Mating-Type Locus Organization and Mating-Type Chromosome Differentiation in the Bipolar Edible Button Mushroom Agaricus bisporus

In heterothallic basidiomycete fungi, sexual compatibility is restricted by mating types, typically controlled by two loci: PR, encoding pheromone precursors and pheromone receptors, and HD, encoding two types of homeodomain transcription factors. We analysed the single mating-type locus of the commercial button mushroom variety, Agaricus bisporus var. bisporus, and of the related variety burnettii. We identified the location of the mating-type locus using genetic map and genome information, corresponding to the HD locus, the PR locus having lost its mating-type role. We found the mip1 and β-fg genes flanking the HD genes as in several Agaricomycetes, two copies of the β-fg gene, an additional HD2 copy in the reference genome of A. bisporus var. bisporus and an additional HD1 copy in the reference genome of A. bisporus var. burnettii. We detected a 140 kb-long inversion between mating types in an A. bisporus var. burnettii heterokaryon, trapping the HD genes, the mip1 gene and fragments of additional genes. The two varieties had islands of transposable elements at the mating-type locus, spanning 35 kb in the A. bisporus var. burnettii reference genome. Linkage analyses showed a region with low recombination in the mating-type locus region in the A. bisporus var. burnettii variety. We found high differentiation between β-fg alleles in both varieties, indicating an ancient event of recombination suppression, followed more recently by a suppression of recombination at the mip1 gene through the inversion in A. bisporus var. burnettii and a suppression of recombination across whole chromosomes in A. bisporus var. bisporus, constituting stepwise recombination suppression as in many other mating-type chromosomes and sex chromosomes.


Introduction
Mating systems are responsible for the degree of selfing/outcrossing in natural populations and impact gene flow, the accumulation of deleterious alleles, and adaptability [1][2][3][4][5][6]. Mating systems can be controlled by, and reciprocally influence, genetic systems governing mating compatibility, such as self-incompatibility systems in plants [7]. In heterothallic fungi, sexual compatibility is governed by a mating-type system, which is most often controlled by two loci segregating independently in basidiomycetes (e.g., rusts, smuts and mushrooms) [8]. To mate and produce offspring successfully, two haploid cells must carry from the JGI genome pages of C. cinerea (http://genome.jgi-psf.org/Copci1/Copci1.home. html on 27 March 2011) and of the Agaricus strains, and by blast, blastp and/or tblastn searches using Agaricus sequences against the GenBank and the JGI C. cinerea site. H97 is one component homokaryon from the heterokaryon Horst U1 [53] used for EST sequencing and genome annotation [51].
The haploid progeny (referred hereafter as H i ) used for the mapping purpose was derived from the intervarietal hybrid JB3-83 x U1-7 [17,38]. JB3-83 is issued from A. bisporus var. burnetti JB3 [17] that belongs to the same Sonoran desert population as JB137 (from which the sequenced strain JB137-S8 originates [51]), and U1-7 is identical to one of the constituent nuclei of Horst U1, i.e., homokaryon H39 [53]. The segregation of the MAT locus in the H i progeny has been described previously [41]. Details on the obtention of mapping data and on the construction of the genetic linkage map are available in [38]. In order to refine the location of the MAT locus, the segregation data of the molecular markers assigned to chromosome 1 [38,54] were recomputed using Mapmaker/exp V3.0b software [55]. Molecular markers of known sequences were assigned to the H97 genome sequence V3.0 by blastn searches. The MAT locus was then placed in the next round of the process of linkage mapping using available progeny data. Marker orders were tested using the ripple command. The multipoint likelihood of alternative orders was computed, and orders with log-likelihoods within same magnitude were considered as equally likely.
For differentiation patterns along chromosomes, we used for A. bisporus var. bisporus the genomes of H97 and H39, the constituent homokaryons with opposite mating types isolated from the cultivated heterokaryon U1. The two genomes of A. bisporus var. burnetti analysed, H119p1 and H119p4, are the two constituent homokaryons of opposite mating types isolated from the heterokaryon 119/9 (JB-13), originating from the Agaricus Recovery Program [31].

Orthology Reconstruction and Synonymous Divergence
Short-read sequences of the genomes of the constituent homokaryons of the var. burnettii 119/9 strain (i.e., H119p1 and H119p4) [31] were assembled with SPAdes v3.11.1 [63,64] with a hybrid assembly approach by using the available PacBio genome of H119p4 [31] as a reference with the "-trusted-contigs" flag.
As gene models were not available for A. bisporus var. bisporus homokaryon H39 and A. bisporus var. burnettii 119/9 genomes (i.e., H119p1 and H119p4), ab-initio gene annotations for these three genomes were obtained by using protein sequences of the var. bisporus homokaryon H97 v3.1 [27] and var. burnettii strain JB137-S8 [51] with BRAKER v2.1.5 [65][66][67][68][69]. The HD1 gene in H119.p1 remained undetected by automatic prediction but could be manually defined by its position in between the mip and the HD2 gene annotations and in structure by blasting with other A. bisporus and C. cinerea HD1 proteins. To find orthologous gene copies in the constituent homokaryon genomes of var. bisporus Horst U1 and var. burnettii 119/9, all-against-all BLAST searches were first performed for each pair of homokaryotic genomes with DIAMOND [67]. The output tsv files of DIAMOND were then used as the input files for orthAgogue [70] to find orthologous pairs of genes for each pair of homokaryotic genomes. Co-orthologous pairs of genes were aligned with macse v2.04 [71]. Synonymous substitution rates (dS) were estimated with the yn00 method of PAML version 4.9f [72]. Custom python and shell scripts were used to work with multiple loci simultaneously, and the plots were made with R version 4.0.3 [73].
As the hybrid assemblies of the H119p1 and H119p4 genomes were highly fragmented, and thus were not assembled into complete chromosomes, the full gene order for the 119/9 strains could not be determined from the assembly. However, the two varieties have highly collinear genomes [31]. We therefore plotted the dS along the gene order of the var. bisporus homokaryon H97 genome.

Detection of Centromeres
Centromeres have been identified and discussed previously [31]; in order to place them on our plots, we used three independent lines of evidence based on their features [31]. First, we downloaded transposable element (TE) sequences annotated for the A. bisporus var. bisporus H97 v3.0 genome [74] and investigated their positions along chromosomes. Second, we investigated the positions of orthologous gene pairs found in H97 and H39 genomes. Lastly, we investigated the 5mC methylation along H97 chromosomes. We considered genomic regions as centromeres following a previous publication [31], i.e., where (1) TEs are found in abundance (i.e., TE islands, (2) methylation is concentrated on TEs in the region and (3) there is a depletion of orthologous genes in the regions. To do these simultaneously, we used circos plots [75] showing TEs, methylation and orthologous genes along H97 and H39 genomes. We used bisulfite sequencing (BSseq) data published previously [76]. We trimmed BSseq data with TrimGalore v. 0.6.5 [77]. We used bismark v. 0.22.3 [78] to map BSseq data to the reference genome (H97 v3.1; [27]) and subsequent extraction of methylation data. We considered a cytosine as methylated only when it has at least 3X coverage, and the majority of the reads remained not converted to uracils.

Linkage Map of Mating-Type Locus of the Variety burnettii
Markers of a previously generated linkage map of the var. burnettii strain 119/9 [31] were used to plot genetic to physical distances of a 1 Mb region on chromosome 1 that contains the HD mating type genes. This map was generated with Genotyping by Sequencing data. The physical distance between adjacent markers (kb) was determined and divided by the genetic distances between these markers (cM). The kb/cM data were subsequently plotted to the physical distance (kb). A few markers were removed because they were in the wrong order, thus generating negative values. Wrong ordering of markers can occur especially in dense maps [79].

