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

An accurately resolved gene tree may not be congruent with the species tree because of lineage sorting of ancestral polymorphisms. DNA sequences from the mitochondrially encoded genes (mtDNA) are attractive sources of characters for estimating the phylogenies of recently evolved taxa because mtDNA evolves rapidly, but its utility is limited because the mitochondrial genes are inherited as a single linkage group (haplotype) and provide only one independent estimate of the species tree. In contrast, a set of nuclear genes can be selected from distinct chromosomes, such that each gene tree provides an independent estimate of the species tree. Another aspect of the gene-tree versus species-tree problem, however, favors the use of mtDNA for inferring species trees. For a three-species segment of a phylogeny, the branching order of a gene tree will correspond to that of the species tree if coalescence of the alleles or haplotypes occurred in the internode between the first and second bifurcation. From neutral theory, it is apparent that the probability of coalescence increases as effective population size decreases. Because the mitochondrial genome is maternally inherited and effectively haploid, its effective population size is one-fourth that of a nuclear-autosomal gene. Thus, the mitochondrial-haplotype tree has a substantially higher probability of accurately tracking a short internode than does a nuclear-autosomal-gene tree. When an internode is sufficiently long that the probability that the mitochondrial-haplotype tree will be congruent with the species tree is 0.95, the probability that a nuclear-autosomalgene tree will be congruent is only 0.62. If each of k independently sampled nuclear-gene trees has a probability of congruence with the species tree of 0.62, then a sample of 16 such trees would be required to be as confident of the inference based on the mitochondrial-haplotype tree. A survey of mtDNA-haplotype diversity in 34 species of birds indicates that coalescence is generally very recent, which suggests that coalescence times are typically much shorter than internodal branch lengths of the species tree, and that sorting of mtDNA lineages is not likely to confound the species tree. Hybridization resulting in transfer of mtDNA haplotypes among branches could also result in a haplotype tree that is incongruent with the species tree; if undetected, this could confound the species tree. However, hybridization is usually easy to detect and should be incorporated in the historical narrative of the group, because reticulation, as well as cladistic events, contributed to the evolution of the group. © 1995 The Society for the Study of Evolution.