Microsatellite variation in the Mexican rockfish Sebastes macdonaldi *

1 Centro de Investigación Científica y Educación Superior de Ensenada, Departamento de Oceanografía Biológica, A.P. 2732, Ensenada, Baja California C.P. 22860, México. E-mail: arocha@cicese.mx 2 Universidad Autónoma de Baja California, Facultad de Ciencias Marinas, A.P. 453, Ensenada, Baja California C.P. 22860, México. 3 National Oceanic and Atmospheric Administration, Southwest Fisheries Science Center, La Jolla Shores Blvd., La Jolla CA 92037, USA.


INTRODUCTION
Rockfishes of the genus Sebastes represent a challenge for the study of marine speciation and evolution.Among their most relevant biological features are their mode of reproduction, their large number of species and their large morphological and ecological diversities.Rockfishes all have internal fertilisation, an unusual feature among marine teleosts, and matrotrophic viviparity has been shown in some species (e.g.Boehlert et al., 1991).The very large number of species is undoubtedly their best known attribute, as Sebastes represents the most numerous genus in the family Scorpaenidae, with more than 110 species worldwide (Eitner et al., 1999;Nelson, 1994).In concert with their species-rich nature, rockfishes are also well known for their conspicuous morphological and ecological diversity.This is reflected in a variety of forms and colorations ranging from small and slender pelagic planktivores, like Sebastes goodei, to bulky and heavily spined sit-andwait predators, like Sebastes nebulosus (Eschmeyer and Herald, 1983).Finally, the antitropical distribu-tion (Fig. 1) of this group of cold-temperate fish has been considered one of the most interesting anomalies among scorpaenids (Eschmeyer and Hureau, 1971).Even though the origin of the genus appears to be among subtropical western Pacific scorpaenids (Washington et al., 1984), their present centre of distribution is on the western coast of North America, where more than 70% of the species are found in very high levels of sympatry (Chen, 1971;Rocha-Olivares et al., 1999c).
Of the seven species of Sebastes found in the Gulf of California (Chen, 1975), only S. macdonaldi is also distributed in the Pacific coast of Baja California (Chen, 1975;Moser, 1971).S. macdonaldi is the Northeast Pacific rockfish with the southernmost distribution, ranging from Point Sur, Central California, to offshore banks off Bahía Magdalena, Baja California (Fig. 1).The species distribution is discontinuous since it has not been reported further south at the entrance to the Gulf of California.Within the gulf, Mexican rockfish is found near Guaymas and in Bahía de Los Angeles (Chen, 1975;Moser, 1971;Thomson et al., 2000).
Populations in the gulf are thus isolated from those in the Pacific coast by the Baja California peninsula and by the influence, throughout most of the gulf, of warm tropical waters flowing North from the Costa Rica Coastal Current.This disjunct distribution raises a couple of fundamental biogeographical and evolutionary questions: (1) What is the level of differentiation of the populations residing in the Gulf of California?(2) Are populations within the Gulf of California vicariant relicts or are they descendants of a founder population that invaded from the Pacific?These questions have remained unanswered for a long time (Chen, 1975) and can be addressed using genetics.
Microsatellites are co-dominant molecular genetic markers inherited following Mendelian rules and consist of DNA tandem repeats dispersed throughout most eukaryotic nuclear genomes.Polymorphisms in these loci are manifested as variations in the number of tandem repeats, e.g.(AC)n, introduced during DNA replication, resulting from their insertions or deletions following mechanisms not entirely understood (Colson and Goldstein, 1999;Falush and Iwasa, 1999).Repeating motifs of microsatellites range from 1-10 base pairs (bp), but most loci frequently used range from 2-6 bp (Hancock, 1999).Due to their high mutation rates and levels of polymorphism, microsatellites are powerful markers to address a variety of genetic, ecological, and evolutionary questions (e.g. Baker, 2000;Dowling et al., 1996;Machugh et al., 1996).
Here we use microsatellites to study Mexican rockfish from Gulf of California and Pacific populations in order to test null hypotheses derived from the questions above: (1) the levels of molecular genetic diversity are the same in the gulf and in the Pacific, and a smaller genetic diversity would be expected if Gulf of California populations were derived from a small set of founder individuals that invaded from the Pacific; and (2) Gulf of California populations of Mexican rockfish are not genetically differentiated from those in the Pacific.

