Comparative Genomics of the Mating-Type Loci of the Mushroom Flammulina velutipes Reveals Widespread Synteny and Recent Inversions

Background Mating-type loci of mushroom fungi contain master regulatory genes that control recognition between compatible nuclei, maintenance of compatible nuclei as heterokaryons, and fruiting body development. Regions near mating-type loci in fungi often show adapted recombination, facilitating the generation of novel mating types and reducing the production of self-compatible mating types. Compared to other fungi, mushroom fungi have complex mating-type systems, showing both loci with redundant function (subloci) and subloci with many alleles. The genomic organization of mating-type loci has been solved in very few mushroom species, which complicates proper interpretation of mating-type evolution and use of those genes in breeding programs. Methodology/Principal Findings We report a complete genetic structure of the mating-type loci from the tetrapolar, edible mushroom Flammulina velutipes mating type A3B3. Two matB3 subloci, matB3a that contains a unique pheromone and matB3b, were mapped 177 Kb apart on scaffold 1. The matA locus of F. velutipes contains three homeodomain genes distributed over 73 Kb distant matA3a and matA3b subloci. The conserved matA region in Agaricales approaches 350 Kb and contains conserved recombination hotspots showing major rearrangements in F. velutipes and Schizophyllum commune. Important evolutionary differences were indicated; separation of the matA subloci in F. velutipes was diverged from the Coprinopsis cinerea arrangement via two large inversions whereas separation in S. commune emerged through transposition of gene clusters. Conclusions/Significance In our study we determined that the Agaricales have very large scale synteny at matA (∼350 Kb) and that this synteny is maintained even when parts of this region are separated through chromosomal rearrangements. Four conserved recombination hotspots allow reshuffling of large fragments of this region. Next to this, it was revealed that large distance subloci can exist in matB as well. Finally, the genes that were linked to specific mating types will serve as molecular markers in breeding.


Introduction
The genes that regulate mating in fungi encode sets of regulatory and signaling molecules that are broadly distributed in eukaryotes. The corresponding pathways, which are often linked to pathogenicity and fruiting body formation, comparably control developmental processes in other organisms, such as pattern formation in development and sexual differentiation in animals. Studies in Agaricomycetes, that include many important mushroom-forming fungi, therefore also inform our collective understanding of cellular development in eukaryotes. Mushroom forming fungi are further important sources for food (mushrooms, fermentation), medicine (pathogens, fibers, health-promoting, anticancer products) and green technologies (waste recycling, fertilizers, bioremediation). As such they represent a massive economical asset and better understanding of their sexual propagation is desirable.
Mating is the beginning step in sexual development of mushroom-forming fungi. Their life cycle is characterized by haploid as well as diploid stages. Sexual, haploid spores that are dispersed by a mushroom develop into monokaryotic mycelia. Upon fusion of two genetically different monokaryons a dikaryotic mycelium is established and the different nuclei of the two mating partners coexist within the cells of the dikaryon during vegetative growth and production of fruiting bodies. Only upon maturation of the mushroom, the different nuclei fuse in specialized reproductive cells termed basidia and new haploid spores with separate mating-types are created (for a review see [1]). The heterothallic fungi restrict their self-mating by use of one (bipolar species) or two (tetrapolar species) incompatibility loci. This system supports out-breeding and helps to promote genetic variability in populations.
Our current knowledge on the molecular genetics of mating in mushrooms is primarily based on studies in the model organisms Schizophyllum commune and Coprinopsis cinerea ( = Coprinus cinereus). The mating-type loci in those species are termed A and B and control different developmental pathways that are required to maintain a fertile dikaryon (for reviews see [2,3]). Each mating-type locus consists of tightly linked subloci and encodes multiallelic genes [4,5,6]. These alleles are highly polymorphic at DNA and amino acid (AA) levels [7,8,9] due to balancing selection that favors mating-type alleles that become rare and thus extends the coalescence time between alleles [10]. The A and B subloci are functionally redundant and heterozygosity at a single sublocus is sufficient to activate the respective A or B pathway.
A mating-type loci encode pairs of divergently transcribed homeodomain genes (HD genes) and are typically accompanied by the Mitochondrial Intermediate Peptidase gene (MIP) and a Beta-flanking gene [11,12]. HD proteins are distinguished based on conserved DNA binding motifs; homeodomain 1 (HD1) and homeodomain 2 (HD2). HD1 proteins further contain two nuclear localization signals, an activation domain and only weakly bind DNA [13,14] whereas HD2 proteins lack these domains but have strong DNA binding properties [15,16]. Both HD1 and HD2 proteins possess N-terminal dimerization motifs that facilitate their interaction. Interaction of HD1 with a compatible HD2 protein generates a heterodimer that serves as a transcription factor for the A pathway [15,17].
The B mating-type loci are comprised of pheromone receptors and pheromones (recently reviewed in [3]) and each pheromone receptor is accompanied by one or several pheromones in a sublocus [9,18]. Pheromone genes encode small precursor proteins with C-terminal CAAX motifs (C = cysteine, A = aliphatic and X is any residue) that are farnesylated. They usually contain an acidic AA pair as well (ER or EH in C. cinerea) about 10-15 AA from the C-terminus. Those acidic residues are highly conserved in C. cinerea and have been speculated to be the site of proteolytic cleavage [5,19,20,21,22]. They are also conserved in other basidiomycete pheromones, though not in all [23,24]. After farnesylation and proteolytic splicing, fungal pheromones typically constitute about 9-11 amino acid peptides [19,25].
Fungal pheromone receptors are classified within the Rhodopsin-like superfamily and typically contain seven membrane spanning regions (7-TM). They are further characterized by a short N-terminal extracellular domain and a long cytoplasmic Cterminal tail [26]. The cytoplasmic domains are presumed to dock trimeric G-proteins that can activate the downstream B pathway after phosphorylation [26]. Phosphorylation of G-proteins is triggered by interaction of specific pheromones with extracellular pheromone receptor domains.
The fundamental composition of the A and B mating-type loci was found to be strongly conserved within basidiomycetes. The redundant subloci are a result of doubling during evolution and tetrapolar mating systems were found in all three major linages, suggesting an ancient origin [27,28]. Moreover, bipolar and tetrapolar species have been shown to contain essentially the same genes [24,29]. On the other hand, the new availability of genome sequences reveals significant variability within mating-type loci of basidiomycetes and their numbers of pheromone receptor, pheromone and homeodomain encoding genes differ greatly between species [12,30,31,32]. To this, a new mating-type system that is not strictly bipolar or tetrapolar was discovered [30].
Flammulina velutipes, also known as Winter Mushroom and Enokitake is one of the major cultivated mushrooms in Asia. Beyond having a tetrapolar mating system with multiple alleles, the genetics of its mating-type system have remained unknown. We decided to elucidate the genetic structure of this important mushroom to map the mating-type genes and use comparative genomics to understand evolutionary relevant distinctions of the F. velutipes loci, as well as to implement this knowledge in our mushroom breeding programs. We obtained a complete map of the mating-type genes from F. velutipes KACC42780, identified the specific matA and matB loci and explain some of the events that caused the significant deviation of the matA region in comparison to the model species. Mating-type defining genes are currently used to construct haploid, monokaryotic, mushroom producing strains and primers for PCR based mating-type identification.

