Study Species

The genus Triturus comprises four lineages: the small-bodied newts (snout-vent length [SVL] up to 60 mm), the big-bodied newts (SVL up to 110 mm), T. alpestris (SVL up to 70 mm), and T. vittatus (SVL up to 70 mm) (Fig. 1; Steinfartz et al., in press). Triturus newts are usually terrestrial for part of the year, but they enter still or, more rarely, slow-flowing water in spring to breed. They feed on a wide variety of small invertebrates, both in and out of the water, and some species occasionally prey on small fish and other amphibians (Arnold and Burton 1992). Although quantitative measurements of locomotor speed in the wild are lacking, newts use fast locomotor bouts to avoid predation both on land (Brodie 1977) and in the water (Griffiths 1996), suggesting that both aquatic and terrestrial locomotor speed are ecologically relevant functions in these animals.

Newts were collected between 1999 and 2002 at different natural sites in Europe. Specimens of one species (T. boscai) were obtained from the pet trade. To minimize possible effects of geographical variation, each species was represented by individuals from one population only. Because males exhibit a high degree of seasonal plasticity in secondary sexual characteristics (presence/absence of dorsal and caudal crests, webbing), we restricted our analyses to adult, nonreproductive females. Mean SVLs of the females used were similar to those reported in the literature (compiled in Malmgren 2001), thus we are confident that, in this respect, our samples are representative for the species studied. Although Triturus species exhibit sexual dimorphism in body size and shape, males follow similar (but less pronounced) patterns of among-species variation as females (Malmgren and Thollesson 1999; Malmgren 2001).

The newts were housed separately or in pairs in plastic recipients (50 × 30 × 25 cm³) holding 15 liter of tap water. Female newts are nonterritorial, and housing newts in pairs is consequently unlikely to influence their locomotor performance. Each container held a piece of styrofoam (10 × 15 cm) and several specimens of Java moss plants (Vesicularia dubyana) that served as hiding places. Kept in these conditions, the newts did not shift to the terrestrial phase, and remained semiaquatic. This phenomenon has also been observed in natural conditions (e.g., Andreone and Giacoma 1989). To reduce possible effects of prior thermal conditions, the newts were kept in a room at 18–22°C with a natural photoperiod for at least three months before testing (July– August). During the whole of this period, the animals experienced body temperatures within the preferred range of all the species considered (Gvoždík 2003, 2005, unpubl. data). All individuals received similar amounts of food items (earthworms, Tubifex worms, or fish meat), once or twice a week.

Morphometrics

We anesthetized the newts by immersing them in 0.08% phenoxyethanol prior to measurement. From each individual, we took the following measurements to the nearest 0.1mm, using digital calipers: SVL (from the tip of the snout to the posterior edge of the cloaca); axilla-groin distance; tail length (from the posterior edge of the cloaca to the tip of the tail) and height (at the highest point); head width (measured as the distance between the corners of the mouth); forelimb length (from the elbow to the tip of the longest finger); hindlimb length (from the groin to the tip of the longest toe). All limb measurements were performed on the left pair of limbs.

Locomotor Performance Measurements

We measured swimming and running speed of the newts in a linear racetrack (200 cm long × 10 cm wide). The central 100 cm of the racetrack was equipped with four pairs of infrared photocells, placed at 25 cm intervals. The sensitivity of the photocells and the positioning of the infrared beams were adjusted to assure that the newts invariably disrupted the beams with the tip of their snout. The photocells were connected to a computer that recorded the times (precision 0.001 sec) at which the newts passed by. During the swimming trials, the racetrack was filled with 3 cm of tap water; during the running trials, a surface of moist paper towel was used. Because optimal temperatures for locomotion may vary among species (e.g., van Berkum 1986; John-Alder et al. 1988; Navas 1996), all performance measurements were carried out in a temperature controlled room, and repeated at 20, 15, 25, 33, 10, and 30°C. The order in which individuals were subjected to terrestrial and aquatic locomotor trials was randomized. Because newts are mostly nocturnal, all experiments were completed between 18:00 and 24:00 hr. The newts were fed Tubifex worms the day following a trial and were allowed to rest for another two days before the next test. Body mass was monitored throughout the performance trials. Three individuals that lost considerable weight (>15% of their total body mass) were excluded from the analyses.

