Current phylogenetic studies of the kingdom Fungi define Ascomycota as a monophyletic group 1213. For an insight into the evolution of fungal SET-domain genes and their relationship with the genes of higher eukaryotes, we reconstructed phylogenetic trees using the highly conserved SET-domain region (~150 amino acids) 18. First, we performed a series of similarity searches including profile hidden Markov models against fourteen genomes: eleven fungal (Ascomycetes), two animal (one mammalian and one invertebrate), and one plant (Arabidopsis) genomes (see Additional file 1). One hundred and eighty two SET-domain sequences were identified (see Additional file 2). Phylogenetic analyses were performed to identify the SET-domain protein families and subfamilies. Maximum likelihood phylogeny reconstructed from the selected 113 representative sequences is shown in Fig. 1. The phylogenetic clustering based on SET-domain sequences reflected presence/absence of group-specific architectural motifs (see Additional file 3). We also note that the SET-domain based phylogeny does not seem to support consistently any particular evolutionary relationships among the three kingdoms: fungi, animals, and plants (see also Fig. 2).

The distribution of the SET-domain proteins across 14 genomes is summarized in Fig. 3. All 108 SET-domain sequences retrieved from the fungal genomes that belong to the group of the histone methyltransferases are included. SET-domain containing proteins from the RuBisCo, cytochrome C, and the recently discovered ribosomal protein lysine methyltransferases 19 are excluded. Among the SET-domain sequences of animal and plant origin, included are only those that are used as reference for the fungal proteins.

The sizes of the fourteen genomes and the total numbers of ORFs are not linearly correlated; the genomes of the filamentous fungi are approximately three times the size of the yeasts (except Y. lipolytica) while the overall number of identified ORFs is only 1.6-to-2 fold higher than in yeasts. On the other hand, the numbers of SET-domain containing genes in the Pezizomycotina species (13–23) are about 3-fold higher compared to those of the yeasts (5–6; Fig. 3). Non-linear increasement in the number of SET-domain genes is even more pronounced in the animal and plant genomes: about 70 predicted SET-domain genes in the human and mouse genomes, around 40 in D. melanogaster 20, and around 40 in A. thaliana 21. Increased numbers and diversification of encoded SET-domain proteins reflect, perhaps, increased diversity and complexity of multicellular programs.

Despite the smaller number of ORFs than most Saccharomycotina fungi 22, the single-living Taphrinomycotina (archiascomycete), S. pombe, carries three SET-domain genes (SET9, MYND-SET, and a Su(var)3-9 homologue). These SET-domain genes are present in the Pezizomycotina and in animals/plant, but remarkably absent from the other yeasts (except Y. lipolytica, which carries a SET9 gene) (Fig. 3). Presence of these genes supports a conclusion that S. pombe and Y. lipolytica share many common features with the filamentous relatives 2223. Lastly, S. pombe and Y. lipolytica carry one copy of a SET-domain gene of unclear origins. These S. pombe-specific and the Y. lipolytica-specific putative SET-domain proteins, unrelated to each other or to any of the 182 sequences analyzed, are annotated as 'unknown' (Fig. 3) and were excluded from further analyses.

Saccharomycotina genomes have been subjected to expansions (whole-genome duplications) and deletions (reductive evolution), which have shaped the genome sizes of extant yeasts 2324 varying between 9.2 Mb (A. gossypii) to 20 Mb (Y. lipolytica) and overall putative gene numbers between ~4,700 and ~7,000 (Fig. 3). However, the total number of the SET-domain genes in all tested Saccharomycotina is remarkably constant: five or six. Most yeast SET-domain genes are single-copy representatives of distinct subgroups. The SET4 gene, an apparent duplication of SET3 (see below for further discussion) accounts for the difference in the number of SET-domain genes of S. cerevisiae and C. glabrata versus other yeasts; SET6 might have been deleted from the genome of Y. lipolytica or, alternatively, a duplication of SET5 to yield SET6 might have taken place, after the separation of Y. lipolytica from the other yeasts 23. We note also that both SET5 and SET6 are absent from S. pombe, indicating that these genes encode functions limited to the specific needs of the Saccharomycotina group.

