Decoupled evolution of ventral and dorsal scales in agamid lizards: ventral keels are associated with arboreality

Abstract

Arboreality has evolved in all major vertebrate lineages and is often associated with morphological adaptations and increased diversification concomitant with accessing novel niche space. In squamate reptiles, foot, claw, and tail morphology are well-studied adaptations shown to be associated with transitions to arboreality. Here, we examined a less well understood trait—the keeled scale—in relation to microhabitat, climate, and diversification dynamics across a diverse lizard radiation, Agamidae. We found that the ancestral agamid had keeled dorsal but not ventral scales; further, dorsal and ventral keels are evolutionarily decoupled. Ventral keeled scales evolved repeatedly in association with arboreality and may be advantageous in reducing wear or by promoting interlocking when climbing. We did not find an association between keeled scales and diversification, suggesting keels do not allow finer-scale microhabitat partitioning observed in other arboreal-associated traits. We additionally found a relationship between keeled ventral scales and precipitation in terrestrial species where we posit that the keels may function to reduce scale degradation. Our results suggest that keeled ventral scales facilitated transitions to arboreality across agamid lizards, and highlight a need for future studies that explore their biomechanical function in relation to microhabitat and climate.

1. Introduction

How organisms transition and adapt to novel environments is a fundamental question of evolutionary biology. Populations or species may evolve specialized phenotypes that facilitate expansion into previous inaccessible ecological space [1,2]. The repeated evolution of such traits associated with the repeated evolution of similar ecologies can provide evidence for adaptive evolution [3–6]. Sometimes these phenotypes not only allow for the exploitation of novel ecological space, but also result in elevated diversification rates [7]. Understanding which traits facilitate evolutionary transitions, when they do so, and if they subsequently increase diversification can provide important insight into fundamental questions of evolutionary biology.

Arboreality has evolved repeatedly across squamate reptiles [8,9]. The exploitation of arboreal microhabitat space is often associated with adaptive morphological transitions such as zygodactyly in chameleons [10] and adhesive toepads in anoles and geckos [9,11]. In some clades, diversification rates might also increase in response to accessing arboreal microhabitat space likely owing to increased ecological opportunity [8]. Conversely, rates of morphological evolution are reduced in arboreal squamate lineages perhaps owing to the constraints on morphologies that allow for exploitation of a specialized microhabitat [12–14]. Although species may be morphologically constrained overall, myriad different traits are able to facilitate transitions to arboreality including claws (e.g. [15–18]), adhesive toepads (e.g. [9,11,19,20]), toe and limb arrangement (e.g. [21–24]), and prehensile tails (e.g. [25]), only some of which appear in any given clade. Thus, understanding the evolution of arboreality requires investigations of a broad range of potentially adaptive traits. Here, we examine the relationship between arboreality and a trait overlooked in this context, the macroscopic morphology of body scales.

Many scale macrostructures appear broadly throughout squamate diversity including tubercles, keels, and osteoderms [26–28]. These scale structures are thought to have a variety of functions including altering optical properties of the scale [26], improving wear resistance [29] and armor [30]. Yet, the function of squamate scale ornamentation is still often unclear. In particular, keels—ridges along the centre of scales—do not have a consistently understood function despite occurring broadly across squamate reptiles. Thus, we tested four hypotheses about the evolution of keeled scales in agamid lizards.

First, we tested the hypothesis that keeled ventral scales are associated with the transition to arboreality in agamid lizards [31]. Keels may interlock with woody substrates in a similar manner to claws to improve overall clinging ability on vertical surfaces [31,32]. Indeed, functional evidence suggests that ventral keels are adaptive in this manner for arboreal snakes [33]. Whether this extends to legged lizards has not been studied to our knowledge. Agamids provide a particularly compelling system to test this hypothesis owing to their repeated transitions to arboreality [12].

Second, we tested the hypothesis that smooth ventral scales are favoured in wetter environments and keeled dorsal scales are favoured in drier environments. Several authors have posited that smooth scales should be favoured in more humid environments as moist soils are more difficult to slough [26,34,35]. Textured scales have the theoretically undesirable quality of allowing ‘sticking’ of soils and other refuse owing to increased surface area [26,34]. These issues should be more prominent in ventral scales as they make more contact with substrates. The structure of dorsal scales has also been associated with precipitation. Several arid-adapted squamates such as the agamid Moloch horridus and Crotalus rattlesnakes have modified scales that collect rain water [36,37]. Although rain harvesting is primarily driven by microstructures of the scale, work on Crotalus indicates that macrostructures like keels can improve water collection by reducing droplet size [37]. As with arboreality, agamids provide a compelling system to study covariation with climate as their diversity spans hot deserts to tropical rainforests [38].

