Unequal Recombination and Evolution of the Mating-Type (MAT) Loci in the Pathogenic Fungus Grosmannia clavigera and Relatives

Sexual reproduction in fungi is regulated by the mating-type (MAT) locus where recombination is suppressed. We investigated the evolution of MAT loci in eight fungal species belonging to Grosmannia and Ophiostoma (Sordariomycetes, Ascomycota) that include conifer pathogens and beetle symbionts. The MAT1-2 idiomorph/allele was identified from the assembled and annotated Grosmannia clavigera genome, and the MAT locus is flanked by genes coding for cytoskeleton protein (SLA) and DNA lyase. The synteny of these genes is conserved and consistent with other members in Ascomycota. Using sequences from SLA and flanking regions, we characterized the MAT1-1 idiomorph from other isolates of G. clavigera and performed dotplot analysis between the two idiomorphs. Unexpectedly, the MAT1-2 idiomorph contains a truncated MAT1-1-1 gene upstream of the MAT1-2-1 gene that bears the high-mobility-group domain. The nucleotide and amino acid sequence of the truncated MAT1-1-1 gene is similar to its homologous copy in the MAT1-1 idiomorph in the opposite mating-type isolate, except that positive selection is acting on the truncated gene and the alpha(α)-box that encodes the transcription factor has been deleted. The MAT idiomorphs sharing identical gene organization were present in seven additional species in the Ophiostomatales, suggesting that the presence of truncated MAT1-1-1 gene is a general pattern in this order. We propose that an ancient unequal recombination event resulted in the ancestral MAT1-1-1 gene integrated into the MAT1-2 idiomorph and surviving as the truncated MAT1-1-1 genes. The α-box domain of MAT1-1-1 gene, located at the same MAT locus adjacent to the MAT1-2-1 gene, could have been removed by deletion after recombination due to mating signal interference. Our data confirmed a 1:1 MAT/sex ratio in two pathogen populations, and showed that all members of the Ophiostomatales studied here including those that were previously deemed asexual have the potential to reproduce sexually. This ability can potentially increase genetic variability and can enhance fitness in new, ecological niches.

(HMG) domain in MAT1-2 (Coppin et al. 1997). Heterothallic (outcrossing) ascomycete fungi carry either one of the two idiomorphs at the MAT locus, and two isolates bearing complementary MAT1 idiomorphs are required to mate. In contrast, homothallic ascomycetes can undergo haploid selfing (completion of the sexual cycle with a clonemate) because MAT genes with a-box and HMG domains are both present in the mating partners (Debuchy et al. 2010;Billiard et al. 2011). This gene arrangement may have evolved under selection for universal compatibility (Billiard et al. 2011(Billiard et al. , 2012. The MAT1-1 idiomorph usually comprises a gene, MAT1-1-1, whereas the MAT1-2 idiomorph has a different gene, MAT1-2-1. However, additional genes such as MAT1-1-2 and MAT1-1-3 have been reported in the MAT1-1 idiomorphs in various sordariomycetous ascomycetes (Turgeon and Yoder 2000;Debuchy et al. 2010). The questions of which mode of sexual reproduction, heterothallism or homothallism, is ancestral and which genetic/evolutionary mechanism mediates the change from one to the other have received much attention from fungal biologists (Debuchy et al. 2010). Therefore, understanding the MAT locus organization can provide insight into the genetics and evolution of mating systems and life history in ascomycetes.
The fungi in Ophiostomatales (Sordariomycetes, Ascomycota) are diverse, with more than 100 species, including many important or aggressive tree pathogens responsible for wilt diseases, blue stain in commercial timber; in addition, they play an important role as insect symbionts and associates, as well as saprophytes (Zipfel et al. 2006). Most species belong to sexual genera Grosmannia, Ophiostoma, and the asexual genus Leptographium. In North America, Grosmannia clavigera is an important conifer pathogen that has a symbiotic relationship with its vector the mountain pine beetle (Dendroctonus ponderosae). Grosmannia clavigera and other fungal pathogens and symbionts, such as Leptographium longiclavatum and Ophiostoma montium, are carried in the mycangia and on the exoskeleton of mountain pine beetles because these fungi produce abundant slimy spores that attach themselves to the insect's body. These fungi can grow into the sapwood and damage the host tree's water transport system (Yamaoka et al. 1995;Solheim and Krokene 1998). The mountain pine beetles and its fungal symbionts have destroyed more than 17.5 million ha of lodgepole pine forests in western Canada in the last decade (http://www.for.gov.bc.ca/hfp/mountain_pine_beetle/ facts.htm), and the magnitude of devastation is the largest in recorded history in Canada (Kurz et al. 2008;Safranyik et al. 2010).
G. clavigera is very aggressive and it may be capable of detoxifying terpenoids that are an important class of defense compounds in pines . Because G. clavigera plays an important role in the pine2beetle2fungus dynamics and epidemics, it is essential to understand its biology, genetics, and population structure. Populations of G. clavigera are polymorphic and form distinct genetic clusters yet with some gene flow and admixture among clusters (Lee et al. 2007;Tsui et al. 2012). Evidence of random mating and linkage equilibrium were also revealed, indicating sexual reproduction occurred in G. clavigera populations (Tsui et al. 2012). However, the sexual fruiting bodies of G. clavigera are rarely observed (Lee et al. 2005), even though G. clavigera is reported to have a life cycle comprising both the asexual and sexual stages. It is important to understand the ability of a fungal pathogen to perform sexual reproduction because such information could provide clues to the population biology of these fungi, which could be useful to further our understanding of the recent unprecedented epidemic.
