Evolution of associations between Cymothoe butterflies and their Rinorea host plants in tropical Africa

This thesis aimed to elucidate the evolutionary history of the associations between Cymothoeforest butterflies (Nymphalidae, Limenitidinae) and their Rinoreahost plants (Violaceae) in tropical Africa. Insects are by far the most diverse group of multicellular organisms on earth. Because most insect species are herbivores, understanding the evolution of interactions between herbivorous insects and their host plants is therefore crucial to comprehend global patterns in terrestrial biodiversity. The Cymothoe-Rinoreasystem is especially suitable for untangling processes shaping patterns of insect-host plant associations because of its high level of specificity (mostly monophagous) and the large number of related species involved (33 insect herbivores and 32 hosts). Obviously, any evolutionary study relies on a solid classification and taxonomy of the organisms under study. Unfortunately, however, in Cymothoeas well as Rinorea, taxonomy and classification is still partly unresolved.

To improve taxonomy of Cymothoeand facilitate efficient identification of immature specimens found on Rinoreahost plants, we generated an extensive dataset of 1204 DNA barcode sequences (Chapter 2). Application of a novel taxonomic decision pipeline for integrating DNA barcodes with morphology and biogeography proved instrumental for solving taxonomic problems in Cymothoeand five taxa within Cymothoecould be confidently raised to species level. In addition, our DNA barcode data set allowed for the identification of 42 immature specimens from six different countries, significantly increasing the data on Cymothoehost plant associations. Nevertheless, our results also demonstrated that not all species of Cymothoecan be confidently delimited or identified. We hypothesize that this is probably due to incomplete lineage sorting and introgression (the latter possibly mediated through Wolbachiaendosymbionts) between recently diverged Cymothoespecies. In order to assess what are the best methods for matching DNA barcodes from recently diverged species, we compared six methods in their ability to correctly match DNA barcodes from selected published empirical data sets as well as simulated data (Chapter 3). Our results showed that, even though recently diverged species pose a significant problem for effective DNA barcoding, sensitive similarity-based and diagnostic methods can significantly improve identification performance compared with the commonly used tree-based methods.

To improve classification and clarify the biogeographic history of Rinorea, we presented an updated phylogenetic tree of Rinoreawith increased taxonomic sampling, using plastid as well as nuclear DNA sequences (Chapter 5). Phylogenetic relationships inferred from nuclear DNA data were generally congruent with those based on evidence from plastid haplotypes from earlier studies of Rinoreaand helped resolve additional clades, some of which warrant further taxonomic study. Divergence time estimations indicated that Rinoreaoriginated in the Neotropics and reached Africa in the Eocene through trans-Atlantic dispersal. From Africa, Rinoreasubsequently dispersed into Asia in the Oligocene or early Miocene, and colonized Madagascar multiple times independently within a relatively recent time scale (Pliocene), suggesting that factors governing the independent colonizations of Rinoreato Madagascar may have been similar.

In Chapter 4 we assessed whether differential rates of net species diversification in the African butterfly sister genera Harma(1 species) and Cymothoe(approximately 82 species) could best be explained by shifts to novel host plants (from Achariaceae to Rinorea) or by environmental factors such as changing climate. We generated the first time-calibrated species-level molecular phylogenetic tree of Harmaand Cymothoeand found that, after their divergence in the Miocene (15 Mya), net species diversification was low during the first 7 Myr. Coinciding with the onset of diversification of Cymothoein the late Miocene (around 7.5 Mya) there was a sharp and significant increase in diversification rate, suggesting a rapid radiation. This increased rate did not correlate with host plant transition from Achariaceae to Rinoreahost plants, but rather with a period of global cooling and desiccation, indicating that tropical forest fragmentation may well have driven the elevated diversification rates in Cymothoe.

Finally, in Chapter 6 we integrated the time-calibrated phylogenetic evidence from Cymothoeand Rinoreapresented in chapters 4 and 5 with updated host association records from the field, with the aim to distinguish between alternative scenarios for the evolution of insect-host plant associations. Our results showed that: (i) divergences among extant Cymothoeare more recent than those among their associated Rinoreahosts, suggesting asynchronous diversification of Cymothoeherbivores onto already diversified clades within African Rinorea; (ii) phylogenetic trees of Cymothoeand their associated Rinoreahost plants are discordant and current associations between Cymothoeherbivores and their Rinoreahosts have developed primarily through a process of host shifting rather than by cospeciation; and (iii) related Cymothoetend to feed on related Rinoreahosts. Based on the available data, we propose a recent origin of Rinorea{}-feeding by Cymothoebutterflies with a single colonization of pre-existing lineages in the late Miocene. Current associations are best explained by a predominance of shifts among related plants, probably due to constraints in larval physiology and oviposition behaviour. Overall, these findings are in agreement with a growing body of substantial evidence to suggest that divergences of herbivorous insects and their host plants are asynchronous, and that evolutionary dynamics of hosts and parasites do not favour cospeciation.

Insect-plant interactions are receiving increasing attention because of their importance in crop production and protection. At the same time, an increasing number of insects and plants that have evolved in separation are currently coming into contact due to human activities and climatic changes. It is therefore tempting to find implications of our findings for insect-host plant associations for agricultural systems (Chapter 7). Based on our results, one might predict that insects will only become pests of crops that are closely related to their natural host. Extrapolating our findings to an agricultural setting is difficult, however, because of the difference in selective pressures between natural and agricultural ecosystems.