SEARCH

SEARCH BY CITATION

Keywords:

  • amplified fragment length polymorphisms;
  • divergent selection;
  • endogenous selection;
  • exogenous selection;
  • gene flow;
  • genome scan;
  • hybrid zone;
  • Littorina ;
  • sympatry

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

Hybrid zones of ecologically divergent populations are ideal systems to study the interaction between natural selection and gene flow during the initial stages of speciation. Here, we perform an amplified fragment length polymorphism (AFLP) genome scan in parallel hybrid zones between divergent ecotypes of the marine snail Littorina saxatilis, which is considered a model case for the study of ecological speciation. Ridged-Banded (RB) and Smooth-Unbanded (SU) ecotypes are adapted to different shore levels and microhabitats, although they present a sympatric distribution at the mid-shore where they meet and mate (partially assortatively). We used shell morphology, outlier and nonoutlier AFLP loci from RB, SU and hybrid specimens captured in sympatry to determine the level of phenotypic and genetic introgression. We found different levels of introgression at parallel hybrid zones and nonoutlier loci showed more gene flow with greater phenotypic introgression. These results were independent from the phylogeography of the studied populations, but not from the local ecological conditions. Genetic variation at outlier loci was highly correlated with phenotypic variation. In addition, we used the relationship between genetic and phenotypic variation to estimate the heritability of morphological traits and to identify potential Quantitative Trait Loci to be confirmed in future crosses. These results suggest that ecology (exogenous selection) plays an important role in this hybrid zone. Thus, ecologically based divergent natural selection is responsible, simultaneously, for both ecotype divergence and hybridization. On the other hand, genetic introgression occurs only at neutral loci (nonoutliers). In the future, genome-wide studies and controlled crosses would give more valuable information about this process of speciation in the face of gene flow.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

The evolution of reproductive isolation could be described as a continuous variation in divergence (Mallet, 2008; Hendry, 2009; Nosil et al., 2009b). Evolutionary processes promote or constrain such divergence along the speciation continuum (Coyne & Orr, 2004; Nosil, 2012). Understanding the relative importance of the opposite forces of diversifying selection and gene flow during speciation constitutes a long-standing evolutionary debate (reviewed in Räsänen & Hendry, 2008). Ecologically based divergent natural selection is known to drive population divergence (e.g. ecological speciation, Rundle & Nosil, 2005; Schluter, 2009; Nosil, 2012). On the other hand, ecological transitions (e.g. ecotones, Endler, 1977; Barton & Hewitt, 1985) could also lead to intraspecific hybridization. Thus, hybrid zones (sensu Harrison, 1990) represent a phase in speciation suitable for studying the interactions between selection and gene flow and consequently the genetics and ecology of reproductive isolation (Harrison, 1993; Jiggins et al., 1996; Rieseberg & Buerkle, 2002). Two main mechanisms could control, independently or simultaneously, introgression across hybrid zones (Barton & Hewitt, 1985; Moore & Price, 1993; Arnold, 1997). The mechanism of endogenous selection promotes reproductive isolation through hybrid unfitness as a consequence of the interaction of incompatible genetic backgrounds, forming a tension zone (sensu Barton & Hewitt, 1985). Exogenous selection might favour different genotypes depending on the environment (e.g. habitat type); this is found in ecotones. Determining the contribution of both mechanisms to reproductive isolation in diverging taxa is fundamental to understand the interaction between natural selection and gene flow during initial stages of speciation. Here, we study the pattern of phenotypic and genetic introgression in parallel hybrid zones of two ecologically divergent ecotypes of a marine snail, to shed light onto the causes and consequences of hybridization in this species.

The rough periwinkle, Littorina saxatilis, is a widespread intertidal snail across North Atlantic shores (Reid, 1996). This species has been considered an interesting model system for the study of ecological speciation in the face of gene flow (Rolán-Alvarez, 2007; Butlin et al., 2008; Schluter, 2009; Johannesson et al., 2010; Nosil, 2012). Different pairs of ecologically divergent ecotypes of L. saxatilis have been studied in detail in shores of Sweden (Panova et al., 2006), United Kingdom (Wilding et al., 2001) and Spain (Rolán-Alvarez et al., 2004; Fernández et al., 2005) (reviewed in Rolán-Alvarez, 2007; Johannesson et al., 2010). In exposed rocky shores of Galicia (NW Spain), two ecotypes [Ridged-Banded (RB) and Smooth-Unbanded (SU), Fig. 1] are found associated with different levels of the intertidal and dominated by different ecological pressures (Rolán-Alvarez, 2007). The SU ecotype inhabits the exposed lower shore on the mussel belt. With its small size, thin shell and wide aperture (accommodating a large foot), it is adapted to survive wave action. The RB ecotype, large, thick-shelled and small-apertured, inhabits the upper shore on the barnacle belt, which is less exposed, but it is characterized by higher predation rates (from Pachygrapsus marmoratus). The RB and SU ecotypes are characterized by a great divergence in morphology (Carvajal-Rodríguez et al., 2005; Conde-Padín et al., 2007, 2008), behaviour (Erlandsson et al., 1998) and protein expression (Martínez-Fernández et al., 2008), while maintaining a reduced neutral genetic differentiation (Rolán-Alvarez et al., 2004; Quesada et al., 2007; Galindo et al., 2009). In fact, for mtDNA, geographically isolated populations cluster preferentially by locality rather than ecotype (Quesada et al., 2007). Despite this reduced neutral divergence, an amplified fragment length polymorphism (AFLP) genome scan between pure RB and SU ecotypes (Galindo et al., 2009) detected loci presumably affected by selection (directly or indirectly). These so-called outlier loci (Nosil et al., 2009a) represent a 3% of the analysed AFLP loci, and they are highly divergent between the ecotypes. Therefore, these evidences and other studies suggest that ecologically based divergent natural selection is mostly responsible for the evolution of partial reproductive isolation between the ecotypes (reviewed in Rolán-Alvarez, 2007).

image

Figure 1. Experimental design. (a) Map of Galicia (NW Spain) showing the location of the sampling sites (Corrubedo, Silleiro and Cetarea). (b) Ridged-Banded (RB) and Smooth-Unbanded (SU) ecotypes of Littorina saxatilis. (c) Sampling design in the mid-shore (sympatric distribution of ecotypes) indicating the distance between localities and transects (transect 1, T1 and transect 2, T2). Sample sizes for amplified fragment length polymorphisms and geometric morphometrics are included within the ecotype symbol (circle, RB; square, SU; rhombus, hybrid ‘HY’).