Grosmannia clavigera lineage Gs sensu (Alamouti et al. 2011) is heterothallic because a gene homologous to MAT1-2-1 is characterized at the MAT locus in the isolate for which the genome was sequenced but no gene coding for the a-box domain was reported . Although fungi in Ophiostomatales are diverse in breeding strategies with worldwide distribution, the MAT locus has been characterized in only a limited number of species, such as Ophiostoma ulmi, O. novo-ulmi, and O. himal-ulmi, which are responsible for the Dutch elm disease epidemic in Europe and North America (Paoletti et al. 2005a;Jacobi et al. 2010). The genes corresponding to opposite mating-types recently were reported in the MAT locus of Ophiostoma quercus, which causes significant sapstain in hardwood (Wilken et al. 2012). Because the MAT1-1 idiomorph has not been characterized in G. clavigera and the MAT locus organization has not been well investigated in other fungi of Ophiostomatales, we used genomics and primer walking approaches to study the organization and evolution of the MAT locus in G. clavigera and eight related fungi. To address the question of whether heterothallism or homothallism is a derived character state, we also compared the MAT locus organization with other ascomycete species in the Sordariomycetes. The aim of the present investigation was (1) to characterize the mating-type locus organization of G. clavigera bearing a MAT1-1 idiomorph; (2) to investigate the evolution of MAT genes and mating systems in fungi belonging to Ophiostomatales with respect to other ascomycetes; and (3) to determine the mating-type ratio in the populations of G. clavigera.
DNA extraction, polymerase chain reaction (PCR) amplification, sequencing DNA was extracted from all isolates using the procedures described in Roe et al. (2011) andTsui et al. (2012). The genome sequencing, assembly, and annotation of G. clavigera has been previously described in DiGuistini et al. (2011). Based on the genome sequence of G. clavigera (GenBank accession number: ACXQ02000000), genes homologous to cytoskeleton assembly protein (SLA), HMG-domain of MAT, and DNA lyase (APN), as well as a few hypothetical proteins without known functions, were identified and characterized ( Figure 1) . To characterize the MAT1-1 idiomorph, we used a long-range PCR amplification approach coupled with primer walking sequencing. Primers ER and UP2-2F targeting the SLA, and a hypothetical protein-encoding gene (CMQ_5208) located downstream from the MAT locus were designed for long-range PCR reaction ( Figure 1A) ( Table 2). Primers UP2-2R and APN3R, which targeting CMQ_5208 and APN, were also designed to amplify a 15 kb fragment of a putative MAT1-1 isolate ( Figure 1A).
Long-range PCR amplifications of DNA were carried out in 50 mL using a PTC-100 thermocycler (MJ Research Inc., Watertown, MA). Reaction mixtures contained 100 ng of DNA, 1· PCR buffer, 200 mM each dNTP, 0.6 mM of each primer (Eurofins MWG Operon, Huntsville, AL), 1.5 mL of dimethyl sulfoxide and 2 U of Phusion DNA polymerase (Finnzymes; New England BioLabs, Ipswich, MA). The PCR amplifications were performed for 30 sec at 98°, followed by 35 cycles of 10 sec at 98°, 30 sec at 60262°and 4 min at 72°, and final extension at 72°for 10 min. Sequencing reactions with primer walking were performed at the Centre de recherche du CHUQ, Québec, Canada. Primers for sequencing are listed in (Supporting Information, Table S1).
After MAT idiomorph characterization, mating-type specific PCR assay was performed by designing primers HMG1 and HMG2 targeting the HMG domain, as well as primers MAT1x1 and MAT1x2 targeting the a-box domain ( Figure 1B, Table 2). Fragments of n

Gene annotation and analyses
The genome and transcriptomes of G. clavigera were assembled and annotated as described in . Gene models were predicted using GLEAN, and the putative gene function assignments were generated from searches of the NCBInr and Swiss-Prot databases using BLAST with PFAM domain assignments . Sequence reads after primer walking were assembled using the Staden package (Staden 1996) and Geneious Pro (http://www.geneious. com/), and they were compared with genes present in GenBank using BLASTx and BLASTn. The assembled sequences were submitted to FGENESH+ within Softberry (http://www.softberry.ru/berry.phtml? topic=fgenes_plus&group=programs&subgroup=gfs) for gene prediction and to determine the location of the coding/non coding regions, and manually annotated with Artemis (Rutherford et al. 2000).
The nucleotide sequences of both MAT idiomorphs were compared using dotplot (matrix) analyses implemented in Geneious. The sequences of the MAT loci were also compared across different species using the online Artemis Comparison Tool (http://www. webact.org/WebACT/home) with BLASTn algorithm.

