Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Biological Journal of the Linnean Society, 2008, 93, 799–812. With 7 figures Morphological divergence patterns among populations of Poecilia vivipara (Teleostei Poeciliidae): test of an ecomorphological paradigm JOSÉ L. GOMES JR and LEANDRO R. MONTEIRO* Laboratório de Ciências Ambientais, CBB, Universidade Estadual do Norte Fluminense. Avenue. Alberto Lamego, 2000. Campos dos Goytacazes, RJ, cep 28013-600, Brazil Received 17 December 2006; accepted for publication 30 June 2007 The structure of body size and shape divergence among populations of Poecilia vivipara inhabiting quaternary lagoons in South-eastern Brazil was studied. This species is abundant throughout an environmental gradient formed by water salinity differences. The salinity gradient influences the habitat structure (presence of macrophytes) and the fish community (presence of large predators). Size and shape variation within and among populations was quantified by geometric morphometrics and analysed by indirect and direct gradient ordinations, using salinity and geography as a framework. Morphological divergence was associated with the salinity gradient. The evolutionary allometries observed were independent of within-group static allometries. Sexually dimorphic patterns were observed in size variation and within-population allometries. Specimens from freshwater (higher predation) sites presented smaller sizes, relatively longer caudal regions, lower anterior regions and a ventrally displaced eye. These features are consistent with an ecomorphological paradigm for aquatic organisms from populations subject to intense predation. A process of directional selection is postulated as the most likely force driving diversification among P. vivipara populations. © 2008 The Linnean Society of London, Biological Journal of the Linnean Society, 2008, 93, 799–812. ADDITIONAL KEYWORDS: body shape – body size – evolutionary allometry – geographical variation – geometric morphometrics – natural selection – predation. INTRODUCTION The study of body shape variation in fish populations subject to environmental gradients has provided interesting examples of natural selection. The most commonly studied agents of selection for body shape divergence among populations of the same species are water flow (streams/lakes) (Brinsmead & Fox, 2002; Langerhans et al., 2003), habitat use/feeding efficiency (Bourke, Magnan & Rodríguez, 1997; Walker, 1997; Svanbäck & Eklöv, 2003, 2004), and predation/escape performance (Walker, 1997; Ghalambor, Walker & Reznick, 2003, Ghalambor, Reznick & Walker, 2004; Langerhans & DeWitt, 2004; Langerhans et al., 2004). Experimental evidence has been reported of feeding *Corresponding author. Current address: Department of Biological Sciences, The University of Hull, Cottingham Road, Hull HU6 7RX, UK. E-mail: [email protected] habits generating phenotypic plasticity in body shape, usually associated with the anterior region and the trophic apparatus (Robinson & Wilson, 1995; Ruehl & DeWitt, 2005). Body size variation is also correlated with environmental differences experimented by populations, particularly with predation, as a consequence of life-history evolution (Endler, 1995; Reznick et al., 1997). On the other hand, environmental variables, such as water salinity, can cause nonselective size variation by influences on the metabolism (Trexler, 1989; Bouef & Payan, 2001). The findings of shared convergent patterns of interspecific shape divergence in different aquatic vertebrate species subject to similar environmental gradients has allowed the elaboration of an ecomorphological paradigm for the correlation of prey species body shape with predation pressure (Langerhans & DeWitt, 2004; Langerhans et al., 2004). This paradigm provides a framework for the interpretation of © 2008 The Linnean Society of London, Biological Journal of the Linnean Society, 2008, 93, 799–812 799 800 J. L. GOMES JR and L. R. MONTEIRO body shape divergence patterns and a functional hypothesis for predation as an agent of selection. This interpretation, along with quantitative genetics theory of genetic drift as a null model (Lande, 1979; Ackermann & Cheverud, 2002; Monteiro & Gomes, 2005), allows for robust assessments of evolutionary processes acting on interpopulational phenotypic divergence patterns, even for species not yet exhaustively studied. The livebearing fish Poecilia vivipara Bloch & Schneider 1801 is abundant and widespread in the recently formed (approximately 5000 years ago) quaternary lagoons of Northern Rio de Janeiro State. Its high tolerance to environmental extremes, particularly salinity and temperature (Trexler, 1989), makes it one of the few species present in all lentic environments of the region (Bizerril & Primo, 2001). The lagoons present an environmental gradient as a result of water salinity, where the water bodies closer to the ocean present brackish to salt water (sometimes more saline than the sea), and the water bodies farther from the ocean present freshwater. The salinity gradient influences habitat structure (absence of macrophytes in high salinity; Suzuki, Ovalle & Pereira, 1998, 2002) and the fish community (freshwater predators absent from higher salinity sites; Bizerril & Primo, 2001). Being omnipresent throughout the environmental gradient, P. vivipara populations are exposed to varied conditions, such as different predation pressures, habitat structures, food availability, and physicochemical water characteristics. A previous study has demonstrated that the magnitude of body shape variation among P. vivipara populations is high, even for sites within the same lagoon, but with different environmental characteristics (Neves & Monteiro, 2003). Divergence rate tests applied to morphological differences among population means with known maximum times of separation indicated directional selection as the most probable process influencing shape diversification among P. vivipara populations (Monteiro & Gomes, 2005). Although suggestive of natural selection, these previous results lack the determination of possible agents of selection to explain the body shape divergence observed. The present study aimed to evaluate possible evolutionary mechanisms and causal factors influencing the morphological variation among populations of P. vivipara exposed to an environmental gradient related to water salinity and predation. Accordingly, we compared observed divergence patterns with evolutionary neutral expectations from spatial analysis and quantitative genetics. We also compared our results with proposed ecomorphological paradigms associated with different agents of selection (predation, habitat structure, food availability) aiming to identify convergent patterns of shape changes that can be used as evidence of natural selection. MATERIAL AND METHODS STUDY AREA