Third, we tested two competing hypotheses: that keeled scales are associated with either hotter or cooler environments. It has been suggested that keels may form protected air pockets against the body facilitating heat retention in colder environments [39]. Conversely, it has also been suggested that the radiation reflecting properties of keels may allow them to reduce overall heating rates [26]. Neither hypothesis has been meaningfully investigated through functional or physiological modelling studies to our knowledge. Thus, it is unclear whether keels should facilitate heat absorption or reflection. Nevertheless, correlative studies may provide evidence supportive of a role for scale macrostructures in thermal physiology either way.

Finally, in addition to our three tested ecological associations, we tested the hypothesis that keeled scales are associated with increased diversification rates. If keels allow agamids to access novel environments, we might also expect subsequent increases in diversification rates as species further partition this novel habitat [7]. We discuss our results for each hypothesis in the context of understanding macroevolutionary trends of scale macrostructures and their potential role in facilitating ecological diversification across squamate reptiles.

2. Methods

(a) . Study system and data collection

Agamids are a diverse clade of lizards (greater than 500 species) that occur throughout Africa, Eurasia, and Oceania. Members of this group occur in a broad range of environments including deserts and tropical rainforests. Additionally, agamids have independently evolved arboreal or semi-arboreal forms repeatedly [12]. Thus, agamids provide a compelling system in which to study the association of scale keels with climate and microhabitat.

We compiled data for 182 agamid species with available phylogenetic data. We determined the presence and absence of keeled scales by searching the literature (electronic supplementary material, table S1). Our literature sources primarily consisted of species descriptions, identification guides and scale morphological studies. For each species, we noted keels separately for dorsal and ventral body scales. We did not collect data on the head and appendages owing to lack of consistent information across species. We compiled microhabitat data from the Global Lizard Trait Database [40] and checked classifications against citations therein. Our study species fell into three microhabitat categories: arboreal, saxicolous and terrestrial. We determined range-wide climate summary data for each species following Yuan et al. [41]. We downloaded the range of each of our study species from the Global Assessment of Reptile Distributions database [38]. We then extracted range-wide annual mean temperature (BIO1), annual precipitation (BIO12), mean temperature of warmest quarter (BIO10), mean temperature of coldest quarter (BIO11), precipitation of wettest quarter (BIO16) and precipitation of driest quarter (BIO17) from the WorldClim2 database [42] at 2.5 arc-minute resolution. We selected bioclimatic variables to represent both the generic climate experienced by each species and climatic extremes which may play a larger role in selection. For downstream analyses, we calculated the natural log-transformed range-wide mean and maximum of each bioclimatic variable for each species. Finally, we downloaded the squamate reptile phylogeny of Pyron et al. [43] time-calibrated by Ramm et al. [44] and pruned to our focal taxa. All analyses were performed in R v. 4.2.1.

(b) . Ancestral state reconstruction

We performed ancestral state reconstruction for our dorsal and ventral keel traits and microhabitat using SIMMAP [45]. We fit equal rates and different rate models for each trait and selected the best fit model using AICc scores. For our best-fit model, we determined the average number of transitions and calculated posterior probabilities for each ancestral state by summarizing across 999 simulations.

(c) . Comparative analyses

To test for phylogenetic signal in keel traits, we used the δ statistic approach [46]. This approach is based on Shannon entropy and is explicitly designed to handle categorical data. We determined significance by comparing our observed δ value with a null distribution generated from 999 random permutations of trait states across the tree. To test for integration between dorsal and ventral keels as well as associations between keels and microhabitat, we fit independent and correlated models using the corHMM R package [47]. We implemented both single-rate models and hidden Markov models. Single rate models were analogous to Pagel's models of correlated and uncorrelated evolution [48]. Hidden Markov models allow for rate heterogeneity which reduces spurious traits correlations that are generated by lack of evolutionary replication in deep time [49]. We determined the best fit model using AICc scores. We tested the relationship between climate (range-wide mean and max of each variable) and both dorsal and ventral keeled scales using phylogenetic ANOVAs in phytools [50]. Given the role of climate in determining vegetation, we also tested for differences in temperature and precipitation across microhabitat categories. For climate variables with significant keel state and microhabitat relationships, we performed pairwise tests for associations between keel state and climate within microhabitat groups.