Evolutionary analyses of MAT genes
The amino acid sequences encoded by the MAT1-1 a-box and the MAT1-2 HMG domain were aligned to sequences of other ascomy-cetes from GenBank with Clustal W (Thompson et al. 1997) implemented in Geneious. Phylogenetic analysis was carried out with the Neighbor-joining (NJ) and maximum likelihood algorithm in MEGA 5 (Tamura et al. 2011), PhyML(Dereeper et al. 2008 implemented in (http://www.phylogeny.fr/version2_cgi/index.cgi), as well as PRODIST and PROML in PHYLIP 3.69 (Felsenstein 2005).
The nucleotide sequences were aligned in Geneious and manually adjusted with Se-Al v2 (Rambaut 1999) and analyzed with PAUP v4. b10 (Swofford 2003) and MEGA 5 (Tamura et al. 2011). Nucleotide data of different genes were subjected to parsimony analysis implemented in PAUP. Bootstrap support for the branches was based on 1000 replicates with TBR branch swapping algorithms and simple sequence addition. The individual data sets were also subjected to NJ implemented in PAUP using GTR model corrections with the proportion of variable sites and gamma shape estimated from Modeltest 3.7 (Posada 2006). Bootstrap values were estimated based on 1000 replicates. Sequences were deposited in GenBank (accession numbers of MAT idiomorphs: JX402930-JX402958, JX402993-JX402996; of other genes: JX402959-JX402992).
Signatures of purifying or positive selection acting on the MAT genes were tested at the codon level. A maximum likelihood analysis was used to fit codon substitution models to the data using the CODEML program within PAML (Yang 2007). Four random site models were used to describe the variation of v (= dN/dS) among codon sites in an alignment containing the full length MAT1-1-1 and truncated MAT1-1-1 genes of six species. Random site models M1A (neutral), M2A (selection), M7 (beta), and M8 (beta and selection) were used to describe the variation of v among codon sites within each MAT1 gene. M1a assumes two site classes in proportions p 0 and p 1 = 1 2 p 0 with 0 , v 0 , 1(purifying selection) and v 1 = 1 (neutral). M2a adds an additional class of site with v 2 as a free parameter, allowing for sites with v 2 . 1 (positive selection) with proportion p 2 . M7 is a flexible null model in which a v ratio for each codon is randomly selected from a beta distribution between 0 and 1. M8 adds one additional site class to M7 allowing for positive selection. A test for positive selection was implemented using likelihood ratio tests that compare models pair M1a/M2a (Yang 2007). Three different starting v values (0.2, 1.0, and 2.0) were implemented for each model fitting, as described in (Joly et al. 2010). Codon sites under positive selection were then identified using the empirical Bayes method to calculate the posterior probability that a particular amino acid belongs to a given selection class (neutral, deleterious, or advantageous) (Yang 2007).
Analysis of mating-type distribution in populations of G. clavigera and L. longiclavatum We used the aforementioned mating-type primers (HMG 1, HMG2, MAT1x1, MAT1x2) to identify mating-types in populations from Canada and the USA (Table 2). We performed PCR on DNA extracted from 335 isolates of G. clavigera characterized in a previous study (Tsui et al. 2012), as well as more than 100 isolates of L. longiclavatum currently being investigated (Farfan et al. 2011). Matingtype distributions were tested for deviation from the expected ratios of 1:1 using chi-square goodness-of-fit tests.
Reverse transcriptase (RT) quantitative PCR amplification RNA was extracted and RT was performed from G. clavigera isolate SL-KW1407 cultured on 1% MEA overlaid with cellophane for 4 d  to determine the level of expression from hypothetical protein encoding gene CMQ_5309 (truncated MAT1-1-1 gene) and MAT1-2-1. RT-quantitative PCR was also carried out with SsoFast EvaGreen Supermix (Bio-Rad) in a volume of 20 mL in a Bio-Rad CFX384 system. Primer pairs were internal to regions of MAT1-2-1 and truncated MAT1-1-1 to yield amplicons of 1002120 bp (Table  2). PCRs were described as follows: 96°for 45 sec, followed by 35 cycles of 95°for 15 sec, and 57.5°for 30 s, followed by a melt-curve analysis. The Cq value for amplification of the b-tubulin gene was used as a reference.

RESULTS
Organization of the mating-type locus in G. clavigera The MAT1-2 idiomorph and flanking genes of G. clavigera isolate SL-KW1407 were identified from the genome ( Figure 1A), and the amplification of the MAT1-2 idiomorph from isolate G. clavigera SS274 demonstrated greater than 99.9% sequence similarity to the sequence of the reference genome. The gene order and orientation near the MAT locus was syntenic with other Sordariomycetes. Two ORFs were predicted in the MAT1-2 idiomorph. One of the translated proteins bearing the HMG domain (285 amino acids; EFX05114) was homologous to the MAT1-2-1 of other ascomycetes and shared 65% similarity to that of Ophiostoma himal-ulmi. In contrast, CMQ_5309, encoding a hypothetical protein (EFX04946.1), had no significant similarity to any genes in the NCBI sequence database. SLA2, which encodes the cytoskeleton assembly control protein (EFX04935.1), was located upstream from the MAT locus, whereas cytochrome c oxidase subunit (EFX05020.1) gene COX13 and DNA lyase (EFX04950.1) gene APN2 were located downstream from the MAT ( Figure 1A). The deduced amino acids of SLA and APN had 68% and 62% identities, respectively, to those of Neurospora crassa, a relative in the Sordariomycetes. However, the intergenic distance between MAT and APN loci in G. clavigera was large, spanning greater than 10 kb with several putative proteins coding genes of unknown functions identified ( Figure 1A). One of the putative proteins (CMQ_5213; EFX04951) had a protease-associated domain and was homologous (35% similarity) to a RING-9 protein in Verticillium albo-atrum (XP_003007854). The functions of these additional genes to mating activity are not yet established. tBLAST searches using the a-box domain containing genes from different ascomycetes as queries did not return any significant matches in the SL-KW1407 genome, confirming the absence of MAT1-1-1 gene in the reference isolate.