Strains and culture conditions
Flammulina velutipes strains were cultivated at 25uC on 100620 mm dishes (SPL, South Korea) containing Potato Dextrose Agar (HIMEDIA Laboratories, India) for two weeks. Stocks were transferred to fresh PDA dishes every three months, sealed (Clean Wrap) and stored at 4uC after three to five days growth (storage up to one year). For genomic DNA isolation, PDA was covered with cellophane (general household) prior to inoculation. For mating experiments, small dishes (60615 mm, SPL) were inoculated with agar plugs (565 mm) at 0.5 cm from the center and grown for one week. Fresh plugs from the center of the plate that contained merged mycelia were subcultured for another four to eight days prior to clamp formation analysis. Mushroom cultivation in small bottle cultures was started with mating of two monokaryons as described above. Confirmed dikaryons were grown on a 30 mm layer of sterilized sawdust (Douglas Fir)/wheat bran (Rice) mixture (4:1) in 25645 mm glass bottles (three mycelial plugs/bottle). After 10 days incubation at 25u the surface of colonized wood was scraped off and bottles were filled with cold distilled water for 1 hr. Following upside down drying for 30 min bottles were incubated at 15uC, 95% humidity for three weeks. Spores were collected by placing the caps of the mushrooms in small dishes for one day and eluted in sterilized water prior to plating on PDA. Strains used in this study were KACC42780 (A3B3), KACC43777 (A4B4) and dikaryon KACC43778 (

Southern hybridization
DNA probes (500bp) were amplified by PCR (primers; Table  S2). Genomic DNA was digested to completion with SacI at 37uC, over night. Agarose gels (0.8%) containing the digested gDNA were soaked in 0.25 M HCl for 20 min for depurination, 30 min in Denaturation buffer (0.5 M NaOH, 1.5 M NaCl pH 7.5) and 10 min in Neutralization buffer (0.5 M Tris-HCl pH 7.5, 1.5 M NaCl). DNA was transferred to a nylon membrane (Amersham) via capillary transfer with 10x SSC (1.5 M NaCl, 0.15 M sodium citrate) for 20 hrs and membranes were baked at 80uC [33]. Prehybridization and hybridization were done in a hybridization solution (5x SSC, 50% formamide, 5x Denhardt's solution, 0.5% sodiumdodecylsulfate and 25 mg/ml denatured salmon sperm DNA). Pre-hybridization was performed for 3 hrs and hybridization was continued for 24 hrs with a probe labeled with [a-32P] dCTP (Ladderman TM Labeling kit, Takara, Japan) at 42uC. The membrane was exposed to an imaging screen (Fuji) for 18 hrs and DNA bands were visualized using a personal molecular imager system (Bio-Rad, USA).

