Computational protocol: Digits Lost or Gained? Evidence for Pedal Evolution in the Dwarf Salamander Complex (Eurycea, Plethodontidae)

Similar protocols

Protocol publication

[…] The dwarf salamanders are one of the three plethodontid taxa characterized by loss of a single digit on the pes (). Distributed throughout the southeastern Coastal Plain, dwarf salamanders were considered to represent a single species, Eurycea quadridigitata , until a distinct color morph was elevated to species status (E. chamberlaini) . Further, one of us (DAB) noted separate topological placements for eastern (South Carolina) versus western (Texas) E. quadridigitata in a phylogeny for the paedomorphic Eurycea that constitute the Edwards Plateau complex . To examine lineage diversity among dwarf salamanders more fully, we generated DNA sequence data on 120 individuals, representing dense geographic sampling (88 localities, ) across the range of the E. quadridigitata complex (). To explore phylogenetic relationships of dwarf lineages relative to the genus overall, we surveyed 15 additional species (representing the remaining four species complexes in Eurycea), including eight species in the Edwards Plateau complex (). Outgroup taxa included Gyrinophilus porphyriticus, Pseudotriton ruber, Stereochilus marginatus, and Urspelerpes brucei , which, together with Eurycea, represent all genera within the tribe Spelerpini . Specimens were maintained and euthanized following standard procedures approved by East Carolina University’s Animal Care and Use Committee, outlined expressly for this survey (Animal Use Protocol # D247).We sequenced portions of two mitochondrial genes–NADH dehydrogenase subunit 2 (Nd2, 1,020 bp) plus an adjacent transfer RNA (tRNAtrp, ∼70 bp), and cytochrome b (Cytb, 1,012 bp)–for all specimens. We also sequenced three additional genes for a subset of dwarf salamanders (n = 23) representing phylogeographic lineages identified by our initial mtDNA dataset. These loci, chosen for slightly to substantially slower evolutionary rates relative to Cytb, included another mitochondrial gene, 16S ribosomal RNA (16 s, 529 bp), and two nuclear genes, pro-opiomelanocortin (Pomc, 536 bp) and recombination activating gene 1 (Rag1, 1131 bp). Amplification primer sets and cycling conditions are listed in . Sequences were generated on an Applied Biosystems 3130 capillary machine and aligned in CLUSTAL X 1.81 . Protein-coding sequences were translated to ensure appropriate reading frames. Regions of the 16 s alignment for which nucleotide position homologies varied across gap parameter settings were excluded, yielding a slightly smaller final dataset (512 bp). Genbank accession numbers are listed in . [...] We analyzed two concatenated datasets (1 =  mitochondrial genes Cytb+Nd2+tRNAtrp, and 2 =  all-genes) using Bayesian inference (BI) and likelihood (ML) methods. We identified nucleotide substitution models for each gene for BI, partitioning protein-coding genes by codon position and assessing gene/codon partitions by the Bayesian Information Criterion . We implemented BI analysis in MrBayes 3.1.2 , , involving two concurrent runs of four simultaneous Markov Chain Monte Carlo (MCMC) chains for ten million generations, with a sample frequency of 1,000 generations. Topologies in the first 25% of the posterior distribution were discarded as burn-in, and the remaining trees were summarized as a majority consensus. Convergence of model parameters and topology were assessed by the program Are We There Yet (AWTY) .ML analyses were executed in RAxMLHPC v7.2.8 , employing the rapid hill-climbing algorithm . Parameters for the analyses incorporated the GTRGAMMA model of evolution, and 100 random addition sequence replicates were conducted. Branch support was computed via 100 non-parametric bootstrap replicates .The degree to which individual gene sequences support (or are discordant with) clades identified by concatenated data reflects gene-tree heterogeneity. Concatenation approaches focused at the level of a species complex can generate misleading results due to incomplete lineage sorting, introgression, or deep coalescences . In some of these cases, species tree inference methods (e.g., Bayesian concordance, coalescent models) can outperform data concatenation analyses . Thus, we also estimated a species tree using a two-step Bayesian concordance analysis (BCA) and a multispecies coalescent model, implemented in BEST ver. 2.3 .For BCA, we generated posterior probability distributions for the gene tree of each locus separately using MrBayes 3.1.2 (4 MCMC chains; 5 million generations). Upon discarding the first 4 million generations from each locus run, we used BUCKy (Bayesian Untangling of Concordance Knots) v 1.2 b to construct a primary concordance tree from the posterior distributions obtained for these loci. BUCKy also generates concordance factors, which represent the proportion of genes supporting a given clade. We conducted two BCA runs; in the first, mitochondrial genes were analyzed separately, for a total of five loci. In light of linkage, however, we combined the mitochondrial gene sequences in the second BCA, for a total of three loci (a single mitochondrial linkage unit; 2 nuclear genes).We used BEST v 2.3 to estimate a species tree that accounts for deep coalescence. Each phylogeographic lineage of the dwarf salamander complex was treated as a separate species for the BEST analysis, which ran for 120 million generations, with a sample frequency of 1000 generations. We used a uniform prior (0, 3) for gene mutation estimate and modeled the effective population size with an inverse gamma distribution (α = 3, β = 0.1). Convergence of model parameters and topology were assessed using AWTY . [...] We examined character state history of digit number (4 vs. 5 toes) using MCMC methods implemented in the program BayesTraits V1.0 (www. To test the contrasting hypotheses of digit loss among the dwarf salamanders (parallelism) versus digit gain in the Edwards Plateau complex (re-evolution), we used the all-genes dataset to reconstruct the ancestral state for the node subtending the dwarf-Edwards clade. Reconstruction involved the reversible-jump model, with an exponential prior seeded from a uniform on the interval 0 to 30. We set the ratedev parameter to 8 (which, in conjunction with the previously identified prior, produced acceptance rates in the desirable (15–40%) range) and ran the analysis for 100 million iterations.For a second assessment (again, using the all-genes dataset), we employed the fossilize command in BayesTraits, implementing two constraint analyses–the first fixing the dwarf-Edwards node at five toes (i.e., the parallelism hypothesis), the second at four toes (re-evolution hypothesis). We compared harmonic means for the two hypotheses using the Bayes factors statistic, where 2(lnL H1 − lnL H2) is the Bayes factor (BF), with a BF >2 interpreted as positive support and BF >5, strong support . […]

Pipeline specifications

Software tools Clustal W, MrBayes, AWTY, BUCKy, BayesTraits
Application Phylogenetics