Using the primer walking approach, we obtained the full sequence (22,910 bp) of the MAT1-1 idiomorph of G. clavigera isolate B101. The gene arrangement was syntenic to the assembled genome of n Table 2 Major primers used in this investigation
Dotplot comparison of the MAT1-1 and MAT1-2 idiomorphs revealed nucleotide sequence similarity (99%) in the SLA encoding gene, regions upstream from the MAT locus, as well as the gene encoding protein CMQ_5208 (Figure 2A). Surprisingly the gene encoding putative protein (CMQ_5309; EFX05047.1) located upstream from the MAT1-2-1 gene in isolate SL-KW1407 was homologous (.80% similarity in amino acids) to the MAT1-1-1 gene in the MAT1-1 idiomorph but shorter in length and without introns. Sequence comparison revealed a-box domain truncation/deletion (89 amino acids) at the N-terminus of the putative protein. The start codon was, however, present in the putative protein coding sequence (now called truncated MAT1-1-1) followed by five codons that are not homologous to the MAT1-1-1 on the MAT1-1 idiomorph ( Figure 2B). This 15bp 59 end sequence was used to search the G. clavigera genome as well as the NCBI database but no significant match was returned.
The MAT loci of additional G. clavigera isolates from different geographic locations were also sequenced using the identical longrange PCR approach to investigate whether the truncated MAT1-1-1 gene could be unique solely to isolate SL-KW1407 (Table 1). Sequencing data confirmed that the isolates bearing the same matingtype have identical idiomorph size (data not shown). The presence of this truncated gene in multiple G. clavigera isolates confirms that it is not a spurious result or a unique feature of the reference isolate.
Organization of the MAT idiomorphs in fungal species related to G. clavigera The presence of the truncated/incomplete gene suggested that the ancestor of G. clavigera may have had both MAT1-1 and MAT1-2 copies and was homothallic. To understand the evolutionary history leading to the presence of the truncated MAT1-1-1 gene, we used the same primer pair ER and UP2-2F (targeting SLA and hypothetical protein-coding gene CMQ_5208) to characterize the MAT loci of several fungi related to G. clavigera (Lim et al. 2004) (Table 1). If the truncated gene has been inherited from a common ancestor, it may be preserved in other related species since it may confer an evolutionary or ecological advantage.
Sequences containing orthologous genes of the MAT locus and flanking regions were characterized from 21 isolates belonging to seven species within the Ophiostomatales: Llo, Lw, Lt, Llun, Ga, Gr, and Gh (Table 1, Figure 1C). Interspecific variations in fragment size ranged from 5787 to 8011 bp in MAT1-1 isolates, and from 5209 to 7727 bp in MAT1-2 isolates among different species. The variations in fragment size are also partly due to incomplete sequencing of the flanking sequences of SLA and the putative protein-coding gene CMQ_5208. The 39end of mating-type loci in Llun and Gh were incomplete because the primers targeting the CMQ_5208 failed to amplify, and a new primer GCM3 was designed to target the conserved regions in the MAT1-1-3 for MAT1-1 idiomorph characterization. Conversely, primer HMG2 targeting MAT1-2-1 gene was used to amplify the partial MAT1-2 idiomorph fragment. Only the MAT1-1 idiomorphs of Lw and Gr were obtained because MAT1-2 isolates were not available (Table 1). The MAT1-1 idiomorph of L. lundbergii isolate DAOM64706 was shorter and incomplete when compared with other taxa, so its sequence was not included in the analyses.
Dotplot comparison of the MAT1-1 and MAT1-2 idiomorphs in each of the species also revealed nucleotide sequence similarity (95-99%) in the SLA protein-coding regions, regions upstream from the MAT locus, as well as the CMQ_5208 ( Figure S1). The organization of the idiomorphs among these seven species was identical, with the presence of three predicted ORFs in the MAT1-1 idiomorph and two ORFs in the MAT1-2 idiomorph, in addition to the truncated MAT1-1-1 gene (the comparison among Gh, Gc, and Llo is illustrated in Figure 3). The nucleotide sequence of the MAT idiomorphs was similar among these seven species within Grosmannia: it ranged from 72% between G. clavigera and G. huntii to 97.5% between G. clavigera and L. longiclavatum (Table S2 and Figure 3, A and B). Sequence similarity was much lower when compared to other ascomycetes.
