Salamander ecomorphology reveals a unique suite of climbing adaptations

Abstract

Many salamanders climb extensively but lack morphological adaptations, such as claws or adhesive toe pads, found in other climbing tetrapods. Here, we compared climbers and non-climbers from the salamander genera Aneides and Plethodon to evaluate potential morphological adaptations for climbing across multiple levels of biological organization. We integrated body shape morphometrics, allometry of the autopods (manus and pes), mechanical advantage of the digits and comparisons of epithelial microstructures. Compared with non-climbers, climbers have longer limbs that likely facilitate faster climbing and have larger and positively allometric autopods that likely yield superior clinging performance. Longer digits increase climbing reach at the expense of grip force, but climbers circumvent this trade-off with phalangeal morphologies that increase the mechanical advantage of their digits. A few species also have pes epithelia that may increase adhesion or friction but epithelial morphology was largely uncorrelated with habitat use. While scansorial species of Aneides and Plethodon share some characteristics with other genera of climbing salamanders, our results reveal subtle differences between rock-climbing and tree-climbing salamanders that reflect distinct selective pressures imposed by their microhabitats. Many-to-one mapping enables alternate strategies to address the challenges of climbing and helps to explain the abundance and diversity of climbing tetrapods.

1. Introduction

Many-to-one mapping explains how different suites of morphological traits can achieve the same performance [1]. Many factors can influence phenotypic variation and limit morphological convergence (e.g. trade-offs, developmental constraints and genetic variation); as a result, animals have evolved diverse strategies to overcome similar challenges [2–4]. For instance, the propensity to climb (scansoriality) exposes animals to more pronounced gravitational and shear forces compared with horizontal walking, which promotes traits that increase friction and adhesion [5–8]. Claws are common adaptations for gripping and have convergently evolved across invertebrates and vertebrates [5,9,10]. Many tree-climbing (arboreal) lizards also have adhesive toe pads that provide access to a wider range of habitats [9,11–13]. However, many climbing tetrapods do not have claws or toe pads [14,15]. Lungless salamanders (Plethodontidae), in particular, have undergone multiple transitions towards arboreality but lack obvious adhesive structures [15–17]. Plethodontids provide the opportunity to investigate climbing strategies in the absence of specialized morphologies.

Climbing salamanders (Aneides spp.) are the only temperate plethodontids that consistently utilize arboreal habitats [16] (figure 1). The genus includes species like Aneides vagrans, which is renowned for its ability to climb redwood trees up to 93 m above the ground [18,19]. Some Aneides species are more saxicolous, in that they often climb rock structures, but venture into trees during the summer [20,21]. A few morphological characteristics of Aneides have been purported as adaptations for climbing, such as long limbs and digits, squared toe tips and recurved terminal phalanges with large ventral processes [22–24]. However, it is not clear how these morphological traits improve climbing performance. Furthermore, there may be nuanced ecomorphological differences between arboreal and saxicolous Aneides species. Lizards often have limb morphologies that match the structural differences of their microhabitats (i.e. substrate width and texture) [11,13,25], but this pattern has not been investigated in Aneides salamanders.

Figure 1.

Phylogenetic relationships of the focal species of Aneides and Plethodon. Tip shapes depict microhabitat preferences, while pies at the internal nodes depict likelihood values from an ancestral state reconstruction. Photos of representative species from top to bottom: A. aeneus, A. hardii, A. lugubris and P. glutinosus (not to scale).

Examination of the autopods (manus and pes) may yield insights into how some Aneides species are adapted for clinging, gripping and ultimately climbing. Salamanders use a layer of mucus covering their bodies to cling to surfaces [26,27]. Biomechanical models and empirical data suggest their clinging performance is predicted by autopod contact area but not autopod shape [17,27,28]. A few cave-dwelling salamanders (Chriopterotrion magnipes and Speleomantes spp.) grow disproportionately large pedes, possibly reflecting increased selection on pes size to improve clinging performance on cave walls [29,30]. Moreover, many arboreal and saxicolous frogs have evolved toe pads with hexagonal epithelial cells separated by deep channels and covered in peg-like nanostructures that disperse mucus across the skin, increasing wet adhesion and friction through more contact with surface asperities [31]. Salamanders may have similar epithelial morphologies on their autopods, which would increase the clinging effectiveness of their autopods [15,26,27].

Lastly, some salamanders augment their attachment to rough surfaces by mechanically gripping the surface with their digits [19,26]. While longer digits extend the reach of climbers, they generate lower gripping forces than shorter digits with an otherwise comparable morphology, presenting a functional trade-off between reach versus force [32]. However, larger ventral processes on the terminal phalanges affect the insertion of the digital flexors and may increase the mechanical advantage and gripping force of the digits in climbing species [22,24]. Scansorial species of Aneides may also exhibit greater dorsoventral curvature of the terminal phalanges to reduce bone strain while gripping [33].