Sampling
Fish were obtained by hook and line from depths of ca.200 m between March 1995 and March 2002.Samples were taken from three localities off the Pacific coast of North America and one from the Gulf of California (Fig. 1).Because of the small sample size, data from Islotes San Benito (n = 8) were pooled for analyses with those from Uncle Sam Bank (henceforth referred to as sample BAJA, see Table 1).Tissue samples (muscle and liver) were dissected from fish and preserved in 95% ethanol (EtOH).Samples were maintained at room temperature in the field and at 4 °C in the laboratory until molecular analyses.

DNA extraction and purification
Total genomic DNA was extracted from 25-100 mg of preserved tissue digested overnight at room temperature in 500 µl DNAzol (Molecular Research Center, Inc.) and 100 µg proteinase K (Gibco BRL).DNA purification was carried out according to the manufacturer's protocol and purified DNA was precipitated in 250 µl 100% EtOH and rinsed in 800 µl 75% EtOH.Purified DNA was finally re-dissolved in 50 µl TE buffer (10 mM Tris-HCl, 1mM EDTA, pH 8.0).The quality and quantity of extracted DNA was visually assessed in a 1% agarose minigel (0.5X TBE) stained with ethidium bromide (0.5 µg/ml).

Microsatellite genotyping
Diluted PCR products (1:2 or 1:10) were electrophoresed in Long-Ranger denaturing 6% polyacrylamide gels (FMC Bioproducts).Amplified alleles were detected with a DNA automated sequencer (Gene Analyzer ABI 377) and their sizes were determined with the program GeneScan 3.3, using a fluorescent ladder (Rox 500) loaded in each experimental lane.

Genetic diversity
Based on a large dataset of Sebastes spp genotypes for these seven loci (R.D. Vetter, C.A. Kimbrell and E.A. Lynn, unpublished data), allele size (bp) was converted into number of tandem repeats.Individual diploid genotypes were represented as the combination of both alleles, expressed as their size in repeat numbers, to compute genotypic and allelic frequencies per locus for each locality.Expected heterozygosity (He) and the average squared difference in allele size (Rho) were also computed for each locus (Robertson and Hill, 1984;Weir, 1996).Goodness of fit to Hardy-Weinberg (H-W) expectations was calculated for genotypic frequencies and significance was determined using a Markov chain approach (Guo and Thompson, 1992) and linkage disequilibrium tests were performed for all loci pairs (Ohta, 1982).The previous analyses were performed with GENEPOP 3.3 (Raymond and Rousset, 1995).Levels of genetic diversity (He) among localities were tested non-parametrically with STATISTICA 5.5 (StatSoft Inc., 1999).

Genetic differentiation
Genetic structure was quantified as the level of heterogeneity in allelic and genotypic frequencies with the program GENEPOP 3.3 (Raymond and Rousset, 1995).The test of allelic differentiation provides an unbiased estimate of the level of significance of the test described in Raymond and Rousset (1995), whereas the genotypic differentiation test is based on the G exact test of Goudet et al. (1996).We also computed fixation indices as F ST , under the approximation of an analysis of variance (Cockerham, 1973;Weir and Cockerham, 1984), and its analogous R ST , for population pairwise comparisons.The latter incorporates the correlation of the weighted mean allele size expressed as the number of tandem repeats (Michalakis and Excoffier, 1996).Sequential Bonferroni correction (Rice, 1989) was used to adjust α-levels for multiple testing.Finally, we performed an analysis of molecular variance (AMOVA, Excoffier et al., 1992) using values of the pairwise number of distinct alleles (F ST ) or the sum of square differences of allele sizes (R ST ) as Euclidean distances with the program ARLEQUIN 2.0 (Schneider et al., 1999).

