Morphological variation in size and shape of organisms has physiological, ecological, and evolutionary relevance (Bookstein et al. 1985; Peters 1986; Coyne and Orr 2004). This variation, induced by genetic and environmental factors, can modify growth processes and be a key factor in species survival (Barlow 1961; Somers 1986). Thus, the analysis of intraspecific variation, at genetic and morphological levels, is crucial in the elucidation of evolutionary processes. In this context, an important step is to identify evolutionary units in a continuous geographic space (Patton and da Silva 1997) at both genetic and morphological levels (dos Reis et al. 2002). Delimitation of evolutionary units is not only a first step in the study of speciation events but also crucial when conservation management of endangered species is required (Ryder 1986; Moritz 1994, 2002). In this sense, endemic taxa are of particular importance as they are considered among the most vulnerable. Hence, delimitation of evolutionary units is interesting when the group under consideration has an intriguing origin (Meyer and Zardoya 2003), a unique morphology (Pritchard 1992), and worldwide conservation problems (Turtle Conservation Fund 2002).

The freshwater turtle genus Hydromedusa is endemic to the Neotropical region. It inhabits water bodies of the Paraná–La Plata basin and coastal streams of Brazil and Uruguay (Iverson 1992; Cabrera 1998). The genus appeared in the fossil record 56 million years ago (de la Fuente and Bona 2002) and consists of 2 extant species. Hydromedusa maximiliani is restricted to the Atlantic Rainforest of Brazil (Iverson 1992), whereas Hydromedusa tectifera presents a wider distribution, from Santiago del Estero in Argentina to the State of Minas Gerais in Brazil (Iverson 1992; Cabrera 1998; de Sousa and Novelli 2009). When they occur in sympatry, H. maximiliani inhabits flowing streams at altitudes above 600 m, whereas H. tectifera occurs in the lowlands (Souza 2005).

The low diversity and the endemic nature of this genus make these species a focus of conservation interest. Analyses of the genetic variation of H. maximiliani have shown that each stream holds genetically differentiated populations (Souza et al. 2002a). Furthermore, Souza et al. (2002b) showed that there is no genetic flow between those populations and that their genetic structure follows a hierarchical pattern of hydrogeographical basins. This coincides with the overall low vagility shown by freshwater turtles (Obbard and Brooks 1981; Kaufmann 1995; Jones 1996). However, there is no information regarding the morphological or genetic variation in the sister species of the genus, H. tectifera.

Hydromedusa tectifera is mainly distributed over 3 major basins, the Uruguay River, the Parana River, and the Los Patos-Merin Lagoon; in addition, it occurs in all minor Atlantic coastal rivers within its area of distribution (Cabrera 1998) (Fig. 1). Some of these basins are currently isolated by the broad estuary of the Rio de la Plata. However, during the marine transgressions that occurred in the Miocene, these basins may have alternated between a contact zone situation and being even more isolated than today because of ingression or regression of marine waters (Sprechmann 1980). If so, reduction of genetic flow among basins caused by the species' low vagility should favor the action of disruptive processes that promotes genetic structuring and concomitant morphological variation.

View Figure

• Download as Powerpoint
• Download Figure

The goal of this study was to analyze the patterns of morphological variation of H. tectifera in a hydrogeographical context. For this, analyzed individuals were grouped according to their watershed basin of origin, which constituted the geographical units (i.e., the Paraná, Salado, and Uruguay basins).

METHODS

All 44 specimens of H. tectifera deposited in the Reptile Collection of Facultad de Ciencias (ZVC-R), Museo de Ciencias Naturales de Argentina Bernardino Rivadavia (MACN), and Museo de Ciencias Naturales de La Plata (MLP) were used in this work. The distribution of the analyzed individuals comprised the Salado, lower Uruguay, and lower and middle Paraná River basins (Fig. 1). Morphological variation was analyzed through linear measurements and geometric morphometry.

Linear Measurements

We recorded 15 straight-line measurements of the carapace for 34 individuals: maximum carapace length (CL), maximum carapace width (CW), head length from the nose tip to cranial posterior border (HL), head width (HW), midline length of nuchal scute (NL), maximum width of nuchal scute (NW), midline length of the first vertebral scute (VL), maximum width of first vertebral scute (VW), midline length of first marginal scute (ML), maximum width of first marginal scute (MW), midline length of the last marginal scute (CAL), maximum width of the last marginal scute (CAW), curvilinear length of carapace (CCL), and width of anterior (AB) and posterior borders (PB) of the first vertebral scute. Measurements were taken with a dial caliper to the nearest 0.1 mm.