Prior to the speed trials, the newts were placed in receptacles containing 2 cm of water that were gradually (0.5°C min⁻¹) cooled or heated until they reached the test temperature. After another 30 min, an individual was placed on the racetrack, 10 cm in front of the first photocell. The newt was induced to move down the racetrack by tapping the tip of its tail with a soft paintbrush. Each individual was tested four times at each temperature, allowing a 60 min recovery period between two consecutive runs. Each trial was judged as “good” or “bad” (e.g., van Berkum and Tsuji 1987) and bad trials (<15%) were discarded. The fastest speed over any 25 cm interval was taken as a measure of maximum running or swimming speed. A previous study, following similar proceedings, demonstrated good short-term repeatability of locomotor performance in Triturus (L. Gvoždík, M. Puky, and M. Šugerková, unpubl. data). We will assume that the differences in laboratory measures of locomotor performance reliably reflect variation in speed under natural conditions. We do not actually have field data to support this assertion, but work on lizards (Irschick et al. 2005) has shown that differences between laboratory and field measures of sprint speed are minor in relatively slow species, and all our study species fall into this category. Because the species used in this study vary considerably in body size, and because it has been argued that body size may confound the effect of speed on escape success (Van Damme and Van Dooren 1999), relative speed (SVL s⁻¹) rather than absolute speed was used in the analyses.

Phylogeny

The topology of the tree used in the comparative analyses follows the results from a Bayesian analysis of mitochondrial DNA data (Steinfartz et al., in press, Fig. 1). This phylogenetic hypothesis was preferred because, unlike previous studies (reviewed by Zajc and Arntzen 1999), it is based on complete taxon sampling of all Triturus species. It provides fully resolved relationships, except for the T. cristatus group, which is considered a hard polytomy (J. W. Arntzen, pers. comm. 2006). Unfortunately, the information on branch lengths provided by Steinfartz et al. is incomplete and/or inconsistent with previously published estimates (e.g., by Oosterbroek and Arntzen 1992; Larson et al. 2003; Babik et al. 2005). We therefore set all branch lengths equal to one, or used computer-generated branch lengths in our comparative analyses (details below).

Phylogenetic Comparative Analyses

Since Felsenstein's (1985) seminal paper, several methods have become available to control for phylogeny in comparative data (Harvey and Pagel 1991; Martins and Hansen 1996). Because they are based on contrasting evolutionary assumptions, different phylogenetic techniques may yield different results, especially when the number of species involved is low (Martins et al. 2002). It has therefore been recommended to use at least two phylogenetic approaches with different assumptions (Martins et al. 2002; Housworth et al. 2004), in addition to a nonphylogenetic analysis (Price 1997; Garland et al. 1999).

Accordingly, we analyzed our data using the raw data (TIPS), with Felsenstein's independent contrasts (FIC, Felsenstein 1985), and with the phylogenetic generalized least squares method (PGLS, Grafen 1989; Martins and Hansen 1997). Felsenstein's independent contrasts method is the most commonly used comparative approach and its assumptions and computational details have been described extensively elsewhere (Felsenstein 1985; Harvey and Pagel 1991; Garland et al. 1992). This method assumes that the data result from an evolutionary process that is similar to Brownian motion, that is, genetic drift or directional selection with randomly changing directions of selection. Phylogenetic generalized least squares is an extension of Felsenstein's method, based on generalized linear model theory (McCullagh and Nelder 1989). We chose to use this method for two reasons. First, it accounts for measurement error, which is important when sample sizes are small (Harmon and Losos 2005). Second, it incorporates the Ornstein-Uhlenbeck model of phenotypic evolution by genetic drift under various constraints (Lande 1976; Felsenstein 1988; Hansen 1997). The strength of constraints acting on the evolution of a phenotype is described by a parameter, a_c. If a_c = 0 and within-species phenotypic variation is negligible, PGLS and FIC yield identical results. However, with larger a_c-values (e.g., a_c > 15, indicating strong evolutionary constraints) and/or substantial intraspecific variation, PGLS will yield results that are more similar to raw data analyses. By applying three comparative methods, we cover a broad range of evolutionary mechanisms and assumptions, which should allow us to assess the robustness of the results.

