Detecting Macroevolutionary Self-Destruction from Phylogenies

Phylogenetic analyses have lent support to the concept of lineage selection: that biological lineages can have heritable traits that influence their capacity to persist and diversify, and thereby affect their representation in biodiversity. While many discussions have focused on "positive" lineage selection, where stably heritable properties of lineages enhance their diversification rate, there are also intriguing examples that seem to represent "negative" lineage selection, where traits reduce the likelihood that a lineage will persist or speciate. In this article, we test whether a particular pattern of negative lineage selection is detectable from the distributions of the trait on a phylogeny. "Self-destructive" traits are those that arise often but then disappear again because they confer either a raised extinction rate or they are prone to a high rate of trait loss. For such a trait, the reconstructed origins will tend to be dispersed across the tips of the phylogeny, rather than defining large clades of related lineages that all share the trait. We examine the utility of four possible measures of "tippiness" as potential indicators of macroevolutionary self-destruction, applying them to phylogenies on which trait evolution has been simulated under different combinations of parameters for speciation, extinction, trait gain, and trait loss. We use an efficient simulation approach that starts with the required number of tips with and without the trait and uses a model to work "backwards" to construct different possible trees that result in that set of tips. We then apply these methods to a number of case studies: salt tolerance in grasses, color polymorphism in birds of prey, and selfing in nightshades. We find that the relative age of species, measured from tip length, can indicate a reduced speciation rate but does not identify traits that increase the extinction rate or the trait loss rate. We show that it is possible to detect cases of macroevolutionary self-destruction by considering the number of tips with the trait that arise from each inferred origin, and the degree to which the trait is scattered across the phylogeny. These metrics, and the methods we present, may be useful for testing macroevolutionary hypotheses from phylogenetic patterns.