Landmark Analysis

A total of 44 individuals were analyzed with geometrical morphometry, and data were collected with an Epson Photo PC 3000Z model GT90A digital camera. Each specimen was photographed in dorsal view using a tripod. Images were digitized with TpsDig software (Rohlf 2003a). Digitized landmarks (12) corresponded to central and nuchal scutes of carapace (Fig. 2). Marginal landmarks were not included because shape changes associated with these anatomical points are strongly affected by carapace depth. To avoid linear dependencies among shape variables that cause statistical difficulties (Bookstein 1996), we recorded landmarks from only one side of the carapace. Landmark configurations for all specimens were aligned by the generalized Procustes superimposition procedure (Bookstein 1991; Monteiro and Reis 1999). The Thin Plate Spline and uniform component approximation were used to project the specimens into a linear tangent space to perform a linear multivariate analysis of shape variation and covariation (Bookstein 1991).

View Figure

• Download as Powerpoint
• Download Figure

Data Analyses

There are 2 common problems in morphometric studies: size and allometry. The former may confound the interpretation of shape changes because morphometric variables usually carry both information about size and shape simultaneously (Zelditch et al. 2004). The latter is an important factor affecting shape variation ontogenetically (Berry and Shine 1980; Lindeman 1999; Shine 2005). For this reason, linear variables are usually transformed with a method that removes size and allometric effects. Hence, we transformed linear measurements into size-free new variables with a transformation that considers allometric effects (Lleonart et al. 2000) with the following equation:

where X₀ is the mean standard length. In this case, we chose straight carapace length, but any measure related with size can be used. The X represents the standard length of each specimen; Y is the variable to being transformed; b is the allometric coefficient of the variable Y with standard length obtained from a linear regression between the logarithms of Y and X; and Z is the new variable that we used in our statistical analyses. This procedure was performed for all linear variables.

Although geometric morphometry filters size prior to analyzing shape changes, the allometric effects could remain. Therefore, we tested the data for allometry by means of Multivariate Regression of Shape on standard length (Monteiro and Reis 1999) with the TpsRegr software (Rohlf 2003b). This was achieved using a permutation test with 1000 iterations simulating the null hypothesis of independence between size and shape by randomly exchanging the value for standard length among individuals (Good 1994). To analyze shape changes filtering allometric effects, we conducted linear regressions between morphometric variables and standard length; residuals of these regressions were used in multivariate analyses. To test basin identity, Discriminant Function Analyses (DFA) were performed independently on the linear measurements matrix and on the regression's residuals of morphometric variables.

RESULTS

Linear Measurements

The Discriminant Function Analysis of transformed linear variables of the carapace resulted in a clear differentiation among the 3 geographic basins (Wilks' Lambda p < 0.05, λ  =  0.11) (Fig. 3). The squared Mahalanobis distance for carapace variables was significant between individuals from the Paraná basin and the other 2 groups with p < 0.05. Nuchal, marginal, and caudal widths were the variables that influenced the discrimination among groups at p < 0.05 (Table 1). P-values obtained in other variables such as upper border of first central, carapace, and caudal width were not significant but marginal.

View Figure

• Download as Powerpoint
• Download Figure

Table 1 Wilks' Lambdas and p-values of linear variables in DFA for carapacial landmarks from Hydromedusa tectifera. Bold values show significant p-level, whereas asterisks indicate variables with a marginal p-level.

View Fullscreen

Landmark Analysis

The set of central landmarks had significant Wilks' Lambda value (p < 0.05, λ  =  0.016) among groups. We found significant values for the squared Mahalanobis distances between individuals from the Uruguay River basin and the Salado River basin, whereas the p-level between Uruguay and Paraná River basins was marginal (Fig. 4). However, individuals from the Paraná River did not demonstrate differences with individuals from the Salado River.

View Figure

• Download as Powerpoint
• Download Figure

We used scores of Root 1 of the DFA, which discriminate between Uruguay and Salado River basins, to analyze shape changes, and we used the Root 2 to visualize shape of individuals of Salado River. The negative values of the Root 1 were associated with an expansion at the level of the nuchal scute and contraction at the level of central scutes, whereas positive values were associated with changes in the opposite direction (Fig. 5). Shape changes associated with size also corresponded to an expansion of the nuchal region for small sizes and a compression of the central region for larger sizes, with more magnitude than basin effect (Fig. 5).