We compared the functional morphology of climbing and non-climbing species of Aneides to assess whether the limbs of scansorial species may be adapted for scansoriality. We also examined three Plethodon species, which are closely related to Aneides and span a similar gradient between climbing versus ground-dwelling. Our objectives were to: (i) assess the relationship between body shape and habitat preference, (ii) contrast autopod allometry between climbers and non-climbers, (iii) quantify the mechanical advantage of the phalanges for gripping, and (iv) examine potential adaptations for adhesion or friction in the microstructure of pes epithelia. We predicted that climbing species exhibit longer limbs, positive autopod allometry, higher mechanical advantage in their digits and pes epithelia with deep intercellular channels and peg-like nanostructures compared with non-climbing species. We also predicted that these traits may differ between arboreal and saxicolous species. We found support for most of our hypotheses, indicating that a suite of subtle morphological changes likely improves climbing abilities in Aneides. Our findings further demonstrate that claws and adhesive toe pads are beneficial for climbing but not required; climbing can evolve in other tetrapods through alternate strategies involving only subtle changes to their morphology.

2. Methods

(a). Habitat categorization and reconstruction

We generated morphological data for all 10 species of Aneides and three species of Plethodon (P. elongatus, P. glutinosus and P. petraeus). Based on the literature, each species was categorized as a climber or non-climber and then subdivided based on its primary habitat preferences (electronic supplementary material, table S1). Half of the Aneides are scansorial (arboreal or saxicolous), while the remaining species are ground-dwellers (terrestrial). Plethodon petraeus is saxicolous, whereas P. glutinosus is primarily terrestrial but facultatively climbs on rocks and logs. These Plethodon species co-exist with A. aeneus and occur syntopically at some localities [34,35]. Meanwhile, P. elongatus is fully terrestrial and overlaps in distribution with some Aneides species in the western United States.

To investigate the origin of scansoriality in Aneides, we reconstructed the evolutionary history of habitat use with a recent time-calibrated salamander phylogeny [36]. We pruned the tree to include 30 plethodontid species, including our focal taxa and 17 additional species that we later used to investigate the evolutionary allometry of pes size across plethodontids more broadly (see below). Then, we performed a stochastic character mapping with 1000 simulations using the ‘make.simmap’ function in the phytools R package v. 2.4 [37]. The analysis was conducted with an equal-rates transition model, which was determined to be a better fit than a symmetrical rates model and all-rates-different based on Akaike Information Criterion (AIC) scores. All analyses in this study were performed in R v. 4.3.2 [38].

(b). Body shape measurements

We examined 411 preserved specimens of Aneides and Plethodon, with an average of 32 ± 8 s.d. (range: 15−47) specimens per species that spanned the size range of each species (electronic supplementary material). Each specimen was photographed in the dorsal and ventral views with its digits splayed out using a Nikon D800E Digital SLR equipped with a 105 mm macro lens, an iPhone 11 or an iPhone 13. We prioritized specimens with minimal distortion, all of their limbs and a complete tail. However, we also used individuals of A. caryaensis and P. elongatus with partial tails due to the limited availability of specimens. We sampled males and females instead of a single sex because our preliminary analyses indicated that Aneides exhibit minimal sexual dimorphism in their limbs (electronic supplementary material).

To compare body shape between climbers and non-climbers, we quantified eight morphological traits often related to locomotion and habitat use [39,40]. These included: snout-vent length (SVL), tail length, forelimb length, hindlimb length, length of the longest (third) digit on the manus, length of the longest (third) digit on the pes, surface area of the manus and surface area of the pes (figure 2a). Most measurements were taken from photographs using FIJI v. 2.3 [41]. The autopod area was measured by tracing its perimeter and calculating the enclosed area in FIJI. In this study, autopod referred to the distal regions of the limbs that included the manus (carpals, metacarpals and phalanges) in the forelimbs or the pes (tarsals, metatarsals and phalanges) in the hindlimbs. Because the limbs were often fixed in variable poses, we used digital callipers to measure the limb segments (e.g. upper arm/leg and lower arm/leg) to the nearest 0.1 mm and then summed the segments to calculate limb length. We also weighed each individual to the nearest 0.01 g and then subtracted the weight of their specimen tag. For specimens with partial tails, we regressed the tail length and body mass of complete specimens against SVL on a log–log scale and used the resulting species-specific equations to estimate the missing tail and mass data.

Figure 2.

