Phylogenetic backbone of Ustilaginales

To understand genetic transitions of mating type loci in a broader evolutionary context, we investigated a representative world-wide set of 25 Ustilaginales species (Table S1, shaded in grey). 18 of these species either collected on field trips (6 specimens) or originating from herbarium material (12 specimens) were cultured for further investigation (Tables S1, S2). From 22 species (Table S1) we amplified the well-established marker genes ef1-α, rpb1, lsu rDNA, ssu rDNA and ITS containing 5.8S rDNA encoding elongation factor 1-alpha, RNA polymerase II subunit 1, large subunit rDNA, small subunit rDNA and internal transcribed spacer containing 5.8S rDNA, respectively. Together with the reference sequences of Cintractia limitata, Malassezia globosa, Mal. pachydermatis, Schizonella melanogramma, Sporisorium reilianum, Ustanciosporium standleyanum and Ustilago maydis (Table S1) we calculated a robust multi-gene phylogeny that represents all major groups of Ustilaginales. The phylogeny was rooted with the non-grass smuts Mal. globosa, Mal. pachydermatis, Melanotaenium euphorbiae and Urocystis eranthidis (Figure 1). Bayesian Markov chain Monte Carlo and Maximum Likelihood (ML) analyses revealed identical topologies supporting the monophyly of Ustilaginaceae as well as Ustilaginales with 1.0 posterior probabilities and 100% bootstrap support each. Within Ustilaginaceae, we found one clade dominated by Sporisorium species including S. reilianum and U. maydis (Figure 1, coloured in red), a second clade dominated by Ustilago species including U. hordei (Figure 1, coloured in green), a third clade that consists of S. consanguineum and U. spermophora (Figure 1, coloured in blue) and a forth clade that consists of Tranzscheliella hypodytes and U. williamsii (Figure 1, coloured in yellow). Macalpinomyces eriachnes is resolved as a sister taxon of the ingroup species of the first three clades (Figure 1). Thus, at least four different clades could be defined in the Ustilaginaceae.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Multi-gene phylogeny and interspecific sexual compatibility of Ustilaginales.

Concatenated Maximum Likelihood (ML) analysis of 2571 bp of ssu, ITS, lsu rDNA, ef1-α and rpb1. Circles next to branches indicate bootstrap support values and a posteriori probabilities of Bayesian and ML analyses, respectively. Branch lengths correspond to substitutions per site and abbreviated branches indicate longer branches. Connected squares illustrate hybrid filament formation (bold lines) or pheromone response (thin lines). Numbers in squares represent respective a mating types. Coloured boxes depict different phylogenetic clades (see text). Host ages refer to [60].

https://doi.org/10.1371/journal.pgen.1002436.g001

Diversity of the PR system of Ustilaginales

To analyse the diversity of the PR system, we pursued two sequencing strategies. We first assessed the occurrence of the pheromone receptor genes pra1, pra2 and pra3 in a set of 104 different species of Ustilaginaceae using PCR amplification. To this end, we designed allele-specific degenerated primers based on available sequences of pheromone receptors of U. maydis, U. hordei and S. reilianum. Primers for pra1, pra2 and pra3 were directed against conserved regions overlapping with trans-membrane domain (TMD) 1 and TMD6, TMD2 and the inner loop between TMD5 and TMD6, as well as TMD1 and the inner loop between TMD5 and TMD6, respectively (see Materials and Methods, Table S3). This initial approach revealed fragments of the expected sizes of about 780 bp, 620 bp and 680 bp from three pra1, five pra2 and two pra3 receptor genes, respectively. Subsequently, these sequences were used in addition to the initial reference sequences to design nested degenerated primers, which were again allele specific and directed against conserved regions (for details see Table S3). Thereby, 20 additional PCR fragments were obtained resulting in a dataset of 36 partial sequences of pheromone receptor genes containing 30 novel sequences (Table 1) and six known sequences.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Species collection as well as accession numbers of the 5-gene phylogeny and PR loci-associated genes.

https://doi.org/10.1371/journal.pgen.1002436.t001

In a second approach, we sequenced complete PR loci of U. cynodontis, U. filiformis, U. xerochloae, Me. pennsylvanicum, S. walkeri and the non-grass smut Us. gigantosporum. To this end, we performed genome walks starting either from genes of PR locus-flanking regions or from the pheromone receptor sequences obtained in the degenerated primer approach. Within flanking regions, we chose the highly conserved genes lba and panC (left border a locus and probable pantoate-beta-alanine ligase). For this purpose, we designed gene-specific degenerated primers based on available sequences of S. reilianum, U. hordei and U. maydis (see Materials and Methods). Since degenerated primers directed against flanking genes were applicable for all tested strains we were able to sequence PR loci of Me. pennsylvanicum (a1 locus), S. walkeri (a1 locus), U. filiformis (a1 locus) and Us. gigantosporum (a1, a2, a3 locus) that escaped the described initial approach. Applying BLAST [51] we predicted complete coding sequences of 10 pheromone receptors within these mating type loci. In sum, the two strategies revealed 42 novel sequences of pra receptors from 34 species (Table S4).

