A genome Tree of Life for the Fungi kingdom

Abstract

Fungi belong to one of the largest and most diverse kingdoms of living organisms. The evolutionary kinship within a fungal population has so far been inferred mostly from the gene-information-based trees ("gene trees"), constructed commonly based on the degree of differences of proteins or DNA sequences of a small number of highly conserved genes common among the population by a multiple sequence alignment (MSA) method. Since each gene evolves under different evolutionary pressure and time scale, it has been known that one gene tree for a population may differ from other gene trees for the same population depending on the subjective selection of the genes. Within the last decade, a large number of whole-genome sequences of fungi have become publicly available, which represent, at present, the most fundamental and complete information about each fungal organism. This presents an opportunity to infer kinship among fungi using a whole-genome information-based tree ("genome tree"). The method we used allows comparison of whole-genome information without MSA, and is a variation of a computational algorithm developed to find semantic similarities or plagiarism in two books, where we represent whole-genomic information of an organism as a book of words without spaces. The genome tree reveals several significant and notable differences from the gene trees, and these differences invoke new discussions about alternative narratives for the evolution of some of the currently accepted fungal groups.

Keywords: alignment-free method; divergence tree; feature frequency profile; fungal phylogeny; proteome tree.

Fig. 1.

A Circos (topological) representation of the proteome tree of Fungi kingdom. The branches of three major groups are colored in light green for group I (Monokaryotic fungi), red for group II (Basidiomycota), and purple for group III (Ascomycota). All protists are in blue. The branches of two sets of outgroups are in black. The names of nine groups at phylum level belonging to the three major groups are shown around the circle. The four marked (by lowercase alphabets in parentheses) groups with dotted-lined branches are the groups whose placements in the proteome tree are significantly different from those in the gene trees, as discussed in Results. The taxon identification numbers can be found in Fig. S1A, and their taxon names can be found in Table S1. For the identities of the outgroups, see Materials and Methods. The branch lengths are relative and not to scale. The figure was prepared using the Interactive Tree of Life (ITOL) (43).

Fig. 2.

Simplified proteome tree of Fungi and Protozoa. The figure shows the proteome tree collapsed at the phylum or equivalent levels with the relative branch lengths from one common ancestor of a clade to its previous common ancestor. (The branch lengths for the two outgroups and uncollapsed species are not shown.) For the statistical support of the collapsed groups, the Jackknife Monophyly Index (5) for each collapsed clade (except the two outgroups) are shown under the branch lines. The branch lengths calculated by JSD are normalized to 1,000 (the scale on top), which corresponds to 500 from the common ancestor of fungi and protists to the terminal leaves. The number of the members in a clade is indicated at the end of the clade name, and the four marked (by lowercase alphabets in parentheses) groups are the groups whose placements in the proteome tree are significantly different from those in the gene trees, as discussed in the Results. The clade colors correspond to those in Fig. 1. For the identities of the out-group, see Materials and Methods. The tree was constructed using ITOL (43).

Fig. S1.

Comparison of the proteome tree and a gene tree of the Fungi kingdom. List of taxonomic identifications (taxonIDs) are included in the linear trees The corresponding names of organisms can be found in Table S1. (A) The proteome tree represented in a linear form corresponds to the tree of the circular form in Fig. 1. The branch lengths calculated by JSD (33) are scaled to 1,000 (which corresponds to 500 from the common ancestor of fungi and protists to the terminal leaves) and shown above the branch lines. All of the Jackknife Monophyly Index values are 1.00, except those shown below the branch lines. (B) A gene tree of Fungi downloaded from MycoCosm of the JGI fungal portal (36, 37). The colored bars in the gene tree indicate the fungal organisms used in our proteome tree of Fig. 1 and A, and uncolored regions are the fungi whose genome sequences are not released to the public at the time of our study, and thus not used in the proteome tree. Each clade of Ascomycota, Basidiomycota, and Monokaryotic fungi (any fungi other than Ascomycota and Basidiomycota) are constructed independently using the “FastTree” program and then joined manually (see genome.jgi.doe.gov/ext-api/mycocosm/clustering/clm/r/fungi.42/all/2014_Mycocosm_All-Fungi_tree.png).

Fig. S2.

Topological stability and branch length distribution in three types of genome trees. (A). Optimization of tree topology as a function of feature length. Topological variation, as represented by the Robinson-Foulds metric (13) between a pair of trees, is shown on the y axis, and feature length, l, on the x axis. The trees are constructed using a divergence matrix calculated by JSD (33) for all pair-wise FFPs, one for feature length of l and the other for feature length of l + 1. Graphs are smoothed using a spline function using statistical R (44). This experiment is done by sampling 200 fungal species from 244 fungi used in this study for computational expediency. This figure shows that, among three types of genomic information (whole-genome DNA sequence, transcriptome RNA sequence, and proteome amino acid sequence), the proteome sequence-based proteome trees converge to the most topologically stable tree, starting from the lowest point in the curve, and remain stable. (B). Distribution of pair-wise JSDs between two fungi using three types of genomic sequence information (whole-genome DNA sequence, transcriptome RNA sequence, and proteome amino acid sequence). A pair-wise divergence between two fungi, as represented by their respective sequence FFPs for an optimal feature length, is calculated by JSD between the two FFPs. The optimal feature lengths used for proteome, transcriptome, and genome sequences are 13, 23, and 23, respectively. Such JSDs are calculated for all pairs in the study population and assembled into a divergence matrix. The distribution pattern of the divergence are represented by the count (in log₁₀ scale) of pairs having a given divergence magnitude on the y axis and the magnitude of the divergence on the x axis. Since highly crowded long distances cause ambiguity in distinguishing long branches, a phenomenon analogous to long-branch attraction effect (45) in gene trees, the topology of the proteome tree will have the least-crowded long branches. Thus, proteome produce the most stable tree topology among the three types of genome trees.