Data from: Coincident transitions across elevation and origins of functional innovations drove the phenotypic and ecological diversity of lungless salamanders
Data files
Dec 31, 2025 version files 52.76 KB
-
Burress_et_al_supporting_data.csv
43.89 KB
-
README.md
8.88 KB
Abstract
Ecological opportunity (EO) is an important catalyst for evolution. Whereas theory often centers around a lineage encountering a source of EO in isolation, in practice, they experience numerous sources of opportunity, either concurrently or sequentially. Such multiplicity can obscure the macroevolutionary signature of EO. Here, we test the effects of elevation (a proxy of the “mountain effect”) and an array of functional innovations on the evolutionary history of plethodontid salamanders, a diverse and charismatic radiation of lungless amphibians. Functional innovations unlock access to novel microhabitats, ultimately enabling sub-lineages to occupy a diverse range of ecological niches, particularly in lowland areas where those niches are more abundant. Consistent with expanded ecological opportunity, such transitions to lower elevation result in rapid phenotypic evolution. At high elevation, by contrast, rates of phenotypic evolution and phenotypic disparity decline, reflecting a loss of phenotypically extreme ecological specialists. Transitions in elevation and the origin of innovations appear largely coincident among lungless salamanders, suggesting myriad sources of EO. The magnitude of the “mountain effect” on evolutionary rates (~10-fold) is on par with or greatly exceeds that of islands, lakes, and coral reefs on other iconic vertebrate radiations. Therefore, we find that elevation acts as a major ecological moderator and, in concert with functional innovations, shapes the ecological and phenotypic diversity of lungless salamanders.
Authors: Edward D. Burress, Meaghan Gade, Eric A. Riddell, and Martha M. Munoz
correspondence: Edward D. Burress; edburress@ua.edu
The associated Dryad repository contains five files (one on Dryad; four on Zenodo)
(1) A .csv file (Burress_et_al_supporting_data.csv) that contains the supporting data used for the analyses in the manuscript
This file contains 13 columns. Column abbreviations are as follows: species names (x), snout-vent length (SVL), snout length (SnL), head length (HL), body width (BW), forelimb length (FLL), hindlimb length (HLL), tail length (TL), mean elevation (ln_elevation_mean), median elevation (ln_elevation_midpoint), microhabitat designation (microhabitat), alternative microhabitat designation (alternative_microhabitat), and notes (notes).
All numerical values are species means.
Column 1 contains species names.
Columns 2-8 are reported in ln-transformed millimeters.
Columns 9-10 are reported in ln-transformed meters.
Columns 11-12 are categorical microhabitat categories (note that not all species have an alternative microhabitat designation)
Column 13 contains any relevant notes pertaining to microhabitat designations (note that not all species have notes)
Supplementary File
(2) a .pdf file (Supplementary_Materials_salamander) that contains the supplementary materials referenced in the manuscript.
Table S1. Model fitting among transition models that describe microhabitat evolution in lungless salamanders. The row in bold is the best-fitting model according to AICc. Abbreviations: Equal rates (ER), symmetrical rates (SYM), all rates different (ARD), sample size corrected Akaike information criterion (AICc).
Figure S1. Species richness (left panel) and elevation (right panel) across the distribution of plethodontid salamanders. Illustrations depict a species from the indicated geographic region. The inset in the right panel depicts species richness across the elevational gradient.
Figure S2. Evolutionary history of the elevational (midpoint) distribution of plethodontid salamanders (A). The ancestral states were estimated with Maximum likelihood in the contMap function in phytools (Revell 2012). Distribution of elevation (midpoint) among species (B). Bars reflect the three discretization schemes used for elevation described in the text: the lower and upper 50th percentiles, lowland (<1068.5m) and highland (>1068.5m); the lower, mid, and upper 33rd percentiles, lowland (<796.3m), midland (7.96.3-1465.6m), and highland (>1465.6m); and four quartiles, lowland (<640m), midlow (641-1069m), midhigh (1070m-1689m), and highland (>1690m). Compare with the results shown in Figures S6-8.
Figure S3. Diagram depicting the assessment of coincident and delayed rate shifts. Hypothetical scenario in which a functional innovation (shown in red) is a synapomorphy of a clade (highlighted in gray). An ancestral state reconstruction would highlight the most recent common ancestor of the clade as the origin of the innovation (delineating the “focal clade”); however, the true origin would naturally lie along the rootward branch (bc). Therefore, we consider rate shifts along bc as coincident with the origination of the functional innovation. To also account for evolutionary lag, we consider delayed rate shifts along up to two tipward branches (bd). The same assessment was applied to microhabitat transitions.
