Parsimony network analysis of rDNA sequences was used to delimit phylogenetic species of yeasts in an objective, formal manner. Many strains assigned to Candida apicola (Starmerella clade), when compared to the type, fell outside the inclusion limits proposed by Kurtzman and Robnett (1998) based on a pair-wise comparison of the large subunit rRNA gene D1/D2 domains. However, when these sequences were analyzed jointly with ITS rDNA sequences by parsimony network analysis, 28 of the 30 strains formed a cohesive set. Two strains, MUCL 45721 and CBS 4353, were excluded from the species, but there was no evident justification to subdivide the rest. A similar analysis of 81 isolates originally assigned to Candida azyma (Wickerhamiella clade) yielded dramatically different results, giving rise to six independent networks corresponding to Candida azyma sensu stricto (18 strains), Candida azymoides (2 strains), a pair of isolates from Australian hibiscus flowers, a single isolate from the same substrate, a single isolate from Malaysian bertam palm nectar, and 57 isolates that are assigned to the new species Candida parazyma (type = UWOPS 91-652.1(T) = CBS 11563(T) = NRRL Y-48669(T)). The strains retained in C. azyma sensu stricto differed from one another by up to four substitutions in their D1/D2 sequences, but their polymorphism at the level of the ITS was considerable and suggested a history of divergence resulting from dispersal. Strains of C. parazyma fell into seven variant haplotypes based on sequences of the rDNA ITS and D1/D2 regions. The most abundant haplotype occurred across the global range of the species. Others were either endemic to Belize, Costa Rica, Rarotonga, or Tennessee, suggestive of vicariance, or occurred across remote localities, offering partial support to the notion of rapid dispersal.

The relationship between mating success and sequence divergence in the internal transcribed spacer (ITS)/5.8S-D1/D2 rDNA region was examined in isolates tentatively assigned to Metschnikowia agaves and Starmerella bombicola. Both species are haplontic and heterothallic, such that the formation of mature asci can be used as a measure of genetic compatibility. Parsimony haplotype network analysis and mating success confirmed that all known isolates of M. agaves are conspecific. The previously reported D1/D2 polymorphism of five substitutions was not corroborated; the maximum divergence observed between any two strains was three substitutions, four with ITS. Of 39 putative S. bombicola strains, 36 formed an ITS-D1/D2 haplotype network using the 95% criterion. Thirty-five strains could mate with one or more compatible partner. The excluded strains did not mate. Mature asci arose from crosses between individuals differing by as many as five, but not six or seven substitutions in the D1/D2 domain. All strains capable of mating formed mature asci with at least one partner and all network members could be linked to another member by three or fewer substitutions. These results support the use of sequence divergence as a criterion for species delineation, but caution against describing poorly sampled species solely on the basis of that criterion.