Ophiostoma is a sister genus to Grosmannia (Zipfel et al. 2006). A genomic DNA library of O. montium populations (R. C. Hamelin, unpublished data) was searched to design primers anchoring the SLA and the flanking regions of the MAT locus in both mating-types. Using a long-range PCR amplification and primer walking approach, we also amplified the MAT locus from four isolates of O. montium including both mating-types (Tables 1 and 2). We found fragments of ca. 6 kb and 5 kb in size that corresponded to both MAT1-1 and MAT1-2 idiomorphs. Dotplot comparison of the MAT1-1 and MAT1-2 idiomorphs were similar to the orthologs from Grosmannia species and revealed nucleotide sequence similarity in the SLA protein and intergenic regions upstream from the MAT locus ( Figure 4A). The MAT1-1 idiomorph had three putative ORFs that corresponded to MAT1-1-1, MAT1-1-2, and MAT1-1-3 genes, and they were similar (53-76% at amino acids level) to those of O. novo-ulmi subsp. novoulmi ( Figure S2). The MAT1-2 idiomorph contained a MAT1-2-1 gene encoding a protein (263 aa) with HMG domain (75% similar to that of O. novo-ulmi subsp. novo-ulmi), as well as a truncated MAT1-1-1 gene without the a-box domain located upstream from the MAT1-2-1 gene. The truncated MAT1-1-1 (224 amino acids) was short when compared with the original MAT1-1-1 (714 aa) in the MAT1-2 idiomorph ( Figure 4B) and the extent of deletion/truncation was greater than that in Grosmannia species.
Evolutionary analyses of the MAT loci Phylogenetic analysis inferred from 129 amino acid characters from HMG-domain of 74 ascomycetes supported the monophyletic origin of fungi in Ophiostomatales, and Grosmannia and Ophiostoma formed a sister relationship with strong statistical support ( Figure  5). Representatives of these two genera also clustered with the members of Neurospora and Sordaria in Sordariomycetes, with strong likelihood support. Within the cluster of Ophiostomatales, G. clavigera, L. longiclavatum, and L. terebrantis together with either G. aurea or L. wingfieldii formed a monophyletic clade with .90% bootstrap support. O. montium formed a sister relationship to a cluster containing O. novo-ulmi and its relatives. Sporothrix schenckii, a human pathogen, also nested within the same cluster with strong support (93%). Sequence analysis of a-box domain from 39 ascomycetes species also supported the same sister relationship between Grosmannia and Ophiostoma but the relationships among the major ascomycete families were not well resolved ( Figure S3).
MAT genes have been useful for evaluating the phylogenetic relationships among different fungal species (Debuchy and Turgeon 2006;Devier et al. 2009). Gene genealogies of MAT1-1-2, MAT1-1-3, and MAT1-2-1 ( Figure S4) were concordant with previous species relationships established based on rRNA and other protein coding genes (Lim et al. 2004). Also, phylogenetic analyses of SLA, intergenic regions upstream of full-length MAT1-1-1 and truncated MAT1-1-1, COX13, and APN, demonstrated that sequences of opposite matingtypes corroborate species phylogenies rather than showing transspecific polymorphism ( Figure S5). There were no strong conflicts in tree topologies inferred from genes at and flanking the MAT locus. Also there was no conflict in topologies among various tree-building algorithms.
Phylogenetic analyses of the nucleotide sequences inferred that recombination occurred in the flanking sequences upstream from MAT1-1-1 and truncated MAT1-1-1 genes ( Figure 6, A and B). The tree topology of flanking sequence did not conflict with that of SLA coding sequences ( Figure S5) because representatives of both matingtypes clustered within a species ( Figure 6A). Without recombination, the flanking sequence on MAT1-1 idiomorph would be expected to cluster separately to the flanking sequence on the MAT1-2 idiomorph in the gene genealogy. The nucleotide sequence of truncated MAT1-1-1 genes diverged from the full-length MAT1-1-1 genes as a result of independent accumulation of mutations ( Figure 6B). Sequences of the opposite mating-type within a species did not cluster together as for the SLA gene, but were divergent even though they had high similarity (83%) at the amino acid level.
Based on the tests of selection using likelihood ratio tests, diversifying (positive) selection was detected in the truncated MAT1-1-1 from the full-length MAT1-1-1 at the intraspecific level (Tables 3, A and B). The truncated MAT1-1-1 genes in the MAT1-2 idiomorph are under positive selection from the full-length MAT1-1-1 genes on the MAT1-1 idiomorph in G. clavigera, L. longiclavatum and G. aurea, but not L. terebrantis (Tables 3, A and B). The truncated/ incomplete MAT1-1-1 gene may go through neutral evolution due to loss of function or adaptive evolution. In contrast, tests of selection on MAT genes (MAT1-2-1, MAT1-1-1, and truncated MAT1-1-1 individually) at the interspecific level revealed purifying selection (Table  S3C), indicating that these genes are preserved for proper function of the sexual cycle.

