Comparative population structure of three snook species (Teleostei: Centropomidae) from the eastern central Pacific Píndaro Díaz-Jaimes1*, Edson Sandoval-Castellanos, and Manuel Uribe-Alcocer Instituto de Ciencias del Mar y Limnología, Universidad Nacional Autónoma de México, Apartado Postal 70-305, México, D.F. 04510, México (e-mail:
[email protected]) Received: September 9, 2005 / Revised: April 20, 2007 / Accepted: April 30, 2007
Ichthyological Research ©The Ichthyological Society of Japan 2007
Ichthyol Res (2007) 54: 380–387 DOI 10.1007/s10228-007-0413-3
Abstract Three snook species, Centropomus viridis, Centropomus medius, and Centropomus robalito, from the eastern central Pacific, representing three of the four proposed phyletic lineages in the genus, were analyzed for genetic variability by means of allozyme and RAPD to evaluate the divergence between populations at different levels of dispersal ability and to evaluate the importance of barriers to dispersal in the population subdivision and genetic diversity. Levels of genetic diversity among species estimated by allozymes were similar and consistent with the observed levels of differentiation in marine fish species. Mean heterozygosity ranged from 0.089 for C. viridis to 0.10 for C. robalito. Genetic diversity for the snook species studied was slightly higher than the mean estimation reported in allozymes for 106 marine fish (0.055) and for anadromous fish species (0.043 to 0.057). Multilocus allele frequency homogeneity tests and population-subdivision estimates for both allozyme and RAPD markers revealed the existence of population structure in C. viridis and C. medius, in coincidence with geographic separation of samples, whereas no divergence was detected in C. robalito. This finding may be attributed to the greater population size of C. robalito, which originated by a recent population range expansion, and hence the potential for dispersal is mediated by larval drift. Fluctuations in population size and population range expansion are used to explain discrepancies between levels of genetic diversity and population structure in the studied species. Key words
Population structure · Genetic divergence · Genetic markers · Population size
Supplementary material The online version of this article contains supplementary material, which is available on Springer’s server at springerlink.com.
S
nook species of the genus Centropomus are important high-level carnivores including euryhaline, semicatadromous, and/or estuarine fish (Chávez, 1963; Rivas, 1986) found in tropical and subtropical waters of North and Central America in both Atlantic and Pacific oceans. Even though the habitat preferences of snook include fresh or low-salinity water, spawning takes place at sea (Chávez, 1963; Fuentes-Castellanos, 1973; Muhlia-Melo et al., 1995), and species show a gradual dependence on the existence of estuarine systems, lagoons, or freshwater outflow to feed and grow juveniles and to complete their life cycles (MuhliaMelo et al., 1995). Likewise, the species show marked differences in abundance among ocean basins and within species because of their vulnerability to temperature and their strong dependence on the presence of freshwater outflow (Muhlia-Melo et al., 1995). Because adults display a solitary behavior, especially the larger species, their catch is difficult and requires considerable fishing effort. Diadromous species tend to display low differentiation patterns similar to marine species (Avise, 2000) and exhibit lower genetic diversities than nonanadromous species (Ward et al., 1994). The absence of obvious barriers at the sea where anadromous species spend a part of their lives represents opportunities for lineage dispersal and pro-
motes gene flow among adjacent populations, inhibiting divergence among geographically isolated populations (McDowall, 2001) and, hence, the disruption of genetic structure. Additionally, some features of these fish species, such as dispersal ability, high reproductive potential, and rapid growth rates, are thought to influence the levels of the population structure (Graves, 1998). However, some degree of population structure should be expected for anadromous species (Ward et al., 1994), with differences related to species dependence on the presence of estuaries, to the existence of barriers to migration caused by lack of freshwater flows, or to the occurrence of oceanographic currents limiting gene flow between populations. Along the Mexican Pacific coast, intermediate shorelines between Sinaloa and Oaxaca are characterized by a short continental platform, and hence by the lack of estuarine systems and by the convergence of two main currents generating a countercurrent flowing along the equatorial margin. This subsurface topography might limit the distribution of snook species in such a way that two main abundance areas occur to the north and to the south in Sinaloa and Oaxaca, respectively, where lagoons are abundant and suitable for reproduction and may represent an additional factor promoting population divergence. For this reason, we