Genetic diversity
Of the seven microsatellite loci, the repeating motifs of four were dinucleotides, one was a trinu- cleotide and two were tetranucleotides.They were all polymorphic in Sebastes macdonaldi with the exception of locus Sra5-9 (Table 2).The number of alleles in polymorphic loci ranged from 2 to 24 (average 13.5).The degree of polymorphism (numbers and frequencies of alleles) was variable (Fig. 2, Table 2).Locus Sra11-103 was bi-allelic and hence the least variable.The rest of allelic distributions were multimodal and bimodal for three of the loci (Fig. 2).Coefficients of variation (CV) in allele size reflect the genic diversity of microsatellite loci.For example, in the case of locus Sra7-2 CV it is the highest (> 6%) due to the wide range of allele sizes, whereas for locus Sra11-103 CV it is an order of magnitude smaller (< 0.6%) (Table 2).Genotypic frequencies of all loci were found to conform to H-W and linkage equilibria within all localities (p > 0.05).Levels of expected heterozy-gosity were variable among loci and among localities (Table 2) and very high (He > 0.9) for loci Sra7-2, Sra16-5 and Sra15-8, whereas the locus with the smallest He was Sra11-103 (He < 0.51) at all sampled localities.Rho was largest for loci Sra7-2, Sra7-25, and Sra15-8 (Rho > 16.8).Levels of molecular genetic diversity (He) were not significantly different among the three localities (Kruskall-Wallis H, p = 0.67) or between the Pacific and the Gulf of California (Mann-Whitney U, p = 0.11).

Population differentiation
The degree of allelic differentiation was tested among the three localities (BLA, BAJA and CAL) as well as between the Gulf of California and the Pacific (BLA and BAJA+CAL).Allelic and genotypic frequencies of all polymorphic loci were not signif-MOLECULAR GENETICS OF SEBASTES MACDONALDI 455 icantly heterogeneous among the three localities (p > 0.059).However, significant frequency shifts were found between the gulf and the Pacific in allele frequencies of loci Sra7-7, Sra7-25 and Sra15-8 (p < 0.05) and in genotypic frequencies of loci Sra7-7, Sra15-8 and all loci pooled together (p < 0.04) (Table 3).Pairwise comparisons showed significant allelic differentiation (individual p < 0.05) at loci Sra7-7, Sra7-25 and Sra15-8 between BLA and BAJA.The same was true for genotypic frequencies, except that for locus Sra7-25 the differentiation was marginally non-significant (individual p = 0.058).After sequential Bonferroni correction for multiple tests, none of the pairwise comparisons of allelic and genotypic frequencies was significant (Table 4).
Fixation indices based on allelic frequencies (F ST ) were very small for all loci and localities (< 3.94%) but significant for loci Sra7-7, Sra7-25 and over all loci among the three localities.Comparison between Gulf and Pacific fish resulted in significant fixation indices for loci Sra7-7 and over all loci, whereas loci Sra7-25 and Sra15-8 were marginally 456 A. ROCHA-OLIVARES et al. non-significant (Table 5).Pairwise F ST and R ST values were also small (F ST <0.7% R ST < 2.0%) and not significant after sequential Bonferroni correction (Table 6).Finally, AMOVA results corroborated that no significant molecular genetic variance (expressed as F ST or R ST ) was associated with subdivision among the three localities (p > 0.1), though significant (R ST , p = 0.03) and highly significant (F ST , p = 0.009) Φ-statistics were obtained between Gulf and Pacific populations (Table 7).