Transcript analysis of MAT genes
The MAT1-1-1 gene with a-box domain is a master regulator of sexual reproduction and is involved in gamete fertilization and the formation of ascogenous hyphae; the deletion of MAT1-1-1 gene can lead to incomplete development of perithecia in some ascomycetes (Debuchy et al. 2010). To determine whether the truncated MAT1-1-1 gene is transcribed, PCR and RT-qPCR experiments were performed on the cDNA. The MAT1-2-1 gene (Cq = 28) and truncated MAT1-1-1 gene (Cq = 32.9) were expressed during the vegetative stage in mycelia (4d culture with cellophane overlaid on MEA plates), even though the expression level was low compared with the reference b-tubulin gene (Cq = 25.5). The data were also consistent with the low level of sex gene expression reported in transcriptomic data in various terpenoid compound treatments (DiGuistini et al. 2011) ( Table S4).
Determination of MAT type ratio in populations of G. clavigera and L. longiclavatum Using specific primers targeting the a-domain and HMG domain, we detected both the MAT1-1 and MAT1-2 idiomorphs in G. clavigera. We tested the null hypothesis of balanced numbers of the two MAT idiomorphs within the epidemic population samples in Tsui et al. (2012). Both mating-types were present in the populations at all of the spatial scales. MAT1-1 and MAT1-2 frequencies did not significantly deviate from a 1:1 ratio in any of the 19 populations, the four genetic clusters inferred from Bayesian estimation during population genetic studies, nor the entire population (P , 0.05) (Table 4A). Similarly both MAT1-1 and MAT1-2 also were detected in populations of L. longiclavatum, and its mating-type ratio did not deviate significantly from 1:1 at small spatial scale (Table 4B). However the MAT ratio was significantly different at larger landscape level, where MAT1-1 isolates appeared more frequent in the Rocky Mountain (Cluster Rocky) whereas MAT1-2 isolates were predominant in the new epidemics area (Cluster North; northern British Columbia and Alberta) (Table 4B).

DISCUSSION
The MAT locus organization indicates heterothallism All fungal species in this study are heterothallic because they have a locus with one of the two alternative single-copy idiomorphs, MAT1-1 or MAT1-2. The organization of opposite MAT loci was highly conserved among L. longiclavatum, L. terebrantis, G. aurea, G. huntii, L. lundbergii, L. wingfieldii, and O. montium, thus supporting a common origin in Ophiostomatales (Butler 2007). Also the general SLA-MAT-APN pattern, as well as the synteny and orientation of MAT1-1-1, MAT1-1-2, and MAT1-1-3 genes on MAT1-1 idiomorph are consistent with other representatives within Sordariomycetes, such as Neurospora, and Podospora (Debuchy and Turgeon 2006). This suggests a common evolutionary origin of the MAT organization/ structure for members of the Ophiostomatales and even the Sordariomycetes.
Whether heterothallism is the ancestral state in ascomycetes has been a major biological question in fungal evolution (Coppin et al. 1997; Butler 2007). Evolution from heterothallism (outcrossing) to homothallism (haploid selfing) could be the most likely scenario based on population genetics models (Nauta and Hoekstra 1992;Billiard et al. 2011). Recent analyses of mating-type evolution have ascertained Neurospora to be heterothallic in ancestry (Strandberg et al. 2010;Nygren et al. 2011;Gioti et al. 2012). We speculate that the representative of Grosmannia and Ophiostoma may have diverged from the same common heterothallic ancestor as Neurospora. However, it is important to determine the mating-type organization from additional species in Ophiostomatales because their mating systems and breeding strategies (Wilken et al. 2012) may be different from other Sordariomycetes.
Another special feature in the MAT locus is that COX13 is located between MAT and APN genes in G. clavigera, and this is similar to the organization found in the human pathogen Histoplasma capsulatum (Fraser et al. 2007). Although the MAT and APN loci were at least 15 kb apart, there was no evidence to suggest the presence of transposable elements, which have been reported in MAT locus expansion in the obligate biotroph Blumeria graminis (Spanu et al. 2010), H. capsulatum (Fraser et al. 2007) and the endophyte Phialocephala fortiniii (Zaffarano et al. 2010).
Unequal recombination defined the evolution of the truncated MAT1-1-1 genes in Grosmannia and Ophiostoma Most fungi in this investigation carried the truncated MAT1-1-1 in the MAT1-2 idiomorphs. The polymorphism between full-length and truncated MAT1-1-1 indicated that the event leading to the current MAT loci organization was ancient-before the radiation of G. clavigera from other species. A number of competing scenarios may account for the evolutionary origin of this unique organization.
