Is Penicillium monophyletic? An evaluation of phylogeny in the family Trichocomaceae from 18S, 5.8S and ITS ribosomal DNA sequence data

Abstract

Using ribosomal DNA sequences from 17 fungi, we evaluated the monophylly of Penicillium and the phylogeny in the ascomycete family Trichocomaceae (= Eurotiaceae). We determined the 5.8S and internal transcribed spacer rDNA sequences for 13 species, and the 18S rDNA sequences for Eupenicillium javanicum and Neosartorya fischeri. After including additional published 18S, 5.8S and ITS rDNA sequences, our data set consisted of 2204 aligned sites for 17 taxa and yielded 16 equally parsimonious trees, each requiring 609 nucleotide substitutions. Indicating that Penicillium is not monophyletic, all the trees show Eupenicillium javanicum (with a Penicillium state) grouping with the Aspergillus-producing taxa Neosartorya fischeri and Eurotium rubrum rather than the Penicillium-producing Talaromyces flavus var. macrosporus or Talaromyces bacillisporus. The branch separating Eupenicillium javanicum from Talaromyces flavus and Talaromyces bacillisporus received 98% bootstrap support. Providing further statistical evidence that Penicillium is not monophyletic, constraint trees forcing the penicillia to appear monophyletic were significantly less likely (using the maximum likelihood program DNAML) than trees allowing the penicillia to cluster into two groups. Aspergillus may be monophyletic, nested among species with Paecilomyces and Penicillium states. Monascus purpureus, although usually placed in a separate family, grouped with Eupenicillium and the Aspergillus species in the Trichocomaceae. Species in the Onygenales are basal to the Trichocomaceae. Coding regions varied little in length, were readily alignable and provided most of the evolutionary “signal” present in the data set, while the spacer regions varied in length from 309 to 447 base pairs and encompassed much unalignable data. We found evidence of higher levels of mutational saturation in the internal transcribed spacer regions than in the coding regions when we compared the pair-wise percent substitution, with and without correction for repeated substitutions at the same site, for the two data subsets.