PCR analyses and sequencing
PCRs were performed using GoTaq flexi kit (Promega Korea Ltd, South Korea) with specific primers (Bioneer, South Korea) for each gene (Table S2). Thermal cycling parameters consisted of 10 min denaturation at 94uC, 30 cycles of 94uC, 30 s denaturation, 30 s annealing (temperature 2uC lower than the specific, lowest melting temperature of primers used), 30 s extension at 72uC and a 10 min final extension at 72uC. Samples for sequencing were amplified with a BigDye Terminator cycle sequencing kit (Applied Biosystems, USA), precipitated with 0.1 volume 3 M Sodium Acetate (pH 5.8) and 2.5 volumes 100% EtOH for 10 min at 270uC and centrifuged for 15 min, 13000rpm at 4uC. Pellets were washed with 70% EtOH, air dried and eluted in Hi-Di Formamide (Applied Biosystems, USA) and analyzed on a ABI3730 DNA Analyzer.

Analysis of the Mat A locus
Tblastn searches were performed against a draft genome of F. velutipes (see above), the genome of S. commune strain H4-8, available at DOE Joint Genome Institute (http://genome.jgi-psf. org/Schco1/Schco1.home.html) and the genome of Laccaria bicolor S238N-H82 (available under the Coprinus cinereus project) using predicted proteins that surround the A mating-type locus of C. cinerea okayama 7#130, available under the Coprinus cinereus sequencing project of the BROAD institute of MIT (http:// www.broadinstitute.org/annotation/genome/coprinus_cinereus/ GenomesIndex.html). Gene homologues that were located on F. velutipes contigs Fv01174, Fv03236 and Fv02632 were used for comparison in Chromomapper 1.0.10.26 [41].

Phylogenetic analyses
Pheromone receptor sequences of L. bicolor were obtained from the Joint Genome Institute (JGI) website according to Niculita-Hirzel et al. [31]. Sequences of S. commune, P. djamor, C. cinerea and Cryptococcus neoformans were obtained from NCBI Genbank using accession numbers as reported by James et al. (Figure 7, in [42]). Sequences of F. velutipes were obtained as described above. The protein sequences of pheromone receptors were aligned using ClustalW 1.64 [43] in combination with Gblocks 0.91b [44] to eliminate remaining poorly aligned sequences with a setting 'allow smaller final blocks'. Ambiguously aligned regions were excluded from the protein sequences of pheromone receptors in the phylogenetic analyses. Maximum likelihood analyses were conducted using RAxML 7.04 [45]. A model of PROTGAMMA-WAGF was selected with an analysis of ProtTest 2.4 [46] and incorporated in the analysis. Branch values from 1,000 nonparametric bootstrap repetitions were used for nodal support [47]. In this study, nodes were considered strongly supported when supported by more than 70% bootstrap value. The resulting tree was visualized with Figtree 1.3.1.

Identification of pheromone receptor genes and pheromone precursors
A draft genome of Flammulina velutipes monokaryotic strain KACC42780 (A3B3) was screened for pheromone receptor homologues and CLA4, a kinase gene shown to be linked to the B locus in other mushroom species [12]. Six contigs, five with a single and one with two pheromone receptor homologues were retrieved. CLA4 was found on a separate contig (Table S1). Pheromone receptor genes were annotated and named FvSTE3.1, FvSTE3.2 and FvSTE3.s one to five (FvSTE3.s1 to FvSTE3.s5), where 's' stands for 'similar' to distinguish non mating-type specific pheromone receptors (see below). Repeated searches with the newly indentified genes did not uncover additional pheromone receptors. Six predicted pheromone receptor proteins contained the seven-transmembrane domains that are characteristic for fungal pheromone receptors [26]. Alignments of the pheromone receptor proteins showed that the sequence of FvSte3.s2 deviated in the region that corresponded with the first transmembrane region in other pheromone receptors. Moreover, seven amino acids were deleted in this protein from a highly conserved motif, truncating its transmembrane sequence (not shown). FvSte3.s2 grouped specifically with FvSte3.s1 in maximum likelihood analyses (not shown), shared 60% base pair identity and was separated from gene FvSTE3.s1 by only 6.7 Kb. All together, this suggests that FvSTE3.s2 is a pseudogene that was derived as a copy from FvSTE3.s1. Three pheromone precursor genes named FvPP1, FvPP2 and FvPP3 were identified on contigs with pheromone receptors. FvPP1 was located 2102bp apart from pheromone receptor FvSTE3.1 and FvPP2 and FvPP3 flanked pheromone receptor FvSTE3.2 at distances of 320bp and 427bp respectively ( Figure 1). Notably, FvPp1 contained a C-terminal Tryptophan (W) behind its CAAX-box (CAAXW). All three FvPp proteins were classified as fungal pheromones according to Pfam 24.0 searches [36] with E-values of 0.0013, 0.45 and 0.035 for FvPp1, FvPp2 and FvPp3. Alignment of the FvPp proteins showed that FvPp2 (54 amino acids) and FvPp3 (46 amino acids) shared 55% and 54% reciprocal similarity in contrast to FvPp1 (54 amino acids) with 37%. The three proteins contained a glutamic acid/ arginine (E/R) motif at amino acid position 13/14 (FvPp1) and 15/16 (FvPp2, FvPp3) counted from the C-terminus ( Figure S1). These residues are conserved in various basidiomycetes and have been speculated to be the site of proteolytic cleavage (see introduction). Proteolytic splicing at the E/R sites, together with C-terminal processing, would result in peptides of nine (FvPp1) and 11 amino acids (FvPp2 and FvPp3) which corresponds well with the size of other fungal pheromones [19,25].