(d) . State dependent diversification rate

We fit hidden state speciation and extinction (HiSSE) models [51] to test for state dependent shifts in diversification rates. We fit models separately for dorsal and ventral keel traits. For each trait, we fit four models. First, we fit a ‘null’ model without hidden states that set all rates equal for speciation (λ). Second, we fit a ‘CID-2’ model that allowed λ to vary as functions of two hidden states. Third, we fit a ‘dependent’ model that allowed λ to vary with keel state. Fourth, we fit a ‘full’ model that allowed λ to vary with both keel state and two hidden states. For all models, we set equal extinction (µ) and allowed all transition rates (q) to vary. In models with hidden states, we did not allow simultaneous transitions between both the observed and hidden states. Lastly, we fit the same set of diversification rate models for microhabitat in which we binned species into arboreal and non-arboreal.

3. Results

(a) . Ancestral state reconstruction

Both dorsal and ventral keels are evolutionarily labile (figure 1). The best-fit model for dorsal scale keel state (ΔAICc = 2.15) and microhabitat (ΔAICc = 27.94) was different rates. Neither model was better supported for ventral scale keel state (ΔAICc = 0.44). Thus, we compared results from both equal and different rate models for ventral keel state, which were qualitatively similar. We report results for the different rates model here for all variables. Our results show weak support that the ancestral agamid had keeled dorsal scales (posterior probability = 0.61) and saxicolous microhabitat (pp = 0.81) as well as strong support for non-keeled ventral scales (pp = 0.93; figure 1,electronic supplementary material, figure S1). We estimated 19.12 ± 0.18 gains and 25.84 ± 0.12 losses of keeled dorsal scales. For keeled ventral scales, we estimated 12.74 ± 0.07 gains and 11.75 ± 0.07 losses. Finally, we estimated 3.84 ± 0.04 transitions from arboreal to terrestrial, 0.68 ± 0.02 transitions from saxicolous to arboreal, 20.38 ± 0.11 from saxicolous to terrestrial, 20.44 ± 0.08 from terrestrial to arboreal, and 10.21 ± 0.08 from terrestrial to saxicolous. There were no transitions from arboreal to saxicolous.

Figure 1.

SIMMAP ancestral state reconstruction of ventral scale keel state across agamid lizards. Nodes are labelled with posterior probabilities for ancestral states. Tips are labelled with ventral scale keel state, dorsal scale keel state, and microhabitat.

(b) . Comparative analyses

We found significant phylogenetic signal for both dorsal and ventral keel state (dorsal: δ = 5.11, p < 0.001; ventral: δ = 10.93, p < 0.001). We did not find evidence of integration between dorsal and ventral scales for keel state. Our best supported model was the hidden Markov independent model indicating that each set of scales evolves independently (table 1). For tests of microhabitat associations, our best-fit model was the hidden Markov independent model for dorsal scales and the single-rate correlated model for ventral scales (table 1; figure 2). Dorsal scale keel state was not significantly associated with any climate variables (all p > 0.050). Comparatively, we found that ventral scale keel state was associated with annual precipitation (mean: F = 75.93, p = 0.005; max: F = 53.68, p = 0.017) and precipitation of the wettest quarter (mean: F = 71.85, p = 0.011; max: F = 46.45, p = 0.037), but not other climate variables (all p > 0.050). Species with keeled ventral scales had higher range-wide annual precipitation than species with non-keeled ventral scales. However, annual precipitation was also related to microhabitat with arboreal species having significantly higher range-wide annual precipitation (mean: F = 56.59, p = 0.003; max: F = 38.18, p = 0.006) and precipitation of the wettest quarter (mean: F = 47.42, p = 0.003; max: F = 28.52, p = 0.032). Our post hoc pairwise comparisons showed that annual precipitation and ventral keel state were related in terrestrial species (t = 3.38; p = 0.003; figure 3), but not arboreal (t = 0.83; p = 0.533) or saxicolous (t = 0.29; p = 0.797) species. These results were consistent for precipitation of the wettest quarter (arboreal: t = 0.56; p = 0.663; saxicolous: t = 0.62; p = 0.560; terrestrial: t = 3.53; p = 0.005).

Table 1.

