Body shape evolution among ploidy levels of the Squalius alburnoides hybrid complex (Teleostei, Cyprinidae)

Authors

  • C. CUNHA,

    1. Departamento de Biodiversidad y Biología Evolutiva, Museo Nacional de Ciencias Naturales, CSIC, Madrid, Spain
    2. Departamento de Biologia Animal, Centro de Biologia Ambiental, Faculdade de Ciências de Lisboa, Universidade de Lisboa, Lisbon, Portugal
    Search for more papers by this author
  • M. BASTIR,

    1. Departamento de Paleobiología, Museo Nacional de Ciencias Naturales, CSIC, Madrid, Spain
    Search for more papers by this author
  • M. M. COELHO,

    1. Departamento de Biologia Animal, Centro de Biologia Ambiental, Faculdade de Ciências de Lisboa, Universidade de Lisboa, Lisbon, Portugal
    Search for more papers by this author
  • I. DOADRIO

    1. Departamento de Biodiversidad y Biología Evolutiva, Museo Nacional de Ciencias Naturales, CSIC, Madrid, Spain
    Search for more papers by this author

Carina Cunha, Departamento de Biodiversidad y Biología Evolutiva, Museo Nacional de Ciencias Naturales, CSIC, C/José Gutiérrez Abascal, 2, 28006 Madrid, Spain.
Tel.: +34 91 411 13 28; fax: +34 91 564 50 78; e-mail: mcncj676@mncn.csic.es, cjcunha@fc.ul.pt

Abstract

Hybridization, ploidy level and genomic constitution may be important to respond to different environments, by producing different phenotypes and thus reducing competitive interaction. Through geometric morphometrics, we examined variation in body size and shape among biotypes of the Squalius alburnoides hybrid complex and their sperm donor (Squalius carolitertii). Results showed that S. carolitertii is significantly larger in size than the biotypes of the complex. No significant relationship was observed between ploidy and body size among S. alburnoides biotypes. Significant variation in body shape was found between S. carolitertii and S. alburnoides, and between tetraploids and the other biotypes. These differences in biotypes may reduce resource competition, highlighting the potential importance of resource availability favouring one biotype over another. In S. alburnoides, the adaptation to different trophic niches through modification of trophic morphology, body shapes, and feeding behaviour, may result from an increase in ploidy and genomic constitution. This adaptation may account also for the formation and maintenance of this nonsexual complex.

Introduction

Evolution means change and one of the most perplexing issues in evolutionary biology is how evolutionary processes produce phenotypic change (Schlichting & Pigliucci, 1998). Hybridization can create new gene combinations, which in turn may create key innovations that originate new phenotypes (Gilbert, 2003; Rieseberg et al., 2003; Grant et al., 2004), which in turn may result in the formation of an entirely new species (Barton, 2001; Seehausen, 2004; Chapman & Burke, 2007). It is widely accepted that there is a close link between hybridization, polyploidy and asexuality (Schultz, 1969), where hybridization seems to be the starting point for the other two phenomena (Bell, 1982; Dawley & Bogart, 1989). Due to their hybrid origin, allopolyploids may be characterized by broad adaptive phenotypic plasticity that may expand the range of phenotypic expression. This expansion of phenotypic expression may also expand the environmental tolerance of allopolyploids beyond that of parental taxa. Moreover, this expansion might avoid intraspecific competition by the colonization of novel habitats (Baack, 2005; Kim et al., 2008). Furthermore, polyploid genomes may undergo rapid changes in their structure and function via genetic and epigenetic changes (Levy & Feldman, 2002; Osborn et al., 2003; Chen, 2007). Changes in genome structure typically have immediate effects on phenotype that often reflect ecological and behavioural differences that directly affect the organism’s fitness (Mable, 2003; Otto, 2007; Gilbert & Epel, 2008).

Squalius alburnoides, a small endemic cyprinid fish that inhabits rivers from the Atlantic slope of the Iberian Peninsula, is among the most complex vertebrate polyploid systems of hybrid origin known. Moreover, S. alburnoides provides a good model for the consequences of polyploidy and hybridization in the phenotype. The hybrid origin and the different ploidy levels of S. alburnoides could be an important source of phenotypic plasticity. The S. alburnoides hybrid complex shows asexual reproduction mechanisms: gynogenesis, hybridogenesis and meiotic hybridogenesis, in which individuals of other Squalius species act as sperm donors (Squalius pyrenaicus and S. aradensis in the southern river basins, and Squalius carolitertii in the northern basins), and contains diploid, triploid and rare tetraploid individuals (reviewed in Alves et al., 2001; Sousa-Santos et al., 2006). However, the Douro River basin (a northern Iberian basin) has populations mainly composed by symmetric allotetraploids that exhibit sexual reproduction (Cunha et al., 2008), thus without the constraints of asexuality.