Segregation analysis of FvSTE3 and FvPP genes
In order to designate the pheromone receptor and pheromone precursor genes to specific matB loci, their distribution was analyzed in dikaryon KACC43778, parental strains KACC43777 (A4B4), KACC42780 (A3B3) and two monokaryotic siblings B2 (A3B4) and B27 (A4B3). PCR with specific primers (Table S2) detected FvSTE3.1 and FvPP1 explicitly in matB3 strains. Pheromone receptors FvSTE3.2, FvSTE3.s1 to FvSTE3.s5 and pheromone precursors FvPP2 and FvPP3 were detected in matB3 and matB4 mating-types ( Figure 2). Correct amplification and PCR patterns of FvSTE3 and FvPP genes were confirmed by sequencing and Southern analysis ( Figure 2). Analysis of 16 additional monokaryons with different mating-types showed that the observed distribution was persistent (Table 1). Monokaryons were derived as single spore colonies from dikaryon KACC43778 and mating-types were assigned based on clamp formation patterns (material and methods). Sequence alignment of the PCR products for each gene that were obtained from the 16 strains revealed two differing copies of genes FvSTE3.s1, FvSTE3.s2 and FvSTE3.s3. This enabled segregation analysis through restriction fragment polymorphism of FvSTE3.s1 and FvSTE3.s2. Subtype FvSTE3.s1-a was invariably detected together with FvSTE3.s2-d while FvSTE3.s1-b was linked with FvSTE3.s2-c. Segregation of the couples was independent from the analyzed mating-type loci ( Table 1). Segregation of FvSTE3.2, FvSTE3.s3 to FvSTE3.s5, FvPP2 and FvPP3 remained undetermined. Taken together, the results show that only a single pheromone receptor and pheromone segregate specifically with the B3 mating-type in F. velutipes KACC42780.

Polymorphism of pheromone receptors
Mating-type specific genes in fungi are characterized by highly polymorphic alleles. We examined the distribution and polymorphism of the FvSTE3 genes in various F. velutipes strains, from different countries (materials and methods). Genomic DNA, of two compatible monokaryons for each strain, was analyzed by PCR with specific primers (Table S2). FvSTE3.s1, FvSTE3.s2, FvSTE3.s3 and FvSTE3.s4 were frequently amplified with six, five, seven and seven products out of 10 strains. FvSTE3.1, FvSTE3.2 and FvSTE3.s5 were not detected outside the control ( Table 2). New primer sets (Table S2, *additional sets) resulted in frequent amplification of FvSTE3.s5 but FvSTE3.1 and FvSTE3.2 remained undetected. Sequences for the amplified genes shared respectively 92-99 percent base pair similarity (Table 2). Low polymorphism of FvSTE3.s1 to FvSTE3.s5 showed that those genes are not matingtype specific pheromone receptors; their amino acid sequences are identical. The absence of FvSTE3.1 and FvSTE3.2 from all tested F. velutipes strains, especially in comparison with the other pheromone receptors, clearly indicates polymorphic alleles for those two genes and therefore supports a mating-type specific role.  Table 1. Distribution of pheromone precursors, pheromone receptors and homeodomain genes in F. velutipes monokaryotic siblings.

Structure of the matB3 locus
The pheromone receptor and pheromone precursor genes together with CLA4 were mapped based on linkage of their respective contigs to a draft genome of F. velutipes KACC42780. The MatB3a locus containing FvSTE3.1 and FvPP1 was located on scaffold 1 and flanked upstream by FvSTE3.s3 and FvSTE3.s5, and downstream by FvSTE3.2, FvPP2 and FvPP3 (Figure 1). Notably, the pheromone receptors were almost evenly spaced by 177 Kb, 181 Kb and 184 Kb which was considerably more distant than has been reported for other basidiomycetes [31,32]. Localization of the genes on the same fragment of scaffold 1 suggested that they might be part of the matB3 locus. FvSTE3.s1 and FvSTE3.s2 were mapped on scaffold 26 and FvSTE3.s4 on scaffold 29 (not shown). CLA4 was located at 0.98 Mb and 1.16 Mb distance from the borders of scaffold 8 (not shown). This demonstrated that FvSTE3.s1, FvSTE3.s2, FvSTE3.s4 and CLA4 were not linked to the matB3 locus.