Log-likelihoods, AICc, ΔAICc for models testing correlations between two categorical variables: (a) dorsal and ventral scale keel state, (b) dorsal keel state and microhabitat, and (c) ventral keel state and microhabitat. We fit four models for each: independent, hidden Markov independent, correlated and hidden Markov correlated. The best-fit models for each comparison are bolded (ΔAICc = 0).

model	log-likelihood	AICc	ΔAICc
(a) dorsal v. ventral keel state
independent	−170.0	348.3	17.3
hidden Markov independent	−154.8	331.0	0
correlated	−165.5	347.9	16.9
hidden Markov correlated	−149.9	340.0	9.0
(b) dorsal keel state v. microhabitat
independent	−207.9	432.6	1.8
hidden Markov independent	−195.3	430.8	0
correlated	−197.8	435.8	5.0
hidden Markov correlated	−186.2	469.2	38.4
(c) ventral keel state v. microhabitat
independent	−208.3	433.4	45.4
hidden Markov independent	−175.6	391.4	3.4
correlated	−173.9	388.0	0
hidden Markov correlated	−164.9	426.5	38.5

Figure 2.

Bar charts showing the proportion of species in each dorsal (left) and ventral (right) keel state. Species are divided by microhabitat: arboreal, saxicolous and terrestrial. Dorsal scale data are shown on the left and ventral scale data on the right. Ventral scale keel states are significantly correlated with microhabitat across agamids.

Figure 3.

Boxplots of mean range-wide annual precipitation in metres for species of each microhabitat type and ventral keel state. Only terrestrial species showed a significant within group relationship between ventral keel state and range-wide precipitation (**p < 0.01).

(c) . State dependent diversification rate

Our best-fit model for both dorsal and ventral keels was CID-2 (table 2). Thus, our data did not support an influence of scale keels on diversification rates in agamid lizards. Instead, we found support for an unmeasured character that drives changes in diversification rate across agamids. Similarly, we did not find support for an association between arboreality and diversification rates in agamids (best-fit = CID-2, ΔAICc = 2.8; electronic supplementary material, table S2)

Table 2.

HiSSE models comparing keeled and non-keeled scales with log-likelihoods, AICc, ΔAICc, and parameter estimates for speciation rate: non-keeled (λ₀), keeled (λ₁), and hidden states (λ_A and λ_B). The best-fit models for each comparison are bolded (ΔAICc = 0).

model	log-likelihood	AICc	ΔAICc	λ0A	λ1A	λ0B	λ1B
(a) dorsal scales
null	−808.0	1624.2	23.2	0.049	0.049	—	—
CID-2	−794.3	1601.0	0	0.075	0.075	2.1 × 10−9	2.1 × 10−9
dependent	−807.4	1625.1	24.1	0.041	0.052	—	—
full	−793.5	1608.3	7.3	0.071	0.003	0.009	0.081
(b) ventral scales
null	−807.7	1624.2	23.8	0.049	0.049	—	—
CID-2	−794.0	1600.4	0	1.3 × 10−8	1.3 × 10−8	0.075	0.075
dependent	−807.5	1625.1	24.7	0.047	0.051	—	—
full	−791.1	1603.5	3.1	2.1 × 10−9	2.1 × 10−9	0.074	0.079

4. Discussion

(a) . Ventral keels facilitate arboreality

Our results supported only one of our tested hypotheses. Specifically, we found evidence that keeled ventral scales facilitated the transition to arboreality in agamid lizards (table 1; figure 2). Ventral keels appear to have also facilitated arboreality in several snakes, although snake keels are often formed by compression of the ventral scale rather than as permanent macrostructures [33]. Regardless of the precise mechanism for forming the keel, our study demonstrates that arboreality has been repeatedly facilitated by a common scale shape in phylogenetically disparate squamates. Still, keeled scales have not universally evolved across arboreal squamates. Notably, chameleons lack such scale macrostructures [52] and keels are not universally present in arboreal geckos [53] and anoles [54]. This may be owing to the presence of alternative adaptations to arboreal microhabitats in such groups, specifically zygodactyly in chameleons [10] and adhesive toepads in geckos and anoles [55,56]. Additionally, the locomotor anatomy of chameleons prevents contact between the venter and substrate [22]. In context, our results support the idea that arboreality is facilitated by a complex mix of traits few to none of which are likely universally convergent across squamates.