Analyses of the distribution of the fungal SET-domain types will be carried out in the context of the transition from unicellular to multicellular filamentous, and to animal/plant multicellular systems. We shall follow genes that are present in: all studied genomes, Saccharomycotina-specific genes, families that are excluded from the Saccharomycotina, and those that are found in specific fungal groups.

Despite a broad range of lifestyles, life cycles, and mating mechanisms 67, the numbers of SET-domain genes in the respective yeast species is remarkably constant and the gene structure is highly conserved. Neither the large Y. lipolytica genome size nor the extensive gene-loss in C. glabrata is reflected by the numbers of the SET-domain genes suggesting that in yeasts, the collection of SET-domain genes has been selectively maintained. The permanently filamentous yeast A. gossypii displays high genomic synteny with S. cerevisiae but has a very different morphology and occupies a specific niche: a plant pathogen. A. gossypii grows as multinucleated hyphae in the subtropics, while S. cerevisiae proliferates as single cells associated with sugar-containing fruits 23676869. C. glabrata, closely related to S. cerevisiae, is a pathogen living on mucosal human tissues, while the more distant pathogenic C. albicans grows as yeast but may switch to hyphal growth when exposed to host serum. Genes from pathogenic species that differ from S. cerevisiae might have a role in the virulence 2370. Despite the strikingly different biology of thee yeasts, their SET-domain genes are fully shared suggesting orthologous functions. In such a context, the yeast SET-domain genes cannot be considered critical for the dimorphic transitions, for hyphae formation, for long-range nuclear dynamics, or as factors contributing to the pathogenicity.

SET-domain genes do not seem to be underlying differences in the life cycles and sexual mechanisms adopted by the individual species, either. D. hansenii is homothallic with essentially haplontic life cycle and Y. lipolytica is heterothallic (self-sterile) with a haplo-diplontic cycle, while sexual cycles are unknown for C. glabrata 2367 and, until recently, for C. albicans 71. Although C. albicans has mating-type-like (MTL) genes that resemble the mating-type genes of S. cerevisiae, C. albicans is able to mate under anaerobic conditions reflecting its adaptation as an anaerobic parasite 72. Finally, D. hansenii and Y. lipolytica have one mating type-locus, whereas C. glabrata possesses three mating type-like loci with configurations similar to that of S. cerevisiae 73. Given the importance of the diversity of sexual mechanisms for the evolution of the species, it is remarkable that these mechanisms are, most likely, not connected with the evolution of the yeast SET-domain genes. We emphasize, however, that this conclusion does not preclude participation of SET-domain genes in these processes; rather, it suggests that the different biology of the species is not connected to specific diversifications of the SET-domain genes per se. It agrees with observations from whole-genome analyses that almost identical gene sets control diverse cellular functions in yeasts suggesting that orthologous genes might not play identical cellular roles in different yeast systems 236769. Differences, like the presence of SET4 in S. cerevisiae and C. glabrata and the absence of SET6 from Y. lipolytica are, most likely, connected to the fermentative lifestyle 74 and to metabolic specificities of organisms, rather than with differences in the morphology, in sexual mechanisms, or in pathogenicity.

