- Research article
- Open Access
A fungal phylogeny based on 82 complete genomes using the composition vector method
BMC Evolutionary Biologyvolume 9, Article number: 195 (2009)
Molecular phylogenetics and phylogenomics have greatly revised and enriched the fungal systematics in the last two decades. Most of the analyses have been performed by comparing single or multiple orthologous gene regions. Sequence alignment has always been an essential element in tree construction. These alignment-based methods (to be called the standard methods hereafter) need independent verification in order to put the fungal Tree of Life (TOL) on a secure footing. The ever-increasing number of sequenced fungal genomes and the recent success of our newly proposed alignment-free composition vector tree (CVTree, see Methods) approach have made the verification feasible.
In all, 82 fungal genomes covering 5 phyla were obtained from the relevant genome sequencing centers. An unscaled phylogenetic tree with 3 outgroup species was constructed by using the CVTree method. Overall, the resultant phylogeny infers all major groups in accordance with standard methods. Furthermore, the CVTree provides information on the placement of several currently unsettled groups. Within the sub-phylum Pezizomycotina, our phylogeny places the Dothideomycetes and Eurotiomycetes as sister taxa. Within the Sordariomycetes, it infers that Magnaporthe grisea and the Plectosphaerellaceae are closely related to the Sordariales and Hypocreales, respectively. Within the Eurotiales, it supports that Aspergillus nidulans is the early-branching species among the 8 aspergilli. Within the Onygenales, it groups Histoplasma and Paracoccidioides together, supporting that the Ajellomycetaceae is a distinct clade from Onygenaceae. Within the sub-phylum Saccharomycotina, the CVTree clearly resolves two clades: (1) species that translate CTG as serine instead of leucine (the CTG clade) and (2) species that have undergone whole-genome duplication (the WGD clade). It places Candida glabrata at the base of the WGD clade.
Using different input data and methodology, the CVTree approach is a good complement to the standard methods. The remarkable consistency between them has brought about more confidence to the current understanding of the fungal branch of TOL.
Fungi make up one of the major Eukaryotic kingdoms besides the Plantae and Animalia. These heterotrophic organisms possess a chitinous cell wall and grow as single cells or as multicellular mycelium made of hyphae. Although some species are not capable of forming specialized reproductive structures and propagate solely by vegetative growth, many fungi reproduce sexually and asexually via spores. To date, around 70 000 fungal species have been described while the total number of species has been estimated at 1.5 million [1, 2].
Since the early 1990s, the introduction of molecular characters has drastically revised the traditional fungi phylogenetic system based on morphology, physiology and sexual states. Numerous works have addressed cladistic relationships among all major groups of the kingdom [2–5]. Molecular characters have shown great power when morphological characters are convergent, reduced, or missing among the taxa. So far most fungal molecular phylogenetic inferences have been established on alignment of single or several orthologous gene loci [3, 6]. When multi-locus data are investigated, the commonly adopted methods are gene concatenation and consensus tree analysis . Since more genes provide more phylogenetic information, many recent phylogenomic studies tried to infer phylogenies for various organisms by combining large datasets of aligned genes (or ESTs) [7–9]. For example, Robbertse et al  and Fitzpatrick et al  built phylogeny from large datasets of protein-coding genes of 17 and 42 genomes, respectively.
These methods have achieved great success in the last two decades. However, some well-documented stochastic or systematic errors in tree reconstruction often lead to incongruent results [11, 12]. Furthermore, their applications depend on manual selection of many parameters and fine adjustments of sequence data. For example, at least at some stages, the standard methods select and process genes (and sites) to avoid systematic errors . These problems broach a question of principle: the phylogeny based on sequence alignment needs an independent verification in order to put the fungal TOL on a more secure footing. Recently, methods based on other strategies such as gene content, gene order and the distribution of oligonucleotides or peptides have been proposed to infer phylogenies (see  and references therein), which have made the verification feasible.
We have constructed a kingdom-wide fungal phylogenetic tree for 82 sequenced genomes using an alignment-free composition vector (CV) method [13–16]. The method has previously been successfully applied to prokaryotic and viral phylogenies [16, 17]. It uses whole-genome data of organisms and excludes artificial selection of genes and sites. In this report we will compare in detail our phylogenetic inferences with those inferred from standard methods. We will show the striking consistency between them and discuss the relationships among controversial lineages. Since our method reconstructs the fungal phylogeny with independent input data and methodology, the CVTree is a strong independent verification and complement to, but not a substitution for the traditional alignment-based analysis.
Results and discussions
Basal splits and the Dikarya
Figure 1 represents the CVTree of the 82 sequenced fungi. The organisms are grouped into 5 phyla or subphyla at the highest level, with the Ascomycota and Basidiomycota forming a monophyletic group Dikarya. Because currently available genomes in Chytridiomycota (2 strains of the same species, Batrachochytrium dendrobatidis), Microsporidia (Encephalitozoon cuniculi only) and Mucoromycotina (2 genera of the same family, Rhizopus oryzae and Phycomyces blakesleeanus) lack diversity, it is inappropriate to fully discuss the relationships among these clades until more organisms are sequenced. The following discussion will focus on the Basidiomycota and Ascomycota.
The phylum Basidiomycota consists of 3 subphyla: Agaricomycotina, Pucciniomycotina and Ustilaginomycotina. Except for Malassezia globosa, the 11 Basidiomycetes are classified into 5 classes, 7 orders, 7 families, 8 genera and 9 species in the scheme of the NCBI taxonomy. M. globosa is marked as Ustilaginomycotina incertae sedis in the NCBI taxonomy browser . The CVTree places it as sister taxon to Ustilago maydis (Figure 1, block H). This topology is supported by recent analyses of rDNA data and concatenated single-copy orthologous proteins [19, 20].