Download figure to PowerPoint

Hybridization between RB and SU ecotypes occurs at the mid-shore, a patchy environment where both ecotypes are distributed sympatrically and incomplete assortative mating produces a variable percentage (2–40%) of morphologically intermediate forms (Rolán-Alvarez et al., 1997, 1999; Cruz et al., 2004; Conde-Padín et al., 2008). Following the description of Johannesson et al. (1993), we will categorize these intermediate individuals as ‘hybrids’ hereafter. Natural hybrids are apparently fertile and they mate at random with both pure ecotypes at the mid-shore (Rolán-Alvarez et al., 1999). However, to a certain extent, the role of endogenous and exogenous barriers between these ecotypes remains controversial because of some contradictory results. For example, Cruz & García (2001) and Cruz et al. 2001) found some reduced fitness for certain intermediate trait values using regression analysis of different shell traits on various fitness components. This could be interpreted as an indication of a hybrid sink effect (Barton, 1980; Barton & Hewitt, 1985) occurring in this hybrid zone. On the other hand, hybrid frequency is relatively stable across seasons and years, and their density is positively correlated with RB and SU density at the mid-shore, but not associated with micro-areas of low density (Carballo et al., 2005), in opposition to the hybrid sink effect. In addition, RB and SU ecotypes do not exhibit post-zygotic isolation under laboratory conditions (Rolán-Alvarez et al., 2004; Saura et al., 2011). However, some authors claim that detecting post-zygotic barriers in natural hybrid zones could be difficult, mainly due to methodological reasons, and therefore biasing the results in many systems towards exogenous selection mechanisms (see Bierne et al., 2011).

In the present study, we performed an AFLP genome scan in parallel hybrid zones of sympatric RB and SU ecotypes from the mid-shore. Outlier and nonoutlier loci from this genome scan were used for phylogeography and genetic structure analyses (including hybrids). The latter was able to determine genetic introgression. We also studied the correlation between genetic and phenotypic variation (shell size and shape), determining for the first time the genetic status of natural hybrids (following the study by Johannesson et al., 1993) in Galician populations of L. saxatilis. Our goal was to demonstrate that the observed phenotypic introgression reflected true genetic admixture of the parental forms. Additionally, genetic and phenotypic variation allowed us to make a rough estimate of the morphological heritability and find potential outlier loci linked to specific quantitative traits.

By comparing introgression patterns across the hybrid zone between localities with a known phylogeography, we are able to determine which selection mechanism (endogenous or exogenous) is most likely to control the hybridization and consequently gene flow. Under endogenous mechanisms, similar patterns of introgression between localities are expected. Although if endogenous mechanisms evolved recently or if exogenous mechanisms are affected by the genetic background, then evolutionarily close localities should present a similar pattern of introgression. Although if exogenous selection is the main mechanism responsible for the divergence, then local patterns of introgression are expected and these would be controlled by the local ecological conditions. We wanted to determine which of these mechanisms is involved in this L. saxatilis hybrid zone. Finally, we isolated and sequenced one of the most differentiated outlier loci detected in this study and then studied sequence variation on the same populations to confirm the former interpretation of genetic variation for AFLPs.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

Experimental Design

Littorina saxatilis samples from the mid-shore were collected in 2003 using a similar experimental design to Galindo et al. (2009), which studied upper- and lower-shore pure ecotypes. The mid-shore is a patchy area (mussels and barnacles) where RB, SU and hybrids are found in sympatry. Briefly, two areas of 1–3 m2 (transects 1 and 2) were sampled in three exposed shores of Galicia (NW Spain): Corrubedo, Silleiro and Cetarea (Fig. 1). Samples from each transect (approximately 300 individuals) were separated a posteriori into three classes (RB, SU and hybrids) based on their diagnostic characters (bands and ridges) as described by Johannesson et al. (1993). Sample sizes (see Fig. 1) are not representative of the frequencies of each class in the sampled area. Frequency of hybrids was always the lowest of the three classes but within the range of percentages observed in previous studies (range 2–40%; Rolán-Alvarez et al., 1999).

From a migration point of view (assuming a stepping stone model), Corrubedo is separated from the other two localities by three rias (i.e. inlets formed by drowned fluvial valleys open to the sea), this represents more than 200 Km of shore compared with the 25 Km between Silleiro and Cetarea, where rias are absent (see Fig. 1).

From an ecological point of view, the localities are somewhat different. The slope of the rocky habitat ranges between 25–30° at Corrubedo, 5–10° at Silleiro and 15–20° at Cetarea (described in Johannesson et al., 1993) and the distance between the upper- and lower-shore habitats ranges between 10.5 and 12 m at Corrubedo, 20–27 m at Silleiro and 16.5–21.5 m at Cetarea (see fig.2 from Rolán-Alvarez et al., 2004). On the other hand, Cetarea was the locality with the lowest density of mussels at the mid-shore (J. Galindo, personal observation).

Morphometric Analysis

Shells were examined using a Leica MZ12 stereoscopic microscope (Leica, Wetzlar, Germany), and digital images were captured using a Leica digital ICA video camera and Leica QWin Lite v2.2 software (Leica). Then, we estimated the centroid size (CS; estimating shell size), uniform (affine; U1 and U2) and nonuniform (nonaffine; relative warps RW1 and RW2) components of shell shape using 11 landmarks (excluding LM3) following previous work on L. saxatilis (Carvajal-Rodríguez et al., 2005; Conde-Padín et al., 2007, 2009). The relative warps were computed excluding the uniform component following the study by Rohlf (1993). Principal component analysis (PCA) of all these morphological variables was accomplished, and then we used the first component (PCA1) as an index that summarizes the morphological variation in our populations. All statistical analyses were performed using spss v.14.0 software (SPSS Inc., Chicago, IL, USA).

Amplified Fragment Length Polymorphism Analysis

Genomic DNA was isolated using a cetyltrimethyl ammonium bromide and chloroform extraction method following the study by Wilding et al. (2001). AFLP procedure and genotyping were performed using the same methodology as Galindo et al. (2009), in the same laboratory and during the same period of time. We analysed only nine (Eco+ACT, Eco+AAG, Eco+AGC, Eco+ACGG, Eco+ATCG, Eco+AGAC with Mse+CAC and Eco+ACGG, Eco+ATCG, Eco+AGAC with Mse+CAA) of the twelve selective primer combinations included in the previous study. Repeatability analysis was performed at the same time as Galindo et al. (2009), and following the same criteria, 94% repeatability was obtained. We finally scored 1995 polymorphic AFLP loci.