To assess the number of different a alleles in our dataset we performed ML analyses of two pheromone receptor sequence alignments comprising either complete coding sequences of 17 pheromone receptors or all available partial sequences, including trimmed sequences from genome walks and published sequences (Table S1). Both phylogenies resolve three mating type-specific clades with 100% bootstrap support for full length sequences and 83, 99 and 100% bootstrap support for partial sequences, showing a very high consistency between the different datasets (Figure 2, Figure S1). Furthermore, each novel gene encoding pheromone receptors, that has been sequenced by use of primers non-specific for certain alleles, groups with sequences of one of the three pra alleles. This suggests that the existence of a fourth PR allele is highly unlikely. In sum, 21 sequences could be assigned to pra1, 13 to pra2 and 13 to pra3. Together with the observed conservation of one receptor per locus this indicates the presence of only three pra alleles in Ustilaginaceae.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Phylogeny of mating type-specific pheromone receptors.

Maximum Likelihood analysis of complete pheromone receptor-coding sequences. Numbers and asterisks next to branches indicate bootstrap (bt) support values and branch lengths correspond to substitutions per site.

https://doi.org/10.1371/journal.pgen.1002436.g002

To support this observation, we also identified pheromone precursor genes in our genome walk data by performing sequence comparisons to mfa genes of U. maydis, S. reilianum and U. hordei. This resulted in the identification of 21 pheromone precursor genes with, at most, two genes per locus. A ML analysis of a pheromone precursor alignment including all 28 available coding sequences confirmed three mating type-specific clades albeit the support in single clades was weaker due to the sparse sequence information of short pheromone sequences (Figure S2). In essence, three pra and three mfa alleles are ancient and unique to Ustilaginales.

To evaluate the occurrence of pra alleles in a phylogenetic background, we mapped species-specific information on a ML phylogeny from partial rDNA sequences (lsu and ITS containing 5.8S) of 108 species of Ustilaginomycotina containing all 104 species that were tested in the degenerated primer approach (Figure S3). All three pra alleles are present in the three major clades of Ustilaginaceae as well as in the non-Ustilaginaceae species Us. gigantosporum showing that these three pra receptors are not restricted to S. reilianum but are apparent in many species. In addition, they do not correlate with phylogenetic groupings. Thus, these data strongly support the hypothesis that the last common ancestor of the Ustilaginaceae had a triallelic PR system whose three alleles are conserved and which gave rise to convergent evolution of biallelic states.

Organisation and genetic diversity of genes at the PR locus of Ustilaginales

So far, we focused on the pheromone receptor and pheromone precursor genes. To examine the precise organisation of the PR locus we analysed 11 a loci of Ustilaginales spanning at least one border gene (Figure 3). Remarkably, there is a high degree of synteny between PR loci of different species regarding genes for pheromones and receptors as well as PRE (pheromone response element) sites. The latter suggests a conserved regulation of pheromone and receptor gene expression via Prf1 homologs. In contrast, the genetic organisation of border genes flanking the PR locus is less conserved. For example, the border genes rba and panC are missing in Us. gigantosporum. In addition, the a1 locus of U. xerochloae is flanked by an inverted homologous gene of um02342 and sr13546 encoding two proteins of unknown function. They locate 106.2 kb upstream on the same chromosome in U. maydis and 43.1 kb downstream in S. reilianum (Figure 3). The first right border a locus genes of Us. gigantosporum represent an inverted sr13582 homolog (protein of unknown function) and two homologous genes that locate at the same chromosome 81.1 kb downstream of the a1 locus of U. maydis (um02414 and um02415; related to dihydrouridine synthase and related to anti-silencing protein 1) and 92.5 kb upstream of the a2 locus of S. reilianum (sr10827 and sr10828; related to tRNA dihydrouridine synthase and related to anti-silencing protein 1). A left border gene of the S. walkeri a3 locus preserved only the first of three introns that were observed in the homologous genes um02380, sr13588 and in a respective homolog of Us. gigantosporum (protein of unknown function). Furthermore, the panC homolog of S. walkeri is inverted (Figure 3). These differences between inner and outer regions of PR loci provide evidence for differential constraints on recombination comprising strong conservation of mating type regions and weak dynamics in the evolutionary history of flanking regions.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Genetic structure of mating type a gene clusters of Ustilaginales.

Shown are three a locus alleles of different Ustilaginales species. Arrows indicate coding regions of respective genes and lines represent non-coding or intron regions. Pheromones and cognate pheromone receptors are depicted in red shades. Homologous border genes are depicted in identical colours or patterns. Strokes represent pheromone response element sites (ACAAAGGGA) with no (black) or one mismatch (grey). Abbreviation signs depict connected regions on respective chromosomes. um and sr gene numbers correspond to gene identifications on MUMDB [104] and MSRDB [105].

https://doi.org/10.1371/journal.pgen.1002436.g003

We next addressed whether interspecific genetic diversity of single genes reflects the differential conservation of gene organisation between PR loci and their flanking regions. For this purpose, we calculated the nucleotide diversity π from all genes of the PR locus and its flanking regions. Since single gene datasets each contain sequences of different species, we considered their individual phylogenetic diversity (pd) based on the five-gene phylogeny described above and divided π by pd. The pd index indicates the proportional branch length in relation to the total branch length of the phylogeny [52]. Genes within the PR locus, namely lga2, rga2 and genes for pheromone receptors and pheromones show significantly increased nucleotide diversity π in comparison with the flanking genes lba, rba, aro4, coding for a probable phospho-2-dehydro-3-deoxyheptonate aldolase, ORF1, panC, as well as the house-keeping genes rpb1, ef1-α, lsu rDNA, ssu rDNA and ITS rDNA including 5.8S (Figure 4). The diversity of the pheromone genes is most probably even higher since gap positions in their alignment are not considered in DnaSP diversity calculations (Figure S4). Unlike other flanking genes, homologs of um02380 revealed increased nucleotide diversity similar to pra3 (Figure 4). Hence, the increased nucleotide diversity of mating type genes and PR locus-flanking genes contrasts the conservation of their gene organisation and suggests accelerated mutation rates for the highly syntenic PR locus genes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. Nucleotide diversity of PR loci-associated and house-keeping genes.