The S. alburnoides complex was characterized based on morphological and cytogenetical analyses, and two morphological forms were defined mainly by differences in the number of gillrakers and pharyngeal teeth: one with a pharyngeal tooth formula of 5–5 and 12–17 gillrakers, mainly represented by triploid females (form I); and another with a pharyngeal tooth formula of 5–4 (rarely 5–5) and with longer, more numerous gillrakers (17–26), represented only by diploid males (form II) (Collares-Pereira, 1983, 1984, 1985). Recent identification of genome composition using molecular markers showed that form II is exclusive to nuclear nonhybrid males (AA); a biotype that has never been found in Douro populations (Alves et al., 1997, 1999; Carmona et al., 1997). The other morphological form should thus be represented by the other biotypes (CA/PA, CAA/PAA, CCA/PPA and CCAA/PPAA). However, the predominant biotypes in nature are triploids CAA, across the distribution range for S. carolitertii, and PAA, across the range for S. pyrenaicus (Pala & Coelho, 2005; Crespo-López et al., 2007; C. Cunha, I. Doadrio, J. Abrantes & M. Coelho, unpublished data).

Spatial segregation among different biotypes of asexual freshwater fish is well documented as a mechanism to reduce competition (Vrijenhoek, 1984a, b; Schenck & Vrijenhoek, 1989; Echelle & Echelle, 1997). Previous results suggest that the sexual tetraploid populations of S. alburnoides reduce competition with hybrid diploids (CA) and triploids (CAA) by choosing different habitats (Cunha et al., 2008). The habitats used by these tetraploid populations are characterized by shallow waters, gravel and cobble beds, and fast streams, whereas hybrid diploids and triploids (mainly CAA females) occur in middle reaches of rivers with abundant riparian vegetation and moderate current, with rocky or gravel bottoms (Carmona, 1997; Cunha et al., 2008).

The body shape of fishes is thought to be of particular ecological and evolutionary relevance (Klingenberg et al., 2003). If groups of individuals of the same species have different niches, then that may have a substantial effect on the evolution of their body shape and size. Recent studies have also shown that allometry, or covariation in size and shape as a consequence of variation in growth, significantly affects overall morphology (Klingenberg, 1998; Zelditch et al., 2004; Rincón et al., 2007). Moreover, size differences between allopatric populations may reflect environmental differences to which the populations have been adapted. Therefore, morphometric studies may provide useful information on the evolution of fishes (e.g. Schluter, 1993; Walker, 1997; Rüber & Adams, 2001; Hulsey & Wainwright, 2002).

Traditional morphometric measures provide relatively little information on shape and some of the information it provides is fairly ambiguous (Zelditch et al., 2004). Consequently, several new methods and statistical procedures, such as geometric morphometrics, have been developed to characterize and analyse body shape (Rohlf & Marcus, 1993). Geometric morphometry (GM) techniques are based on the analysis of bi- or tri-dimensional (3D) landmark coordinates, which may be used to statistically explore variation in the spatial (geometric) relationships of morphological structures (Rohlf & Marcus, 1993). Another important feature of GM is that the results of these statistical analyses can be visualized by powerful diagrams based on the thin-plate spline (Bookstein, 1989, 1991).

With this study of geometric morphometrics, we wanted to understand whether phenotypic changes in body shape and size exist in the Squalius alburnoides hybrid complex, among its different ploidy levels and genomic constitutions. Furthermore, we would like to know whether such morphological changes would allow the exploitation of different habitats, thus avoiding competition. We used geometric morphometrics to address the following questions: (i) Are there any morphological size and shape differences between members of the S. alburnoides complex and their sperm donor S. carolitertii?; (ii) Are there any size and shape differences among ploidy biotypes of the S. alburnoides complex?; (iii) How does polyploidization affect variation in size and shape?; and (iv) How does allometry affect variations in shape? To answer these questions, we analysed S. alburnoides populations from the Douro basin with CA, CAA, CCA and CCAA biotypes. The C genome is provided by the sperm donor S. carolitertii and the A genome (the clonal part) came from the extinct paternal ancestor. The answers to the proposed questions clarify the effects that an increase in ploidy and the origin of a hybrid might have had on the phenotypic variation and ecology of this complex. Moreover, we may be able to model how many more morphological types could be generated, and how these variations can be produced and which structures are most likely to be affected.