Defining the Exact Location of the MAT Locus on Chromosome 1
A new order of mating-type-linked molecular markers assigned to chromosome 1 [38,54] has been established ( Figure 1). This order is statistically as likely as the one previously defined [38], with a similar magnitude of log-likelihoods: the difference between the two log-likelihood values is 0.63, i.e., the likelihood of the new order is 10 0.63 = 4.26 greater than the older one. The total distance of the linkage group is unchanged. Except for the marker PR057, the positions of the markers are consistent with the expected order according to the genome sequence. Thus, the available genome sequence allowed defining the location of the sequenced marker AbSSR23 (GenBank X92961) to position 1077972-1081182 on chromosome 1 of strain H97 and of the marker ACM5C (GenBank GS376049) to position 831821-832017. The primers defining the CAPS marker PR006 (CAATCTCAAGCTTGCCTGC; AGGTGACATGTCA-GAAGCGC; [38]) match on chromosome 1 to positions 862656-862675 (+ strand) and 863836-863854 (− strand) ( Figure 1). Another closely linked marker, PR067 (AB126057), locates at position 1097918-1100046 on chromosome 1 (Figure 1). Genes 2021, 12, x FOR PEER REVIEW 6 of 28 Figure 1. Mapping the mating-type (MAT) locus on chromosome 1 of A. bisporus var. bisporus strain H97 with genetic data on markers previously published [38,54]. Only markers that made possible to anchor the scaffold 1 sequence are located. PRxxx, ACMxx are CAPS (cleaved amplified polymorphism sequence) markers; AbSSRxx are microsatellites. When available, the GenBank reference number is given in bracket. The position of the BSN locus controlling the presence of two or four spores in basidia is also indicated. bisporus strain H97 with genetic data on markers previously published [38,54]. Only markers that made possible to anchor the scaffold 1 sequence are located. PRxxx, ACMxx are CAPS (cleaved amplified polymorphism sequence) markers; AbSSRxx are microsatellites. When available, the GenBank reference number is given in bracket. The position of the BSN locus controlling the presence of two or four spores in basidia is also indicated.
The MAT locus resides in the mapping interval between the marker AbSSR23 and the marker ACM5C. With our marker segregation data, we are unable to map more precisely the MAT locus, as the probabilities of having the marker order AbSSR23-Mat-PR006-ACM5C or the alternative one, AbSSR23-PR006-Mat-ACM5C, were comparable, with a log-likelihood difference between the two orders of ca 0.35. However, the identification of HD genes within the same interval (see below) strongly supports the location of the mating-type locus in A. bisporus.

Identification and Structure of the Mating-Type Homeodomain Loci
Indeed, tblastn searching with HD2 protein b2-1 of C. cinerea (GenBank CAA56132) on the JGI page of strain H97 identified in the 214 kb region in between markers ABSSR23 and PR006 on chromosome 1 two putative genes (at positions 890959-892499 and 893233-894863). Their respective products contained a classical three-helical homeodomain motif of 60 amino acids [80] as found in basidiomycete HD2 mating-type transcription factors ( Figure 2A). These two HD2 genes are arranged in tandem ( Figure 3). Upstream of the two genes is another gene (at position 895062-897100; Figure 3) coding for a putative transcription factor with a TALE-homeodomain type [81] with three extra amino acids in between the first two helixes and a poorly conserved DNA-recognition motif as typical for basidiomycete HD1 domains ( Figure 2B). Congruent with this, the b1-1 mating-type protein of C. cinerea (GenBank CAA44210) matched in tblastn searches the DNA region encoding the TALE homeodomain.
The MAT locus resides in the mapping interval between the marker AbSSR23 and the marker ACM5C. With our marker segregation data, we are unable to map more precisely the MAT locus, as the probabilities of having the marker order AbSSR23-Mat-PR006-ACM5C or the alternative one, AbSSR23-PR006-Mat-ACM5C, were comparable, with a log-likelihood difference between the two orders of ca 0.35. However, the identification of HD genes within the same interval (see below) strongly supports the location of the mating-type locus in A. bisporus.

Identification and Structure of the Mating-Type Homeodomain Loci
Indeed, tblastn searching with HD2 protein b2-1 of C. cinerea (GenBank CAA56132) on the JGI page of strain H97 identified in the 214 kb region in between markers ABSSR23 and PR006 on chromosome 1 two putative genes (at positions 890959-892499 and 893233-894863). Their respective products contained a classical three-helical homeodomain motif of 60 amino acids [80] as found in basidiomycete HD2 mating-type transcription factors ( Figure 2A). These two HD2 genes are arranged in tandem ( Figure 3). Upstream of the two genes is another gene (at position 895062-897100; Figure 3) coding for a putative transcription factor with a TALE-homeodomain type [81] with three extra amino acids in between the first two helixes and a poorly conserved DNA-recognition motif as typical for basidiomycete HD1 domains ( Figure 2B). Congruent with this, the b1-1 mating-type protein of C. cinerea (GenBank CAA44210) matched in tblastn searches the DNA region encoding the TALE homeodomain. In the classical situation of a basidiomycete HD locus, HD1 and HD2 genes are divergently transcribed in pairs [82][83][84]. The H97 HD1 gene a1-1 is divergently transcribed to the two HD2 genes and forms with gene a2-1 a typical divergently transcribed HD1-HD2 gene pair ( Figure 3). Downstream of the HD1 gene transcribed on the opposite strand (at position 897420-899760) follows the mip1 gene, coding for a putative mitochondrial endopeptidase (ID 239372), that has high levels of identity (72-74%) to orthologous proteins from other Agaricomycetes including C. disseminatus (Q6Y5M6), C. cinerea (EAU92790) and Laccaria bicolor (EDR15793). In these three species also, mip1 flanks one side of the HD mating-type loci, whereas gene β-fg encoding a conserved fungal protein of unknown function flanks the other side [48,85,86]. In A. bisporus var. burnettii JB137-S8, downstream to the HD2 genes of the MAT locus, encoded on the oppo- site strand, are two copies of β-fg genes (at positions 887425-889480 and 883864-884895), while a single copy has been observed so far flanking HD mating-type genes in other Agaricomycetes. The two β-fg gene copies are also found in the A. bisporus var. bisporus H97 strain (as noted in a preliminary analysis [52]) and display several differences. The structure of the shorter gene β-fg1 with six introns (one after the start codon) and the length and amino-acid sequence of its protein are highly similar to the genes and proteins in C. cinerea [87], C. disseminatus [48], L. bicolor [86], Pholiota microspora [88] and Heterobasidion annosum [49]. The 220 aa-long product (ID 1137415) of β-fg1 shows 57-58% identity to the products of β-fg genes in C. disseminatus (AAZ14920), C. cinerea (EAU92782) and L. bicolor (EDR15176). β-fg2 is much longer, with 14 introns (also one after the start codon), due to multiplication of a 5 intron-exon unit, and encodes a protein of 406 amino acids (ID 1181595) with an N-terminal half composed of 7 1/2 repeats of a 20 amino acid motif (N 8/8 K 7/8 T 6/8 G 7/8 V 6/8 E 7/8 G 8/8 H 6/8 N 8/8 A 5/8 P 6/8 Q 7/7 S/T 5/7 G 7/7 H 6/7 P 7/7 Q 7/7 S/ T 7/7 G 5/7 V 6/7 ). In Rhodonia (Postia) placenta, the mating-type locus with the homeodomain transcription factor genes is also flanked by a gene encoding an enlarged β-fg protein possessing N-terminal repeats (R. placenta protein ID 124109 [89]).  In the classical situation of a basidiomycete HD locus, HD1 and HD2 genes are divergently transcribed in pairs [82][83][84]. The H97 HD1 gene a1-1 is divergently transcribed to the two HD2 genes and forms with gene a2-1 a typical divergently transcribed HD1-HD2 gene pair (Figure 3). Downstream of the HD1 gene transcribed on the opposite strand (at position 897420-899760) follows the mip1 gene, coding for a putative mitochondrial endopeptidase (ID 239372), that has high levels of identity (72-74%) to orthologous proteins from other Agaricomycetes including C. disseminatus (Q6Y5M6), C. cinerea (EAU92790) burnettii JB137-S8. Shared genes are shown by black boxes, and HD1 and HD2 genes by boxes striped from bottom left to top right and from top left to bottom right, respectively. Closed arrows beneath genes indicate direction of transcription. Genes of DNA similarity or high identity between the two strains are marked with percentages of identity. Insertions are shown by the insets (note the different relative size of the inset in strain JB137-S8). Direct or inversely repeated sequences are marked by arrowheads, and grey boxes indicate DNA sequences that in blastx searches in the NCBI GenBank hit proteins with transposon/retrotransposon activities.
The gene order mip1-HD1 and HD2-β-fg with an HD1-HD2 gene pair divergently transcribed is known from the HD mating-type locus of several other Agaricomycetes [42,48,50,86,90,91]. The gene content and order of the region and its location between molecular markers AbSSR23 and PR006 ( Figure 1) firmly linked to mating-type function strongly supports that the single A. bisporus MAT locus corresponds to these HD genes. The genetic distance between PR067-AbSSR23, AbSSR23-MAT, MAT-PR006 and PR006-ACM5C are 1.9 cM, 4.3 cM, 1.7 cM and 4.8 cM, respectively. For these same intervals, physical to genetic distances correspond to 8.81, 41.44, 11.77 and 6.38 kb/cM, respectively. In the intervarietal cross A. bisporus var. bisporus x var. burnettii, the recombination rate is on average 26.32 kb/cM [38], and the MAT locus thus appears to reside at a region with little recombination [38], similar as the HD mating-type locus in C. cinerea and other basidiomycetes.
We also searched for potential genes coding for proteins with pheromone activity and with pheromone receptor activity in the H97 genome. Four genes for putative precursors of lipid-modified pheromones (Table S1) were found by tblastn searching with the sets of mating-type pheromone precursors and the precursor for pheromone-like non-mating-type specific peptides identified from genomes of C. cinerea, L. bicolor and Schizophyllum commune [8,86,87,92]. Three of them localize on chromosome 1 to the regions that flank the two ends of a large island rich in transposable elements (about 270 kb in length), corresponding to the centromere [31], at a chromosomal position that links only loosely with the MAT locus ( Figure 1; distance > 50 cM Kosambi). The fourth gene is present on chromosome 2 [51] (Table S1). We found four genes encoding putative STE3-like pheromone receptors (Table S1) in tblastn searches using the sequences of all putative STE3-like receptors of the C. cinerea Okayama 7 genome [87]. One of the four H97 STE3-like genes (ste3.1) localizes on chromosome 5 and one (ste3.4) on chromosome 12. Two others (ste3.3, ste3.2) are present on chromosome 13 (Table S1). The scattering of pheromone precursor genes and pheromone receptor genes over several different scaffolds together with lack of linkage or, in the best case, only loose linkage to MAT activity further supports that none of these genes has a mating-type function.