Bars indicate nucleotide diversity (π) estimates divided by the phylogenetic diversity (pd) of respective datasets. Black bars: a locus genes, dark grey bars: a locus-flanking genes, light grey bars: house-keeping genes. Numbers above bars indicate the quantity of analysed sequences.

https://doi.org/10.1371/journal.pgen.1002436.g004

Increased nucleotide diversity could relate to adaptive changes that were driven by specific evolutionary constraints. In order to compare the evolutionary constraints of PR locus-associated genes we used seven codon site models of variable ratios of ω values across sites, which are implemented in PAML v4.3 (see Material and Methods), and calculated likelihood ratio statistics for each dataset of Figure 4. In addition, we analysed datasets from partial sequences of lba and panC. Datasets of lga2 and the pheromones were excluded from the analysis because of the small dataset with only three sequences for lga2 and shortness of the pheromone sequences. The analysis revealed that in each gene ω varied among codons (except ORF1) as the Nsites Model M3 rejects M0 (Table S4). For the datasets of pra2 and panC model M8 (beta&ω), which allows for positive selection, fitted the data better than model M7 (beta), which does not allow for positive selection. As model M8a is not rejected by M8 the identified divergence of both genes rather accounts to relaxed purifying selection than positive selection. In summary, these results support the hypothesis that the investigated PR-flanking regions do freely recombine (except one flanking region of Ustilago hordei).

To investigate whether the increase in nucleotide diversity is linked to specific sites within the encoded pheromone receptors, we predicted transmembrane domains for pheromone receptor sequences and performed sliding window analyses of the nucleotide diversity π and the ratio of non-synonymous and synonymous substitutions (dN/dS ratios) for each allele-specific pheromone receptor alignment. The analyses revealed several diversity peaks within pra1, pra2 and pra3 that slightly resemble each other but neither nucleotide diversity nor dN/dS ratios suggest prominent sites (Figure S5). This shows that diversity peaks and species-specific substitutions scatter almost randomly on the pra genes without hints to differential selection of single sites. In summary, we observed strong synteny of PR loci whose genes accumulated significantly more substitutions than PR locus-flanking and house-keeping genes.

Homologs of lga2 and rga2 in Ustilaginaceae

Besides pheromone- and receptor-encoding genes, a2 loci of U. maydis and S. reilianum contain two additional genes, namely lga2 and rga2. As shown for U. maydis they encode mitochondrial proteins, whose concerted action is responsible for the uniparental inheritance of mitochondria [50]. Sequence comparison applying BLAST [51] furthermore identified homologs of rga2 in respective regions of a2 loci of Ma. eriachnes and Us. gigantosporum. Surprisingly, a3 loci of U. xerochloae, S. walkeri and Us. gigantosporum also encode a homolog of rga2 that locates between homologs of pra3 and mfa3.1 (Figure 3). To assess the homology of these putative rga2 genes, we performed a multiple amino acid alignment of all predicted Rga2 proteins and the reference proteins of U. maydis and S. reilianum (Figure S6). All predicted genes of different species revealed the same intron structure and encoded proteins comprised comparable amino acid sequence identities of 30 to 53% in relation to Rga2 of U. maydis and S. reilianum. Additionally, we applied iPSORT prediction [53] revealing mitochondrial target signals for the putative Rga2 proteins as was reported for Rga2 of U. maydis [54].

Compared to rga2, lga2 is significantly less conserved between U. maydis and S. reilianum. To identify homologs within respective regions of a2 and a3 loci we conducted gene predictions based on U. maydis intron characteristics using the Augustus prediction server [55]. To verify homology of the identified genes to lga2, we furthermore predicted targeting peptide signals in the respective proteins applying iPSORT prediction [53] and screened for functional domains applying SMART [56], [57]. Since lga2 is a direct target of the bW/bE homeodomain transcription factor we additionally searched for promoter sequence identity upstream of the putative lga2 genes. Importantly, in the a2 locus of Ma. eriachnes we found a putative lga2 gene that showed homology to known sequences. This gene displays 32% sequence identity to lga2 of S. reilianum, shows the same intron structure and the gene product contains a mitochondrial target signal and an F-box-like motif. Although this domain does not completely overlap with the predicted F-box-like motif of Lga2 of U. maydis [54] (Figure 5A), both F-box-like motifs are located within a protein region that contains the most shared amino acids (12 out of 20) for all three species (Figure 5A). Based on information of the promoter sequence of lga2 in U. maydis [58], [59], we identified a sequence with high similarity to the His-Kon8 binding site within the 5′ region of lga2 of S. reilianum and Ma. eriachnes indicating the same regulation via bW/bE transcription factors (Figure 5B). In particular, out of 29 nucleotides 18 and 15 nucleotides overlap in S. reilianum and Ma. eriachnes, respectively. However, even lowering stringency and gene predictions based on intron characteristics did not identify a clear lga2 homolog in other species. We only identified an ORF with four introns and a mitochondrial target signal in the a3 locus of S. walkeri. Thus, lga2 homologs are likely lacking in the a loci of Us. gigantosporum. In conclusion, rga2 is not restricted to the a2 locus but also occurs in the a3 locus, where it does not pair with lga2. This indicates a complex mechanism of parental inheritance of mitochondria within Ustilaginales.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. Multiple alignment of Lga2 homologs and their regulatory regions.