Although each of the three subphyla is widely accepted as monophyletic group, their relationships are not well-resolved . Previous cytological, biochemical and molecular analyses [2, 21] have suggested a topology like (Pucciniomycotina, (Agaricomycotina, Ustilaginomycotina)). With highly restricted taxon sampling in the Pucciniomycotina and Ustilaginomycotina, the CVTree recovers the same topology, but the bootstrap value is rather low (Figure 1, blocks in blue). Broader taxon sampling in each subphyla and further investigations are necessary to address this difficult question.
The 65 Ascomycetes come from three subphyla: the Pezizomycotina, Saccharomycotina and Taphrinomycotina. Although the monophyly of the Taphrinomycotina has not been fully agreed [22–25], the fission yeasts Schizosaccharomycetes (Taphrinomycotina) have been widely taken as a basal lineage of the Ascomycota [2, 3, 5, 25]. Our results support the early divergence of Schizosaccharomyces and a close relationship between the Pezizomycotina and Saccharomycotina (Figure 1). In the current dataset, organisms from the Taphrinomycotina and Saccharomycotina come from only one order in either subphylum, so it is proper to focus our higher-than-order level discussion on the relationships within the subphylum Pezizomycotina, where we have 39 organisms distributed in 4 classes.
The CVTree recognizes 4 class-level clades within the Pezizomycotina: the Dothideomycetes, Eurotiomycetes, Leotiomycetes and Sordariomycetes. They all come from the well-supported "Leotiomyceta" clade but their relationships have not been well resolved . Lutzoni et al  placed the Sordariomycetes and Dothideomycetes as sister clades in their four-gene analysis. However, subsequent works based on single- and multi-gene(s) found that it was the Leotiomycetes rather than Dothideomycetes that should be sister to the Sordariomycetes and they formed a clade [4, 10, 25, 26]. Our CVTree confirms the latter topology (Figure 1).
Although many accept the close relationship between the Sordariomycetes and Leotiomycetes, the relationships among the Dothideomycetes, Eurotiomycetes and Sordariomycetes-Leotiomycetes clade were less clear. At least three hypotheses have been proposed: (1) the Dothideomycetes and Sordariomycetes-Leotiomycetes group together to the exclusion of Eurotiomycetes . (2) The Eurotiomycetes and Sordariomycetes-Leotiomycetes group together to the exclusion of Dothideomycetes [4, 10, 25]. (3) The Dothideomycetes and the Eurotiomycetes form a clade that is sister group to the Sordariomycetes-Leotiomycetes clade [4, 10]. By including in their analyses only one species (Stagonospora nodorum) of Dothideomycetes, both Fitzpatrick et al  and Robbertse et al  reported conflicting results: depending on the tree construction model used, their results supported either hypothesis (2) or (3). By contrast, the current work adds in four more species (i.e., Cochliobolus heterostrophus, Pyrenophora tritici-repentis, Mycosphaerella fijiensis and Mycosphaerella graminicola) and our result unambiguously places the Dothideomycetes sister to Eurotiomycetes, supporting hypothesis (3) (Figure 1, block A and B).
Within the Sordariomycetes, Hypocreales and Sordariales are two well-supported order-level clades. However, it is uncertain in which order M. grisea (family Magnaporthaceae) should be placed (e.g. both Index Fungorum  and NCBI taxonomy browser  categorize it as Sordariomycetes incertae sedis). The CVTree suggests that it is closely related to the Sordariales (Figure 1, block D). This placement concurs with some recent phylogenomic analyses [3, 4, 10, 28]. Similarly, the order-level classification of two Verticillium species (family Plectosphaerellaceae) is not fully resolved either. Our tree places the Plectosphaerellaceae sister to the Hypocreales (Figure 1, block D), and this relationship is supported by the 4-gene phylogeny of Zhang et al  as well as Index Fungorum. Index Fungorum categorizes the family into the subclass Hypocreomycetidae, which includes in the Hypocreales. Regarding the other three classes, none of them have more than two ordinal members, so the order-level relationships within them could only be trivial.
Comparing to higher-level relationships, there have been more disagreements regarding the classification of taxa lower than order. This was partially caused by the difficulty to recognize various sexual states of one and the same species. Even the International Code for Botanic Nomenclature (Article 59)  permits to give anamorph a separate name from the corresponding teleomorph. However, molecular characters are capable to reveal more definite relationships.
In our dataset there are 7 orders (i.e., the Schizosaccharomycetales from the Taphrinomycotina; the Eurotiales, Hypocreales, Onygenales, Pleosporales and Sordariales from the Pezizomycotina; and the Saccharomycetales as the unique order in the Saccharomycotina) in which the number of sequenced organisms is greater than two. Within the Pleosporales, the relationships among the 3 organisms are less controversial: P. tritici-repentis and C. heterostrophus are from the family Pleosporaceae while S. nodorum is from the Phaeosphaeriaceae (Figure 1, block B). In what follows we discuss the phylogeny within the other 6 orders one by one.
All the three organisms in this order belong to the genus Schizosaccharomyces. Before whole-genomes were available for members of the Schizosaccharomyces, their phylogenetic relationships were inferred from mitochondrial genomes . Mitochondrial phylogeny shows that Schizosaccharomyces pombe and Schizosaccharomyces octosporus are more closely related to the exclusion of Schizosaccharomyces japonicus. The CVTree recovers the identical topology from whole-genome data (Figure 1, block F).