Identification of homeodomain genes
One F. velutipes HD1 gene (FvHD1-1), two HD2 genes (FvHD2-1, FvHD2-2) and the Mitochondrial Intermediate Peptidase (MIP) gene were identified in the F. velutipes genome located on a single contig; Fv01174 ( Figure 4, Table S1). MIP was included because this gene is closely linked to HD genes in all Agaricomycetes [11,12]. Analysis of the FvHD2-1 and FvHD2-2 gene models (accession codes in materials and methods) showed intron-exon distributions and long C-terminal exons similar to that of homeodomain gene a2-1 and b2-1 of C. cinerea [14]. The first and the second intron interrupted the homeodomain at the exact same locations in F. velutipes and C. cinerea. The second introns have been reported to be conserved in several other basidiomycetes as well [7,14,48,49]. FvHD1-1 (six predicted introns) showed no intron-exon resemblance to C. cinerea HD1 genes associated with a2-1 and b2-1. No reliable coiled coils (less than 20-30% drops in the probability between weighted/non-weighted analysis) that could indicate dimerization motifs were found in the F. velutipes HD genes but nine-amino acid transactivation domains were detected in the Ntermini of FvHd1-1, FvHd2-1 and FvHd2-2 (Table S3). These domains are generally found in mammalian and yeast transcrip- Figure 3. Phylogenetic tree of mating-type specific and non mating-type specific pheromone receptor protein sequences. The tree shows phylogeny amongst pheromone receptor proteins from F. velutipes (Fv), C. cinerea (Cc), L. bicolor (Lb), S. commune (Sc), P. djamor (Pd) and C. neoformans (Cn). Nodal supports with more than 70% bootstrap values are considered strongly supported and displayed in the tree. Known matingtype specific pheromone receptors are depicted in blue. Two major clades are distinguished, labeled A and B. The four non mating-type specific pheromone receptors of F. velutipes FvSte3.s1 to FvSte3.s5 (pseudogene FvSTE3.s2 was excluded) form a separate group (shaded purple) within clade B that is supported by strong branch values. The clade including FvSte3.1 and two other known mating-type specific pheromone receptors is shaded in orange. The clade that contains FvSte3.2 is shaded in yellow. Both these clades are supported by strong branch values. FvSte3.1 and FvSte3.2 group closest with SCBbr1 and SCBbr2, respectively. Clades that contain known mating-type specific pheromone receptors are strong evidence for mating-type specificity of other clade members. LbSte3.5 is a notable exception. doi:10.1371/journal.pone.0022249.g003 tion factors [40]. FvHd1-1, FvHd2-1 and FvHd2-2 were further predicted to contain single and or bipartite nuclear localization signals (Table S3).

Segregation of FvHD genes
In order to link the FvHD genes to specific matA loci we determined their distribution in matA3 and matA4 mating-types. Specific primers for FvHD2-1, FvHD2-2 and FvHD1-1 were designed for exon regions that flanked the conserved homeodomain (Table S2). FvHD2-1 and FvHD2-2 showed specific linkage to the matA3 locus whereas FvHD1-1 was detected both in strains with matA3 and matA4 loci (Table 1). PCR products (average 450bp) obtained from matA3 and matA4 strains with FvHD1-1 specific primers, showed 97% base pair similarity and 100% amino acid identity (no polymorphism) demonstrating presence of a copy instead of different alleles. This means that the matA3 locus lacks a specific HD1 gene in comparison with matA4.
Structure of the matA locus MIP was located 201bp distant from FvHD2-1 which is in contrast to most known basidiomycetes where MIP is directly flanked by a HD1 gene that is part of a HD1/HD2 couple [12]. Detailed screening of the sequence adjacent to MIP and FvHD2-1 did not reveal a FvHD1 gene but instead, a hypothetical gene with no orthologues near matA subloci in L. bicolor (373Kb distant), C. cinerea (different chromosome) and S. commune (absent). Gene FvHD1-1 and FvHD2-2 that are 199bp apart, showed divergent, outward transcription directions ( Figure 5) which is typical for HD1/HD2 couples in basidiomycetes [12]. FvHD1-1 and FvHD2-2 were separated by 73 Kb from MIP and FvHD2-1. MatA subloci at large distance have so far only been reported for S. commune [4,32]. No Beta-flanking gene was detected near the HD genes, which was unusual [12]. These deviations showed that the matA locus had been subjected to several rearrangements relative to all other Agaricomycetes.