(A) Amino acid alignment of Lga2 sequences from reference species (S. reilianum and U. maydis) and proposed sequences of Ma. eriachnes. Dots indicate predicted mitochondria target signals. Arrowheads indicate positions of introns in the respective genes. Dashes represent alignment indels. Grey shades mark positions with two (light grey) or three (dark grey) identical amino acid residues. Bold letters indicate predicted F-box-like motifs. (B) Nucleotide sequence alignment of the lga2 b-binding site and its flanking regions of U. maydis with 5′ sites of lga2 of S. reilianum and Ma. eriachnes. Grey shades mark sites with two (light grey) or three (dark grey) identical aminoacids and nucleotides, respectively.

https://doi.org/10.1371/journal.pgen.1002436.g005

Interspecific compatibility in Ustilaginales

The dimension of intercompatibility within grass smuts is still unclear and a representative dataset that gives an overview of the whole Ustilaginaceae and beyond is currently missing. Consequently, we screened a representative set of seven species for interspecies sexual compatibility (summarised in Figure 1). Firstly, we monitored the development of conjugation hyphae in liquid media indicating an active PR locus-dependent pheromone response. Secondly, we examined the formation of hybrid filaments on plates containing potato dextrose and charcoal (PD-CC), indicating plasmogamy and the activity of compatible HD alleles. Finally, we illustrated interspecific sexual fusion and filament formation for two examples using scanning electron microscopy (SEM).

Initially, we optimised mating conditions under which all tested species showed an adequate intraspecific mating behaviour. Whereas each species efficiently formed filaments on PD-CC plates, the mating reaction in liquid media distinctly varied between species. Although most compatible strains of one species developed mating structures in water and liquid PD, the reaction in PD was significantly weaker (data not shown). However, since U. cynodontis and U. xerochloae only mated in liquid PD, each mating assay applying liquid media was performed in water and in liquid PD (Table S5).

In the first two series, 720 single mating tests were performed comprising two replicates of 120 different mating tests under the three conditions described above (water, liquid PD, PD-CC plates). The 120 different mating tests consisted of 11 intraspecific and 109 interspecific confrontations. From 109 different interspecific confrontations 18 resulted in a distinct mating reaction (Figure 1, Tables S5 and S6). Figure 6A and 6B exemplify the three recognised interaction types comprising conjugation hyphae formation (Figure 6A v) followed by filament formation (Figure 6B v), conjugation hyphae formation (Figure 6A vi) without filament formation (Figure 6B vi) and yeast-like growth without any reaction (Figure 6A, 6B vii). In three cases, namely S. scitamineum MAT2 confronted with U. xerochloae a1 or a3 and S. reilianum a1 confronted with U. xerochloae a3, we detected only very few hybrid filaments without respective observations of conjugation hyphae in liquid media (Tables S5, S6) which, most probably, is the result of a very low mating rate.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 6. Interspecific mating reactions between different species of Ustilaginales.

(A) Differential interference contrast (DIC) images of mating assays in liquid potato dextrose. Images i–iv show yeast cultures of respective species. Images v, vi and vii show confrontations of S. reilianum with S. scitamineum, Us. gigantosporum and U. cynodontis, respectively. All figures are scaled equally. bar: 10 µm, b?: unknown b allele. (B) Filament formation on charcoal-containing potato dextrose media. Images i–vii correspond to sample descriptions in A. Figure width represents 3 mm. (C) SEM images of mating assays of S. reilianum and S. scitamineum (i–iii) and U. cynodontis and U. xerochloae (iv–vi). Single yeast cells (i, iv) form conjugation hyphae that fuse (arrowheads in ii, v), expand and form empty sections by the insertion of basal septa (arrows in iii, vi). bar: 4 µm, dotted bar: 1 µm.

https://doi.org/10.1371/journal.pgen.1002436.g006

Each tested species revealed intercompatibility at least with two other species including matings between pairs of closely and distantly related species. All five interspecies matings with Us. gigantosporum, that means across the Ustilaginaceae family border, stimulated the formation of conjugation hyphae that did not fuse (Figure 1, Tables S5 and S6). Within Ustilaginaceae hybrid filament formation was observed for all interspecific crossings except for the crossing of S. reilianum a3b1 with U. maydis a1b1.

SEM revealed that haploid sporidia of different species (Figure 6C i, iv) form conjugation hyphae that fuse through a thickened fusion site (arrowheads in Figure 6C ii, v). Conjugation hyphae of S. scitamineum are significantly thicker than those of S. reilianum (Figure 6C ii). Upon fusion, hybrids of S. reilianum and S. scitamineum as well as U. cynodontis and U. xerochloae form filaments that expand at the apical growth cone and form characteristic empty sections via insertion of retraction septa at the basal pole (arrows in Figure 6C iii, vi). This clearly confirms the sexual compatibility between different grass smut species and emphasises their increased potential for hybridisation that, considering the phylogenetic background of their hosts [60], has been preserved for more than 100 million years of evolution.

