Should I stay or should I go? Pollinator shifts rather than cospeciation dominate the evolutionary history of South African Rediviva bees and their Diascia host plants.

Plant-pollinator interactions are often highly specialised, which may be a consequence of co-evolution. Yet when plants and pollinators co-evolve, it is not clear if this will also result in frequent cospeciation. Here, we investigate the mutual evolutionary history of South African oil-collecting Rediviva bees and their Diascia host plants, in which the elongated forelegs of female Rediviva have been suggested to coevolve with the oil-producing spurs of their Diascia hosts. After controlling for phylogenetic nonindependence, we found Rediviva foreleg length to be significantly correlated with Diascia spur length, suggestive of co-evolution. However, as trait correlation could also be due to pollinator shifts, we tested if cospeciation or pollinator shifts have dominated the evolution of Rediviva-Diascia interactions by analysing phylogenies in a cophylogenetic framework. Distance-based cophylogenetic analyses (PARAFIT, PACo) indicated significant congruence of the two phylogenies under most conditions. Yet, we found that phylogenetic relatedness was correlated with ecological similarity (the spectrum of partners that each taxon interacted with) only for Diascia but not for Rediviva, suggesting that phylogenetic congruence might be due to phylogenetic tracking by Diascia of Rediviva rather than strict (reciprocal) co-evolution. Furthermore, event-based reconciliation using a parsimony approach (CORE-PA) on average revealed only 11-13 cospeciation events but 58-80 pollinator shifts. Probabilistic cophylogenetic analyses (COALA) supported this trend (8-29 cospeciations vs. 40 pollinator shifts). Our study suggests that diversification of Diascia has been largely driven by Rediviva (phylogenetic tracking, pollinator shifts) but not vice versa. Moreover, our data suggest that, even in co-evolving mutualisms, cospeciation events might occur only infrequently.