Materials and methods

Sample collection and ploidy level determination

A total of 116 specimens were sampled and captured by single-pass electric fishing (300 V, 3–4 A, DC) along an 80-m section, with stop nets to close the electric fishing area. We sampled five stations along the Douro basin, all of which located in the northern distribution area of the S. alburnoides complex (see Fig. 1). The sample stations of Manzanas, Rabaçal and Tâmega rivers correspond to the type of habitat where hybrid diploids and triploids (CAA) are commonly found, whereas Lodeiro and Paiva rivers correspond to the type of habitat where sexual tetraploids are found. For all specimens, the ploidy level was determined through flow cytometry as described elsewhere (Dawley & Goddard, 1988; Collares-Pereira & Moreira da Costa, 1999).

Figure 1.

 Map of the Iberian Peninsula showing collection sites and distribution area of the Squalius alburnoides (represented in grey) and Squalius carolitertii (represented by marbled texture). 1 – Manzanas River; 2 – Rabaçal River; 3 – Tâmega River; 4 – Lodeiro River; 5 – Paiva River.

Genotyping

We used seven microsatellite loci: four isolated from the genus Luxilus (Turner et al., 2004) and three from the genus Squalius (Mesquita et al., 2003; Pala & Coelho, 2005), according to the procedures described in Pala & Coelho (2005). These loci were used to identify the different hybrid biotypes of S. alburnoides (CA, CAA, CCA and CCAA) (Pala & Coelho, 2005; Crespo-López et al., 2006, 2007). The methodology used to identify the genome copy number of the intergeneric hybrid complex (e.g. CAA, CA) consisted of the identification of alleles specific to the A and C genomes, in heterozygous microsatellite loci, and in the use of flow cytometry. Previous work has shown the accuracy of this method for S. alburnoides (Pala & Coelho, 2005; Crespo-López et al., 2007; Cunha et al., 2008), and for other asexual complexes (Christiansen, 2005; Lampert et al., 2005; Ramsden et al., 2006).

Geometric morphometrics

Digital photos of the specimens were taken with a ruler included to provide the scale. TpsDig (Rohlf, 2003) was then used to digitize the x- and y-coordinates of 16 homologue landmarks (lms). The landmarks were measured on each specimen as illustrated in Fig. 2. Shape data were obtained by several procedures: Generalized Procrustes Analysis (GPA) (Rohlf & Slice, 1990), thin-plate splines (Bookstein, 1991), and principal components analysis (PCA) of Procrustes residuals (O’Higgins, 2000). GPA uses iterative translation, rotation, and scaling to unit centroid size (Rohlf & Slice, 1990; Dryden & Mardia, 1998) until the Procrustes distance (d) among homologue landmarks between the mean and each specimen is minimized. The Procrustes distance is defined as the square root of the summed squared distances between two or more Procrustes-registered landmark configurations (Bookstein, 1991). After the scaling procedure, a scaling factor, centroid size (CS) is obtained, which is the square root of the summed distances between the centroid (centre of gravity of the object) and each landmark (Bookstein, 1991). In the absence of allometry, centroid size is the only size measurement that is uncorrelated with shape (Bookstein, 1991, 1996). Shape data were obtained by PCA of Procrustes shape coordinates for form space analysis (Mitteroecker et al., 2004, 2005; Bastir et al., 2007) and using thin-plate splines (Bookstein, 1989, 1991) to obtain partial warps and uniform component scores (Rohlf, 1993; Rohlf et al., 1996).

Figure 2.

 Positions of 16 landmarks used in this study. All landmarks were digitized from the left lateral view of the fish.

Statistical analyses

Shape variation and group differences were assessed by several methods. The hypothesis of species specific mean shape differences was addressed using manova permutation tests (n = 999) and conducted using Morpheus et al. software (Slice, 1998). Permutation tests are preferable to parametric statistics because no normal distribution is required, sample sizes may differ, and because landmarks with different degrees of freedom (type1–type3, Bookstein, 1991) may be combined. Differences in centroid size were determined by anova and post hoc comparisons.