To find out whether the development of interspecific mating structures is directly linked to pheromone signalling we used two haploid strains of U. maydis (a1b1 and a2b2) that express Gfp under the control of the mfa1 promoter. In these strains Gfp expression is specifically increased in response to pheromone recognition ([46], Materials and Methods), thereby serving as a molecular readout for active pheromone signalling. Both Gfp strains were confronted with 14 different haploid strains of six Ustilaginales species and screened for Gfp fluorescence. As a positive control, we used two haploid wild type strains of U. maydis (a1b2 and a2b1) and the respective compatible Gfp strain. For quantification of Gfp expression three independent experiments were performed. Consistent with the results of mating assays described above, only combinations of the U. maydis a1b1 Gfp reporter strain with S. scitamineum and S. reilianum induced mating structures (Figure 7A). The quantification of the fluorescence revealed that interspecific confrontations with S. scitamineum (MAT2) and S. reilianum (a2+a3) induced significantly less fluorescence than intraspecific confrontations with compatible wild type strains of U. maydis (Figure 7B). These differences are consistent with the quantitative differences of sexual structures observed in interspecific matings. Thus, reporter gene expression illustrates that similar to intraspecific crossings interspecific mating also induces pheromone signalling, indicating the deployment of the same physiological and molecular network in both events.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 7. Interspecific induction of mating via pheromone signalling in U. maydis.

(A) Differential interference contrast (DIC) and fluorimetric images from positive pheromone response reactions in liquid potato dextrose. Conjugation hyphae are formed by both mating partners (DIC images). All figures are scaled equally. bar: 10 µm. (B) The diagram illustrates fluorimetric measurements (relative fluorescence units, RFU) from mating assays of U. maydis P_mfa1-egfp strains FB1 (a1b1) and FB2 (a2b2) confronted with different mating types (a1, a2 and a3) of different smut species in water. Black and grey bars refer to RFUs of confrontations with strain FB1 P_mfa1-egfp and strain FB2 P_mfa1-egfp, respectively. U. maydis wild type strains FB6a (a2b1) and FB6b (a1b2) were used as positive controls. The white bar depicts RFU of the mating of FB1 P_mfa1-egfp and FB2 P_mfa1-egfp. Error bars indicate standard deviations of three independent experiments.

https://doi.org/10.1371/journal.pgen.1002436.g007

Genetic organisation and evolution of a mating type loci in Ustilaginales

Sexual identity in basidiomycetes is determined by a few genes that reside at two specific genomic regions, the so-called mating type loci [61]. Studies on many model organisms, e.g. Coprinopsis cinerea [16], Cryptococcus sp. [62], Microbotryum violaceum [23], Schizophyllum commune [63] and Ustilago maydis [45], revealed that the mating type genes and the mating-dependent signalling network are conserved across large phylogenetic distances. By contrast, the genetic structure of both sex-determining regions is remarkably diverse resulting in bipolar and tetrapolar species with two or multiple alleles of mating type loci [18]. The mating type locus encoding the pheromone-receptor (PR) system shows diverse genetic determinations that have supposedly evolved from rather simple ancestral types via individual translocations, gene duplications and fusions to the second mating type locus [61].

Studying a representative set of different species of Ustilaginales we could show that three PR alleles, which were until now only described for S. reilianum [42], are conserved among members of Ustilaginaceae and most likely represent the plesiomorphic character state of this group (Figure 1 and Figure 2). In consequence, the less frequently observed biallelic states as reported for S. scitamineum [36], U. maydis [39] or for species of the conspecific group of U. hordei and its close relatives [36] should have evolved independently from triallelic states at least three times. With opposite mating obligatory for the pathogenic development of grass smuts it is highly unlikely that the genetic transitions of their PR system from tetrapolarity to bipolarity occurred as a result of loss of function as shown for Coprinellus disseminatus [26]. However, it remains unknown whether those transitions in the mating type loci followed degeneration processes as proposed for U. maydis [39] or whether the linkage of the PR and HD locus as shown for U. hordei [41] predominates in Ustilaginaceae. Since most basidiomycetes with a biallelic PR locus are bipolar and several fungal examples propose an evolutionary trajectory from tetrapolarity to bipolarity via chromosomal linkage of both mating type loci [64]–[66], we propose that most of the biallelic species of Ustilaginaceae have chromosomally linked mating type loci.

The necessity for Ustilaginales to mate in order to conserve their parasitical niche as well as to assure sexual recombination imposes strong selection pressure towards successful mating. In general, diversity levels of reproductive genes in many taxonomic groups show rapid diversification of sex-related genes [67]. Although the precise selective forces driving this diversification and their functional consequences for mating biology are poorly understood, accumulating evidence suggests an adaptive co-evolutionary process as a main driving force for increased diversification of reproductive genes [67]–[69]. Consistently, in Ustilaginales the mating type-specific genes pra1 to 3, mfa1 to 3, rga2 and lga2 revealed increased interspecific diversity compared to either a locus-flanking or house-keeping genes (Figure 4). At least for pheromones and their cognate receptors a co-evolutionary scenario is likely since interacting genes reside on different alleles and their expression patterns are similar [70]. This would suggest similar constraints on their evolutionary rate as shown for Saccharomyces cerevisiae [71]. An additional aspect that could promote diversification of mating type genes is the functional plasticity and broad specificity of the PR system. Since small changes within pheromone and receptor genes do not necessarily lead to loss of function, they are rather under relaxed than strict purifying selection favouring their rapid diversification.

