Ethics Statement

This work was conducted using puye (Galaxias maculatus) as model study, a common freshwater and brackish fish species along the Chilean coast. The species is not protected and is considered as vulnerable in Central Chile and as out of danger in the Magellanic Province. Permission to undertake field works and collect specimens was issued by the Chilean Fisheries Service Director (Pablo Galilea Carrillo), under the technical memorandum (427/2011). The Instituto de Ecología y Biodiversidad (IEB/12-2011) and Chilean Fisheries Service (SERNAPESCA 427/2011) ethic committees approved sampling protocols and experiments. For this, we compiled with local legislation and the Convention on Biological Diversity.

Sample collection, DNA extraction, PCR amplification and sequencing

Individuals were collected between 2007–2014 in 13 estuarine localities along two main biogeographic areas in the Chilean Coast. Firstly, we included six localities from the Intermediate Area, between 38°S and 42°S, outside the direct influence of the LGM Patagonian Ice Sheet: Intermediate Area localities were: 1) Moncul River (38°42’S; 73°24’W); 2) Queule River (39°17’°S; 73°10’W); 3) Lingue River (39°26’S; 73°09’W); 4) Valdivia River (39°53’S; 73°24’W); 5) Chaihuín River (40°01’S; 73°27’W), and 6) Maullín River (41°35’S; 73°38’W). Secondly, we included seven localities from the Magellanic Province, between 43° S and 53° S, within the influence of the LGM Patagonian Ice Sheet. Magellanic localities were: 1) Yelcho River (43°03’S; 72°33’W), 2) Concoto Island (44°06’S; 73°43’W); 3) Williams Channel, Rivero Island (45°25’S; 74°21’W); 4) María Eugenia Bay (45°55’S; 73°31’W); 5) Tortel, Calén Fjord (47°45’S; 73°32’W); 6) Pascua River, Calén Fjord (48°15’S; 73°18’W); and 7) Strait of Magellan (53°14’S; 70°59’W; Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Sampling localities of Galaxias maculatus along the Chilean coast.

https://doi.org/10.1371/journal.pone.0131289.g001

Whole specimens were fixed in ethanol (95%) and DNA was prepared from muscle tissue using a salting-out method [40]. A partial fragment of the mitochondrial Control Region (D-loop) was amplified using specific primers GAL-F 5’–TAA CTC TCA TTA ACT AAA G– 3’ and GAL-R 5’–TGA TAG TAA AGT CAG CAA GCC– 3’) designed from the complete mitochondrial genome of the species (ACN: AP004104) [41]. PCR amplifications were performed in a 25 μl reaction containing 2.5 μl 10X Buffer (50 mM KCl, 10 mM Tris-HCl, pH 8.0), 1.0 μl of 50 mM MgCl₂, 200 mM dNTPs, 0.5 μl of each primer (10 pg/μl), 1 U Taq (Invitrogen), 17.5 μl of double-distilled water and 5 ng of DNA. Thermal cycling parameters included an initial denaturation step at 94°C for 5 min, followed by 35 cycles at 94°C for 90 sec, 60.7°C for 90 sec, and 72°C for 90 sec, and a final 10 min extension at 72°C. PCR amplification products were purified using QIAquick Gel Extraction Kit (QIAGEN) and sequenced in both directions with an Automatic Sequencer ABI3730 x 1 at Macrogen Inc. (Seoul, Korea).