Unequal recombination/crossover at the MAT locus between opposite mating-type idiomorphs during the pairing of chromosomes should be the most favorable mechanism to account for the truncated MAT1-1-1 (Gioti et al. 2012) (Figure 7). The ancestral MAT1-2 idiomorph of Grosmannia and Ophiostoma members contained one ORF corresponding to the MAT1-2-1 gene bearing the HMG domain. A fragment of the MAT1-1 idiomorph, for instance the MAT1-1-1 gene and the 59end flanking sequence, could have become integrated into the ancestral MAT1-2 idiomorph during the crossover in sexual reproduction ( Figure 7). Afterward the a-box domain had been deleted over evolutionary time (Figure 7). Unequal recombination at the MAT locus is not unique to members of Grosmannia and Ophiostoma. This event has clearly happened many times, and its footprints have been demonstrated in other ascomycete species (Yokoyama et al. 2003;Paoletti et al. 2005b;Seidl et al. 2009;Amselem et al. 2011;Wilken et al. 2012) (Figure 5). Fragments of the MAT1-1-1 and MAT1-1-3 genes were reported in 10 isolates of O. quercus that contained the full, complete MAT1-2-1 gene in their MAT1-2 idiomorphs (Wilken et al. 2012). Similarly, truncated MAT1-1-1 genes were identified in the MAT1-2 idiomorphs of at least five Phialocephala species (Zaffarano et al. 2010) and the one in Hypocrea jecorina (Seidl et al. 2009). In contrast, a partial MAT1-2-1 sequence was found in the MAT1-1 idiomorph of Aspergillus fumigatus (Paoletti et al. 2005b). Also fragments of homologous MAT1-2-1 and MAT1-1-1 genes were detected bordering the matingtype idiomorphs in Botrytis cinerea isolates (Amselem et al. 2011).
Recombination is supposed to be suppressed and rare at the MAT locus in ascomycetes (Idnurm 2011), but recombination or crossover events have been reported. Homologs of MAT1-1-2 and MAT1-1-3 also were reported in the MAT1-2 idiomorph of Diaporthe W-and Gtype species, possibly due to recombination of idiomorphs (Kanematsu et al. 2007). The MAT1-2-2 gene in the MAT1-2 idiomorph of Magnaporthe oryzae was partially homologous to the MAT1-1-3 gene in opposite mating-types (Kanamori et al. 2007). Recombination breakpoints and unequal crossover events have been revealed in the MAT loci of Neurospora, Cochliobolus, Stemphylium, Ascochyta and Phoma during the comparison of MAT idiomorphs between n heterothallic to homothallic representatives (Yun et al. 1999;Inderbitzin et al. 2005;Gioti et al. 2012;Woudenberg et al. 2012). Acquisition of MAT genes by transposition has been independently reported in several ascomycetes (Rydholm et al. 2007;Poggeler et al. 2011;Gioti et al. 2012). Therefore, the intronless MAT1-1-1 in MAT1-2 idiomorph may have arisen by retrotransposition from the MAT1-1-1 cDNA from the ancestral MAT1-1 idiomorph, but the MAT loci do not contain transposon-related and repetitive sequences. Interspecific introgression of MAT1-1-1 gene and vegetative incompatibility genes from Ophiostoma ulmi into O. novo-ulmi has been proposed (Paoletti et al. 2006). Introgression or non-random acquisition of MAT genes was also demonstrated in the evolutionary histories among multiple Neurospora species Strandberg et al. 2010). However, the molecular data do not support introgression to account for the presence of the truncated MAT1-1-1 gene because the gene genealogies of MAT and flanking genes do not contradict the species phylogeny of Grosmannia and Ophiostoma known from the housekeeping genes (Lim et al. 2004), suggesting all these genes shared the same evolutionary history.
The occurrence of truncated mating-type genes in opposite MAT idiomorphs may suggest a common homothallic ancestor carrying a complete set of MAT genes, for instance all MAT1-1-1, MAT1-1-2, MAT1-1-3, and MAT1-2-1 located at a single locus in the same order as Sordaria macrospora [ Figure 15.2 in (Debuchy and Turgeon 2006)]. The MAT1-1 and MAT1-2 idiomorphs could have arisen from the loss of HMG and a-box domain sequences as a result of multiple translocation breaks and segregations (Paoletti et al. 2005b). This scenario had been proposed to explain the MAT locus evolution in Cordyceps takamontana (Yokoyama et al. 2003), members of Botrytis and Sclerotinum (Amselem et al. 2011), as well as the Aspergilli (Galagan et al. 2005;Paoletti et al. 2005b). However, this model of evolution is very unlikely based on the criterion of parsimony. Also G. clavigera and related fungi have syntenic order of MAT genes and may share a common heterothallic ancestor within the Sordariomycetes as discussed above.
Deletion of the a-box domain due to evolutionary degeneration of MAT1-1-1 genes The deletion/removal of the a-box domain in the truncated MAT1-1-1 gene may have been selected to avoid universal compatibility or haploid selfing (self-fertility) that involves no sexual recombination in reproduction (Billiard et al. 2011(Billiard et al. , 2012. In fact, the expression of additional a-box domain may interfere or compete with the signal from the "resident/ original" HMG domain at the same locus (Coppin et al. 1997), therefore the "additional" MAT1-1-1 gene may have been "inactivated" for functions in mating (Coppin et al. 1997) after the integration into the MAT locus as a result of unequal recombination. Under laboratory conditions, artificial association of both mating-type loci in the same nucleus of a heterothallic Neurospora crassa isolate led to inhibition of growth and the formation of barren perithecia in crosses with a tester (Perkins 1972). Transgenic dual maters (carrying genes of both mating-types) were also unable to produce progeny in isolates of N. crassa , Podospora anserina (Coppin and Debuchy 2000) and Cochliobolus heterostrophus (Turgeon et al. 1993).