Lga2- and Rga2-dependent inheritance of mitochondria

In most sexually reproductive eukaryotes, stochastic and deterministic processes induce uniparental inheritance (UPI) of mitochondria. Both UPI and biparental inheritance (BPI) of mitochondria entail advantages and disadvantages regarding mitochondrial recombination, evolutionary conflicts and energy balance [72]–[75]. In U. maydis, UPI is a deterministic process depending on the interplay of the mating type-specific proteins Lga2 and Rga2 [76]. Whereas Lga2 blocks the fusion of parental mitochondria and mediates their uniparental elimination, Rga2 protects against Lga2-dependent elimination [50]. In S. reilianum and U. maydis lga2 and rga2 genes are restricted to the a2 allele [39], [42]. We could show that in the case of Ustilaginales with three PR alleles this restriction to the a2 allele rather constitutes an exception since rga2 genes of S. walkeri, U. xerochloae and Us. gigantosporum additionally reside within a3 alleles (Figure 3). Since a homolog of lga2 is missing in the three mating type loci of Us. gigantosporum and the a3 allele of S. walkeri contains a predicted gene coding for a novel protein with a mitochondrial targeting signal, our data strongly suggest a more complex or even species-specific mechanism of mitochondrial inheritance in grass smuts, involving different combinations of a mating type-specific genes. Referring to the role of Lga2 and Rga2 in U. maydis [50] and based on our sequence data (Figure 3), one mechanism could encompass sexual fusions of species with three PR alleles resulting in either UPI or BPI of mitochondria depending on the combination of the two a mating types. In particular, mating of a1 and a2 strains would result in UPI whereas mating of a1 and a3 strains as well as a2 and a3 strains would result in BPI, thereby uniting uniparental and biparental inheritance of mitochondria in one species.

Interspecific sex and hybridisation-based speciation in Ustilaginales

Hybridisation can lead to substantial genomic changes and thereby gives rise to various novel phenotypes [77]. In consequence, hybridisation has been frequently discussed with regard to its role in evolutionary adaptation and diversification for various organismic groups (reviewed in [78], [79]) including fungi [80]. In order to hybridise parental species have to overcome pre- and postzygotic barriers requiring interspecific sexual compatibility [81]. Using a set of seven species we demonstrated that sexual intercompatibility up to the stage of plasmogamy (Figure 6) is common within Ustilaginales bridging more than 100 million years of evolutionary differentiation (Figure 1). In addition, in the investigated crossings interspecific sex activates the same signalling machinery as intraspecific sex (Figure 7), emphasising the functional redundancy of self and non-self pheromones and their cognate receptors. Referring to studies in various animals and plants (reviewed in [82]), this broad intercompatibility between closely as well as distantly related species of Ustilaginales could lead to hybridisation events more frequently than previously expected. In most cases, hybridisation effects introgression but sometimes it also initiates hybrid speciation [83]. The more frequent emergence of newly combined genotypes would increase the probability for one genotype to arise that exhibits a higher fitness compared to its parental species or that enables the exploitation of a novel ecological niche [78]. Such niche differentiation is known from homoploid hybrids of several smut species including closely related species, e.g. the conspecific group of U. hordei and its close relatives as well as distantly related species like S. reilianum and S. cruentum [36]. In addition, co-phylogenetic studies of Ustilaginaceae and their hosts revealed evidence for hybridisation events in Ustilaginaceae. In particular, there is much incongruity between both topologies [84] that, referring to the strong host specificity of grass smuts, was assumed to result from common host jumps and/or hybridisation events [38]. Although it is not clear how these host jumps occurred, as in highly adapted species this might involve complex genetics, our data and several observations of natural hybrids (summarised in [36]) highlight the potential relevance of hybridisation in grass smut speciation. Nevertheless, there are reproductive barriers between intercompatible grass smut species as shown for U. maydis and S. reilianum that independently established on the same host and coexist without evidence for “natural” hybrids [85].

Thus, it remains unclear whether mating specificities are directly linked to host specificities or if mating specificities and mating efficiency change after the establishment of new host specificities. However, single outbreaks, as the rust fungus hybrid Melampsora xcolumbiana on Populus hosts [86], emphasise the ecological relevance of novel hybrid-based genotypes. Hence, the future challenge will be to track the distribution of hybrids among natural populations and to examine their individual ecological potential.

Species selection, fungal cultures, and growth conditions

For phylogenetic analyses 104 species of Ustilaginales were analysed in total (Table S1, Figure S3). Seven of the species, namely Melanopsichium pennsylvanicum, Urocystis eranthidis, Ustilago avenae, U. cynodontis, U. filiformis, U. williamsii and Ustanciosporium gigantosporum were collected in field trips for this study. For Cintractia limitata, Malassezia globosa, Mal. pachydermatis and Schizonella melanogramma we used sequence information from GenBank [87].

From 24 species we used cultures that either originate from collaborators (5 species) or were cultured from herbarium material in this work (19 species, Table S2). The strains of 22 species were deposited at CBS. To increase the germination success of spores and to separate sporidia, spores were germinated in three different liquid media (complete media (CM) [88]; potato dextrose (PD); water) with shaking at 16°C and 28°C. If necessary, kanamycin was added to the media (100 µg/ml). Single haploid yeast cultures were isolated from streak plates (PD) of liquid cultures with germinated spores. U. maydis strain FB2 P_mfa1-egfp was constructed by transformation of progenitor strain FB2 (a2b2) with linearised plasmid pmfa1-egfp-cbx [46]. Homologous integration event at the ip locus was verified by Southern analysis [89].

