INFERRING PHYLOGENIES FROM mtDNA VARIATION: MITOCHONDRIAL-GENE TREES VERSUS NUCLEAR-GENE TREES REVISITED.

Abstract

The paper by Moore (1995) provided a theoretical basis for preferring mitochondrial-DNA(mtDNA) gene trees over nuclear-gene trees when estimating species phylogenies. This commentary examines conditions that contradict Moore's analysis, suggesting that mtDNA-gene trees for clades exhibiting female philopatry and male dispersal, or a polygynous mating system, may be less reliable indicators of a species phylogeny than nuclear-gene trees. Moore (1995) provides a clear discussion of the differences between gene trees and species trees. The fundamental reason for this distinction is that new alleles, or haplotypes, can originate at any point in time and are not constrained to begin evolutionary divergence at the time of speciation. Indeed, the fact that mutation generates polymorphism within a species is the foundation of population genetics. This point alone allows true branch lengths on a gene tree to differ from true branch lengths on the species tree for the same set of taxa. Of greater concern is that the true branching topology of a gene tree can also differ from that of the species tree when polymorphisms are maintained for long periods relative to internodal distances on the species tree (Moore 1995, fig. 1). If polymorphisms are resolved relatively quickly by the extinction and fixation of alternative alleles (i.e., lineage sorting sensu Avise et al. 1987), then the topology of the species tree will be imposed on the gene tree, because homologous gene sequences will coalesce in the most recent common ancestor of any two species. If, however, ancient polymorphisms coexisted for periods longer than the time between speciation events then the order of speciation may not be reflected in the gene phylogeny. This is expected because loss of particular lineages from descendant taxa should be random with respect to the age of the lineage. The loss of genetic lineages over a time period that includes a speciation event can occur by subsampling of ancestral polymorphisms in the original descendant population or by the subsequent extinction of alleles in descendant taxa. For the purpose of data analysis, undersampling extant intrataxonomic variation also mimics the effects of lineage sorting. Two strategies have been suggested to deal with the potential discordance between gene trees and species trees. The first recommends the simultaneous analysis of multiple unlinked genes (Pamilo and Nei 1988; Wu 1991). This strategy, however, is based on the assumption that the pattern of loss of alleles in descendant taxa is not correlated across loci. Although this may often be the case, there are also circumstances that would invalidate this assumption. For example, the subsampling of ancestral alleles during speciation may not be independent across unlinked loci if genetic variation was geographically structured in the ancestor (Hoelzer and Melnick 1994). Even when this critical assumption is true, the number of unlinked genes that must be examined to give one confidence in the topology can be prohibitive (Moore 1995). The second strategy, which is promoted by Moore (1995), is to use a gene that exhibits short coalescence times. This will maximize the chance that the historical durations of polymorphism retention will not exceed internodal distances on the species tree. Moore (1995) goes on to show that mtDNA is expected to exhibit a coalescence time that is 25% as long as for nuclear genes. This assertion is based on the expectation that effective population size of mtDNA (Nemit) is one-fourth that of nuclear genes (Nenuc) because the mitochondrial genome is both haploid and transmitted only through females. In a random mating population with an equal number of males and females, there will be four times more nuclear-gene copies passed on to the next generation because the nuclear genome is diploid, and alleles are contributed by both sexes. This is an excellent point with important implications for our interpretations of results in molecular systematics and our choices of approach to future studies; however, there are also other factors that can increase Nemit relative to Nenuc. Consideration of these factors shows that Nem it can exceed Nenuc and render a mtDNA tree a less reliable estimate of the species phylogeny than a nuclear-gene tree.