Computational protocol: How common is ecological speciation in plant-feeding insects? A 'Higher' Nematinae perspective

Similar protocols

Protocol publication

[…] Modeltest 3.5 [] was implemented in conjunction with PAUP* 4.0b10 [] to identify the least complex substitution model for use in Bayesian phylogenetic analyses in MrBayes 3.1.2 []. Hierarchical likelihood ratio tests indicated a GTR+I+Γ4 model as optimal for each of the three genes. A separate, unlinked substitution model was allowed for each gene in a three-partition analysis. A single run employing default priors was run for eight million generations with eight incrementally heated (t = 0.1) chains; tree sampling was done from the current cold chain every 100th generation, and the first 10,001 trees recovered prior to reaching stationarity were discarded as a burnin. The consensus tree showing all compatible groupings (Fig. ) was calculated on the basis of the remaining 70,000 trees. A corresponding maximum-likelihood (ML) analysis was performed using RAxML 7.0.4 []. This analysis employed a separate GTR+I+Γ4 model for each gene, but branch lengths were estimated jointly for the whole data (Additional file ). Clade support was estimated on the basis of 500 bootstrap replicates of the data matrix (Fig. ).BEAST 1.4.8 [] was used to estimate the relative ages of various nematine groups based on a Bayesian relaxed molecular clock method. The topologically unconstrained analysis allowed a separate GTR+I+ Γ4 model of substitution for each gene and employed an uncorrelated relaxed lognormal clock model for rate variation among branches, a Yule prior on speciation, and default priors for other parameters except for the mean of branch rates (ucld.mean), which was fixed to 1. Three independent runs with automatic tuning of operators were run for 80 million generations, and parameters and trees were sampled every 1,000 generations (the XML file is available as Additional file ). After inspection of adequate convergence of runs and effective sample sizes of the parameters in Tracer 1.4.1 [], the tree files were combined in LogCombiner 1.4.8 (part of the BEAST package). The first 40,000 trees from each file were discarded as a burnin, and the tree file was subsequently thinned by resampling trees every 3,000 generations; the maximum clade credibility (MCC) tree showing mean branch lengths (Fig. ) is based on the 40,001 post-stationarity trees that remained after thinning. [...] To reconstruct ancestral host-plant families and feeding habits, these traits were treated as unordered multistate characters and maximum-parsimony optimized on the phylogenetic trees using Mesquite 2.6 []. Oligo- and polyphagous taxa were coded with all used host families.To estimate the number of ecological shifts that have occurred during the radiation of the Higher Nematinae, we first identified all distinct ecological niches (feeding habit × host plant(s)) found in the ingroup species included in the phylogenetic analysis, and coded each niche with a separate state within a single character (outgroup states were coded as unknown). Because the aim was to calculate numbers of changes, the typical number of steps between two different states was 1. However, we also created 'generalist' states for species that utilize multiple plant taxa and then used the step-matrix option in Mesquite to define the cost to these states, from the plant taxa that are included within the generalist host range, as being zero. By doing so, we essentially assumed that a clear overlap in the host ranges of different species implies that they have not speciated ecologically (theoretical models of resource-based speciation typically assume distinct, non-overlapping niches as the cause of divergent selection [,]). Phylogenetic uncertainty in the estimate was taken into account by recording the numbers of steps in the niche character across the 70,000 post-burnin trees that were sampled by MrBayes during the phylogenetic analysis [].As a separate estimate of the proportion of lineage splits accompanied by a shift in resource use, we indentified all terminal sister-species pairs across the MCC tree (Fig. ), and then separated these 35 pairs into those in which both species have identical or overlapping niches, and into those in which the species have different niches. Thereafter, we performed a logistic regression in SPSS for Windows 17.0 (SPSS, Inc., 233 S. Wacker Drive, Chicago, IL 60606-6307, USA) to test whether the probability that sister species have a different niche depends on the time elapsed since their most recent common ancestor (= relative node height in the MCC tree).Proportions of higher nematine species feeding on different plant genera (Fig. ) were extracted from Lacourt's [] list of host-plant affiliations of sawflies of the Western Palearctic region. Only species with known hosts were included, and proportions were calculated separately for the tribe Pristiphorini and for the Nematini+Mesoneurini clade (see Figs. , and ). Oligo- and polyphagous species were counted as an additional species for each plant genus on which they feed (for example, the oligophagous Craesus latipes (Villaret) was treated as one species on Alnus and another on Betula). […]

Pipeline specifications

Software tools ModelTest-NG, PAUP*, MrBayes, RAxML, BEAST, Mesquite
Application Phylogenetics