In our analysis the order Sordariales is represented by 3 species: Chaetomium globosum, Neurospora crassa and Podospora anserina. According to the NCBI taxonomy they belong to the Chaetomiaceae, Sordariaceae and Lasiosphaeriaceae families, respectively. Previous analysis based on LSU rDNA (and other genes) have shown that many traditional families in this order do not form clades, e.g., the Chaetomiaceae and Lasiosphaeriaceae are paraphyletic groups [28, 31, 32]. Despite such inconsistency, recent 18S rDNA and phylogenomic analyses agreed that the relationships among the three species are ((C. globosum, P. anserina), N. crassa) [3, 4, 25, 28, 33]. Our CVTree phylogeny supports this topology as well (Figure 1, block D).
We mention in passing that so far the rice blast fungus M. grisea has been denoted as Sordariomycetes incertae sedis, i. e., not being designated to an existing order. From the CVTree and papers just cited above it comes out as ((Hypocreales, Plectosphaerellaceae), (M. grisea, Sordariales)), hinting on the feasibility of putting this species in a new order.
In the order Eurotiales all sequenced organisms are Aspergillus species, including Neosartorya fischeri, the teleomorph of Aspergillus fischerianus. The CVTree phylogeny (Figure 1, block A) of the 8 species is identical to the recent result of 30-gene phylogenomic analysis by Rokas et al (2007)  and to that shown in the BROAD-FIG Aspergillus Comparative Database . In all of these trees A. nidulans is the basal lineage while in the previously widely accepted LSU rDNA phylogeny by Peterson (2000)  it did not ([see Additional file 1], Figure S1(a)). Another difference between Peterson's result and ours is the placement of Aspergillus niger and Aspergillus terreus: the former gave that A. niger was sister taxon to the Aspergillus oryzae-Aspergillus flavus clade and A. terreus diverged early. In contrast, our result supports the sister relationship of A. terreus and the A. oryzae-A. flavus clade and the early divergence of A. niger.
In our dataset there are 4 species with 11 strains belonging to the Onygenales. The two Coccidioides species as well as Uncinocarpus reesii form one clade and the three Paracoccidioides as well as Histoplasma capsulatum form the other. Deeper in the tree, these two clades are sister taxa (Figure 1, block A). For a long time, Paracoccidioides brasiliensis was considered an imperfect fungus. In recent years, it has been considered a member of the family Onygenaceae and placed in a common group with C. immitis, H. capsulatum and U. reesii . More recently, a clade distinct from Onygenaceae has been proposed as a new family Ajellomycetaceae to encompass Histoplasma and Paracoccidioides but not C. immitis and U. reesii [38, 39]. The current work supports the suggestion of the Ajellomycetaceae by placing the Onygenaceae as its sister group.
In the Hypocreales and Saccharomycetales, taxonomy shows inadequate resolution and conflicts with current phylogeny derived from standard methods, i.e., some traditional families and genera turn out to be non-monophyletic. We discuss phylogeny of these two branches in the following two sections.
Although the molecular phylogenetic studies have helped in solving many problems that morphology could not, the classification of many members of Hypocreales, especially Fusarium spp. and Trichoderma spp. are far from being settled [40, 41]. Regarding the 7 organisms in our dataset in the order, the key difference between the CVTree (Figure 1, block D) and NCBI taxonomy ([see Additional file 1], Figure S1(b)) is the position of Fusarium oxysporum. In the NCBI taxonomy, it is placed in the group mitosporic Hypocreales, while Fusarium verticillioides and Fusarium graminearum belong to the family Nectriaceae. In the CVTree, however, the three species form the Nectriaceae clade: F. oxysporum is grouped with F. verticillioides and F. graminearum is sister taxon to them. The monophyly of the clade is supported by Index Fungorum , which classifies F. oxysporum into the Nectriaceae. Moreover, the same topology can be found at BROAD-FGI Fusarium Comparative Database .
The Saccharomycetales is a unique order in the Saccharomycetes which in turn is a unique class in the subphylum Saccharomycotina according to Hibbett et al (2007) . These species have been studied extensively and some members are model organisms. Two distinct events in their evolutionary history have been well-documented: (1) some species have undergone whole-genome duplication more than 100 million years ago. They form the so-called WGD clade; (2) some species translate CUG codon into serine instead of leucine. They form another branch called the CTG clade. Any reasonable phylogeny should clearly resolve the two clades among yeasts. The CVTree does.
Our current dataset includes 23 organisms from Saccharomycetales, as compared to 19 in Fitzpatrick et al (2006)  and 12 in James et al (2006) . Therefore, we are in a position to perform a more detailed comparison of the CVTree with other phylogenies.
Within the Saccharomycetales, Yarrowia lipolytica is the early-diverging lineage and the other organisms consists of two groups covering the WGD and CTG clade, respectively (Figure 1, block E). This is a common feature of the 12-, 19 and 32-organism trees. Within the CTG clade, the CVTree gives a structure identical with that of Fitzpatrick et al  if Pichia stipitis is not included in. Our tree places P. stipitis and D. hansenii as sister taxa. This placement is consistent with the result of the 94 single-copy genes analysis , which suggests a close relationship between D. hansenii and P. stipitis to the exclusion of C. lusitaniae (Figure 1, block E). Our results further confirms two features in the CTG clade : (1) Candida guilliermondii is closely related to Debaryomyces hansenii to the exclusion of Candida lusitaniae and (2) Lodderomyces elongisporus is closely related to Candida parapsilosis and is likely to be its sexual form.
The other group further splits into two clades by an ancient whole-genome duplication (WGD) event. The WGD clade includes six Saccharomyces sensu stricto species and two Saccharomyces sensu lato species (Saccharomyces castellii and C. glabrata). The ladderized topology within the Saccharomyces sensu stricto organisms is consistent with previous phylogenomic results [44–46]. So far the base of the WGD clade has not been confidently resolved. Some proposed that S. castellii diverged from the Saccharomyces sensu stricto species earlier than C. glabrata by comparing synteny among species and multi-gene analysis [4, 47]. However, other phylogenomic analyses argued that C. glabrata diverged earlier [4, 44, 45]. According to the CVTree, C. glabrata is the likely basal lineage of the WGD clade (Figure 1, block E).