Orthologous Mating-Type Genes in the A. bisporus var. burnettii JB137-S8 Genome
The genes described above in A. bisporus var. bisporus H97 (Table S1) were used to screen by blast and tblastn searches the genome of A. bisporus var. burnettii JB137-S8 on the respective JGI data page. All genes of strain H97 for putative pheromone precursors and for putative pheromone receptors had orthologs in the JB137-S8 genome at homologous chromosomal positions (Table S2; see below) with products sharing high identity (97-100%) with the H97 gene products (Table S1).
The mip1 gene and the two β-fg gene copies in JB137-S8 had products showing high identities (98-99%) with the H97 proteins (Table S1). In sharp contrast, the HD1 and HD2 mating-type proteins were little conserved in sequence, consistent with expectations given their typical high rate of evolution [8,93]. The product of HD1 gene a1-2 of strain JB137-S8 revealed only 76% identity to protein a1-1 that comes from the HD1 gene localized in the H97 locus at the position corresponding to gene a1-2 in JB137-S8 (Table S1; Figure 3). At the DNA level, they share an identity of 88% over their whole length, making it very likely that they are alleles of the same HD1 gene. Strain JB137-S8 has only a single HD2 gene. b2-1 corresponds in location to the HD2 gene b2-2 in strain H97 (Figure 3), while the two genes do not show significant DNA similarity with the exception of a short stretch of DNA in a region encoding part of the HD2 homeodomain (73% identity). The products of b2-1 and b2-2 share, however, 26% identity in protein sequence (Table S1), suggesting that also these could be alleles of one gene. In contrast, identities (35%) of b2-1 to the other H97 HD2 protein a2-1 were restricted to only a 118 aa-long region with the HD2 homeodomain ( Figure 2A). We found no significant similarity at the DNA level between the two genes.
Whereas no further HD2 gene was found in the JB137-S8 genome, there was a second HD1 gene in its MAT locus. b1-1 forms with gene b2-1 a divergently transcribed HD1 and HD2 gene pair (Table S1; Figure 3). The first 231 aa of its product show 33% identity to the region in the H97 HD1 protein a1-1 which covers the HD1 homeodomain ( Figure 2B) and, by analogy to other species, likely the N-terminal specificity domain required for heterodimerization with compatible HD2 proteins [43,93]. High similarity (72% DNA identity) between the two genes are restricted to the DNA stretch encoding the conserved helix III of the homeodomain, making it unlikely that b1-1 is an allele of a1-1.
In summary from the comparison of the genes in the MAT alleles of the two sequenced strains, it appears that two paralogous HD1-HD2 gene pairs (a.k.a. HD subloci) exist in A. bisporus, although in neither of the two A. bisporus varieties both gene pairs were complete. Analogous to the paralogous HD1-HD2 gene pairs in C. cinerea [83,94], we called these the a pair and the b pair ( Figure 3). The A. bisporus var. bisporus strain H97 has a complete a pair and misses an HD1 gene of the b pair, and the A. bisporus var. burnettii strain JB137-S8 has only an HD1 gene of the a pair but a complete b pair.
Interestingly, the length of DNA between gene mip1 and the two β-fg genes differs much in length between the two sequenced strains of different varieties ( Figure 3). The region in strain H97 is 10 kb long versus ca. 42 kb in strain JB137-S8. A sequence comparison revealed 98% DNA identity for the coding regions of the mip1 alleles, 97% DNA identity in the downstream 387 bp long intergenic regions, 88% DNA identity for the alleles of the a1 gene and 77% DNA identity for the 559 bp-long sequence fragments upstream to the start codons of the a1 alleles (Table S1). The data support that, as in other species [48,85,90,95], there is a sharp decline in DNA identity from the end of the mip1 gene (including its terminator region) to the start of the MAT locus. Interestingly, the 559 bp-long fragment upstream of the genes a1-1 and a1-2 covers the promoter region plus 386 bp of a start of a divergently transcribed HD2 gene (77% DNA identity to a2-1), suggesting that there is a short remnant of a former HD2 gene a2-2 in strain JB137-S8 (Table S1, Figures 3 and 4a). Furthermore, there is a 207 bp sequence in close vicinity in strain JB137-S8 that appears to represent a footprint of the 3 end of a former mip1 gene (77% DNA identity to the 3 end of mip1 of strain H97; Figure 4B) that had the same direction of transcription than the extant mip1 gene (Table S1, Figure 3). It is possible that this mip1 footprint originated from a duplication of a complete HD1 and HD2 gene pair, represented in strain JB137-S8 by the genes b1-1 and b2-1 ( Figure 3).
HD2 protein a2-1 were restricted to only a 118 aa-long region with the HD2 homeodomain ( Figure 2A). We found no significant similarity at the DNA level between the two genes.
Whereas no further HD2 gene was found in the JB137-S8 genome, there was a second HD1 gene in its MAT locus. b1-1 forms with gene b2-1 a divergently transcribed HD1 and HD2 gene pair (Table S1; Figure 3). The first 231 aa of its product show 33% identity to the region in the H97 HD1 protein a1-1 which covers the HD1 homeodomain ( Figure 2B) and, by analogy to other species, likely the N-terminal specificity domain required for heterodimerization with compatible HD2 proteins [43,93]. High similarity (72% DNA identity) between the two genes are restricted to the DNA stretch encoding the conserved helix III of the homeodomain, making it unlikely that b1-1 is an allele of a1-1.
In summary from the comparison of the genes in the MAT alleles of the two sequenced strains, it appears that two paralogous HD1-HD2 gene pairs (a.k.a. HD subloci) exist in A. bisporus, although in neither of the two A. bisporus varieties both gene pairs were complete. Analogous to the paralogous HD1-HD2 gene pairs in C. cinerea [83,94], we called these the a pair and the b pair ( Figure 3). The A. bisporus var. bisporus strain H97 has a complete a pair and misses an HD1 gene of the b pair, and the A. bisporus var. burnettii strain JB137-S8 has only an HD1 gene of the a pair but a complete b pair.
Interestingly, the length of DNA between gene mip1 and the two β-fg genes differs much in length between the two sequenced strains of different varieties (Figure 3). The region in strain H97 is 10 kb long versus ca. 42 kb in strain JB137-S8. A sequence comparison revealed 98% DNA identity for the coding regions of the mip1 alleles, 97% DNA identity in the downstream 387 bp long intergenic regions, 88% DNA identity for the alleles of the a1 gene and 77% DNA identity for the 559 bp-long sequence fragments upstream to the start codons of the a1 alleles (Table S1). The data support that, as in other species [48,85,90,95], there is a sharp decline in DNA identity from the end of the mip1 gene (including its terminator region) to the start of the MAT locus. Interestingly, the 559 bp-long fragment upstream of the genes a1-1 and a1-2 covers the promoter region plus 386 bp of a start of a divergently transcribed HD2 gene (77% DNA identity to a2-1), suggesting that there is a short remnant of a former HD2 gene a2-2 in strain JB137-S8 (Table S1, Figs. 3 and  4a). Furthermore, there is a 207 bp sequence in close vicinity in strain JB137-S8 that appears to represent a footprint of the 3′ end of a former mip1 gene (77% DNA identity to the 3′ end of mip1 of strain H97; Figure 4B) that had the same direction of transcription than the extant mip1 gene (Table S1, Figure 3). It is possible that this mip1 footprint originated from a duplication of a complete HD1 and HD2 gene pair, represented in strain JB137-S8 by the genes b1-1 and b2-1 ( Figure 3). We found a long stretch of extra DNA (35 kb estimated in length) in the MAT locus of the A. bisporus var. burnettii strain JB137-S8 between the mip1 footprint and the b gene pair. This 35 kb DNA fragment contains several sets of direct and inverted repeats as well as several corroded sequences for reverse transcriptases, integrases, maturases, gagpol-proteins, a superfamily 21 transposase and a protein with a hAT family dimerization domain ( Figure 3) and should thus originate from multiple insertions of transposable elements into the MAT locus.
In the A. bisporus var. bisporus strain H97 on the other hand, there is a 1.6 kb long insertion in between the 1 kb shared sequences of the intergenic region (97% DNA identity) of the two copies of the highly identical β-fg genes (98 and 94% DNA identity between the alleles; Table S1; Figure 3). This insert is flanked by inverted repeats of 65 bp (Figure 3), and there are other copies in the genomes of the two strains indicating it also to be a transposable element. As in other species [48,85,90], the sharp increase in DNA identity from homeodomain transcription factor genes to the location of the neighboring β-fg gene marks the end of the MAT locus ( Figure 3). The chromosomal region of A. bisporus defined here as the MAT locus fulfills thus the prerequisite of basidiomycete mating-type loci [82,93] of containing unique DNA sequences with highly diverged alleles of genes with potential functions in control of sexual development and mating compatibility.