Figure S4. Statistical independence of rate and state using limb length as an example (reduced limbs). We interrogated the non-independence of phenotype and rate directly following the same workflow used in the main text (i.e., the rate-by-state analyses depicted in Figure 2). The multivariate evolutionary rate of forelimb length is not correlated with limb length (residuals) using phylogenetic contrasts or tip rates (all P>0.446). Therefore, although limb length is a trait used to calculate the multivariate evolutionary rate in the main text, its evolutionary rate is independent of its state (i.e., it is statistically unproblematic as an innovation; ‘reduced limbs’).
Figure S5. Trait correlations for all pairs of traits, as estimated with MuSSCRat (May & Moore, 2020). These correlations were used to inform the simulated datasets (see main text). During this procedure, we arbitrarily used the values from the model with a prior of 100 rate shifts; trait correlations from the model with the lowest and highest prior on the number of rate shifts are shown to demonstrate that the estimates are equivalent. The dashed line depicts a 1-to-1 relationship.
Figure S6. Rates of salamander phenotypic evolution across elevation quartiles, estimated with a random local clock model in MuSSCRat (May and Moore 2020) across different prior numbers of rate shifts. Colors depict the elevation quartile: lowland (blue), midlow (yellow), midhigh (orange), and highland (red). All models have posterior probability = 1.0.
Figure S7. Rates of salamander phenotypic evolution across elevation quartiles, estimated with an uncorrelated lognormal model in MuSSCRat (May and Moore 2020). Colors depict the elevation quartile: lowland (blue), midlow (yellow), midhigh (orange), and highland (red). Posterior probability = 1.0. Compare with the results using a Random Local Clock model depicted in Figure 2a in the main text.
Figure S8. Sensitivity of the MuSSCRat analyses to different numbers of states in the discrete character (i.e., different solutions when discretizing elevation). Using 2, 3, or 4 discrete states to represent elevation resulted in unanimous support for faster rates of phenotypic evolution in lowlands. Posterior probability (PP) that the rates are state-dependent is inset for each model.
Figure S9. Branch-specific background rates of phenotypic evolution across the plethodontid phylogeny. Background rates reflect rate heterogeneity not attributable to elevation (compare with Figure 1 in the main text).
Figure S10. Trait-specific rates of evolution in lungless salamanders, as estimated with MuSSCRat. Distributions reflect the range of estimates across 100,000 iterations. Rates are summarized from a single model, but the trait-specific rates were visually identical across all models.
Figure S11. The origins of functional innovations across the plethodontid phylogeny. Posterior probability (PP) of innovation-associated rate shifts is labeled with text. Reference Figure 1 for tip names. Note that the rate shift associated with miniaturization in Thorius (light blue; PP=0.86) is not labeled because the shift occurred in rates of background variation (not rates as they respond to elevation as shown here).
Figure S12. Posterior probability (PP) of rate shifts across models with different priors on the number of shifts (A). The support for rate shifts waned with priors of fewer rate shifts, but rate shifts along innovation-associated branches were concentrated in the 99th and 95th percentiles across models with different priors (A). Posterior support for most branches followed this pattern, as they fell below the 1:1 line (dashed lines in B). Compare to Figure 3 in the main text.
Figure S13. Microhabitat and phenotypic diversity across the elevational gradient ent summarized (averaged) across six different coding schemes presented by Baken and Adams (2019): the number of microhabitat specialists (a) and phenotypic disparity (b). The x-axis depicts the four elevation quartiles, from low (Q1 to high (Q4).
Figure S14. Distribution of traits across microhabitat types (mean±95% C.I.). With only one exception (wide bodies in arboreal species), the phenotypic extremes of each trait are found in a microhabitat type that is lost at high elevation.
Figure S15. Ancestral state reconstruction of microhabitat use in lungless salamanders (A). Pies at internal nodes denote the proportion of character states across 1000 stochastic histories. Note that “cave” is painted the same color across the plot, but was coded as a polymorphic state (i.e., associated with several microhabitats; see Figure 1). Transition matrix showing the relative transition rates among character states (B; warmer colors reflect higher transition rate).
Code/Software
(3 a .Rev file (musscrat_script_multivariate_RLC.Rev) including the script to run the multivariate, state-dependent musscrat model with a random local clock.
(4a.Nex file (musscrat_script_multivariate_UCLN) including the script to run the multivariate, state-dependent musscrat model with an uncorrelated lognormal clock.
(5 a .R file (RevBayes_log_reader_script.R) including the script to extract parameters from a RevBayes output log file (for use in other programs such as R).