The monophyly of Kluyveromyces waltii, Saccharomyces kluyveri, Eremothecium gossypii and Kluyveromyces lactis is unsettled either. Some authors suggested that the four species are paraphyletic but they together with the WGD species constitute a clade [46, 48]. For example, Kurtzman (2003)  proposed a topology like (((WGD, (K. waltii, S. kluyveri)), K. lactis), E. gossypii). In contrast, others proposed that the 4 species themselves form a clade [4, 45]. The CVTree supports the monophyly of the four yeasts and the sister-relationship between this clade and the WGD group (Figure 1, block E).
The relationships among these 4 organisms are again controversial. Although many works agreed that (K. waltii, S. kluyveri) should be closer to each other, the grouping of E. gossypii and K. lactis are not widely accepted. Kurtzman  and Suh et al  proposed that they are paraphyletic, while Jeffroy et al  and Fitzpatrick et al  suggested that they form a clade and placed this group as sister branch to the (K. waltii, S. kluyveri) clade. The CVTree, unlike studies mentioned above, places K. lactis and (K. waltii, S. kluyveri) as sister group to the exclusion of E. gossypii (Figure 1, block E). As different materials and methods give controversial results, more genomes and analyses are required to confidently resolve this incongruence.
To the best of sequenced fungi available, we have inferred their phylogeny using the alignment-free composition vector method and discussed their relationships. The above detailed comparison has shown remarkable consistency between the CVTree and the recent results of standard methods, the consistency actually holds at all levels. We can now give an overall picture of the CVTree phylogeny: the Microsporidia is placed in fungi and the Ascomycota and Basidiomycota are resolved as sister taxa that together constitute a clade named Dikarya. Moreover, we also investigated the position of the kingdom by adding gene repertoires of animals, plants and Protozoans. Our results suggested the sister-relationship between fungi and the animal-choanoflagellate clade (data not shown). In the Ascomycota, 3 subphyla are recognized: the Taphrinomycotina is the early-diverging group; the Pezizomycotina includes 4 clades from the "Leotiomyceta" and the Saccharomycotina encompasses the WGD and CTG clades. In the Basidiomycota, monophyly of three subphyla are supported as well.
The novelty of this work can be viewed from the following four aspects: First, the CVTree uses different data from standard methods. Standard methods construct trees using subsets of proteomes or rRNAs as input data. The number is from one to a few hundred. By contrast, the CV method uses all the information of nuclear protein-coding genes. In addition, as we have explained in Introduction, most of the standard multi-gene analyses manually select genes and sites from input data. In contrast, our approach does not need such adjustment, thus circumvent the ambiguity of choosing genes and sites.
Second, the CVTree uses an independent methodology to automatically construct phylogenetic tree. In standard methods, the alignment algorithm, evolutionary model and numerous parameters are needed to be selected and set case-by-case according to the heterogeneity of input data. So far there is no general rule to guide these selections and settings. By contrast, the string-counting strategy of CVTree minimizes arbitrary factors in tree construction. It has only one parameter K ([see Additional file 2] and Methods) and the construction process is automatic. Furthermore, the algorithm is rapid comparing to many standard methods. For instance, it takes only about 1 hour to construct one 82-organism tree on a computer of 2.3 GHz CPU and 4 G RAM for various K values. ([see Additional file 2]).
Third, the CVTree gives novel and stronger supports to the results of standard methods through remarkable consistency between them. We note that such consistency not only confirms the validity of our methods but also strongly support the fungi phylogeny based on the standard methods because such supports come from an independent input dataset and methodology. The supports from CVTree put the current understanding of the fungi phylogeny on a more secure footing. Current kingdom-wide fungi phylogeny has been established on numerous works, each of which investigated a major or minor group and contributed a piece (i.e., local phylogeny) to the whole picture. These works constructed local phylogenies using different input data, site selection strategies, alignment algorithms, evolutionary models and parameters. In other words, the kingdom tree was assembled from local phylogenies that complied with different criteria. However, the feasibility of assembling these pieces together is not self-evident and needs validations. Our CVTree provides such a proper support because all of the relationships therein are inferred under the same criteria. Whenever a group is supported it is reinforced by a "global" picture.
Fourth, novel phylogenetic findings of this work. To the best of our knowledge, the CVTree is so far the only successful kingdom-wide fungi phylogeny constructed by a strategy other than the standard methods. Besides consistency, the current study has shed light on many controversial cases. For example, the CVTree suggests that Dothideomycetes and Eurotiomycetes are sister clades using broader species sampling than previous works [4, 10]; that M. grisea and the Plectosphaerellaceae group with the Sordariales and Hypocreales, respectively [3, 4, 10, 28]; that the Ajellomycetaceae is a distinct clade from the Onygenaceae [38, 39]; and that A. nidulans is the earliest diverged among the 8 aspergilli [34, 35]. In all of the above examples, contradictions are found between researches done in different periods and it is interesting to note that the CVTree tends to support more recent results. CVTree's support to a certain relationship actually adds weight to it.
The CVTree is robust, to a certain degree, to variations of gene models used in genome annotation: (1) Our experiments showed that randomly adding or dropping ≤ 30% proteins from the whole gene products rarely changed major relationships in the tree ([see Additional file 3]). (2) The gene models of some fungal genomes have been changed rapidly because of lacking of evidences of transcripts and these changes have been reflected in different versions of genome annotation. This enable us to test the topological stability of the CVTree by using different versions of genome annotation to construct trees. We downloaded multiple annotation versions of some organisms, i.e., two versions of M. grisea and two of F. graminearum and used their combinations to construct trees. Our results showed that all possible combinations generated identical topology ([see Additional file 3] for details). However, the stability was not absolute and differences in gene annotations might alter the tree topology. We found an example (A. niger) that different gene annotations of the same species led to two slightly different positions, but it did not affect other species in the tree ([see Additional file 3]).