Outlier Analysis

Amplified fragment length polymorphism genome scans for the detection of outlier loci (i.e. loci that exhibit FST values that exceed the neutral expectations under a certain evolutionary model) have been successfully applied in a wide range of organisms (reviewed by Nosil et al., 2009a) including L. saxatilis (Wilding et al., 2001; Galindo et al., 2009). These outlier loci are supposed to be under divergent natural selection or linked to loci under selection (Nosil et al., 2009a), but other processes might be involved (Bierne et al., 2011).

We used two different methods to detect outlier loci, both widely used but known to perform differently (see Pérez-Figueroa et al., 2010). We implemented a less stringent analysis, using the program dfdist (Beaumont & Nichols, 1996; http://www.maths.bris.ac.uk/~mamab/stuff/), with the goal of obtaining a data set without outliers (‘nonoutliers’). Using a method that detects more outliers (including potential false positives, Caballero et al., 2008), we make sure that the chances of including a locus influenced by selection in our nonoutlier data set are low. For this same reason, we considered outliers those loci with < 0.05 and we did not apply multiple test correction. The methodology and the parameters are described in the study by Galindo et al. (2009). dfdist input files were created using AFLP convert program (http://webs.uvigo.es/acraaj/tools.htm). Additionally, to compare our percentage of outlier loci detected with a previous genome scan (Galindo et al., 2009), we performed Sequential Goodness of Fit (SGoF) multiple test correction (Carvajal-Rodríguez et al., 2009; Carvajal-Rodríguez & de Uña-Álvarez, 2011).

The second method, more stringent, is based on the program bayescan (Foll & Gaggiotti, 2008), which is an extension of Beaumont & Balding (2004), that uses population- and locus-specific FST coefficients. For each locus, two alternative models are defined: one that includes the effect of selection and another that excludes it. Then, the posterior probabilities of these two models are estimated using a reversible-jump MCMC approach. The Bayesian approach also deals with the problem of multiple testing of a large number of loci, as this number is taken into account through the prior distribution. We performed ten short pilot runs (10 000 iterations) to automatically tune the model parameters, followed by additional 400 000 iterations (20 thinning interval, 20 000 sample size) with a burn-in of 100 000 iterations. AFLP loci were considered outliers when log (Bayes Factor, BF) > 1.5 (posterior probability > 0.97), which indicates ‘very strong’ evidence for selection (Jeffreys' scale of evidence for BF, see bayescan manual). False Discovery Rate (FDR) correction is applied by default under this approach. A simulation study showed that bayescan outperforms other methods (e.g. dfdist) in determining outlier loci (Pérez-Figueroa et al., 2010).

We used the same three population pairwise comparisons in both analyses: Corrubedo, Silleiro and Cetarea. We combined transects and compared ecotypes within locality. Hybrids were excluded from these analyses.

Isolation of Outlier Loci

Our goal was to validate the pattern of variation found for AFLP outlier loci using sequence information. We chose one of the most differentiated outlier loci (767 locus, FST = 0.741, RB-SU Cetarea) from the analysis and selected four to six individuals containing the desired locus (band), and then we reran the selective PCRs from those individuals on a denaturing sequencing gel (6% acrylamide) alongside a 100-bp ladder (Fermentas, Vilnius, Lithuania). Bands of the correct size (bp) were excised from the gel and re-amplified using AFLP selective primers and PCR conditions from the study by Galindo et al. (2009), and then we cloned the PCR products using pGEM-T easy vector (Promega, Fitchburg, WI, USA) following manufacturer's instructions. Clones were re-amplified using AFLP primers and analysed on an ABI310 Genetic Analyzer (Applied Biosystems) alongside GeneScan ROX 500 (Applied Biosystems, Foster City, CA, USA). We sequenced the clones that matched the size of the outlier locus using BigDye Terminator v3.1 (Applied Biosystems) following the manufacturer's instructions. Sequencing reactions were analysed on an ABI310 Genetic Analyzer (Applied Biosystems). Primers were designed from sequences of the outlier locus using primer 3 (Rozen & Skaletsky, 2000), and PCR products were used as probes for screening the filters of the L. saxatilis BAC library (CHORI-317; http://bacpac.chori.org/home.htm) following the methodology of Wood et al. (2008). DNA extraction of one of the identified BACs was performed using a rapid alkaline lysis miniprep method as recommended by BACPAC resources. Then, we performed genome walking using the TOPO walker kit (Life Technologies, Carlsbad, CA, USA) and following manufacturer's instructions. Finally, primers were designed for a fragment of 820 bp (CET-365 locus), which contained part of the 767 outlier locus. PCRs to amplify CET-365 locus in population samples were performed in 20 μL final volume as follows: 2.5 mm MgCl2, 125 μm of each dNTP, 10 pmol direct and forward primers (CET-365 D 5′-TGTGAAAGGAAACCACATGT-3′; CET-365 F 5′-GCATTGGTTTACTTCCATAC-3′), 0.5 U of Taq polymerase (Bioline, London, UK) and 40 ng of DNA template in 1× PCR buffer (Bioline). Cycling was as follows: 32 cycles of 94° 30 s, 52° 30 s and 72° 1 min, with an initial step of 94° 5 min and a final extension of 72° 7 min. Sequencing reactions were carried out and analysed as previously described. Sequences were analysed following the study by Wood et al. (2008). Briefly, Chromas Lite 2.1 (Technelysium, Queensland, Australia) was used to correct the sequences and detect heterozygous profiles and alignment and trimming were performed using bioedit v.7.1 (Hall, 1999). Samples with multiple polymorphic sites were phased using the haplotypes from homozygous individuals. In this analysis, we used RB and SU individuals (excluding hybrids) from the AFLP analysis presented in this study (mid-shore, sympatric distribution) and individuals from the populations analysed in the previous study (Galindo et al., 2009), which correspond to pure RB from the upper-shore (UPPER) and pure SU from the lower-shore (LOWER) of the same localities. Sample size was five individuals (ten alleles) for all the populations except for Silleiro upper-shore RB (n = 6) and Silleiro mid-shore SU (n = 4). All these individuals belonged to transect 1 of each locality.

