Phylogenetic inferences of pollinator and Ficus sequence data were used for relative divergence timing estimation and for tests of congruence between them. To generate phylogenetic inferences among the wasps, fragments of mtDNA cytochrome oxidase one (COI ~ 620 bp), cytochrome b (Cytb ~ 380 bp), and nuclear DNA (nDNA) elongation factor-one alpha F2 copy (EF-1α ~ 500 bp) were sequenced. A comprehensive explanation of DNA extraction, PCR, and alignment protocols is given in [81 (link)]. The species delimitation test requires an (mtDNA) ultrametric tree. Substitution rates of COI mtDNA tend to be higher than in nuclear protein coding genes and therefore more prone to saturation bias that impedes deep node resolution. In order to implement reasonable prior tree topology constraints for ultrametric pollinator tree inference and for co-phylogenetic analyses, we used species for which multiple gene fragments including nDNA were available to infer a phylogeny using Bayesian and parsimony approaches. Sequence data of up to 767 bp's of a ribosomal internal transcribed spacer (ITS) and up to 479 bp of external transcribed spacer (ETS) were used to infer Ficus species phylogenies under Bayesian and parsimony methods. Analyses presented in this work assume the Ficus species phylogeny is fully resolved and does not consider population-level genetic variation influence on host associations. Species-level appraisal of host lineages does provide evidence of comparative genetic distances for instances of departures from one-to-one species specificity. Sequences of ITS and ETS were amplified using the protocol outlined in previous work [80 (link)].
A Bayesian approach implemented in MrBayes 3.1.1 [82 (link)] was used to partition the COI, Cytb, and EF-1α data into gene fragments and also codon positions. The Ficus sequence data was partitioned into ITS and ETS for the Bayesian phylogenetic analyses. A general time reversible DNA substitution model was used with gamma distributed (+G) rates with a proportion of invariant sites (+I). Posterior probabilities and mean branch lengths were derived from 15000 post-burnin trees sampled every 1000 trees from generations 5 to 20 million. Four separate Bayesian reconstructions were run to verify consistency of phylogenetic outcomes. The consensus trees were derived from post-burnin generations of Markov chains that had reached apparent stationarity. Convergence was assessed using the MCMC Tracer Analysis Tool v.1.4.1 [83 ] by plotting the log likelihoods to assess the point in the chain where stable values were reached and with the standard deviation of split frequencies of all runs. Parsimony bootstrap analysis implemented using PAUP version 4.0b10 [84 ] was used to assess the robustness of the Bayesian consensus phylogeny. The parsimony bootstrap consensus trees were derived from a search consisting of 500 bootstrap replicates using a full heuristic search. To calculate bootstrap support, we used branch-swapping by stepwise addition on best trees only, 100 random additional sequences holding 10 trees at each step, and the TBR search algorithm.
A Bayesian approach implemented in MrBayes 3.1.1 [82 (link)] was used to partition the COI, Cytb, and EF-1α data into gene fragments and also codon positions. The Ficus sequence data was partitioned into ITS and ETS for the Bayesian phylogenetic analyses. A general time reversible DNA substitution model was used with gamma distributed (+G) rates with a proportion of invariant sites (+I). Posterior probabilities and mean branch lengths were derived from 15000 post-burnin trees sampled every 1000 trees from generations 5 to 20 million. Four separate Bayesian reconstructions were run to verify consistency of phylogenetic outcomes. The consensus trees were derived from post-burnin generations of Markov chains that had reached apparent stationarity. Convergence was assessed using the MCMC Tracer Analysis Tool v.1.4.1 [83 ] by plotting the log likelihoods to assess the point in the chain where stable values were reached and with the standard deviation of split frequencies of all runs. Parsimony bootstrap analysis implemented using PAUP version 4.0b10 [84 ] was used to assess the robustness of the Bayesian consensus phylogeny. The parsimony bootstrap consensus trees were derived from a search consisting of 500 bootstrap replicates using a full heuristic search. To calculate bootstrap support, we used branch-swapping by stepwise addition on best trees only, 100 random additional sequences holding 10 trees at each step, and the TBR search algorithm.
Free full text: Click here