The quaternary plains at Northern Rio de Janeiro State (Fig. 1) were formed by sediment deposition during formation of the Paraíba do Sul River delta, and sea level fluctuations associated with palaeoclimatic modifications during the Holocene (starting approximately 5100 years ago; Martin et al., 1997). The construction of the river delta continuously formed river arms, which were later abandoned and became long, narrow lagoons scattered all over the region (Martin et al., 1997; Soffiati, 1998; Primo, Bizerril & Soffiati, 2002). Historically, there was considerable contact among the lagoons and the main river body (particularly during floods). However, the construction of a large network of drainage and irrigation channels by the National Department of Work Against Drought (DNOS) after 1950, considerably decreased the water level in the region and isolated most lagoons from all others (Primo et al., 2002). The distance from the sea determined a considerable amount of variation in environmental conditions (related to water salinity differences) of different lagoons, as well as environmental gradients within lagoons (Suzuki et al., 1998, 2002). The Feia and Campelo lagoons are the oldest in the region. Their origin is related with river sediments deposited during the first development of the Paraíba do Sul River delta southwards, starting 5000 years ago (Martin et al., 1997; Soffiati, 1998). Subsequently, the main river mouth was displaced towards the east and grew in that direction, mostly by marine sediment deposition. The eastern lagoons (such as Açu, Grussaí, and Iquipari), closer to the ocean were mostly formed by abandonment of ancient river arms (Primo et al., 2002). These are younger lagoons (up to 2000 years) and present a salinity gradient, decreasing from the sites close to the ocean (the sand bars) towards the interior of the continent. The salinity gradients influence the composition of the fish community in the region (Bizerril & Primo, 2001). Hoplias malabaricus, the main predator of P. vivipara (Mazzoni & IglesiasRios, 2002), is a freshwater species absent from higher salinity sites (Garcia et al., 2003). The marginal vegetation (macrophytes) is usually absent from higher salinity sites and abundant in freshwater sites (Suzuki et al., 1998, 2002). The mean salinities of collection sites are listed in Table 1. FIELD AND LABORATORY METHODS The specimen collections were carried from May 2002 to September 2004. Specimens were collected with a © 2008 The Linnean Society of London, Biological Journal of the Linnean Society, 2008, 93, 799–812 MORPHOLOGICAL DIVERGENCE AND ECOMORPHOLOGY 801 Brazil o 40 W 41oW 22oS Rio de Janeiro 4 7 18 22 15 23 9 26 8 6 10 12 16 17 11 24 19 20 25 3 13 2 1 21 5 14 0 5 10 N 20 km Figure 1. Map showing the geographical distribution of sampling sites in Northern Rio de Janeiro State. Numbering corresponds to Table 1. pole seine (1-mm2 mesh), humanely killed in iced water, fixed in 10% formalin an stored in 70% alcohol. From each site, a number of specimens (N = 34–60) of each sex were used (Table 1), summing up to 1376 males and 1538 females. Only adult animals (sexually mature) were included in the sample. All females were carrying developing embryos and the males had a fully developed gonopodium and visible secondary sexual characters. Where the lagoons presented a salinity gradient, two or more sites were sampled at different positions along the gradient. Salinity in each site was measured with a densimeter (Seatest, © 2008 The Linnean Society of London, Biological Journal of the Linnean Society, 2008, 93, 799–812 802 J. L. GOMES JR and L. R. MONTEIRO Table 1. Samples of Poecilia vivipara collected in the lagoons Sample sizes Number Site name Females Males Salinity (ppt) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 Açu Lagoon - Interior Açu Lagoon - Bridge Açu Lagoon - Sandbar Barra do Itabapoana Barra do Furado Bom Jardim Buena Cacimba Campelo Lagoon Cataia Lagoon Cima Lagoon Fazenda Frecheira Feia Lagoon - Ponta Grossa Feia Lagoon - Canto do Sobrado Três Vendas Grussai Lagoon - Sandbar Grussai Lagoon - Interior Imburi Iquipari Lagoon - Sandbar Iquipari Lagoon - Interior Lagamar Praia Lagoon - Site 2 Praia Lagoon - Site 1 Taí Grande Lagoon Taí Pequeno Lagoon Limpa Lagoon 60 60 60 60 60 60 60 60 60 59 60 60 60 60 60 60 59 60 60 51 59 60 60 52 59 59 46 60 50 60 60 60 35 60 60 60 59 60 59 60 60 34 60 60 50 54 60 57 60 40 52 – 13* 26.9* 30.5* 12 16 0 0 0 0* 0 0 0 0 0 0 7* 2.3* 0 11.1* 1.7* 5 0 0 0 0 0 Numbers shown correspond to numbering in figures. Salinities marked with an asterisk represent the means of a 2-year time series. Aquarium Systems). For some lagoons (Table 1), a 2-year monthly time series was available from an independent study (M. S. Suzuki, unpubl. data), and a mean salinity was calculated for each site. The specimens were photographed by a digital camera (Pixera Professional) and the coordinates of 13 landmarks (Fig. 2A, B) were registered for each individual, using the TpsDig software (Rohlf, 2006). Because of the conspicuous sexual dimorphism in body size, coloration and the shape difference caused by the anal fin modification in males (to form the gonopodium), sexes were analysed independently. GEOMETRIC AND MULTIVARIATE METHODS Body size was estimated as centroid size: the square root of the summed squared distances from each landmark to the configuration centroid, which is the average of all landmarks (Bookstein, 1991; Monteiro & Reis, 1999). Body size variation among sites was assessed by analysis of variance (ANOVA) and a Figure 2. Landmarks digitized on specimens for males (A) and females (B). © 2008 The Linnean Society of London, Biological Journal of the Linnean Society, 2008, 93, 799–812 MORPHOLOGICAL DIVERGENCE AND ECOMORPHOLOGY regression of mean centroid sizes on the mean salinity values for each site. The landmark configurations were superimposed separately for males and females by Procrustes superimposition (Rohlf & Slice, 1990). This least squares fit effects linear transformations of similarity (scaling, translation and rotation), to minimize the summed squared distances among corresponding landmarks in a sample, allowing for the estimation of a mean shape (Monteiro & Reis, 1999). The aligned co-ordinates were projected to the space spanned by partial warps and uniform components (Bookstein, 1991), calculated using the grand mean shape (over all samples) as reference. The among-site variation was assessed by submitting the matrix of shape variables (partial warps and uniform components) to a canonical variates analysis (Monteiro & Reis, 1999). Minimum spanning trees were calculated from Mahalanobis distances and combined to the scatterplot of canonical scores to provide a more accurate view of the shape similarities and to check the efficiency of dimensionality reduction. The importance of allometric patterns