MAT Genes and Their Products
EST sequences were available for all three homeodomain transcription factor genes of strain H97 (https://mycocosm.jgi.doe.gov/Agabi_varbisH97_2/Agabi_varbisH97_2. home.html accessed on 27 March 2011) indicating that all three identified mating-type genes are expressed. a1-1 and a2-1 both have three introns, and b2-1 has two. Resulting mature mRNAs encode proteins of the lengths of 627, 464 and 491 aa, respectively ( Figure S1 in the Supplementary Materials). The lengths of the JB137-S8 proteins a1-2, b1-1 and b2-1 are 624, 613 and 455 aa, respectively ( Figure S1). As in C. cinerea [85] and in L. bicolor [86], the HD1 proteins are therefore longer than the HD2 proteins. C-terminal to the HD1 homeodomains, the proteins contain a putative bipartite nuclear localization signal (NLS; Figure S1a in the Supplementary Materials) at corresponding positions to NLSs in the C. cinerea and L. bicolor HD1 proteins [86,96]. The regions C-terminal to the homeodomain in the HD2 proteins have no recognizable domain but in all three proteins the C-terminal halves are serineand threonine-rich (23-26% of all aa; Figure  110 to 120 aa in HD1 proteins; ca. 130-150 aa in HD2 proteins) and predicted structure, and have been shown to confer mating-type specificity by discriminative heterodimerization between HD1 and HD2 proteins [85,97,98].
The A. bisporus HD1 proteins as well as the HD2 proteins in blastp searches of the NCBI GenBank hit the respective HD1 and HD2 homeodomains of basidiomycete matingtype proteins and often also their sequences for the N-terminal dimerization domains (in HD1 proteins defined as pfam12731 = Mating_N domain). We used the N-terminal sequences up to the last aa of the homeodomain to perform phylogenetic analyses of allelic and paralogous HD1 and HD2 proteins of basidiomycete mating type proteins ( Figure 5). The two orthologous HD1 sequences a1-1 of var. bisporus and a1-2 of var. burnettii branched together ( Figure 5A). As expected, the paralogous HD1 sequence b1-1 was placed farther away, together with orthologous sequences from other species ( Figure 5A). The two paralogous HD2 proteins of the A. bisporus var. bisporus H97 (a2-1 and b2-2) branched together but not with b2-1 of A. bisporus var. burnettii JB137-S8 ( Figure 5B), possibly because of the N-terminal truncation of protein b2-1 that might hinder a more specific clustering of the protein.

Genes for Pheromone-Like Peptide Precursors
The amino-acid sequences deduced from the four genes in each of the two sequenced A. bisporus varieties for putative precursors of pheromone-like peptides are given in Figure 6.
EST data indicate that at least three of the four precursor genes are expressed, all three with transcripts that contain an intron in the 3 untranslated region (https://mycocosm.jgi.doe. gov/Agabi_varbisH97_2/Agabi_varbisH97_2.home.html accessed on 2011-04-01).
The two orthologous HD1 sequences a1-1 of var. bisporus and a1-2 of var. burnettii branched together ( Figure 5A). As expected, the paralogous HD1 sequence b1-1 was placed farther away, together with orthologous sequences from other species ( Figure 5A). The two paralogous HD2 proteins of the A. bisporus var. bisporus H97 (a2-1 and b2-2) branched together but not with b2-1 of A. bisporus var. burnettii JB137-S8 ( Figure 5B), possibly because of the N-terminal truncation of protein b2-1 that might hinder a more specific clustering of the protein.
.  bisporus H97 b2-1, Tremella mesenterica Sxi2) in the ClustalW alignment were manually adjusted to assign them appropriately in position. Extra aa were added to the short N-terminal sequence of C. cinerea a2-2 from EAU92788 since the non-functional gene contains a preliminary stop codon in the 5 end of the gene shortly after the normal ATG start codon [85]. Proteins from Tremellomycetes served as outgroup. Accession numbers refer either to NCBI GenBank or to protein IDs at the fungal JGI databases.