Synteny of matA regions
To unravel the events that altered the organization of the matA region of F. velutipes we mapped synteny of 200 successive genes surrounding the matA locus of C. cinerea (chromosome 1, bp2471986-2934538) with that of L. bicolor, F. velutipes and S. commune. The latter was included because this fungus also has separated matA subloci and is taxonomically closer to F. velutipes than C. cinerea and L. bicolor [50]. Since there had been no annotation of the F. velutipes genome, C. cinerea genes were acknowledged as ''syntenic'' in F. velutipes when similar protein sequences with expect values equal to or smaller than 10 29 were obtained by tblastn. Loss of syntenic genes in the matA region from C. cinerea was indicated by 11 unique genes, and 21 (L. bicolor), 17 (F. velutipes) and 21 (S. commune) genes from C. cinerea that were not detected in one or two of these species. The selected C. cinerea region covered the matA containing F. velutipes contig Fv01174 (304 Kb) as well as parts of contig Fv02632 and Fv03236, the last which was found to contain the missing Beta flanking gene. The three contigs were linked to scaffold 3 (3.8 Mb) of the F. velutipes draft genome where Fv01174 was flanked upstream at 2.1 Kb by Fv03236 and downstream by Fv02632 at 28 Kb (Figure 4). Our analysis identified synteny between C. cinerea and L. bicolor over a remarkably large segment of 350 Kb (Figure 4, panel A), much larger than previously demonstrated [31]. In addition, we identified the boundaries of this syntenic segment, which are clearly denoted by genes from C. cinerea that have no local homologue in L. bicolor, were differently distributed in L. bicolor, or were inverted (Figure 4, panel A). Those boundaries coincide with the ends of the syntenic part of the matA region from F. velutipes (Figure 4). The matA region in F. velutipes showed highest synteny over the first 250 Kb of the 350 Kb segment and somewhat lower in the last 100 Kb (Figure 4, black line, 250 Kb; start to red mark, 100 Kb; red mark to end). Though the specific gene order of F. velutipes in comparison to C. cinerea was changed by inversions, the overall gene order as found in C. cinerea was shown to be strongly preserved. None of the inverted gene groups was translocated (Figure 4). In S. commune the syntenic 350 Kb matA region was also recovered, albeit in three different sections with 140 Kb and 200 Kb interval distances (Figure 4, purple and dashed light purple lines). Synteny was highest in the two segments corresponding to the first 250 Kb of the 350 Kb matA region and lower for the third fragment, resembling F. velutipes. Moreover, synteny of the fragments in S. commune ended at the same relative locations that represented the synteny boundaries in C. cinerea, L. bicolor and F. velutipes. Finally, the gene order within the three segments was strongly conserved, indicating a high level of gene conservation in this entire region for all Agaricomycetes. Inside the 350 Kb segment, two spots showed high recombination in all four species. The first spot was the matA locus itself, the second spot formed a small gap in synteny located between C. cinerea genes CC1G_01875.3 and CC1G_01877.3 (both hypothetical genes) and their respective orthologues. Exemplary, one of the few genes that was repositioned in L. bicolor in comparison to C. cinerea was inserted in this gap. In F. velutipes and S. commune gene clusters were separated by inversion or translocation at both of these locations (Figure 4, 5, 6). Detailed comparison of the F. velutipes gene order to that of C. cinerea showed that the matA locus of F. velutipes was separated by inversion of two 70 Kb fragments directly left and right of FvHD1-1 and FvHD2-2 (Figure 4, 5). Modeling a reversion of those clusters reunited the Beta-flanking gene, FvHD2-1, MIP, FvHD1-1 and FvHD2-2 in a similar distribution as found in C. cinerea and L. bicolor. Notably, FvHD2-2 and FvHD1-1 were never moved during the rearrangements ( Figure 5). Synteny mapping also revealed important differences between the matA regions of F. velutipes and S. commune. First, all S. commune homeodomain genes were repositioned, whereas FvHD2-2 and FvHD1-1 retained their position during rearrangements. This shows that the respective matA loci were split between different genes. Second, the S. commune fragments representing the first part of the high syntenic 350 Kb region, and whose rearangement caused separation of the matA subloci, constituted 140 Kb and 260 Kb (Figure 4, purple lines). This was significantly larger than the 70 Kb in F. velutipes (Figure 4, blue lines) and shows that different fragments were rearranged. Third, the S. commune fragments were wedged by sections not syntenic to the 350 Kb C. cinerea matA region (Figure 4, dotted light purple lines), which was not observed in F. velutipes.
To specify the events that split the S. commune matA locus we extended the synteny map between C. cinerea and S. commune in both directions, using on average each next 5 th gene on the C. cinerea genome ( Figure 6). Model A ( Figure 6A) showed highest similarity to the gene order in C. cinerea, with all major S. commune gene clusters in an equal orientation. Model B ( Figure 6B) with the S. commune chromosome in an opposite orientation, showed inversion of all but two major gene clusters. Both models indicated transposition of two (model A) or three (model B) high syntenic fragments corresponding to the 350 Kb region (Figure 6, blue and purple lines respectively). In addition, the two transposed fragments in model A were exchanged in order in comparison to C. cinerea. Repositioning of the two fragments representing the high syntenic 250 Kb region (discussed above) in model A, between the genes flanking the 350 Kb region and the genes representing the last 100 kb of the 350 Kb region on the other side ( Figure 6A, C, black lines), reconstituted a matA region mainly identical to C. cinerea. Model B clearly indicated inversions as a cause for separation of the S. commune matA subloci. Indeed reversion of the genes underscored by the first (390 Kb) and second (210 Kb) purple line reunited the matA subloci ( Figure 6D). However, the S. commune matA region that resulted from those inversions still contained a large non syntenic fragment ( Figure 6, red dotted line) and differed considerably in gene order from C. cinerea.