Diagram of the morphological measurements. (a) Body shape measurements: SVL, tail length (TL), forelimb length (FLL), longest finger length (FDL), manus area (MA), hindlimb length (HLL), longest toe length (HDL) and pes area (PA). (b) Measurements of X, Y and Z were used to measure curvature in the phalanges, depicted in dorsal and lateral views. (c) Anatomical model based on [32]; used to measure mechanical advantage at the proximal interphalangeal (PIP) joint, distal interphalangeal (DIP) joint and at the tip (TIP) as the ratios between the lengths of the moment arm (M1−3) and load arm (L1−3).

To visualize intraspecific and interspecific variation in body shape, we performed a principal component analysis (PCA). We first log-transformed the limb and tail measurements, regressed them against log(SVL) and used the residuals from the regressions as size-corrected variables. We performed the PCA with a covariance matrix on the residuals using the ‘prcomp’ function in the stats R package v. 4.4.1 [38]. We plotted intraspecific variation and mean values for each species with the phylogeny overlaid to show the phylogenetic relationships.

(c). Allometry of autopod size

We investigated the static allometry of the manus and pes area (A) relative to body mass (W) for each species, assuming the allometric equation

$${\displaystyle A=b{W}^{\alpha },}$$

where α represents the relative growth parameter and b represents the relative autopod area. On a log–log scale, the allometric relationship is linear: log(A) = log(W) × α + log(b), where α is the static allometric slope and log(b) is the static allometric intercept. We hereafter refer to α and b as the slope and intercept, respectively. We estimated these coefficients using ordinary least squares regressions with the smatr R package v. 3.4 [42]. Scaling relationships were considered positively allometric if the 95% confidence interval (CI) of the calculated slopes was above the expected isometric relationship between area and mass (slope = 0.66). We used the non-phylogenetic ‘procD.lm’ function in the geomorph R package v. 4.0.8 [43] to compare the average static slope and autopod size between habitat groups.

We compared the scaling coefficients for Aneides and Plethodon with published values for the pes of 17 additional plethodontid species (Chiropterotriton magnipes, 10 Bolitoglossa and 6 Speleomantes) [29,30] (electronic supplementary material, table S1). We re-analysed the Speleomantes data from [30] to extend their reported coefficients at the genus level to species-specific coefficients. The maximum SVL, locomotor mode (climbing versus non-climbing) and habitat preferences were assessed for each species (electronic supplementary material, tables S1 and S8). We used the SVLs to estimate the maximum body mass of all 30 species using a length–mass equation for plethodontid salamanders [44]. Then we estimated the maximum pes area of each species using the species-specific scaling coefficients. We used the ‘procD.pgls’ function in geomorph to perform phylogenetic least squares regressions to compare the evolutionary allometric slopes and relative pes size between habitat groups, while accounting for phylogenetic non-independence. While static allometry refers to pes size relative to body mass within a species, evolutionary allometry refers to the relationship across species.

(d). Functional morphology of the phalanges

To investigate the gripping potential of climbers and non-climbers, we compared the phalanges of our 13 focal taxa with micro-computed tomography (µCT) scans. We acquired one µCT scan per species from MorphoSource.org [45,46] or by obtaining new scans (electronic supplementary material, table S2). New scans were generated using a General Electric Phoenix v|tome|x m 240 at the Smithsonian National Museum of Natural History Imaging Facility, with a voxel size between 27 and 45 μm, 80 kV, 230−240 μA and an exposure time between 250 and 333 ms.

We selected one of the longest fingers and toes from each specimen, favouring digits that were straighter and undamaged. We segmented and measured the three phalanges of each digit in three-dimensional (3D) Slicer v. 5.6.2 [47,48]. From the ossified portions of the proximal and penultimate phalanges, we took six measurements following published methodologies [32]. These traits included phalangeal length, midshaft diameter, proximal end height, proximal end width, distal end height and distal end width (figure 2b). We also took five measurements from the terminal phalanx that included phalangeal length, proximal end height, proximal end width, distal end width and curvature (figure 2b). Using the equation commonly used to calculate claw curvature [49], we calculated phalangeal curvature as:

$${\displaystyle \mathrm{c}\mathrm{u}\mathrm{r}\mathrm{v}\mathrm{a}\mathrm{t}\mathrm{u}\mathrm{r}\mathrm{e}=57.296\times (2\mathrm{s}\mathrm{i}{\mathrm{n}}^{-1}(2{\mathrm{X}}^2{\mathrm{Y}}^2+2{\mathrm{Y}}^2{\mathrm{Z}}^2+2{\mathrm{X}}^2{\mathrm{Z}}^2-{\mathrm{X}}^4-{\mathrm{Y}}^4-{\mathrm{Z}}^4{)}^{0.5}/2\mathrm{X}\mathrm{Y}),}$$

where X is the distance from the ventral base of the phalanx to the vertex of the ventral curve, Y is the distance from the ventral vertex to the tip of the phalanx and Z is the distance from the ventral base of the phalanx to the tip of the phalanx. Measurements of X, Y, Z and the terminal phalanx length were collected from lateral photos of the phalanx in FIJI. All linear measurements were divided by digit length to reduce the effects of size. We analysed variation in the size-independent phalangeal measurements (ratios and curvature) from the manus and pes with a PCA using the ‘prcomp’ R function, performed with a correlation matrix to account for the different units of these measurements.