of shape differentiation to the observed among-site divergence patterns was assessed by a multivariate analysis of covariance (MANCOVA), using partial warps and uniform component scores as dependent variables and centroid size as covariate. The allometric patterns influencing major axes of variation within-samples were assessed by regressions of relative warps (principal components of the shape space spanned by partial warps and uniform components) on centroid size. The spatial pattern of morphological differentiation was assessed by a comparison of shape distances (Procrustes distances) among sites with geographical distances. The Procrustes distances represent the metric of the shape space used for statistical comparisons using geometric morphometric methods (Dryden & Mardia, 1998). The Procrustes and geographical distances were compared directly by Mantel (matrix correlation) tests and by multivariate Mantel correlograms (Legendre & Fortin, 1989; Legendre, 1993). The correlograms divide the distance matrices by classes belonging to different intervals of geographical distances (lags) and calculate a matrix correlation for each interval. Bootstrap confidence intervals for the matrix correlation estimates were calculated with 10 000 replications. Partial Mantel tests and correlograms were calculated by removing the influence of salinity from shape distances. A matrix of salinity differences was calculated using the module of the salinity difference between each pair of sites. The matrices of Procrustes distances obtained for different sexes were also compared by matrix correlation. The association of shape variables (mean site scores of partial warps and uniform components) with mean 803 water salinity and mean centroid size was assessed by a two-block partial least squares analysis (Rohlf & Corti, 2000). This analysis searches for pairs of vectors (shape space-variable space) explaining maximum covariance between the two blocks of variables (Monteiro, Duarte & Reis, 2003). To assess the existence of a linear spatial gradient, the latitude and longitude for each sampled site were included in the variable matrix. The significance of covariance explained was tested by a permutation test (10 000 replications). The proportion of shape variation explained by the partial least squares vectors was assessed by the coefficient of determination for the multivariate regression (Monteiro, 1999) of shape variables (partial warps and uniform components) on partial least squares vector scores. A test for the hypothesis that a neutral process (genetic drift) would be sufficient to explain the amongsite shape variation patterns was performed (Ackerman & Cheverud, 2002; Marroig & Cheverud, 2004). This test is based on the prediction from quantitative genetic theory that, under genetic drift, the amount of divergence among populations should be proportional to the variation in the ancestral population (Lande, 1979), which can be estimated by the pooled withingroups additive genetic covariance matrix or its surrogate, the pooled within-group phenotypic covariance matrix. A large amount of evidence is available in the literature supporting the use of phenotypic covariance matrices as surrogates for genetic covariance matrices (Cheverud, 1988; Roff, 1996). To compare the proportionality between magnitudes and patterns of variance distribution in among and within-group matrices, we calculated eigenvalues of within-group and amonggroup covariance matrices of shape variables. A regression of log-transformed among-group eigenvalues on log-transformed within-group eigenvalues was performed and 95% confidence intervals (CIs) for the regression coefficients were calculated (Ackerman & Cheverud, 2002). Regression coefficients equal (or not statistically different than) to 1 indicate an adjustment to the expectation of genetic drift. Regression coefficients significantly larger than 1 indicate a process of directional selection influencing morphological diversification because there is more variation amonggroups than expected by the genetic drift model (Marroig & Cheverud, 2004). The software used for the geometric analyses were the TPS series (TPSRelw, TPSPLS, TPSRegr; Rohlf, 2006). The canonical variates analyses and the calculations of among-groups and within-group covariance matrices were performed on NTSYS-pc (Rohlf, 2000). ANOVAS and MANCOVAS were performed with the package ‘stats’, which is part of the R-system (R Development Core Team, 2006). The matrix correlations and correlograms were calculated with the © 2008 The Linnean Society of London, Biological Journal of the Linnean Society, 2008, 93, 799–812 804 J. L. GOMES JR and L. R. MONTEIRO add-on package ‘ecodist’ (Goslee & Urban, 2006) in the R-system. RESULTS BODY SIZE VARIATION 10 The ANOVAs showed that mean differences in body size, were highly significant for males (F = 168.22, P < 0.0001) and females (F = 144.67, P < 0.0001) from different sites. Body sizes were larger in lagoons with salinity gradients (Açu, Iquipari, and Grussai Lagoons). For both sexes, samples from sand bar sites showed larger mean body sizes than samples from interior sites (Fig. 3). Within- and among-group size variation was larger for female than for male samples (Fig. 3). A positive association of mean body size and mean salinity was observed in both sexes. The estimated regression equations (using only site means 8 6 4 2 Centroid Size A 0 5 10 15 20 25 30 25 30 10 Salinity (ppt) 8 6 4 2 Centroid Size B 0 5 and depicted in Fig. 3) were: female mean size = 3.73 + 0.09 ¥ Salinity (F = 63.52, R2 = 0.7342, P < 0.0001); and male mean size = 2.96 + 0.07 ¥ Salinity (F = 60.4; R2 = 0.7242, P < 0.0001). 10 15 20 Salinity (ppt) Figure 3. Scatterplots showing the association of centroid size with salinity for different sites. A, female data set. B, male data set. Dashed line indicates the least squares regression. BODY SHAPE VARIATION The ordination of samples by canonical variates was significant for both sexes. In the female data set (Wilks’ lambda < 0.0001; F = 26.572; P < 0.0001), the first two canonical axes depict 44.87% of variation among sites (Fig. 4). The first canonical axis (representing 32.97% of variation) ordinate samples from freshwater sites with positive scores and saltwater sites with negative scores. Shape differences associated with freshwater sites in the first axis are a more ventrally positioned eye, a more dorsally positioned pectoral fin, an anteriorly positioned (and relatively larger) mouth, a lower body height, and a longer caudal peduncle (the quadrilateral formed by landmarks (#10, #11, #12, #13; (Fig. 4). The second canonical axis (representing 11.90% of variation) does not ordinate sites in a clear spatial or environmental pattern (Fig. 4). The minimum-spanning tree connects samples which are closest to each other in the space defined by Mahalanobis