Genes for Pheromone-Like Peptide Precursors
The amino-acid sequences deduced from the four genes in each of the two sequenced A. bisporus varieties for putative precursors of pheromone-like peptides are given in Figure 6. EST data indicate that at least three of the four precursor genes are expressed, all three with transcripts that contain an intron in the 3′ untranslated region (https://mycocosm.jgi.doe.gov/Agabi_varbisH97_2/Agabi_varbisH97_2.home.html accessed on 2011-04-01).  [8,93] and the typical CAAX motif at the C-terminus for farnesylation and carboxymethylation [99] are underlined.
The four precursors are between 44 and 54 aa long ( Figure 6). Although all allelic pheromone precursors but the Phl2 precursors differed by one amino-acid from each other, the mature peptides would be alike between the strains of the two varieties and should thus confer no different pheromonal activity. Pheromone precursors are expected to be N-terminal processed first for secretion by proteolytic cleavage after the outmost conserved MDS/AF motif and then by a peptidase cleaving at a charged dipeptide motif such as ER/DR/EH/NH/EE/DE [8,93,100]. The C-terminal CAAX motif (C = cysteine, a = aliphatic aa, X = any aa) of precursors for lipid-modified fungal pheromones is modified by farnesylation of the cysteine residue, removal of the three terminal amino acids and carboxymethylation of the resulting carboxy-terminal cysteine [99,100]. All detected pheromone precursors possess internal charged amino acid motifs in expected distances to the characteristic C-terminal CAAX motif ( Figure 6). The predicted four mature peptides of a strain are about 11-13 aa long and differ from each other in sequence by one or more aa (46-91% identity/similarity). All mature pheromones are G-rich (Figure 6), and three of them resemble non-mating-type pheromone-like peptide motifs of C. cinerea (consensus GGGSGNGAYC) and L. bicolor (consensus GGGSGNNAYC) [8,86].

Genes for Pheromone Receptors
Analyses with the TMpred and the SPLIT 4.0 servers supported for all four predicted Agaricus STE3-like pheromone receptors a seven-transmembrane structure, while the TMHMM Server v. 2.0 predicted with good probability for STE3.4 only six transmembrane helices (the third helix being not well supported), but for the pheromone receptors STE3.1, STE3.2 and STE3.3 also seven. The sequences for the seven-transmembrane helices were used with those from proteins of C. cinerea, L. bicolor, R. placenta and S. commune to generate a phylogenetic tree (Figure 7). In C. cinerea, three paralogous subgroups of a pheromone receptor gene plus pheromone precursor genes in the PR mating-type locus were functionally defined [101][102][103]. The positions of the pheromone receptor genes within the PR mating-type locus and at non-mating type chromosomal sites have been previously found reflected in trees with proteins of C. cinerea, L. bicolor and R. placenta [8,86,89,104,105], which is confirmed in this study (Figure 7). Of the Agaricus proteins, only Stretches of charged amino acids for possible N-terminal processing [8,93] and the typical CAAX motif at the C-terminus for farnesylation and carboxymethylation [99] are underlined.
The four precursors are between 44 and 54 aa long ( Figure 6). Although all allelic pheromone precursors but the Phl2 precursors differed by one amino-acid from each other, the mature peptides would be alike between the strains of the two varieties and should thus confer no different pheromonal activity. Pheromone precursors are expected to be N-terminal processed first for secretion by proteolytic cleavage after the outmost conserved MDS/AF motif and then by a peptidase cleaving at a charged dipeptide motif such as ER/DR/EH/NH/EE/DE [8,93,100]. The C-terminal CAAX motif (C = cysteine, a = aliphatic aa, X = any aa) of precursors for lipid-modified fungal pheromones is modified by farnesylation of the cysteine residue, removal of the three terminal amino acids and carboxymethylation of the resulting carboxy-terminal cysteine [99,100]. All detected pheromone precursors possess internal charged amino acid motifs in expected distances to the characteristic C-terminal CAAX motif ( Figure 6). The predicted four mature peptides of a strain are about 11-13 aa long and differ from each other in sequence by one or more aa (46-91% identity/similarity). All mature pheromones are G-rich (Figure 6), and three of them resemble non-mating-type pheromone-like peptide motifs of C. cinerea (consensus GGGSGNGAYC) and L. bicolor (consensus GGGSGNNAYC) [8,86].

Genes for Pheromone Receptors
Analyses with the TMpred and the SPLIT 4.0 servers supported for all four predicted Agaricus STE3-like pheromone receptors a seven-transmembrane structure, while the TMHMM Server v. 2.0 predicted with good probability for STE3.4 only six transmembrane helices (the third helix being not well supported), but for the pheromone receptors STE3.1, STE3.2 and STE3.3 also seven. The sequences for the seven-transmembrane helices were used with those from proteins of C. cinerea, L. bicolor, R. placenta and S. commune to generate a phylogenetic tree (Figure 7). In C. cinerea, three paralogous subgroups of a pheromone receptor gene plus pheromone precursor genes in the PR mating-type locus were functionally defined [101][102][103]. The positions of the pheromone receptor genes within the PR mating-type locus and at non-mating type chromosomal sites have been previously found reflected in trees with proteins of C. cinerea, L. bicolor and R. placenta [8,86,89,104,105], which is confirmed in this study (Figure 7). Of the Agaricus proteins, only the STE3.2 receptors group with proteins by mating type, all belonging to subgroup 2 of mating-type pheromone receptors (Figure 7). In contrast, the STE3.3 genes and the STE3.1 and STE3.4 genes cluster with two distinct groups of non-mating type pheromone receptors encoded by two genes found in all four other species in close distance to the PR mating-type loci [86,87,89,92]. In C. cinerea, these are the genes for CcSTE3-2151 and CcSTE-2153 [86,87].
The tree thus supports the inference from previous comparative analyses [87] that the ortholog of the pheromone receptor gene CcSTE3.2a/rcb2 of the C. cinerea PR mating-type locus is in A. bisporus on the chromosome 13. From 105 analysed genes from chromosome 13 of A. bisporus var. bisporus H97, 71 had related genes on chromosome 10 of C. cinerea (Table S2). Importantly, the gene ste3.2 sits in a block of four genes and gene ste3.3 in a block of five genes showing a conserved chromosomal order to the respective C. cinerea gene blocks including the PR mating-type receptor gene CcSTE3.2a/rcb2 and the non-matingtype pheromone receptor gene CcSTE-2153, respectively. These data and the tree shown in Figure 7 strongly suggest that the location of the locus orthologous to the PR mating-type locus has been found on the chromosome 13 in A. bisporus (ste3.2).
Genes 2021, 12, x FOR PEER REVIEW 15 of 28 the STE3.2 receptors group with proteins by mating type, all belonging to subgroup 2 of mating-type pheromone receptors (Figure 7). In contrast, the STE3.3 genes and the STE3.1 and STE3.4 genes cluster with two distinct groups of non-mating type pheromone receptors encoded by two genes found in all four other species in close distance to the PR mating-type loci [86,87,89,92]. In C. cinerea, these are the genes for CcSTE3-2151 and CcSTE-2153 [86,87].  [87], L. bicolor [86], R. placenta [89] and S. commune [92]. The α mating type pheromone receptor Ste3p from the ascomycete Saccharomyces cerevisiae served as an outgroup. Black bars mark proteins whose genes are localized within PR mating type loci (in the bipolar P. placenta, the orthologous PR locus with pheromone precursor and pheromone receptor genes seem to have lost mating function [89]. Bα and Bβ locus and B subgroup 1 to 3 indicate paralogous mating type proteins within a species as defined by classical genetics and transformation analyses [8,93]. Grey bars mark genes Figure 7. Phylogenetic analysis of the seven-transmembrane regions of putative mating-type-specific and non-mating-type STE3-like pheromone receptors from the Agaricomycetes A. bisporus var. bisporus H97, A. bisporus var. burnettii JB137-S8, C. cinerea [87], L. bicolor [86], R. placenta [89] and S. commune [92]. The α mating type pheromone receptor Ste3p from the ascomycete Saccharomyces cerevisiae served as an outgroup. Black bars mark proteins whose genes are localized within PR mating type loci (in the bipolar P. placenta, the orthologous PR locus with pheromone precursor and pheromone receptor genes seem to have lost mating function [89]. Bα and Bβ locus and B subgroup 1 to 3 indicate paralogous mating type proteins within a species as defined by classical genetics and transformation analyses [8,93]. Grey bars mark genes that are present on the same chromosome than the PR mating-type locus. Accession numbers refer either to NCBI GenBank or to protein IDs at the fungal JGI databases. m denotes adjustments done to C. cinerea gene models [8].