Species identity of new cultures was checked by ITS rDNA sequencing (Table S1, see below). In mating assays we used only verified single yeast cultures of 18 haploid cultures of 7 different species in total, namely Sporisorium reilianum, S. scitamineum, U. cynodontis, U. hordei, U. maydis, U. xerochloae and Ustanciosporium gigantosporum (Table S2). For further experiments isolated strains were stored at −80°C in PD-glycerine and re-grown at 28°C on PD or CM agar plates.

PCR conditions and sequencing

Genomic DNA from yeast cultures was isolated by the method of [90]. Genomic DNA from herbarium material was isolated with DNeasy96 Plant Kit (Qiagen, Hilden). ITS rDNA containing 5.8S was amplified using the primers ITS1 and ITS4 [91]. Partial ssu rDNA, lsu rDNA, rpb1 and ef1-α were amplified using the primers NS23 and NS24 [92], LR0R and LR6 [93], [94], RoK157 and RoK158 (Table S3) and 987F and 1567R [95], respectively. Detailed primer descriptions are given in Table S3. Primer properties were evaluated with OligoCalc [96], [97] or Clonemanager v9.0 (Sci-Ed Software). Primers were obtained from SIGMA-ALDRICH (Hamburg). All PCR amplifications were performed on a PTC-200 Thermo Cycler (MJ Research). For DNA amplification ≤5 kb Phusion® High-Fidelity Polymerase (Finnzymes, Espoo) or peqGold Taq DNA Polymerase (Peqlab) and for >5 kb KOD Xtreme™ Polymerase (Merck Biosciences, Nottingham) were used following manufacturer's instructions. PCR products were purified directly or through gel purification using my-Budget Double Pure Kit (Bio-Budget). Purified fragments were sequenced on an Abi 3130XL sequencer (Applied Biosystems) by the sequencing service of the Biochemistry department at the Ruhr-Universität Bochum or by GATC Biotech AG Konstanz. Nucleotide sequences of ITS, lsu, ssu, ef1-α, rpb1, pra1, pra2, pra3 and mating type loci have been deposited at GenBank under the accession numbers JN367287 - JN367447 as listed in Tables S1 and S4.

Genome walking procedure

In order to obtain complete a loci sequences, degenerated primers were used to amplify pra1, pra2, pra3, lba and panC in different species. Initial primer design was based on published sequences of U. maydis (MUMDB), Sporisorium reilianum (MSRDB) and U. hordei (GenBank; [87]). Genome walks started from amplified regions applying the GenomeWalker™ Kit (Clontech Laboratories, Mountain View) and following manufacturer's instructions. Completed loci were checked by long-range PCR and enzymatic digestion. A detailed primer list is given in Table S3.

Phylogenetic reconstructions

Sequences were quality checked and hand edited using Sequencher 4.8 (Gene Codes Corporation). Nucleotide alignments were performed with MAFFT 6.707 [98] in default mode using a maximum number of 1000 iterations. Amino acid alignments were performed with BioEdit [99] applying ClustalW [100]. Afterwards, leading and trailing gaps were removed manually from the alignments except for ITS alignments which were trimmed using Gblocks v0.91 on the MABL server (http://www.phylogeny.fr) applying all less stringent settings. Maximum Likelihood (ML) [101] analyses were performed with RAxML 7.0.4 [102]. RAxML 7.0.4 conducted 1000 bootstrap replicates using a rapid bootstrap algorithm [103] applying GTRMIX approximation. The more accurate GTRCAT approximation was applied in the subsequent ML search for the best scoring ML tree starting from each 5^th bootstrap tree. Bootstrap support values were drawn at the best scoring ML tree. In multi-gene ML analyses, sequences were concatenated. For each partition RAxML estimated and optimised individual α-shape parameters, GTR-rates and empirical base frequencies. With the partitioned multiple alignment of ITS, lsu, ssu, ef1-α and rpb1 we additionally performed a Bayesian analysis using MrBayes v3.1 [104]. In order to allow the overall evolutionary rate to be different across partitions, the evolutionary model was applied individually and parameter estimations were unlinked. Monte Carlo Markov chains (MCMC) were run over one million generations under the GTR+I+G model. Trees were sampled every 100 generations leading to 10,000 trees. To check for overall convergence, this approach was repeated four times with random starting trees. After examination in Tracer v1.5 (http://tree.bio.ed.ac.uk/software/tracer), a burn-in of 2500 was chosen for each run. Out of the remaining trees a majority rule consensus was calculated to obtain estimates for a posteriori probabilities. All trees were visualised and edited in FigTree v1.3.1.

Gene identification and sequence analyses

In order to identify homologous genes, sequences were compared with GenBank [87] and the genome databases from Ustilago maydis (MUMDB, [105]) and Sporisorium reilianum (MSRDB, [106]) applying BLAST [51]. Furthermore, SMART [56], [57] and iPSORT [53] were used to identify functional domains and subcellular localisation signals in the corresponding amino acid sequences. Since lga2 homology between U. maydis and S. reilianum is weak except for the number of introns and BLAST [51] did not identify homologs, we ran gene predictions in respective mating loci regions using Augustus (http://augustus.gobics.de). Coding sequences of homologous genes were determined manually according to reference sequences from U. maydis, S. reilianum and U. hordei.