Population structure of three snook species
can hypothesize that differences in the levels of population structure should be expected in snook species and may be related to their dependence on estuaries and/or their migratory ability. Speciation in Pacific snook species occurred after the closure of the isthmus in Central America, about 3.5 million years ago (Donaldson and Wilson, 1999), and several vicariance processes have been hypothesized (Tringali et al., 1999). Four different snook phyletic groups, which have been reported by Rivas (1986) and confirmed by Tringali et al. (1999), show differences in migration ability, growth rates, longevity, feeding habits, and habitat preferences (Fuentes-Castellanos, 1973; Taylor et al., 1998). Representative species from the three main phyletic groups distributed along the Mexican Pacific coast were selected to test if dispersal ability and dependence on estuaries are related to any divergence pattern. Centropomus viridis, representative of the C. undecimalis group, is characterized by long-bodied specimens with sizes ranging from 60 to 120 cm and an apparent high migratory ability and marine habitat preference (Tringali and Bert, 1996). Centropomus medius belongs to the C. pectinatus group, ranges in size from 25 to 60 cm, and has moderate dependence on permanent freshwater outflow (Rivas, 1986), with no reports of occurrence in the nearshore or oceanic environment (Tringali et al., 1999). A third species, Centropomus robalito, included in the C. ensiferus group, was selected because of its higher dependence on freshwater, small size ranging from 25 to 35 cm, and its limited migratory ability resulting in a marked preference to live in lagoons and estuaries (Muhlia-Melo et al., 1995). Evaluation of population isolation within these phylogenetically related species with different dispersal abilities might contribute to the understanding of speciation processes in this group after the formation of the Tehuantepec isthmus. Furthermore, along the Pacific coast off Mexico, snook species represent an important multispecies fishery, with reported catches averaging 500–900 t for the past 10 years, in contrast with the relatively low abundances of the species. Because of the high morphological and morphometric similarities of the species, statistics are based on a pool of data for all captured species; consequently, catch statistics are unclear in determining the actual contribution of each species to the fishery. Therefore, a stock definition based on genetic criteria represents a first approach to define management strategies directed to preserve the genetic variability of snook species.
Materials and Methods Sampling.—A total of 537 samples including 87 Centropomus viridis, 77 Centropomus medius, and 93 Centropomus robalito snook specimens were collected over 5 years from the most separated Mexican locations abreviated as Sonora (Son), Sinaloa (Sin), Nayarit (Nay), and GuerreroOaxaca (Goax), coinciding with significant abundance areas and in relationship to the existence of estuaries, along the Mexican Pacific coast. The sampling sites, years, and sample
381
sizes are C. viridis (Son, 1995–2000, n = 11; Sin, 1998–1999, n = 23; Nay, 1999–2000, n = 23; Goax, 1998–2000, n = 30), C. medius (Sin, 1995–1999, n = 27; Nay, 1999–2000, n = 23; Goax, 1999–2000, n = 27), and C. robalito (Sin, 1998–1999, n = 29; Nay, 1999, n = 30; Goax, 1998–2000, n = 44). The difficulty of obtaining significant sample sizes was caused mainly because by differences in abundance between areas among species, especially for C. viridis and C. medius. Identification of specimens in the field was based on the diagnostic characters proposed by FAO (1978). A small piece of muscle of every specimen was kept in liquid nitrogen and stored at −70°C until processed. Allozymes.—About 0.5–1.0 g of fresh muscle tissue was manually ground in 1 ml homogenizing solution [0.01 M Tris, 0.001 M ethylenediaminetetraacetic acid (EDTA), and 40 mg NADP, pH 6.8], centrifuged at 4000 g at 4°C, and electrophoresis was done on horizontal 12% starch gels (Sigma; S-5651) overnight. Two buffer systems, Tris-glycine (Shaklee and Kennan, 1986) (0.025 M Tris, 0.192 M glycine, pH 8.5; a) and Tris-citrate II (Selander et al., 1971) (0.687 M Tris, 0.157 M citric acid, pH 8.0; b), were used to analyze 11 enzymes that resolved 15 loci, 7 of which showed polymorphism. Enzymes assayed with buffer a were as follows: glucose-phosphate-isomerase (GPI), two loci; l-leucil-lalanine (LA), two loci; l-leucil-glycil-glycine (LGG), two loci; those with buffer b were lleucil-l-proline (PAP), one locus; phosphoglucomutase (PGM), one locus; and phosphogluconate dehydrogenase (6-PGD), one locus. The enzymes aspartate aminotransferase (AAT), glyceraldehyde-3-phosphate dehydrogenase (GAPDH), isocitrate dehydrogenase (IDH), lactate dehydrogenase (LDH), and malate dehydrogenase (MDH) displayed no apparent polymorphism. Enzyme assays used the procedures of Harris and Hopkinson (1976). Random amplified polymorphic DNAs (RAPDs).