28.4 n&v 1075 MH

eses.Trees built using single genes from many species tend to support L–E–D, but analyses using many genes from a few complete genomes support A–P–C. The number of species represented in a phylogenetic study can have two effects on tree reconstruction. First, without genomes to represent most animal phyla, genome-based trees provide no information on the placement of the missing taxonomic groups. Current genome studies do not include any members of the Lophotrochozoa. More notably, if a species’ genome is evolving rapidly, tree reconstruction programs can be misled by a phenomenon known as long-branch attraction. In long-branch attraction, independent but convergent changes (homoplasies) on long branches are misconstrued as ‘shared derived’ changes, causing artefactual clustering of species with long branches. Because these artefacts are systematic, confidence in them grows as more data are included, and thus genome-scale analyses are especially sensitive to long-branch attraction. Long branches can arise in two ways.One is when a distantly related organism is used as an ‘outgroup’ to root the tree of the organisms of interest. The other is when one organism of interest has a very different, accelerated pattern of evolution compared with the rest. Unfortunately for whole-genome studies,the usual outgroup,yeast, is very distantly related to animals, and C. elegans is a long-branch species. Long-branch attraction will therefore tend to result in nematodes moving to the base of the tree, generating erroneous support for A–P–C. Not all whole-genome studies are tainted: analysis of rare insertions and deletions of genomic features (introns) in some animal genomes,characters that may be immune to the insidious charms of longbranch attraction,does not support A–P–C. Philippe et al. have overcome these problems by using data from ‘expressed sequence tags’ (ESTs) in addition to complete genome sequences. Sequencing ESTs efficiently samples just the genes in any genome, avoiding the non-coding parts. The vastly lower cost of an EST project compared with sequencing a complete genome means that large numbers of ESTs have been generated for a much wider range of organisms, and we and others have been decorating the animal tree with EST data, including data from the neglected Lophotrochozoa. Using this expanded data set, Philippe et al. find convincingly in favour of L–E–D (Fig. 1). They include many more data than previous non-genomic studies (35,371 amino acids from 146 genes) and more species than genome studies (35 species representing 12 animal phyla and 14 outgroups including choanoflagellates, thought to be the protozoan phylum most closely related to animals). When only a distant outgroup (yeast) was used, nematodes emerge at the base of the tree. But with closer outgroups (protozoans related to animals, and jellyfish), nematodes cluster with arthropods, as predicted by the L–E–D hypothesis. In the complete data set, however, lophotrochozoan flatworms cluster with the ecdysozoan nematodes, and not with their supposed lophotrochozoan relatives (the molluscs and annelid worms). Suspecting that this was another longbranch artefact, Philippe et al. selectively eliminated genes expected to contribute most to long-branch attraction — those with a greater evolutionary rate in some species (such as nematodes) compared with others (such as deuterostomes). Indeed, as the most biased genes were removed, support for Ecdysozoa and Lophotrochozoa increased. Will this be the last, defining statement in the controversy? There remain some unresolved problems with Philippe and colleagues’ analysis, such as the position of the phylum Tardigrada (water bears). Tardigrades are unquestionably close to arthropods (they have eight stumpy legs), but appear as the sister phylum to the Nematoda. Have the nematodes lost the legs they once had,or are tardigrades misplaced? Additionally, only 12 of the 35 animal phyla are currently represented: will addition of more phyla — particularly lophotrochozoan phyla — change the tree significantly? Have coeloms in protostome and deuterostome animals very different developmental news and views