The Bucculatricid problem: Phylogeny of the Gracillarioidea inferred from 21 protein-coding genes. A problem of compositional heterogeneity?

Gracillarioidea (approximately 2,000 described species) is the most diverse superfamily of leaf-mining moths, with many economically important agricultural pests. While the majority of species are leaf miners, the group shows a diversity of other life-history strategies, such as fruit mining, stem mining, leaf rolling, boring, and galling. Despite their economic importance and wealth of life-history strategies, relationships among the four gracillarioid families (Bucculatricidae, Douglasiidae, Gracillariidae, and Roeslerstammiidae) remain uncertain. Fifty-seven taxa, including twelve outgroups, were initially sequenced for ten nuclear protein-coding genes (8,436 bp). An additional 11 genes (6,375 bp) were sequenced for 27 taxa and combined with the original ten to create a data set of 14,811 bp. The concatenated, all taxa, all-gene data set and three other data sets of different taxa and gene sampling design were analyzed with parsimony, maximum-likelihood, and Bayesian inference. Partially or fully augmenting a data set with more characters tended to increase branch support for particular deep nodes, and this increase was dramatic when non-synonymous changes were analyzed alone. Supporting a recent study, there was strong evidence for the exclusion of Douglasiidae from Gracillarioidea, as monophyly of the superfamily was statistically rejected in eight of nine analyses (AU test, P ≤ 0.009). Results support the monophyly of Gracillariidae, Lithocolletinae + Leucanthiza, and the Acrocercops and Parectopa groups. There was strong support for the ‘G.B.R.Y.’ clade, a group comprising of the Gracillariidae + Bucculatricidae + Roeslerstammiidae + Yponomeutidae, when analyzed with non-synonymous changes only. Base compositional heterogeneity may explain the spurious position of Bucculatricidae when synonymous changes are included.