Genetic diversity
Even though the loci and primers used in this study were originally described and designed for a congener (Sebastes rastrelliger), they all have worked satisfactorily in S. macdonaldi as well as in other Northeast Pacific species (R.D. Vetter, C.A. Kimbrell and E.A. Lynn, unpublished data, Buonaccorsi et al., 2002).This remarkable level of crossspecies microsatellite conservation is a clear indication of the close molecular relatedness of species of Sebastes that has surfaced in other microsatellite studies (e.g.Miller et al., 2000;Roques et al., 1999;Wimberger et al., 1999) as well as other molecular investigations (Rocha-Olivares et al., 1999a;Rocha-Olivares et al., 2000;Rocha-Olivares et al., 1999b;Rocha-Olivares et al., 1999c).Most loci analysed were polymorphic, except for one.A total of 78 alleles were found in polymorphic loci (average per locus = 13.5),though all loci were not equally variable.Loci Sra7-2, Sra16-5, and Sra15-8 display greater allelic diversity as well as high levels of heterozygosity.All genotypic frequencies were in H-W equilibrium, suggesting that null alleles were unlikely to have been encountered, biasing genotypic frequencies.Heterozygosity levels were large (He > 0.50) for loci  Molecular genetic diversity in fish has been found to be associated with life history traits reflecting habitat types.Marine species generally possess significantly higher levels of genetic diversity (average He = 0.79) than freshwater (average He = 0.46) or anadromous (average He = 0.68) species (DeWoody and Avise, 2000).Therefore, the high heterozygosities found in S. macdonaldi (0.74 ≤ He ≤ 0.78) are not atypical.A very close value (mean He = 0.72 in a set of shared loci) was found among 400 grass rockfish S. rastrelliger sampled along the Pacific coasts of Oregon and California (Buonaccorsi et al., unpublished data).Buonaccorsi et al. (2002) studied exactly the same loci analysed here (except for Sra7-2) in the copper rockfish Sebastes caurinus finding 0.63 ≤ He ≤ 0.71 in fish sampled throughout its geographic range.Overlapping levels of microsatellite heterozygosity (0.5-0.87) were found at six loci of the Pacific ocean perch Sebastes alutus (Miller et al., 2000).In contrast, only three of five polymorphic loci described for the quillback rockfish S. maliger possessed comparable levels of He (0.66-0.79), whereas the others showed low polymorphism (He < 0.5) (Wimberger et al., 1999).Eight microsatellite loci described for the four North Atlantic redfish (S. fasciatus, S. mentella, S. marinus, and S. viviparus) were also highly polymorphic (0.5 ≤ He ≤ 0.96) and at least half the loci had He > 0.8 in all species (Roques et al., 1999).
Germane to the zoogeographical hypothesis is the fact that no significant difference in molecular diversity was found between Pacific (mean He = 0.75) and Gulf of California populations (mean He = 0.78).A high genetic diversity in the Gulf of California is not consistent with a recent founder event.At face value, very similar levels of heterozygosity in the gulf and Pacific populations suggest that: (1) Gulf of California populations are not reproductively isolated from those in the Pacific; (2) there was an ancient dispersal event and invasion of the gulf from the Pacific that may have resulted in a founder MOLECULAR GENETICS OF SEBASTES MACDONALDI 457 effect, but the signal has been eroded by the rapid mutation rate of microsatellite loci; (3) there was never a genetic founder effect and contemporary gulf populations were separated from the Pacific by vicariance; or finally (4) the invasion is too recent and/or the colonisers too numerous for random genetic drift to cause a detectable drop in genetic diversity.Admittedly, some of these hypotheses have specific predictions regarding the pattern and extent of genetic differentiation to be found among geographic regions.