We measured mechanical advantage as a proxy for gripping force, assuming a neutral and straightened position. We accounted for postural variation by straightening the digits using the ‘Curved Planar Reformat’ module in 3D Slicer. This method maintains spacing between the phalanges but causes minimal distortion to the shape of the phalanges. Thus, we used the ‘FastModelAlign’ module [48] to map the original phalanges onto the position of the distorted bones. Minor adjustments to phalangeal position and orientation were made using the ‘Transforms’ module. To measure mechanical advantage, we assumed a simple anatomical model with a single flexor tendon that inserts on the ventral process of the terminal phalanx (figure 2c) [32]. We measured the tendon moment arms at each joint and allowed the load arm to increase along a proximodistal gradient [32]. Mechanical advantage at the proximal interphalangeal (PIP) joint, distal interphalangeal (DIP) joint and the tip of the digit (TIP) were then calculated by taking the ratio between the moment arm and load arm.

(e). Relationship between habitat and morphology

We tested for associations between morphology and habitat with one-way ANOVAs. A separate phylogenetic ANOVA was performed for each size-independent body shape trait (averaged by species), size-independent phalangeal measurement, mechanical advantage at each joint and total mechanical advantage summed across the joints. These analyses were performed using the ‘procD.pgls’ function in geomorph. For significant tests (alpha level = 0.05), pairwise comparisons between habitat groups were conducted with a Holm–Bonferroni correction to reduce the probability of type I errors associated with conducting multiple statistical tests.

(f). Scanning electron microscopy of the pes

To investigate whether climbing species have epithelial morphologies on their pedes that aid in adhesion, we examined the ventral surface of the pes in nine focal species using scanning electron microscopy. We dissected pedes from specimens in natural history collections or the authors’ personal collection (electronic supplementary material, table S3). All specimens were fixed in 10% neutral-buffered formalin and stored in 70% ethanol. The authors’ specimens were frozen for less than a month and thawed before fixation, but appeared undamaged. Samples were dehydrated with an ethanol series and moved into 100% ethanol 24 h prior to being critical point dried (Tousimis Autosamdri 815) and sputter coated with gold palladium (Denton Vacuum Desk V). Whole pedes were imaged using a Zeiss Evo 10 scanning electron microscope (Carl Zeiss AG). We also estimated the density of potential mucus pores on the soles and toes using micrographs with magnifications between 300× and 450× (electronic supplementary material, figure S1).

3. Results

(a). Habitat evolution in Aneides

Our evolutionary reconstruction of habitat use indicated that the ancestor of Aneides was most likely scansorial. Specifically, it had the highest likelihood of being saxicolous (46%), compared with being arboreal (15%) or terrestrial (39%; figure 1 and electronic supplementary material, figure S2). The reconstruction also indicated that arboreality may have independently evolved from terrestriality twice, once in the ancestor of A. lugubris and a second time in the ancestor of A. ferreus+A. vagrans.

(b). Ecomorphological variation in body shape

Arboreal, saxicolous and terrestrial salamanders occupied unique regions of morphospace (figure 3). The first principal component (PC1; 80.5% of the variation) described variation in limb length, digit length and autopod area; while PC2 (10.6% of the variation) reflected variation in tail length (electronic supplementary material, table S4). Arboreal and saxicolous species had longer limbs, longer digits and larger autopods than terrestrial species, but saxicolous species had slightly longer limbs and larger extremities than arboreal species. The sampled Plethodon species had longer tails than most Aneides, regardless of their habitat preferences. The saxicolous P. petraeus had forelimb morphologies that were, on average, not as long nor as large as those in the saxicolous Aneides but still longer than those of terrestrial Plethodon species. Plethodon glutinosus had longer limbs than the terrestrial Aneides, while P. elongatus had the shortest limbs of all the species examined in our study.

Figure 3.

Phylomorphospace showing variation in body shape. Each small, unfilled symbol represents an individual, while the convex hulls depict intraspecific variation. Larger filled symbols represent species means (n = 15–47 individuals per species). The black branching lines depict the phylogenetic relationships between species.

Phylogenetic ANOVAs supported these results and revealed significant variation related to habitat use in all traits (p ≤ 0.050; electronic supplementary material, table S5). Saxicolous species had significantly longer tails than arboreal species (p = 0.038), but no other pairwise comparison was significant between the two groups (p ≥ 0.522). In contrast, all comparisons between saxicolous species and terrestrial species were significant (p ≤ 0.039). Arboreal species only had significantly longer fingers (p = 0.024) and toes (p = 0.048) than terrestrial species.

