our collection from Brookings, OR which had a high standard deviation in body lengths relative to neighboring collections. Interestingly, Miller et al. (2005) also found that a collection from Brookings, OR showed the widest variance in ages which further supports that a sweepstakes event was less likely at this particular location. Nonetheless, in general, sweepstakes recruitment is a possible cause for black rockfish genetic disruptions given that the period of parturition (when larvae are extruded) occurs between mid-January and mid-March with a peak in February and older females tend to release larvae earlier than younger females (Bobko and Berkeley 2004). It also appears that faster growth and higher survival was linked to older females likely due to contributing a larger energy storage oil globule to larvae . In this way, localized areas may receive segments of larval recruitment originating from a relatively small numbers of adults.

The influence of seascape on genetic structure
The three locations of steep genetic clines from this study (two in Central California close to Point Arena and Cape Mendocino and one in the region between Alaska and Continental U.S.) have all been discussed previously as locations of putative mesoscale oceanographic dispersal barriers in population genetic studies of rockfish (reviewed by Gunderson and Vetter 2006;Hyde and Vetter 2009). The complete list of barriers includes Alaska Gyre, Queen Charlotte Sound, Puget Sound, the Cape Mendocino jet (e.g., yellowtail rockfish, Hess et al. 2011), and the Southern California Eddy (e.g., cowcod rockfish, Hess et al. 2014; sunset rockfish, Longo et al. 2022) and these locations have all been shown to correlate to genetic discontinuities in rockfishes (Gunderson and Vetter 2006). Even outside of the rockfish genus, other marine taxa (fishes and invertebrates) also show genetic discontinuities at these locations (Hare and Avise 1996;Arndt and Smith 1998;Lecomte et al. 2004;Marko 2004;Cimmaruta et al. 2005;Hickerson and Cunningham 2005;Hickerson et al. 2006;Wilson 2006;Petersen 2007). Besides correlative evidence from genetic studies, Point Conception and Cape Mendocino (Briggs 1974;Williams and Ralston 2002;Cope 2004) are observed to be biogeographic breaks, where species ranges terminate. In addition, Cape Mendocino and Cape Blanco are regions, where groundfish species assemblage structure has been found to shift (Tolimieri and Levin 2006).
The possible mechanisms responsible for the way in which these locations could act to limit dispersal may be due to currents (Magnell et al. 1990;Marchesiello et al. 2003;Longhurst 2007) or bathymetric features such as submarine ridges and canyons (Williams and Ralston 2002). Longhurst (2007) defined a series of geographic compartments within the California Current Ecological Province and these coast upwelling regions explained genetic patterns in the vermillion rockfish species complex (Hyde and Vetter 2009). At various capes along the west coast and particularly at Cape Mendocino, there are major upwelling quasi-permanent eddies (Marchesiello et al. 2003) and they are strongest in spring and summer when rockfish juvenile recruitment occurs. For regions around Monterey and Cape Mendocino California there is also a decrease in total habitat in waters shallower than 200 m, which affects the common depths inhabited by both yellowtail and black rockfish (Williams and Ralston 2002).
The collections that were genetically divergent from neighboring collections (collection 8 at Brookings, OR and collection 9 at Charleston, OR) do not appear to be a result of a major dispersal barrier that disrupts connectivity in a way that splits this species into two separate stocks. However, Cape Blanco is located in between Brookings and Charleston, OR and there may be a localized isolating effect of currents in the vicinity of Cape Blanco that are sufficient in cutting off recruitment from neighboring areas to produce a recruitment sink. In addition, a large swath of sandy habitat between Charleston and Newport separates rocky habitat that is more suitable for black rockfish. Copper rockfish also show a genetic discontinuity between Charleston and Newport, Oregon (Buonaccorsi et al. 2002;Johansson et al. 2008). Aside from rockfish, four species of Pacific salmonids, Oncorhynchus mykiss (steelhead trout), O. clarki (cutthroat trout), O. kisutch (coho salmon), and O. tshawytscha (Chinook salmon) share similar Evolutionarily Significant Unit (ESU) designations and the northern border of ESU III is demarcated at Cape Blanco, OR.

Implications for fisheries management of black rockfish
Stock assessments are critical for management of marine fishes including black rockfish, as their purpose is to collect and analyze demographic data to characterize changes in abundance of fishery stocks. In the most recent stock assessment of black rockfish (Cope et al. 2015), Black rockfish was divided into three stocks stratified by the borders of California and Oregon and the borders of Oregon and Washington along the continental U.S. coast. The northern portion of the range is assessed and managed by Alaska Department of Fish and Game (ADFG) and by Department of Fisheries and Oceans Canada for the portions of the range in Alaska and Canada, respectively. ADFG assesses the species as three separate stocks in the Western Region: Kodiak, Chignik, and the South Alaskan Peninsula Area-Eastern District (Mattes and Sagalkin 2007).
The continental U.S. black rockfish assessments have in the past used a tagging study and allozyme genetic study (Wallace et al. 2008) as a basis for assessing black rockfish as two different stocks separated by Cape Falcon, OR (between collections 10 and 11). The more recent assessment considers three stocks and uses the state borders (CA and OR border is between collections 7 and 8; OR and WA border is between collection 11 and 12); however, our genetic results provide basis to break California into two stocks at Point Arena (between collections 4 and 5). In addition, our genetic results within Oregon provide support for another stock boundary near Cape Blanco (between collections 8 and 9). A previous genetic study of black rockfish by Lotterhos et al. (2014) also supports a stock boundary at Cape Blanco, because a collection at the extreme southern end of their study region, near Cape Blanco, was divergent from the rest of the range. Alaska is already managed separately from the continental U.S., and until more work can be performed, results from Seeb and Seeb (2005) showing a stock boundary west of Yakutat seem reasonable. A greater level of stratification along these suggested boundaries would create five black rockfish stocks along the U.S. continental coastline which may complicate stock assessments. However, the potential for gene-flow barriers identified as genetic breaks in this study may indicate that abundance trends of rockfish populations on either sides of these genetic breaks may be behaving in independent ways and could benefit by being assessed separately.

Concluding remarks
Our population genetic study on black rockfish establishes this species as an example of a marine species having more genetic structure on both fine and coarse spatial scales than would be expected given the large dispersal potential of their pelagic larval and juvenile life stages. We demonstrated how evolutionary processes such as restricted gene-flow with isolation by distance, range expansion, and extinction and recolonization appear to have influenced the observed genetic patterns. While black rockfish and yellowtail rockfish have overlapping ranges, these sister taxa are bathymetrically segregated. The hypothesis that black rockfish population genetic structure would resemble that of yellowtail rockfish was partially supported by this study. Both mitochondrial and microsatellite genetic structure shows discontinuities in similar regions of California near Cape Mendocino. These data, as well as a previous study (Seeb and Seeb 2005), support northern range expansion. However, black rockfish did not have a substantially higher degree of population structuring based on F ST as might be expected given a preference for shallower depths as compared to yellowtail rockfish. Black rockfish may, however, be more prone to extinction and recolonization events resulting in localized genetic differentiation between neighboring collections to a degree not observed in yellowtail rockfish.