Genetic Structure Analysis

Nei's distance (after Lynch & Milligan, 1994) was calculated with the program aflp-surv v.1.0 (Vekemans et al., 2002). A neighbour-joining (N-J, Saitou & Nei, 1987) tree was constructed based on Nei's distance using the neighbor routine in phylip (Felsenstein, 1981). One thousand bootstraps, performed over loci, of Nei's distance were estimated in aflp-surv, and consense routine in phylip was used to determine the percentage of bootstraps supporting each branch of the tree. Population structure was also determined using the Bayesian clustering method implemented by the program structure v.2.3 (Pritchard et al., 2000; Falush et al., 2003, 2007). We performed five replicate runs of 500 000 iterations (burn-in 100 000) from one to nine clusters (K), assuming an admixture model, correlated allele frequencies and without prior population information. We used the Evanno method of ΔK (Evanno et al., 2005) as implemented in structure harvester (Earl & vonHoldt, 2012) to determine K. Results were plotted using the replicate with the highest likelihood value, but all the replicates were previously plotted and evaluated (results not shown). N-J trees and structure analyses were performed on two data sets: nonoutlier and outlier. For the nonoutlier data set (dfdist analysis), we removed all the outliers detected in any of the three pairwise comparisons (Corrubedo, Silleiro and Cetarea). For the outlier data set (bayescan analysis), we included all the outlier loci detected in any of the three comparisons. Notice that such strategy produces two conservative subgroups of potential outlier and nonoutlier AFLP loci.

Pairwise FST (Weir & Cockerham, 1984) was calculated with the program aflp-surv v.1.0 (Vekemans et al., 2002). We analysed the loci corresponding to each of the nine primer combinations (ranging 166–256 loci) independently. With these results, a randomization anova (Peres-Neto & Olden, 2001) was performed to determine the significance of the variation in divergence between ecotypes across localities.

The sequences from the outlier locus (CET-365) were analysed using dnasp v.5 (Librado & Rozas, 2009), and sequence diversity (π) and FST were calculated. Phylogenetic network was constructed using network v.4.6 (www.fluxus-engineering.com; Bandelt et al., 1999).

Introgression Analysis

Population admixture was determined within each locality (combining the two transects) with structure v.2.3, using the same methodology as in the genetic structure analysis. In this case, K from 1 to 5 was tested. To infer hybridization within localities, we also used a different Bayesian clustering method (newhybrids v.1.1; Anderson & Thompson, 2002). This program assumes that the samples come from a population where hybridization may have occurred and that a random sample of individuals is genotyped at multiple unlinked loci. It considers six genotype categories: pure species 1, pure species 2, F1 hybrids, F2 hybrids, and the first backcross generation to pure species 1 or pure species 2, with the results giving the estimated probabilities that each individual belongs to each of the different genotype categories. Simulations were run for 800 000 iterations (burn-in 100 000) and Jeffreys priors were assumed. Structure analyses were performed using the locality-specific outliers and also the locality-specific nonoutliers. We only used locality-specific outliers for the NewHybrids analysis.

Principal coordinate analysis (PCoA) was used to calculate, within locality, pairwise similarities (Jaccard's similarity coefficient) between AFLP multilocus genotypes (representing L. saxatilis individuals) for locality-specific outliers and locality-specific nonoutliers. Analyses were computed using the software PCO (http://www.stat.auckland.ac.nz/~mja/Programs.htm). Correlations between the first component of the PCoA and the morphological PCA1 (see above) were performed using spss v.14.0 software (SPSS inc.).

Heritability Estimation

The basis of the estimator proposed by Ritland (1996, 2000) is as follows. Let the value of a quantitative trait Y for two individuals i and j be Yi for the first and Yj for the second. Their shared phenotypes are measured by the cross-product.

  • display math

where U and V are the sample mean and variance of Y, respectively, in the population. Among all pairs, the average Zij equals the phenotypic correlation. If shared phenotypes are determined by shared genes and environments, then

  • display math

where rij is the regression relatedness coefficient for dominant markers (see Ritland, 2005 for further details), re is the correlation due to shared environments (assumed constant for all pairs), and eij is the random error. This is a linear regression equation, so over several pairs of individuals, the heritability can be estimated as

  • display math

where Cov(Zij, rij) is the covariance between the estimated relatedness and phenotypic similarity, and Var(rij) is the actual variance of relatedness. If the actual variance of relatedness is statistically not significant, then at least, the mere presence of genetic variation can be ascertained by testing for positive Cov(Zij, rij).

Ritland's heritability estimator is implemented in the program MARK, available at http://genetics.forestry.ubc.ca/ritland/programs.html. Heritability and covariance estimates were obtained for five quantitative traits (CS, U1, U2, RW1 and RW2) using RB, SU and hybrid individuals and the locality-specific outlier loci detected with bayescan. One hundred bootstraps were performed, and the results were averaged over bootstrapped data sets and localities.

Detection of Locus–Trait Associations

We performed a stepwise regression to identify associations between the outlier loci and quantitative traits (CS, U1, U2, RW1 and RW2), considered as independent and dependent variables, respectively. The adjusted R2 indicates the proportion of the trait variance explained by a given locus, and a significant regression value determines an association between loci and traits. Statistical analyses were performed using SPSS v.14.0 software (SPSS Inc.).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

Genetic Structure

The results of the outlier analysis with dfdist and bayescan are presented in Fig. 2. The percentage of bayescan outliers within locality was approximately 2%. When SGoF multiple test correction is applied to the dfdist analysis, we detect 3.2 ± 0.44% of outliers within locality, compared with 2.8 ± 0.26% from the study by Galindo et al. (2009) using the same methodology. The outlier loci data set (bayescan) contained 92 loci, which are the combination of the outliers found in each locality. From those outliers, 3 were shared between all the localities, 2 were shared between Corrubedo and Silleiro, other 2 outliers were shared between Corrubedo and Cetarea, and the highest number of shared outliers (7) were detected between Silleiro and Cetarea (see Fig. 2).

image

Figure 2. Number of outlier loci detected with bayescan (BF > 1.5) and dfdist (< 0.05, without multiple test correction in italics). The number of outliers shared between localities for bayescan and dfdist (italics) is also shown.