Population structure
The patterns of allelic and genotypic differentiation suggest that significant differentiation is associated with comparisons between fish from the Pacific and the Gulf of California.On the other hand, the absence of differentiation between Pacific samples (i.e., CAL and BAJA) may reflect high levels of gene flow along the Pacific coast of California and Baja California, but the small sample size of CAL (n = 14) limits the statistical power.A larger sample size is required from California to corroborate this observation.Three of the seven loci (Sra7-7, Sra7-25 and Sra15-8) were mainly responsible for the observed significant differentiation (Tables 3 and 4).These significant allelic and genotypic frequency shifts were sufficient to translate into a weak but significant partitioning of the genetic variance among localities (overall F ST = 0.0066, p = 0.03) and an even stronger differentiation between gulf and Pacific organisms (overall F ST = 0.0113, p = 0.004).In a study that included the same loci used here, except for Sra11-103, Buonaccorsi et al. (unpublished data) also found a low and marginally significant partitioning (F ST = 0.0013, p = 0.0497) in the grass rockfish S. rastrelliger distributed continuously along the coasts of California and Oregon.The same loci, however, were all significantly differentiated (overall F ST = 0.036, p < 0.001) among populations of copper rockfish (Buonaccorsi et al., 2002).
A small, albeit statistically significant, level of genetic differentiation is consistent with the de facto isolation of the Gulf of California populations from the Pacific.Ecological studies and direct observations suggest that bottom dwelling Sebastes, such as Sebastes macdonaldi, are unlikely to engage in long distance migrations but rather are relatively sedentary around rocky outcrops and other hard structures they inhabit (e.g., Love et al., 1999;Love et al., 2000), and even some midwater Sebastes species appear to have small radii of movement in short time scales (Pearcy, 1992).Thus, it seems unlikely that S. macdonaldi in particular would travel thousands of kilometres around the tip of the Baja California peninsula into the Gulf of California.This unlikely journey would have to entail some kind of southern submergence, as surface and subsurface waters are presumably too warm for the survival of the species larvae and juveniles, in order to reach regions marked by strong upwelling in the Gulf of California, where established populations are viable.Thus, hypothesis number one above lacks any positive evidence, is unparsimonious, and is not supported by the data.Hypothesis number two, involving an ancient founder event, is also inconsistent with the data.Presumably-isolated populations of Puget Sound copper rockfish S. caurinus were found to be highly divergent from nearby samples taken from the Gulf Islands (F ST = 0.037) and from the Pacific coast (F ST = 0.087) at the same loci studied here.The age of Puget Sound populations cannot be more than 11,500 years, a time at which the sound became marine after being filled with fresh water as the glaciers melted in the area (Buonaccorsi et al., 2002).The degree of genetic partitioning between Gulf of California and Pacific populations of the Mexican rockfish is only one third to one seventh of that amount.The large genetic differentiation observed in the copper rockfish may be partly explained by natural selection (Buonaccorsi et al., 2002), as Pacific and Puget Sound fish may face different selective pressures.However, it is unlikely that all loci would have been equally affected by genetic hitchhiking to a selected locus, and one can also argue that populations of the Gulf of California are subject to a very different physical environment than those in the Pacific (Alvarez-Borrego, 2001;Chen, 1975).Thus, an ancient invasion of the gulf is not consistent with the low level of genetic differentiation.Distinguishing between hypotheses three and four above may prove to be difficult with the data at hand.Both hypotheses may be consistent with the main findings of this study, namely comparable levels of genetic diversity and little differentiation between gulf and Pacific populations.The small level of genetic differentiation would favour a recent dispersal since vicariance would entail a separation that is at least a million years old, according to the hypothesised middle Pleistocene midpeninsular seaway connecting the gulf and the Pacific ocean (Riddle et al., 2000).Analysis of another genetic marker amenable to phylogenetic analyses, such as mito-chondrial or nuclear DNA sequences, may help to elucidate the historical scenario.In conclusion, we have found small but significant levels of genetic differentiation between allopatric and isolated populations of the Mexican rockfish in the Gulf of California from those in the Pacific coast of North America.The evidence argues against contemporary gene flow between the two regions and an ancient invasion of the gulf with a founder effect.A larger data set including DNA sequences amenable to phylogenetic analyses will help to test alternative hypotheses of dispersal versus vicariance.
FIG. 2. -Allele size frequency distributions of the seven polymorphic microsatellite loci of Sebastes macdonaldi used in this study.

TABLE 3 .
-Significance levels (S.E.) in test of allelic and genotypic differentiation for microsatellite loci of S. macdonaldi.

TABLE 4 .
-Significance levels in test of pairwise allelic and genotypic differentiation for microsatellite loci of S. macdonaldi.
ST ) of microsatellite loci of S. macdonaldi.Significance (in parenthesis) is shown only for values < 0.055.

TABLE 7 .
-Analysis of molecular variance (AMOVA) of microsatellite loci of S. macdonaldi using the number of distinct alleles (F ST ) and the sum of square differences of allele sizes (R ST ) as Euclidean distances.