As a method in development, the current CVTree has some restrictions: (1) The relationship between the branch length of CVTree and that of alignment-based tree is currently unclear. Our simulation experiments have revealed that distance between two CVs is proportional to traditional evolutionary distance when substitutions between two sequences are rare (data not shown). However, because the distance between two CVs is not estimated from site substitutions, the branch length of the CVTree in general does not have simple relationship with the traditional evolutionary distance. As a result we currently can not compare the CVTree with alignment-based tree at the level of scaled branch length. Therefore the current work constructs unscaled tree and only discuss topological relationships. (2) The CVTree may suffer from long-branch attraction and amino acids composition bias. The core of the CV method lies in that it provides an alternative way to construct the distance-matrix. Once the matrix is established, the following tree construction process is performed by NJ method. Many have reported that long-branch attraction and amino acids composition bias may affect the NJ tree topology [49, 50]. However, it is not clear how and to what extent the matrix construction process is affected by these errors. (3) Our method is based on whole-genome data, so its sampling scope is restricted by the number of sequenced genomes.
Having been successfully applied to the prokaryotic branches of the TOL, the present research extends the CVTree method to the Eukaryotic kingdom of fungi and provides independent verification of traditional phylogenetic approaches. Further study to cover the whole TOL including all Eukaryotic branches is underway.
The CVTree method
The composition vector method used in this study has been described in previous publications [13–15] so we only give a brief account here. An organism is represented first by a raw composition vector whose components correspond to the number of various overlapping K-peptides (for a fixed K) in the collection of all protein products encoded in the genome. These 20Kcomponents are put in lexicographic order. In order to highlight the shaping role of selective evolution, the components of a CV are then modified by subtracting a statistical background reflecting the viewpoint of the neutral theory of evolution that mutations happen randomly at molecular level. Our substraction procedure is based on a (K - 2)-th order Markov prediction and therefore the minimum K starts from 3. The dissimilarity of two species is measured by a correlation distance derived from the corresponding modified CVs. Finally, from the distance matrix thus obtained, a neighbor-joining tree is produced by the PHYLIP package . In the CVTree construction the fixed peptide length K controls the resolution of the method. The best choice of K depends on the length of input genomes. Our experiments revealed that K = 6 and 5 are the best for prokaryotes and viruses, respectively [13, 17]. For fungi, we constructed trees of K = 3 to 10 and evaluated their robustness by 100 bootstrap replicates ([see Additional file 2] and Figure 1). Since the K = 7 topology shows better bootstrap values than others, our discussions in this study are mainly based on it.
Fungal and outgroup genomes
The collection of protein products from 82 fungal genomes was used in this study (Table 1). We relied on the genome annotations provided by the corresponding sequencing project with A. niger being the only exception. This species was annotated in house ([see Additional file 3] and Table 1) by BGF [52, 53], an ab initio gene prediction tool developed in our laboratory. The last column "Source" in Table 1 indicates the origin of the data: BROAD Institute Fungal Genome Initiative (BROAD-FGI) , Department of Energy Joint Genome Institute (JGI) , National Center for Biotechnology Information (NCBI) ftp-site , Resources for Fungal Comparative Genomics (RFCG)  and Fungal Genome Research website (FGR) . A protist choanoflagellate (Monosiga brevicollis) and two metazoans (Caenorhabditis elegans and Drosophila melanogaster) genomes were included as outgroup. The RFCG data used in this study covers all the fungal species which were used in Fitzpatrick et al (2006) analysis  except for Candida dubliniensis because its sequences are not available at RFCG.
Fungal phylogeny references
In order to integrate the multi-laboratory efforts in fungal molecular phylogeny and taxonomy, an Assembling the Fungal Tree of Life (AFTOL) project was launched in 2002 . As a preliminary but successful outcome of AFTOL, Hibbett et al. have published a higher-level fungal phylogenetic classification . It is a classification as more than two lower taxa belonging to a higher taxon may be juxtaposed in this scheme and seven Linnaeus ranks (kingdom, subkingdom, phylum, subphylum, class, subclass, and order) are used. It is phylogenetic as well because emphasis has been put on each included taxon being a monophyletic group.
This classification allows us to use terms of taxonomy when analyzing phylogenetic relationships. However, it is worth noting that our work concerns phylogeny, but not taxonomy. Accordingly, taxonomic terms such as "higher-level" or "class-level" clades, "three subphyla" and "the order Sordariales" should be understood phylogenetically. In other words, taxa in this study are not used as labels of hierarchical taxonomic ranks but only as names of clades. Higher or lower levels actually mean major or minor clades. A clade being called a class or an order merely means that it forms a monophyletic branch and has been named in that way in some previous publications.
The present study infers phylogenetic relationships from whole-genome data of fungi without using sequence alignment. We use the Hibbett et al paper  as a major reference to phylogenetic classification at higher-than-order levels. When it comes to compare the branching pattern of originally juxtaposed taxa we refer to recent phylogenetic studies that utilize the standard methods.
Regarding phylogeny at lower-than-order levels, we used the following three publicly accessible databases as references. Index Fungorum  provides an up-to-date phylogenetic classification of fungi. Focusing on the Ascomycota, Myconet  regularly updates its classification. Another popular taxonomy database is NCBI taxonomy browser , which integrates classifications from various sources and gives useful information as well. If at a certain node the resolution is not high enough or when discrepancy between the CVTree phylogeny and current classification was found, we refer to recent publications for more information.
the composition vector
the phylogenetic tree obtained by using CV method
Tree of Life
the Assembling the Fungal Tree of Life Project
ribosomal RNA-encoding DNA.