We used the PGLS module in the COMPARE 4.6 package (Martins 2004) to examine relationships among traits, with (1) a_c = 0 (i.e., the FIC approach); (2) a_c estimated from the data, using a maximum-likelihood grid search (i.e., the PGLS approach); and (3) a_c set at a large value (i.e., the TIPS approach). The model with the highest log-likelihood value (logL; e.g., Rice 1995) was considered the most appropriate (for details, see Martins and Hansen 1997). Depending on the model selected, we used independent contrasts, PGLS-transformed data (Butler et al. 2000) or raw data to visualize the relationships in plots. To evaluate the effect of unknown branch lengths on results of comparative analyses we repeated calculations for 200 possible trees with the same topology (Fig. 1) but randomly generated branch lengths. We thus obtained 95% confidence intervals for correlation coefficients and regression slopes (Martins 1996) that included sampling variance and variance due to uncertain phylogeny.

Hypotheses on ancestral character states can be useful in understanding evolutionary trajectories. Several methods have been developed to reconstruct ancestral phenotypes (Farris 1970; Maddison 1991; Martins and Hansen 1997; Schlutter et al. 1997; Garland et al. 1999). We chose the approach implemented in the COMPARE package, because it provides standard errors to the estimates, and because it allows estimating ancestral values under different microevolutionary assumptions: Brownian motion (linear PGLS model) or constrained evolution (exponential PGLS model) (Martins and Hansen 1997; Martins 1999).

Statistical Analyses

Because the newt species vary considerable in body size (Malmgren 2001), we corrected all morphological variables for size prior to further analyses, using Mosimann's (1970) multivariate method. This method avoids problems associated with the use of residuals from ordinary least squares regression and with other methods controlling for body size effects (Klingenberg 1996; Darlington and Smulders 2001; Butler and Losos 2002). Mosimann's (1970) method removes the effects of size for each observation using a directly measured index of individual size (here SIZE, the geometric mean of the seven morphometric traits measured). Each individual was adjusted for size by taking the difference of each log-variable with logSIZE.

Species means for size-corrected (i.e., relative) variables and independent contrasts were checked for normality (normal quantile plot). Because no substantial deviations from normality were detected, we used Pearson product-moment correlations to test for associations between variables. The relationships between relative axilla-groin distance and forelimb and hindlimb length were estimated using ordinary least squares regression. We used backward stepwise multiple regression (with P for removing a variable = 0.1) to evaluate associations between relative axilla-groin distance, tail length, and tail height on relative swimming speed. Because multicolinearity may confound results of multiple regression (Hair et al. 1998), we first checked for intercorrelations among predictor variables. All regressions with independent contrasts were forced through the origin (Garland et al. 1992). A significance level of α = 0.05 was used for all statistical tests. All means are reported ± standard error. Statistical analyses were performed in Statistica 6.1 (StatSoft 2000).

Body Elongation

The present study reveals considerable interspecific variation in body elongation among female Triturus newts. Body elongation in newts is associated with the number of trunk vertebrae (Arntzen and Wallis 1999), a trait that responds plastically to developmental temperature in a variety of vertebrate taxa (e.g., Barlow 1961; Lecyk 1965; Osgood 1978), including salamanders (Orska and Imiolek 1962; Peabody and Brodie 1975; Jockusch 1997). Hence, genotype-environment interactions might affect the interspecific variation in relative axilla-groin length observed in this study. However, it seems unlikely that such effect would be strong enough to confound our results. First, the plastic response in vertebrate number is generally small, even when a wide range of experimental temperatures is applied (see references above). Second, female newts actively select oviposition sites with favorable water temperatures (Gvoždík 2005), thus reducing the variation in thermal conditions during egg development (Huey 1991) and, hence, the variation in trunk vertebrate number. Finally, within-species variation in the number of trunk vertebrae is small (Arntzen and Wallis 1999). Therefore we assume that the effect of developmental plasticity on axilla-groin distance variation is minor relative to the interspecific, genetic variation in this trait.