Genetic diversity and population structure in Galaxias maculatus

D-Loop sequences were edited using Proseq v. 3.5 [42] and aligned with ClustalW [43]. New D-loop haplotypes recorded in the species were deposited in GenBank under the Accession Numbers KP298433—KP298673 and KP336557—KP336665. We performed a DNA saturation analysis following Xia et al. [44] to evaluate how saturation of transitions is accumulated in relation to nucleotide divergence in the whole data set. Levels of genetic polymorphism were estimated using standard diversity indices including: the number of haplotypes (k), the number of segregating sites (S), haplotypic diversity (H), the average number of pairwise differences (П), and nucleotide diversity (π) for each locality, for each main area, and for the whole D-loop data set using DnaSP v.5.00.07 [45]. We performed neutrality statistical tests (Tajima’s D and Fu’s F_S) for each locality, area, and for the whole data set to estimate whether D-loop sequences in G. maculatus deviate from expectations under a neutral model. Considering the high levels of genetic diversity previously estimated in the species [22,39,41] we determined the levels of genetic differentiation between the analyzed localities using mean pairwise differences (N_ST) following Pons and Petit [46] in Arlequin v. 3.5 [47]. The statistical significance of genetic differences between localities was estimated using permutation tests using 20,000 iterations of haplotype identities. Similarly, we estimated the levels of genetic differentiation of subpopulations using the nearest-neighbor statistic (S_nn) that measure how often nearest neighbors (in sequence space) of sequences are from the same locality in geographic space [48]. The statistical significance of S_nn analyses was determined through a permutation test using 10,000 iterations. We inferred the spatial genetic structure in the species by estimating the number and composition of groups that were most differentiated based on sequence data with an Analysis of Molecular Variance (AMOVA) following Excoffier et al. [47] in Arlequin. AMOVA is a popular method that uses multiple spatial scales in statistical methods to characterize spatial genetic structure by partitioning it into: within populations, among populations within groups and among groups. Finally, we performed a test for isolation by distance using a Mantel test with 1,000 permutations in Arlequin to determine the correlation between Slatkin’s linearized localities genetic differentiation [49] and the linear geographic distance (km) between populations measured using FOSSIL [50].

Demographic inference in G. maculatus

We constructed genealogical relationships in G. maculatus using Maximum Parsimony Networks in Hapview (http://www.cibiv.at). To estimate the pattern of demographic history in the species, we compared the distribution of pairwise differences between haplotypes (mismatch distribution) for each locality, main area, and for the whole data set, to the expected distribution under the sudden expansion growth model of Rogers and Harpending [51]. This analysis is a popular method since the amount of nucleotide differences between haplotypes depends on the length of time since they diverged. Alternatively, we reconstructed past population dynamics through time using a Bayesian Skyline Plot method implemented in BEAST v. 1.7 [52]. This method is fundamentally based on coalescent theory, which quantifies the relationship between a genealogy of the sequences and the demographic history of a population. In contrast to the previous method, it does not assume any demographic model. For comparison purposes, three substitution models (strict clock, uncorrelated lognormal and uncorrelated relaxed clock) were computed for the main areas here analyzed and compared statistically using a Bayes factor test [53] with TRACER v. 1.5 (http://beast.bio.ed.ac.uk/Tracer). The uncorrelaed lognormal model was the best fit for D-loop data in G. maculatus. We conducted three independent Bayesian MCMC runs using the GTR + I + G model, previously estimated using MrModeltest v. 2.3 (http://www.abc-se/~nylander), and a specific population level mutational rate estimated for G. maculatus by Salinas [54]. Two independent runs per area (Intermediate Area and Magellanic Province) were made for 50 x 10⁶ generations (sampled every 1000 step), discarding a 10% of the trees as burn-in. The convergence of runs was confirmed with Tracer ensuring a minimum of 1000 effective sampling for each statistics (ESS). The results of the multiple runs were combined using LogCombiner [52]. The median and corresponding credibility intervals of the Bayesian skylines plots were depicted with Tracer.

Gene flow and connectivity

We compared different models of gene flow between the main areas of interest (Intermediate Area and Magellanic) with the software MIGRATE v.3.5 [55], to test for different scenarios. For this, we defined four candidate models constraining the presence of directionality of gene flow between the main areas. The first model allowed bidirectional gene flow between the Intermediate Area and the Magellanic Province (full island model). Models 2 and 3 were defined by asymmetric patterns of gene flow from the Intermediate Area to the Magellanic Province, and from the Magellanic Province to the Intermediate Area, respectively. Finally, model 4 assumed both main areas as the same panmictic population. All the analyses were run using a GTR + I + G substitution model and transition-transversion ratio of 2.9818 as previously estimated by jModeltest v2 [56]. The specific substitution rate for the selected marker in the species was set to constant, as suggested by the manual. Analysis consisted of one long chain with 500.000 recorded parameter steps, a sampling interval of 100 and a burn-in of 100.000 (10%), running multiple replicates (10 independent chains). A heated scheme (1.00, 1.50, 3.00, and 1000000.0) was used to calculate the marginal likelihoods for model comparisons. Comparisons among the different models were done using a Bayes Factors calculated by substracting the highest value of the log marginal likelihoods (lmL; Bezier curve approximation) from the lmL values of each models. Finally, the associated probability of each model in relation to others was estimated following Kaas and Rafery [57].