The SET9 (SUV4-20), SET-MYND, and Su(var)3-9 genes, found in S. pombe, in filamentous fungi, in animals, and in plants, but absent from the Saccharomycotina, suggest that the encoded activities are not involved in 'core' functions. They are involved in generating marks for silencing mechanisms implying that these silencing mechanisms do not exist in the Saccharomycotina. SET9 (SUV4-20) tri-methylates H4K20, a mark for pericentric heterochromatin 61 and MYND-SET family members (studied only in animals so far) generate H3K36 marks involved in suppression of cell proliferation 4849, while SUV39 methylates H3K9 critical for heterochromatin formation 52. Recent findings have suggested that epigenetic 'ON' marks are reduced in evolution, while the 'OFF' signs have significantly increased; unicellular organisms contain more marks associated with transcriptional activation, whereas mammals contain more modifications associated with repression 56. Indeed, the only known histone methylation marks in S. cerevisiae (H3K4, H3K36 and H3K79) and the enzymes generating them (SET1, SET2, and DOT1, an unrelated to SET domain that methylases K79) are associated with gene-activation 75. Absence of SET9, MYND- SET, and Su(var)3-9 genes from the Saccharomycotina is consistent with this evolutionary trend. However, their presence in S. pombe (and of SET9 in Y. lipolytica) suggests that the 'silencing' marks established by the encoded activities are not signature features of multicellular genomes; moreover, the epigenetic marks might be 'read' differently. For example, SET9 of S. pombe methylates H4K20 but it is a sign for DNA damage response in S. pombe 62 rather than for heterochromatin 6061. Absence of genes encoding silencing marks in the Saccharomycotina, thereby, raises an important question described next.

Answers may be found in the remarkable ability of yeasts to adopt available means to achieve ends that are functionally similar but molecularly different from mechanisms employed by other systems. The most striking example is the loss of the Su(var)3-9 gene from the Saccharomycotina because loss of SUVR39 protein entailed loss of the epigenetic H3K9me mark, and of the entire machinery involved in making heterochromatin 56. Nonetheless, silencing processes involving the MAT cassettes, the ribosomal loci, and the telomeric regions are achieved through mechanisms similar to the assembly of heterochromatin 76. Even more surprising is that close S. cerevisiae relatives do not use the same tools but have evolved species-tailored mechanisms for achieving effects functionally similar to heterochromatin. There is no Sir1 (Silent information regulator 1) in C. glabrata, no Sir1 and Sir3 in A. gossypii; neither Sir nor the RNAi-pathways are conserved in D. hansenii and none of the S. cerevisiae heterochromatin factors was found in Y. lipolytica 67. Collectively, the data illustrate the great evolutionary divergence of 'invented' mechanisms producing silencing effects in systems that have lost the epigenetic H3K9me mark.

Furthermore, loss of SET9 and MYND-SET related genes encoding gene-silencing functions has resulted in evolving mechanisms that take advantage of the degree (mono-, di-, or tri-) methylation of the lysine NH2- groups to achieve different transcriptional outcomes for pertinent genes. In S. cerevisiae, di- (H3K4me2) or tri- (H3K4me3) methylated lysines are associated with non-active, or actively transcribed sequences, respectively 7778; in Arabidopsis, H3K4me2 marks are found at genes, independently of whether they were transcribed or not, while the H3K4me3 marks were enriched at active loci 79. In metazoa, both modifications are associated with actively transcribed genes but are differentially distributed along the gene sequence with the H3K4me3 marks accumulated at the transcription start-sites 2033.

Thereby, the amount of methyl tags on the same lysine residue may function as repressive or activating signs in S. cerevisiae, while the fission, filamentous, and higher multicellular systems use additional signs carried by a larger number and diversity of methylation marks.

Consistent with their greater morphological and developmental complexity, the filamentous fungi have a greater number of genes than their unicellular relatives. Conserved proteins among the filamentous fungi may be implicated in their morphogenesis, while non-conserved, species-specific genes might be associated with pathogenic capabilities 1617808182838485. According to these criteria, the different types and numbers of SET-domain genes in the Pezizomycotina compared to the yeasts (Fig. 3) might encode functions linked with hyphal growth, mycelia development, or other processes underlying cellular morphogenesis and virulence. Furthermore, the genes from the JmjC-SET, TPR-SET, and SET5/6 related filamentous fungal genes encode fungal functions specific for the entire filamentous group, while the cluster of five-related genes present in M. grisea, but not in N. crassa, might be associated with life as a pathogen. Whether the M. grisea-specific SET-domain cluster is related to its pathogenicity is unknown. Variations in the JmjC-SET and the TPR-SET gene numbers in particular genomes (Fig. 3) are, most likely, due to species-specific deletion/duplication events. Whether these differences in copy numbers are linked with the pathogenicity of the species has not been demonstrated.