Discussion
With the fast growing range of sequenced genomes, new information of mating-type genes quickly becomes available. However, complete analyses of the genetic structure of matingtype loci in mushroom forming fungi remain rare. Our study focused on the genetic structure of the mating-type loci from Flammulina velutipes (Winter Mushroom or Enoki). This led to the finding of several characteristics for F. velutipes as well as new facts that will help gaining insight in mating-type locus evolution, especially in Agaricales.

The B mating-type genes
Three pheromone precursors were identified near pheromone receptors in F. velutipes. As we limited our screen for pheromones to the six contigs that contained pheromone receptors (total 446 Kb), more pheromones can be expected to be found in the genome [31]. However, those pheromones will not be closely associated with the pheromone receptors and are unlikely to perform a mating-type specific role. FvPp1, the only pheromone precursor that was specifically linked to the matB3 mating type is unique, ending with a tryptophan (W) after the CAAX-box. We suggest that this extra tryptophan originates from the former STOP codon (Stop = TAA, TAG or TGA). A single base pair substitution would suffice to change TAG or TGA to TGG (TGG = W). It is uncertain if the additional tryptophan impairs pheromone processing; yet deviant farnesylation signals have previously been demonstrated to be functional [23].
We identified two mating-type specific pheromone receptors (accompanying pheromones, phylogenetic grouping with other mating-type specific pheromone receptors and high sequence polymorphism) named FvSTE3.1 and FvSTE3.2, which represented two of the three fungal pheromone receptors families. No genes of the third pheromone receptor family were detected in the matB3 strain, though this family might be expected in other F. velutipes matB loci.The identification of multiple non mating-type specific pheromone receptors in F. velutipes is in line with recent discoveries in L. bicolor, C. cinerea and S. commune [3,31,32]. We demonstrated a clear distinction between the two pheromone receptor types based on phylogenetic distribution and sequence polymorphism. Different sequence polymorphism indicates that these genes are subjected to different selection mechanisms. Somewhere in evolution, non mating-type specific pheromone receptors must have been functionally and selectively separated from mating-type specific ones. At the moment it is uncertain if non mating-type specific pheromone receptors are functional and what role they perform in the fungus. At least, their role is not mating-type specific (this study, [31,32]).

Structure of the matB locus
Mating-type specific pheromone receptor FvSTE3.2, together with pheromone precursor FvPP2 and FvPP3, was positioned in the same region of scaffold 1 as the matB3a locus containing FvSTE3.1 and FvPP1 (Figure 3). Arguably, FvSTE3.2 and the accompanying pheromone precursors comprise a functional second sublocus (yet our strain is just homozygous) since this specific receptor was not detected in any of the F. velutipes strains from different locations. The 177 Kb distance between both subloci makes the matB locus of F. velutipes exceptionally large in comparison to other higher basidiomycetyes [20,31,32]. To this, non mating-type specific pheromone receptors have been demonstrated to be linked to matB loci [3,32] meaning that FvSTE3.s3 and FvSTE3.s5 could be part of matB3 as well. This would increase the matB3 locus to over 500 Kb. Identification of additional F. velitupes matB loci should reveal if this large distance is consistent in F. velutipes and if FvSTE3.s3 and FvSTE3.s5 are truly connected.
The presence of FvSTE3.1 and FvSTE3.2 on different subloci corresponds with their phylogenetic separation into two clades that were derived through duplication [12]. It is surprising that their closest homologues, ScBBR1 and ScBBR2, are alleles on the same locus (matBb) in S. commune [18,23]. The diversity between F. velutipes and S. commune, as well as the large distance in the first, shows that the genomic organization of matB loci should be considered more flexible than previously has been assumed.
The A mating-type genes Three homeodomain genes distributed over two distant subloci were identified in F. velutipes. Both subloci were specifically linked to mating-type A3 (FvHD2-1 and FvHD2-2) yet the only present HD1 gene (FvHD1-1) was found in matA3 and matA4 mating types. MatA4 thus contains a different matAa allele, consisting of FvHD1-1 either combined with another HD2 gene than FvHD2-2 or as a single gene. This means that FvHD1-1 or FvHD2-2, that form a mating-type gene couple in matA3, have been independently recombined in other mating types. Lack of a matA3 specific HD1 gene dictates existence of at least one other HD1 gene in matA4, presumably in the matAb locus.

