Eusociality, Darwin’s special difficulty, has been widely investigated but remains a topic of great debate in organismal biology. Eusocial species challenge existing theories, and the impact of highly integrated societies on diversification dynamics is controversial with opposing assertions and hypotheses in the literature. Here, using phylogenetic approaches in termites – the first group that has evolved eusociality – we assessed the fundamental prediction that eusocial lineages have higher diversification rates than non-eusocial clades. We found multiple lines of evidence that eusociality provided higher diversification as compared to non-eusociality. This is particularly exacerbated for eusocial species with ‘true’ workers as compared to species with ‘false’ workers. Because most species with ‘true’ workers have an entirely prokaryotic microbiota, the latter feature is also related to higher diversification rates, but it should be investigated further, notably in relation to angiosperm diversification. Overall, this study suggests that societies with ‘true’ workers are not only more successful at ecological timescales but also over millions of years, which further implies that both organism- and species-level traits act on species selection.
Figure S1 - Partitioning Isoptera final
Figure S1. Partitioning the phylogeny of Dictyoptera into evolutionary scenarios. The null scenario is the entire dictyopteran tree. Then each time an evolutionary scenario is created, it consisted into a subclade and the rest of the tree composed of all dictyopterans without the subclade (i.e. here called the backbone). The numbers at nodes denotes all seven subclades. The first scenario excludes the entire termite (Isoptera, node 1). The second scenario excludes the Euisoptera (node 2). The third scenario excludes the Kalotermitidae + Neoisoptera (node 3). The fourth scenario excludes the Neoisoptera (node 4). The fifth scenario excludes the Neoisoptera minus the subfamily Rhinotermitinae (node 5). The sixth scenario excludes the Termitidae + Coptotermitinae + Heterotermitinae (node 6). The seventh scenario excludes the family Termitidae (node 7).
Figure S2 - Prior and posterior probabilities for shifts
Figure S2. Frequency distribution of distinct macroevolutionary rate regimes estimated for the Dictyoptera using BAMM with a Poisson prior of 10. A scenario including five shifts of diversification has the highest posterior probability Similar results are obtained with different values of the Poisson prior (Fig. S6).
Figure S3 - Marginal probabilities and marginal odds ratios of the BAMM run
Figure S3. Two different metrics of weighing the relative evidence of a diversification shift occurring along any individual branch. (a) The phylogenetic tree has its branch lengths been replaced by the branch-specific marginal shift probabilities, i.e. the length of a given branch is equal to the percentage of samples from the posterior that contain a rate shift on that particular branch. (b) The phylogenetic tree has the branch lengths scaled to equal the corresponding marginal odds ratio accounting for the effects of the prior and branch length. The longest branch length, in both trees, is labelled for reference. In both cases, the longest branch is within the termites.
Figure S4 - Credible set of diversification shifts in Dictyoptera
Figure S4. Credible set of configuration shifts for net diversification of Dictyoptera inferred with BAMM. Phylogenies show the distinct shift configurations with the highest posterior probability. For each shift configuration, the locations of rate shifts are shown as black circles, with circle size proportional to the marginal probability of the shift. Text labels (e.g. f=0.11) denote the posterior probability of each shift configuration.
Figure S5 - Best shift configuration in diversification shifts in Dictyoptera
Figure S5. The best shift configuration for net diversification shifts inferred with BAMM. The phylogeny indicates five core rate shifts (indicated by red-filled circle) within the termites, two within the Mantodea, and two within the cockroaches (notably in Blaberidae, Panesthiinae). The first shift in mantises occurred early (166.8 Ma) and includes most of the mantises but leaves out all ancient and depauperate lineages near the root of the tree. The second shift within mantises is located at the base of a clade including notably the plant-mimicking mantises (Empusidae), the flower mantises (Hymenopodidae), and the core of praying mantises (Mantidae). The first shift in cockroaches is ancient (243 Ma), just after the Permian-Triassic extinction, corresponds to the divergence of the termite lineage (including Cryptocercidae and Lamproblattidae, two species-poor families of cockroaches) and the lineage leading to Blattidae. The second shift occurred at the K-Pg event (66 Ma), and involves an offshoot of the blaberid radiation (Panesthiinae, ca. 150 species). The most important shift occurred within termites, 66 Ma, at the node sustaining the radiation of Termitidae and a part of Rhinotermitidae. Interestingly this lineage is composed only by eusocial species with true workers (Fig. S1).
Figure S6 - BAMM analyses to check for prior effect
Figure S6. All the results for the BAMM analyses using a range of values for the Poisson prior process (0.1/0.5/1/5/10/50). The figures are ranked as follows: (1) prior and posterior distribution, and (2) best configuration. Overall, the results indicate a similar macroevolutionary mixture of rate shifts and diversification rates, with a significant shift within termites and a higher net diversification for termites.