We note also the phylogenetic relationship between the filamentous fungi-specific TPR-SET and SET5/6 related proteins, and the Saccharomycotina-specific SET5 and SET6 proteins (Fig. 1). It illustrates the divergent evolutionary paths of a common ancestor at the separation of the two Ascomycota subgroups. Similarly, the phylogenetically related SET domains of the JmjC-SET and SET3/SET4 subgroups suggest that they result from subphylum-specific divergence of a shared ancestor. The SET3-lineages has been conserved throughout the evolution and its duplication in S. cerevisiae and C. glabrata has produced SET4, while duplication in the filamentous group has produced the JmjC-SET cluster (Fig. 2c).

SET-domain genes of multicellular organisms not found in any unicellular fungi may be involved in multicellular, rather than in 'core' functions. Members of the ASH1, the G9A subfamilies, and of the E(z) family are such candidates.

In contrast to the 'core' function encoded by SET2, the ASH1-related copy found in the filamentous fungi has relatives only in multicellular organisms. The descendants of the ancestral ASH1 lineage in extant animal and plant genomes has evolved further by acquiring additional structural motifs. This process could be linked with the evolution of more complex morphology and communication systems operating in the higher multicellular systems.

No paralogues of the E(z) genes have been recognized in the genomes of unicellular yeasts suggesting a later occurrence in the evolution. The structure of the E(z)-type SET-domain peptide may illustrate a discrete step in the evolution of the SET-domain protein structure/function. A Zn-finger involved in substrate specificity of histone methyltransferases 86 is formed by the Cys in the consensus NHXC sequence and the post-SET domain motif (CxCxxxxC) conserved in a subset of SET-domain proteins. Loss of this structure is, most likely, a secondary event triggered by a single substitution of the C in the NHXC box leading to the loss of the post-SET domain and subsequent divergence of the biochemical specificity.

Occurrence of E(z) might be linked with the transition to multicellularity, as suggested by the lack of K27 marks and E(z) genes from the unicellular yeasts, Dictyostelium, and Chlamydomonas. However, abundant H3K27me marks were found in Tetrahymena 5, suggesting that H3K27me marks have evolved for species-specific needs as well. It will be very informative to establish the roles of the E(z) genes in Tetrahymena, particularly in view of the ability of animals and plant E(z) paralogues to assemble specific Polycomb-group complexes 87888990. Furthermore, H3K27me and H3K4me marks establish a bivalent chromatin state at the nucleosomes of developmentally regulated genes in animal stem cells 91 and in Arabidopsis 92, defining them as histone marks for differentiation processes.

Particularly important is the presence of conserved peptide domains related to the SANT-domain family, present in the E(z) proteins of animal and plant origins. Furthermore, in the reconstructed phylogeny based solely on SET-domain sequences, the E(z) animal and plant proteins cluster together with very high bootstrap values (99%). The conserved structure suggests that the last shared ancestor for the E(z)-lineage carrying these domains has existed before the divergence of the animal and plant genes. This putative ancestor has either originated after the separation of the fungal lineage, or has been deleted in the fungal ancestor. A similar pattern was noted for the evolution of the TRX genes, suggesting that the PcG/TrxG mechanism has been evolutionarily conserved in the animal and plant kingdoms but not in the fungal kingdom.

