Computational protocol: Diversification and Demography of the Oriental Garden Lizard (Calotes versicolor) on Hainan Island and the Adjacent Mainland

Similar protocols

Protocol publication

[…] Nucleotide sequences were checked and aligned by eye. No premature stop codons were observed in the ND2 and COI regions, suggesting that the sequences were not nuclear copies of mitochondrial sequences. We used DnaSP v5.0 to calculate the number of haplotypes and haplotype diversity (h) and to estimate nucleotide diversities (π) for the total sample and for each population ().To estimate the phylogeny and divergence times for the mitochondrial haplotypes, we conducted a Bayesian analysis using the program BEAST v1.5.4 . Using a relaxed molecular clock enabled us to estimate divergence times in the presence of rate heterogeneity among lineages . Estimating molecular divergence times requires a priori assumptions about the age of one or more clades to calibrate the substitution rate. This is most commonly achieved by using fossil calibrations. In the absence of fossil calibrations for C. versicolor, we expanded the data set to include various agamid taxa from GenBank (). We also included sequences from the C. versicolor group from the Central Dry Zone of Myanmar . We partitioned the data set by gene and by codon position, producing seven partitions. The most appropriate evolutionary model was selected for each partition using the hierarchical likelihood-ratio test implemented in Modeltest v3.7 .We estimated the phylogeny and divergence times of 28 outgroup taxa and two randomly sampled haplotypes that spanned the root of the C. versicolor tree (). The use of deep fossil calibrations can lead to overestimates of recent divergence times and molecular rates, probably owing to an excess of deleterious mutations at the population level , . Nonsynonymous sites show stronger time-dependent pattern of rates than synonymous sites, resulting in overestimates of molecular rates on short timescales , . Therefore, we used a two-step approach similar to that employed by Pepper et al. . We began by analysing the third codon sites of mitochondrial protein-coding sequences, which reduces the impact of purifying selection on estimates of evolutionary timescales . Our first step was to use a calibration for the divergence of Laudakia microlepis and L. caucasia. This split was constrained to be 8–10 Ma, assuming vicariance based on mountain uplifts caused by the collision of the Indian and Arabian plates . A lognormal prior distribution (mean = 0, standard deviation = 0.354, offset = 8) was chosen to reflect the uncertainty in the fossil calibration age. Posterior distributions of parameters were estimated using Markov chain Monte Carlo (MCMC) sampling. Samples were drawn every 1,000 steps over a total of 5×107 MCMC steps. We used a Yule tree prior with an uncorrelated lognormal relaxed clock .In the second step of our analysis, we ran a partitioned BEAST analysis including all 212 sequences of C. versicolor, using the age estimate for the most recent common ancestor of the two C. versicolor lineages from the first step (). This calibration was implemented as a lognormal prior, with mean = 0.57, log (stdev) = 0.72, and offset = 0.03. A constant-size coalescent prior was used for the tree. The optimal substitution model was selected for each data partition. Samples from the posterior were drawn every 1,000 steps over a total of 2×107 MCMC steps. We conducted four replicates of each BEAST analysis and compared their results to check for MCMC convergence and acceptable mixing. The results were found to be satisfactory and samples from the multiple runs were combined. We regarded posterior probabilities ≥0.95 as being indicative of strongly supported nodes . [...] To characterize population structure and to define groups of populations using genetic criteria, we conducted spatial analysis of molecular variance with the program SAMOVA 1.0 . The analysis was run for values of K from 2 to 20 and the significance of fixation indices was tested using 1,000 random permutations. We chose the optimal number of groups based on two criteria. First, from a plot of FCT as a function of K, we determined the value of K necessary for FCT to reach a plateau. Second, we excluded the configurations of K that had one or more single-population groups, because this indicated that the group structure was disappearing , .To test further for geographical genetic structure, analyses of molecular variance (AMOVA) with 10,000 permutations were assessed in Arlequin v3.1 , according to the degree of differentiation between regions (FCT), between populations within regions (FSC), and between all populations (FST). We examined several hypotheses, including: (i) no overall regional structure; (ii) structure between Hainan Island and its adjacent mainland (the “Qiongzhou Strait isolation” hypothesis); (iii) structure between the mountain and plain provinces on Hainan Island (the “two zoogeographical regions” hypothesis); and (iv) structure caused by the Changhua-Wanquan River () presenting a barrier between the northwestern and southwestern groups. Separate AMOVA analyses were performed for the full data set and for the different lineages identified in the phylogenetic analysis, using the substitution model selected by the hierarchical likelihood-ratio test to calculate the distance matrix. Population groupings for each analysis are detailed in , with locations shown in .To test for associations between genetic and geographical distance among populations or within lineages, isolation-by-distance analyses were performed using IBDWS v3.16 . Significance was evaluated using 10,000 random permutations. Considering the two-dimensional habitats of C. versicolor across the sampled region, pairwise values of FST/(1−FST) were plotted against the logarithm of geographical distance (two-dimensional stepping-stone model) . FST values were calculated with Arlequin 3.1 using the TrN model of evolution selected by the hierarchical likelihood-ratio test. Reduced major axis regression was used to estimate the slope and intercept of the isolation-by-distance association. To include geographical structuring among populations, a third matrix (indicator matrix) was included so that populations were grouped by regions based on the results of AMOVA and SAMOVA tests and the “Qiongzhou Strait isolation” hypothesis. Partial Mantel tests took these groups into account while determining the significance of genetic and geographical distance relationships. […]

Pipeline specifications

Software tools DnaSP, BEAST, ModelTest-NG, SAMOVA, Arlequin, IBDWS
Applications Phylogenetics, Population genetic analysis
Organisms Calotes versicolor