There are various reasons keeled ventral scales might be advantageous for arboreal agamids. Ventral keels may reduce wear on the scale imposed by rough surfaces such as tree bark. Specifically, the keel may act as a buffer minimizing direct contact between the bulk of the scale and the substrate [29]. However, rocky surfaces are also likely to impose substantial wear, yet we do not observe an association between ventral keels and saxicolous agamids. Alternatively, ventral keels may facilitate attachment to vertical surfaces by interlocking with the substrate similarly to claws [31–33]. In this case, we would expect that keel prominence should increase with degree of arboreality mirroring the pattern of increased claw curvature in more arboreal lizards [16,18,57]. Unfortunately, our dataset precludes examination of this question at present. Thus, detailed scale surface morphometrics and finer-scale microhabitat data would likely prove fruitful for future research in this area. Nonetheless, our results support the hypothesis that keeled scales facilitated transitions to arboreal microhabitats in agamids.

Unlike other convergently evolved arboreal traits, like adhesive toepads, scale keels do not appear to have increased diversification rates despite facilitating transitions to a novel environment. Specifically, we did not observe increased diversification in agamids associated with the evolution of keeled scales (table 2). Like snakes [13,14], agamids also did not show elevated diversification rates associated with microhabitat in general (electronic supplementary material, table S2). Our sampling (186 species) covers approximately a third of agamid diversity limited by species included in the time-calibrated phylogeny. Thus, it is possible that a more complete sampling would influence our results. Still, our observed results may be because keeled scales do not allow for fine-scale microhabitat specialization in a manner similar to adhesive toepads in Anolis lizards and geckos [9,11]. That is, despite improving performance in arboreal ecological space for agamids, keels did not facilitate further division of that niche space into a greater diversity of microhabitat specialists rather than fewer arboreal generalists. Thus, although multiple phenotypic avenues to arboreality have evolved in squamates, it appears only a limited subset of these traits also produce elevated diversification rates.

(b) . Relationships with climate

For climate, we only observed a relationship between keeled ventral scales and higher precipitation. This patterns appears to have been driven by differences in terrestrial species as well as the tendency for arboreal species to occur in higher precipitation environments (figure 3). This climatic relationship is not one that we predicted a priori. Our initial hypothesis was that if keeled ventral scales were associated with precipitation, they would be less prevalent in more humid environments. Previous authors have speculated that because it is more difficult to remove debris from textured scales, wetter environments might worsen this effect [26,34]. However, recent evidence challenges the idea that species in more humid environments experience more debris pressure [35]. In any case, no hypothesis for why keeled ventral scales should be more common in wetter environments has been proposed to our knowledge.

We speculate that keeled ventral scales may be more common in wetter environment agamids, albeit only for terrestrial species, to reduce scale degradation. In snakes, keels have been shown to reduce wear on scales associated with their locomotion [29]. Furthermore, wear and decomposition rates are generally higher across biological materials in higher rainfall environments [58,59]. For example, bird feathers degrade faster in more humid environments which has selected for greater melanization to resist bacteria [60,61]. Thus, it is possible that keels in terrestrial agamids also protect their ventral scales from wear during contact with the substrate. This would likely be more advantageous in wetter environments where bacterial pressure on scale integrity is likely already higher [58–60]. Functional studies of scale macrostructures are required to clarify their relationship with precipitation.

5. Conclusion

We provide strong evidence supporting a link between the evolution of ventral keels and arboreality across agamids. Ventral scales are likely part of a diverse repertoire of traits that can be independently modified to improve performance in arboreal microhabitats across squamate reptiles. However, we found no evidence that keels are associated with temperature, nor do they appear to influence diversification rates in agamids. Thus, although there are many phenotypic paths to improve performance in a given ecological space, it appears not all phenotypes allow for subsequent diversification within that space. We also do not find evidence that dorsal keels are more prevalent in arid species despite their possible role in rain harvesting in at least some reptiles [36,37]. Future functional studies of scale macrostructures will be important to better understand their relationship with the environment.

Ethics

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

Data accessibility

All data in this study are available in the electronic supplementary material and cited published data sources.

Supplementary material is available online [62].

Declaration of AI use

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

Authors' contributions

M.L.Y.: conceptualization, data curation, formal analysis, investigation, methodology, project administration, visualization, writing—original draft, writing—review and editing; E.P.W.: conceptualization, data curation, formal analysis, investigation, methodology, project administration, visualization, writing—original draft, writing—review and editing.

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

Conflict of interest declaration

We declare we have no competing interests.

Funding

E.P.W. was supported by the National Science Foundation Graduate Research Fellowship (grant no. 1752814).