distances (i.e. the whole set of canonical variates). The connection of sites that appear distant in the two-dimensional scatterplot, such as sites #13 (Ponta Grossa) and #14 (Canto do Sobrado), reflects the loss of information during the projection from a multidimensional space to a two-dimensional scatterplot. In the ordination of the male data set (Wilks’ lambda < 0,0001; F = 21 055; P < 0,0001), the two first canonical axes explain 45.15% of variation among sites. The first canonical axis (explaining 33.39% of variation) ordinates samples from saltwater sites with positives scores and freshwater sites with negative scores (Fig. 5). Samples from freshwater sites (negative scores present a more anteriorly positioned mouth, ventrally displaced eye, dorsally displaced pectoral fin, lower body height, and longer caudal peduncle. The second axis (explaining 11.76% of variation) does not ordinate sites in a clear spatial or environmental pattern (Fig. 6). The minimum spanning tree suggests that the reduction of dimensionality caused a smaller loss of information than in the female data set. The importance of allometric patterns to the observed differences among sites was assessed by a MANCOVA, using centroid size as a covariate and partial warps and uniform components as variables. The significant results indicate that the differences in shape for males and females from different sites were not fully allometric (Table 2). However, the interaction term of Size ¥ Site (test for parallelism) is significant, indicating that the allometric trajectories in © 2008 The Linnean Society of London, Biological Journal of the Linnean Society, 2008, 93, 799–812 MORPHOLOGICAL DIVERGENCE AND ECOMORPHOLOGY 805 2.58 8 Canonical Variate 2 (11.90%) 22 15 5 1.19 11 26 4 23 12 19 1 17 -0.21 6 7 13 21 18 3 16 2 9 14 20 24 -1.60 10 25 -3.00 -4.00 -2.03 -0.05 1.92 3.89 Canonical Variate 1 (32.97%) Figure 4. Canonical variates analysis for the female data set. Numbers follow the sequence in Table 1 and Fig. 1. Grey lines correspond to minimum spanning tree from Mahalanobis distances. Grey dots correspond to brackish and saltwater sites, whereas black dots correspond to freshwater sites. Larger grids below the scatterplot correspond to deviations in shape associated with positive (right) and negative (left) scores of the first canonical axis. Smaller grids on the side correspond to deviations in shape associated with positive (upper) and negative (lower) scores of the second canonical axis. Table 2. Multivariate analysis of covariance models for among-site shape differences (partial warps and uniform components) using centroid size as covariate Females Size Site Size ¥ Site Males Size Site Size ¥ Site d.f. Wilks’ lambda F Numerator d.f. Denominator d.f. P (> F) 1 25 25 0.13 0.001520 0.35 437.45 21.24 2.93 22 550 550 1 465 24 340 24 340 < 2.2 ¥ 10-16 < 2.2 ¥ 10-16 < 2.2 ¥ 10-16 1 24 24 0.161 0.004 0.452 309.127 16.672 2.020 22 528 528 1 305 21 300 21 300 < 2.2 ¥ 10-16 < 2.2 ¥ 10-16 < 2.2 ¥ 10-16 d.f., degrees of freedom. shape space are not parallel among sites (Table 2). The significant interaction impedes an interpretation of the Sites term independently from Size. Separate regressions were adjusted for each relative warp with centroid size. For the female data set, the single axis with a significant relationship with size was the first relative warp, which explained 29.87% of total shape variation (i.e. considering the entire data set without regard of sample site). No relationship between the regression coefficients and salinity was observed (Fig. 6). The male data set presented association only between the second relative warp (explaining 13.21% of total shape variation) and centroid size. Again, although there is considerable variation on steepness of trajectories among sites, no relationship between allometric coefficients and water salinity was © 2008 The Linnean Society of London, Biological Journal of the Linnean Society, 2008, 93, 799–812 806 J. L. GOMES JR and L. R. MONTEIRO 2.83 25 Canonical Variate 2 (11.76%) 24 1 1.55 3 4 13 7 15 14 0.27 9 18 17 10 20 21 12 23 -1.02 11 6 5 8 2 16 19 22 -2.30 -3.00 -1.25 0.50 2.25 4.00 Canonical Variate 1 (33.39%) Figure 5. Canonical variates analysis for the male data set. Numbers follow the sequence in Table 1 and Fig. 1. Grey lines correspond to minimum spanning tree from Mahalanobis distances. Grey dots correspond to brackish and saltwater sites, whereas black dots correspond to freshwater sites. Larger grids below the scatterplot correspond to deviations in shape associated with positive (right) and negative (left) scores of the first canonical axis. Smaller grids on the side correspond to deviations in shape associated with positive (upper) and negative (lower) scores of the second canonical axis. observed (Fig. 6). Because the samples were composed by adults, the allometric patterns observed should be considered static allometries (non-ontogenetic). The comparison of shape distances (Procrustes) with geographical distances by matrix correlation was marginally significant for both sexes. The full matrix correlation for the female data set was r = 0.187, P = 0.0601 whereas, for the male data set, the correlation was r = 0.171, P = 0.0584 (tests based on 10 000 permutations). The Mantel correlograms (not shown) presented significant spatial autocorrelation only at small geographical scale (less than 16 km for females and less than 6 km for males). This geographical range limits the significance of shape-spatial correlation to sites within the same lagoon. The observed 95% CI for matrix correlations in the significant distance class was in the approximate range 0.05–0.20 in both sexes. The partial matrix correlations and correlograms were all nonsignificant, after removing the influence of salinity differences among sites from shape distances. The matrix correlation between shape distances and salinity differences was r = 0.418, P = 0.0019 for the female data set and r = 0.379, P = 0.0001 for the male data set. The comparison of Procrustes distances among sites obtained for males and females was significant (r = 0.324, P = 0.0013), suggesting a moderate correlation among divergence patterns based on different sexes. The test for genetic drift performed, using amonggroup and within-group covariance matrices for shape variables showed regression coefficients significantly higher than 1 both for females (b = 1.867, 95% CI = 1.692–2.041) and males (b = 1.862, 95% CI = 1.563–2.161). This result is inconsistent with the expectation of the genetic drift model, suggesting a process of natural selection behind the morphological diversification among sites. The partial least squares analysis for among-site shape variation showed a significant pattern of association of shape with size, salinity, and geographical coordinates for both sexes. For the female data set, 