Differentiation along Mating-Type Chromosomes
We analysed the differentiation along mating-type chromosomes by plotting the pergene synonymous divergence in the heterokaryotic strains (dS, Figure 8). The dS level is expected to increase with time because of the linkage with the mating-type locus, as the cessation of recombination allows the accumulation of different substitutions in the alleles associated with the different mating-type alleles [11]. In the cultivated U1 strain A. bisporus var. bisporus, we found high dS levels along most chromosomes and including edges of chromosomes ( Figure 8A, Figure S2 in the Supplementary Materials). Unexpectedly, we found some genomic regions with zero dS within some chromosomes, not extending up to edges ( Figure 8A), indicating that the two parents used to generate the cultivated strain were related, as previously suggested [27]. Selfing in this variety with recombination being restricted at chromosome ends would indeed not generate such islands of homozygosity in the middle of chromosomes. The dS pattern in the cultivated U1 strain A. bisporus var. bisporus is thus consistent with this variety resulting from an inbreeding event and being clonally replicated for cultivation [27]. The very high differentiation between the β-fg1 alleles ( Figure 8A), much higher than the 99th percentile values of the dS distribution (Supplementary Figure S3), indicates an ancient event of recombination suppression at the mating-type locus and encompassing the β-fg1 gene, which is consistent with its linkage to HD genes in most Agaricomycota. The alleles at the HD genes were also highly differentiated between mating types, much higher than the 95 th percentile values of the dS distribution (Supplementary Figure S3), as expected given their role in mating-type determinism ( Figure 8A). The alleles at the mip gene were less differentiated, indicating more recent complete linkage to the mating-type HD genes. An even more recent event of recombination suppression occurred in A. bisporus var. bisporus, suppressing recombination across whole chromosomes, as indicated by linkage map analyses [31,38]. This event is more recent as it is found only in the variety bisporus and maybe only in the cultivated strain. As expected given their lack of role in mating-type determinism, the STE3-like genes were not particularly highly differentiated between mating types ( Figure 8A).
For A. bisporus var. burnettii, we studied dS levels in the heterokaryotic strain called 119/9, because the genomes of the constituent homokaryotic strains (H119p1 and H119p4) were available. In these strains, we also detected two copies of β-fg genes, the mip1 gene as well as an HD1-HD2 pair (Figure 9). We could not assemble the mating-type locus in a single contig in the H119p1 genome. We could nevertheless detect a ca. 140 kb-long inversion between mating types, trapping the HD genes, the mip1 gene and fragments of about 130 kb of additional genes, further supporting recombination suppression in several successive steps. We did not find evidence of any additional HD gene in this 119/9 heterokaryotic strain. One breakpoint of the inversion was placed between the HD1-HD2 gene pair and the gene β-fg2 and the other breakpoint within a conserved Agaricomycotina gene of unidentified function, splitting the gene coding for Agabi119p4_1168 in strain H119p4 into g2660 and g4769 in strain H119p1 (Figure 9).
We plotted the synonymous differentiation between H119p1 and H119p4 alleles along the well-assembled H97 genome and found high dS levels along most chromosomes ( Figure 8B, Figure S2 in the Supplementary Materials), as expected in an outcrossing species. Some regions with zero dS were nevertheless detected in some parts of chromosomes including edges. The dS plot thus seems mosaic as in the var. bisporus, with regions of high dS and regions of zero dS, but with an opposite pattern: the zero dS regions are at the edges of the chromosomes in the var. burnettii, while they are more in the middle of chromosomes in the var. bisporus. The zero dS regions extending up to chromosome edges in the var. burnettii are expected after a few selfing or inbreeding events when recombination is not restricted along chromosomes. Selfing may have occurred in the laboratory before sequencing this heterokaryotic strain [31]. The mating-type chromosome seemed to be the only one to have retained heterozygosity along its whole chromosome. Strikingly, the limit zero/non zero dS seemed often to occur at STE3-like genes ( Figure 8B), although we cannot see any reason for such a pattern. The alleles at the HD genes were highly differentiated, but also the alleles of the β-fg2 gene, further supporting an extension of recombination suppression in several steps.  . The Y-axis shows the synonymous divergence levels between the alleles found in the homokaryotic genomes of H97 and H39 of alternative mating types, issued from the U1 cultivar. The X-axis shows the gene order along the H97 genome. Yellow highlights on the background show the putative centromeric regions. Genes found in and around the mating-type loci are highlighted with different colours (blue for HD genes, orange for β-fg1 gene, brown for mip1 gene and magenta for STE3-like putative pheromone receptor genes). Other chromosomes are shown in Supplementary Figure S2.
single contig in the H119p1 genome. We could nevertheless detect a ca. 140 kb-long inversion between mating types, trapping the HD genes, the mip1 gene and fragments of about 130 kb of additional genes, further supporting recombination suppression in several successive steps. We did not find evidence of any additional HD gene in this 119/9 heterokaryotic strain. One breakpoint of the inversion was placed between the HD1-HD2 gene pair and the gene β-fg2 and the other breakpoint within a conserved Agaricomycotina gene of unidentified function, splitting the gene coding for Agabi119p4_1168 in strain H119p4 into g2660 and g4769 in strain H119p1 (Figure 9). We plotted the synonymous differentiation between H119p1 and H119p4 alleles along the well-assembled H97 genome and found high dS levels along most chromosomes ( Figure 8B, Figure S2 in the supplementary material), as expected in an outcrossing species. Some regions with zero dS were nevertheless detected in some parts of chromosomes including edges. The dS plot thus seems mosaic as in the var. bisporus, with regions of high dS and regions of zero dS, but with an opposite pattern: the zero dS regions are at the edges of the chromosomes in the var. burnettii, while they are more in the middle of chromosomes in the var. bisporus. The zero dS regions extending up to chromosome edges in the var. burnettii are expected after a few selfing or inbreeding events when recombination is not restricted along chromosomes. Selfing may have occurred in the laboratory before sequencing this heterokaryotic strain [31]. The mating-type chromosome seemed to be the only one to have retained heterozygosity along its whole chromosome. Strikingly, the limit zero/non zero dS seemed often to occur at STE3-like genes ( Figure 8B), although we cannot see any reason for such a pattern. The alleles at the HD genes were highly differentiated, but also the alleles of the β-fg2 gene, further supporting an extension of recombination suppression in several steps. The mating-type genes as well as flanking genes are found in three distinct contigs in the homokaryon H119p1 genome (indicated by black bars, the β-fg2 gene being split onto two contigs), while they are all found (β-fg genes in tandem) in one contig in the H119p4 genome. A ca. 140 kb-long inversion was identified between the two genomes of different mating types, figured by the pair of dark grey triangles. The inversion breakpoints are indicated by light grey symbols within the involved genes and non-coding DNA regions.