Download figure to PowerPoint

The N-J tree obtained for the nonoutlier data set (1598 loci) grouped samples by locality rather than ecotype, even for the closest localities (Silleiro and Cetarea) (Fig. 3a). Populations within locality grouped by ecotype, and divergence between transects is lower than between ecotypes. Corrubedo is clearly differentiated from the other two localities, this is also corroborated by the structure analysis where = 2 was assessed (see Table S1). The N-J tree obtained for the outlier data set (92 loci) (Fig. 3b) grouped samples by ecotype, RB and SU. The same result was obtained in the corresponding structure analysis (= 2; Fig. 3b and Table S1).

image

Figure 3. (a) Neighbour-joining (N-J) tree using Nei's distance for nonoutlier loci (1598 loci). Structure analysis using nonoutlier loci for = 2. (b) N-J tree for outlier loci (92 loci). Structure analysis using outlier loci for = 2. Branches show bootstrap values in percentage (1000 replicates). Corrubedo (COR), Silleiro (SIL) and Cetarea (CET). The number after the name of the locality represents the transect (1, transect 1; 2, transect 2). Circle (Ridged-Banded ecotype), square (Smooth-Unbanded ecotype). Hybrids were not included in these analyses.

Download figure to PowerPoint

We calculated the FST values between ecotypes within each transect (6 comparisons) and for each primer combination (9 combinations, ranging from 166 to 256 loci). Then, we used these values (18 FST values per locality) in a randomization anova, and we found that Silleiro (FST = 0.042 ± 0.003) had a significantly (< 0.001) lower FST between ecotypes compared with Corrubedo (FST = 0.074 ± 0.004) or Cetarea (FST = 0.064 ± 0.006), but these two were not significantly different between them. These suggest that genetic differentiation between ecotypes was significantly lower (and therefore gene flow larger) in Silleiro than in the other two localities.

Introgression Analysis

Figure 4 shows the results of the structure analyses performed for each locality independently, including hybrid individuals, and using the nonoutlier loci specific of each locality. Two clusters (= 2) were detected in all these comparisons (see Table S2). In Corrubedo and Cetarea, each ecotype grouped separately. Hybrids were similar to SU ecotype in Corrubedo and similar to RB in Cetarea, although showing certain levels of introgression. In Silleiro, the result was very different, and ecotypes were not clearly differentiated, showing much more introgression than in the other two localities. Figure S1 shows the analysis performed with locality-specific bayescan outliers for = 2 (Table S2). Again, hybrids from Corrubedo were more similar to SU and to RB in Cetarea. Hybrids in Silleiro showed great variation in the level of introgression, being some of them truly intermediate. Cetarea was the locality with less introgression in general.

image

Figure 4. Structure analysis of introgression (= 2) for locality-specific nonoutlier loci.

Download figure to PowerPoint

The coefficient of variation for PCoA1 in outlier and nonoutlier loci is given in Table 1, and it was used as a way to estimate the genetic variation for both classes of loci across ecotypes. The results clearly show that there is a significant difference in the coefficient of variation of PCoA for outliers (and caused by hybrid individuals) but not for nonoutliers. This result indicates that the group of hybrid individuals is genetically more variable (exclusively for outliers) than RB or SU ecotypes, for all the localities. This implies that hybrids are the result of introgression for outlier loci. A similar conclusion can be achieved by studying the relationship between genetic variation in outlier loci and morphology (Fig. 5). The X-axis represents the first coordinate axis from the PCoA, explaining 59.5% of overall variation in outlier loci in Corrubedo, 45.6% in Silleiro and 57.5% in Cetarea. The Y-axis corresponds to the first PCA axis of the morphometric variables (CS, U1, U2, RW1 and RW2; 37.5% of the overall variation in morphology). The relationship was investigated in two ways: (i) using the initial phenotypic classification that we performed based on the shell characters (ridges and bands) following the study by Johannesson et al. (1993) (Fig. 5a) and (ii) using the program NewHybrids to categorize the individuals based on the genetic information and classifying them as RB, SU, F1, F2, backcross RB and backcross SU (Fig. 5b). Results are shown for each transect and locality in both cases. We found that the morphometric variables are highly correlated with the genetic variation for outlier loci. Corroborating the structure analyses of admixture, we found that hybrids are more similar to SU in Corrubedo and more similar to RB in Cetarea. Silleiro was the locality showing more introgression and with more intermediate individuals (morphologically and genetically). The main difference between Fig. 5a,b is that in Silleiro, most of what we classified phenotypically as RB (also considered RB with the first PCA axis of the morphometric variables) was classified as backcross RB with the NewHybrids analysis. This is just emphasizing the high levels of introgression found in Silleiro, even for outlier loci. Nevertheless, in all cases, the result supports a different introgression level across localities but not across transects within locality.

Table 1. Coefficient of variation in PCoA1 for outlier and nonoutlier loci across individuals within ecotype [Ridged-Banded (RB), hybrids and Smooth-Unbanded (SU)], localities (Corrubedo, Silleiro and Cetarea) and transects (T1 and T2). A randomization anova (1000 randomizations) was used to determine if differences between ecotypes were significant, using different estimates from localities and transects as pseudoreplicates
 Coefficient of outlier variationCoefficient of nonoutlier variation
RBHybridsSURBHybridsSU
  1. ns, nonsignificant.

  2. a

    < 0.05.

Corrubedo
T13.6825.214.237.0317.5912.10
T23.408.804.688.146.2012.52
Silleiro
T18.5385.864.0132.8719.0276.36
T29.2222.533.1445.1950.30
Cetarea
T16.057.222.303.5020.073.19
T24.3233.718.6817.915.14109.68
Average5.8730.564.5113.8918.8744.03
anova  4.57a  1.96 ns 
image

Figure 5. Relationship between the PCoA1 for outlier loci (abscise axis) and the PCA1 of the morphological variation (ordinate axis) for each locality (Corrubedo, Silleiro and Cetarea) and transect (T1 and T2). (a) Individuals grouped by ecotype classes [Ridged-Banded (RB), Smooth-Unbanded (SU) and hybrids] based on morphological characters (ridges and bands; see text). (b) Individuals grouped by genotypic classes according to the NewHybrids analysis with outlier loci (see text): RB (pure RB), SU (pure SU), F1, F2, backRB (backcross RB), backSU (backcross SU).