93.78% of total covariances among the two groups of variables was explained by the first pair of partial least squares vectors (PLS vectors). The amount of covariance explained was significant in the permutation test (P = 0.0012). For the male data set, the first pair of PLS vectors explain 92.52% of total covariance among variable blocks (permutation test P = 0.0062). This pair of vectors suggest an interpopulational divergence pattern associated with mean body size and water © 2008 The Linnean Society of London, Biological Journal of the Linnean Society, 2008, 93, 799–812 MORPHOLOGICAL DIVERGENCE AND ECOMORPHOLOGY 807 2 4 6 8 10 A 25 26 19 20 0.05 0.00 -0.05 21 22 23 24 0.05 0.00 Relative Warp 1 -0.05 13 14 15 16 17 18 7 8 9 10 11 12 0.05 0.00 -0.05 0.05 0.00 -0.05 1 2 3 4 5 6 0.05 0.00 -0.05 2 4 6 8 10 2 4 6 8 10 2 4 6 8 10 Centroid Size 2 B Relative Warp 2 0.04 0.02 0.00 -0.02 -0.04 -0.06 0.04 0.02 0.00 -0.02 -0.04 -0.06 3 4 5 6 2 3 4 5 6 21 22 23 24 25 16 17 18 19 20 11 12 13 14 15 6 7 8 9 10 1 2 3 4 0.04 0.02 0.00 -0.02 -0.04 -0.06 0.04 0.02 0.00 -0.02 -0.04 -0.06 5 0.04 0.02 0.00 -0.02 -0.04 -0.06 2 3 4 5 6 2 3 4 5 6 2 3 4 5 6 Centroid Size Figure 6. Within-group association of major axes of variation and centroid size. A, female data set comparing the first relative warp and centroid size. B, male data set comparing the second relative warp and centroid size. Lines correspond to least squares regressions. salinity differences. The shapes associated with positive values (higher salinity sites) in the first PLS vector pair present a relatively smaller mouth (and turned upwards), dorsally displaced eyes and opercles, ven- trally displaced pectoral fin, higher bodies, and a shorter caudal peduncle than the shapes from freshwater sites, with negative scores (Fig. 7). These shape differences are similar to the first canonical axis. The © 2008 The Linnean Society of London, Biological Journal of the Linnean Society, 2008, 93, 799–812 808 J. L. GOMES JR and L. R. MONTEIRO Females Males Easting Size Salinity Northing Figure 7. Partial least squares (PLS) results for the association of site mean shapes and geographical-environmental variables. The grids to the left represent the shape deviations along the first PLS shape vector for females. The grids to the right represent the shape deviations along the first PLS shape vector for males. Upper grids represent positive deviations and lower grids represent negative deviations. Arrows in the middle represent the correlations of variables with the first PLS variable vector, scaled so that the distance of box frames to the horizontal line in the middle correspond to a correlation of 1. PLS vector for variables associated with the shape vector indicates a linear geographically orientated (southeast–north-west) salinity gradient, influencing mean body sizes and shapes (Fig. 7). A multivariate regression of shape variables (partial warps and uniform components) on the PLS shape vector scores was used to calculate the percentage of shape variation explained by the PLS (as the multivariate coefficient of determination R2 based on Procrustes distances). For the female data set, the scores of the first PLS shape vector explain 41.1% of total shape variation among mean site shapes whereas, for the male data set, the scores of the first PLS shape vector explain 39.1% of total shape variation among mean site shapes. DISCUSSION The pattern of morphological diversification of P. vivipara populations during colonization of holocenic lagoons northern Rio de Janeiro is complex and certainly influenced by the mosaic of environmental differences as observed previously (Neves & Monteiro, 2003; Monteiro & Gomes, 2005). The environmental differences among lagoons are mainly influenced by water salinity with noticeable consequences on habitat structure (absence of macrophytes on sites with higher salinity) and the structure of fish com- munities (Bizerril & Primo, 2001). The largest and more abundant predator sharing the same habitat as P. vivipara is the trahira or wolf fish Hoplias malabaricus (Erythrinidae). Juvenile trahiras (sometimes small adults) were regularly collected along in the same seines we used for P. vivipara, but only in freshwater sites (where trahiras are even exploited commercially by local fishermen). The association of H. malabaricus and P. vivipara as predator and prey has been reported before, and P. vivipara is one of the most frequent items in H. malabaricus diet in Rio de Janeiro (Mazzoni & Iglesias-Rios, 2002). In a fish community in the venezuelan llanos, juveniles and subadult trahiras were the most important predators of Poecilia species (Winemiller, 1989). Hoplias malabaricus is considered a primary freshwater species and usually absent from sites with brackish to saltwater (Garcia et al., 2003). The only possible predators found in our brackish water collections were small cichlids (Geophagus brasiliensis and Cichlasoma facetum), which were observed to eat young, but not adult P. vivipara in the laboratory (J. L. Gomes Jr & L. R. Monteiro, pers. observ.). Poecilia vivipara populations from freshwater sites are certainly exposed to a high predation environment (that partly explain the size differences observed). On the other hand, water salinity can have a direct influence on fish © 2008 The Linnean Society of London, Biological Journal of the Linnean Society, 2008, 93, 799–812 MORPHOLOGICAL DIVERGENCE AND ECOMORPHOLOGY growth, by reduction of basal metabolic rates, increase in feeding and food conversion rates and optimization of osmoregulation in certain salinity levels (Bouef & Payan, 2001). Therefore, it is important to distinguish between direct and indirect salinity effects on size variation and evolutionary allometry. Because the observed patterns of morphological differentiation were generally similar in males and females, we first discuss the shared patterns between sexes and later point out the dimorphic patterns BODY SIZE VARIATION Our results show that mean body size is positively associated with water salinity, which could be a direct salinity effect on growth (Boeuf & Payan, 2001), an indirect salinity effect because of predation pressure in different fish communities, or both. In other livebearing species, individuals from low predation sites are usually larger than those from high predation sites (Endler, 1995; Johansson, Turesson & Persson, 2004). Results from a common garden experiment with P. vivipara populations (J. L. Gomes Jr & L. R. Monteiro, unpubl. data) indicate that individuals from a brackish water lagoon presented higher growth rates (and mean body size) than individuals from a freshwater lagoon even when raised in freshwater (the observed size differences are not due to plasticity alone). BODY SHAPE VARIATION The major axes of shape variation among sites were similar in males and females with respect to