Figure S7 - Differences BiSSE MCMC
Differences between speciation (a), extinction (b), and net diversification rates (c), computed from the MCMC analyses using the best BiSSE model. No posterior distribution overlaps with the red line, indicating that the differences between all rates are significant. Eusocial lineages diversified faster than non-eusocial lineages.
Figure S8 - Differences MuSSE worker states macroevolutionary rates MCMC
Differences in rates computed from the MCMC analyses using the best MuSSE model for the ‘true’ workers. The compound figure shows: 1) the differences between speciation rates as made by pair comparisons; 2) the differences between extinction rates; and 3) the differences between net diversification rates. Only the lineages with ‘true’ workers have a significant different speciation, extinction, and net diversification rates (i.e. no posterior distribution overlaps with the red line). Eusocial lineages with ‘true’ workers diversified faster than non-eusocial lineages and eusocial lineages with pseudergates.
Figure S9 - Differences MuSSE hindgut microbiota macroevolutionary rates MCMC
Differences in rates computed from the MCMC analyses using the best MuSSE model for the gut microbiota. The compound figure shows: 1) the differences between speciation rates as made by pair comparisons; 2) the differences between extinction rates; and 3) the differences between net diversification rates. Only the lineages with an entirely prokaryotic gut microbiota have a significant different net diversification rates (i.e. no posterior distribution overlaps with the red line). Eusocial lineages with an entirely prokaryotic hindgut microbiota diversified faster than other lineages (cellulolytic flagellates or no specialized microbiota for lignocellulose digestion).
Figure S10 - Simulations with SSE models
Randomization tests for (a) BiSSE, (b) MuSSE on the worker state, and (c) MuSSE on the hindgut microbiota. The difference of fit between the best model and the reference model is shown with the red vertical line for real data, and in vertical coloured bars for the distribution of simulated data. The tests show that our results are robust to type-I error.
Figure S11 - MuSSE hindgut microbiota Wood-feeding non-eusocial
Results for the MuSSE analyses with the wood-feeding non-eusocial lineages re-coded as cellulolytic flagellates along with plots of the differences in rates computed from the MCMC analyses using the best MuSSE model. The compound figure shows: 1) the differences between speciation rates as made by pair comparisons; 2) the differences between extinction rates; and 3) the differences between net diversification rates. Only the lineages with an entirely prokaryotic hindgut microbiota have a significant different net diversification rates (i.e. no posterior distribution overlaps with the red line). Eusocial lineages with an entirely prokaryotic hindgut microbiota diversified faster than other lineages (cellulolytic flagellates or no specialized microbiota for lignocellulose digestion).
Figure S12 - MuSSE hindgut microbiota rates Macrotermitinae
Results for the MuSSE analyses with the Macrotermitinae re-coded with ‘lower’ termites, due to their simple hindgut structure, along with plots of the differences in rates computed from the MCMC analyses using the best MuSSE model. The compound figure shows: 1) the differences between speciation rates as made by pair comparisons; 2) the differences between extinction rates; and 3) the differences between net diversification rates. Only the lineages with a complex hindgut structure have a significant different net diversification rates (i.e. no posterior distribution overlaps with the red line). Eusocial lineages with a complex hindgut structure diversified faster than other lineages (eusocial and non-eusocial organisms with simple hindgut structure)
Table S1 - BAMM analyses
Summary of diversification models in BAMM compared across a gradient of values for the Poison process governing the number of rate shifts. The analyses show that a scenario with either four, five or six shifts of diversification best explained the macroevolution of Dictyoptera. Although it might suggest that the Poison prior has some effect on the analyses, it is important to note that all shifts are in common between the analyses (see Fig. S6 for more details). Based on the effective sample size (ESS) and the highest posterior distribution per number of shifts for the Poisson rate prior, we selected the model with a Poison prior of 10 (highlighted in bold). In addition this model represents an intermediary among the analyses (i.e. we did not favour a model with more shifts or a model with few shifts).
Table S2 - SSE analyses
Supports for eusociality, eusocial lineages with ‘true’ workers, and eusocial lineages with entirely prokaryotic microbiota as driver(s) of diversification. Tables (a), (c) and (e) report the means for each value estimated based on 100 dated trees, for each diversification model applied under the BiSSE approach, their number of parameters (NP), the log-likelihood (logL), the corrected Akaike Information Criterion (AICc), the difference of AICc (ΔAIC) between the best model (lowest AIC) and a given model. The best model is highlighted in bold and determined by ∆AIC. Tables (b), (d) and (f) report the standard errors for each parameter estimated by the diversification models and listed in the tables (a), (c) and (e).