We estimated different demographic indices and evaluated two asymmetric isolation-with-migration models between main areas of interest (Intermediate Area and the Magellanic Province) using IMa2 software [58,59]: i) from the Intermediate Area to Patagonia (m _{N > S}) and from Patagonia to the Intermediate Area (m _{S > N}). We carried out separately several preliminary runs in the M mode (Markov Chain Monte Carlo; MCMC mode) of the software to set the best set of priors that ensure mixing and convergence. Uniform priors were used to estimate effective population size (Θ₁, Θ₂, and ancestral Θ_a, Θ = 3000) and splitting time (t = 300), whereas an exponential prior (mean = 1) for gene flow (m) was adopted. We performed 80 x 10⁶ MCMC steps sampling every 100 generations, with a burn-in of 8 x 10⁶. D-loop sequences were assumed to mutate under the HKY mutation model following the author guidelines [59]. Once convergence was achieved under the M-mode, we used the same simulated genealogies under the L-Mode (Load Tree mode) to estimate the log maximum-likelihood and credibility intervals (95% under HPD) estimates for migration parameters using a Likelihood Ratio Test. Under the L-Mode we compared our data with a null model of no migration [58,59]. Finally, estimated parameters were re-scaled into years (t/μ) and effective rate of migration ((Θ_x/ m_x)/2) (rate at which genes come into population, per generation) using an overall substitution rate per year (m = 1.1 x 10⁻⁴) estimated on previous studies [54].

Main patterns of genetic diversity and structure in G. maculatus

Even if coastal and Andean South American samples of G. maculatus have been previously included in phylogeographic and demographic studies [22,39], here we present the first population-level study in the species along two biogeographical provinces that were differentially affected during the glacial cycles of the Quaternary. In contrast to previous studies that incorporated mainly river and lake specimens, our study focuses on estuarine and coastal populations. As previously determined in the species [22,39,41], G. maculatus harbors extremely high levels of genetic diversity along its distribution in the Chilean Pacific coast. In fact, our results further corroborate previous molecular studies in the species [22,41,72], pointing towards an extreme within-species genetic diversity among coastal and estuarine populations. In this study, out of 353 individuals we recognized 321 different haplotypes. Similarly, Zemlak [22] recorded a total of 273 different haplotypes, out of 299 individuals, while Waters et al. [72] recognized 139 different haplotypes out of 144 individuals from New Zealand. In spite of the high levels of genetic polymorphism recorded in this study, we also detected marked and significant differences in the average number of nucleotide differences and nucleotide diversity between main biogeographical areas here included. Therefore, higher levels of genetic diversity along formerly unglaciated areas (Intermediate Area) than in those located within the limits of the Patagonian Ice Sheet (Magellanic Province) detected in the species, supports the first prediction of the Expansion-Contraction model. Main differences recorded in terms of genetic diversity between the main biogeographical areas here included were also supported by population dynamics reconstructions showing a more recent population expansion in the Magellanic Province than in the Intermediate Area. Therefore, our results support the hypothesis that the Intermediate Area represents an older region and less affected by glacial advances and retreats than the Magellanic Province. Together with main patterns of genetic diversity recorded in the species we also found contrasting patterns of genetic structure between the main biogeographic areas here analyzed. Each one of the analyzed rivers along the Intermediate Area, expanding less than 250 km, represented a different genetic unit that generally coincided with the defined haplogroups of Zemlak et al. [22] at the same localities. Magellanic populations of G. maculatus, expanding for more than 1000 km, showed low levels and even absence of genetic differentiation. Again, major differences in terms of the genetic structure between formerly glaciated and unglaciated areas are adjusted to the predictions of the Expansion-Contraction model. Nevertheless, high levels of genetic diversity, measured as П and π, among Magellanic localities and the absence of dominant broadly distributed haplotypes in this area do not fit with the expectations of recent postglacial expansion from refugial areas at lower latitudes. Moreover, the presence of a markedly differentiated and highly diverse haplogroup (Hap04) in this Province does not support the idea of a recent postglacial recolonization. Such results contrast to those observed in other South American freshwater [24,25,73–76] and marine fishes [67,68]. In fact, Alò et al. [75] recorded low levels of genetic diversity and a clear genetic continuity between Intermediate Area and Magellanic province populations of Aplochiton (A. zebra, A. taenius and A. marinus). Most of these studies have demonstrated the important role of the last glacial cycle over the demography of Magellanic fishes with the presence of typical star-like haplotype genealogies as a consequence of recent colonization processes associated to strong founder effects. In this context, population genetic patterns in Patagonian terrestrial plants and vertebrates indicate that different processes and directional range shifts have generated a mosaic of phylogeographical patterns that are far more complex than a simple north to south described one [71]. In the particular case of G. maculatus, Zemlak et al. [22,39] proposed that the main patterns of genetic diversity and structure in the species suggest a low amount of influence of the last glacial cycle. In fact, levels of genetic diversity along the Magellanic Province likely reflect the existence of marine in situ or periglacial refugia during cooling periods of the Quaternary.