Recombination Suppression at the Mating-Type Locus in A. bisporus var. burnettii
Recombination suppression across whole chromosomes was clearly shown from a previous linkage map analysis in the U1 strain of A. bisporus var. bisporus except chromosome edges [31]. Recombination was more evenly distributed along chromosomes in A. bisporus var. burnetti, but the previous publication did not report finely recombination across the mating-type chromosome in this variety. We, therefore, re-analysed here the data focusing on the mating-type locus region. Markers of a 1 MB region of this map containing the HD mating type genes were used to plot the genetic to physical distance, showing a clearly higher genetic distance between the markers flanking the HD genes than between other marker pairs. That indicates a strong recombination reduction in the mating-type locus region ( Figure 10).

Recombination Suppression at the Mating-Type Locus in A. bisporus var. burnettii
Recombination suppression across whole chromosomes was clearly shown from a previous linkage map analysis in the U1 strain of A. bisporus var. bisporus except chromosome edges [31]. Recombination was more evenly distributed along chromosomes in A. bisporus var. burnetti, but the previous publication did not report finely recombination across the mating-type chromosome in this variety. We, therefore, re-analysed here the data focusing on the mating-type locus region. Markers of a 1 MB region of this map containing the HD mating type genes were used to plot the genetic to physical distance, showing a clearly higher genetic distance between the markers flanking the HD genes than between other marker pairs. That indicates a strong recombination reduction in the mating-type locus region ( Figure 10).

Discussion
A single mating-type locus governs mating compatibility in the button mushroom A. bisporus. Using linkage analysis with a genetic map, we could assign the MAT locus on chromosome 1 to a DNA region that contains HD1 and HD2 homeodomain transcription factor genes and that is flanked by the mip1 gene, corresponding to a putative mitochon-