the ordination patterns and shape changes described. Because the ordination of sites is influenced by water salinity, which has been shown to influence body size, it is important to assess the importance of allometric patterns to the observed among-site shape variation. The results of the MANCOVA indicate that differences among sites are maintained even after removing the statistical influence of size (using a separate slopes model). However, the within-group allometric patterns were not homogeneous (nonparallel allometric trajectories), which suggests an uncoupling of within-group (static) and among-group (evolutionary) allometries. The allometric differences observed among sites, where specimens presented different mean sizes, were not just simple extrapolations of a common allometric curve. We used the term static allometry (Cheverud, 1982) for the within-group allometry because the samples were composed exclusively of reproductive adults, but it is possible that the within-group allometric trajectories incorporate some ontogenetic allometry because growth continues after sexual maturation (although at a much smaller rate; Snelson, 1989). 809 The shape changes depicted by the canonical axes are known to be associated with locomotion and foraging patterns. Differences in the relative size of the caudal region and body depth were reported to influence escape velocity (Walker, 1997). Langerhans & DeWitt (2004) and Langerhans et al. (2004) showed that individuals of Gambusia affinis, Poecilia reticulata, and Brachyrhaphis rhabophora from populations subject to predation present a relatively longer caudal peduncle, a shallower anterior body, and a lower position of the eye. This predation morphotype also presented higher burst swimming speeds (Langerhans et al., 2004) but, for a different result regarding body depth, see also Royle, Metcalfe & Lindström (2006). The lower position of the eye is presumably involved with enhanced predator detection (Langerhans & DeWitt, 2004). In the present study, the mean shapes corresponding to freshwater populations presented a phenotype convergent with the high predation morphotypes shared by P. reticulata, B. rabdophora, and G. affinis (Langerhans & DeWitt, 2004). Considering that the contrast of freshwater and brackish water environments probably also represents a difference in predation regimes (as discussed above), we can consider our results to be consistent with the ecomorphological paradigm for predation pressure in aquatic vertebrates proposed by Langerhans et al. (2004). Further support to this conclusion is given by the fact that the evolutionary allometries are not similar to static within-group allometries, and the observed patterns of morphological divergence among sites are not consistent with the expectation of genetic drift, according to quantitative genetic models for the comparison of shape variation patterns within and among groups (results from present study). Rate tests for morphological divergence performed in a different study of the same system (Monteiro & Gomes, 2005) also rejected the hypothesis of genetic drift. The tests of neutral evolution and the convergent patterns of shape divergence among P. vivipara and the other three Poeciliids suggest directional selection by predation as the process behind the major axes of shape variation. Other shape differences among sites include changes in the anterior region, such as the position of the pectoral fin, and the mouth angle. Such shape differences are likely to be influenced by feeding modes (scraping ¥ pelagic feeding and benthic ¥ surface feeding) (Robinson & Wilson, 1995; Ruehl & DeWitt, 2005). The changes in mouth angle associated with the canonical axes (most apparent for males) might reflect differences in food availability in different lagoons. The icon shape associated with positive scores in the first canonical axis in Figure 6 (the male ordination) shows a mouth angle (line between landmarks 1 and 2) directed upwards, whereas the icon associated with negative scores for the same axis shows a mouth angle © 2008 The Linnean Society of London, Biological Journal of the Linnean Society, 2008, 93, 799–812 810 J. L. GOMES JR and L. R. MONTEIRO more anteriorly directed. Considering that the sites associated with the positive scores present higher salinity and less macrophytes (open water) for scraping, the observed pattern of shape differences is consistent with the one obtained experimentally by Ruehl & DeWitt (2005) for G. affinis with attached food versus free prey diets. The spatial analysis showed a local pattern of shape divergence (autocorrelated sites were only those within the same lagoon) for both sexes. The lack of larger scale autocorrelations reflect the way the salinity environmental gradient is organized. As shown by the PLS analysis, the most important geographical factor influencing the salinity gradient is distance from the sea (with similar environments among distant lagoons and differences among closely spaced lagoons). Even the small scale autocorrelation is lost when the salinity effect is removed from the shape distances, suggesting that environmental differences are more important than geographical proximity to determine shape differences among sites. The divergence among specimens from different sites can be expected to be large, even within the same lagoons, because the migration rates for such small-sized livebearing species are quite small (even for pools separated by a few meters) (Chapman & Warburton, 2006), and gene flow is greatly influenced by geographical distance and barriers (Crispo et al., 2006). The lack of spatial pattern provides further evidence with respect to the evolutionary processes at stake in the morphological divergence among populations. Stochastic processes are expected to generate monotonically decreasing correlograms and higher matrix correlations (Telles & Diniz-Filho, 2000). SEXUALLY DIMORPHIC PATTERNS The results were generally similar for males and females, even with the strong sexual dimorphism characteristic of livebearing fish species. There are, however, differences regarding the observed allometric patterns, which reflect differences in within-site body size variation that might arise from different selective forces acting on the sexes. The males present consistently smaller means and variances than females, possibly because of selection for small males in opportunistic matings (Bisazza & Pilastro, 1997; Pilastro, Giacomello & Bisazza, 1997). Opportunistic matings play an important role in P. vivipara reproduction, as commonly observed both in laboratory and the field (J. L. Gomes Jr & L. R. Monteiro, pers. observ.). The males lack bright colour spots, with the exception of a temporary carotenoid-related orange coloration in the throat region (more common in saltwater lagoons). Females, on the other hand, can