View Figure

• Download as Powerpoint
• Download Figure

DISCUSSION

We report morphological variation of carapace shape in H. tectifera associated with developmental and historical factors. Our results support the hypothesis that at least part of the morphological variation in H. tectifera is associated with variation between basins; possibly as a result of reduced gene flow between their populations. This was supported by both linear and geometric morphometry analyses. However, different analyses grouped individuals from different basins. Linear data discriminated individuals of Paraná basin from those of the other 2 basins, whereas morphometric variables only discriminated individuals from the Uruguay basin from those of the Salado River basin. Disparity between methods in basin discrimination could be related to differences in their sensitivity to shape variation (Monteiro and Reis 1999) and the region of body shape that they recorded. Also, these results show that these methods could be complementary and should be used together. Moreover, both methods demonstrate that most differences observed correspond to the anterior region of the carapace (nuchal and first central scutes). In geometric morphological data, negative values of the Root 1 are associated with individuals from the Uruguay River basin. Shape changes corresponded to an expansion of the nuchal scute and a compression of the central scutes (Fig. 5). The opposite shape changes are presented in the positive values of the Root 1, associated with individuals from the Paraná River basin.

Because individuals considered in this study come from a discontinuous range, it is not possible to affirm that the variation found is a consequence of a continuous geographic pattern that shows isolation by distance. However, in this study, individuals from the same basin, but from distant locations (i.e., Paraná), are more similar to each other than to individuals from different basins (i.e., Paraná and Uruguay) that are geographically closer (Figs. 3 and 4). Thus, morphological differences between Paraná and Uruguay basins are not associated with geographic distance, and they might be considered different evolutionary units (Karl and Bowen 1999; Roman et al. 1999). Hydromedusa tectifera inhabits streams and rivers; therefore, geographic distance should not restrict gene flow among populations as in terrestrial organisms. In aquatic organisms, migration to lower portions of the basins is more likely to occur than is dispersal between basins.

Basins analyzed here are in close geographic proximity, at least in their lower sections. Thus, the recovered morphological differences among populations could reflect past events that limited gene flow among them. During the last 20 million years, the sea has progressed into the South American continent through the Salado and Paraná Rivers (Sprechmann 1980), probably as a consequence of increments in sea level caused by melting of the Antarctic Ice Sheet (Jansen et al. 2007). The more extensive of these marine transgressions lasted from 15 to 7 million years ago. Consequently, a large sea called the “Mar Entrerriense” separated southern Buenos Aires Province from central Argentina (Martínez and del Rio 2002), and the entire territory of Entre Rios Province was under water. During the Miocene and the whole Quaternary period, there were several transgressions of minor magnitude. These transgressions may have separated the populations and limited the gene flow between the basins, resulting in the pattern of differentiation currently observed. Morphological and genetic differentiation as well as speciation of freshwater and estuarine fishes has already been proposed as a result of fluctuations of the sea level in this region (Beheregaray and Levy 2000; Beheregaray and Sunnucks 2001; Loureiro and García 2006).

Hydromedusa tectifera appeared in the fossil record 5 million years ago (de la Fuente and Bona 2002), and it has wider a distribution than other freshwater turtles from the region (Souza 2005). However, morphological differences found in this study are minor compared with morphological variation found in some other turtle species (Houseal et al. 1982; Valenzuela et al. 2004). The other extant species of the genus, H. maximiliani, presents 2 populations isolated for about 16 million years, with molecular evidence of gene flow absence between them (Souza et al. 2002a). However, the 2 populations are morphologically similar (Souza et al. 2002a). Our findings are similar to those of Souza et al. (2002b), suggesting that H. tectifera presents a slow differentiation rate, supporting the hypothesis of relatively low vagility and slow differentiation rate in this genus.

On the other hand, de la Fuente and Bona (2002) described a new fossil species of Hydromedusa from the Paleocene in Patagonia. Diagnostic characters proposed in their article were nuchal and first central scute width. The fossil species has the widest first central scute, whereas H. tectifera has the narrowest, and H. maximilliani has an intermediate width. It is particularly interesting to note that the present analysis found that this same central scute region harbors most of the variation of the carapace within H. tectifera populations.

Studies on phylogeography of freshwater turtles performed in other regions have found concordance in patterns of genetic structure between turtles and freshwater fish from the same region (Walker and Avise 1998). López et al. (2005) proposed that the ichthyogeographic Paranaense and Salado regions are independent from each other. However, other authors suggested that the lower Salado basin is totally independent from other biogeographic regions, whereas the upper Salado basin is strongly associated with the Paraná unit (MacDonagh 1934; Ringuelet et al. 1967; Ringuelet 1975). Our study shows ambiguous evidence given that different analyses showed different patterns. Analyses of the genetic structure of H. tectifera, as well as other freshwater turtles and fish of the region, are strongly needed to provide additional information that could resolve this controversy.

In conclusion, our results indicate that Hydromedusa possesses a tendency to vary a specific portion of its carapace in all species, extant and extinct. This is of particular interest considering that this is the region used as a diagnostic character for the genus. Also, this study provides further evidence that morphological variation in freshwater turtles is associated with hydrogeographic basins; hence, these animals could be important in assessing biogeographical regions because they have an intermediate vagility between strictly aquatic organisms such as fish and terrestrial organisms.