(c). Allometry of autopod size

The autopods of all P. petraeus and all Aneides species scaled with positive static allometry relative to body mass, while those of P. elongatus and P. glutinosus scaled with isometry (electronic supplementary material, table S6). Habitat had a significant effect on the static slope of the manus (F₂ = 2.767, p = 0.048) but not the pes (F₂ = 1.151, p = 0.306). Saxicolous species exhibited a shallower pes slope (slope = 0.73) than arboreal species (slope = 0.79; Z = 1.92, p = 0.016; figure 4a), while all other pairwise comparisons were non-significant (p ≥ 0.134; electronic supplementary material, table S7). Habitat had a significant effect on the relative size of the manus (F₂ = 208.34, p = 0.001) and pes (F₂ = 247.72, p = 0.001), with all pairwise comparisons being significant (p ≤ 0.001; electronic supplementary material, table S7). Saxicolous species had proportionally the largest autopods, followed by arboreal species and then terrestrial species (figure 4a,b). Overall, there was considerable variation in the static allometric slope and intercept of the pes across 30 species of plethodontid salamanders (slope = 0.66−0.946; intercept = 0.06−0.21; figure 4c and electronic supplementary material, table S8).

Figure 4.

Allometry of autopod size (A) relative to body mass (W). Static allometry of the (a) manus and (b) pes for species in Aneides and Plethodon, grouped by habitat. (c) Variation in static scaling coefficients for 30 species of plethodontid salamanders. (d) Evolutionary allometry of estimated pes area and maximum body mass across five genera of salamanders. In all panels, the black dotted line represents the isometric slope (0.66).

The common evolutionary allometric slope across all 30 species was 0.64 (95% CI = 0.55−0.72), indicating that salamander species exhibit evolutionary isometry (figure 4d). Habitat did not have a significant effect on the evolutionary allometric slope (F₂ = 0.825, p = 0.441; electronic supplementary material, table S9). However, habitat did have a significant effect on relative pes size (F₂ = 20.192, p = 0.001). Saxicolous species had significantly larger pes than arboreal (Z = 1.852, p = 0.021) and terrestrial species (Z = 3.30, p = 0.001; electronic supplementary material, table S9). Arboreal species also had larger pes sizes than terrestrial species (Z = 1.797, p = 0.033; figure 4d).

(d). Functional variation of phalangeal morphology

Phalanges of the fingers and toes were characterized by similar axes of variation that distinguished scansorial species from terrestrial species (figure 5a). Along PC1 (70.5% of the variation) and PC2 (10.6% of the variation), arboreal and saxicolous species occupied overlapping regions of the phalangeal morphospace (electronic supplementary material, table S10). Scansorial species differed from terrestrial species by having smaller proximal phalanges (in all anatomical directions), narrower and shorter penultimate phalanges, and slightly longer penultimate and terminal phalanges (figure 5a). The terminal phalanges of arboreal and saxicolous species were more recurved, laterally expanded at the tip and had a larger ventral process (figure 5a). In contrast, the terminal phalanges of P. elongatus were dorsoventrally flatter and more pointed at the tip. Most of the terrestrial species were intermediate to the scansorial species and P. elongatus, in that the former exhibited some curvature and lateral expansion at the tip. The phalangeal morphology of P. petraeus was most similar to members of the A. flavipunctatus complex but notably different from other Plethodon species.

Figure 5.

Morphological and functional variation of the phalanges. (a) Morphospace of the proximal (PP), penultimate (MP) and terminal (TP) phalanges in the longest finger and toe of each species. Insets depict the dorsal and lateral view of the terminal phalanges in representative species: (i) A. aeneus, (ii) A. lugubris, (iii) A. niger, (iv) A. iecanus and (v) P. elongatus. (b) Mechanical advantage (MA) of the longest finger at the PIP joint, DIP joint and the distal TIP. (c) Mechanical advantage of the longest toe at the PIP, DIP and TIP.

However, the phylogenetic ANOVAs did not indicate a strong association between habitat and phalangeal morphology (electronic supplementary material, tables S11 and S12). Significant variation in the finger phalanges associated with habitat (p ≤ 0.049) was found in the width and height of the proximal and penultimate phalanges, as well as the size of the ventral process on the terminal phalanges (electronic supplementary material, table S11). The toes exhibited similar patterns in the proximal and penultimate phalanges as well as significant variation in the curvature of the terminal phalanges (p ≤ 0.029; electronic supplementary material, table S12).