maximize reproductive fitness by attaining a larger size (Reznick, 1983; Cheong et al., 1984); hence, the larger mean sizes and within-group variation among females. Whenever the sexes are subject to different selection pressures, it is expected that their responses differ (Hendry et al., 2006). In the present study, the patterns of divergence were similar, but the correlation between Procrustes distances among sites using different sexes was moderate. These results suggest that the shared patterns of among site divergence are important and probably caused by the same effects, but the differential selective pressures still cause variation in responses of males and females. The dimorphism of shape variation patterns in P. vivipara is, nevertheless, smaller than the dimorphism in P. reticulata (Hendry et al., 2006), where males are much more conspicuous than females and present courtship behaviours. The most obvious dimorphic pattern is the difference in within-group allometries. The first relative warp (the major axis of within-group variation) is the principal component associated with body size for females whereas, for the male data set, the axis associated with size is the second relative warp. The within-group allometric patterns are the major axis of shape variation for females, but a secondary axis of shape variation for males. This result is probably a reflection of the smaller magnitude of within-group size variation observed for males, which, in turn, is a consequence of dimorphic growth patterns (Snelson, 1989; Hendry et al., 2006), and different selection pressures. CONCLUSIONS Livebearing fishes are extremely tolerant to environmental variation, easily kept in captivity, and usually abundant and widespread. They are therefore excellent models for the study of morphological, behavioural, and life-history adaptation. Poecilia vivipara is abundant in all lentic environments in the holocenic lagoons in Northern Rio de Janeiro (particularly in salt or brackish water), facing physical and biological differences in their environments (a magnitude of variation seldom found in other poeciliid models). This system should allow for the understanding of the combined and unique effects of different environmental variables on body shape, size, and life history. The pattern of size and shape divergence observed among populations of P. vivipara inhabiting different sites is influenced by the gradient formed by water salinity. However, the evolutionary allometry observed seems to be caused by an independent combination of size and shape divergence in response to exposure to predators in freshwater sites. The major axis of among-site shape differences depicts a convergence with contrasts between predator and predator-free sites in the literature for three other poeciliid species. © 2008 The Linnean Society of London, Biological Journal of the Linnean Society, 2008, 93, 799–812 MORPHOLOGICAL DIVERGENCE AND ECOMORPHOLOGY The observed convergence, the different within-group allometries, the neutral hypothesis tests, and the lack of a spatial structure, along with previous published rate tests, suggest an evolutionary process of directional selection as responsible for the among-site divergence pattern. The ecomorphological paradigm proposed by Langerhans et al. (2004) provides both a functional hypothesis relating a possible adaptation to its selection agent, and the comparative test by shared convergences with different lineages (Langerhan & DeWitt, 2004). Further research in this system should be directed to an assessment of the genetical divergence structure among populations, common garden experiments (partially accomplished), field work with the postulated major predators, and the life-history evolution of P. vivipara populations. ACKNOWLEDGEMENTS The authors want to thank A. C. Pessanha for assistance in field and laboratory work. Previous versions of this manuscript have been greatly improved by the comments of C. R. Ruiz-Miranda, M. S. Suzuki, M. J. Cavalcanti, M. C. Gaglianone, and M. R. Nogueira. The collection permits were issued by The Instituto Brazileiro do Meio Ambiente e dos Recursos Naturais Renováveis (IBAMA). This work was funded by research grants and fellowships from the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq), Fundação de Amparo à Pesquisa do Estado do Rio de Janeiro (FAPERJ), and Universidade Estadual do Norte Fluminense. This is contribution number 103 from the Graduate Program in Ecology and Natural Resources, Universidade Estadual do Norte Fluminense. REFERENCES Ackermann RR, Cheverud JM. 2002. Discerning evolutionary processes in patterns of tamarin (genus Saguinus) craniofacial variation. American Journal of Physical Anthropology 117: 260–271. Bisazza A, Pilastro A. 1997. Small male mating advantage and reversed size dimorphism in poeciliid fishes. Journal of Fish Biology 50: 397–406. Bizerril CRSF, Primo PBS. 2001. Peixes de águas interiores do Estado do Rio de Janeiro. Rio de Janeiro: FEMARSEMADS. Bookstein FL. 1991. Morphometric tools for landmark data geometry and biology. New York, NY: Cambridge University Press. Bouef G, Payan P. 2001. How should salinity influence fish growth? Comparative Biochemistry and Physiology C 130: 411–423. 811 Bourke P, Magnan P, Rodríguez MA. 1997. Individual variation in habitat use and morphology in brook charr. Journal of Fish Biology 51: 783–794. Brinsmead J, Fox MG. 2002. Morphological variation between lake- and stream-dwelling rock bass and pumpkinseed populations. Journal of Fish Biology 6: 1619–1638. Chapman P, Warburton K. 2006. Postflood movements and population connectivity in gambusia (Gambusia holbrooki). Ecology of Freshwater Fish 19: 357–365. Cheong RT, Henrich S, Farr JA, Travis J. 1984. Variation in fecundity and its relationship to body size in a population of the Least killfish, Heterandria formosa (Pisces: Poeciliidae). Copeia 1984: 720–726. Cheverud JM. 1982. Relationships among ontogenetic, static, and evolutionary allometry. American Journal of Physical Anthropology 59: 139–149. Cheverud JM. 1988. A comparison of genetic and phenotypic correlations. Evolution 42: 958–968. Crispo E, Bentzen P, Reznick DN, Kinnison MT, Hendry AP. 2006. The relative influence of natural selection and geography on gene flow in guppies. Molecular Ecology 15: 49–62. Dryden IL, Mardia KV. 1998. Statistical shape analysis. New York, NY: John Wiley and Sons. Endler JA. 1995. Multiple-trait coevolution and environmental gradients in guppies. Trends in Ecology and Evolution 10: 22–29. Garcia AM, Raseira MB, Vieira JP, Winemiller KO, Grimm AM. 2003. Spatiotemporal variation in shallowwater freshwater fish distribution and abundance in a large subtropical coastal lagoon. Environmental Biology of Fishes 68: 215–228. Ghalambor CK, Reznick DN, Walker JA. 2004. Constraints on adaptive evolution: the functional trade-off between