Download figure to PowerPoint

Outlier Locus: 365-CET

We analysed 60 individuals for 714 bp and detected six different haplotypes (deposited in GenBank with accession numbers KC527045- KC527050). Figure S2 shows the haplotype network and the haplotype frequencies per population. Haplotype 5 was only found in Silleiro and Cetarea and haplotype 4 was only found in Corrubedo. The populations with the lowest (0.000) haplotypic diversity were the SU populations in Silleiro and Cetarea where haplotype 5 was fixed. The population with the highest (0.733) diversity was Corrubedo mid-shore SU. As expected, Cetarea was the population with the highest ecotype divergence (FST = 0.656); these results further corroborate the former analyses. Silleiro was the locality with the lowest divergence in agreement with the results of highest introgression in this locality (see Figs 4 and 5). Corrubedo showed high divergence between pure ecotypes (FST = 0.293) but lower in the mid-shore (FST = −0.077). Shared haplotypes between northern and southern localities were mostly present in RB populations. On the other hand, haplotype 5 was fixed in southern SU populations, whereas it was not found in the north. Finally, haplotype 4 (responsible for most of the ecotype divergence) in Corrubedo was not found in Silleiro or Cetarea.

Covariance and Heritability Analyses

The covariance between the similarity of phenotypic traits and the molecular relatedness obtained from locality-specific bayescan outliers averaged over pairs of individuals is presented in Fig. S3 (with bootstrapped data sets and localities). Significant and positive covariance was detected for CS, U1, U2, RW1 and RW2 (average 0.122 ± 0.020), indicating the presence of genetic variation for these traits. Heritability estimates were very low, although significant, mean heritability was 0.021 ± 0.004 for several traits (CS, U1, U2, RW1 and RW2).

Locus–Trait Association Analyses

Significant associations, within locality, between bayescan outlier loci and each quantitative trait of shell shape (CS, U1, U2, RW1 and RW2) are shown in Table 2. Some loci explained the phenotypic variation simultaneously for different traits within or between localities, but in other cases the loci were unique in trait and locality. The average percentage of the total explained phenotypic variance across traits and localities was 64.9 ± 2.9. From this total phenotypic variance, 23.8 ± 0.9 corresponded to the common explained phenotypic variance suggesting that an important proportion of the locus–phenotype associations are specific of trait and locality, perhaps suggesting different genetic architecture across traits (see 'Discussion').

Table 2. Detected informative outlier loci for different morphological traits (CS, U1, U2, RW1 and RW2). In brackets: percentage of the phenotypic variance explained by the locus on the basis of stepwise regression analysis. In bold: loci detected for more than one trait and percentage of the common phenotypic explained variance
 CS (%)U1 (%)U2 (%)RW1 (%)RW2 (%)
  1. P: significance level

Corrubedo1471 (20)391 (18) 1691 (16) 1691 (25) 1691 (28)
1360 (15)680 (16) 230 (10) 1422 (13)962 (23)
1702 (13) 1368 (12) 1365 (7)1240 (6)
1203 (11) 230 (7)   34 (5)
766 (9) 690 (7)   36 (5)
605 (7)   1517 (5)
1452 (5)   662 (4)
24 (3)    690 (4)
742 (3)   632 (3)
758 (3)    1702 (3)
Total explained (%)8960264586
Common explained (%) 13 14 26 25 35
P 0.0230.0270.0310.0400.022
Silleiro 508 (18) 605 (15)24 (16)693 (20)680 (18)
343 (12)1702 (13)289 (13) 1422 (11) 196 (15)
1810 (12) 841 (10) 485 (13) 1810 (10) 1452 (9)
1143 (8) 1422 (10) 276 (8) 230 (6) 1365 (8)
421 (6)1135 (8) 841 (7) 1365 (6) 1368 (8)
318 (6) 508 (7) 758 (5)1569 (5)1635 (8)
1224 (4)   1224 (5) 1781 (5)1360 (5)
   743 (4) 36 (3)151 (4)
    632 (4)
     276 (3)
     743 (3)
     1224 (2)
Total explained (%)6663716687
Common explained (%) 34 27 24 27 16
P 0.0470.0120.0300.0340.034
Cetarea841 (14)1420 (47) 151 (13) 24 (26) 151 (36)
1471 (9) 742 (13)  1143 (16) 962 (14)
 577 (9) 743 (11) 1168 (12)
 371 (5)  1240 (7) 352 (5)
  353 (4)  1422 (7) 1240 (5)
 391 (4) 308 (5)384 (4)
 289 (3)  962 (4) 690 (4)
 693 (3) 64 (3) 742 (3)
 1810 (3) 318 (3)680 (2)
    1168 (3) 1635 (2)
   1216 (3)1409 (1)
    353 (2)  
   1279 (2) 
   1581 (2) 
   1707 (2) 
   1224 (1) 
   1368 (1) 
   1496 (1) 
   1812 (1) 
Total explained (%)23911310088
Common explained (%) 0 17 13 16 70
P 0.0320.0030.0170.0330.041

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

Hybrid zones between taxa undergoing ecological speciation represent interesting systems to study the interaction between selection and gene flow (Nosil, 2012). Littorina saxatilis has been claimed as one of these systems (Rolán-Alvarez, 2007; Butlin et al., 2008; Schluter, 2009; Johannesson et al., 2010; Nosil, 2012), and there have been a wealth of studies in recognition of this. Previously, intermediate individuals (i.e. potential hybrids or introgressed individuals) have been genetically studied in other populations of L. saxatilis. For example, Panova et al. (2006) used neutral microsatellite markers to perform assignment tests to determine the hybrid composition of Swedish populations, including shell shape analyses. Grahame et al. (2006) studied samples along transects in UK populations, comprising the exposed and sheltered habitats and included intermediate individuals around the habitat transition zone. They studied the shell shape variation and genetic variation (outlier AFLP loci between pure ecotypes, Wilding et al., 2001) along these transects. However, in these populations (Sweden and United Kingdom), hybrids and pure ecotypes rarely appear sympatrically like in the Galician hybrid zone (range of ecotype frequency at the mid-shore; RB = 10–34%; hybrid = 2–40%; SU = 31–84%; Rolán-Alvarez et al., 1999). Therefore, the phenotypic and genetic analysis of sympatric populations of RB and SU ecotypes, including intermediate individuals (i.e. hybrids sensu Johannesson et al., 1993), carried out in this study is key to give insights into the role of hybridization during the initial stages of the speciation continuum in L. saxatilis.