Mechanical advantage also differed between habitat groups (electronic supplementary material, tables S11 and S12; figure 5b,c). At the PIP joints for both the fingers and toes, terrestrial species had the highest mechanical advantage, but there was substantial overlap between groups. At the DIP joints, saxicolous species had significantly higher mechanical advantage than terrestrial species (p ≤ 0.042), while arboreal species were intermediate between them. Similar patterns were observed at the tips of the digits, but only the toes of saxicolous species differed significantly from terrestrial species (Z = 1.96, p = 0.036). When mechanical advantage was summed across the joints, values for saxicolous species were significantly 1.2−1.3× higher than those for arboreal and terrestrial species (electronic supplementary material, table S12).

(e). Surface epithelium of the pes

Species examined herein had smooth surfaces on the ventral side of the pes and toe pads on the macroscale but variable skin morphology at the micro- and nanoscale (figure 6). The epithelia consisted of irregular polygonal cells in all species (figure 6). Most species lacked distinct intercellular channels, except for A. aeneus and A. hardii (figure 6a–g). Both species had channels that were about 1 µm wide and relatively shallow on their soles but nowhere else (figure 6e–g). Large pores, presumed to be mucus pores, were common on the soles (11–25 pores/mm²) and ventral surface of the toes (8–26 pores/mm²) but rare on the toe pads of all species (0−2 pores per pad; electronic supplementary material, table S3). One A. lugubris specimen had twice the number of large pores on its sole and numerous smaller pores (figure 6h) than the other A. lugubris specimen. Variation in the epithelial morphology and pore density did not appear to correlate with habitat use.

Figure 6.

Scanning electron micrographs of the epithelia on the ventral surface of the pes. Smooth sole and toe pad of (a) P. elongatus, (b) A. vagrans, (c) P. glutinosus and (d) A. aeneus showing the lack of intercellular channels between the polygonal cells. Intercellular channels were observed on the soles of (e–f) A. aeneus and (g) A. hardii. (h) The sole of an A. lugubris specimen with a high density of pores. Nano-morphologies observed in (i) A. niger, (j) A. lugubris (same as in (h)), (k) A. vagrans (same as in (b)), and (l) P. petraeus. Abbreviations: IC, intercellular channel; p, pore. Scale bar = 10 µm (a–h); scale bars = 2 µm (i–l).

We categorized the nano-morphologies into four discrete types but note the presence of continuous variation (figure 6i–l). Type 1 was composed of numerous spherical structures that appeared unorganized (figure 6i) and were found on the toe pads of most species. Type 2 consisted of papillous protrusions that were spaced out (figure 6j) and found on the soles of A. hardii, A. lugubris (but only in the specimen with abundant pores) and P. elongatus. In contrast, type 3 had dense clusters of papillae separated by narrow channels (figure 6k) and was found most notably on the soles of A. vagrans. Other species—including A. aeneus, A. flavipunctatus, A. niger, the other A. lugubris specimen and P. petraeus—had sole morphologies that were intermediate between types 2 and 3, in that the papillae varied in size and density. Lastly, type 4 was distinct and consisted of small, densely packed, peg-like structures found on the toes of P. petraeus and on the toes and sole of P. glutinosus (figure 6l). The pegs were loosely polygonal in shape and had central depressions on top (as did the papillae in type 3; electronic supplementary material, figure S3).

4. Discussion

Plethodontid salamanders lack the common morphological adaptations for overcoming gravitational and shear forces observed in other climbing tetrapods [15,16,19]. Instead, we found that scansorial species within Aneides and Plethodon possess alternate suites of morphological traits that likely augment their clinging, gripping and climbing performance on inclined surfaces. Scansorial species had longer limbs and digits than non-climbers, which is often associated with faster and more stable climbing on broad surfaces [50–52]. Climbers also had disproportionately large autopods and phalangeal morphologies that likely facilitate larger clinging and gripping forces, respectively, compared with non-climbers. Some also had epithelial morphologies that may increase adhesion and friction when clinging. Observations that A. aeneus occupy rock crevices further off the ground and cling to steeper angles than other syntopic plethodontids, including P. glutinosus support the hypothesis that these traits confer superior climbing abilities [53–55]. However, not all climbers share the characteristics of scansorial Aneides, emphasizing the existence of multiple climbing phenotypes [15,17]. Many-to-one mapping may contribute to the abundance of scansorial plethodontids and weak patterns of convergence [16,17] by availing more viable evolutionary pathways towards scansoriality [1,3].