In the absence of purifying selection, genes that have lost their original functions accumulate mutations and degenerate to become pseudogenes. Our data illustrated that the truncated MAT1-1-1 gene in O. montium is highly eroded (31% in length compared with its 'full length' homolog) and could be a pseudogene. Similarly, the truncated MAT1-1-1 gene reported in Cordyceps takamontana (known as Isaria tenuipes) was a pseudogene with accumulated mutations and stop codons (Yokoyama et al. 2005). The 39-end of the MAT1-1-1 gene in the MAT1-2 idiomorph of H. jecorina was also considered to be nonfunctional because the flanking region contained multiple stop codons with no translational start point detected (Seidl et al. 2009). Also, the duplicated homeodomain transcription factors at A matingtype locus in Coprinopsis cinerea (Basidiomycota) were deleted and inactivated progressively (Kues et al. 2012). The MAT locus in yeasts was considered a "deletion hotspot" with a continued process of evolutionary deletions, gene truncation and transpositions on chromosomal genes located beside the MAT locus after the mating-type switching event (Gordon et al. 2011).
The truncated MAT1-1-1 genes in G. clavigera and relatives could have lost their original function in sexual reproduction but the degree of degeneration is not the same as the homolog in O. montium. However, these genes do not contain any stop codons and introns. The CODMEL tests indicated purifying selection at the inter-specific level due to accelerated rate of amino acid changes. The deletion of the a-box domain in the truncated MAT1-1-1 genes also did not prevent its expression as it has already been demonstrated in Magnaporthe orzyae (Kanamori et al. 2007). It is possible that the truncated genes have evolved new functions through adaptive evolution because the organization has been maintained in the entire phylogenetic clade. In contrast, the 39-terminal truncated SXI1a gene at the MAT locus of Cryptococcus neoformans serotype-AD hybrid (Basidiomycota) is still functional in sexual reproduction and may even promote cell fusion (Lin et al. 2007). Additional in-depth molecular genetics studies and experiments are necessary to elucidate the possible biological functions of the truncated MAT1-1-1 gene.

Implications for the mating strategies and breeding systems in fungi
Fungi within the Ophiostomatales have complex mating behavior that range from strict outcrossing (heterothallism) to haploid-selfing (homothallism) (Gorton and Webber 2000;Carlier et al. 2006). Also, Ophiostoma ulmi had been thought to perform 'pseudoselfing' by a process involved in mutation at the MAT locus or introgressed MAT genes that led to a mating-type switch (Brasier and Gibbs 1975). Our results reflected that most members in Grosmannia are heterothallic in genetic makeup and they require a partner (outcrossing) to produce perithecia in life histories. These molecular data are largely congruent to the classical data from pairing-cultures, except that G. robusta was reported to produce perithecia readily in culture without pairing cultures (Jacobs and Wingfield 2001). Also evidence of incongruence between molecular data and classical data were reported in O. quercus, as MAT1-2-1 genes appeared to be present in both mating partners that are able to cross (Wilken et al. 2012). Previous phylogenetic data demonstrated that homothallism has evolved multiple times independently from within heterothallic ascomycetes (Yun et al. 1999;Strandberg et al. 2010;Billiard et al. 2011). Further characterization of the MAT loci from selfing species in Ophiostomatales based on classical mating studies could verify if homothallic members have been derived from a heterothallic ancestor, infer the phylogenetic relationships between the MAT locus organizations and reveal the mechanisms underlying the lifestyle changes.
Heterothallic (outcrossing) fungi gain benefits from recombination by increased genetic diversity and repaired mutation (Heitman et al. 2007). The presence of both mating-type isolates in G. clavigera at different spatial scales is consistent with high levels of genotypic diversity found in this fungus (Tsui et al. 2012). The finding of linkage equilibrium among microsatellite markers further indicated that sexual reproduction is a major factor influencing the population genetic structure and epidemiology of G. clavigera (Tsui et al. 2012). If G. clavigera undergoes a sexual cycle regularly, the ascospores and ascomata should be discovered without difficulty. It was proposed that the inability to find the ascomata in nature is due to the inappropriate sampling methodology, or collecting plant material at the wrong stage in the life cycle (Sommerhalder et al. 2006). Unfortunately, several attempts to cross complement isolates in vitro proved unsuccessful. Other possible explanations may be mutations in genes regulating the sexual development or environmental factors that reduce ascomata formation (Bennett et al. 2003).
Finally, our results revealed the presence of homologous MAT genes in Llo, Lt, Lw, and Llun, which have long been considered to be asexual (Jacobs and Wingfield 2001). We also found 1:1 matingtype ratio in L. longiclavatum populations. Evidence of purifying selection on the MAT genes at the inter-specific level indicated that sexual reproduction is important in nature or occurs regularly (López-Villavicencio et al. 2010). Our data showed that these fungi previously deemed asexual have the potential to reproduce sexually, as was demonstrated for other asexual fungi such as Penicillium, Aspergillus, and Fusarium (Butler 2007;Ropars et al. 2012). This ability potentially increases genetic variability and can enhance fitness of fungal pathogens in new, ecological niches (Coppin et al. 1997).