Genetic Structure and Introgression

AFLP genome scans have been previously performed between divergent ecotypes of L. saxatilis (Wilding et al., 2001; Galindo et al., 2009). Wilding et al. (2001) studied populations from north-east UK, and Galindo et al. (2009) analysed pure RB and SU ecotypes (upper- and lower-shore populations) from the same localities studied here. In this study, we investigated mid-shore populations where RB and SU ecotypes present an overlapping distribution. The two studies used a similar experimental design and sample sizes. In the mid-shore AFLP genome scan, we detected 3.2% of outliers and Galindo et al. (2009) detected 2.8% using the same methodology. This reveals a similar proportion of outlier loci irrespective of the shore level or rate of hybridization. This comparison has never been made in L. saxatilis, and it could be explained whether ecologically based divergent natural selection is strong enough to maintain (directly or indirectly) the divergence for outliers in the face of gene flow (e.g. hybridization). Even in the presence of selection, hybridization promotes recombination between the parental gene pools and consequently should reduce the proportion of outliers due the loss of the transitory linkage to the genomic region under selection. However, variation in recombination rates due to the genetic architecture of reproductive isolation (e.g. outlier loci within regions of restricted recombination) could overcome this effect. For example, Roesti et al. (2012) found outlier loci close to centromeres and Jones et al. (2012) in inversions, both in stickleback populations. Additionally, divergence hitchhiking (Via, 2009, 2012) could be another possible mechanism contributing to the pattern observed in our study. Mapping information is necessary to assign outlier loci to regions of reduced recombination or to detect divergence hitchhiking; thus, this should be taken into account in future studies.

In this study, the phylogeography analysis (Fig. 3) showed different results for outlier and nonoutlier data sets. This was expected because outlier and nonoutlier loci are controlled, to a great extent, by different evolutionary forces (divergent natural selection and genetic drift, respectively). Nonoutlier loci (Fig. 3a) showed a clear geographic pattern, with populations grouped by locality. A similar pattern has been previously found for mtDNA between RB and SU ecotypes from other localities of the Galician shores (NW Spain) (Quesada et al., 2007). This is also the case for United Kingdom (Wilding et al., 2001) and Swedish (Panova et al., 2006) L. saxatilis populations. The structure analysis (Fig. 3a) showed two groups: northern (Corrubedo) and southern populations (Silleiro and Cetarea). We hypothesize that these two groups are formed by rather different gene pools due to drift. Outlier loci (Fig. 3b) showed a clear ecological pattern, grouping populations by ecotype (RB and SU). This result was also found by Wilding et al. (2001), analysing 15 outlier AFLP loci. In both studies, this result reflects that similar ecological conditions in different localities control (directly or indirectly) the genetic variation at outlier loci through divergent natural selection (see Galindo et al., 2009). These results, a geographic pattern obtained for putatively neutral loci (nonoutliers) and ecological subdivision (RB vs. SU) for outlier loci, have been found in other examples undergoing ecological speciation (Nosil et al., 2009a and references therein).

Here, we also demonstrated the presence of current gene flow through the genetic corroboration of hybridization between the ecotypes (RB and SU) and introgression in the parental neutral genetic backgrounds. Our results suggest that the presence of gene flow during these initial stages of divergence does not promote homogenization of putatively adaptive (phenotypic and genetic) divergence (Fig. 5a), as ecotypes remain clearly differentiated morphologically and for outlier loci genetic variation. We also tried to discern between historical and ecological factors responsible for the rate of hybridization and/or introgression between RB and SU ecotypes by studying three replicate hybrid zones. These hybrid zones showed a rather different composition of hybrids (Figs 5a and S1). In Corrubedo, hybrids are similar to SU, in Cetarea similar to RB, and in Silleiro they are very variable; some are similar to RB, others to SU and some are intermediates. From the hybridization point of view, Silleiro is the locality with a greater level of gene flow and also presents greater introgression for nonoutlier loci (see Fig. 4). With these results, we demonstrated that hybridization in these populations lacks of a clear phylogenetic signal, as the pattern of hybridization did not show more similarity between the populations sharing a more recent ancestor (Silleiro and Cetarea) (Fig. 3a). These rather different introgression patterns observed in this study are in agreement with hybridization being mainly caused by the specific microhabitat conditions in each locality (e.g. local ecological factors). Comparisons between transects within locality (Fig. 5a) did not show differences. As we described in the experimental design (see Material and Methods), Corrubedo was the locality with the shortest average distance between the pure RB and SU ecotypes, consequently the RB ecotype is found closer to the lower-shore habitat. Corrubedo is also the locality with the steepest slope in the rocky habitat, which is translated into greater wave exposure (SU-like habitat). On the other hand, Cetarea had the lowest density of mussels at the mid-shore, decreasing the area of SU-like habitat (J. Galindo, personal observation). Slope values were smaller and distance between pure ecotypes longer, compared with Corrubedo. Finally, the mid-shore in Silleiro had the smallest slope, making this habitat less exposed. Moreover, the mosaic of patches of mussels and barnacles occupied the biggest area of all three localities and densities of RB, SU and hybrids at the mid-shore were high (J. Galindo, personal observation). These descriptions and the fact that hybridization consistently differed between localities but not within would suggest that the pattern of hybridization might be mainly influenced by ecological rather than historical factors (exogenous vs. endogenous selection).

Differential introgression patterns across hybrid zones have been previously found for example in sculpins (Nolte et al., 2009). Different hypotheses have been discussed to explain this pattern; these include different local ecological conditions, different genetic architecture of reproductive barriers, stochasticity (e.g. drift) and different population structure. In our case, the role of ecology in hybridization is clear, and in addition, we achieved this conclusion by an independent line of evidence. In hybrids, we observed a clear correlation between multivariate indexes of outlier AFLP loci and morphological variables, both presumably adaptive (see Table 1). All these results are in agreement with a nonallopatric origin of the populations analysed similarly to previous studies (Rolán-Alvarez et al., 2004; Quesada et al., 2007; Galindo et al., 2009). However, we cannot rule out that endogenous selection could have a minor role in these hybrid zones. In summary, ecologically based divergent natural selection (exogenous selection) seems to be the main factor responsible for both the ecotype divergence (reviewed in Rolán-Alvarez, 2007) and the pattern of hybridization (this study) in Galician populations of this species.