Our findings indicate clear ecomorphological divergence in body shape between the arboreal, saxicolous and terrestrial species examined herein. However, when sampling across Caudata, many climbers do not appear phenotypically distinct from non-climbers [17], suggesting the effects of habitat may be lineage dependent. Nevertheless, the morphologies of arboreal and saxicolous Aneides appear fine-tuned for their respective microhabitats. Arboreal species climb on tree bark that is often rugose and with large footholds, whereas saxicolous species commonly climb on sandstone and limestone outcrops that are comparatively smoother (asperity size = 100−200 µm) [20,56]. Frequently climbing on rocks may exert stronger selection for larger autopods and gripping forces because salamander clinging performance decreases on coarse surfaces (asperity size = 100−350 µm) compared with completely smooth (0 µm) and rough surfaces (1000−4000 µm) [15,26]. Indeed, other saxicolous plethodontids also had larger autopods than arboreal species. Saxicolous Aneides also had longer tails than arboreal species, but the function of salamander tails is not well-known. Many plethodontids, including Aneides, have prehensile tails [57] that could be used like a fifth limb to increase stability and prevent falling [8,58–60], or to grasp the substrate when crossing narrow perches or bridging gaps [57,61]. Additionally, arboreal Aneides use their tail in controlled descents after jumping or falling [62]. To better understand the relationship between body shape and habitat, future studies should compare the climbing performance of morphologically and ecologically diverse salamanders on substrates that vary in texture and width.

To maintain consistent clinging performance, morphological traits for attachment should ideally scale proportionally with body mass and the gravitational forces acting to dislodge them (slope = 1) [63]. Although less than one, the static slopes of climbing species of Aneides and Plethodon exhibited positive allometries that indicate they grow disproportionately large autopods (slope = 0.73–0.92). These findings are unusual because within groups of closely related frogs and lizards, adhesive pads scale closer to isometry (slope = 0.66) and vary more in relative size [63]. Selection can act on static slopes and intercepts, but the latter are often more responsive to selection [63–65]. Static slopes are considered to be constrained by stabilizing selection, so positive allometry is often interpreted as a response to strong directional selection [65,66]. Therefore, selection is likely acting on autopod size in climbing salamanders to increase clinging performance. The positive allometries in terrestrial species of Aneides probably represent plesiomorphic conditions and support our finding that the ancestor of Aneides was likely scansorial (figure 1).

Moreover, A. ferreus and A. vagrans exhibited steeper allometries (slope = 0.88−0.92) than the other scansorial species of Aneides (slope = 0.73−0.83). Aneides vagrans dwell in the canopy of redwood trees as high as 93 m, which is 1.5−5× higher than has been documented for other Aneides species [18,21,67,68]. Aneides vagrans can also manoeuvre and parachute in the air after jumping to glide down the tree, which is facilitated by their long limbs and large autopods that increase drag [62,69]. Aneides ferreus has only been observed 61 m above the ground but, in laboratory settings, they readily jump onto nearby objects, including vertical surfaces [18,67,70]. Environmental challenges imposed by gliding may be aligned with those for climbing, thereby producing a steeper allometric relationship in these species. Alternatively, larger autopods may be attributable to phenotypic plasticity, which could be addressed in future studies that examine the genotype–environment interactions affecting autopod size.

Despite the range of static intercepts and slopes recorded, scansorial plethodontids spanning multiple genera all have relatively larger pes than non-climbing relatives. We propose that scansorial salamanders obtain larger autopods and, thus, superior clinging performance through many-to-one mapping of scaling coefficients [28]. For instance, the arboreal Bolitoglossa franklini has isometric pes growth but a higher intercept, which yields relatively large pes and exceptional clinging abilities comparable to those of scansorial Aneides [71]. Previously, the pes allometries of Bolitoglossa were considered non-adaptive based on the assumption that webbed autopods should exhibit steeper static slopes than unwebbed autopods [29]. In reality, autopod shape is a poor predictor of salamander clinging performance [71], and the webbed autopods of Bolitoglossa are the by-product of a shift towards paedomorphic development [29]. Thus, variation in the scaling coefficients enables larger autopod sizes to evolve through selection despite potential constraints on autopod shape. Interestingly, no species examined herein exhibits both a high slope and intercept, potentially reflecting a functional trade-off that limits autopod size.

A few salamander species had pes epithelia that could confer some degree of increased adhesion and friction based on their similarities with frog toe pads [72–74]. The intercellular channels on the soles of the saxicolous A. aeneus and terrestrial A. hardii were narrower (approximately 1 µm wide) than the channels on the toe pad of the rock frog Staurios parvus, but like those found on its subarticular tubercles (0.06−1.6 µm)—secondary adhesive structures on frog toes [73]. Because the channels of A. aeneus and A. hardii were relatively shallow, they may have a larger role in friction than adhesion [74]. Furthermore, P. petraeus and P. glutinosus had peg-like nanostructures that resembled those in frogs. The pegs of the two Plethodon species were of similar width (approximately 0.33 µm) as the pegs of S. parvus (0.2−0.3 µm) and had similar concave depressions at the distal tips of the pegs [72,73] (electronic supplementary material, figure S3). These similarities suggest that the Plethodon nanopillars may aid in adhesion or friction. However, thick mucus layers on the pedes could reduce the effectiveness of salamander nanostructures (channels or pillars) by preventing interactions with the surface [75]. That said, all species examined herein as well as terrestrial Bolitoglossa have relatively low densities of putative mucus pores across their pedes (8–26 and 13−21 pores/mm², respectively) compared with arboreal Bolitoglossa (47–68 pores/mm²) and a frog (Hyla cinerea) toe pad (53 pores/mm²) [76,77]. That suggests scansorial Bolitoglossa and at least some scansorial frogs utilize mucus adhesion more than scansorial Aneides or Plethodon.