Allometric variation was analysed by form space analysis (Mitteroecker et al., 2004, 2005), which uses Procrustes registration but re-introduces the natural logarithm of CS (lnCS) into the data matrix (which now comprises both shape coordinates and lnCS) before PCA is carried out (Mitteroecker et al., 2004). The natural logarithm of CS usually has the largest variance in any column of this matrix, and the first PC of the size-shape (= form) distribution is often closely aligned with size. In this situation, this method is well suited to compare scaling trajectories (Mitteroecker et al., 2004, 2005; Bastir et al., 2007) and 3D PC-plots well illustrate the similarity or dissimilarity of allometric trajectories. This analysis was performed with the program morphologika2, version 2.5 (O’Higgins, 2000).

Finally, to assess phenetic relationships between different genetic groups, we did a cluster analysis using NTSysPC 2.1 (Rohlf, 1997) on a Procrustes distance matrix of species mean shape configurations. These results are represented as an unweighted pair-group method with arithmetic mean (UPGMA)-tree. In addition, a PCA of the shape space of the mean shape configurations was performed and the results projected on the first three PC axes, and a minimum spanning tree calculated to show the nearest neighbour according to the Procrustes distance in full shape space.

Results

Overall morphology and genotyping

Overall morphological and meristic characters were used only to discriminate between S. carolitertii and S. alburnoides. Specimens of S. alburnoides and S. carolitertii were identified by their pharyngeal tooth formulae, number of gillrakers, number of branched dorsal fin rays, the presence or absence of a black line around the dorsal fin base, and the presence or absence of a pigmented lateral band. However, based on this set of morphological characters, eight individuals of S. alburnoides were misidentified as S. carolitertii rather than S. alburnoides. These misidentified individuals had eight branched dorsal fin rays, 5 + 2 pharyngeal teeth in two rows and no black line around the dorsal fin base and no pigmented lateral band, instead of seven branched dorsal fin rays, five pharyngeal teeth in one row and black line around the dorsal fin base (albeit not always observed) and a pigmented lateral band. Cytometry analysis showed that these individuals were all triploids, and microsatellites showed that they were all CCA. Therefore, in some cases the morphological approach to distinguish the sperm donor S. carolitertii from the S. alburnoides complex, might lead to misclassifications and might provide misleading estimations of their frequencies in natural populations. Moreover, among the S. alburnoides hybrids the classic morphological approach failed to discriminate among the different biotypes (CA, CAA, CCA and CCAA), and in these cases cytometry and microsatellite markers were required for their discrimination. We only identified tetraploids (CCAA) and triploids (CCA) at the Lodeiro and Paiva stations. In the remaining sample stations, we only identified hybrid diploids and triploids (mainly CAA females).

Geometric morphometrics analysis

Size analysis

Size analysis (anova) showed highly significant differences among the different genetic groups (F4,111 = 28.058, < 0.00000). Table 1 provides descriptive statistics for centroid size, Fig. 3 shows the results of the anova, and Table 2 the post hoc comparisons (MS = 267.83, d.f. = 111.00). These data showed that the largest specimens correspond to the diploid sperm donor group S. carolitertii (CC) (CS = 110.8 mm), whereas the smallest sizes were recorded in the tetraploid group of the S. alburnoides hybrid complex (CS = 72.2 mm). Specimens in the triploidC (CCA) group were larger than all the other biotypes, and post hoc comparisons (Duncan’s test) showed that this size difference was significant compared to the diploid and tetraploid groups. When compared to the triploidA (CAA) specimens, the larger size of the triploidC group individuals was only close to being statistically significant (P = 0.06).

Table 1.   Descriptive statistics of centroid size.
GroupnMean CSSE95−95+
Diploid772.926.1960.6685.18
TriploidA3377.022.8571.3882.67
TriploidC1088.355.1878.0998.60
Tetraploid3472.202.8166.6477.76
Sperm donor32110.802.89105.07116.53
Figure 3.

anova of centroid size. Circles indicate means and vertical bars the 95% confidence intervals.