Consistent with our prediction and with previous findings (Arntzen and Wallis 1999), the evolution of body elongation concurred with a reduction of head width and limb length in female newts. A similar trend has been described in a variety of fossil and extant groups of vertebrates (Gans 1975; Greer 1987, 1991; Wiens and Slingluff 2001; Adriaens et al. 2002; Bejder and Hall 2002; Caldwell 2003). Surprisingly however, the trend varied between Triturus lineages. Whereas “big-bodied” newts follow the general trend of body elongation with concurring head width and limb reduction, head width in “small-bodied” newts was only marginally associated with axilla-groin length, and limb length was not. The disparity in elongation trends between the two Triturus lineages echoes earlier findings in other vertebrate groups (Gans 1975; Lande 1978; Caldwell 2003; but see Wiens and Slingluff 2001), in which the evolution towards limblessness proceeded through two stages of elongation (Gans 1986): (1) a change in body diameter relative to body length and (2) an ongoing reduction of relative body diameter with a concomitant reduction in limb dimensions. The difference in body elongation between the two Triturus lineages suggests that they may be representative of these two stages of evolution toward extreme elongation and perhaps limblessness. Species with intermediate morphologies are lacking from modern groups of elongated, obligatorily aquatic salamanders (Amphiuma, Proteus, and Siren), and the fossil record of these groups is currently poor (Rieppel and Grande 1998). Triturus newts may provide a valuable alternative system to study the developmental mechanisms and adaptive significance of body elongation, and its role in the aquatic lifestyle of caudate amphibians.

Although these studies represent a major trend in various vertebrate groups (see references above), little is known about the factors driving the evolution of body elongation. From a biomechanical perspective, body elongation and its concurrent morphological changes seem advantageous to animals moving in confined places (e.g., burrows, crevices, dense vegetation) or to animals that specialize in undulatory locomotion (Gans 1975). In undulatory swimmers, more elongate animals can recruit greater proportions of their body to propel themselves through the water (Gillis 1996). Accordingly, body elongation in members of the T. cristatus group with prolonged aquatic phases has been considered an adaptation enhancing swimming capacity (Arntzen and Wallis 1999). However, our performance measurements demonstrate that body elongation reduces, rather than aids, maximal swimming speed in female Triturus. This surprising result suggests that our expectations of how body shape should influence swimming speed in newts were naive, and beckons for information on the kinematics and mechanics of aquatic locomotion in these animals. More specifically, we need to know to what extent body undulations contribute to useful forward thrust. If it turns out that the tail, rather than the body, drives swimming, than this could explain the inverse relationship between relative axilla-groin length and maximal swimming speed. Indeed, having a rather short trunk with few vertebrae may increase body stiffness (Long and Nipper 1996) and this would help reduce the recoil of the anterior part of the body and, ultimately, decrease the amount of drag experienced during swimming (Webb 1988; Domenici and Blake 1997). Our results corroborate the idea of an important role for the tail in swimming (for theoretical considerations, see also Bainbridge 1963; Vogel 1994; Gillis 1996 and for similar empirical results, see Bennett et al. 1989; Gillis 1997).

If body elongation is an adaptation to aquatic life, it must benefit the animals through some function other than swimming speed. Possibly, body elongation aids aquatic acceleration capacity (Webb 1984; Domenici and Blake 1997; Brainerd and Patek 1998). However, fast acceleration requires a relatively high body and tail, in addition to a flexible body. In Triturus newts, body elongation may enhance flexibility through increased vertebrae numbers, but it is not associated with increased tail dimensions. This suggests that a need for high acceleration capacity or swimming speed has not been a primary selective agent in the evolution of body elongation in female newts. Increased maneuverability seems a more likely asset of long bodies and one that may be of vital importance to species living in aquatic habitats with high structural complexity (dense vegetation). Also, one cannot rule out the possibility that body elongation serves a function outside locomotion (e.g., increased space for storing clutches).