The inconsistent distribution of presumably advantageous epithelial morphologies among scansorial salamanders raises doubt about whether they are adaptive. Intercellular channels are also absent in terrestrial and arboreal Bolitoglossa but present in some semi-aquatic Desmognathus salamanders that also have keratinous toe tips, possibly for station holding in streams [76,78]. Seeing how torrent frogs also have intercellular channels that they use to drain excess fluid from under their pads [79], wet environments may promote the evolution of channels, but A. niger, which exhibits some semi-aquatic tendencies [80], lacks channels. Meanwhile, A. vagrans, and likely other species of Aneides, can potentially modulate the compliance of their toe pads and adhesion through fine control over the perfusion of blood into their digits [24]. Draining their digits would enable the distal tips to deform around small surface asperities and increase friction [81,82]. Indeed, it was observed that the toe epithelium of A. vagrans folded and formed large pleats when the animal was dragged across a surface [24]. These observations suggest the material properties of autopod epithelium may be more important for climbing than its microstructures.

Some salamanders can compensate for lower clinging performance on rough surfaces by gripping surface asperities with their digits [19,26]. We found that scansorial species of Aneides and Plethodon have larger ventral processes on their terminal phalanges than terrestrial species. That changes the insertion angle of the flexor tendon, increases force production, and mitigates the functional trade-off between gripping force and digit length. Climbers also had shorter proximal phalanges and slightly longer distal ones, another strategy for enhancing gripping force without sacrificing reach [32]. Furthermore, climbers had more recurved terminal phalanges that were laterally expanded at the distal tip, which may decrease bone strain when grasping and increase the stiffness of the digit [24,33]. Other salamanders also have expanded terminal phalanges but it appears unrelated to habitat preferences [83]. Instead, locomotor mode (climbing versus non-climbing) may be a better predictor of phalangeal morphology. Plethodon cinereus has phalanges resembling those of P. elongatus (flat and pointed) [84], which might explain its inability to climb on coarse vertical surfaces [15]. Climbing frogs also have expanded phalanges, providing additional evidence for their broad adaptive significance [85]. A more comprehensive comparison of phalangeal morphology might reveal that climbing salamanders have repeatedly evolved phenotypes that support gripping.

5. Conclusions

Integrating data across multiple levels of biological organization, from whole organisms to cells, helped us identify potential morphological adaptations that assist some scansorial salamanders. Because many-to-one mapping is prevalent across Plethodontidae, it is unclear to what extent our findings describe the ecomorphological patterns of scansorial lineages across the family. There are other scansorial lineages (Eurycea spp. and Nyctanolis pernix) with long limbs and large autopods that seem to resemble Aneides and Plethodon. However, many genera (i.e. Bolitoglossa, Chiropterotriton and Speleomantes) have webbed autopods and truncated digits [28,29] that contrast the dexterous digits in scansorial species of Aneides. The different suites of climbing morphologies likely reflect multiple factors, including separate evolutionary origins of scansoriality, different developmental pathways or preferences for microhabitats with different structural properties (i.e. width, texture and compliance). Yet, scansorial species of Aneides and Bolitoglossa have relatively large pedes and expanded terminal phalanges [28], suggesting that some traits are convergent across a broader phylogenetic scale than others. We propose that the traits studied herein represent axes of variation by which salamanders can vary their arsenal of climbing morphologies. Future studies should assess the role of constraints and environmental demands in structuring patterns of morphological convergence across climbing salamanders.

Ethics

This work did not require ethical approval from a human subject or animal welfare committee.

Data accessibility

The morphological data and script files used for this study can be found on Dryad [86]. All CT scans have been uploaded to MorphoSource (see electronic supplemental material).

Supplementary material is available online [87].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors’ contributions

J.M.H.: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, visualization, writing—original draft; G.P.: data curation, investigation, writing—review and editing; S.K.: conceptualization, funding acquisition, methodology, resources, supervision, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interests

We declare we have no competing interests.

Funding

This work was supported by a US National Science Foundation Graduate Research Fellowship (DGE-1746914), a Washington Biology Field Club Award and American Society for Ichthyology and Herpetology Gaige Fund Award to J.M.H.; as well as an American Association of University Women Research Publication Grant (015943) to S.M.K.