Completely sequenced pheromone receptors and the deduced protein sequences were characterised with respect to their predicted transmembrane domains, nucleotide diversity and their dN/dS ratios along the protein sequence. Transmembrane domains were predicted using TMpred [107], as well as MEMSAT and MEMSAT-SVM [108] on the PSIPRED server (http://bioinf.cs.ucl.ac.uk/psipred, [109]). Nucleotide diversity π was calculated with DnaSP v5.10.01 [110] applying Jukes-Cantor correction. To compare nucleotide diversity of different gene datasets, values were divided by the phylogenetic diversity (pd) of respective species subsets. pd was calculated with phylocom 4.1 [52] based on the multi-gene phylogeny (Figure 1).

To analyse how genetic variation distributes along pheromone receptor genes, we performed sliding window analyses of π in DnaSP v.5.10.01 (windowlength: 25, step size: 5). Differential selection of single sites within pheromone receptors was tested applying codeml (implemented in PAML v4.3 [111]–[113]). Since we found no significance for positive selection, we illustrated the proportion of non-synonymous substitutions along the receptor alignments. Therefore, we estimated the average behaviour of each codon for all pairwise comparisons for synonymous and non-synonymous mutations using SNAP of the HIV database website (http://www.hiv.lanl.gov, [114]) and illustrated the ratio of non-synonymous and synonymous values along the amino acid sequence.

Mating assays on PD-CC and in liquid media

Mating assays were performed with haploid strains that were isolated from different species. To determine the mating type and to validate haploidy, we performed mating tests, amplified pheromone receptors and stained nuclei with DAPI. Mating tests were performed with cultures of the same species and with cultures of different species. Densely grown liquid pre-cultures (PD, 200 rpm shaking at 28°C) were diluted in liquid media (PD, pH 8.0) and grown over night (28°C, 200 rpm,) to an optical density OD₆₀₀ between 0.4 and 0.8. Cells were harvested by centrifugation (1000 g, 5 min at room temperature) and pellets were resuspended in distilled water (pH 8.0) or PD (pH 8.0) to a final OD₆₀₀ of 1.0. 150 µl cell suspensions of each strain were mixed and added into 24 well plates. a mating type compatibility and conjugation hyphae formation were screened after 6 and 12 hours of incubation at 28°C using a Zeiss Axiostar microscope. To test for b mating type compatibility, 3 µl cell suspensions were dropped on PD charcoal plates, incubated at 28°C and screened for filament formation after 18 hours using a Zeiss Stemi 2000-C binocular. Mating type-specific primers that locate within pheromone receptors were used to identify and validate opposite mating types of fungal strains (Table S3). For DAPI staining, cells were fixed in 2% formaldehyde for 30 min, transferred to mounting media containing DAPI (Linaris, Wertheim-Bettingen) and analysed using a Zeiss Axio Observer microscope.

Light microscopy

Cell suspensions were dropped on glass slides that were covered with a thin layer of agarose (2% w/v) and analysed using a Zeiss Axio Observer microscope equipped with objective lenses of 40-fold (Plan-Neofluar, 1.3 NA), 63-fold (Plan-Apochromat, 1.4 NA) and 100-fold (Plan-Apochromat, 1.4 NA) magnification. Epifluorescence microscopy was conducted using Gfp filter sets (ET470/40BP, ET495LP, BP525/50) and DAPI filter sets (HC 387/11, HC 447/60, BS 409). Filters were obtained from AHF Analysentechnik (Tübingen). Frames were taken with a CCD camera CoolSNAP HQ2 (Photometrics, Tucson). Microscope and camera were controlled by MetaMorph 7.5 (Molecular Devices, Ismaning). The same software was used for measurements and image processing including adjustment of brightness, contrast and γ values, as well as correction of background unevenness.

SEM microscopy

Cell suspensions were dropped on Poly-L-Lysine coated glass slides. Immediately after drying samples were fixed with 4% formaldehyde - 1% glutaraldehyde (v/v) in 0.2 M phosphate buffer (pH 7.4, modified from [115]) for one hour. Afterwards, samples were rinsed three times in 0,2 M phosphate buffer (pH 7.4), dehydrated in ethanol (50/75/100/100%), transferred to formaldehyde dimethyl acetal (FDA), critical-point dried, sputter-coated with gold-palladium for 200 s and analysed using a DSM 950 scanning electron microscope (Zeiss, Oberkochen, Germany).

Fluorimetric measurements of pheromone induced Gfp

U. maydis mutant strains expressing Gfp under the mfa1 promoters were confronted with a collection of haploid strains from different species and screened for Gfp fluorescence. Due to differences in mating behaviour of different species, matings were performed under two different conditions, in distilled water (pH 8.0) and in liquid PD (pH 8.0), for six hours at 28°C in 24 well plates. After incubation, 200 µl cell suspension was transferred to black-walled 96 well plates and relative fluorescence units (RFU) were measured at room temperature with excitation and emission wavelengths of 485 nm and 520 nm, respectively (bandwidth 9 nm and 20 nm, respectively) using a monochromator fluorescence reader (Tecan, Männedorf). Three independent experiments were performed. In order to compare different measurements of one experiment and due to differences in base fluorescence and optical densities between different fungal species and strains, OD₆₀₀ dependent base fluorescence was subtracted from measured RFUs.