— Genomic DNA from a minimum of 24 tissue samples of each snook species was isolated by the phenol–chloroform protocol (Sambrook et al., 1989), resuspended in 50–100 µl TE buffer (10 mM Tris, 100 mM EDTA, pH 8.0), and quantified with a Hoefer DyNA quant 200 fluorometer. Genomic DNA was amplified using the randomly amplified polymorphic DNA methodology (RAPD; Williams et al., 1990) by using 40 random primers from Operon (Qiagen, Valencia, CA, USA) from the series OPF and OPA, from which only the primers OPF-9 (5′-CCAAGCTTCC-3′) for C. viridis, OPF-10 (5′-GGAAGCTTGG-3′) for the three analyzed species, and OPF-13 (5′-GGCTGCAGAA-3′) for C. medius and C. robalito were assayed for polymorphisms. Amplifying reactions were made in a GeneCycler thermal cycler (Bio-Rad, Hercules, CA, USA) and consisted of 25 µl final volume reaction-mixture containing 1.5 ng/µl DNA in amplification buffer, 100 mM Tris-HCl, pH 8.4, 500 mM KCl (2.5 µl), 15 mM MgCl2 (1.6 µl), 9.5 pmol primer, 10 mM dNTPs (2.0 µl), and 1 U Taq polymerase (Promega Biosciences, San Luis Obispo, CA, USA). Settings of the program for amplifications consisted of a denaturalization cycle of 3 min at 94°C followed by 44 cycles of 1 min at 94°C, 1 min at 37°C, 1 min 20 s at 72°C, and a final extension cycle of 15 min at 72°C.
382
Amplified fragments were separated by electrophoresis on 1.4% agarose gels (Invitrogen, Grand Island, NY, USA) for 1–2 h at 90 mA (100 V). A 100-bp DNA ladder (Invitrogen 15628-019) was used as a size standard. Agarose gels were stained with ethidium bromide, visualized in a UVlight transilluminator, and photographed for subsequent analyses. Data analyses.—The Genepop version 3.3 software (Raymond and Rousset, 1995) was used to obtain allele frequency estimations, to test the conformation of each locus to Hardy–Weinberg genotypic proportions, and to obtain genetic diversity estimates in terms of mean population heterozygosity. The program Genepop was also used to estimate heterozigosity deficit and excess and to make exact tests to evaluate spatial homogeneity of allele frequencies. Before proceeding with population-differentiation analyses, the temporal homogeneity of allele frequencies in samples was tested, comparing temporal collections (where available) by using the exact probability tests (Guo and Thompson, 1992). If no heterogeneity was found, those collections were joined into a single sample and homogeneity tests made between spatially separated locations. Estimation of allele frequencies for RAPDs was based on the procedures described in Lynch and Milligan (1994) and by using the software TFPGA (Tool for Population Genetics Analyses; Miller, 1998). Estimations of allele frequency were based on the Taylor expansion (Kendall and Stuart, 1977, cited in Le Corre et al., 1997) as implemented in TFPGA to decrease the bias resulting by calculating allele frequencies from null-allele genotypic proportions (see details in Lynch and Milligan, 1994). RAPD fragments were interpreted using the following assumptions: (1) fragments were considered to behave as dominant genes (Williams et al., 1990); (2) every polymorphic fragment was considered derived from a two-allele locus; (3) the Hardy– Weinberg equilibrium was assumed for all genotypes; and (4) each fragment was considered as an independent locus. The Fstat 1.2 program (Goudet, 1995) was used to obtain estimates of population differentiation based on Wright’s F-statistics (FST) using the procedures from Weir and Cockerham (1984) and its respective significance probabilities and 95% confidence intervals. Pairwise sample comparisons of F-statistics and multilocus population differentiation by exact tests were made to look for a spatial differentiation pattern between geographic locations. Significance levels were adjusted for multiple testing by using the Bonferroni sequential method (Rice, 1989). We also used a Bayesian approach to find the optimal partition of the populations for the three species studied using the software BAPS (Bayesian Analysis of Population Structure, version 2.0; Corander et al., 2003). This software allows the determination of the groups into which the populations most naturally fall and the later calculation of populationspecific parameters such as F-statistics. Because the number of populations to be analyzed for each species was less than nine, we used the enumerative-calculation option to estimate the population structure, as recommended by the authors.