Table 2. Post hoc mean comparisons (P-values; Duncan’s test, significant P-values in bold).
GroupsDiploidTriploidATriploidCTetraploidSperm donor
Diploid 0.4941860.0150920.9044450.000046
TriploidA0.494186 0.0609330.4521180.000057
TriploidC0.0150920.060933 0.0133580.000374
Tetraploid0.9044450.4521180.013358 0.000029
Sperm donor0.0000460.0000570.0003740.000029 

Shape analysis

Mean shape differences were addressed by performing two manovas: one simultaneously comparing all groups (P < 0.001), and the other conducting pairwise comparisons (Table 3). This analysis showed a significant difference in shape of the sperm donor group (S. carolitertii) when compared to all members of the S. alburnoides hybrid complex. The tetraploid group differed significantly in shape from the other groups of the S. alburnoides complex (Fig. 4). These analyses showed that the sperm donor group (S. carolitertii) differed in vertical body height (lms 3, 10, 4, 9) and orientation of the mouth (lms 1, 14), which pointed downward (Fig. 4). Furthermore, the ventral fin (lm 10) was shifted posteriorly and the insertion of the anal fin (lms 8, 9) was relatively expanded compared to that of the S. alburnoides biotypes.

Table 3. manova mean shape permutation tests (n = 999), pairwise comparisons (P-values; permutation test, significant P-values in bold).
GroupsDiploidTriploidATriploidCTetraploidSperm donor
Diploid00.170.570.0470.001
TriploidA 00.090.0240.001
TriploidC  00.0090.001
Tetraploid   00.001
Sperm donor    0
Figure 4.

 Mean shape comparisons represented as thin-plate splines-transformation grids. Significant differences are indicated in bold. Different ploidy levels of Squalius alburnoides: 2n – diploids, 3na – triploids CAA, 3nc – triploids CCA, 4n – tetrapolids. SD, sperm donor.

The tetraploids (CCAA) differed more from the diploid (CA) and triploidC than from the triploidA biotypes. Compared to the diploids (CA), the snout pointed upward (lms 1, 14), the insertion of the ventral fin (lm 10) was elevated and shifted anteriorly whereas the dorsal fin (lms 3, 4) was shifted posteriorly.

Tetraploids were characterized by having a lowered posterior corner of the operculum (lm 12), a longer anal-caudal peduncle (lms 7, 8), and a dorsal fin origin (lm 3) posterior to the insertion of the pelvic fin, when compared with the triploidA individuals. Furthermore, in tetraploids the orientation of the mouth (lms 1, 14) points slightly downward (Fig. 4).

Transformation of triploidC to tetraploids reveals a clearly upward-pointing mouth (lms 1, 14; Fig. 4). Body height (lms 3, 10, 4, 9) at the central point between dorsal and ventral fins was relatively reduced and these fins are sheared against each other, that is, the dorsal fin is shifted posteriorly, the ventral anteriorly. The difference between the biotypes triploidA and triploidC was not significant (P = 0.09), but there was a trend towards different mouth orientations and body heights.

Allometry

Principal components analysis of form space suggested no differences in overall allometric patterns (Table 4). The first three PCs account for almost 99% of the variance. Figure 5 shows roughly parallel trajectories and similar orientations of scatter plots both within hybrids and among the hybrids and the parental group. When compared to the hybrids, plots for the S. carolitertii group show a larger size but a similar scatter orientation. Larger specimens were relatively taller (increased body height) and showed a relatively shorter snout than smaller specimens in both intra- and inter-group comparisons.

Table 4.   Descriptive statistics for the PCA (eigenvalues, per cent variance, cumulative per cent variance).
Principal componentEigenvalue% VarianceCumulative % variance
PC16.06E-029.82E-019.82E-01
PC22.41E-043.91E-039.86E-01
PC32.15E-043.48E-039.89E-01
PC41.07E-041.74E-039.91E-01
PC58.50E-051.38E-039.92E-01
Figure 5.

 Principal components analysis (PCA) of form space. Note the overall parallel orientations of the scatters. Inset TPS-grids show allometric changes common to all groups. Different ploidy levels of Squalius alburnoides: 2n – diploids, 3nA – triploids CAA, 3nC – triploids CCA, 4n – tetrapolids.

Phenetic distances

Procrustes distances between mean shapes appear in the UPGMA-tree in Fig. 6. This tree shows the tetraploid and triploidA biotypes of S. alburnoides clustering far from the S. carolitertii group. The other cluster is less distant in shape to the S. carolitertii group and consists of diploids and triploidC biotypes. The projections of mean shapes onto the first three PC axes (97% of variance) and the minimum spanning tree (Fig. 7) show the triploidC shape connecting to all others except to the tetraploids.

Figure 6.