Recent large-scale comparison of sequence data has found over 40 N. crassa genes with no identifiable S. cerevisiae homologues 16. It was suggested that these "orphan" genes might have resulted from either loss of related genes from the S. cerevisiae lineage, from horizontal gene transfer into the N. crassa lineage, from newly generated 'innovative' genes, or from exceptional gene-divergence in S. cerevisiae 16. Our analysis is congruent with processes of deletion from the Saccharomycotina genomes and with the assembly of novel genes in the filamentous fungi, as credible fates for the evolution of orphan SET-domain genes.

The SET9, SET-MYND, and Su(ver)3-9 genes appear to have been lost from the Saccharomycotina genomes or from the lineage leading to the Saccharomycotina. Supporting evidence comes from the finding of highly conserved sequences for each of these three genes in Rhizopus orizae, which is an outgroup to Ascomycota and a member of the basal group, Zygomycetes. Existence of the orthologous protein candidates in the R. orizae genome (RO3G_14129: 48% identical and 66% similar to S. pombe's SET9, RO3G_13695: 25% identical and 41% similar to S. pombe's MYND-SET, and RO3G_16553: 43% identical and 59% similar to S. pombe's Su(ver)3-9) suggests that ancestors of these genes have existed before the divergence of the Ascomycetes, and subsequently have been lost from the Saccharomycotina. The yeast SET9-gene lineage has been lost even later, after the divergence from Y. lipolytica.

An interesting feature in the evolution of the Su(var)3-9 gene is the presence of the chromodomain in animal and S. pombe proteins. Because a chromodomain is absent from the proteins of the filamentous fungi and plants, it was suggested that the ancestral form, before the divergence of animal and fungal lineages, has carried the combination of the two motifs but the chromodomain has been lost from the filamentous fungi after their separation from the fission yeast 50. However, absence of a chromodomain in the Su(var)3-9 orthologue in R. orizae and in Ustillago (a member of the sister group Basidomycetes) suggests an alternative scenario: the ancient fungal Su(var)3-9 lineage did not have a chromodomain and acquisition of chromodomain encoding sequences by the S. pombe and animal Su(var)3-9 genes were two independent events. Su(var)3-9-related genes have remarkably proliferated in Arabidopsis, but no plant homologue carries a chromodomain.

Homologues of E(z), SET2(ASH1)-like, G9a-like, JmjC-SET, and TPR-SET genes were not found in the R. orizae genome. It is plausible that these genes have occurred in the Pezizomycotina, required by multicellularity-related functions. The mechanism of genetic innovation in the filamentous lineage is unclear. Predicted trends underlying the occurrence of novel proteins may include evolution of new protein architecture from preexisting domains (including reshuffling of existing domains) and/or expansion of particular domain families by series of duplications, followed by specialization, to meet the specific needs of a species. It is clear that the JmjC-SET and TPR-SET genes have been retained in the Pezizomycetes for uniquely fungal processes, while the E(z), SET2 (ASH1), and G9a-lineages have been inherited in the animal and plant ancestors, where they have evolved further (Fig. 2).

The limited number of animal and plant SET-domain genes included here represent descendants of lineages found in the unicellular fungal group (SET1, SET2, SET3, Su(var)4-20, Su(var)3-9, and MYND-SET) and of lineages found in the filamentous group (ASH1, G9a, and E(z)). Distinct subgroups, shared by animals and plants, have evolved within the larger families existing in the fungi (i.e., Trithorax, E(z)), as well as plant- or animal-kingdom specific lineages 21939495. Furthermore, some animal-specific SET-domain families (SET8) are distributed throughout metazoan genomes, while others (SET7) are found only in vertebrates.