Hawksworth DL: The magnitude of fungal diversity: the 1.5 million species estimate revisited. Mycol Res. 2002, 105 (12): 1422-1432. 10.1017/S0953756201004725.
Lutzoni F, Kauff F, Cox C, McLaughlin D, Celio G, Dentinger B, Padamsee M, Hibbett D, James T, Baloch E, et al: Assembling the fungal tree of life: progress, classification, and evolution of subcellular traits 1. Am J Bot. 2004, 91 (10): 1446-1480. 10.3732/ajb.91.10.1446.
James TY, Kauff F, Schoch CL, Matheny PB, Hofstetter V, Cox CJ, Celio G, Gueidan C, Fraker E, Miadlikowska J, Lumbsch HT, Rauhut A, Reeb V, Arnold AE, Amtoft A, Stajich JE, Hosaka K, Sung GH, Johnson D, O'Rourke B, Crockett M, Binder M, Curtis JM, Slot JC, Wang Z, Wilson AW, Schüssler A, Longcore JE, O'Donnell K, Mozley-Standridge S, Porter D, Letcher PM, Powell MJ, Taylor JW, White MM, Griffith GW, Davies DR, Humber RA, Morton JB, Sugiyama J, Rossman AY, Rogers JD, Pfister DH, Hewitt D, Hansen K, Hambleton S, Shoemaker RA, Kohlmeyer J, Volkmann-Kohlmeyer B, Spotts RA, Serdani M, Crous PW, Hughes KW, Matsuura K, Langer E, Langer G, Untereiner WA, Lücking R, Büdel B, Geiser DM, Aptroot A, Diederich P, Schmitt I, Schultz M, Yahr R, Hibbett DS, Lutzoni F, McLaughlin DJ, Spatafora JW, Vilgalys R: Reconstructing the early evolution of Fungi using a six-gene phylogeny. Nature. 2006, 443 (7113): 818-822. 10.1038/nature05110.
Fitzpatrick DA, Logue ME, Stajich JE, Butler G: A fungal phylogeny based on 42 complete genomes derived from supertree and combined gene analysis. BMC Evol Biol. 2006, 6: 99-10.1186/1471-2148-6-99.
Hibbett DS, Binder M, Bischoff JF, Blackwell M, Cannon PF, Eriksson OE, Huhndorf S, James T, Kirk PM, Lücking R, Lumbsch HT, Lutzoni F, Matheny PB, McLaughlin DJ, Powell MJ, Redhead S, Schoch CL, Spatafora JW, Stalpers JA, Vilgalys R, Aime MC, Aptroot A, Bauer R, Begerow D, Benny GL, Castlebury LA, Crous PW, Dai YC, Gams W, Geiser DM, Griffith GW, Gueidan C, Hawksworth DL, Hestmark G, Hosaka K, Humber RA, Hyde KD, Ironside JE, Köljalg U, Kurtzman CP, Larsson KH, Lichtwardt R, Longcore J, Miadlikowska J, Miller A, Moncalvo JM, Mozley-Standridge S, Oberwinkler F, Parmasto E, Reeb V, Rogers JD, Roux C, Ryvarden L, Sampaio JP, Schüssler A, Sugiyama J, Thorn RG, Tibell L, Untereiner WA, Walker C, Wang Z, Weir A, Weiss M, White MM, Winka K, Yao YJ, Zhang N: A higher-level phylogenetic classification of the Fungi. Mycol Res. 2007, 111 (Pt 5): 509-547. 10.1016/j.mycres.2007.03.004.
Bruns TD, Vilgalys R, Barns SM, Gonzalez D, Hibbett DS, Lane DJ, Simon L, Stickel S, Szaro TM, Weisburg WG: Evolutionary relationships within the fungi: analyses of nuclear small subunit rRNA sequences. Mol Phylogenet Evol. 1992, 1 (3): 231-41. 10.1016/1055-7903(92)90020-H.
Dunn CW, Hejnol A, Matus DQ, Pang K, Browne WE, Smith SA, Seaver E, Rouse GW, Obst M, Edgecombe GD, Sørensen MV, Haddock SHD, Schmidt-Rhaesa A, Okusu A, Kristensen RM, Wheeler WC, Martindale MQ, Giribet G: Broad phylogenomic sampling improves resolution of the animal tree of life. Nature. 2008, 452 (7188): 745-749. 10.1038/nature06614.
Savolainen V, Chase MW: A decade of progress in plant molecular phylogenetics. Trends Genet. 2003, 19 (12): 717-724. 10.1016/j.tig.2003.10.003.
Bourlat SJ, Juliusdottir T, Lowe CJ, Freeman R, Aronowicz J, Kirschner M, Lander ES, Thorndyke M, Nakano H, Kohn AB, Heyland A, Moroz LL, Copley RR, Telford MJ: Deuterostome phylogeny reveals monophyletic chordates and the new phylum Xenoturbellida. Nature. 2006, 444 (7115): 85-88. 10.1038/nature05241.
Robbertse B, Reeves JB, Schoch CL, Spatafora JW: A phylogenomic analysis of the Ascomycota. Fungal Genet Biol. 2006, 43 (10): 715-725. 10.1016/j.fgb.2006.05.001.
Philippe H, Telford MJ: Large-scale sequencing and the new animal phylogeny. Trends Ecol Evol. 2006, 21 (11): 614-620. 10.1016/j.tree.2006.08.004.
Delsuc F, Brinkmann H, Philippe H: Phylogenomics and the reconstruction of the tree of life. Nat Rev Genet. 2005, 6 (5): 361-375. 10.1038/nrg1603.