 Unweighted pair-group method with arithmetic mean tree based on the Procrustes distance matrix. Different ploidy levels of Squalius alburnoides: 2n – diploids, 3na – triploids CAA, 3nc – triploids CCA, 4n – tetrapolids.

Figure 7.

 Principal components analysis (PCA) of mean shapes (shape space). The shape projections are linked by a minimum spanning tree based on full shape space. Different ploidy levels of Squalius alburnoides: 2n – diploids, 3na – triploids CAA, 3nc – triploids CCA, 4n – tetrapolids.

Discussion

In this study, we examined variation in body size and shape among biotypes of the S. alburnoides complex. Our results showed the importance that hybridization, ploidy level and genomic constitution might have in the response to different environments by producing different phenotypes, reducing the competitive interaction among biotypes.

Size effect

Ecological traits may have interacting ontogenetic and evolutionary effects on body size (Walker, 1997 and references therein), and size differences may also have a genetic component (Stearns, 1992; Roff, 1997). Polyploidy in fishes has been associated with traits such as large body size, fast growth rates, a long life span and increased ecological adaptability (Uyeno & Smith, 1972; Schultz, 1980). Often, polyploidy has been associated with an increased body size like in Euchlanis dilata (Walsh & Zhang, 1992). However, in Poeciliopsis monacha-lucida a negative relationship between ploidy level and body size has been observed, where triploids have smaller body sizes than diploids (Schultz, 1982). Squalius alburnoides meets none of these previously reported descriptions. As in many other polyploids, especially vertebrates, we found no relationship between ploidy level and body size, possibly due to an overall decrease in cell number and in cell surface area to volume ratio (Otto & Whitton, 2000; Leggatt & Iwama, 2003; Otto, 2007).

In our analysis, tetraploids were the smallest in size, whereas the sperm donor, S. carolitertii, showed the greatest centroid size due to its much larger size, measuring up to 25 cm (Doadrio, 2001). It has been suggested that the paternal ancestor of S. alburnoides was a small cyprinid similar to the reconstituted nuclear diploid nonhybrid males (AA), and with a body size similar to that of Anaecypris hispanica (Alves et al., 2001; Gromicho et al., 2006). Anaecypris hispanica is thought to be related to the paternal ancestor, and does not reach a standard length of 7 cm (Doadrio, 2001). Therefore, it is possible that the hybrid biotypes show an intermediate size between those of their parental species, and indeed the S. alburnoides individuals have an average size of 13 cm (Doadrio, 2001).

The smaller body size of diploids and tetraploids allows them, not only to occupy less space, but also to reduce turning resistance by reducing body mass and entrained water distance from the turning axis (Webb, 1982). Both of these features are useful for the type of habitat they occupy, particularly for the tetraploids that inhabit shallow waters with fast streamflow.

The body size of the S. alburnoides biotypes only differed significantly when one compared the triploidC (CCA) biotype with all the others but the triploidA. A dosage effect could be the most plausible explanation for the differences found. Indeed, the number of copies of each genome, especially in odd-number ploidies, may have consequences on gene regulatory processes, where the balance between the expression of paternal and maternal genes is crucial (Chen, 2007; Mable, 2007; Pala et al., 2008). The C genome duplication makes triploidsC more similar to the sperm donor (S. carolitertii), whereas the smaller triploidA biotypes may reflect the characteristics of the paternal ancestor due to duplication of the A genome. Recent studies have, in fact, demonstrated the importance of gene regulation in the morphological changes (Carroll, 2000; Campbell & Barwick, 2006; Jeong et al., 2008).

Shape effect

Our results reveal that the variation in body shape in the S. alburnoides complex corresponds to two main morphological shifts: (i) orientation of the mouth and (ii) vertical body height. The sperm donor and the tetraploids showed significant variation in body shape. Variation in body shape may indicate ecological and behavioural differences (Webb, 1984; Klingenberg et al., 2003), but may reflect also a phylogenetic signal. Squalius carolitertii displayed a greater body height, shorter and truncated caudal peduncle, larger dorsal fin and displacement of the ventral fin insertion (Fig. 4). Because other members of Squalius share these features, we interpreted them as plesiomorphic. The S. alburnoides, however, does not share those traits (Doadrio, 2001; Miranda & Escala, 2005 and this study). Reflecting their hybrid origin, they have a more slender body and narrower caudal peduncle, more reminiscent of the Alburninae clade (Zardoya & Doadrio, 1999). These differences might have enhanced the survival of the S. alburnoides complex by allowing them to exploit new resources and avoid competition between biotypes and their sperm donor.

