Ontogeny Discombobulates Phylogeny: Paedomorphosis and Higher-Level Salamander Relationships

Top Cited Papers
Open Access
Publisher Website
Google Scholar
Abstract
Evolutionary developmental biology (“evo-devo”) has revolutionized evolutionary biology but has had relatively little impact on systematics. We show that similar large-scale developmental changes in distantly related lineages can dramatically mislead phylogenetic analyses based on morphological data. Salamanders are important model systems in many fields of biology and are of special interest in that many species are paedomorphic and thus never complete metamorphosis. A recent study of higher-level salamander phylogeny placed most paedomorphic families in a single clade based on morphological data. Here, we use new molecular and morphological data to show that this result most likely was caused by the misleading effects of paedomorphosis. We also provide a well-supported estimate of higher-level salamander relationships based on combined molecular and morphological data. Many authors have suggested that paedomorphosis may be problematic in studies of salamander phylogeny, but this hypothesis has never been tested with a rigorous phylogenetic analysis. We find that the misleading effects of paedomorphosis on phylogenetic analysis go beyond the sharing of homoplastic larval traits by paedomorphic adults, and the problem therefore is not solved by simply excluding suspected paedomorphic characters. Instead, two additional factors are critically important in causing paedomorphic species to be phylogenetically “misplaced”: (1) the absence of clade-specific synapomorphies that develop during metamorphosis in nonpaedomorphic taxa and allow their “correct” placement and (2) parallel adaptive changes associated with the aquatic habitat of the larval stage. Our results suggest that the effects of paedomorphosis on phylogenetic analyses may be complex, difficult to detect, and can lead to results that are both wrong and statistically well supported by parsimony and Bayesian analyses.