Species are often treated as evolutionarily significant units of diversity that reflect patterns of gene flow and selection. In contrast, higher taxa are mostly regarded as convenient labels for levels in the tree of life, which reflect evolutionary history if defined cladistically but are assumed to have no real significance for ongoing evolution. We test the alternative hypothesis that some higher taxa are evolutionarily significant units with coherent patterns of selection on their constituent species. Specifically, we ask whether the big 4 orders of holometabolous insects, namely Coleoptera, Diptera, Hymenoptera, and Lepidoptera, display divergent, but internally conserved patterns of selection acting on protein-coding genes. Analyzing orthologous genes from whole genome sequence data for multiple species per order, we find that, in most genes, selection on roughly one fifth of codons is conserved within each order but differs significantly among orders. The shift is associated with variation in GC content among orders, but primarily at codon 2nd positions hence due to selection rather than mutational or repair bias. Comparison of alternative models assigning different taxonomic levels (either more lumped or divided than orders) shows that best models always specify Hymenoptera and Lepidoptera as coherent units, whereas patterns of selection on protein-coding genes within Coleoptera and especially Diptera are better explained by subdividing them further. We hypothesise that some aspect of the general lifestyle, body plan or genetic makeup of orders (or of nested clades within Coleoptera and Diptera) leads to conserved patterns of selection across protein-coding genes within them, whereas constraints differ among them. The emergence of whole-genome data for broad and deep phylogenetic samples will allow this hypothesis of evolutionarily significant higher taxa versus more evenly dispersed shifts in selection across genes to be tested further.
