Hemiplasy: A New Term in the Lexicon of Phylogenetics

Abstract

Homoplasy (trait similarity due to evolutionary convergence, parallelism, or character reversals) is a well-appreciated form of phylogenetic noise that systematists strive to identify and avoid when reconstructing species phylogenies. However, another source of phylogenetic “noise” is often neglected: the idiosyncratic sorting of gene-tree lineages into descendant taxa from character-state polymorphisms retained across successive nodes in a species tree. Here we introduce a term (hemiplasy) that formalizes a category of outcomes that can emerge from this evolutionary lineage-sorting phenomenon, and we make a case for why a wider recognition of hemiplasy (and attempts to ameliorate its complications) can play an important role in phylogenetics. The word homoplasy, meaning shaped (-plasy) in the same (homo-) way, refers to any trait correspondence or similarity not due to common ancestry. A central challenge in phylogenetic reconstruction is thus to distinguish the phylogenetic noise of homoplasy from the phylogenetic signal of homology (similarity in biological features due directly to shared ancestry). However, homology itself bears a subtle relationship to phylogeny, as emphasized by Willi Hennig (1950) more than a half-century ago. Hennig introduced the critical distinction between shared ancestral homology (symplesiomorphic similarity) and shared derived homology (synapomorphic similarity), noting that only the latter is indicative of monophyly within an organismal phylogeny. Hennig's cladistic insights fostered a fundamental revolution in phylogenetic principles and methodologies. The molecular revolution in biology that began at about that same time added further nuances to the homology concept. For example, DNA sequence homology in a multigene family can be due either to paralogy (similarity tracing to a gene duplication event) or to orthology (similarity tracing to an allelic separation within a particular locus). Orthology and paralogy are both genuine forms of genetic homology, but a failure to distinguish them in comparisons of DNA sequences can lead to errors in phylogenetic reconstruction. Phylogenetic jargon is already extensive but also important because words such as homoplasy, synapomorphy, and orthology capture and convey sophisticated evolutionary concepts that otherwise might remain opaque or underappreciated. In this spirit, here we formally define a new term—hemiplasy—for how the well-known phenomenon of idiosyncratic lineage sorting can lead to fundamental discordances between gene trees and organismal (species) trees. As will be described, hemiplasy is a bona fide form of homology (allelic orthology in this case) that nonetheless can give the illusion of homoplasy in an organismal tree. No other word or simple phrase currently exists to encapsulate the phenomenon that we will define under the suggested term. The nature of Mendelian heredity in sexually reproducing taxa ensures that alleles at unlinked loci transmit through an organismal pedigree via noncoincidental genealogical pathways across multiple generations. Thus, both within and among related species, the true topologies of gene trees inevitably differ somewhat from locus to unlinked locus (Ball et al., 1990). Furthermore, gene genealogies can in principle differ in basic topology from the overall population tree or species tree of which they are a part, if for no other reason than stochastic lineage sorting across successive evolutionary nodes in an organismal phylogeny. These concepts and their corollaries have been available for more than two decades (Hudson, 1983; Tajima, 1983; Takahata and Nei, 1985; Neigel and Avise, 1986), and they are encapsulated today in the widely appreciated distinction between gene trees and organismal trees (Doyle, 1992; Maddison, 1997; Nichols, 2001). Nevertheless, as noted by Liu and Pearl (2007), “the current molecular phylogenetic paradigm still reconstructs gene trees to represent the species tree.” Nei (1987: 401–403) summarized the theoretical probability of qualitative discordance between a gene tree and an organismal tree for the simple case of three related species (for more complex situations, see Rosenberg, 2002 and Degnan and Rosenberg, 2006). For selectively neutral alleles, the probability of the topological discordance illustrated in Fig. 1b is given by (2/3)e−T/2N, where T is the number of generations between the first and second speciation events (T = t1 – t2) and N is the effective population size (Fig. 1). This probability can also be interpreted as the percentage of unlinked neutral loci expected to show topological disagreement with the species tree. The formula supports general intuition by showing that a gene tree is most likely to “misrepresent” the topological structure of a species phylogeny when internodal times are short relative to effective population sizes. For example, the discordance probability is approximately 50% when T/2N = 0.3, but it is infinitesimally smal