The observed differences in relative body length between more aquatic and more terrestrial members of a clade may also result from more stringent selection for shorter bodies on land than in the water. In our study system, relative axilla-groin length (and not relative limb length) is the principle structural component explaining variation in running speed. The fact that species with relatively short trunks run faster seems to contradict the idea that longer bodies should allow more lateral flexion, larger stride lengths, and hence higher running speeds (Sukhanov 1968). Perhaps having a short, stiff trunk reduces excessive flexion and torsion of the body, thereby decreasing the internal work necessary during locomotion, and hence allowing higher speeds (Van Damme and Vanhooydonck 2002).

No Trade-Off between Swimming and Running Speed

We found no evidence for a trade-off between maximum swimming and running speed in female newts. This is consistent with results obtained from intraspecific studies on semiaquatic frogs, salamanders, and snakes (Bennett et al. 1989; Navas et al. 1999; Shine and Shetty 2001; Nauwelaerts et al. 2005). A failure to detect a negative association between putatively conflicting performance functions in intraspecific studies is typically ascribed to low interindividual variation, confounding effects of “overall quality” differences among individuals, or lack of motivation of the study animals (Vanhooydonck et al. 2001; Van Damme et al. 2002). These explanations seem less likely in interspecific comparisons, where differences in morphology, performance, and ecology are generally more pronounced.

In this case, the hypothesis of a functional trade-off seems an invalid assumption, originating from an incomplete understanding of the kinematics and mechanics of two putatively conflicting functions. Clearly, more work is needed on the question of exactly how musculoskeletal systems can accommodate to cope with different environments (Biewener and Gillis 1999; Ashley-Ross and Bechtel 2004). The unpredicted antagonistic effects of body elongation and tail length on swimming speed, combined with the weak association between these two traits, may provide a functional explanation for the absence of a trade-off between swimming and running capacity in female Triturus newts.

It has been argued that the evolution of two performance traits is rarely compromised if the traits are functionally decoupled, that is, propelled by different structures (Domenici and Blake 1997; Blake 2004). The decoupling of aquatic locomotion (driven by the tail) from terrestrial locomotion (driven by the body/limbs), may help explain how newts succeeded in maximizing both types of performances. In addition, the antagonistic effects of a particular feature on different functions may be “swamped” by effects of other traits that influence the two functions in the same direction (Garland 1994; Van Damme et al. 2002). Many biochemical, physiological, and morphological characters (e.g., enzyme activity, muscle and blood buffering capacity, glycogen storage, respiratory functions, relative muscle mass) may enhance (or decrease) both running and swimming capacity (Biewener and Gillis 1999; Brainerd and Simons 2000; for a similar reasoning on the endurance-speed trade-off, see Garland 1994).

Alternatively, the idea of a trade-off between aquatic and terrestrial speed may be valid, but difficult to demonstrate with the current dataset. A possible factor is the limited amount of variation among species in the relative use of aquatic and terrestrial habitats. All species in the study can be considered “generalists” because they make use of both environments. Detecting trade-offs in generalists is deemed more difficult because selection regimes may not be sufficiently different to invoke extreme adaptations in one or the other direction (Vanhooydonck and Van Damme 2001).

In conclusion, our study reveals several unexpected results regarding the evolution of form and locomotory function in newts and exposes shortcomings in our understanding of how a single musculo-skeletal system may cope with different environmental conditions. Understanding the selection pressures that acted during evolutionary transitions between major environments will require a combination of biomechanical and kinematic analyses, and measurements of different types of performance as well as their ecological relevance. In this respect, and because of their resemblance with the first tetrapods (Edwards 1989), caudate amphibians seem excellent models to study the water-land transition (Clack 2002).