Structure of the matA region
We mapped the synteny between F. velutipes, S. commune, L. bicolor and C. cinerea based on blast searches with successive genes of the latter. Consequently, few genes that are missing in C. cinerea but that might be syntenic between other species remained undetected. The high detail of our maps however, showed that the applied method was accurate for both annotated and non annotated species. F. velutipes and S. commune showed different gene orders when compared, yet both followed the overall gene order of C. cinerea and L. bicolor. This, together with non separated matA loci in C. cinerea and L. bicolor shows that they represent an ancestral organization. We identified a 350 Kb matA region that is strongly conserved amongst Agaricales. Notably, synteny of genes belonging to this region is preserved even if parts become separated by chromosomal rearrangements as was shown in S. commune. The borders of the 350 Kb region, as well as two internal hotspots (one of which is the matA locus) are conserved sites of recombination in two major clades of the Agaricales as classified by Matheny et al. [50]. They mark the edges of important rearranged segments in F. velutipes as well as S. commune. Reasonably, one might expect rearrangements of the strongly conserved 350 Kb matA region in other Agaricales, especially in the Marasmioid clade that contains F. velutipes and S. commune.
Our segregation experiments showed that both matA3 subloci are linked despite their 73 Kb distance. Until now, far distant matA subloci in Agaricales were only known in S. commune (,450 Kb) and considered to be an exception. The matA subloci of F. velutipes were demonstrated to be separated by inversion of two (,70 Kb) gene clusters and are clearly derived from an ancestral locus as represented by C. cinerea. It has been generally assumed that the matA subloci of S. commune have followed a similar course leading to separation. However, our analysis showed significant differences between rearrangements in F. velutipes and S. commune. What is more, the analyses of the extended synteny map between S. commune and C. cinerea strongly indicate that the matA subloci of S. commune were separated by transposition of gene clusters instead of inversions. As shown in model A ( Figure 6A, C) a two step transposition of two high syntenic segments corresponding to the conserved 350 Kb matA region, directly results in a gene order mostly similar to that in C. cinerea. Though it is possible to reconstitute the ancestral matA locus by two inversions of considerably larger fragments as shown in model B, many rearrangements including transpositions remain ( Figure 6B and D). Both the smaller sizes of the rearranged fragments and the fewer steps needed to reconstitute the ancestral gene order support the transposition model.
F. velutipes provides a phylogenetically diverse species with an unusual mating type system (one of two components of the matA3 locus, HD2, is variable relative to matA4 while the HD1 is identical and matB3 contains a unique pheromone precursor). This enables comparative genomics to identify trends in mating type locus evolution. Studies of mating-type genes in Agaricales have shown that subloci are typically closely linked (10-20 kb) with the directly surrounding genes, especially in the matA locus, being highly syntenic. In our study we determined that the Agaricales in fact have very large scale synteny at matA (,350 Kb) and that this synteny is maintained even when parts of this region are separated through chromosomal rearrangements (S. commune). Four conserved recombination hotspots allow reshuffling of large fragments of this region which resulted in separation of the matA subloci of F. velutipes as well as of S. commune, by different events. This implies that separation of matA loci is not exceptional and might be expected in other Agaricales. In addition to matA, we determined that also matB loci can exist over large distances (,180 Kb) and that non matingtype specific pheromone receptors and mating-type specific ones are controlled by different selection mechanisms. Finally, the genes that were linked to specific mating types will serve as important molecular markers for breeding.

Supporting Information
Figure S1 Alignment of the three pheromone precursor proteins from F. velutipes KACC42780. Conserved amino acids in the pheromone sequences are marked with *. The conserved putative proteolytic site represented by E and R is indicated by a bold line. The CAAX-box of each protein is designated by a dashed line. The C-terminal halves of FvPp2 and FvPp3 are highly similar (boxed).

(TIF)
Table S1 Contigs of F. velutipes KACC42780 that were used in this study. The table enlists the contigs of F. velutipes KACC42780 that were used in this study, describing their respective size and important genes that are located on those contigs. Genes that were specifically linked to mating type loci in this study are indicated with *. Pseudogene FvSTE3.s2 is italicized. The STE3 prefix 'Fv' is added for Flammulina velutipes, the small 's' preceding STE3 numbers is added to distinguish non mating-type specific pheromone receptors (FvSte3 similar) from mating-type specific pheromone receptors. Gene accession numbers for pheromone receptors, pheromones, homeodomain genes and MIP are given in the material and methods. (DOC) Table S2 Primers used in this study. List of specific primers for each pheromone receptor, pheromone and homeodomain gene from F. velutipes KACC42780. The first primer sets for each gene were used to determine distribution of genes for segregation analysis. The '*additional sets' for FvSTE3.1, FvSTE3.2 and FvSTE3.s5 were newly designed after amplification in additional F. velutipes strains failed. Primer 'Fl matA 1-1 rv 500bp' for gene FvHD1-1 is exceptional and anneals in an intron. (DOC) Table S3 Amino acid positions of domains that were detected in proteins FvHd1-1, FvHd2-1 and FvHd2-2. Homeodomain proteins were analyzed for 9 amino acid transactivation domains (9AA TAD) with 9aaTAD [40] and nuclear localization signals (NLS) by WoLF PSORT [38]. Numbers in the table refer to the amino acid (AA) numbers of the domains in the respective proteins. (DOC)