Qi J, Wang B, Hao B: Whole Proteome Prokaryote Phylogeny Without Sequence Alignment: A K-String Composition Approach. J Mol Evol. 2004, 58: 1-11. 10.1007/s00239-003-2493-7.
Qi J, Luo H, Hao B: CVTree: a phylogenetic tree reconstruction tool based on whole gemomes. Nucleic Acids Res. 2004, W45-W47. 10.1093/nar/gkh362. 32 Webserver
Hao B, Qi J: Prokaryote phylogeny without sequence alignment: from avoidance signature to composition distance. J Bioinform Comput Biol. 2004, 2: 1-19. 10.1142/S0219720004000442.
Gao L, Qi J, Sun J, Hao B: Prokaryote phylogeny meets taxonomy: an exhaustive comparison of composition vector trees with systematic bacteriology. Sci China C Life Sci. 2007, 50 (5): 587-599. 10.1007/s11427-007-0084-3.
Gao L, Qi J: Whole genome molecular phylogeny of large dsDNA viruses using composition vector method. BMC Evol Biol. 2007, 7: 41-10.1186/1471-2148-7-41.
National Center for Biotechnology Information. [ftp://ftp.ncbi.nih.gov/genomes/]
Xu J, Saunders CW, Hu P, Grant RA, Boekhout T, Kuramae EE, Kronstad JW, Deangelis YM, Reeder NL, Johnstone KR, Leland M, Fieno AM, Begley WM, Sun Y, Lacey MP, Chaudhary T, Keough T, Chu L, Sears R, Yuan B, Dawson TL: Dandruff-associated Malassezia genomes reveal convergent and divergent virulence traits shared with plant and human fungal pathogens. Proc Natl Acad Sci USA. 2007, 104 (47): 18730-18735. 10.1073/pnas.0706756104.
Begerow D, Stoll M, Bauer R: A phylogenetic hypothesis of Ustilaginomycotina based on multiple gene analyses and morphological data. Mycologia. 2006, 98 (6): 906-916. 10.3852/mycologia.98.6.906.
McLaughlin DJ, Frieders EM, Lü H: A microscopist's view of heterobasidiomycete phylogeny. Stud Mycol. 1995, 38: 91-109.
Nishida H, Sugiyama J: Archiascomycetes: detection of a major new lineage within the Ascomycota. Mycoscience. 1994, 35 (4): 361-366. 10.1007/BF02268506.
Lumbsch HT, Huhndorf SM: Outline of Ascomycota – 2007. Myconet. 2007, 13: 1-58.
Liu YJ, Hodson MC, Hall BD: Loss of the flagellum happened only once in the fungal lineage: phylogenetic structure of kingdom Fungi inferred from RNA polymerase II subunit genes. BMC Evol Biol. 2006, 6: 74-10.1186/1471-2148-6-74.
Spatafora JW, Sung GH, Johnson D, Hesse C, O'Rourke B, Serdani M, Spotts R, Lutzoni F, Hofstetter V, Miadlikowska J, Reeb V, Gueidan C, Fraker E, Lumbsch T, Lücking R, Schmitt I, Hosaka K, Aptroot A, Roux C, Miller AN, Geiser DM, Hafellner J, Hestmark G, Arnold AE, Büdel B, Rauhut A, Hewitt D, Untereiner WA, Cole MS, Scheidegger C, Schultz M, Sipman H, Schoch CL: A five-gene phylogeny of Pezizomycotina. Mycologia. 2006, 98 (6): 1018-1028. 10.3852/mycologia.98.6.1018.
Lumbsch HT, Schmitt I, Lindemuth R, Miller A, Mangold A, Fernandez F, Huhndorf S: Performance of four ribosomal DNA regions to infer higher-level phylogenetic relationships of inoperculate euascomycetes (Leotiomyceta). Mol Phylogenet Evol. 2005, 34 (3): 512-524. 10.1016/j.ympev.2004.11.007.
Index Fungorum. [http://www.indexfungorum.org/names/Names.asp]
Zhang N, Castlebury LA, Miller AN, Huhndorf SM, Schoch CL, Seifert KA, Rossman AY, Rogers JD, Kohlmeyer J, Volkmann-Kohlmeyer B, Sung GH: An overview of the systematics of the Sordariomycetes based on a four-gene phylogeny. Mycologia. 2006, 98 (6): 1076-1087. 10.3852/mycologia.98.6.1076.
International Code for Botanic Nomenclature. [http://ibot.sav.sk/icbn/main.htm]
Bullerwell CE, Leigh J, Forget L, Lang BF: A comparison of three fission yeast mitochondrial genomes. Nucleic Acids Res. 2003, 31 (2): 759-768. 10.1093/nar/gkg134.
Huhndorf SM, Miller AN, Fernández FA: Molecular systematics of the Coronophorales and new species of Bertia, Lasiobertia and Nitschkia. Mycol Res. 2004, 108 (Pt 12): 1384-1398. 10.1017/S0953756204001273.
Miller AN, Huhndorf SM: Multi-gene phylogenies indicate ascomal wall morphology is a better predictor of phylogenetic relationships than ascospore morphology in the Sordariales (Ascomycota, Fungi). Mol Phylogenet Evol. 2005, 35: 60-75. 10.1016/j.ympev.2005.01.007.
Berbee M: The phylogeny of plant and animal pathogens in the Ascomycota. Physiol Mol Plant Pathol. 2001, 59 (4): 165-187. 10.1006/pmpp.2001.0355.
Rokas A, Galagan JE: Aspergillus nidulans Genome and a Comparative Analysis of Genome Evolution in Aspergillus. The Aspergilli: Genomics, Medical Aspects, Biotechnology, and Research. Edited by: Osmani SA, Goldman GH. 2007, New York: CRC Press, 43-54.
