Phylogenetic evidence for the evolution of ecological specialization in Timema walking‐sticks

We used phylogenetic and ecological information to study the evolution of host‐plant specialization and colour polymorphism in the genus Timema , which comprises 14 species of walking‐sticks that are subject to strong selection for cryptic coloration on their host‐plants. Phylogenetic analysis indicated that this genus consists of three main lineages. Two of the lineages include highly generalized basal species and relatively specialized distal species, and one of the lineages comprises four specialized species. We tested for phylogenetic conservatism in the traits studied via randomizing host‐plant use, and the four basic Timema colour patterns, across the tips of the phylogeny, and determining if the observed number of inferred changes was significantly low compared to the distribution of numbers of inferred changes expected under the null model. This analysis showed that (1) host‐plant use has evolved nonrandomly, such that more closely related species tend to use similar sets of hosts and (2) colour pattern evolution exhibits considerable lability. Inference of ancestral states using maximum parsimony, under four models for the relative ease of gain and loss of plant hosts or colour morphs, showed that (1) for all models with gains of host‐plants even marginally more difficult than losses, and for most optimizations with gains and losses equally difficult, the ancestral Timema were generalized, feeding on the chaparral plants Ceanothus and Adenostoma and possibly other taxa, and (2) for all models with gains of colour morphs more difficult than losses, the ancestral Timema were polymorphic for colour pattern. Generation of null distributions of inferred ancestral states showed that the maximum‐parsimony inference of host‐plant generalization was most robust for the most speciose of the three main Timema lineages. Ancestral states were also inferred using maximum likelihood, after recoding host‐plant use and colour polymorphism as dichotomous characters. Likelihood analyses provided some support for inference of generalization in host‐plant use at ancestral nodes of the two lineages exhibiting mixtures of generalists and specialists, although levels of uncertainty were high. By contrast, likelihood analysis did not estimate ancestral colour morph patterns with any confidence, due to inferred rates of change that were high with respect to speciation rates. Information from biogeography, floristic history and the timing of diversification of the genus are compatible with patterns of inferred ancestral host‐plant use. Diversification in the genus Timema appears to engender three main processes: (1) increased specialization via loss of host‐plants, (2) retention of the same, single, host‐plant and (3) shifts to novel hosts to which lineages were ‘preadapted’ in colour pattern. Our evidence suggests that the radiation of this genus has involved multiple evolutionary transitions from individual‐level specialization (multiple‐niche polymorphism) to population‐level and species‐level specialization. Ecological studies of Timema suggest that such transitions are driven by diversifying selection for crypsis. This paper provides the first phylogeny‐based evidence for the macroevolutionary importance of predation by generalist natural enemies in the evolution of specialization.