The phylogenetic analysis allowed us to trace clear distinctions between species-, subphylum-, and kingdom-specific SET domains, as well as to recognize factors involved in core-cellular roles versus those likely to be associated with multicellular requirements. Surprisingly, the collection of SET-domain genes within yeast did not appear critical for differences in lifestyles, abilities to morph, sexual mechanisms, and pathogenicity of hemiascomycetes. However, SET4, SET5, and SET6 encode Saccharomycotina-specific functions and appearance of SET4 parallels the genome duplication of Saccharomycetes believed to be important for their fermentative abilities 42. SET-domain genes found in the filamentous species, but absent from the unicellular sister group, reflect two evolutionary events: deletion from the yeasts genomes (SET9, MYND-SET, and Su(var)3-9) and appearance of novel structures. The latter group involves genes (JmjC-SET and TPR-SET) originating for the subphylum Pezizomycetes-specific roles and genes apparently connected with the occurrence of multicellularity; descendants of these genes are found also in animal and plant genomes (E(z), SET2(ASH1), and G9a). There is no Ascomycota-specific SET-domain family or gene (present in fungal genomes but absent from animals and plants), while there are families found exclusively in plant and animal genomes, as well as plant-specific or animal-specific subgroups. Animal and plant SET-domain genes are ancestrally related with complex fungal proteins with diverse modular elements. A combinatorial assembly of various peptide domains generates enormous possibilities for variation and precision required for functioning and adaptation of multicellular organisms.

The SET-domain sequences were searched from fourteen complete genomes listed in Additional file 1. Twenty-four SET-domain sequences used as queries were collected from various sources, as listed in Additional file 6.

Four search methods were used to mine new SET-domain proteins from the fourteen genomes: BLAST protein similarity searches were conducted by BLASTP 96 using the 24 SET-domain sequences as the queries against the non-redundant database available at National Center for Biotechnology Information (NCBI) with the default settings. To find similar protein regions from unannotated genomic regions, TBLASTN 96 was used to perform similarity searches against nucleotide sequences of the fourteen genomes translated in all six frames. More sensitive searches were performed using the position specific iteration BLAST (PSI-BLAST) 97. Each query was used against individual genomes with the inclusion E-value threshold of 0.001 and four search iterations.

CLUSTALX (version 1.83) 104 was used to generate multiple alignments of SET-domain sequences (with the GONNET series protein weight matrices and gap opening penalty = 10 and gap extension penalty = .20). Due to the highly variable length of the SET-I region, poorly conserved sites across the sequences were removed. Some other highly variable positions were also removed and the alignments were adjusted manually (see Additional file 8).

A draft phylogeny was reconstructed using all of the 182 SET-domain sequences found in this study using the maximum likelihood method implemented in PHYML (version 2.4.4) 105. This draft phylogeny (see Additional file 2) was used in the further analyses. Protein domain architectures for each protein group are shown in Additional file 3. In order to produce a more reliable multiple alignment and phylogenies, we reduced the number of sequences by choosing representative SET-domain sequences using the draft phylogeny as a guide tree. Poorly aligned sequences and those not clustering clearly with any known major SET-domain families were removed. All fungal sequences were retained, while only one each representative hit from plants and animals was chosen from each SET-domain cluster of SET1, SET2, Su(var)3-9, and E(z). The selected representative sequences are indicated in the phylogeny in red in Additional file 2. The final multiple alignment including 113 sequences is shown in Additional file 5. Phylogenetic reconstruction was done using the maximum likelihood method (implemented in PHYML version 2.4.4) and the maximum parsimony method (implemented in PHYLIP version 3.65). The neighbor-joining method was not used because the estimated distance matrix using the JTT substitution model (implemented in PHYLIP 3.65 106) generated estimation errors due to too many substitutions. For the maximum likelihood method, two sets of trees were reconstructed: one with no invariable site and a constant substitution rate among sites, and the other with the proportion of invariable sites and the gamma shape parameter estimated from the data. Reconstructed phylogenies were largely consistent. In our further analysis, we used the maximum likelihood phylogeny using the estimated proportion of invariable sites and gamma shape parameter. For the maximum parsimony method, the input sequence order was jumbled 10 times (see Additional file 9). Phylogenetic confidence was estimated by the bootstrap analysis 107 with 500 pseudoreplicates for all phylogenetic analysis.