P. Díaz-Jaimes et al.
Results Allozymes. Interspecies variability and Hardy–Weinberg expectations.—Of the 15 analyzed loci, 6 (40%) displayed polymorphism for C. viridis under the 0.95 criterion for the most common allele, whereas for C. medius and C. robalito polymorphism was estimated in 47%. Observed mean heterozygosity estimates were 0.089 for C. viridis, 0.088 for C. medius, and 0.10 for C. robalito (see electronic appendix, Table S1). Genetic diversity within populations among species was similar for C. medius and C. robalito populations, but marked differences in genetic diversity estimations were observed between C. viridis populations (0.101, 0.1164, and 0.089 for Sonora, Sinaloa, and Nayarit, respectively) when compared with the south samples, Goax (0.058). One locus (Pap) deviated significantly from Hardy– Weinberg genotypic proportions for C. robalito after correcting for multiple testing (P > 0.007; initial α = 0.05/7 = 0.007), whereas in C. viridis and C. medius no departures were observed at any locus after Bonferroni correction (see electronic appendix, Table S1). Intraspecies genetic divergence.—Allele frequency heterogeneity was observed at two of eight loci (P < 0.006; initial α = 0.05/8 loci = 0.006) for C. viridis, at four of seven loci analyzed for C. medius (P = 0.005; initial α = 0.05/7 = 0.007; data not shown), while no heterogeneity was observed for C. robalito from overall (P = 0.105) nor from individual loci for allele frequency homogeneity-exact tests. Pairwise sample comparisons (Table 1) revealed allele frequency heterogeneity between the Sinaloa and the Guerrero-Oaxaca samples (Sin-Goax; P = 0.0003; initial α = 0.05/6 comparisons = 0.008) for C. viridis, whereas for C. medius highly significant heterogeneity was observed for comparisons between Goax-Sin and Goax-Nay samples after Bonferroni correction (P < 0.016; initial α = 0.05/3 = 0.016). Comparisons for C. robalito showed no significant genetic differentiation at any loci from allele frequency homogeneity tests after correcting for multiple tests (initial α = 0.05/3 = 0.017; see Table 1). Significant genetic divergence as revealed by overall FSTestimates was detected in C. viridis (mean 0.031; P = 0.008; initial α = 0.05/8 = 0.006), and for C. medius (mean 0.078; P = 0.005; initial α = 0.05/7 = 0.007), whereas no population structure was detected for C. robalito (mean −0.002; P = 0.446; Table 2). Analysis of genetic structure with BAPS also supported the presence of genetic divergence between populations among species. In C. virids, the optimum partition (P = 0.955) generated a group consisting of the northern samples from Son-Sin-Nay and a second group consisting of the southern sample Goax. Later FST estimates based on the groups generated were slightly lower (0.027) than the FST-estimate (0.031) based on the preassigned structure. For C. medius, genetic structure generated a northern group with samples Sin-Nay separated from the south sample Goax (P = 1.000) and a lower FST estimate (0.04) than the estimation (FST = 0.078) based on the preassigned structure. In contrast, no partition into groups was observed using BAPS software for C. robalito, with each of the three ana-
Population structure of three snook species
383
Table 1. Significance probabilities values for overall multilocus and pairwise sample comparisons of allele frequency homogeneity tests (P) and population subdivision FST, estimates (PFST) for Centropomus species Species
Comparisons Overall
C. viridis P PFST C. medius P PFST C. robalito P PFST
Son-Sin
Son-Nay
A
R
A
R
A