Our data further show that tetraploids are more heterogeneous than the other biotypes, suggesting that this may be a result of their sexual reproduction. The diploidization of S. alburnoides has generated a new incipient polyploid species through chromosome pairing and segregation (Cunha et al., 2008) and allowed rearrangements in noncoding genomic DNA, homologous recombination, sequence elimination and changes in DNA methylation (Osborn et al., 2003; Levy & Feldman, 2004). Epigenetic changes and the regulation of gene expression allow the coexistence of two different genomes in the same nucleus (Kashkush et al., 2002). These genome duplications and changes in genomic structure confer new opportunities for adaptation. For instance, the more fusiform body shape and the smaller body size of the tetraploids provide reduced drag and lower energy requirements for maintaining their position in flowing water (Plaut, 2000; Langerhans et al., 2003). These features suggest an adaptation to fast flowing streams. Moreover, among hybrids the significant shape differences of tetraploids may be associated with their reproductive isolation from other biotypes through habitat selection, assortative mating and apparently nonviable embryos (Cunha et al., 2008). The sexual reproduction of the tetraploids has originated an emergent polyploid neospecies.

The relative orientation of the mouth (e.g. superior vs. terminal vs. inferior) shows at what water depth feeding normally occurs (Winemiller, 1991). In contrast with S. alburnoides, the subterminal mouth of S. carolitertii suggests that they never feed on the surface, but instead consume arthropods and the fry of other fishes (see Doadrio, 2001; Kottelat & Freyhof, 2007). The observed differences in mouth position among tetraploids, diploids and triploidsA biotypes also reflect different foraging abilities. Gomes-Ferreira et al. (2005) reported that triploid hybrids (mainly triploidsA) and diploid hybrid females select different food types, suggesting that they feed at different water depths thus reducing competition. Diploid hybrid females tend to feed near the bottom and submerged vegetation, whereas triploid hybrid females feed mainly in the middle of the water column. In this study, differences in mouth position shown by tetraploids (Fig. 4) (slightly more upward pointing than in diploids and slightly more downward pointing than in triploidsA) most likely reflect differences in foraging mode and diet composition. TriploidsC and the symmetric tetraploids live in sympatry for these triploids are consequence of sporadic crosses between symmetric tetraploid females and S. carolitertii males (Cunha et al., 2008). However, the slight differentiation found in their relative mouth orientations (Fig. 4) shows that tetraploids might normally feed higher in the water column than triploidsC, thus reflecting resource partitioning. The body shape differences observed among the biotypes might be the result of the combined effects of biotic interactions and habitat differentiation to reduce competition.

There were clear morphological differences between the sperm donor and S. alburnoides specimens (Figs 6 and 7). The morphological plasticity found in S. alburnoides biotypes, and consequently their greater ecological adaptability, might result from the ploidy increase and from its genomic constitution. Such plasticity seems to arise from gene sequence duplication through heterosis and gene redundancy, which could explain the improved performance of S. alburnoides. The improved performance might be a consequence of the ability to diversify gene functions by altering the number of redundant copies of relevant genes. In S. alburnoides, the adaptation to different trophic niches through the modification of trophic morphology, body shapes and feeding behaviour, might account for the formation and maintenance of this nonsexual complex. Moreover, the relationship found among morphology, biotypes and ecology could be the result of directional natural selection acting on instantaneously generated adaptive morphs. The adaptability of this group could also arise from the many levels of ploidy and modes of reproduction. The broader implications of these findings emphasize the value of geometric morphometrics as both an exploratory and analytical approach. Our results suggest that in this complex an ecological study of resource availability could test the hypothesis that resource availability favours one biotype over the others.

Acknowledgments

We thank A. Perdices for assistance in sample collection and Ana Filipa Filipe for permission to publish the scientific illustration of Squalius alburnoides. We thank Prof. J. Amaral, Prof. D. Brooks and A. Burton for the English revision of the manuscript and helpful comments. We thank A. Perdices, I. Pala, C. Luís, and two anonymous reviewers for helpful comments. This study was supported by POCTI/BSE40868/2001, REM2001-0662/GLO 07 M/0109/2002, CGL2004-00077/BOS and by a PhD fellowship from the Fundação para a Ciência e a Tecnologia given to C. Cunha (SFRH/BD/8637/2002).

Ancillary