Dispersal is a critical factor determining the spatial scale of speciation, which is constrained by the ecological characteristics and distribution of a species' habitat and the intrinsic traits of species. Endogean taxa are strongly affected by the unique qualities of the below-ground environment and its effect on dispersal, and contrasting reports indicate either high dispersal capabilities favoured by small body size and mediated by passive mechanisms, or low dispersal due to restricted movement and confinement inside the soil. We studied a species rich endogean ground beetle lineage, Typhlocharina, including three genera and more than 60 species, as a model for the evolutionary biology of dispersal and speciation in the deep soil. A time calibrated molecular phylogeny generated from >400 individuals was used to delimit candidate species, to study the accumulation of lineages through space and time by species-area-age relationships, and to determine the geographical structure of the diversification using the relationship between phylogenetic and geographic distances across the phylogeny. Our results indicated a small spatial scale of speciation in Typhlocharina and low dispersal capacity combined with sporadic long distance, presumably passive dispersal events that fuelled the speciation process. Analysis of lineage growth within Typhlocharina revealed a richness plateau correlated with the range of distribution of lineages, suggesting a long-term species richness equilibrium mediated by density dependence through limits of habitat availability. The interplay of area and age dependent processes ruling the lineage diversification in Typhlocharina may serve as a general model for the evolution of high species diversity in endogean mesofauna.
Phylogeny_of_Typhlocharina_RAxML_BestTree
ML trees were obtained using RAxML 7.2.7 (Stamatakis 2006). Data sets were partitioned by gene and the protein-coding genes (cox1-a and cox1-b fragments) were additionally partitioned by separating the third codon positions (Andújar et al. 2012). An independent GTR + G model was applied to each data partition. The best scoring ML tree was selected among 200 searches on the original alignment with different randomized parsimony starting trees. Support values were obtained with 1000 bootstrap replicates (Felsenstein 1985). BI was run in MrBayes 3.2.3 (Huelsenbeck et al. 2001) on the concatenated data, partitioned by gene and codon as before, and for each partition the optimal substitution model was selected using the Akaike information criterion (AIC) in jModelTest 2.1.7 (Darriba et al. 2012).
Phylogeny_of_Typhlocharis_RAxML_BestTree.tree
Phylogeny_of_Typhlocharina_MrBayes_MajorityRuleConsensus
50% majority rule consensus tree obtained in MrBayes. Bayesian inference consisted of two independent runs, each with three hot and one cold chain, for 10 million generations, whereby trees were sampled every 1000 generations. The standard deviation of split frequencies was checked to assess the convergence of results, as well as the mean and effective sampled size (ESS) of likelihood values computed with TRACER 1.6 (Rambaut et al. 2013). The 50% majority rule trees were calculated excluding 50% of the initial trees as a conservative burn-in, ensuring that the plateau in tree likelihood values had been reached.
Phylogeny_of_Typhlocharis_MrBayes_MajorityRuleConsensus.tree
Phylogeny_of_Typhlocharis_BEAST_dated_tree_(Rates_ULN_bd)_Complete
Dated phylogenetic tree obtained with BEAST 1.8.1 (Drummond et al. 2012). For the analyses the original concatenated dataset was collapsed to haplotypes and the outgroups were excluded, resulting in a dataset of 330 terminals. We used as calibration prior for each gene an uniform function on the mean substitution rate encompassing the 95% confident interval values obtained for the same DNA fragments (cox1-a, cox1-b, rrnL, LSU and SSU genes) in the confamilial genus Carabus (Andújar et al., 2014, 2012) and the tribe Anillini (Andújar et al., 2016) (Table 1 in original paper). We additionally applied a gamma prior on the root age for the most recent common ancestor (crown group) of Typhlocharina. We used TreeStat 1.6.1 (Rambaut & Drummond 2010) to recover the node age of the ancestor of Typhlocharina from the sample of the MCMC search of the BEAST analysis favoured in Andújar et al. (2016), and we used the “fitdistr” option of the R package MASS to obtain the values of a gamma function adjusting the distribution of sampled ages, thus accounting for the uncertainty associated to the age estimations (Table 1 on the original paper).
Analyses were conducted under the model of substitution best fitting to each gene and codon partition as above, and were run for 50 million generations sampling one tree in every 5,000 generations. In the molecular clock settings gene partitions were uncorrelated, and two independent analyses were conducted applying an uncorrelated lognormal (ULN) to all genes. Analyses were conducted under a Birth-Death model (BD) and with the inclusion of the constraint on the age for the ancestor of Typhlocharina (Table 1 on the original paper)
Phylogeny_of_Typhlocharis_BEAST_dated_tree_(Rates_ULN_bd)_Pruned_to_species
Time calibrated tree pruned to keep a single tip branch per species.