Discussion
A single mating-type locus governs mating compatibility in the button mushroom A. bisporus. Using linkage analysis with a genetic map, we could assign the MAT locus on chromosome 1 to a DNA region that contains HD1 and HD2 homeodomain transcription factor genes and that is flanked by the mip1 gene, corresponding to a putative mitochondrial endopeptidase, and copies of the β-fg gene, with no predicted function, as described for several HD mating-type loci in Agaricomycetes [8,46,48,50]. A. bisporus is thus another case reported in the Agaricomycetes where bipolarity is found caused by the loss of mating-type function by the pheromone-pheromone receptor system at the PR locus, after C. disseminatus [48], P. microspora [46,50], Phanerochaete chrysosporium [47], Volvariella volvacea [106] and the antrodia clade of the Polyporales with R. placenta and Wolfiporia cocos [89,107]. Analyses of multiple Agaricomycetes supported that tetrapolarity is the ancient situation and bipolarity the derived one [107]. In other basidiomycete clades, such as in the Tremellomycetes, Trichosporonales, Microbotryales and Ustilaginales, bipolarity has also repeatedly evolved, but this occurred through chromosomal rearrangements and linkage by recombination suppression of the two MAT loci rather than the loss of the mating-type role by the PR locus [8,10,12,14,44,106,[108][109][110].
The high genetic differentiation between alleles at the β-fg genes at the mating-type locus together with the presence of an β-fg gene near the HD genes across Agaricomycetes [86] altogether indicate an ancient recombination suppression event trapping these genes at the mating-type locus. The inversion at the mating-type locus in the 119/9 heterokaryotic A. bisporus var. burnettii strain trapped the mip1 gene, while the differentiation between alleles at the mip1 gene was low, indicating another, more recent, recombination suppression event. We found duplications of the β-fg gene, as noted before [52], and, in addition, footprints of two HD subloci but with the deletion of one HD gene in one pair, although not the same in the reference genomes of the two varieties. Such gene duplications and losses are also a hallmark of regions with ancient recombination suppression [111,112]. Such an extension of recombination suppression to a few other genes near mating-type genes has also been reported in other fungal clades, with different genes involved depending on the clades [11]. The linkage of these genes to mating-type genes may sometimes be adaptive, as in the case of genes involved in mitochondria inheritance when mitochondria are transmitted by a single mating-type, and the mip1 gene may be involved in this phenomenon [113]. However, in most cases, as for the β-fg gene in Agaricus, it is not clear whether and why it would be beneficial to link different alleles to the mating-type genes; it has therefore been proposed that the extension of recombination suppression occurred for other evolutionary causes than antagonistic selection, such as for sheltering deleterious mutations or by the spread of transposable element methylation marks that accumulate in non-recombining regions [11]. In multiple fungi, the recombination suppression has extended across large regions distal to the mating-type genes without any genes involved in mating-type determinism, which seems to occur particularly often in pseudo-homothallic fungi [10,14,15]. Recombination suppression has extended across the whole mating-type chromosomes in the cultivated strain of the pseudo-homothallic A. bisporus var. bisporus variety and intriguingly also across whole autosomes [27,31]. Recombination suppression has thus extended stepwise, with at least two steps, in the A. bisporus var. bisporus mating-type chromosome, with first the β-fg gene given its high genetic differentiation, and then across the whole chromosome, as indicated by linkage map analyses. Such stepwise extension of recombination suppression fits the general pattern in pseudo-homothallic fungi [10,11,14,15,114]. As reported in other ancient non-recombining regions around mating-type genes [10,115], transposable elements have accumulated near the HD, β-fg and mip1 genes in the reference genomes of the two A. bisporus varieties. The evolutionary and proximal causes of such stepwise recombination suppression are still debated [11], and the A. bisporus var. bisporus case adds a nice biological model for testing hypotheses. It will be interesting to assess whether recombination suppression across whole chromosomes occurred only in the cultivated strain or occurs in the whole A. bisporus var. bisporus variety. Another sequenced commercial var.
bisporus strain had just a simple MAT locus restricted to only a single 4 kb-long HD1-HD2 gene pair, comparable to the H97 a pair [52]. Crosses between homokaryons of the U1 cultivar and wild A. bisporus var. bisporus homokaryons still displayed recombination restricted to chromosome ends [116], but recombination patterns have not been studied in wild strains of this variety. Stepwise recombination suppression seems to have occurred in several fungal mating-type chromosomes [11,14,114], as in many sex chromosomes [117]. The reasons why are still not understood [11], but several hypotheses have been invoked, such as the sheltering of deleterious mutations through permanent heterozygosity [118]. In the case of the U1 cultivar, the final step of recombination suppression along all chromosomes, and not only the mating-type chromosome, may alternatively be due to a strong selection under domestication for propagating the same heterozygous genotype without recombination, just as selfing annual crops or fruit tree crops propagated by grafting.
In the reference genomes of the two A. bisporus varieties, we found the genes orthologous to the PR mating-type locus scattered across the genome and lacking a mating-type role. The pheromone receptor gene ste3.2 orthologous to PR mating-type pheromone receptor genes in other Agaricomycetes locates on chromosome 13, and no pheromone precursor gene was found linked to this gene. The alleles at these genes were not differentiated between mating types, while in basidiomycetes the PR alleles usually show relatively high differentiation due to very ancient recombination suppression [119,120]. As in A. bisporus, PR genes do not control mating type anymore; pheromone precursor and receptor genes do not need to be physically linked or differentiated between mating types, while they could still perform important functions in the sexual cycle. Four pheromone-precursor-like genes were detected in the Agaricus genomes that relate to multiple similar genes in C. cinerea and L. bicolor, and they could activate the pheromone receptor STE3.2 independently of mating. Alternatively, mutations in the PR pheromone receptor gene could set active the downstream signalling system independently of mating. Indeed, in C. cinerea, self-compatible STE3 receptors originating from the B6 mating type have been described that by a single amino acid change (Q229P in the sixth transmembrane helix of rbc2 6 ), they constitutively activate the pheromone signalling pathway [100,121]. Another self-compatible mutant receptor (of rcb3 6 ) had a R96H substitution at the junction of the third transmembrane domain and the second intracellular loop that caused a change in specificity and allowed the binding of a self-pheromone, i.e., a pheromone encoded in the same PR mating-type locus [100,122]. Amino acid V183 from the third intracellular loop was lost in the rbc2 43 receptor in a self-compatible C. cinerea mutant [100]. Loss of two amino-acids (K75L76) in the third transmembrane helix of the PR mating-type receptor bbr2 of S. commune changed the spectrum of pheromone recognition [123].
Although non-mating type STE3-like receptor genes are broadly found in Agaricomycetes close to PR mating-type genes [47,86,87,89,90,92,104], their function and whether they interact with pheromones are still unknown. In the Tremellomycete Cryptococcus neoformans, a constitutively active non-mating-type receptor (CPR2) has been characterized that competes with the mating type-specific Ste3α pheromone receptor for signalling [110]. In A. bisporus, one of the non-mating type ste3 receptor genes discovered in the genome still remained close to the proposed ortholog of the PR mating-type locus (ste3.3), whereas a second gene has likely been translocated from the neighbourhood of the former mating type locus into two other places (ste3.1 and ste3.4), onto two different chromosomes and near the centromere for ste3.4. This might be part of a still ongoing degeneration process following loss of function but could also indicate that these genes have an important yet unknown cellular function that needs to be retained. Recent studies of natural expression and overexpression of non-mating type pheromone receptor-like genes in S. commune have suggested possible roles in vegetative filamentous growth and/or self-recognition [124].
In the tetrapolar Agaricomycetes species, the HD and the PR mating-type genes control different parts of the sexual cycle [43,93]. The HD locus encodes two types of transcription factors that act upon heterodimerisation as master regulators directly on genes in the nucleus [43,93]. Interactions between pheromones and their receptors start a signalling pathway that ultimately will target at regulating genes in the nucleus [82], including a kind of epistatic fine-tuning of HD-regulated genes that can explain the PR-improved frequency of HD-induced fruiting in C. cinerea [125]. In bipolar species with the PR mating-type locus not controlling mating type anymore, such fine-tuning, if still occurring, may follow another way. Overexpression of HD genes in the bipolar P. microspora caused clamp cell fusion [88], being in tetrapolar species a PR-controlled phenotype [43,93]. Clamp cell fusion occurs only after HD-induced clamp cell formation [93]. Therefore, the HD mating-type genes may well be involved in the molecular process occurring within the cells due to PR mating-type function, and the visible PR-regulated phenotype of clamp cell fusion could be mediated via the HD mating-type proteins. In the bipolar P. chrysosporium lacking also a functional mating type-determined pheromone-pheromone receptor-signalling system, nuclear migration has been observed between mating colonies [47], a phenotype controlled in tetrapolar species by the PR mating-type genes also in absence of compatible HD mating-type proteins [43,93]. Confrontation tests of different P. chrysosporium isolates indicated that the HD genes at the MAT locus can have a positive influence on nuclear migration, but the positive MAT control on nuclear migration is not exclusive and never observed to be bidirectional. Not all different MAT combinations resulted in nuclear migration, and unilateral nuclear migration was sometimes also observed between strains with identical MAT alleles [47]. Numerous mutant "modifier" genes outside of the PR mating-type locus in the tetrapolar S. commune were found to influence unilateral nuclear migration [126] which offers one explanation as how to bypass the control by PR mating-type genes in nuclear migration [47]. Another is that a kind of coordination is still exerted by the still available non-mating-type-specific pheromone and pheromone receptors in the pheromone response pathway [47]. It is conceivable that the HD genes in the single MAT locus take a positive influence on nuclear migration in P. chrysosporium and other bipolar species with observed mating-type effects on nuclear migration [47] by modelling the expression of pheromone and pheromone receptor genes and the pheromone response pathway. Any type of molecular communication between HD and PR genes has not yet been demonstrated in Agaricomycetes but is well established in the tetrapolar basidiomycete maize smut Ustilago maydis from the Ustilaginomycotina [127]. Nuclear pairing in mycelium of tetrapolar Agaricomycetes is HD gene-controlled [93,100]. Such control could be sufficient for consorted postmeiotic nuclear migration from basidia into spores of the pseudo-homothallic var. bisporus [26]. For the heterothallic var. burnettii, nuclear migration between mating homokaryons is rare [26] in accordance with a missing PR control.
An important question to address in evolution from tetrapolarity to bipolarity by loss of PR mating type function is therefore: are there phenotypes related to mating and/or sexual development that are under sole control of the PR genes? There is no nuclear migration or nuclear pairing in the vegetative mycelial cells in A. bisporus, and there are no clamp cells at hyphal septa of need to be fused [22,26]. Crucial for the success of mating and sexual reproduction are nevertheless heterologous nuclear recognition and maintenance of the heterokaryotic stage (heterokaryon stability). In the A. bisporus var. bisporus pseudohomothallic variety, nuclear recognition at basidiospore formation and/or appropriate positioning of the four haploid nuclei after meiosis ensure that nuclei of different mating types migrate in pairs together into one of the two spores of a basidium [26]. Formation of bisporic basidia has been suggested to be under the control of a main locus, BSN, on chromosome 1 but distinct from the MAT genes [40,41]. When such basidiospores germinate or when in the heterothallic variety two mating-compatible strains fuse, nuclei of different mating types stay together, although they do not show such tight pairing as it is seen in typical Agaricomycete dikaryons [22,26]. In the typical dikaryotic Agaricomycetes, nuclear pairing and synchronized nuclear divisions occur under regulation of the HD mating type genes [128]. As pointed out for S. commune [129], nuclear positioning and heterokaryon stability can be further enforced by the PR mating-type genes. Loss of matingtype function of the PR genes might thus contribute to the random distribution of nuclei in Agaricus suggesting the absence of pheromone-mediated nuclear communication. A more relaxed nuclear recognition in the hyphal cells might be antagonized by producing a high number of nuclei per cells, making it statistically less likely that upon cellular divisions one type of nuclei will be fully lost [22]. Under such conditions it could be sufficient for heterokaryon maintenance and sexual reproduction just to retain the HD genes' functionality within a MAT locus.

Conclusions
In this study, we refined the exact location of the mating-type locus on chromosome 1 of A. bisporus and report the structure of the HD genes and their organization at the mating-type locus for the reference strains of the two varieties bisporus and burnettii. We found interesting similarities and differences with the HD mating-type locus of other Agaricomycetes, and with striking islands of transposable elements. We also identified the gene structure and locus of a pheromone receptor gene orthologous to the PR matingtype locus in other Agaricomycetes but having lost its role in mating-type determinism in A. bisporus. The high differentiation between the β-fg alleles near the HD mating-type genes indicates an ancient event of recombination suppression, followed more recently by a suppression of recombination by an inversion trapping the mip1 gene in at least a strain of A. bisporus var. burnettii and across whole chromosomes in at least the U1 A. bisporus var. bisporus cultivar, thus constituting stepwise evolution of recombination suppression. We found evidence of gene duplication and gene losses around the mating-type locus, which is also a hallmark of regions with ancient recombination suppression. Overall, this study thus brings insights into mating-type locus evolution and mating-type gene structures, and can have applied interest for button mushroom breeding.