reproduction and fast start swimming performance in the Trinidadian guppy (Poecilia reticulata). American Naturalist 164: 38–50. Ghalambor CK, Walker JA, Reznick DN. 2003. Multi-trait selection, adaptation, and constraints on the evolution of burst swimming performance. Integrative and Comparative Biology 43: 431–438. Goslee S, Urban D. 2006. Ecodist: dissimilarity-based functions for ecological analysis. R package, version 1.01. Available at http://www.r-project.org. Hendry AP, Kelly ML, Kinnison MT, Reznick DN. 2006. Parallel evolution of the sexes? Effects of predation and habitat features on the size and shape of wild guppies. Journal of Evolutionary Biology 19: 741–754. Johansson J, Turesson H, Persson A. 2004. Active selection for large guppies, Poecilia reticulata, by the pike cichlid, Crenicichla saxatilis. Oikos 105: 595–605. Lande R. 1979. Quantitative genetic-analysis of multivariate evolution, applied to brain–body size allometry. Evolution 33: 402–416. Langerhans RB, DeWitt TJ. 2004. Shared and unique features of evolutionary diversification. American Naturalist 164: 335–349. Langerhans RB, Layman CA, Langerhans AK, DeWitt © 2008 The Linnean Society of London, Biological Journal of the Linnean Society, 2008, 93, 799–812 812 J. L. GOMES JR and L. R. MONTEIRO TJ. 2003. Habitat-associated morphological divergence in two Neotropical fish species. Biological Journal of the Linnean Society 80: 689–698. Langerhans RB, Layman CA, Shokrollahi AM, DeWitt TJ. 2004. Predator-driven phenotypic diversification in Gambusia affinis. Evolution 58: 2305–2318. Legendre P. 1993. Spatial autocorrelation: trouble or new paradigm? Ecology 74: 1659–1673. Legendre P, Fortin M. 1989. Spatial pattern and ecological analysis. Vegetatio 80: 107–138. Marroig G, Cheverud JM. 2004. Did natural selection or genetic drift produce the cranial diversification of neotropical monkeys? American Naturalist 163: 417–428. Martin L, Suguio K, Dominguez JML, Flexor JM. 1997. Geologia do quaternário costeiro do litoral Norte do Rio de Janeiro e do Espírito Santo. Belo Horizonte: CPRM. Mazzoni R, Iglesias-Rios R. 2002. Distribution patterns of two fish species in a coastal stream in Southeast Brazil. Brazilian Journal of Biology 62: 171–178. Monteiro LR. 1999. Multivariate regression models and geometric morphometrics: the search for causal factors in the analysis of the shape. Systematic Biology 48: 192–199. Monteiro LR, Duarte LC, Reis SF. 2003. Environmental correlates of geographical variation in skull and mandible shape of the punaré rat Thrichomys apereoides (Rodentia: Echimyidae). Journal of Zoology 261: 47–57. Monteiro LM, Gomes JL Jr. 2005. Morphological divergence rates for natural selection: uncertainly of parameter estimation and robustness of results. Genetics and Molecular Biology 28: 345–355. Monteiro LR, Reis SF. 1999. Princípios de Morfometria Geométrica. Ribeirão Preto: Holos Editora. Neves FM, Monteiro LR. 2003. Body shape and size divergence among populations of Poecilia vivipara in coastal lagoons of Southeastern Brazil. Journal of Fish Biology 63: 928–941. Pilastro A, Giacomello E, Bisazza A. 1997. Sexual selection for small size in male mosquitofish Gambusia holbrooki. Proceedings of the Royal Society Series B, Biological Sciences 264: 1125–1129. Primo PBS, Bizerril CRSF, Soffiati A. 2002. Lagoas do Norte Fluminense-Perfil Ambiental. Rio de Janeiro: SEMADS. R Development Core Team. 2006. R: a language and environment for statistical computing. Available at http:// www.r-project.org Reznick D. 1983. The structure of guppy life histories tradeoff between growth and reproduction. Ecology 64: 862–973. Reznick DN, Shaw FH, Rodd FH, Shaw RG. 1997. Evaluation of the rate of evolution in natural populations of guppies (Poecilia reticulata). Science 275: 1934–1937. Robinson BW, Wilson DS. 1995. Experimentally induced morphological diversity in trinidadian guppies. Copeia 1995: 294–305. Roff DA. 1996. The evolution of genetic correlations: an analysis of patterns. Evolution 50: 1392–1403. Rohlf FJ. 2000. NTSYS-Pc: numerical taxonomy and multivariate analysis system, Version 2.0. Setauket: Applied Biostatistics Inc. Rohlf FJ. 2006. Morphometrics at SUNY Stony Brook. Available at http://life.bio.sunysb.edu/morph/ Rohlf FJ, Corti M. 2000. Use of two-block partial least squares to study covariation in shape. Systematic Biology 49: 740–753. Rohlf FJ, Slice DE. 1990. Extensions of the Procrustes method for the optimal superimposition of landmarks. Systematic Biology 11: 1–16. Royle NJ, Metcalfe NB, Lindström J. 2006. Sexual selection, growth compensation and fast-start swimming performance in green swordtails, Xiphophorus helleri. Functional Ecology 20: 662–669. Ruehl CB, DeWitt TJ. 2005. Trophic plasticity and finegrained resource variation in populations of western mosquitofish, Gambusia affinis. Evolutionary Ecology Research 7: 801–819. Snelson F. 1989. Social and environmental control of life history traits in poeciliid. In: Meffe A, Snelson FF, eds. Ecology and evolution of poeciliid fishes (poeciliidae). Princeton, NJ: Prentice Hall, 149–161. Soffiati A. 1998. Aspectos históricos das lagoas do Norte do Estado do Rio de Janeiro. In: Esteves FA, ed. Ecologia das Lagoas Costeiras do Parque Nacional da Restinga de Jurubatiba e do Município de Macaé, RJ. Rio de Janeiro: NUPEM-UFRJ, 1–35. Suzuki MS, Figueiredo RO, Castro SC, Silva CF, Pereira EA, Silva JA, Aragon GT. 2002. Sand bar opening in a coastal lagoon (Iquipari) in the northern region of Rio de Janeiro State: hydrological and hydrochemical changes. Brazilian Journal of Biology 62: 51–62. Suzuki MS, Ovalle ARC, Pereira EA. 1998. Effects of sand bar openings on some limnological variables in a hypertrophic tropical coastal lagoon of Brazil. Hydrobiologia 368: 111–122. Svanbäck R, Eklöv P. 2003. Morphology dependent foraging efficiency in perch: a trade-off for ecological specialization? Oikos 102: 273–284. Svanbäck R, Eklöv P. 2004. Morphology in perch affects habitat specific feeding efficiency. Functional Ecology 18: 503–510. Telles MPC, Diniz-Filho JAF. 2000. Null expectations of spatial correlograms under a stochastic process of genetic divergence with small sample sizes. Genetics and Molecular Biology 23: 739–743. Trexler JC. 1989. Phenotypic plasticity in poeciliid life histories. In: Meffe A, Snelson FF, eds. Ecology and evolution of poeciliid fishes (poeciliidae). Princeton, NJ: Prentice Hall, 201–213. Walker JA. 1997. Ecological morphology of lacustrine threespine stickleback Gasterosteus aculeatus L. (Gasterosteidae) body shape. Biological Journal of the Linnean Society 61: 3–50. Winemiller KO. 1989. Ontogenetic diet shifts and resource partitioning among piscivorous fishes in the Venezuelan llanos. Environmental Biology of Fishes 26: 177–199. © 2008 The Linnean Society of London, Biological Journal of the Linnean Society, 2008, 93, 799–812