Patterns of genetic structure in G. maculatus along the analyzed areas

Low genetic differentiation along the Magellanic Province, a complex geographic and oceanographic area, may be a response of particular characteristics of the species. The southern fjords and channels ecosystems along the Pacific coast of Patagonia are considered as one of the largest estuarine ecosystem in the planet [77,78]. Subantarctic waters masses are mixed with freshwater ones from abundant precipitation, river flow, and glacial meltwater [79]. A low salinity surface layer generates a highly stable water column within the fjords whereas a well-mixed water column occurs in the gulfs and open channels. In this context, epipelagic G. maculatus larvae have been reported in ichthyological samples along the Magellanic Province [78]. Considering these findings and the pattern of genetic structure recorded in the species it seems like the Magellanic Province constitutes a continuous habitat for G. maculatus. In contrast, genetic structure along the Intermediate Area suggests low levels of connectivity between river basins. In this context, open sea areas between the coastal rivers here analyzed seem to be preventing the genetic continuity of the species along the Intermediate Area. Such results are quite surprising considering the high dispersal potential described for New Zealand coastal populations of G. maculatus [80]. In this context, mechanisms associated to larval retention or physiological limitation associated to salt tolerance may explain the complex pattern of genetic structure along the Intermediate Area. Otherwise, philopatric behavior could also generate small-scale genetic structure within the Intermediate Area localities, despite the high dispersal potential of the species [34].

Major differences in terms of genetic diversity and structure recorded in G. maculatus agree with the 42°S biogeographic break described along the Chilean coast resulting from different oceanographic, geologic, and climatic processes [38,77]. Along the Chilean coast, this break represents a boundary for major environmental changes in terms of topography, currents, and water salinity that represent a major transition for marine [38] and freshwater species [81]. Different biogeographic scenarios altered the species assemblages across this limit and likely generated distinct patterns of genetic diversity [82]. First, this break coincides with the separation of the West Wind Drift into two main oceanic currents along the Pacific coast [39]. Intermediate Area localities are washed by the northward flow of the Humboldt Current System [83] while the Magellanic Province is located within the influence of the Cape Horn Current, the southern branch of the West Wind Drift. In this context, the presence along Intermediate Area localities of haplotypes belonging to the Magellanic diversity (Hap 04) points towards asymmetric northward gene flow that could be associated to a latitudinal displacement of the West Wind Drift during glacial periods [84,85]. A concordance between phylogeographic and biogeographic patterns has been recorded in many areas of the globe indicating that the forces determining species distributions are also related to the spatial patterns of population genetic structure [86–90]. Therefore, local adaptation processes and distinctive characteristics of main biogeographic areas along the Chilean coast in terms of Quaternary glacial histories and oceanographic regimes are responsible for the general pattern of genetic structure recorded in G. maculatus.