Bromham L: Statistical tests between competing hypotheses of Hox cluster evolution

Abstract.—The Hox genes encode transcription factors that play vital roles in the anterior-posterior patterning of all bilaterian phyla studied to date. Additionally, the gain of Hox genes by duplication has been widely implicated as a driving force in the evolution of animal body plans. Because of this, reconstructing the evolution of the Hox cluster has been the focus of intense research interest. It has been commonly assumed that an ancestral four-gene ProtoHox cluster was duplicated early in animal evolution to give rise to the Hox and ParaHox clusters. However, this hypothesis has recently been called into question, and a number of alternative hypotheses of Hox and ParaHox gene evolution have been proposed. Here, we present the first statistical comparisons of current hypotheses of Hox and ParaHox gene evolution. We use two statistical methods that represent two different approaches to the treatment of phylogenetic uncertainty. In the first method, we estimate the maximum-likelihood tree for each hypothesis and compare these trees to one another using a parametric bootstrapping approach. In the second method, we use Bayesian phylogenetics to estimate the posterior distribution of trees, then we calculate the support for each hypothesis from this distribution. The results of both methods are largely congruent. We find that we are able to reject five out of the eight current hypotheses of Hox and ParaHox gene evolution that we consider. We conclude that the ProtoHox cluster is likely to have contained either three or four genes but that there is insufficient phylogenetic signal in the homeodomains to distinguish between these alternatives. [Bayesian; Homeobox; maximum likelihood; phylogenetics; ProtoHox; ParaHox.