Until now, not many studies have analysed AFLP outlier loci in more detail across population samples (Wood et al., 2008; Midamegbe et al., 2011; Nunes et al., 2012; Paris & Despres, 2012). This study and the study by Wood et al. (2008) corroborated through sequencing of outlier loci the results obtained with the outlier AFLPs in L. saxatilis populations. The sequence analysis of the outlier locus (365-CET) (Fig. S2) supported the role of genetic drift between northern (Corrubedo) and southern localities (Silleiro and Cetarea), which shared most of the haplotypes. On the other hand, as expected, we detected great divergence between ecotypes at Cetarea (even at the mid-shore). Moderate divergence was found in Corrubedo, but a different haplotype was responsible for such divergence (haplotype 3). Silleiro lacked divergence between the ecotypes. The different patterns observed at distinct localities could be the result of local ecological conditions controlling for ecotype divergence; consequently, different loci or alleles would show divergence in different localities. For example, if one allele at Cetarea has been hitchhiked by a recent adaptive mutation at a different locus, this effect would not be observed in a different locality. Another possible explanation would be that this locus is under selection (directly or indirectly) in the southern localities. Although, as it has been shown with the AFLPs, Silleiro presents high levels of introgression, hybridization would dilute the divergence between the ecotypes for this locality. Because we have already shown that outliers are not influenced by the hybridization level, the first hypothesis seems more likely. According to previous studies isolating and sequencing outlier loci from AFLP genome scans, the chances of detecting the direct target of selection are very low; these loci are more likely to represent a loosely linked marker (see Nosil et al., 2009a). Wood et al. (2008) and Nunes et al. (2012) for example did not find a candidate gene or a promoter in their sequences. This was also the case for locus 365-CET. Then, there is the open question of whether it is worthwhile to sequence these outlier loci as discussed by Nunes et al. (2012). The simple answer is probably no, if we are interested in finding the target or targets of selection. Nowadays, next-generation sequencing technologies facilitate these tasks (Stapley et al., 2010; Ekblom & Galindo, 2011), and good examples of this are three-spine stickleback studies (Hohenlohe et al., 2010; Jones et al., 2012; Roesti et al., 2012) where in a single experiment, they can map signatures of selection directly onto the stickleback genome, determining the genetic basis and genetic architecture of reproductive isolation.

Phenotype–Genotype Relationship

For divergent natural selection to act on a quantitative trait, heritable genetic variation is necessary. This could be quantified by estimating the heritability. In L. saxatilis, the heritability of the morphological variables (CS, U1, U2, RW1, RW2) studied here has been estimated in a previous study and using wild families (h2 = 0.5; Conde-Padín et al., 2007). In the former study, shells from embryos within the brood of wild females were used to estimate the heritability. This approach could be importantly biased because full-sibs were assumed (multiple paternity is present in this species; Panova et al., 2010) or due to potential environmentally confounded effects (see discussion in Conde-Padín et al., 2007). Here, we used the method of Ritland (1996) as an independent corroboration of our previous estimates of heritability for these morphological variables. In the present study, genetic variation obtained from Ritland's heritability estimator was identified for these morphological traits, corroborating previous findings and confirming the existence of significant genetic variation under the morphological variation observed in the wild and presumably associated with the adaptive change (see also Conde-Padín et al., 2007). However, heritability estimates were significant although low (< 0.1), probably due to the absence of familiar structure and small population sizes (Rodríguez-Ramilo et al., 2007). The precise estimation of traits' heritability may still be open to future experimental refinement.

In stable hybrid zones, parental lines are not expected to be inbred. However, this is not a complexity for Quantitative Trait Loci (QTL) mapping (Besnier et al., 2010). In fact, QTL studies have been conducted recently within natural populations (Kruuk et al., 2008; Slate et al., 2009). Although one of the basic requirements for QTL mapping is a genetic map of variable markers, in the present study numerous transitory locus–trait associations have been identified for different quantitative traits. However, the percentage of the common explained phenotypic variance was low, suggesting that different outlier loci are responsible for the genetic variability of different quantitative traits, and therefore, these outliers may not be the direct target of selection. The detection of diverse informative outlier loci for different morphological traits also points towards complex selective forces acting in this system.

Conclusions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

We showed that phenotypically intermediate individuals represent extensive genetic introgression for neutral loci (nonoutliers) and partially for outlier loci, a signature of ongoing gene flow during its divergence. Different localities, with different local ecological conditions, presented different levels of introgression and in different directions, and this was not linked with population history. Therefore, exogenous selection mechanisms are likely to be responsible for the maintenance of this hybrid zone, but a minor role of post-zygotic mechanisms cannot be excluded and further studies are necessary in this direction. This L. saxatilis hybrid zone showed its potential to study the role of hybridization (i.e. gene flow) in the early stages of ecological speciation and the role of ecology in hybridization. Future studies should focus on determining more precisely the genetic architecture of adaptive and speciation traits by including mapping information of the used markers.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

We thank John W. Grahame and Roger K. Butlin for use of the Littorina BAC library, and Henry Wood for assistance with its use. We also thank Paloma Morán for her input in the design of the experiment and genetic analyses, and Nerea González-Lavín and Nieves Santamaría for technical assistance. This work has been funded by Ministerio de Ciencia e Innovación (MCI) (CGL2008-00135/BOS), Fondos Feder: Unha maneira de facer Europa and Xunta de Galicia (INCITE09 310 006 PR; 10PXIB 310044PR). Juan Galindo is currently supported by an ‘Isidro Parga Pondal’ fellowship (Xunta de Galicia) and Mónica Martínez-Fernández by a ‘Juan de la Cierva’ research fellowship (JCI-2010-06167) from MICINN.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information
FilenameFormatSizeDescription
jeb12152-sup-0001-FigS1.pdfapplication/PDF104KFigure S1 Structure analysis of introgression (= 2) for locality-specific outlier loci.
jeb12152-sup-0002-FigS2.pdfapplication/PDF104KFigure S2 Sequence variation for the outlier locus 365-CET.
jeb12152-sup-0003-FigS3.pdfapplication/PDF36KFigure S3 Covariance between phenotypic similarity and relatedness.
jeb12152-sup-0004-TableS1-S2.docxWord document80K

Table S1 Structure analysis overall populations: ΔK and log likelihood values.

Table S2 Structure analysis within locality: ΔK and log likelihood values.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.