BROAD-FGI Aspergillus Comparative Database. [http://www.broad.mit.edu/annotation/genome/aspergillus_group/MultiHome.html]
Peterson SW: Phylogenetic relationships in Aspergillus based on rDNA sequence analysis. Integration of Modern Taxonomic Methods for Penicillium and Aspergillus Classification. Edited by: Samson RA, Pitt JI. 2000, Amsterdam: Harwood Academic Publishers, 323-355.
San-Blas G, Niño-Vega G, Iturriaga T: Paracoccidioides brasiliensis and paracoccidioidomycosis: molecular approaches to morphogenesis, diagnosis, epidemiology, taxonomy and genetics. Med Mycol. 2002, 40 (3): 225-242. 10.1080/714031110.
Bagagli E, Theodoro RC, Bosco SMG, McEwen JG: Paracoccidioides brasiliensis: phylogenetic and ecological aspects. Mycopathologia. 2008, 165 (4–5): 197-207. 10.1007/s11046-007-9050-7.
Untereiner W, Scott J, Naveau F, Sigler L, Bachewich J, Angus A: The Ajellomycetaceae, a new family of vertebrate-associated Onygenales. Mycologia. 2004, 96 (4): 812-821. 10.2307/3762114.
Druzhinina I, Kubicek CP: Species concepts and biodiversity in Trichoderma and Hypocrea: from aggregate species to species clusters?. J Zhejiang Univ Sci B. 2005, 6 (2): 100-112. 10.1631/jzus.2005.B0100.
Zhang N, O'Donnell K, Sutton DA, Nalim FA, Summerbell RC, Padhye AA, Geiser DM: Members of the Fusarium solani species complex that cause infections in both humans and plants are common in the environment. J Clin Microbiol. 2006, 44 (6): 2186-2190. 10.1128/JCM.00120-06.
BROAD-FGI Fusarium Comparative Database. [http://www.broad.mit.edu/annotation/genome/fusarium_group/MultiHome.html]
Jeffries TW, Grigoriev IV, Grimwood J, Laplaza JM, Aerts A, Salamov A, Schmutz J, Lindquist E, Dehal P, Shapiro H, Jin YS, Passoth V, Richardson PM: Genome sequence of the lignocellulose-bioconverting and xylose-fermenting yeast Pichia stipitis. Nat Biotechnol. 2007, 25 (3): 319-326. 10.1038/nbt1290.
Rokas A, Williams BL, King N, Carroll SB: Genome-scale approaches to resolving incongruence in molecular phylogenies. Nature. 2003, 425 (6960): 798-804. 10.1038/nature02053.
Jeffroy O, Brinkmann H, Delsuc F, Philippe H: Phylogenomics: the beginning of incongruence?. Trends Genet. 2006, 22 (4): 225-231. 10.1016/j.tig.2006.02.003.
Kurtzman CP: Phylogenetic circumscription of Saccharomyces, Kluyveromyces and other members of the Saccharomycetaceae, and the proposal of the new genera Lachancea, Nakaseomyces, Naumovia, Vanderwaltozyma and Zygotorulaspora. FEMS Yeast Res. 2003, 4 (3): 233-245. 10.1016/S1567-1356(03)00175-2.
Scannell DR, Byrne KP, Gordon JL, Wong S, Wolfe KH: Multiple rounds of speciation associated with reciprocal gene loss in polyploid yeasts. Nature. 2006, 440 (7082): 341-345. 10.1038/nature04562.
Suh SO, Blackwell M, Kurtzman CP, Lachance MA: Phylogenetics of Saccharomycetales, the ascomycete yeasts. Mycologia. 2006, 98 (6): 1006-1017. 10.3852/mycologia.98.6.1006.
Foster PG, Hickey DA: Compositional bias may affect both DNA-based and protein-based phylogenetic reconstructions. J Mol Evol. 1999, 48 (3): 284-290. 10.1007/PL00006471.
Bruno WJ, Socci ND, Halpern AL: Weighted neighbor joining: a likelihood-based approach to distance-based phylogeny reconstruction. Mol Biol Evol. 2000, 17: 189-197.
Felsenstein J: PHYLIP-Phylogeny Inference Package (Version 3.2). Cladistics. 1989, 5: 164-166.
Li H, Gao L, Fang L, Liu T, Li HH, Li Y, Fang LJ, Xie HM, Zheng WM, Liu JS, Xu Z, Jin J, Li YD, Xing ZX, Gao SG, Hao BL: Test Data Sets and Evaluation of Gene Prediction Programs on the Rice Genome. J Comput Sci & Technol. 2005, 20 (4): 446-53. 10.1007/s11390-005-0446-x.
Beijing Gene Finder. [http://tlife.fudan.edu.cn/bgf/]
Board Institute Fungal Genome Initiative. [http://www.broad.mit.edu/annotation/fgi/]
Department of Energy Joint Genome Institute. [http://genome.jgi-psf.org/]
Resources for Fungal Comparative Genomics. [http://fungal.genome.duke.edu/]
Fungal Genome Research. [http://fungalgenomes.org/]
Assembling the Fungal Tree of Life Project. [http://aftol.org]
NCBI taxonomy browser. [http://www.ncbi.nlm.nih.gov/Taxonomy/]
The authors thank Dr. David M. Geiser and Dr. Ning Zhang for providing valuable suggestions on recent literature of Aspergilli and Hypocreales phylogeny, respectively. We thank all four anonymous referees whose comments have significantly improved the manuscript. This work was supported by the Basic Research Program of China (The 973 Program No. 2007CB814800) and Shanghai Leading Academic Discipline Project (Project Number: B111).
BH and HW designed the study, carried out the molecular phylogenetic analysis, and drafted the manuscript. HW and LG collected the genome dataset. ZX and LG developed the CVTree program. All authors read and approved the final manuscript.